Understanding water resources conditions in data scarce river basins using intelligent pixel information Case: Transboundary Indus Basin
M.J.M. Cheema
Understanding water resources conditions in data scarce river basins using intelligent pixel information Case: Transboundary Indus Basin
Proefschrift
ter verkrijging van de graad van doctor aan de Technische Universiteit Delft, op gezag van de Rector Magnificus prof. ir. K.C.A.M. Luyben, voorzitter van het College voor Promoties, in het openbaar te verdedigen op dinsdaag 29 mei 2012 om 15:00 uur
door
Muhammad Jehanzeb Masud CHEEMA Master of Science University of Agriculture Faisalabad geboren te Sargodha, Pakistan
Dit proefschrift is goedgekeurd door de promotor: Prof. dr. W.G.M. Bastiaanssen
Samenstelling promotiecommissie: Rector Magnificus Prof.dr. W.G.M. Bastiaanssen, Prof.dr.ir. N.C. van de Giesen, Prof.dr. S. Uhlenbrook, Prof.dr.ir. P. van der Zaag, Prof.dr.ir. H.H.G. Savenije, Dr. F. van Steenbergen, Dr. W.W. Immerzeel,
voorzitter Technische Universiteit Delft, promotor Technische Universiteit Delft Technische Universiteit Delft en UNESCO-IHE Technische Universiteit Delft en UNESCO-IHE Technische Universiteit Delft Meta Meta Universiteit Utrecht
The research described in this dissertation was performed at the Water Resources Section, Faculty of Civil Engineering and Geosciences, Delft University of Technology, Delft, the Netherlands. The Higher Education Commission (HEC), Pakistan is thanked for providing funds to carry out this research. The International Water Management Institute, Pakistan is also thanked for providing financial support for additional months.
Copyright by M.J.M. Cheema, 2012 (
[email protected]) All rights reserved. No part of this publication may be reproduced or utilized in any form or by any means, electronic or mechanical, including photocopying, recording or by any information storage and retrieval system, without the prior written permission of the author.
ISBN: 90-6562-299-3
Published by . VSSD, Delft, the Netherlands
Keywords: Indus Basin, land use, surface soil moisture, ETLook, evaporation, transpiration, groundwater depletion
To my family
Contents Acknowledgements………………………………………………………..…….ix Symbols and Abbreviations…………………………………………………….xi 1
Introduction ............................................................................................................... 1 1.1 Transboundary river basins ................................................................................. 1 1.2 Water conflicts and treaties ................................................................................. 2 1.3 Indus water treaty................................................................................................ 3 1.4 Transboundary aquifer ........................................................................................ 9 1.5 Data availability and sharing issues .................................................................. 11 1.6 Redefinition of water resources management ................................................... 12 1.7 Remote sensing in hydrology and water management ...................................... 12 1.8 The research justification .................................................................................. 14
2
Study area ................................................................................................................ 17 2.1 Geographical description .................................................................................. 17 2.2 Hydro-climatology ............................................................................................ 18 2.3 Indus river, major tributaries and doabs ............................................................ 19 2.4 Groundwater ..................................................................................................... 20 2.5 Agriculture and cropping pattern ...................................................................... 22
3
Land use and land cover classification in the irrigated Indus Basin using growth phenology information from satellite data to support water management analysis ..................................................................................................................... 25 3.1 Introduction ....................................................................................................... 25 3.2 Study area ......................................................................................................... 26 3.3 Methodology ..................................................................................................... 28 3.4 Results and discussion ...................................................................................... 30 3.4.1
Phenology ........................................................................................... 30
3.4.2
Effect of physical condition on LULC ................................................ 34
3.4.3
Accuracy assessment .......................................................................... 38
3.5 Conclusions ....................................................................................................... 46
4
Local calibration of remotely sensed rainfall from the TRMM satellite for different periods and spatial scales in the Indus Basin ........................................ 48 4.1 Introduction ....................................................................................................... 48 4.2 Materials and methods ...................................................................................... 50 4.2.1
Study area ........................................................................................... 50
4.2.2
Rainfall systems over the Indus basin ................................................. 51
4.2.3
TRMM retrieval algorithm ................................................................. 51
4.2.4
Data availability .................................................................................. 54
4.2.5
Methodology ....................................................................................... 55
4.3 Results and discussion ...................................................................................... 57 4.3.1
Technique -1 ....................................................................................... 57
4.3.2
Technique -2 ....................................................................................... 59
4.3.3
Validation ........................................................................................... 62
4.3.4
Temporal and spatial deviation analysis ............................................. 63
4.3.5
Agricultural landuse – rainfall relationship......................................... 65
4.4 Conclusions ....................................................................................................... 67 5
Validation of surface soil moisture from AMSR-E using auxiliary spatial data in the transboundary Indus Basin .............................................................................. 70 5.1 Introduction ....................................................................................................... 70 5.2 Materials and methods ...................................................................................... 71 5.2.1
Study area and landuse patterns .......................................................... 71
5.2.2
Remote sensing data ........................................................................... 72
5.2.3
Methodology ....................................................................................... 72
5.3 Results and discussion ...................................................................................... 76 5.4 Summary and conclusions ................................................................................ 87 5.5 Appendix: Soil moisture retrieval algorithm ..................................................... 88 6
The surface energy balance and actual evapotranspiration of the Transboundary Indus Basin estimated from satellite measurements and the ETLook model ......................................................................................................... 91 6.1 Introduction ....................................................................................................... 91 6.2 Study area ......................................................................................................... 93 6.3 Material and methods ........................................................................................ 94 6.3.1
vi
Satellite data and pre-processing......................................................... 94
6.3.2
Meteorological data ............................................................................ 96
6.3.3
Theoretical background of ETLook .................................................... 97
6.3.4
Calibration and validation approaches .............................................. 101
6.3.5
Sensitivity and uncertainty analysis .................................................. 102
6.4 Results and discussion .................................................................................... 103 6.4.1
Surface energy balance ..................................................................... 103
6.4.2
Actual evapotranspiration estimates ................................................. 105
6.4.3
Validation ......................................................................................... 108
6.5 Summary and conclusions .............................................................................. 113 7
Spatial quantification of groundwater abstraction for irrigation in the Indus Basin using pixel information, GIS and the SWAT model ................................ 115 7.1 Introduction ..................................................................................................... 115 7.2 Material and methods ...................................................................................... 116 7.2.1
Study area ......................................................................................... 116
7.2.2
Soil and Water Assessment Tool ...................................................... 117
7.2.3
Data ................................................................................................... 119
7.2.4
ETLook ............................................................................................. 121
7.2.5
Model calibration procedure ............................................................. 122
7.2.6
Pixel based groundwater abstraction data ......................................... 123
7.3 Results and discussion .................................................................................... 124 7.3.1
Model calibration .............................................................................. 124
7.3.2
Spatial patterns of water supply and consumption ............................ 127
7.3.3
Accuracy assessment ........................................................................ 132
7.3.4
Water balance ................................................................................... 134
7.4 Conclusions ..................................................................................................... 135 8
Summary and conclusions .................................................................................... 137 8.1 Rationale ......................................................................................................... 137 8.2 Pixel land use .................................................................................................. 138 8.3 Pixel rainfall .................................................................................................... 139 8.4 Pixel surface soil moisture .............................................................................. 139 8.5 Pixel evapotranspiration .................................................................................. 140 8.6 Pixel groundwater abstraction ......................................................................... 142 8.7 New data sources ............................................................................................ 144 vii
8.8 Development of applications .......................................................................... 146 8.9 Conclusions ..................................................................................................... 147 9
Samenvatting ......................................................................................................... 151 9.1 Motivatie ......................................................................................................... 151 9.2 Pixel landgebruik ............................................................................................ 152 9.3 Pixel neerslag .................................................................................................. 153 9.4 Pixel oppervlak bodemvocht ........................................................................... 154 9.5 Pixel verdamping ............................................................................................ 155 9.6 Pixel grondwateronttrekking ........................................................................... 156 9.7 Nieuwe gegevensbronnen ............................................................................... 158 9.8 Toepassingsontwikkeling ................................................................................ 160
10 References .............................................................................................................. 163 Curriculum vitae………………………………………………………………………..185 Publications……………………………………………………………………………...186
viii
Acknowledgements I give honor and thanks to Almighty Allah, the source of knowledge and wisdom, who endowed me with the abilities for successful execution of this PhD research. I have been fortunate to work under dynamic supervision of Prof. Dr. Wim Bastiaanssen. His intellectual inspiration, valuable guidance, encouragement and sparing time from his busy schedules for lengthy stimulating discussions have been invaluable to me. I have learned a lot from him professionally as well as personally, which has significantly improved my professional capabilities. Thank you for all this and especially for arranging my difficult administrative requests. I am also extremely grateful to Dr. Walter Immerzeel for his helpful discussions on the SWAT model application. Thanks for significantly contributing to this research and for consistent encouragement and pushing me to wrap things up. Funds for this research were generously provided by Higher Education Commission (HEC), Pakistan and I am greatly indebted. Additional funds were made available by IWMIPakistan to support me for a few months of additional stay at TUDelft to complete the PhD conveniently. I also thank University of Agriculture Faisalabad (UAF) for granting me leave to enable me pursues this research. These funding institutes and their donors are gratefully acknowledged. Special thanks go to Rao Azhar (HEC, Pakistan), Loes Minkman (NUFFIC) and Franca Post (CICAT, TUDelft) for making all administrative and logistic work in the Netherlands possible. I am thankful to Dr Vladmir Smakhtin and Dr Asad Sarwar for their gentle and highly professional attitude, which greatly facilitated to successfully complete this study. For this study, secondary information was collected from various government agencies in Pakistan, including the Pakistan Meteorological Department (PMD), the Punjab Irrigation Department (PID), the SCARP Monitoring Organization (SMO) and the Indus Water Commission (IWC). Here I would like to thanks Engr.Sheraz Jamil Memon and Engr. Faris Kazi of IWC, Habib Ullah Bodla of PID and Dr Muhammad Arshad of UAF for their positive attitude and making it possible to get precious databases. Many thanks go to the colleagues in the section of Water Resources at TUDelft for the great assistance I received from them. Although I am grateful to everybody for the pleasant time, I would like to mention some colleagues in specific. Hanneke de Jong and Betty Rothfusz, thank you both for all administrative assistance you provided. Martine Rutten, Ilyas Masih, Saket Pande and Zheng Duan thank you all for good discussions. Reeza, Jacqueline and Congli for providing a friendly environment in the office. Thanks Miriam for your friendly and caring attitude and also for being my paranimf. Special thanks to Atiq, Naveed and Faisal for providing an atmosphere that always give me a feeling as I am in my homeland. I am also extremely thankful to Annemarie Klaasse and Henk Pelgrum of Water Watch for providing necessary support in collecting satellite data and understanding ETLook algorithm. I want to thank my friends who have made sure that not my whole life consisted of doing a PhD. In particular, I want to mention Bilal Ahmad, Faisal Nadeem, Fakhir, Seyab, Shah Muhammad, Atif, Laiq, Malik Aleem and Iftikhar Faraz who were always ready to play cricket and arrange dinners. Sarfaraz Munir and Syed Iftikhar Kazmi are specially thanked for the nice company which provided me an excellent opportunity to share my feelings and
ix
concerns more openly with someone from my own country, Pakistan. How can I forget the pleasant gupshup with Zahid Shabbir and fight with high velocity opposing winds while riding bikes from Rotterdam to TUDelft and back. It was a great adventure of my life, which I will not forget. Of course, this list is not complete and I want to thank all my friends but I would prefer to rather do this in person than in the form of an exhaustive list. I wish to express my gratitude to my family for their love, good wishes, inspirations and unceasing prayers for me, without which the present destination would have been mere a dream. The dream of my father, Masud Ata Cheema, to see me a doctor comes true. Today, I am missing my loving mother, but I am sure she will be happy in heaven. I would like to thank my uncles Dr Zahid Ata Cheema and Mr. Muhammad Aftab Mehmud who motivated me to start my PhD study. I also want to thank my aunts, brothers (Jehangir Masud Cheema and Mughees Aftab) and sisters (Kshif and Adeela) for their prayers and well wishes. Finally, I thank my wife and children for their patience and perseverance during long period of our separation and care and support while our stay in the Netherlands. Raheela, without you I would not have been able to finish this thesis as you always ask on which paper I am working, how many are submitted and how many are published? This kept me focused on the final goal. Final thanks to my little fairies, Shaiza and Hamima for their prayers and love. The sweet company of you made this tough journey a very pleasant and memorable experience of my life and I will never forget these moments.
x
Symbols and Abbreviations List of symbols α αo βw cp Cr DEPgw Δe ΔSus E ET ETo ETSWAT ETETLook ε εs εfw G H I IRRcw IRRgw IRRRS IRRSWAT Ksf λE Ln LOSScw Φ Ψ Qgw Qsurf Qlat Qperc ρ R R↓ r R2 ra,soil ra,canopy rcanopy
Shape factor Surface albedo Water use distribution parameter Specific heat of dry air Capillary rise in the unsaturated zone Net groundwater depletion Vapor pressure deficit Change in storage of the unsaturated zone Evaporation Evapotranspiration Reference crop evapotranspiration Actual evapotranspiration modeled by SWAT Actual evapotranspiration estimated by ETLook Dielectric constant Dielectric constant of soil solids Dielectric constant of free water Soil heat flux Sensible heat flux Interception Canal water supplied at farm gate Gross groundwater abstraction Total irrigation estimated by remote sensing Total irrigation applied in SWAT Ability of plant to extract soil moisture Latent heat flux Net longwave radiation Canal water losses Available water capacity of soil Soil evaporation compensation factor Return flow from shallow aquifer Surface runoff Lateral flow through unsaturated zone Percolation to saturated zone Air density Rainfall Incoming shortwave radiation Pearson’s product moment correlation Coefficient of determination Aerodynamic resistance for soil Aerodynamic resistance for canopy Canopy resistance
xi
Rn Rn,soil Rn,canopy rop Rr rs rs,min rsoil rsp RSWAT Rtoa RTRMM SeFC Sesub Setop Sm Sr St Sv T Tair Tb Tp τa τc τo τMODIS τr τsw U2 θsat θo θAMSRE θsat,xy θres,xy wup,z ω Λ Z z zd zroot
xii
Net radiation Net radiations at soil surface Net radiations at canopy Reflectivity from smooth soil surface Rainfall rate Spearman’s rank correlation coefficient Minimum stomatal resistance Soil resistance Reflectivity from rough soil surface Rainfall from SWAT Top of atmosphere radiation Satellite rainfall Effective saturation at field capacity Subsoil effective saturation Topsoil effective saturation Soil moisture stress Radiation stress Temperature stress Vapor pressure stress Transpiration Air temperature Brightness temperature Potential plant transpiration Atmospheric optical thickness Vegetation optical thickness Oxygen opacity at nadir Short wave transmissivity from MODIS Precipitation optical thickness Shortwave transmissivity Wind speed Saturated soil moisture content Volumetric water content AMSRE surface soil moisture Saturated moisture content at 1 km pixel (x,y) Residual moisture content at 1 km pixel (x,y) Plant water uptake factor Single scattering albedo Evaporative fraction Radar reflectivity factor Depth from soil surface Damping depth Depth of root development in the soil
List of Abbreviations ALOS amsl AMSR-E APHRODITE AVHRR AWR CAMS CCA CERES CMAP CRU CWR DAAC DEM DOY DN EROS ESA ETLook FAO FY-2 GDA GHz GIS GOP GPCC GPCP GRACE GLC ha hr HRU IB IB-IN IB-PK IBIS IBSP ICID ICIMOD IDW IGBP IN IRSA ISODATA
Advanced Land Observing Satellite Above Mean Sea Level Advanced Microwave Scanning Radiometer – EOS Asian Precipitation Highly Resolved Observational Data Integration Advanced Very High Resolution Radiometer Australian Water Resources Climate Assessment and Monitoring System Canal Command Area Clouds and Earth’s Radiant Energy System Climate Prediction Center’s merged Analysis of Precipitation Climatic Research Unit Crop Water Requirement Data Active Archive Centers Digital Elevation Model Day of Year Digital Numbers Earth Resources Observation and Science European Space Agency Evapotranspiration Look Food and Agriculture Organization Feng Yun 2 (Earth Observation System) Geographical Differential Analysis Giga Hertz Geographic Information System Government of Pakistan Global Precipitation Climatology Centre Global Precipitation Climatology Project Gravity Recovery and Climate Experiment Global Land Cover Hectare Hour Hydrological Response Unit Indus Basin Indus Basin Indian part Indus Basin Pakistani part Indus Basin Irrigation System Indus Basin Settlement Plan International Commission on Irrigation and Drainage International Centre for Integrated Mountain Development Inverse Distance Weighted International Geosphere-Biosphere Program India Indus River System Authority Iterative Self Organizing Data Analysis Technique xiii
IWASRI IWC IWT IWMI JAXA Km3 KPK LAI LAIeff LIS LT-1 LULC MERIS mha MINFAL MODIS NASA NCDC NDVI NGU NIR nm NOAA NSE NSIDC PARC PID PK PMD PR RA RE RFI RH RMSE ROI RS SC SEBAL SEE SI SMMR SPOT SRTM SSM/I SWAT
xiv
International Water Logging and Salinity Research Institute Indus Water Commission Indus Water Treaty International Water Management Institute Japanese Space Agency Cubic kilometer Khyber Pakhtunkhwa Leaf Area Index Effective leaf area index Lightning Imaging Sensor Length per Time Land Use and Land Cover Medium-spectral Resolution Imaging Spectrometer Million Hectares Ministry of Food, Agriculture and Livestock Moderate Resolution Imaging Spectro-radiometer National Aeronautics and Space Administration National Climatic Data Center Normalized Difference Vegetation Index Net Groundwater Use Near Infrared Nano Meter National Oceanic and Atmospheric Administration Nash-Sutcliffe Efficiency National Snow and Ice Data Center Pakistan Agricultural Research Council Provincial Irrigation Department Pakistan Pakistan Metrological Department Precipitation Radar Regression Analysis Relative Error Radio Frequency Interference Relative Humidity Root Mean Square Error Regions of Interest Remote Sensing Sensitivity Coefficient Surface Energy Balance Algorithm for Land Standard Error of Estimates Scattering Index Scanning Multi-channel Microwave Radiometer Satellite Probatoire d’Observation dela Terre Shuttle Radar Topography Mission Special Sensor Microwave/Imager Soil and Water Assessment Tool
SWIR TRMM TMI UN USGS VC VIRS VWC WAPDA WMO yr
Shortwave Infrared Tropical Rainfall Measurement Mission TRMM Microwave Imager United Nations United States Geological Survey Vegetation Cover Visible-Infrared Radiometer Scanner Vegetation Water Content Water and Power Development Authority World Meteorological Organization Year
xv
1 Introduction The most precious resource on earth, vital for human sustainability is water. Exponential increase in global population and unconstrained water resource utilization threatens the spatial and temporal availability of the world’s freshwater resources. The threat is more severe in developing countries where the majority of the population practices agriculture. Agriculture accounts for instance for 30% of the economy in a country such as Pakistan. Surface water and groundwater (separately or in combination) are used to fulfill the crop water requirements. The declining water resources need to be managed in an integrated way at basin scale. In fact, water management has to undergo significant improvements in terms of vision, targets, decision-making, and accounting. Hydrological simulation models can be used as analytical tools for determining the water flow paths (e.g. Andersson et al., 2006) and the impact of water management measures on irrigation systems (e.g. Droogers et al., 2000). The transboundary nature of river basins and the limited availability of data is however a hindrance for good modeling. Spatial information of topography, land use, rainfall, soil moisture, evapotranspiration, and leaf area index, derived from remote sensing can be used for, and will enhance spatially distributed modeling. Such data can also be used to validate and calibrate hydrological models. This thesis aims to improve the knowledge base of river basins by using satellite measurements and advanced hydrological models in data scarce environments to support short term and long term planning and water allocation processes. The transboundary Indus Basin is used as a case study.
1.1
Transboundary river basins
The river basin is the basic geographic unit which collects and provides water for the basin ecosystems itself, but also for agriculture, industry, and socio-economic development within the basin and downstream. The water of a basin flows across and underneath international boundaries to sustain agro ecosystems, whose boundaries do not coincide with the political boundaries. Such situations complicate the study of water flows and resource management. Nearly half the world is situated in 263 international river basins bearing 40% of the world’s population(Wolf et al., 1999). These 263 international rivers generate 60% of global fresh water. Most of these rivers are situated in Europe (69), followed by Africa(59), Asia (57), North America (40) and South America (38) as illustrated in Figure 1-1.
Figure 1-1 International river basins situated in the continents. (Source: Oregon State Uni)
The transboundary nature of the river basins has resulted in acrimonious disputes over water. Any change in upstream water use can severely affect the downstream users, although they may be thousands of kilometers apart from each other. The effects of land and water use planning in one part of the basin is vital for the users in another part of the basin (e.g. Molden et al., 2001). One example of such a river basin is the Indus Basin where the riparian countries have strained relationships over water flows. The Indus is a transboundary basin encompassing Pakistan, India, China and Afghanistan (Figure 1-2). Therefore, flow commitments by means of water treaties between the co-basin states are necessary for sharing and better utilizing of the resources.
Figure 1-2 Location of the Indus Basin
1.2
Water conflicts and treaties
All water users are hydrologically connected in a river basin. Upstream water use has a direct effect on the downstream users even thousands of kilometers away, or in another country. By promoting water supplies, upstream water users can cause dramatic consequences for downstream water users and their environments. Upstream riparian proprietors should not deprive downstream water users access in terms of quantity and quality. Even environmentally endorsed and acceptable practices to improve biodiversity and reduce soil loss upstream can lead to extermination of flora and fauna in downstream flood plains and estuaries. During wars, manipulation of the river waters can also be used as an offensive or defensive military weapon (Gleick, 2008). Water flowing across political boundaries has resulted in various conflicts and agreements of cooperation (treaties) among the riparian countries. During the last 60 years, 37 incidents of conflicts among the riparian countries over water are reported by Wolf (1998). For example, partitioning of India and Pakistan in 1947 divided the Indus Basin, which caused a continuous threat of war over water flows. India, occupying the upstream portion of the basin, had control of the barrages and diverted water to its own lands. It caused a serious 2
environmental threat to the lower riparian areas in Pakistan. The construction of new storage and hydropower facilities during 1947 to 1960 made the situation even worse. The Nile Basin is shared by ten countries. For many years, there was tension among the countries over the use of the Nile. For example, Tanzania vowed to use water from Lake Victoria (that feeds the Nile) for its domestic use; thus causing tension with lower riparian country Egypt. The tension between Egypt and Sudan over the water rights of the Nile increased in 1958 (Mandel, 1992). The plan by Egypt to divert water for use in the Sinai desert was strongly opposed by Ethiopia and the two countries were at the verge of war in 1980. The conflict between Syria and Iraq over the water of the Euphrates River stems from 1975 and is basically an un-solved issue. Construction of the Ataturk dam on the Euphrates river by Turkey has substantially reduced the flows to Syria (Zawahri, 2006). Crossfire between Israel and Syria over the water rights in the Huleh Basin occurred during 1951-53. Conflict arose between India and Bangladesh over the use of the Ganges River water and the conflict intensified in 1975 when India started to construct the Farakka barrage to unilaterally control the flow of the Ganges River. In the early 1990s Hungary and Slovakia started with the Gabukovo-Nagymaros barrage system along the Danube River. Conflict arose between the two countries and Hungary deployed troops to keep the system inoperative(Fuyane and Madai, 2001). To resolve such conflicts, the riparian countries have to come up with treaties defining water rights. The Food and Agriculture Organization (FAO) of United Nations reported that 3600 treaties on the use of international waters have been formed between 805 A.D. to 1984 A.D. Historically, the treaties to resolve water conflicts date back to 2500 B.C. when the two states of Lagash and Umma signed an agreement to end conflict along the Tigris River. A water treaty was signed between Mexico and United States in 1944 which defined the water rights and delivery responsibilities associated with the Colorado and the Rio Grande/Rio Bravo basins (Gastélum et al., 2010). Similarly, the Mekong River Commission was established in 1995 to efficiently utilize the resources of the Mekong River. In 1959, Egypt and Sudan signed an agreement to fully utilize the Nile water and established a Permanent Joint Technical Commission (Mandel, 1992). In 1996, India and Bangladesh signed a 30-year-treaty to share the flows of the Ganges which ended the dispute over Indian unilateral water diversions from the Farakka barrage. A similar effort was made between India and Pakistan in 1960, to resolve their water conflicts in the Indus Basin.
1.3
Indus water treaty
Development of irrigation systems in the Indus Basin dates back to the Harrapan civilization 2300 B.C. to 1500 B.C. (Fahlbusch et al., 2004). During the 2nd millennium, various Mughal emperors constructed limited canal systems to irrigate dry lands along the Ravi, Chenab and Sutlej rivers (Thatte, 2008). The systematic development of irrigation canals with weir-controlled structures started during British rule in 1850, when the 395 km long Upper Bari Doab canal (UBDC) was constructed. The headwork was constructed on the Ravi River at Madhopur in 1873.The next large project was the development of the Sirhind canal from the Sutlej River at Ropar to irrigate the districts of Ludhiana, Ferozpur 3
and Hissar, etc. It became operational in 1882. Afterwards, a network of canals was developed all over the Indus Basin, the Sidhnai canal taking off from the Ravi River was constructed in 1886. The Lower Chenab Canal (LCC) from Khanki headwork in the Chenab River became operational in year 1900. The Lower Jhelum Canal (LJC) commenced in 1901 from the left bank of the Jhelum River at Rasul barrage. A schematic diagram of irrigation system is shown in Figure 1-3 to give an idea of the location of various barrages and link canals in the basin. During the late 19th century, severe famine occurred that resulted in the establishment of the 1st Irrigation Commission of India in 1901. It came up with a proposal to transfer west flowing rivers eastwards to cope with the severe famine in the eastern parts. This proposal seems to be the base of the Indus Water Treaty (IWT) (Thatte, 2008). Development of The Triple Canal Project (Upper Jhelum canal: UJC, Upper Chenab canal: UCC, and Lower Bari Doab canal: LBDC) was proposed by the commission in 1905. It comprised a system of linked canals, including irrigation systems, starting from the Jhelum, through Upper Jhelum canal, to the Chenab River, and then to the Ravi River through the Upper Chenab canal. The project was completed in 1915.The gigantic Sutlej Valley Project (1921) was designed to replace the old-shutter type weirs with gate-controlled barrages. Four weirs at Ferozpur, Sulemanki, Islam and Punjnad were constructed. The former three were completed in 1927 and the latter one in 1933. Four canals, the Pakpattan, Dipalpur, Eastern and Mailsi canals were constructed in 1933 as part of this project. The Haveli and Rangpur canals were then completed in 1939, taking off from the Trimmu headworks, downstream of the confluence of the Jhelum and Chenab. In 1947, independence from the British rule resulted in the partitioning of the two riparian countries (Pakistan and India, sharing the major portion of the basin). Two major headworks, one at Madhopur on the Ravi and the other at Ferozpur on the Sutlej(rivers flowing eastward from India to Pakistan) went under Indian control. Irrigation in the Pakistani part of the Punjab province was dependent on these headworks. The Indian possession of the headworks resulted in administrative problems to regulate and supply water. To cope with the foreseen water crisis due to the stopping of east flowing rivers, Pakistan started to construct various barrages and linked canals to divert west flowing rivers eastward. For example, the Balloki-Sulemanki link (BSL:1954), the Marala-Ravi link (MRL:1956) and the Bombanwala-Ravi-Badian-Dipalpur link (BRBD: 1956) canals were constructed for this purpose. The Taunsa barrage was constructed in 1958. The Abbasia canal was extended, and the Thal canal project was undertaken. Meanwhile, India started construction of the Ferozpur and Rajasthan feeders in 1947. The Bhakra Nangal project started in 1948. The Harike barrage was completed in 1952. The Madhopur-Beas link canal was constructed in 1955 to divert waters of the Ravi to the Beas (Thatte, 2008). The Bhakra canal (remodeling of the Ropar headworks) and the Sirhind canal system were completed in 1955. The Rajisthan canal project was initiated in 1958.
4
Figure 1-3 Schematic diagram of the Indus Basin Irrigation Systems. C denotes canals, F feeders and L linking canals.
5
Due to disagreement on water use between the two countries, India diverted all flows from the east flowing rivers (the Ravi, the Beas and the Sutlej). It created water scarcity and an environmental threat in the eastern part of the Indus Basin located in Pakistan. Various conflicts arose between the two countries on water distribution of the rivers in the Indus Basin. To resolve these issues, water rights were defined under World Bank and United Nations auspices in 1960, by the signature of the famous Indus Water Treaty (IWT) between India and Pakistan. Nine articles with seven annexure were defined in the treaty. According to the IWT, the flows of three main west flowing rivers (the Indus, the Jhelum, and the Chenab) were available to Pakistan, while India had exclusive rights to waters of rivers flowing east. The treaty prohibited both countries from undertaking any structures that may change the volume of daily flows (Article II). Article III restricted India from constructing storage facilities on west flowing rivers. However, India was allowed to construct incidental limited storage on the western rivers. This was allowed only if the design was communicated to Pakistan six months in advance. The design needed to be approved by Pakistan and storage of structures should not exceed the defined volume. A permanent Indus Water Commission (IWC) was established under the IWT article VIII for smooth implementation of the treaty. The commission was to meet once in the year alternately in Pakistan and India. The functions of IWC were to establish and maintain cooperative agreements for IWT implementation, provide a report at the end of each year, inspection of rivers once in five years, and to settle disputes. The commission was also responsible to share data on agricultural use, hydro-electric power generation, water storage, and flows in the rivers. Under the Article VI, both countries were supposed to share daily gauge and discharge data, reservoir extractions, canal withdrawals and escapes. The IWT was successfully implemented in the first few decades and a number of reservoirs and a network of inter-river linking canals were constructed in the Indus Basin under the Indus Basin Settlement Plan (IBSP). The details of the linking canals along with their year of construction are provided in the Table 1.1. Table 1.1 Linking canals constructed in the Indus Basin before and after IWT S. No 1 2 3 4 5 6 7 8 9 10 11 12 13 6
Linking canal Upper Chenab Upper Jhelum Balloki-Sulemanki Marala-Ravi BRBD Madhopur-Beas Trimmu-Sidhnai Sidhnai-Mailsi Mailsi-Bhawal Rasul-Qadirabad Qadirabad-Balloki Chashma-Jhelum Taunsa-Punjnad
Off taking Linked rivers Construction Country Barrage year Marala Chenab-Ravi 1912 Pakistan Mangla Jhelum-Chenab 1915 Pakistan Balloki Ravi-Sutlej 1954 Pakistan Marala Chenab-Ravi 1956 Pakistan Marala Chenab-Ravi 1956 Pakistan Madhopur Ravi-Beas 1955 Pakistan Trimmu Chenab-Ravi 1965 Pakistan Sidhnai Ravi-Sutlej 1965 Pakistan Sidhnai Ravi-Sutlej 1965 Pakistan Rasul Jhelum-Chenab 1967 Pakistan Qadirabad Chenab-Ravi 1967 Pakistan Chashma Indus-Jhelum 1970 Pakistan Taunsa Indus-Chenab 1970 Pakistan
Length (km) 142 142 63 101 175 20 71 132 16 48 129 101 61
14 15 16
Beas-Sutlej Sutlej-Yamuna SutlejHaryanaAlwar
Pandoh Nangal Ferozpur
Beas-Sutlej Sutlej-yamuna
1977 U.C P
India India India
37 214
U.C: Under construction; P:Proposed; Sources:(Thatte, 2008; Wilson, 2011) After signing the IWT, the government of Pakistan started some mega projects. These included construction of two large dams, the Mangla dam (1966) on the Jhelum River and the Tarbela dam (1976) on the Indus River, construction of eight large capacity linking canals, six barrages and remodeling of three of the existing inter-river linking canals. There was no big irrigation canal project implemented after these developments. However, construction of three new irrigation canals: the Raini canal, the Greater Thal canal, and the Kachhi canal was approved in 2002. Amongst these, the former two are under construction. Construction of the large capacity multi-purpose Diamer-Basha dam on the Indus about 315 km upstream of Tarbela dam was initiated and is expected to be completed in 2018. The Kurramtangi dam on the Kurram River and the Munda dam on the Swat River are also proposed for construction. Detail of the major reservoirs constructed in the Indus Basin is provided in the Table 1.2. Table 1.2 Major reservoirs constructed in the Indus Basin S.No 1 2 3 4 5 6 7 8 9 10 11 12 13
Reservoir Tarbela Mangla Chashma Diamer-Basha Kurramtangi Munda Bhakra Pong Pandoh Thein Salal Baglihar
River Indus Jhelum Indus Indus Kurram Swat Sutlej Beas Beas Ravi Chenab Chenab Indus
Country Pakistan Pakistan Pakistan Pakistan Pakistan Pakistan India India India India India India India
Construction year 1976 1966 Under construction Under construction Under construction 1963 1974 1977 1995 2004 Under construction
In India, the Bhakra dam was completed in 1963 while the Rajasthan feeder canal was finished in 1964. The Pong dam (1974) and the Pandoh dam (1977) were constructed on the Beas River. The Beas-Sutlej link canal was constructed in 1977 to divert water from the Beas to the Sutlej. In 1985 a lift irrigation scheme was completed in the Haryana district. The Indira Gandhi Nahar Phase I was constructed in 1999. The dam on the Ravi was completed in 2001. Phase II of the Indira Gandhi Nahar project was completed in 2006. The Wullar barrage/Tulbul navigational project in the states of Jammu and Kashmir was proposed by India in 1984. The Salal dam project on the Chenab in the Jammu and Kashmir states was started in 1970 and step-wise completed in 1995. Another mega project downstream of Salal dam is the Baglihar dam. The construction began in 1999 and the first phase was completed in 2004. The locations of the various structures constructed after IWT are shown in Figure 1-4. 7
Figure 1-4 Location of reservoirs and barrages constructed on the Indus River and its tributaries. During the last decade, several issues have arisen, on which IWC is working to resolve within its mandate. India has for example started construction of storage structures on the tributaries of the Indus, whose rights were given to Pakistan. The Wullar barrage/Tulbul hydropower projects on the Jhelum, and the Kishan Ganga hydropower project on the Kishan Ganga river, a tributary of Jhelum, are few examples (Zawahri, 2009). Construction of the Baglihar dam on the Chenab with storage capacity of 37 million cubic meter (MCM) is considered a violation of IWT. Afghanistan is also planning to control the water of the Kabul River (Lashkaripou and Hussaini, 2007) with financial and technical support from India. Similar structures are proposed upstream of the Jhelum and Chenab (Khan, 2009). It is argued that, although there is a provision in IWT to construct hydro-power generation projects, storage structures must not exceed 12.35 MCM capacities. The construction of these storage structures on western rivers will have catastrophic consequences for Pakistan
8
as reduced flows resulting from filling of these dams during low flow season could destroy the rabi seaon crops in Pakistan (PILDAT, 2010). The average annual flows of major rivers in the basin are provided in Table 1.3. These flows represent the pre-treaty (1922-61) and post-treaty (1985-2002 and 2007-2010) situations. The flows show decreasing trends for both west and east flowing rivers. The average flow of eastern rivers into Pakistan was reduced by 75% to 92% during the years 1985-2002 and 2007-2010, respectively. Pakistan can utilize only residual flows from these east flowing rivers. However, these flows are variable and available only during the monsoon season. About 17% reduction in the average flow of the west flowing rivers is also observed. Climate change and its variability may cause reduction in flow of the west flowing rivers (Ahmad, 2009). However, the upstream interventions could also be the cause of reduced flows. Table 1.3 Average flows in major rivers of the Indus basin before and after IWT River
Rim station Average Annual Average Annual Flow Flow (1922-61) (1985-2002) (km3) (km3) Kalabagh 114.4 94.1 West Indus 28.3 23.7 flowing Jhelum Mangla rivers Chenab Marala 31.9 24.5 Ravi Below 8.6 4.0 East Madhopur flowing Below 17.2 2.2 rivers Sutlej Ferozepur Total 200.4 148.5
Average Annual Flow (2007-10) (km3) 101.9 19.3 23.9 1.1 0.8 147.0
Source: (Khan, 1999; GOP, 2011; IUCN, 2011) An integrated approach to manage transboundary water resources can lead to development and revision of water treaties between states, and prevent potential conflicts and resolve disagreements. Provision of objective information to facilitate negotiations between various fellow states requires tools that can monitor spatial and temporal changes in water demand and water use over vast areas.
1.4
Transboundary aquifer
Continues population growth in the Indus Basin resulted in mounting pressure on increased food production. It is estimated that to feed the increasing population, 40% more food will be required by the year 2025. The surface water resources in the basin (especially in Pakistan) are limited and variable. The upstream interventions by India have also threatened the timely availability of surface water downstream. Reduction of storage facilities in Pakistan can result in up to 50% shortfall of crop water requirements by the year 2025(Alam and Bhutta, 1996).
9
The Indus Basin aquifer has large groundwater reserves. The development of this reserve started 30 years ago (Sarwar, 2000). Inadequate and variable surface supplies forced farmers to start irrigating with groundwater. Local and readily available groundwater makes irrigation more productive compared to surface water irrigation. Currently 40–50 % of agricultural water needs in the basin are met through groundwater (Sarwar and Eggers, 2006). Both in Pakistan and India large numbers of irrigation wells have been added every year, which resulted in 20-30% increase in groundwater abstractions during the last 20 years (Qureshi et al., 2010b). The groundwater withdrawal exceeds annual recharge causing imbalance in groundwater reserves. This process is accelerating in the province of Pakistani and Indian Punjab, Haryana and Rajasthan states(Shah et al., 2000). Groundwater is also used in conjunction with surface water. Conjunctive use is in practice on more than 70% of the irrigated areas within the Indus Basin(Qureshi et al., 2010b). Figure 1-5 shows that the 29% more area came under conjunctive irrigation during the last 20 years. These values represent the irrigated area in Punjab, Sindh and Khyber Pakhtunkhwa (KPK) provinces of Pakistan only. The situation in the Indian part of the basin is not different. Sustainability of major crops in the basin is now heavily dependent on groundwater. Uncontrolled and unregulated use (over exploitation) of groundwater in many areas of the Indus Basin resulted in saltwater intrusion into the aquifer (Kijne, 1999). Water yield of wells is declining and pumping cost is increasing due to deepening of the water table. Salinization associated with the use of poor-quality groundwater for irrigation has raised the severity of the problem (Qureshi et al., 2010a).
Figure 1-5 Annual trends of surface, ground and conjunctive water use in the Pakistani part of the Indus Basin
10
1.5
Data availability and sharing issues
A major obstacle in transboundary river basin water resources management is that the fundamental information on water flows, sources of water, and water demand is either missing or not accessible. The downstream riparian countries depend on their upstream neighbors for data collection and sharing. If this does not happen the downstream countries cannot prepare themselves to cope with floods and droughts or generate hydropower (Zawahri, 2008). This problem is more severe in basins in developing countries and the Indus Basin is an example. The vastness of the basin, budget constraints, political distrust, and its transboundary nature is a hindrance in establishing a comprehensive measurement network. Rainfall is an important component of the hydrological cycle but cannot be used in water management studies if measurement stations are scarce. In the case of the Indus Basin, less than four rain gauge stations are available for an area of 10,000 km2, which is insufficient for basin scale studies. The situation is even worse for in-situ soil moisture and evapotranspiration measurements. There is no flux station (to the author’s knowledge) available in the whole basin that provides continuous information on soil moisture status and actual evapotranspiration. The same applies for land use and crops grown in the entire basin. Some spatial databases are available describing the land uses of the basin but these databases are outdated or/and coarse with little detail on cropping patterns. River flows are monitored by the Water and Power Development Authority (WAPDA) of Pakistan. The discharge data is collected by a network of manual and automated observation stations installed at various points along the rivers especially in upstream areas. It is the only data available. Apart from collection, the accessibility of the data is also not straightforward. Acquisition of long term data series is difficult and involves a series of bureaucratic permissions. Accessibility is also hampered by the fragmented structure of governmental institutions designated with various water management roles and tasks. There is seldom any coordination among the departments involved in data collection and system planning. Due to lack of coordination and institutional problems, the data collected by these departments is of little use to decision makers and water resources planners in order to manage water flows effectively. Recently, the Provincial Irrigation Departments (PIDs) and the WAPDA initiated several projects to integrate databases. The successful completion of these projects will be a big step forward in achieving a comprehensive hydrological database. Moreover, the continued political turmoil and distrust between the two countries make it difficult to carry out basin scale integrated water resources management. There is also no trust in the quality of data shared between the two countries because it cannot be verified, and is politically biased. Regional cooperation on water issues and comparisons between fellow states in a basin requires a standardized description of the water flows and the emerging processes. One possible way of promoting a climate of confidence and favorable political will is by building adequate databases for water accounting on basin scale. It must be admitted that nobody has reliable data related to water resources conditions, as data
11
gathering and definition procedures greatly differ. Satellite data can resolve this dilemma (Bastiaanssen, 2000).
1.6
Redefinition of water resources management
Water scarcity is not only due to the physical shortage of water but also to poor management; or in the words of the World Water Council: Today’s water crisis is not about having too little water to satisfy our needs. It is a crisis of managing water so badly that billions of people and the environment suffer badly (Cosgrove and Rijsberman, 2000). Conventional water resource planning and management is mainly focused on blue water (water in streams, rivers, aquifers, lakes and reservoirs).There is a need to incorporate rainfall, especially in arid and semi arid basins, that infiltrates naturally into the soil and on its way back to the atmosphere in the form of evapotranspiration (green water)(Falkenmark and Rockström, 2006).Managing non-beneficial evaporation will result in a significant reduction in water use that can be re-allocated to other users. Planning and management of surface water resources is important. However, under the current situation where groundwater utilization is upto 50% of total irrigation supplies, there is a need to plan and manage groundwater resources to maximize basin level efficiency. Groundwater can be a primary buffer against drought, as its response to short term climate variability is slower than surface water systems. The mismanagement of this buffering system can lead to serious impacts on the environment and ultimately on food security (Ahmad, 2002). Sustainable management of groundwater is considered a more serious challenge than development (Shah et al., 2000). The challenge is complex and management is not straightforward. The absence of a robust knowledge base is a major hindrance to sound management. In general, the integrated system, correctly managed, will yield more water at more economic rates than separately managed surface and groundwater systems.
1.7
Remote sensing in hydrology and water management
Transboundary river basin water resources management gains trust and faith if rainfall, diverted water, soil moisture, crop evapotranspiration and vegetation growth data is (i) collected at a range of scales, (ii) adequate, and (iii) available and accessible throughout the basin. Hydrologists cannot (in a relatively short time span) diagnose the water flow path at the regional scale if hydrological data is poor or incomplete. It requires considerable time to thoroughly quantify or model the hydrological processes and cycles in a river basin using other parties’ data. Satellite data is an attractive alternative for data required by hydrological models and to provide spatial information to decision makers. Satellites provide objective data for database building (for various applications, see Table 1.4), which is politically neutral and cannot be manipulated. Satellite measurements reflect the land surface features and the observable landscape patterns resulting from socio-economic development, prevailing jurisdiction, agricultural practices, hydrological processes, and irrigation management. Because they are direct measurements, satellite observations are often more reliable than secondary data. For instance, the irrigated area in the Gediz River Basin in Western Turkey appeared from the satellite images to be 60% larger than from the secondary data obtained 12
from governmental statistics (Bastiaanssen and Prathapar, 2000). It is obvious that if such types of secondary data are used to establish intra-basin water cooperation, disputes and conflicts can potentially worsen and trust will fade away. Table 1.4Satellite measurements for possible applications in transboundary river basins.(Source: Bastiaanssen and Prathapar, 2000) Discipline
Application
Hydrology
Snow cover, rainfall, soil moisture, evapotranspiration
Agriculture
Irrigated areas, rainfed areas, crop identification, biomass growth, crop yield, irrigation performance
Environment Forest area, wetlands, rangelands, water logging, salinization, water quality Geography
Digital elevation, land slope, land aspects, land cover, land use
There are large numbers of satellites in the earth’s orbit which are being used to acquire information on hydrological and biophysical parameters. Pixel size varies from few metres to kilometres and temporal resolution varies from 3-hours to months. For example, the Tropical Rainfall Measuring Mission (TRMM) provides 3-hour rainfall rate estimates at 25 km pixel resolution since 1997. The Advanced Microwave Scanning Radiometer-Earth Observing System (AMSR-E) observes atmospheric, land and oceanic parameters. Daily soil moisture estimates at 25 km pixel resolution are available through AMSR-E. Daily evapo transpiration can be estimated using AMSR-E and MODIS satellites at 1 km grids. NDVI, LAI, land use, albedo, biomass at 1 km resolution can also be estimated from MODIS, SPOT vegetation etc. Ground water levels can be estimated using the GRACE satellite that provides monthly changes in storage change at 400 km grids. Spatially distributed hydrological models are in use to compute rainfall-runoff processes, river flow, erosion and sediment transport, land-atmospheric interaction, water allocation planning, irrigation supply, groundwater recharge and ecological responses to land and water resources management. Beven and Fisher (1996) recognized that remotely sensed soil moisture, ET and snow cover estimates are necessary for scaling the hydrological processes in basin scale hydrological models. Many researchers have used satellite data in hydrological models in un-gauged or data scarce regions (Droogers and Bastiaanssen, 2002; Immerzeel et al., 2008b; Winsemius et al., 2008; Wipfler et al., 2011). Calibration and validation of these models need long term data series obtained from dense measurement networks. However, in the basins like Indus, such data is meager, thus causing a high level of uncertainty in the model results. Spatially variable information describing topography, crop types, land use, climatic data, and leaf area index, derived from remote sensing can be used for modeling across basins. This presents a new way to study the hydrological processes, water resources depletion, food security and environmental development in international river basins. It has opened a new protocol where central governmental bodies and internationally controlled agencies get uniform information. Remotely sensed information has a public domain status, and everybody can have access to raw satellite data due to the Earth Observing System with unlimited access to Data Active Archive Centers (DAACs). Federal Governments and the
13
UN can inspect land and water resources management issues, either by hiring their own experts or by involving commercial consultants.
1.8
The research justification
An integrated holistic approach to international river basin management is needed, in which the basin is accepted as the logical unit of operation. A multi-sectoral, integrated system, complemented by information sharing, transparency and wide participation is therefore best suited to encompass all these elements. Such an integrated system approach to evaluate the interaction between the hydrological processes in the mountains, river flow generation, water retention in reservoirs, groundwater pumping and agricultural water use in the Indus Basin is largely lacking. In the past, most scientific modeling research concentrated on the parts with well-established databases (e.g. Sarwar, 2000; Ahmad, 2002; Arshad, 2004; Habib, 2004; Hussain, 2011). These studies are valuable to test hypothesis and to construct local scale hydrological knowledge. However, a complete understanding of the hydrological processes can only be obtained if the research focus is to establish a solid basis for solving real life problems on the entire basin. A huge number of hydrological models are available to use in exploration of different hydrological processes. These models need input data that is limited or have inaccuracies. They must be estimated either by some relationship with physical characteristics or by tuning the parameters in order to have responses close to observed ones, a process known as calibration. Calibration of physically based, distributed models is complex given the limitations of data, the complexity of the mathematical representation of hydrological processes, and the incomplete knowledge of basin characteristics. Model calibration is usually based on the comparison between modeled and observed values from a few gauging stations. The problem of parameterization and lack of data for sound validation of modeling of large basins can be overcome by hydro-meteorological information from earth observation satellites. Land use, rainfall, soil moisture, water levels, total water storage changes, evapotranspiration, etc. are examples of data that can be obtained via satellites. These spatially distributed parameters can be used for distributed hydrological modeling and validation. There is no satellite dedicated to water management application. Various vegetation and water parameters are derived from different space borne spectral radiometers. Complex algorithms are used to transform original satellite measurements into spatially and quantified pixel information. Pixels need to be trained and made intelligent by scientists because spectral radiance (W m-2 sr-1 m-1) is a signal only. Uncertainty also exists in this conversion process. The overall objective of this thesis is “The development of methodologies to efficiently utilize satellite measurements in hydrology and to model the conjunctive water use for data scarce river basins”. This study is unique in that it combines ET and rainfall with water available from reservoirs to determine water balances and determine water flows with complex patterns of conjunctive use. The following innovations will result from this study: 14
(1) Simple calibration and validation techniques for spatial data in data scarce conditions will be developed. (2) A distributed pixel knowledge base on water flow paths and groundwater interactions for the entire basin will be constructed. (3) A hydrological model suitable for providing near real time data and capable of testing alternative solutions to combat over-exploitation and verify IWT agreements will be designed. This PhD study will prove that intelligent pixels combined with hydrological models will generate reliable data to effectively deal with water allocation issues such as (i) tempered groundwater exploitation, (ii) definition of volumetric water rights, including compulsory return flows, (iii) efficient irrigation systems, and (iv) vulnerability to climate changes. While climate change is on the international radar screen, the real challenge is to improve current manage of water resources, and to control conjunctive use in a sustainable manner.
15
16
2 Study area 2.1
Geographical description
The study area selected for this study is the Indus Basin which lies between latitude 24°38′ to 37°03′ N and longitude 66°18′ to 82°28′ E. The Indus Basin is located in four countries (Figure 2-1). The lifeline of the Indus Basin is the Indus River that traverses China (upstream), Afghanistan, India and Pakistan (downstream). The total size of the basin is 1.162 million km2. The largest area of the basin is in Pakistan (53% of total). The area in India is 33% followed by China and Afghanistan with 8% and 6%, respectively. Elevations range from 0 to 8600 m above mean sea level (a.m.s.l). The Basin has complex hydrological processes due to variability in topography, rainfall, land use, and water use.
Figure 2-1 Location of the Indus Basin showing main tributaries and provinces/states of different countries in the basin. 17
2.2
Hydro-climatology
The climate of the basin varies spatially and is characterized by large seasonal fluctuations in temperature and rainfall. The major part of the basin is dry and located in arid to semi arid climatic zones. The upper (northern and north-eastern) parts have harsh winters with significant snowfall while the middle and lower parts have comparatively mild winters but hot summers. The average annual rainfall varies from less than 200 mm in the desert area to more than 1500 mm in the north and north-east parts of the basin. The 30 years (1961-90) average reference crop evapotranspiration (ETo) varies between 650 mm and 2000 mm in the northern parts and southern desert areas of the basin, respectively. These values were obtained from the International Water Management Institute (IWMI) world water and climate atlas (http://www.iwmi.cgiar.org/WAtlas/Default.aspx). The temporal variation of rainfall and ETo within the year also varies markedly (Figure 2-2). The ETo is higher during the months of May and June, corresponding with the prerainy season. Most of the rainfall occurs during the months of July, August, and September.
Figure 2-2 Monthly variation of average rainfall and reference evapotranspiration rate (ETo) in the Indus Basin. There are two sources of rainfall in the Indus Basin: the Monsoon and the Western Disturbances. The former takes place from June to September and the latter from December to March (Lang and Barros, 2004; Bookhagen and Burbank, 2006). The Monsoon season is caused by moist air currents from the Arabian Sea and Bay of Bengal. Monsoon rainfall occurs mainly due to heat difference of the land and sea. The heat difference creates pressure gradients causing wind fluxes from ocean to land (Muslehuddin et al., 2005). The moist air from the ocean moves towards the north, passing through the hot basin plain (Houze et al., 2007). Most of the rainfall in summer is due to this phenomenon, causing intensive convective rainfall (Singh and Kumar, 1997). It is intensive in the months of June, July and August. 18
The weather systems responsible for winter rainfall are mid latitude Western Disturbances (Thayyen and Gergan, 2009). They originate over the Caspian Sea and move from the west to east (Singh and Kumar, 1997).These are formed due to large scale interaction between the mid latitude and the tropical air masses. The interaction process results in the formation of westerly troposphere synoptic scale waves. These disturbances cause stratiform rainfall. The orographic effect may cause intensification, resulting in extensive cloudiness, heavy precipitation and strong winds. However, sometimes their movement slows down causing local heavy snowfall over the hilly areas (Dimri, 2006).
2.3
Indus river, major tributaries and doabs
The Indus River originates in Mount Kailash in Tibet (China) on the north side of the Himalayas at an altitude of 5,486 m (Jain et al., 2007). The Indus is fed by 24 tributaries with eight as major tributaries. The Jhelum, the Chenab, the Ravi, the Sutlej and the Beas Rivers are east flowing tributaries, while the Kabul, the Gomal and the Gilgit Rivers flow west and north, respectively. The Jhelum River originates in the upper end of Kashmir valley and joins the Chenab River near Trimmu barrage in Pakistan. The origin of the Chenab is in the Himalayas and flows into the Himachal Pradesh (India) and Jammu and Kashmir states. Further down, the Chenab enters Pakistan upstream of the Marala barrage. The Ravi River originates near the Kangra district of Himachal Pradesh and joins the Chenab in Pakistan. The Sutlej River arises from the lakes of Mansarover and Rakastal in the Tibetan Plateau at an elevation of about 4,570 m. The Sutlej joins the Chenab at Panjand (Pakistan). The Beas River originates in the Rohtang Pass in Himalayas at an elevation of 3,960 m and joins the Sutlej above Harike in India before entering into Pakistan. The Chenab then flows into the Indus above Guddu barrage (Pakistan). The Gilgit River arises in the northern areas of Pakistan with upper reaches mostly glaciated and covered with permanent snow. The Kabul River originates in the south-eastern slopes of the Hindu Kush range in northern Pakistan. It flows through the Chitral valley of Pakistan and then enters Afghanistan to meet the Indus further down, above the Kalabagh barrage near Attock in Pakistan. All these tributaries of the Indus are generally fed by snowmelt and monsoon rains in the summer (85%) and partially by rains in winter (15%). The average seasonal flows of the major rivers in the Indus Basin with their source of origin are given in the Table 2.1. Table 2.1 Average seasonal flows of the Indus River and its tributaries Rivers
Indus Jhelum Chenab Ravi
Origin
Mount kalash, Tibet(China) Jammu & Kashmir state Himachal Pardesh Himachal Pardesh
Length Catchment area (km) (km2)
Average flow 3 -1 (km yr )
3,180
288,000
83.15
816
39,200
28.7 (Mangla)
1,232
41,760
29.0 (Marala)
880
24,960
4.46 (Madhopur)
Major Reservoirs
(Tarbela) Tarbela Wular, Mangla Salal, Baglihar Thein
19
Beas
Himachal Pardesh Mount kalash, Tibet(China) Hindukush range Lake Shandur, Pakistan
Sutlej Kabul Gilgit
464
9,920
16.0 (Mandi)
1,536
75,369
18.0 (Ropar)
700
12,888
21.4 (Warsak)
Pong, Pandoh Bhakra, Nangal
Sources: (Thatte, 2008; FFC, 2009; ICID, 2009) The Indus plain consists of relatively flat zones between the Indus River and its major tributaries i.e. Jhelum, Chenab, Ravi, Beas and Sutlej. Each flat zone is called a doab, meaning a land bounded by two rivers (Thatte, 2008). There are five doabs in the Indus Basin namely the Thal doab (land between the Indus and Jhelum rivers), the Chaj doab (between the Jhelum and Chenab rivers), the Rechna doab (between the Chenab and Ravi rivers), the Bari doab (between the Ravi and Beas rivers) and the Bist doab (between the Beas and Sutlej rivers). These plains produce little runoff compared to the hilly areas which contribute the major portion of the runoff. Table 2.2 summarizes the total and irrigated areas in each doab in the basin as well as Pakistan’s part of the basin. Table 2.2 The doabs in the Indus Basin and area under irrigation in each doab No
Basin’s Area Total Irrigated area area (mha) (mha) 1 Thal Indus, Jhelum 3.2 1.25 2 Chaj Jhelum, Chenab 1.05 0.85 3 Rechna Chenab, Ravi 3.12 2.80 4 Bari Ravi, Beas 3.87 3.50 5 Bist Beas, Sutlej 1.02 0.83 Sources: (Kureshy, 1977; Ullah et al., 2001; Qureshi et Bastiaanssen, 2010)
2.4
Doab
Encompassing rivers
Pakistan’s Area Total Irrigated area area (mha) (mha) 3.2 1.25 1.05 0.85 2.97 2.29 3.01 2.73 − − al., 2002; Cheema and
Groundwater
A basin level study conducted by WAPDA Pakistan in 1965 described the nature of the aquifers in the basin. According to WAPDA (1965), “the Indus plain is underlain by deep, mostly over 300 m deposit of unconsolidated, highly permeable alluvium consisting primarily of fine to medium sand, silt, and clay. Fine-grained deposits of low permeability generally are discontinuous so that sands, making up to 65 to 75 percent of the alluvium, serve as a unified, highly transmissive aquifer”. The use of groundwater for irrigation and low levels of replenishment of the aquifers resulted in high levels of depletion. The groundwater within the basin varies spatially in terms of its water table and water quality, depending on usage (agricultural and domestic). Before inception of irrigation systems in the basin, the groundwater table varied between 20-30 m. Recharge from earthen canals and irrigated fields resulted in a significant rising of the water table in certain 20
locations, while the conjunctive use of ground water with surface water has resulted in lowering of the water table at other areas. The seasonal fluctuations of water table before and after monsoon for the year 2002 are provided in the Figure 2-3. These water table maps were provided by the International Water Logging and Salinity Research Institute (IWASRI).
Figure 2-3 Pre and post monsoon depth to water table in irrigated areas of the Pakistani part of the Indus Basin for the year 2002. The water table depth in the irrigated areas of the Punjab and Sindh provinces is more than 30 m before monsoon except in a few pockets. Some areas of the aquifer have depths even more than 120 m. A rise in the water table is observed after the monsoon season especially in the Sindh province. These areas are under serious threat of water logging because water rises significantly after rains. In general, a continuous trend of water table decline is observed, especially in the Punjab province, which points to a serious imbalance between abstractions and recharge. Figure 2-4. for example, shows how the areas (with groundwater table depth of 30 m or more) increased between 1982 and 2002 in different canal commands in the Pakistani part of the basin. The canal command in the Punjab province showed a significant increase in the area with water table depths of 30 m or more over a 20 year period (1982-2002). The canal commands in the Sindh province showed a reduction in areas with a 30 m depth to water table.
21
Figure 2-4 Change in area with groundwater table depth 30 m and more between 1982 and 2002 in different canal commands of the Indus Basin. The era of unlimited dugwell and tubewell installations has encouraged farmers to augment shortages in surface water with groundwater (Shah et al., 2000). Excessive pumping resulted in deteriorating groundwater quality and diminishing phreatic surfaces across the Indus Basin. Poor quality groundwater and water logged soils occur in the downstream areas. A persistent flow of about 12.3 km3 is required below Kotri barrage (the last gauged structure on the Indus) to meet environmental flow requirements of the river, reduce salinity, and control sea water intrusion (PILDAT, 2003). Flow can be as low as 0.36 km3in drought years, to as high as 113 km3 in wet years.
2.5
Agriculture and cropping pattern
The Basin provides food for 200 million inhabitants. Irrigated agriculture is practiced in large parts of the basin (~22.6% of total area) to meet food requirements. Rainfall is not sufficient to meet the crop water requirements. Monthly crop water requirement (CWR: difference between ETo and effective rainfall) for two selected stations (Lahore and Hyderabad) is provided in Figure 2-5. Hyderabad is located in a relatively drier part of the basin with low rainfall and higher ETo values resulting in higher CWR as compared to Lahore. Lahore receives sufficient rainfall especially in the monsoon; thus limiting CWR in the months of July and August.
22
Figure 2-5 Monthly variation in crop water requirement at two selected location (Hyderabad and Lahore) representing the lower and middle part of the Indus Basin. There are normally two agricultural growing seasons: the rabi covering November, December, January, February, March, April; and the kharif covering May, June, July, August, September and October. Rainfed agriculture is practiced in upstream parts of the Indus Basin. “Savanna deciduous” (11.1%), “pastures deciduous alpine” (6.7%), “pastures deciduous” (6.5%), and “bare soil” (6.3%) are other dominant land use classes in the basin(Cheema and Bastiaanssen, 2010). Cropping pattern is defined as the sequence in which crops are grown in a given area over a period. A specific cropping pattern is in practice in the basin and farmers rarely change it. The growing season is sufficiently long for two crops and double cropping is widely practiced. There are varieties of crops grown, but wheat is the dominant crop in rabi and rice and cotton in kharif. There are also tracts of sugarcane that is a full year crop. Orchards are also grown on 3.6% of the basin area mixed with other crops (Figure 2-6 ). Seasonal fodder crops are also grown to meet the needs of livestock. Historical data show no significant change in cropping patterns.
Figure 2-6 Rice and Orchard fields in the irrigated areas of the Indus Basin
23
24
3 Land use and land cover classification in the irrigated Indus Basin using growth phenology information from satellite data to support water management analysis Chapter based on: Cheema, M.J.M. and Bastiaanssen, W.G.M., 2010. Land use and land cover classification in the irrigated Indus Basin using growth phenology information from satellite data to support water management analysis. Agricultural Water Management 97, 1541-1552.
3.1
Introduction
Advanced modeling of hydrological processes in vast river basins and continental scale land-atmosphere interactions requires land cover and land use information. Note that land use and land cover are not interchangeable. The term land cover defines the physical surface conditions. Examples of land cover classes are water, bare soil, grass, crops, forests etc. Land use reveals the type of application that humans have created to their own benefits. Examples of land use variation for the same land cover class are industrial settlements vs. residential areas (land cover: build up), irrigated vs. rainfed cropland (land cover: cropland), reservoirs vs. natural lakes (land cover: water), recreation vs. pastures (land cover: grass), timbering vs. environment (land cover forests). In case of a mixture of land cover and land use classes on the same map, it is common to refer to Land Use and Land Cover (LULC) classes. LULC information is often used to define the physical land surface properties such as curve numbers (for runoff), surface roughness (for evapotranspiration ET), albedo (for ET), vegetation cover (for biodiversity), rooting depth (for available soil moisture), drought resistance, etc. Hence, modeling of the water balance and preparation of water accounting (Molden, 1997) requires LULC to be known. The process of water accounting is generally based upon the amount of water used by different land uses. A number of frameworks have been developed for water accounting (Niblack and Sanchez, 2008; Turner et al., 2008). These frameworks distinguish beneficial and non-beneficial water use by land use classes. This is only possible with accurate LULC classification. Several global scale land cover databases have been developed since the early nineties. e.g. Global Land Cover Characteristics (GLCC) database carried out under the flag of the International Geosphere-Biosphere Program (IGBP) ,International Water Management Institute (IWMI) land cover database, Global Land Cover (GLC) 2000 and Global land Cover (Glob Cover) developed by European Space Agency (ESA). GLCC database was developed using 1 km resolution Advanced Very High Resolution Radiometer (AVHRR) data of monthly Normalized Difference Vegetation Index (NDVI) using unsupervised classification and subsequent refinements were carried out to develop land cover database for the year 1992-93 (Loveland et al., 2000). IWMI database was based on the year 2001-02 and used Moderate Resolution Imaging Spectroradiometer (MODIS) 500 m NDVI datasets. It was developed using unsupervised classification and refined by ground truth data. GLC2000 was global land cover database developed using 9 months Satellite Probatoire d’Observation delaTerre (SPOT) vegetation data for the year 1999-2000. Glob Cover was
25
developed recently using the 300 m resolution Medium Resolution Imaging Spectrometer (MERIS) satellite data for the year 2005-06 (Arino et al., 2008). The mapping accuracies of these global products are reasonable (i.e.63% – 83%) but these were developed for use in global climate studies. All these data sets are most useful for analyzing general land cover patterns at a continental or large scale. These global datasets are, however, inappropriate to support basin water management analysis and applications because they lack details on land use information. Such global databases cannot distinguish specific crops and only detects dominant land covers leading to a large percentage of mixed classes with natural vegetation (Portmann et al., 2010). Since within irrigated land uses, different crops are grown which have specific crop water requirements. Therefore, more specific crop based classifications are needed to implement proper water allocation plans. General land cover information by means of a few classes, without the land use functioning is thus insufficient and novel techniques to determine land use classes need to be developed (Meyer and Turner, 1994; Schwarz and Zimmermann, 2005). Therefore, in this paper a methodology has been presented to derive LULC for the vast Indus Basin. This is done by integrating satellite derived information on NDVI time series with ground information and expert knowledge on the growing patterns (phenology) of the crops. This information is further applied to identify different crop rotations in growing seasons in order to get real pattern of water use. The goal is to develop an up to date, regionally consistent and detailed LULC map with affordable efforts that could also be applied to obtain LULC information for other basins with water resources problems in the world. This database will be used in further hydrological studies on the Indus Basin. The Indus Basin was selected for this study because the water resources of this international river basin are under pressure. There is not enough water to supply all water use sectors with sufficient quantity. Although good efforts have been undertaken (Habib, 2004), there is a genuine interest to better understand the hydrology and water management issues of the Indus Basin.
3.2
Study area
The international Indus Basin is located in four countries (Figure 3-1). The basin lies in between24° 38′ to 37° 03′ N latitude and 66° 18′ to 82° 28′ E longitudes. The life line of the Indus Basin is the Indus River that traverses China, Afghanistan, India and Pakistan, when moving from upstream to the downstream end of the basin. The total size of the basin is 116.2 million hectares (mha). The vast area of the basin is located in Pakistan (53% of total). The area in India is 33% followed by China and Afghanistan with 8% and 6%, respectively. The basin is bounded by the Karakoram and the Hindukush ranges on the north, the Himalayas on the north-east, the Sulaiman and the Kirther ranges on the west and the Arabian Sea on the south.
26
Figure 3-1 Location of Indus Basin and provinces of different countries in the basin with information on main tributaries of the Indus river.PK stands for Pakistan and IN for India. The Indus river originates from Mount Kailash in Tibet (China) on the north side of the Himalayas at an altitude of 5,486 m (Jain et al., 2007). The Indus river is fed from 7 major tributaries. The Jhelum, the Chenab, the Ravi, the Sutlej and the Beas rivers are the eastern tributaries, while the Kabul and the Gilgit rivers being western and northern tributaries, respectively (Figure 3-1). The Jhelum River originates from the upper end of Kashmir valley and joins the Chenab River near Trimmu barrage in Pakistan. Origin of the Chenab River is in the Himalayas and flows into Himachal Pradesh (India) and Jammu and Kashmir state. Afterwards, the Chenab River enters in Pakistan at upstream of Marala barrage. The Ravi River rises near the Kangra district of Himachal Pradesh and joins the Chenab River in Pakistan. The Sutlej River rises from the lakes of Mansarover and Rakastal in the Tibetan Plateau at an elevation of about 4,570 m. The Sutlej River joins the Chenab River at Panjand (Pakistan). The Beas River originates from Rohtang Pass in Himalayas at an elevation of 3,960 m and joins the Sutlej River in India before entering into Pakistan. The Chenab River then flows into the Indus River near Guddu Barrage (Pakistan). The Gilgit River arises in the Northern areas of Pakistan with upper reaches mostly glaciated and covered with permanent snow. All these tributaries of the Indus River are generally fed with snow melt and monsoon rains in the summer (85%) and partially with rains in winter (15%).
27
The climate of the basin varies spatially and is characterized by large seasonal fluctuations in temperature and rainfall. The upper (northern and north-eastern) parts experiences harsh winter with significant snowfall while the middle and lower part has comparatively mild winters but hot summers. The average annual rainfall varies from less than 200 mm in the desert area to more than 1500 mm in the north and north-east of the basin. Two growing seasons (rabi and kharif) normally prevails in the basin. The rabi season comprises months of November, December, January, February, March and April while May, June, July, August, September and October represents kharif season.
3.3
Methodology
Every LULC class has particular phenological variations of vegetation cover throughout the year. Vegetation cover describes the foliage covered fraction of a natural land surface. The duration of various cropping seasons ranges from 90 days to 150 days and multi-temporal images of vegetation cover capture these dynamics. Vegetation indices are based on differential absorption, transmittance, and reflectance of spectral radiance by the vegetation in the red and near-infrared electromagnetic radiation. Green leaves absorb most of the radiation in the visible part and reflect in the near-infrared spectrum range. The green leaf density increases the photosynthetic activity. The unique behavior of green leaf development in terms of duration and peak of phenological stages in various agroecosystems makes these ecosystems distinguishable using time series analysis. The most common vegetation index is the NDVI derived from the visible and near infrared channel reflectance (Tucker, 1979). The NDVI time series can be helpful in vegetation detection (Hansen et al., 2000; Wardlow and Egbert, 2008). For the current study, multitemporal NDVI images were obtained from the Vegetation (VGT) sensor, which is on board of the SPOT satellite. The spatial resolution of the sensor is 1 km while its 2250 km swath width enables the sensor to acquire data on a daily basis. The daily global coverage and spatial resolution, makes it a very suitable sensor for large basin vegetation mapping (Mucher and de Badts, 2002). The SPOT VGT sensor has four spectral bands: blue (0.43-0.47 μm), red (0.61-0.68 μm), near-infrared (NIR: 0.78-0.89 μm) and shortwave infrared (SWIR: 1.58-1.75 μm). The red and NIR bands are used to characterize vegetation. The raw NDVI data set covering the Indus Basin for the year 2007 were downloaded from the Processing and Archiving Image Center, hosted by VITO, the Flemish Institute for Technological Research, Belgium (http://free.vgt.vito.be). This site provides geometrically and radiometrically corrected 10day synthesis SPOT VGT NDVI (V2KRNS10) products. Thirty six NDVI images from January 2007 to December 2007 were obtained in order to include full phenological information of one complete annual cycle. The SPOT vegetation data were originally available in Digital Numbers (DN) that were then converted into NDVI using the following equation. 3.1 Since no ground observations were available initially, an unsupervised classification (e.g. Giri and Jenkins, 2005) was performed on the stacked image. Unsupervised classification identifies clusters by their spectral similarities (in this case 36 spectra; one for each 10 days) and allows the feature space to segment into similar spectral clusters (Rashid, 2007). 28
Different clustering methods are available for unsupervised classification like k-mean method and Iterative Self Organizing Data Analysis Technique (ISODATA). ISODATA is based on Euclidean distance, in which spectral distances between candidate pixels are compared to each cluster mean. A pixel is assigned to the cluster whose mean is closest to the candidate pixel. New cluster centers are computed by averaging the locations of all the pixels assigned to that cluster (Campbell, 2002). An ISODATA technique was used with 95% convergence threshold for land use classification. The classification was refined by applying expert knowledge of the cropping patterns adopted in the basin (Figure 3-2). Multi-cropping systems can be observed in irrigated areas. The knowledge of crop phenology helped to identify crops with specific NDVI temporal profiles. The beginning of a growing season for a particular crop was considered when there was a significant increase in NDVI. The onset of the NDVI is a result of increased photosynthetic activity. Similarly, a decrease in the NDVI reflects the end of growing period. Hence, NDVI time profiles can be used to identify multi-crops during the growing season.
Figure 3-2 Cropping calendar adopted in the Indus Basin. Rabi expresses the winter crop and kharif the summer crop. The possible effects of physical conditions (e.g. temperature, elevation and slope) on land use classes were also studied. For this, a 90 m resolution Digital Elevation Model (DEM) was obtained from the Shuttle Radar Topographic Mission (SRTM) database (Jarvis et al., 2008). The average temperatures were attained from temperature datasets (period 1961-90) developed by the Climatic Research Unit, University of East Anglia, United Kingdom (New et al., 2002). The mean temperature, elevation and slope were extracted for individual class to examine the spatial variability of land use due to physical conditions. Four types of accuracy assessment studies were performed in order to assess the quality of the generated LULC map. First is the classical error matrix approach, which uses independent classification and reference data to have a precise knowledge of the ground situation (Latifovic and Olthof, 2004). A ground truthing survey was conducted during September-October to capture peak kharif cropping season and in January to coincide with the rabi cropping season conditions. Due to the vast dimension of the Indus Basin, ground truthing was focused on the middle and lower reaches that have different agro-ecological regions. The class with 70% dominance of certain land use was selected. (e.g. if 70% of the observed area was wheat, then the area was classified as wheat). Existing global and regional (IGBP, IWMI,GLC2000) land cover maps were compared with the newly developed map as a second accuracy assessment test. One map was produced by Boston University, Department of Geography using MODIS land cover product MOD12Q1 V004 based on 17 classes IGBP scheme in 2004 (MODIS, 2004). The 29
map was accessed from http://duckwater.bu.edu/lc/mod12q1.html. The other map was developed by (Thenkabail et al., 2005) using MODIS 500 m NDVI time series for the year 2001, and GLC2000 (Agrawal et al., 2003) used 1 km SPOT-Vegetation NDVI time series. A subset covering the Indus Basin was extracted from those maps. The subset maps were compared after rescaling to a single 1 km resolution. For the sake of comparison, some classes were merged to make analysis easier. The third accuracy test was carried out by using local studies. These local studies (Hussain, 1998; Bastiaanssen et al., 2003; Singh et al., 2006), focus on the southern and eastern parts of the Indus Basin. Furthermore, ancillary data was used for refinement as well as fourth accuracy test. Ancillary data included agro-ecological regions and cropping pattern maps developed by Pakistan Agricultural Research Council (PARC), Islamabad, district based Agricultural Statistics of Pakistan compiled by Ministry of Food, Agriculture and Livestock (MINFAL), Government of Pakistan, Food and Agriculture Organization (FAO) and International Centre for Integrated Mountain Development (ICIMOD) publications. The limitation in data collection from Indian side of the basin restricted to use data from Pakistan only. The area fractions of different crops grown in different administrative units reported by MINFAL and estimated by remote sensing were compared. The coefficient of determination (R2) was used to check the reliability of the estimates.
3.4 3.4.1
Results and discussion Phenology
Initially clustering of five classes was applied for the separation of the total Indus Basin into (i) water/ice, (ii) barren, (iii) shrubland /grassland, (iv) natural vegetation (forests) and (v) cropland. The division of the basin into these five classes provided a first understanding of spatial variability in land cover (Table 3.1). The class “Barren” appears to be a major class for all countries. The areal extent is followed by the class “croplands” for Pakistan and India. Afghanistan has a small area under “croplands”, while it is negligible in China. The area in the Chinese part of Indus Basin is located at a higher altitude where no agriculture is possible. The Chinese part is mostly comprises of “water/ice” and “barren” land covers. By absence of vegetation and the dynamics of snow and snowmelt processes, it is not straightforward to differentiate between “water/ice” and “barren” with the information used. Table 3.1 Major LULC classes and their spatial distribution among the countries of Indus Basin. LULC type Water/ice Barren Shrubland/grassland Natural vegetation Croplands Total
30
Pakistan Area (mha) 6.33 21.46 10.84 9.77 13.13 61.53
India Area(mha) 7.68 10.51 4.26 3.95 12.52 38.92
China Area(mha) 4.66 4.22 0.01 − − 8.89
Afghanistan Area (mha) 0.82 4.51 1.30 0.51 0.04 7.18
The number of classes was then increased to identify the land use. This resulted into a first round of 45 classes which provided the basis for further refinement and analysis. This classification was made by taking into consideration cropping calendar and dominance of a particular crop in the area. The class cropland was partitioned using NDVI temporal profiles and expert knowledge of cropping patterns. Some classes were merged on the basis of information obtained during ground truthing. NDVI profile similarities were also considered during this merging. This procedure reduced the number of classes from 45 to 27. The resulting mean NDVI time profiles of the final agricultural classes are shown in Figure 3-3. Figure 3-4 shows only natural land cover classes. These NDVI time profiles reflect the mean values for each individual class. A three period moving average filter is used to smooth the profile as described by Reed et al.(1994). It can introduce time lag which is avoided by using one period before and one period next from the time of analysis.
Figure 3-3 Mean NDVI temporal curves for irrigated and rainfed crops in Indus Basin for the year 2007.A moving average of 3 periods has been used to smooth the lines.
31
Figure 3-4 Mean NDVI temporal curves for forests, pastures and savannas in the Indus Basin for the year 2007. Figure 3-3 and Figure 3-4 confirm that these 27 clusters have a high degree of seperatability: none of the final classes has a similar phenology. The ISODATA clustering technique met the expected goal to separate phenological differences such as the start of growing season, the end of the growing season and growing length of a particular crop. The two peaks in one annual cycle (Figure 3-3) represent multi cropping systems with intensive irrigation practices. The start and end of these peaks distinguish major crop types grown in the study area. The class “irrigated rice, wheat rotation” has two distinct peaks, one at the end of February and other at the end of August. This matches with the cropping period of wheat from November to April (i.e. rabi) and rice from June to October (i.e. kharif). Moreover, the area in this class was completely cultivated with only a very small percentage of fallow land thus showing a higher NDVI from all other crop types. The rainfed crops have lower NDVI values and their timing of peak may differ due to pattern difference in natural climatic conditions as compared with irrigated area. The recession of NDVI curves for rainfed crops like “rainfed crops mixed cotton, wheat rotation/fodder” and “rainfed crops general” show similar trends as the one’s of irrigated crops but with lower NDVI values. The crop labeled as “rainfed crops wheat/grams” shows a different trend than other rainfed crops. Under this class, only rabi crops (wheat and grams) were grown. NDVI values for the rest of the year is flat showing no distinct vegetation.
32
Similarly, Figure 3-4 clearly differentiates between forests, pastures, savannas, snow and desert. The evergreen forests (both needleleaf and broadleaf) in the northern region of the Indus Basin have temporal stable NDVI values throughout the year. However, the class “forests deciduous alpine” shows a sudden decline in NDVI during the months of December until March. Presence of snow, during winter, has resulted in suppressed foliage cover. Emergence of new leaves in April, improves greenness and higher NDVI values are observed in the remaining months. The NDVI time series is effective to separate seasonal and permanent snow cover in the study area. Usually snow has very low NDVI values (< 0) because the near-infrared reflectance of water is smaller than for red reflectance. For permanent snow, the NDVI values are lower than zero for the whole year while the class “snow and ice temporary” has NDVI values lower than zero during the winter season only. The resulted LULC map is presented in Figure 3-5.
Figure 3-5 Land use and land cover map of the Indus Basin developed from 36 SPOTVegetation based NDVI values covering the annual vegetation phenological cycle of 2007. The international boundaries are superimposed on the map. The areal extent of the LULC classes is summarized in Table 3.2. Deciduous savanna with 12.94 mha is the most dominant land use type, followed by rainfed crops in general (11.71 mha). Deciduous savanna appears in the Kabul river sub basin and in the Himalaya mountain ranges. The dominant class in the Indus Basin is irrigated agriculture (23%), followed by pastures (16.2%), rainfed crops (16.1%), savannas (15.9%) and forests (10%). Irrigated rice, wheat rotation appears to be the most common irrigated crop (9.69 mha). An area of 3.25 mha of irrigated rice, wheat rotation class is located in Pakistan. This is 8% 33
higher than reported by MINFAL (2007). This can be explained by the fact that the gross area of one pixel is 100 ha (1 km×1 km) and not all land within on pixel consists of agricultural fields. Table 3.2 LULC types and their areal distribution in the Indus Basin
3.4.2
Effect of physical condition on LULC
The physical condition such as elevation, temperature and slope can affect the spatial distribution of land use classes as indicated by Mahajan et al. (2001), Wang et al. (2003) and Fang et al. (2005). They conducted their studies in an Indian watershed, USA and North China, respectively. Such effects of physical conditions are also significant in the Indus Basin. The elevation traverses longitudinally across the entire basin with maximum in the north and minimum towards south. The temperature is inversely related with elevation. Higher temperatures are usually observed in the plain of Indus Basin with very mild slope and average elevation 100 m to 300 m above Mean Sea Level (a.m.s.l). The different LULC classes with respect to orography are shown in Table 3.3.
34
Table 3.3 LULC classes distribution in Indus Basin with respect to (i) elevation a.m.s.l (ii) slope and (iii) temperature. No
Class name
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Snow and ice permanent Snow and ice temporary Bare soil Very sparse vegetation Pastures deciduous Pastures evergreen lowland Pastures deciduous alpine Savanna evergreen open Savanna evergreen closed Savanna deciduous Forests evergreen needleleaf Forests evergreen broadleaf Forests deciduous alpine Forests/cropland alpine Irrigated mixed cotton,wheat rotation/orchards Irrigated mixed cotton,wheat rotation/sugarcane Irrigated rice,wheat rotation Irrigated mixed rice,wheat rotation/cotton Irrigated wheat,fodder rotation Irrigated rice,fodder rotation Irrigated mixed rice,wheat rotation/sugarcane Rainfed crops wheat/grams Rainfed crops mixed cotton,wheat rotation/fodder Rainfed crops general Rainfed crops and woods Urban and industrial settlements Water bodies
16 17 18 19 20 21 22 23 24 25 26 27
Average elevation (m) 5220 4900 690 2000 450 1250 4450 1210 1250 3540 1085 1715 3370 2175 200
Average slope (Degree) 23.0 23.8 2.7 8.2 3.5 8.8 14.3 11.4 14.6 20.3 15.3 20.3 26.8 22.6 0.9
Average annual 24 h T (°C) -4.2 -2.7 24.6 13.6 24.2 19.1 2.2 19.8 16.9 4.5 16.6 13.8 4.0 9.2 25.7
225
0.8
24.0
210 165
0.6 0.7
26.0 25.6
175 100 150
0.6 1.2 2.4
25.6 26.5 25.3
185 200
0.6 1.2
26.2 27.2
375 840 350 430
3.5 7.9 2.4 1.6
24.7 21.0 23.5 24.3
Low air temperatures are observed at the higher altitudes in the basin. This region is dominated by snow which suppresses vegetation. Majority of LULC falls under the classes “snow and ice permanent” and “snow and ice temporary”. The class “snow and ice permanent” occurs at higher elevations (5220 m) and 23 degree average slope. This class covers most of the glaciated part of the basin. The low temperature allows snow to sustain permanently in the form of glaciers. An increase in temperature during summer is not sufficient to cause the snow to melt. The second class in this region is “snow and ice temporary”, that attains an average elevation of 4900 m with steep slopes (~24 degrees). At
35
this elevation, the temperature rises over the freezing point during summer. This causes the snow to melt, thus allowing vegetation to grow on the moist soils. Moreover, the slope for the second class is relatively higher which enhances snow melt. It provides a sound reason to divide snow and ice into two classes. Figure 3-6(a) & (b) better interprets the effect of temperature on NDVI values for these two classes.
Snow and ice temporary
Snow and ice permanent 15
15
1.0 Temperature
1.0 Temperature
NDVI
NDVI 0.9
0.9 10
10
0.8
0.8 0.7
5
0.7
5
0.6 0
r
r
em be
0.4
D ec
ob er ct
N ov
O
te
-5
em be
t
r m be
ly
us Au g
Se p
ne
Ju
Ju
il
ay M
ar c
Ap r
h
0.5 M
r
r
em be D ec
ob er ct
em be
O
N ov
t
r m be te
ly
us Au g
Se p
ne
Ju
Ju
il
ay M
h ar c
Ap r
M
0.4
NDVI
0.5
-5
Temperature C Ja nu ar y Fe br ua ry
0
NDVI
Temperature C Ja nu ar y Fe br ua ry
0.6
0.3
0.3 -10
-10
0.2
0.2 0.1
0.1 -15
-15
0.0
0.0 -20
-0.1
-20
-0.1 Months
Months
(a)
(b) Savanna deciduous
Pastures deciduous alpine 20
20
1.0 Temperature
1.0 Temperature
NDVI
NDVI 0.9
0.9
15
15
0.8
0.8
10
0.5 0 r
r
D ec em be
N ov em be
r
ob er O
ct
m be te
ly
us t
Ju
e Ju n
ay M
il Ap r
Au g
Se p
-5
h
ua r Ja n
0.3
ar c
y
0.4
r
r
em be D ec
em be
ct O
Se p
N ov
m be te
ob er
r
t
ly
us Au g
ne
Ju
Ju
il
ay M
Ap r
h ar c
ua ry
M
nu ar Ja
-5
Fe br
y
0.4
ua ry
0
0.6
5
0.2 -10
NDVI
0.5
Fe br
5
0.7
M
0.6
Temperature C
0.7
NDVI
Temperature C
10
0.3 0.2
-10 0.1
-15
0.0
0.1 -15
0.0
Months
Months
(c)
(d) Forests deciduous alpine
Forests evergreen needleleaf
30
1.0 Temperature
30
1.0
NDVI
Temperature
NDVI
0.9
25
0.9
25 0.8
0.8
20
20 0.7
0.7
0.5 0.4
5
15
0.6 NDVI
10
NDVI
Temperature C
0.6
Temperature C
15
10
0.5 0.4
5 0.3
0
0.3
Months
(e)
36
r
r
em be D ec
em be
r
ob er
m be
ct O
N ov
t
ly
us
Ju
ne Ju
ay M
il Ap r
h ar c
ua ry
0.2 0.1 0.0
Months
(f)
te
-10
Se p
0.0
Au g
-5
M
0.1
Fe br
y nu ar Ja
r
r
em be D ec
em be
ob er ct O
m be te
N ov
r
t us Au g
ly
-10
Se p
ne
Ju
Ju
ay M
il Ap r
h ar c
ua ry
M
nu ar Ja
-5
Fe br
y
0 0.2
Irrigated mixed cotton,wheat rotation/orchards
Irrigated rice,wheat rotation
40
1.0 Temperature
1.0
40
NDVI
Temperature 0.9
35
NDVI 0.9
35
0.8
0.8 30
30 0.7
0.6 0.5 0.4
15
25
0.6 0.5
20
0.4
15
0.3
0.3 10
10 0.2
0.2 5
0.1
0.0
Months
r
r
em be
O
D ec
ob er
em be
ct
m be te Se p
N ov
t
r
ly
Au g
us
ne Ju
Ju
il
ay M
h ar c
Ap r
nu ar Ja
ua ry
y
r em be
r
0.1
0
D ec
ob er ct O
em be
m be te
N ov
t
r
ly
us Au g
Se p
ne
Ju
Ju
il
ay M
h
Ap r
ar c
ua ry
M
Fe br
Ja
nu ar
y
0.0
M
0
Fe br
5
Months
(g)
(h) Rainfed crops wheat/grams
Rainfed crops mixed cotton,wheat rotation/fodder
40
1.0 Temperature
40
1.0
NDVI
Temperature 0.9
35
NDVI 0.9
35
0.8
0.8
30
30 0.7
0.6 0.5 0.4
15
25
0.6
20
0.5 0.4
15
0.3 10
0.3 10
0.2
0.1
Months
r De c
em be
r em be No v
r
ob er
m be
ct
te
O
t us
ly
ne
Ju
Ju
ay M
il Ap r
h ar c
Au g
Se p
Ja
nu ar
y
0.0 ua ry
0
r em be
r De c
ob er
em be No v
r m be
ct O
t us
te
Au g
Se p
ly
ne
Ju
Ju
ay M
il
h
Ap r
ar c M
Fe br
nu ar
ua ry
0.0 y
0
0.2
M
0.1
5
Fe br
5
Ja
NDVI
20
Temperature C
25
NDVI
Temperature C
0.7
(i)
NDVI
20
Temperature C
25
NDVI
Temperature C
0.7
Months
(j)
Figure 3-6 Average temperature and NDVI relationship for (a) snow and ice permanent (b) snow and ice temporary (c) pastures deciduous alpine (d) savanna deciduous (e) forests deciduous alpine (f) forests evergreen needleleaf (g) irrigated mixed cotton, wheat rotation/orchards (h) irrigated mixed rice, wheat rotation (i) rainfed crops wheat/grams (j) rainfed crops mixed cotton, wheat rotation/fodder. The class “pastures deciduous alpine” comes next at an average elevation of 4450 m a.m.s.l and 14 degrees average slope, followed by “savanna deciduous” (3540 m) but with more slope (20 degrees). Higher altitudes and severe temperatures in winter suppress vegetation. Therefore, Figure 3-6(c) & (d) show lower NDVI value in winter for these classes. NDVI gradually increases in summer due to rise in temperature, favorable for vegetation growth. “Forests deciduous alpine” and “forests evergreen needleleaf” are usually found at an altitude of 3370 m and 1085 m with average slopes of 27 degrees and 15 degrees, respectively. Low temperatures in the winter cause defoliation. Thus, NDVI for “forests deciduous alpine” in Figure 3-6(e) is lower in the winters, and improves significantly, in summer due to favorable temperature. The “forests evergreen needleleaf” grows at a lower altitude, and don’t become dormant. Temperature in these altitudes is mild and favorable for vegetation. Usually evergreen plantation is observed in these areas. The NDVI values in Figure 3-6(f) shows constant trend across the whole year. Irrigated agriculture is mostly concentrated in the plain areas of the Indus Basin with an average slope between 0.6 to 2.4 degrees. The elevation ranges from 100 m in lower Indus Basin to 225 m in the middle parts of the basin. Extensive food requirements of about one
37
billion inhabitants are met from this irrigated plain. The common food and cash crops are wheat, rice, sugarcane, cotton, orchards. The mild temperatures during the winter and higher temperatures in summer, allows two growing seasons. Different growth stages of crops need specific temperature. For example, wheat needs mild temperature during early growth stages while requires warmer climate at ripening. Figure 3-6(g) & (h) clearly depict these phenological variations of crops with temperature trends. Rainfed crops also help to fulfill the food requirements. The rainfed crops are usually grown at a higher altitude as compared to irrigated crops (180 to 840 m, a.m.s.l). The average slope ranges from 0.6 degree (flat areas) to 7.9 degrees. The rainfall varies spatially for different altitudes. The NDVI curves for “rainfed crops wheat/grams” and “rainfed crops mixed cotton, wheat rotation/fodder” are shown in Figure 3-6(i) & (j), respectively. The temporal changes in NDVI values with temperature for different growth stages of crops reinforce our argument to use phenological cycle to identify land use classes. 3.4.3
Accuracy assessment
The usability of the LULC classification for water management analysis depends on the reliability of the developed LULC map. A number of accuracy assessment approaches has been developed (e.g. Congalton, 1991; Friedl et al., 2000; Cihlar et al., 2003). In this study, a 4-step accuracy assessment was carried out. 3.4.3.1 Ground truthing As a first step, the error matrix approach described by Campbell (2002) is used to express the accuracy. An error matrix constructed by plotting LULC classes from the unsupervised classification algorithm against the LULC information gathered from ground truth data is shown in Table 3.4. Only classes with ground inspection data are included in the analysis.
38
Irrigated mixed cotton,wheat rotation/sugarcane
Irrigated rice,wheat rotation
Irrigated mixed rice,wheat rotation/cotton
Irrigated wheat,fodder rotation
Rainfed crops wheat/grams
Rainfed crops mixed cotton,wheat rotation/fodder
Rainfed crops general
Rainfed crops and woods
Urban and industrial settlements
Water bodies
21 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0
22 0 0 0 0 0 0 0 0 0 0 0 0 0 0 10 0 0 0 0 0
23 0 0 0 0 0 0 0 1 0 1 0 0 0 0 0 8 0 0 0 0
24 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 2 35 0 2 1
25 0 0 0 0 0 0 0 0 0 0 1 0 2 0 0 0 0 15 0 1
26 0 1 0 1 1 0 0 5 4 2 0 1 1 1 2 1 1 2 43 2
27 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 9
Tot 3 10 3 4 15 3 8 88 107 42 13 10 19 2 13 11 47 22 50 14
4 75
13 62
89 75
110 69
42 88
12 83
6 83
16 94
1 100
10 100
10 80
41 85
19 79
68 63
9 100
484
3 5 6 8 9 10 11 15 16 17 18 19 20 21 22 23 24 25 26 27
3 3 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0
5 0 9 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0
6 0 0 3 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
8 0 0 0 3 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
9 0 0 0 0 11 0 0 0 1 0 0 0 0 0 0 0 0 1 0 0
10 0 0 0 0 0 3 0 0
Tot U.A.
4 75
10 90
3 100
4 75
13 85
P.A = Producer’s accuracy; U.A. = User’s accuracy
Irrigated rice,fodder rotation Irrigated mixed rice,wheat rotation/sugarcane
Irrigated mixed cotton,wheat rotation/orchards
20 0 0 0 0 0 0 0 0 0 0 0 0 15 0 0 0 0 1 0 0
Savanna deciduous
19 0 0 0 0 0 0 0 1 0 0 0 5 0 0 0 0 0 0 0 0
Savanna evergreen closed
18 0 0 0 0 0 0 0 1 0 0 10 0 0 0 0 0 1 0 0 0
Savanna evergreen open
17 0 0 0 0 0 0 0 1 3 37 0 1 0 0 0 0 0 0 0 0
Pastures evergreen lowland
16 0 0 0 0 1 0 0 10 76 2 2 3 1 0 0 0 8 2 4 1
Pastures deciduous
15 0 0 0 0 0 0 0 67 21 0 0 0 0 0 0 0 0 1 0 0
Bare soil
0 0 0 0 0 0 0 1 0 0 0
11 0 0 0 0 1 0 8 2 2 0 0 0 0 0 0 0 0 0 0 0
Ground classes
Map Classes
Forests evergreen needleleaf
Table 3.4 Error matrix for the accuracy assessment of LULC classification.
P A. 100 90 100 75 73 100 100 76 71 88 77 50 79 50 77 73 74 68 86 64
40
A total of 484 points were collected during the ground truthing survey and used in the error matrix analysis. Twenty LULC classes were visited and has minimally three observation points. The error matrix in Table 4 shows an overall accuracy of the classified map is 77%. The average producer’s and user’s accuracy is 78% and 83%, respectively. Hence, 83% of all land use classes identified on the map agree with the field observation. Considering that the satellite resolution is coarse (100 hectares) and the field sizes (0.4 hectares) are small, this accuracy is rather satisfactory and in agreement with accuracy levels achieved in different land use and crop identification studies (e.g. Bastiaanssen, 1998). He concluded that with extensive field work, crops can be identified with an average accuracy of 86%. However, this accuracy level fluctuates from 49% to 96%, depending upon the spatial coverage of the satellites and field size. Thunnissen and Noordman (1997) suggested 70% minimum classification accuracy should be attained at regional scale. Giri and Jenkins (2005) and Wardlow and Egbert (2008) attained 77.3% and 84% accuracies from MODIS 500 m and MODIS 250 m imagery, respectively. An accuracy of 73% for heterogeneous and 89% for homogeneous pixels were attained by Knight et al. (2006) using MODIS 250 m NDVI data. Keeping in view the heterogeneity of different classes in the present study, the classification accuracy attained is in fair agreement with the previous work. The producer’s accuracy of two classes namely “irrigated wheat, fodder rotation” and “irrigated mixed rice, wheat rotation/sugarcane” is 50% while their user’s accuracy is 83 and 100%, respectively. The apparent reason for this lower producer’s accuracy is the mixed agricultural cultivation processes that occur at the 100 ha scale. In addition to that, a smaller number of ground truth points are available for the class “irrigated mixed rice, wheat rotation/sugarcane” resulting in lower (50%) producer’s accuracy and 100% user’s accuracy. The three irrigated classes “irrigated mixed cotton, wheat rotation/orchards”, “irrigated mixed cotton, wheat rotation/sugarcane” and “irrigated rice, wheat rotation” constitute the major irrigated cropland systems of the Indus Basin. These are distinct with user’s accuracy of 75%, 69% and 88% while producer’s accuracy 76%, 71% and 88%, respectively. The user’s accuracy of three rainfed crop classes, “rainfed crops mixed cotton, wheat rotation/fodder”, “rainfed crops general” and “rainfed crops and woods” is 80%, 85% and 79%, respectively. The producer’s accuracy of these classes is 73%, 74% and 68%, respectively. However, the class “rainfed crops wheat/gram” shows 100% user’s accuracy and 77% producer’s accuracy while remaining 23% of this class is falling in “rainfed crop general” and “urban and industrial settlements”. The user’s and producer’s accuracy of irrigated and rainfed crops is more than 70%, in general. These are key classes for water management analysis. Therefore, it is important that LULC classes with a large acreage have an acceptable accuracy. The accuracy assessment using error matrix is normally dependent on sample points. Limited sample points can lead to assigning the correct class by chance(Foody, 2002). Such chance agreement effects can be assessed using Kappa (k) coefficient suggested by Cohen(1960).The Kappa coefficient incorporates the off-diagonal elements of the error matrices. It represents agreement obtained after removing the proportion of agreement that could be expected to occur by chance (Congalton, 1991). Kappa coefficient is well suited
for accuracy assessment of LULC maps (Vliet, 2009). Jonathan et al. (2006) attained 0.72 Kappa coefficient in a study conducted in Brazil using MODIS 250 m data. Wang and Tenhunen (2004) reported lower values of Kappa coefficient (0.14 and 0.20) for IGBP-DIS released land cover map and map developed by unsupervised classification using AVHRR NDVI temporal profiles, respectively. Melesse and Jordan (2003) achieved 79% (Kappa = 0.79) agreement between on ground and Landsat derived land use map of Florida using unsupervised classification. Thapa and Murayama (2009) tested different classification approaches using ALOS (Advanced Land Observing Satellite) in Japan and reported Kappa coefficient ranging from 0.71 to 0.87. Kappa coefficient attained in this study was 0.73, which is in moderate agreement range as suggested by Congalton and Green (1999). 3.4.3.2 Existing databases The second check was performed against existing LULC maps. These maps encompass continental or global scale and have single legend classes. For the sake of comparison, some classes were merged to make analysis easier. A summary of major classes obtained from different studies is shown in Table 3.5. Table 3.5 Comparison of major classes derived from different sources. Class Irrigated Rainfed Forest Bare soil Pastures Snow and ice Perm./temp Total
Present study,2007 (mha) 29.10 18.61 8.73 7.37 37.40 9.77 110.98
IGBP,2004 (mha) 34.36 0.23 3.11 28.65 44.48 2.03 112.86
GLC,2000 (mha) 43.32 0.88 9.34 6.83 37.85 12.58 110.80
The statistics show some marked differences in multi cropping, rainfed and desert land use classifications. The barren class defined in IGBP map is re-expressed into bare soil and rainfed classes. In some desert areas (e.g. Thal desert), there is an increasing trend of NDVI during rabi indicating increase in vegetation, which must be related to “rainfed crops wheat/grams” in the area. The Eastern side of the Indus Basin is found similar to our map. Urban areas were difficult to demark using NDVI data. MODIS data has the ability to better discriminate urban areas (Giri and Jenkins, 2005). Therefore the urban area class was taken from the IGBP urban area mask. The irrigated classes in GLC2000 map is classified as “intensive irrigated agriculture” and “irrigated agriculture”. No information can be found on dominant cropping patterns in the area. The class “pastures/savanna” in the present study is in accordance with GLC2000. However, GLC2000 has reported more area under the classes falling in general “forest class” of the two maps. The LULC map developed by IWMI (Thenkabail et al., 2005) focused more on the Ganges Basin, as well as the lower and eastern Indus Basin. The land use classification developed in the present study is in accordance to that map. It facilitated to improve the quality of the 42
produced map for the eastern side of the basin, as in present study, no ground truthing was carried out in this part. The irrigated area in the canal commands of the lower Indus Basin is classified as a general class named “irrigated water logged crops (Indus) rice, shrubs” which cannot properly discriminate different crops grown in the area. The IWMI map appears to be more generalized in the Indus Basin. 3.4.3.3 Localized studies Some small scale local studies were also analyzed to check the accuracy of classes falling in southern and eastern parts of the Indus Basin. Bastiaanssen et al. (2003) prepared a LULC classification in the Sirsa region of Haryana state, India representing eastern parts of Indus Basin. They used two Landsat-7 images, one for kharif and other for rabi season, for the year 2002. Ground truthing was also performed in India for accuracy assessment. These images were originally in 30m resolution but were re-scaled to 1 km to match with the current map resolution. Wheat being major rabi crop, is grown at 47% of the area while 42% area is classified as bare soil. In reality, when using 1 km resolution images, wheat and soil could be classified as mixed. By putting wheat and soil together, the difference between their maps and the LULC map presented in this paper for the Sirsa area was 2.3% only. Similarly, kharif crops are grown at 54% of total study area while about 33% area is classified as desert. Clouds and shadow are also present in the kharif image. The present study classified desert into rainfed crop classes. Considering kharif crops and desert mixed for 1 km resolution, both maps has 12.6% difference. 3.4.3.4 Ancillary databases Ancillary data can be useful to further improve pixel based recognition of land surface objects. Therefore, agricultural census of this area can provide a good estimate of the accuracy on mapping agricultural land use classes. The crop area statistics are used to compute cropped area fractions of crops in different administrative units in the Indus Basin. The cropped area fraction of wheat, rice, cotton and grams representing irrigated and rainfed crops is given in Figure 3-7(a),(b),(c) and (d), respectively. The results in Figure 3-7(a) show that the wheat area (comprising major portion of rabi cropping) reported by MINFAL(2008) is in close agreement with the area estimated in present study. Fairly high coefficient of determination (R2 = 0.87) between the area fractions reported and estimated is observed. However, this is also a fact that other rabi crops especially mustard are grown in the area.
43
b
a
c
d
Figure 3-7 District-wise (a) wheat area fraction (b) rice area fraction (c) cotton area fraction and (d) gram area fraction for rabi 2006-07, kharif 2007,kharif 2007 and rabi 2006-07,respectively. The results shown in Figure 3-7(b) &(c) are for rice and cotton areas, respectively. These are the major kharif crops grown in the region. The R2 values between the reported and estimated area fractions are 0.82 and 0.63 for rice and cotton crops, respectively. In different areas large tracts of millet, sorghum, maize, groundnut and tobacco have also been grown during kharif. The cropped area of these scattered crops is hard to estimate because of similarity in the growing season. This creates mixed pixels and thus directly affects the results. This is a drawback of using coarse resolution data. Cropped area fraction of one rainfed crop i.e. grams is shown in Figure 3-7(d). The results of this class are in good agreement (R2 = 0.91) with the distribution of grams from the 1 km NDVI data. Despite these complex cropping patterns, the present study provided an overall general picture that agrees with the cropping pattern of major crops.
44
The LULC results were also verified against non-agricultural land use information (MINFAL, 2007). FAO and ICIMOD were also consulted for forests, snow and ice classes. Total cultivated area calculated in the present study is 13% more than that reported by MINFAL (2007). This may be due to the fact that rainfed crops are not reported separately in MINFAL report. The irrigated area reported by FAO (2009) and Habib (2004) is 18.84 mha and 14.80 mha for the year 2005 and 2004, respectively. Present study estimates irrigated area as15.73 mha. These deviations are certainly related to the accuracy of the LULC mapping procedure proposed, they should also be ascribed to the field work and interpretation procedures for such vast regions. The typical variations of 5 to 10% are, however, acceptable. The forested area classified in the present study is 12.7% more than MINFAL statistics. The area reported by present study under food and cash crops (wheat, cotton, rice, sugarcane, orchards) is similar to that reported by MINFAL (2007). The area of class “snow and ice permanent” is compared with ICIMOD (Jianchu et al., 2007) for the whole Indus Basin. The results are 20.9% lower than ICIMOD. In fact, the snow and ice class in present LULC classification is divided into “snow and ice permanent” and “snow and ice temporary”. This difference in reported area can be due to the fact that ICIMOD has taken temporary snow as permanent. Detailed comparison of areal extent of different classification statistics are shown in Table 3.6. Table 3.6 Classification statistics of LULC in Indus Basin (Pakistan) compared with different sources. S.No
1 2
Total cultivated area Irrigated area
3 4 5
Rainfed crops Forests Food and cash crops (Wheat, rice, sugarcane, cotton, orchard) Rainfed wheat Rainfed grams Fodder Snow and ice permanent*
6 7 8 9 *
Classification Statistics
Present Study Area (mha) 26.89 15.73
MINFAL
Others Source
Area (mha) 23.39 17.37
11.16 4.80
4.19
18.23
18.22
1.45 1.33 2.74 4.18
1.24 1.05 2.50 -
Area (mha) 27.07 18.84 14.80 7.63
FAO,2009 FAO,2009 Habib,2004 Habib,2004+FAO,2009
5.28
ICIMOD
For whole Indus Basin
45
3.5
Conclusions
Spatial and updated land use and land cover information is essential to support basin scale water management. Due to their lack of details, existing global land cover data sets are inadequate to fulfill this role. This research has demonstrated that it is possible to discern LULC classes with managed land use and water use. Temporal profiles of NDVI from SPOT-Vegetation were used to capture the seasonal phenological variations of 27 classes. The ground information and expert knowledge of cropping patterns can efficiently be used to distinguish different crop classes. The overall accuracy attained in this research is 77% with user’s and producer’s accuracy being 83% and 78%, respectively. The Kappa coefficient of 0.73 is in moderate agreement range. This is reasonably good and similar to the accuracy reported by different global scale studies. The vegetation phenology appears to be strongly coupled to elevation, slope and temperature regimes. The overall crop statistics of food and cash crops show good agreement with the statistics reported by governmental organizations. Coefficient of determination for the cropped area fractions reported and estimated for wheat, rice, cotton and gram crops is 0.87, 0.82, 0.63 and 0.91, respectively. The mixed pixels especially for rice and cotton crops is a cause of lower R2 for these two crops as other crops are also grown within the same season. The accuracy is less satisfactory for agricultural plots with a small size and major focus on food and cash crops. The comparative advantage is that the classification is more suitable for hydrological, water resources, agricultural, forestry and environmental studies. The knowledge of dominant crop rotation schemes can play an essential role in the planning of food security and rural development. The crop water requirements vary for specific crops. Therefore, crop based classification helps water policy analysts and managers to formulate better plans with an improved knowledge base on the extent of the natural resources, certainly when the basin has an international character. The technique of classifying different water consumers in the Indus Basin offers a robust method to categorize beneficial and non-beneficial water use, being one of the key ingredients of water accounting procedures.
46
47
4 Local calibration of remotely sensed rainfall from the TRMM satellite for different periods and spatial scales in the Indus Basin Chapter based on: Cheema, M.J.M. and Bastiaanssen, W.G.M., 2012. Local calibration of remotely sensed rainfall from the TRMM satellite for different periods and spatial scales in the Indus Basin. International Journal of Remote Sensing 33, 2603-2627.
4.1
Introduction
Rainfall is an important component of the water cycle and for food production. Its accurate estimate is therefore vital for hydrological, water accounting and water withdrawal studies in the catchments and international shared river basins (Verdin and Klaver, 2002; Tobin and Bennett, 2010). The majority of the world’s cropland depends on rainfall as the major source of water. Historically, rainfall is measured by rain gauges, which provided reasonably accurate measurements at one point or field plot only. Estimating rainfall at basin level at appropriate spatial and temporal scales is still a challenge (Sawunyama and Hughes, 2008). Point measurements are not a true representation of areal average rainfall values (Draper et al., 2009). Rain gauges are subject to various systematic and random errors. Systematic errors are commonly a result of wind, wetting losses, evaporation from containers, splashing of water, and blowing or drifting of snow (Nespor and Sevruk, 1999). The magnitude of these errors can range from 2-50%, 2-10%, 0-4%, 1-2% and 10-50%, respectively (Rubel and Hantel, 1999). Observational and instrumental errors are major random errors in rain gauge measurements. These (systematic and random errors) may result in up to 30% difference between measured and actual rainfall (WMO, 2006). Moreover, a sparse rain gauge network cannot reflect rainfall variability caused by topography and orography, and will result in erroneous estimates of areal rainfall (Andréassian et al., 2001). There are global rainfall databases with data merged from rain gauges, radar observations, numerical weather prediction models, and satellite estimates. Some examples are: the Global Precipitation Climatology Project (GPCP), Global Precipitation Climatology Centre (GPCC), Climatic Research Unit (CRU) precipitation database, Climate Prediction Center’s merged Analysis of Precipitation (CMAP), Asian Precipitation Highly Resolved Observational Data Integration towards the Evaluation of Water Resources (APHRODITE) and the ERA re-analysis data set. But their spatial scales (e.g. 100 km×100 km, 250 km×250 km, and 250 km×375 km) are too coarse to carry out hydrological and water management studies. The temporal coverage is also insufficient. Such global datasets of rainfall are often based on point measurements that are spatially interpolated with geo-statistical procedures. The geo-statistical interpolations can only be useful if a very dense rain gauge network is available. In absence of such a network, the resulted rainfall maps are very general and do not reflect orographical, land surface and atmospheric processes. The World Metrological Organization (WMO) is connected for instance with about 40 observatories in the Indus Basin. Such a rain gauge network in the 48
Indus Basin is insufficient for supporting studies and applications. It is also insufficient for flood warnings, as has been witnessed during July and August 2010. The need for more accurate spatially distributed rainfall estimates can be met by satellite based sensors (Huffman et al., 2001). Advancements in remote sensing make it practically possible to adopt rainfall estimates from satellites as an alternative source of information (Din et al., 2008) as long as proper ground truthing occurs. Spaceborne sensors provide continuous monitoring of rainfall both spatially and temporally. Such data are generally readily available over longer periods and cover large areas (Immerzeel et al., 2009). The Tropical Rainfall Measuring Mission (TRMM) provides regional coverage at higher temporal resolution as compared to other gridded products, but at the cost of a low spatial resolution. The indirect measurement of precipitation by onboard sensors also has uncertainties (Hong et al., 2006; Hossain et al., 2006). These uncertainties are associated with lack of rainfall detection as well, false detection and bias (Tobin and Bennett, 2010). Both temporal errors (± 8 to ±12% per month) and sampling errors (~ 30%) can be expected in TRMM rainfall estimates (Franchito et al., 2009). Such errors can result in erroneous applications if applied without calibration (AghaKouchak et al., 2009; Gebremichael et al., 2010). Therefore, TRMM satellite estimates need area specific calibration to reduce such errors. The correction of satellite products are normally carried out by comparing satellite data with in-situ rainfall measurements. Many efforts have been made in this regard (e.g. Ji and Stocker, 2003; Chokngamwong et al., 2005; Dinku et al., 2007). These studies were done at global and regional scales (Thailand and Africa respectively). TRMM product has shown varying accuracies in different regions and for different adopted methods. Ji and Stocker (2003) and Chongamwong et al. (2005) observed correlation of 0.56 and 0.86 between satellite and rain gauge measurements, respectively. Dinku et al. (2007) observed Nash-Sutcliffe efficiency of 0.81 and 25% root mean square error between satellite and rain gauge data averaged over 2.5° grid boxes. Villarini and Krajewski (2007) tested accuracy for a single 25 km × 25 km pixel containing 23 rain gauges in Oklahoma and found 0.55 correlation between the satellite and rain gauges values. The mismatch between spatial rainfall estimates from satellites (Grimes et al., 1999) and point measurements by rain gauges unavoidably led to distrust in both data sets (Xie and Arkin, 1996; Ciach et al., 2000; Omotosho and Oluwafemi, 2009). All these studies reported area specific bias in the satellite estimates. Therefore, the uncertainty in TRMMbased rainfall in the transboundary Indus Basin needs to be inspected carefully prior to its use in hydrological and agricultural water management studies. Rainfall controls the renewable water resources, affects safe withdrawals for irrigation and provides moisture directly to thirsty crops during the rainfall seasons. This paper is novel because it integrates scarce measurements from rain gauges with time series of TRMM data to generate advanced information systems that is otherwise not available in vast irrigated basins such as the Indus. The aim of this paper is to develop a calibration protocol for TRMM rainfall data at different spatial and temporal scales. Two research questions were addressed to investigate the adoptability of satellite derived rainfall:
49
12-
4.2 4.2.1
What is the deviation between satellite rainfall estimates and rain gauge measurements at different temporal and spatial scales? How could geographical influence on satellite-based rainfall be described and correct the TRMM data?
Materials and methods Study area
The study area is Indus Basin, with large variation in climate and topography. The Indus Basin lies between 24° 38′ to 37° 03′ N latitude and 66° 18′ to 82° 28′ E longitude. It covers Pakistan, India, China (Tibet) and Afghanistan (see Figure 4-1). The basin has a surface area of 116.2 million ha (mha), with elevations ranging from 0–8600 m above mean sea level (a.m.s.l). The basin is bordered by the mighty Himalayas in the north-east; the Karakoram and the Hindukush mountain ranges in the north; the Sulaiman and the Kirther ranges in the west and the Arabian Sea in the south.
Figure 4-1 Location of the Indus Basin (marked in green) showing elevation and location of rain gauge stations used in this study.
50
4.2.2
Rainfall systems over the Indus basin
The major part of the basin is dry and located in arid to semi arid climatic zones. The annual rainfall is higher in the north and the north eastern tracts (>1200 mm yr-1) and lower toward the middle and the southeast ( 0.30 cm3cm-3 soil moisture values on DOY 168 – 171, 176 – 184 and 170 – 171, respectively. These DOYs correspond to the rainy season (Monsoon) in the basin. The maximum value (~0.38 cm3cm-3) is observed in “irrigated rice-fodder rotation” land use which comprises of severely water logged areas in the lower Indus Basin especially in North West, Fuleli and Begari canal commands (Aslam and Prathapar, 2001). Some higher values are also observed in “rainfed crop general” and “bare soil” land uses which occurred on DOY 181 – 183. These higher values are normally due to the rainfall occurrence before the satellite overpass in those areas. The flood year 2010 has been investigated in addition because large portions of land were flooded. The maximum values that a pixel attained during this period are given in figure 10. About 6.5% of all pixels have shown values ≥ 0.30 cm3cm-3 while 1.8% of all have attained soil moisture values ≥ 0.37 cm3cm-3. These values were observed during months 84
of July and August, the duration coinciding with the period of severe flooding in the region as well as basin irrigation for rice fields. Pixels fulfilling ≥ 0.37 cm3cm-3 reach close to their physical upper limit and these areas coincide with areas being flooded. These pixels have reached the θsat values for particular soil type (see Figure 5-7 and Figure 5-9). This brief analysis confirms that the higher end range of AMSR-E values is appropriate. Although not systematically investigated, the lower range of soil moisture (see Figure 5-10) reveal values in the range of 0.01 to 0.10 cm3cm-3. For sandy soils and bare land, the values are often θ 0.60 was found for 75% of the cases with zero to 40 days lag. For the wet kharif season, it was found for 81% cases but with a lag of 20 to 60 days. A strong spatial correlation exists between AMSR-E mean soil moisture and NDVI (rs=0.85) at annual and seasonal periods. The maximum values that a pixel could attain in a single day during 2007 ranged from 0.08 to 0.38 cm3cm-3. The maximum values of 0.37 cm3cm-3 and more is similar to the theoretical expected top layer saturated moisture content and was during 2010 indeed reported as being flooded. The high end values occur also on irrigated rice fields with flood irrigation practices. This suggests that absolute values AMSR-E estimates are plausible for wet land surfaces. The complete analysis was conducted for one year cycle only, and the behavior of the relationships between soil moisture and rainfall, NDVI and saturated moisture content is acceptable for creating trust in the AMSR-E values. It is recommended to expand future analysis with longer time series as shorter time series may reflect low significance. The overall conclusion is that the AMSR-E soil moisture product from the public domain NSIDC data centre is realistic, not only in relative terms, but also in absolute terms. Since 87
the AMSR-E soil moisture product has an acceptable accuracy, it can be used to describe hydrological conditions and water management practices in large scale river basins.
5.5
Appendix: Soil moisture retrieval algorithm
Soil moisture data was obtained from the AMSR-E sensor due to its frequent revisit time and the operational access of data. AMSR-E is one of the six sensors onboard Aqua satellite, which was launched in 2002. Aqua crosses the equator at a local solar time of 01:30 and 13:30 for ascending and descending passes, respectively. This information is for a surface layer of maximum 5 cm (Ray and Jacobs, 2007). AMSR-E measures microwave radiations emitted by the earth’s surface expressed in terms of brightness temperature. Soil moisture is then retrieved using microwave radiative transfer models. Several algorithms are available for retrieval (e.g. Njoku et al., 2003; Owe et al., 2008). One of them has been developed by NASA’s NSIDC following the method outlined in Njoku et al. (2003). Subsequent refinements were then proposed by Njoku and Chan (2006) to account for vegetation surface roughness. This method uses 10.7 GHz frequency to retrieve soil moisture. Alternatively, interpreted AMSR-E data is available from the radiative transfer models developed by the Free University of Amsterdam in conjunction with NASA (Owe et al., 2001; de Jeu et al., 2008). They provide soil moisture data retrieved at lower frequency of 6.9 GHz. After inspection of the measured surface soil moisture data in the dry areas of the Indus Basin, the data from NSIDC seemed more plausible and in agreement with expectations for dry soils in the semi-arid and arid climate of Pakistan and India. The analysis in this paper is therefore based on the NSIDC interpretation models using 10.7 GHz frequency. The retrievals using this frequency are less susceptible to RFI problems (Njoku et al., 2003). Natural surfaces emit radiations in the microwave region, and the total signal is a function of both land surface and atmospheric attenuation. The surface brightness temperature observed for these natural surfaces is given by Njoku et al.(2003) as:
[
]
TBp = Tu + exp(− τ a ) {Td rsp exp(− 2τ c )}+ TS {(1 − rsp )exp(− τ c ) + (1 − ω p ){1 − exp(τ c )}{1 + rsp exp(− τ c )}} 5.3 where, “TBp” is the brightness temperature (K), “Tu” and “Td” are the upwelling and down welling emission, respectively. “TS”is soil surface temperature (K). “τa”and “τc”are atmospheric and vegetation opacity, respectively., “ωp”is the single scattering albedo that depends upon vegetation structure and water content. The value of ωp is very small (Njoku and Entekhabi, 1996) and its effect is minimal (Jackson et al., 1982). The values vary from 0.04 to about 0.13 for different crops (Owe et al., 2008). Since experimental data for this parameter is limited, it can be considered to be negligible (ωp≈ 0)(Njoku and Chan, 2006). The parameter “rsp” is the surface reflectivity (subscript “p” denotes vertical (V) or horizontal (H) polarization).
88
The opacity τa along atmospheric path is dependent on the viewing angle “β”, precipitable water “qv” and vertical column cloud liquid water path “ql”. It can be expressed as given by Njoku and Li, (1999):
τ a = (τ o + av qv + al ql ) Cosβ
5.4
where, “τo”is oxygen opacity at nadir, “av” and “al” are the water vapor and cloud liquid water coefficients, respectively. The vegetation opacity depends upon vegetation water content and has approximately linear relationship (Bolten et al., 2003).
τ c = b × VWC Cosβ
5.5 -2
where, “VWC” is vegetation water content (kg m ) and “b” is an experimentally derived vegetation parameter (Jackson and Schmugge, 1991). The values may vary for different crops, normally taken 0.12 in absence of experimental data. “β” is the incidence angle that can be affected by topographic features and land cover on the ground (Bartalis et al., 2006; Friesen, 2008). The reflectivity from rough soil surface (rsp) can be empirically related to the equivalent smooth surface (rop) (Wang and Choudhury, 1981), through the expression:
{ = {(1 − Q )r
} }exp(− h )
rsV = (1 − Q )roV + QroH exp(− h ) rsH
oH
+ QroV
5.6 5.7
where, parameter “h” is related to the surface height standard deviation “σ”, theoretically given as h = (4πσ λ × Cosβ )2 while “Q” is polarization mixing parameter related to h. The term in bracket describe mixing of the co and cross polarized scattered radiation. The “Q” and “h” parameters are determined experimentally and vary spatially. If surface roughness conditions are not known then “Q” is assigned 0 and “h” is assumed between 0 and 0.3 (Jackson, 1993). The soil surface reflectivity depend upon the soil dielectric constant, which in turn depends on soil moisture content (Dobson et al., 1985). The dielectric constant is an electrical property of matter that measures medium response to an applied electric field (de Jeu, 2003). Large differences between dielectric constants of dry soil (~3.5) and water (~80) permits retrieval of soil moisture from microwave emissivity (Schmugge et al., 1992). The effective dielectric constant is computed by inverting Fresnel equations (Ulaby et al., 1986). These equations estimate surface reflectivity as a function of the dielectric constant “ε” of the medium and the incidence angle “β” is based on vertical or horizontal polarization of the sensor:
89
ε cos β − ε − sin 2 β roV = ε cos β + ε − sin 2 β
2
cos β − ε − sin 2 β = r o cos β + ε − sin 2 β H
2
5.8
Finally, the dielectric mixing model (Dobson et al., 1985) based on soil texture, soil bulk density “ρb” and specific density “ρs” gives the volumetric soil moisture “θo” at each pixel of AMSR-E.
ε α = 1+
(
)
b α ε s − 1 + θ o℘ε αfw − θ o s
5.9
where, “θo” is volumetric water content (cm cm ). “εs”is dielectric constant of soil solids depending upon ρs , “α” is shape factor and “γ” is empirical constant depends on soil textural composition. “εfw”is dielectric constant of free water calculated by Debye equation and more description on this mixing model can be found in Dobson et al.(1985). 3
90
-3
6 The surface energy balance and actual evapotranspiration of the Transboundary Indus Basin estimated from satellite measurements and the ETLook model Chapter based on: Bastiaanssen, W.G.M., Cheema, M.J.M., Immerzeel, W.W., Miltenburg, I. and Pelgrum, H., 2012. The surface energy balance and actual evapotranspiration of the transboundary Indus Basin estimated from satellite measurements and the ETLook model. Water Resources Research, (under review).
6.1
Introduction
Planning and monitoring of consumptive water use is necessary for sound management of scarce water resources. Consumptive use influences social, economic, agricultural, and environmental development. Water is consumed mainly through evaporation (E) and transpiration (T) (jointly termed evapotranspiration (ET)) from crops, soil, forests, urban areas, and natural vegetation, amongst others. If precipitation over a specific land cover exceeds ET (e.g. forests), such a land cover class is a net producer of water resources. Non-consumed water from precipitation feeds streams, rivers and aquifers. If, however, ET exceeds precipitation, such a land cover class will be a net consumer of water resources. Irrigated lands are a typical example of a net consumer of water. ET information can be used for irrigation management (Bastiaanssen et al., 1996; Allen et al., 2007), drought detection (e.g. Calcagno et al., 2007), real water savings (e.g. Seckler, 1996), water accounting (e.g. Molden and Sakthivadivel, 1999), water productivity (e.g. Zwart et al., 2010), virtual water trade (e.g. de Fraiture and Wichelns, 2010), model calibration (e.g. Immerzeel and Droogers, 2008), hydrological model applications (Droogers et al., 2010a) and groundwater management (Ahmad et al., 2005). A number of techniques are in use to measure ET, ranging from conventional point measurements to modeling and spatially distributed remote sensing estimates. At individual plant and field scales, lysimeters, heat pulse velocity, Bowen ratio, scintillometry, surface renewal, and eddy correlation are commonly used (e.g. Meijninger et al., 2002; Nagler et al., 2005). Field scale ET measurements are generally considered accurate, however the accuracy of these traditional methods are often less than 90% (Twine et al., 2000; Teixeira and Bastiaanssen, 2011). The equipment cost, extensive labor, and coverage issues restrict use of these techniques at large scale (Elhaddad and Garcia, 2008). At the regional scale, earth observations by means of satellite data are gradually becoming more accepted (e.g. Courault et al., 2005; Anderson et al., 2007; Mu et al., 2007; Kalma et al., 2008; Guerschman et al., 2009; Wu et al., 2012) although operational data provision remains rare. This paper aims at contributing to the development of operational systems that could be applied on a daily time step for areas with limited ground data. Routine weather data is assumed to be available. Evapotranspiration computations are often based on surface energy balances (e.g. Price, 1990; Mu et al., 2007; Senay et al., 2007; Tang et al., 2009; Long and Singh, 2012). Many of these energy balance models require thermal infrared radiation from cloud free images and atmospheric corrections in order to produce accurate land surface temperature maps
91
(Jia et al., 2009). Cloud free surface temperature images for large areas in basins with monsoon climates are not straightforward to obtain (e.g. Bastiaanssen and Bandara, 2001). Thermal infrared radiation is more sensitive to atmospheric water vapor absorption than visible and near-infrared radiation (Lillesand and Kiefer, 2000), and it is thus more challenging to acquire land surface temperature maps not being thwarted by clouds. For instance, the surface temperature product (MOD 11A2) available through Moderate Resolution Imaging Spectro radiometer (MODIS) is thwarted by cloud cover for the entire period of monsoon 2007 (June – September). About 50% of the basin area was found without or with limited surface temperature data from day of year (DOY) 161 to 241. This illustrates the difficulty in getting continuous information for ET computations in irrigated areas from thermal infrared data. While it is generally accepted that thermal infrared data provide reliable results based on sound physics (e.g. Bastiaanssen et al., 2008; Allen et al., 2010; Allen et al., 2011), the cloud cover is a serious hindrance to routine applications in various parts of the world. To circumvent these problems, the current study deployed the ETLook algorithm that was first introduced by Pelgrum et al. (2010). Soil moisture derived from passive microwave sensors is the driving force for calculation of the surface energy balance in ETLook. Surface soil moisture relates typically to a depth of 2 to 3 cm, and the number of surface soil moisture databases is growing due to an increasing number of operational passive microwave sensors. This is a good moment to explore and develop ET models that are based on these data sets. Future soil moisture data layers will be based on active Synthetic Active Radar (SAR) measurements, once this data become available easily and free of charge. Microwave radiometry is less affected by cloud cover (Ulaby et al., 1981; Fily et al., 1995) and can thus provide continuous surface soil moisture information even in monsoon periods. Li et al.(2006) have shown the value of using microwave derived near-surface soil moisture in a two-source energy balance model over an agricultural area in central Iowa (USA). The ETLook algorithm is a two-source model and surface soil moisture is used for the computation of E, and a parameterization is introduced to compute sub-soil moisture content for the determination of T. Accurate ET information is of paramount importance for the 116.2 million hectares (mha) Indus Basin, with high elevation water source areas, a distinct monsoon climate with cloud covered regions, and declining water tables due to over-exploitation. This study was a first attempt to use microwave technologies to accurately estimate ET over the Indus Basin, and to detect areas with excessively high ET rates using a spatial resolution of 1 km. Such a resolution is thought to be good enough for regional scale applications. The main objective of this study was to demonstrate the validity of a combined optical and microwave based energy balance model (ETLook) in a vast river basin with large irrigation systems. Another objective was to use public domain data to estimate ET in the areas where field data are not available, and to show water managers that spatially discrete ET information is the basis for describing the major water flow path in ungauged basins.
92
6.2
Study area
The study area is the Indus Basin, which lies between latitude 24°38′ to 37°03′ N and longitude 66°18′ to 82°28′ E. The total area of the basin is 116.2 mha and encompasses four countries (Pakistan: 53%, India: 33%, China: 8% and Afghanistan: 6%) (Figure 6-1). The basin exhibits complex hydrological processes due to variability in topography, rainfall, and land use. The elevations range from 0–8000 m above mean sea level (a.m.s.l) and mean annual rainfall varies between approximately 200 to 1500 mm. The basin-wide average rainfall for 2007 was 383 mm yr-1 (Cheema and Bastiaanssen, 2012). The basin has two distinct agricultural seasons, being the wet kharif monsoon season (May to October) and the dry rabi season (November to April). Wheat is the major rabi crop while rice and cotton are major kharif crops. The basin provides food for more than half a billion inhabitants. To meet the water demand for such food production, and because rainfall is inadequate for meeting the full crop water requirements, the world’s largest contiguous irrigation system was built in the Indus Basin. The irrigation system supplies surface water to the middle and lower parts of the Indus. Irrigated agriculture is practiced in 26.02 mha (22.6%) area of the basin (see Figure 6-1). The era of tube well installations with subsidized rates and direct access to water has motivated farmers to augment shortages in surface water with groundwater resources (Shah et al., 2000). Currently 40–50 % of agricultural water needs are met through groundwater used in conjunction with surface water (Sarwar and Eggers, 2006). Groundwater quality is decreasing and phreatic surfaces diminishing across the Indus Basin.
Figure 6-1 Location of the Indus Basin and provinces of different countries in the basin. PK stands for Pakistan and IN for India. The irrigated areas in the basin are also shown.
93
6.3 6.3.1
Material and methods Satellite data and pre-processing
Key input data for ETLook are: surface soil moisture, spectral vegetation index, surface albedo, atmospheric optical depth, land use and land cover (LULC), soil physical properties, and routine weather data. Surface soil moisture was obtained from the Advanced Microwave Scanning Radiometer (AMSR-E) on the Aqua satellite. Daily soil moisture datasets with 25 km foot print (ascending and descending path) covering the Indus Basin were downloaded from the National Snow and Ice Data Center (NSIDC) website (http://nsidc.org/data/ae_land3.html) for the complete year of 2007 (Njoku, 2008). The year 2007 was selected because it was the last year prior to the start of this study, and all required auxiliary data were available. The actual spatial resolution of C-band AMSRE soil moisture is large (approximately 70 km×40 km). AMSR-E collects 60 km resolution C-band brightness temperature with a sampling interval of 10 km, which allows AMSR-E C-band data to be gridded at 25 km resolution. The operational character of surface soil moisture in NSIDC contributes to the construction of a routine provision of spatial ET data bases. A comprehensive soil moisture data validation study in the Indus Basin was performed by Cheema et al., (2011). It was found that both the behavior as well as the absolute values are realistic and provide sufficient information on the spatial and temporal changes of topsoil moisture in the Indus Basin. The daily layers were in the current study averaged to obtain 8-day soil moisture layers to be compatible with the MODIS optical satellite data. This Indus Basin ETLook study required topsoil moisture at 1 km scale. Various sophisticated methods are documented in the literature to downscale the available coarse resolution soil moisture data to 1 km pixels (Hemakumara et al., 2004b; Merlin et al., 2006; Friesen et al., 2008; Merlin et al., 2008; Gharari et al., 2011). All these downscaling methods require a number of parameters and have an empirical character related to the physiographical setting of a specific area. More research studies are required to find more generic solutions to this problem. Due to the absence of detailed soil moisture data in the Indus Basin, a simple method of downscaling based on effective saturation has been adopted in this study. Each AMSR-E pixel was downscaled to 1 km using a bilinear resampling technique first. This is simplistic, but is necessity to remove step changes in the data layers due to the texture of the large scale AMSR-E pixels. The information on saturated and residual moisture content (θsat and θres, respectively) for each soil type was used to calculate effective saturation (Setop xy) at 1 km grid using the definition proposed by van Genuchten (1980) as:
S etop , xy =
θ AMSRE − θ res , xy θ sat , xy − θ res , xy
6.1
where Se ,xy , θAMSRE , θsat,xy and θres,xy represent the effective saturation, AMSRE soil moisture, saturated and residual moisture content at 1 km pixel (x,y), respectively. The values for θsat,xy and θres,xy were inferred from the Food and Agriculture Organization (FAO) soil map (FAO, 1995) using pedo-transfer functions (Droogers, 2006). Soils with a top
94
large pore volume (θsat,) contain more air and have a lower degree of saturation. Their drier conditions reduce soil evaporation because soil moisture is retained stronger to the soil matrix, and the volume with water filled pores that are needed to transport water will be lower under dry conditions. In addition, the saturation of the subsoil (Sesub) is required for the computation of root water uptake and subsequent crop transpiration. The preprocessing of saturation of the subsoil was done by using an empirical relationship between Setop, the vegetation photosynthetically activity (that reflect soil water availability) and Sesub. The following relationship is imbedded in ETLook 1.0 and was applied in the current study:
[
]
S esub = 0.1LAI + (1 − 0.1LAI ) 1 − exp{S etop (− 0.5LAI − 1)}
6.2
where LAI is the Leaf Area Index. The basic assumption is that the degree of saturation of the subsoil is exceeds the saturation of the topsoil when vegetation is photosynthetically active, and that Setop affects the level of Sesub under all conditions. The green LAI reflects the access of vegetation to soil water. In absence of green plants, moisture in the sub-soil holds a direct analytical relationship with the moisture in the top-soil (e.g. Hillel, 1998). Hence, passive microwave data in combination with LAI describes the daily variation of root zone soil water content. The Normalized Difference Vegetation Index (NDVI) is an undisputed indicator of active vegetation and was used to compute LAI as explained further down. It has been demonstrated by, for instance Nagler et al. (2005) and Burke et al.(2001), that NDVI is an indicator of soil wetness and ET fluxes, which is in line with Eq. (6.2). NDVI data are distributed by the Land Processes Distributed Active Archive Center (LP DAAC), located at the U.S. Geological Survey (USGS) Earth Resources Observation and Science (EROS) Center (lpdaac.usgs.gov). Two 16-day NDVI datasets (MOD13A2 and MYD13A2 (collection 5) starting from day 1 and day 9 respectively) at 1 km were used to create 8day NDVI layers. The vegetation cover (VC) was derived from NDVI following Jiang et al., (2006) as:
NDVI fv − NDVI VC = 1 − NDVI − NDVI fv bs
0.7
6.3
Threshold values of NDVI = 0.8 and 0.125 were used as boundary condition for full vegetation cover (NDVIfv) and bare soil (NDVIbs), respectively. The LAI was computed from NDVI values using standard asymptotic relationships between LAI and VC (e.g. Curran and Steven, 1983; Carlson and Ripley, 1997):
1 − VC LAI = − ln kl
6.4
where kl is the light extinction coefficient with a value range of 0.40 to 0.65. An average value of 0.5 was taken for all representative vegetation types (e.g. Kale et al., 2005). The LAI (VC) relationship was similar for all land use classes because ki values for all classes
95
were not available and we assumed that ki differences between classes were small enough to justify the use of a few selective ki values, for all classes. Surface albedo was also derived from standard MODIS products. The 8-day albedo data product MCD43B3 (collection 5) at 1 km resolution was downloaded from (https://wist.echo.nasa.gov/~wist/api/imswelcome/) server provided by LP DAAC. Solar radiation is classically computed from the extra-terrestrial radiation in association with an atmospheric transmissivity in the solar spectrum. The atmospheric transmissivity of shortwave radiation can be inferred from optical depth information provided by the MODIS cloud product (King et al., 1997). One km resolution MYD06_L2 values of the optical depth product were downloaded from https://wist.echo.nasa.gov/~wist/api/imswelcome/ to estimate atmospheric transmissivity for the Indus Basin. The cloud optical depth from MODIS products was used to infer atmospheric transmissivity of shortwave radiation τMODIS (Barnard and Long, 2004). A detailed LULC map of the Indus Basin developed by Cheema and Bastiaanssen (2010) was used to infer information on different LULC classes in the basin. Twenty-seven LULC classes were identified. This LULC classification was used to create look-up tables for the definition of certain bio-physical parameters required for ET computations, such as minimum stomatal resistance, moisture sensitivity and maximum obstacle height. Rainfall (R) data are used to determine interception evaporation. Interception (I) is computed on a daily scale with the classical von Hoyningen model following von Hoyningen (1983) and Braden (1985).
1 I = 0.2 LAI 1 − (VC ) R 1 + 0.2 LAI
6.5
which assumes that maximum a water film of 0.2 mm is stored per unit LAI. This coefficient can be modified. ET cannot exceed R without being augmented by additional water resources. Rainfall is therefore a good measure to validate against. By absence of sufficient rain gauges, rainfall was obtained at spatial resolution of 25 km using Tropical Rainfall Measuring Mission (TRMM) processing algorithms described by Huffman et al. (2007). The global rainfall algorithm (3B43 V6) available through NASA website (http://neo.sci.gsfc.nasa.gov/Search.html?group=39) was used. It provides monthly accumulated rainfall data, which has been calibrated and validated according to the Geographical Differential Analysis (GDA) as outlined in Cheema and Bastiaanssen (2012). 6.3.2
Meteorological data
The major portion of the Indus Basin (53%) lies within the administrative boundaries of Pakistan. Most of the meteorological data (e.g. air temperature, relative humidity and wind speed) were therefore obtained from 65 meteorological stations under the aegis of the Pakistan Meteorological Department (PMD). Weather station data for India, China and 96
Afghanistan were extracted from the National Oceanic and Atmospheric Administration (NOAA) National Climatic Data Center (NCDC). The NCDC collects meteorological data from real time reporting stations worldwide in agreement with World Meteorological Organization regulations (ftp://ftp.ncdc.noaa.gov/pub/data/gsod/). Data from 16 stations with complete datasets were downloaded. Hence, air temperature, relative humidity and wind speed data from 81 stations collected at standard height of 2 m were obtained. ETLook requires gridded meteorological data for air temperature (Tair), relative humidity (RH) and wind speed (U2) at 1 km resolution. Topography, land use, sun angle, and distance from water bodies directly affects the spatial variability of near surface meteorological parameters (Brutsaert, 1982; Schulze et al., 1993). Ordinary geo-spatial interpolation techniques do not take these variables into account. The meteorological distribution model (Daymet) described by Thornton et al. (1997) was therefore used to convert point data to spatial meteorological data. Daymet uses a truncated Gaussian weighting filter for regional distribution of climatic variables in relation with topography. A 1 km Digital Elevation Model (DEM) obtained from GTOPO30 database (http://eros.usgs.gov/#/ Find_ Data/ Products_ and_ Data_ Available/gtopo30_info) was used to establish relationships between the climatic variables and topography. The weather grids for 2007 were independently validated against values from the International Water Management Institute (IWMI) world water and climate atlas that is based on long term field measurements and specific spatial interpolation procedures (http://www.iwmi.cgiar.org/ WAtlas/Default.aspx). The atlas provides monthly summaries of rainfall, temperature, humidity, wind speed, and sun shine hours at 18 km grid averaged over the period 1961-1990, as produced by the University of East Anglia (New et al., 1999). The 8-day Daymet estimates were aggregated to monthly values in order to make them comparable with the IWMI atlas. A high coefficient of determination (R2 > 0.85) was obtained for air temperature estimates. However, for relative humidity and wind speed, moderate coefficients of determination (R2 =0.70–0.80 and 0.60–0.70 respectively) were achieved. These correlations are considered reasonable because IWMI atlas values are monthly averages for 1961-1990, which may be different for different years. The IWMI atlas values are also interpolated and extrapolated, and associated with a certain uncertainty. Nevertheless, the impression is that air humidity and wind speed values display more uncertainty than air temperature. 6.3.3
Theoretical background of ETLook
The surface energy balance can be written as:
Rn − G = λ E + H
(Wm-2)
6.6
where Rn is net radiation, G is soil heat flux, λE is latent heat flux and H is the sensible heat flux. λE is associated with ET. The ETLook algorithm uses a two layer approach to solve the Penman – Monteith equation (Pelgrum et al., 2010). The Penman – Monteith equation for E and T can be written as:
97
∆ ∆ ( Rn , soil − G ) + ρ c p e ra , soil E= r ∆ + γ 1 + soil ra , soil
6.7
∆e ∆ ( Rn ,canopy ) + ρ c p r a ,canopy T= r ∆ + γ 1 + canopy r a , canopy
6.8
where E and T are evaporation and transpiration respectively in Wm-2; Δ (mbar K-1) is the slope of the saturation vapor pressure curve, which is a function of air temperature (Tair, °C) and saturation vapor pressure (es, mbar); Δe (mbar) is vapor pressure deficit, which is the difference between the saturation vapor content and the actual vapor content; ρ (kg m3 ) is the air density, and cp is specific heat of dry air =1004 J kg-1 K-1; γ (mbar K-1) is the psychometric constant; Rn,soil and Rn,canopy are the net radiations at soil and canopy respectively; rsoil and rcanopy are resistances of soil and canopy, while ra,soil and ra,canopy are aerodynamic resistances for soil and canopy respectively. All resistances are in s m-1. The E and T fluxes (W m-2) are converted to rates (mm d-1) using a temperature dependent function of the latent heat of vaporization. The LAI can be used to partition the net radiation into net radiation of the soil (Rn,soil) and the canopy (Rn,canopy) (Shuttleworth and Wallace, 1985). The increase in LAI results in an exponential decrease in the fraction of radiation available for the soil, and vice versa for the canopy. The energy dissipation due to interception losses is subtracted from the total net radiation. This energy is computed from the actual interception evaporation rates and the latent heat of vaporization being associated with that. The net radiation at the soil and canopy can be calculated using Beer’s law as follows:
Rn , soil = Rn ,canopy
{(1 − α ) R − L − I } exp ( −aLAI ) = {(1 − α ) R − L − I }{1 − exp ( −aLAI )} ↓
o
n
↓
o
n
6.9 6.10
where αo is surface albedo (–); R↓ (Wm-2) is the incoming shortwave radiation; Ln (Wm-2) is the net longwave radiation; I is the interception of water by leaves expressed in Wm-2; and a is the light extinction coefficient for net radiation. The incoming shortwave radiation can be calculated using daily measurements of shortwave transmissivity (τsw) and the theoretical extraterrestrial radiation (Rtoa). The parameterization for R↓ and Ln is taken from the FAO Irrigation and Drainage Paper 56 (Allen et al., 1998). The sum of Rn,soil and Rn,canopy constitute total net radiation Rn , after being corrected for interception losses.
98
The surface resistances in equations 6.7 and 6.8 describe the influence of the soil on evaporation or canopy transpiration. The soil resistance (rsoil) is a function of the topsoil effective saturation (Setop), estimated using equation 6.1. A power function defines this relationship (e.g. Clapp and Hornberger, 1978; Camillo and Gurney, 1986; Wallace et al., 1986; Dolman, 1993):
rsoil = b(S etop )
c
6.11
where b and c are soil resistance parameters, which can vary with soil type and are taken here as 30 and –3, respectively. Equation (6.11) describes the transfer of water in the liquid phase through the topsoil. Such transfer processes can also be computed with an unsaturated Darcian flow type of equation, but the accuracy is not necessarily better, due to high and uncertain gradients of moisture and suction in the upper 5 cm of the unsaturated soil. The soil hydraulic conductivity of a porous soil being exposed to anthropogenic influences under dry conditions is rather uncertain and reliance on detail soil physical data would jeopardize the potential of ETLook. The format of Eq. (6.11) is attractive for calibration purposes. Canopy resistance describes the resistance of vapor flow through the transpiring vegetation and is a function of the minimum stomatal resistance rs,min (s m-1), in association with a number of reduction factors and the leaf area that integrates the vaporization process from leaf to canopy scale. The value rs,min represents the resistance to transpiration from canopy under ideal conditions (no moisture stress, enough sunshine etc.). The resistance rs,min can have different values for the different land use classes. The rs,min is defined for a single layer of leaves, therefore effective leaf area index LAIeff, which describes the actual transpiring leaf mass, was used for integration from leaf to canopy. The following equation, as described by Mehrez et al. (1992) and Allen et al. (2006), was used to infer LAIeff:
LAI eff =
LAI 0.3LAI + 1.2
6.12
The canopy resistance under actual growing conditions can be computed using the common Jarvis-Stewart parameterization (Jarvis, 1976; Stewart, 1988). The Jarvis-Stewart parameterization describes the response of stomata to environmental factors in the form of minimal resistance multiplied by the product of interacting stresses on plants, and is computed as follows:
r rcanopy = s ,min LAI eff
1 St S v S r S m
6.13
where St is temperature stress, and a function of minimum, maximum and optimum temperatures, as defined by Jarvis (1976); Sv is vapor pressure stress induced due to persistent vapor pressure deficit; Sr is radiation stress induced by the lack of incoming shortwave radiation; and Sm is soil moisture stress originating from the root zone. The Jarvis-Stewart parameterization is common in many soil-vegetation-transfer models and has not been disputed in this work. It describes the joint response of soil moisture and LAI 99
on transpiration fluxes in a bio-physically justified manner. Sm is defined using a sinusoidal relationship with sub soil effective saturation (Sesub) and tenacity factor (Ksf) defined in ASCE (1996) as:
S m = K sf S
sub e
sin (2πS esub ) − 2π
6.14
where Ksf describes the ability of plants to extract soil moisture under different moisture conditions. It ranges from 1 for sensitive plants to 1.5 for moderately sensitive plants to 3 for insensitive (tenacious) plants. The aerodynamic resistance for soil (ra,soil) and canopy (ra,canopy) can be computed (Holtslag, 1984; Choudhury et al., 1986; Allen et al., 1998) as:
ra ,soil
z ln obs z0,soil =
ra ,canopy
zobs ln 0.1z 0 , soil 2 k uobs
z − d zobs − d ln ln obs z 0,m 0.1z 0,m = k 2 uobs
6.15
6.16
where k is von Karman constant = 0.41[-], uobs is the wind speed at observation height [ms1 ], d is displacement height [m], z0,soil is the soil surface roughness = 0.001 m. z0,m is the surface roughness. The land use map is used to prescribe values for z0,m. Research is in progress to derive surface roughness from radar imagery. The soil heat flux (G) for land surface is calculated using a sine function as described by Allen et al.(1998). The maximum value for G is recorded in May for northern latitudes, which coincides with a phase of–π/4. For southern latitudes the phase is –π/4 +π.
G
2π J π − 2 At , year k sin 4 P exp ( − aLAI ) zd
6.17
where At,year is the yearly amplitude for air temperature; J is the Julian day measured in seconds; k is the soil thermal conductivity (W m-1 K-1), which has a linear relationship with top soil moisture; a is the same light extinction coefficient as used in Beers law, see Eqs. 6.9 and 6.10; zd.(m) is the damping depth that is calculated as:
100
zd =
2kP 2πρ c p
6.18
where P is the period in seconds; and ρcp is the volumetric heat capacity (a function of the porosity and Setop). Eq. (6.17) includes light interception effects on soil heat flux. 6.3.4
Calibration and validation approaches
The cloud optical depth measures the attenuation of solar radiation passing through the atmosphere due to scattering and absorption by cloud droplets. The cloud optical depth can be defined as the negative algorithm of the fraction of the incoming radiation that is not scattered or absorbed in the atmosphere (Kitchin, 1987). Maximum and minimum threshold atmospheric transmissivity values were taken into consideration to account for latitude, zenith angle and diffuse radiation. The resulting atmospheric transmissivity (τMODIS ) was checked and calibrated using the simplified - but doable - field methods suggested by Angstrom (1924) and Hargreaves and Samani (1985). Records of sunshine hours were used for the Angstrom equation. Sunshine records were available from 24 stations in the study area. The same 24 stations were used to get diurnal air temperature differences for the Hargreaves equation. Results from the Angstrom and Hargreaves methods were used to determine a linear fit through the origin for each time interval of 8 days, to obtain calibrated short wave transmissivity (τsw):
τ sw = e.τ MODIS
6.19
where e is the regression coefficient. Minimum stomatal resistance values rs,min for each LULC were used to fine-tune ETLook. The rs,min values for agricultural classes were accepted to be between 40 to 140 s m-1 (ASCE, 1996; Radersma and de Ridder, 1996; Bastiaanssen and Bandara, 2001). Various researchers (e.g. Monteith, 1981; Sharma, 1985; Vanderkimpen, 1991; Allen et al., 1998) suggested a value of rs,min = 100 s m-1 for various agricultural crops like wheat, rice, beans, etc. Firstly, this default rs,min value was used for all agricultural classes (n=11). The following values were assigned to the remaining classes: pastures 125 s m-1, savannas 150 s m-1, forests 150 and 300 s m-1 (for broad leaf and needle leaf forests, respectively), sparse vegetation 200 s m-1, and urban and industrial settlements 60 s m-1. Water bodies were assigned 0 s m-1 because water vapor molecules can be transported into the atmosphere without physical barriers. During the second run of ETLook, all LULC classes with irrigated crops were assigned 80 s m-1 and the rainfed crops were assigned a value of 150 s m-1. During the third run, adjustments were made to the urban and industrial settlements land use, and a rs,min value of 500 s m-1 was assigned. During the third run, the soil resistance functions were also adjusted to improve congruity of the seasonal and annual total rainfall for terrain with sparse vegetation and desert surfaces. The annual bare soil evaporation must be lower than rainfall. After three runs, the ETLook results provided acceptable agreement between rainfall and ET. As expected, areas with ETR appeared to be irrigated areas and water bodies. Hence, calibrated
101
rainfall patterns were used to tune the minimum stomatal resistances and the coefficients in the formulation of the bare soil resistance. The ET output data cannot be used for water resources management without testing its accuracy. Results from previous studies based on soil moisture and lysimeter experiments were used for validation. Pakistan Agricultural Research Council (PARC) measured actual ET at Peshawar, Bhalwal, Faisalabad, Bhakkar, Mian Channu, and Tandojam representing upper, middle and lower parts of the basin (PARC, 1982). The ET results of PARC are for the years 1975-80 following an internationally funded study. Data collection discontinued when the project ended, yet it seems to be one of the most basic databases in Pakistan. More recent field measurement study was conducted by Ahmad (2002) at the Soil Salinity Research Institute, Pindi Bhattian (31°52′34.2˝N, 73°20′50.2˝ E) and Ayub Agricultural Research Institute, Faisalabad (31°23′26.2˝N, 73°02′49.8˝E). As part of a field investigation program during 2000 and 2001, he measured actual ET in rice/wheat and cotton/wheat systems by a temporarily installed Bowen ratio energy balance system. ETLook estimates were also checked against previously conducted remote sensing and modeling studies. The ET estimates provided by Bastiaanssen et al. (1999) for the Sirsa irrigation circle in India were checked. Other studies e.g. Shakoor et al. (2006), Sarwar and Bill (2007), Ahmad et al. (2009) and Shakir et al. (2010) determined ET in selected areas within the basin for different years. Previous studies were synthesized and used to compare with ETLook estimates. The coefficient of determination (R2), Root Mean Square Error (RMSE) and Relative Error (RE) were calculated to estimate the difference of the ETLook estimates with the previous studies. 6.3.5
Sensitivity and uncertainty analysis
A sensitivity analysis was performed to check the contribution of selected main input parameters to the output results. The sensitivity of ET was tested for a number of input parameters, i.e. θAMSRE, NDVI, rs,min, rsoil and θsat. Annual mean climatic conditions were assumed for the analysis. One factor at a time methodology was adopted to check the variance in the outputs due to input variability (e.g. Pitman, 1994). The analysis was conducted on two representative land uses i.e. “bare soil” and “irrigated rice - wheat rotation” at locations 71°22'54.123"E, 28°38'50.042"N and 75°23'53.59"E, 30°40'37.719"N, respectively. Randomly generated uniform distribution of AMSR-E based soil moisture values (n = 100) were used while keeping other parameters constant to check the variations in E, T and ET. The analysis was performed using representative NDVI values of 0.05 for bare soil and 0.67 for irrigated land use. For the parameters for which only a range was known, the defined parameter change was used to estimate the sensitivity between input parameters and output E, T and ET. A complete sensitivity analysis representing the change in the response variable caused by a unit change of an explanatory variable, while holding the rest of parameters constant, was performed. A sensitivity coefficient (SC = ∆out/∆in) was then calculated for each input parameter as described by Gu and Li. (2002). The sensitivity coefficient was normalized by the mean values representing the range of each pair of output and input variable. This normalized sensitivity coefficient is called sensitivity index (SI) and can be positive or negative. SI makes it feasible to compare the results of different input parameters. A 102
higher absolute value indicates higher sensitivity. A negative SI indicates an inverse relationship between input parameter and response variable. SI can be represented as:
SI =
M in
∆ out ∆ in M out
6.20
where, Min and Mout are the mean values of the input and output range, respectively. In addition, a stochastic uncertainty analysis was performed. A Monte Carlo simulation experiment using 1000 pairs of randomly generated input parameters was performed to investigate the model uncertainty. The values of the sensitive parameters were varied, while other climatic variables were kept constant.
6.4 6.4.1
Results and discussion Surface energy balance
The temporal variation of each component of the surface energy balance of the Indus Basin for the hydrological year 2007 is presented in Figure 6-2. The values represent the spatial averages for the whole Indus Basin. The average values attained by the surface energy fluxes with their standard deviations (SD) are provided in Table 6.1. A high variability from the mean is observed for the year, especially for net shortwave radiation (R↓), net radiation (Rn) and sensible heat flux (H). The large variation in climate during summer and winter is the probable cause of the high SD. Table 6.1 The minimum, maximum and average values of surface energy fluxes in the Indus Basin attained during the year 2007. The entire basin is covered and the values represent average flux densities for periods of eight days including daytime and nighttime. Fluxes Minimum Maximum Mean SD Net shortwave radiation(Wm-2) 95.70 237.50 170.10 45.10 Net longwave radiation(Wm-2) –75.60 –36.90 –57.80 10.80 -2 Net radiation(Wm ) 46.20 177.40 112.30 46.70 Soil heat flux(Wm-2) –7.10 8.10 0.34 5.20 -2 Sensible heat flux(Wm ) 37.20 131.60 79.50 29.80 Latent heat flux(Wm-2) 10.90 57.00 32.40 14.30 Evapotranspiration(mm d-1) 0.39 2.10 1.20 0.50 Evaporative fraction (–) 0.19 0.36 0.28 0.05 Rn is the dominant source of energy for land surface processes. The annual average value for Rn was 112.3 Wm-2 with a standard deviation of 46.7 Wm-2. The lower Rn values (150 Wm-2) was observed during DOY 113–225 while the average was 163.5 Wm-2. For the same period, Rn values fluctuated considerably with sudden depressions during DOY 145–225, corresponding to the monsoon season with 103
clouds. Afterwards, the net radiation decreased gradually and reached its minimum values again in winter.
Figure 6-2 Temporal variation of components of the surface energy equation during 2007 in the entire Indus Basin (116.2 mha). The dashed lines represent 24 days moving average values. The dry arid environment of the Indus Basin (annual rainfall is 383 mm) causes the net radiation to dissipate mainly into sensible heat flux (H). H followed the same temporal pattern as that of the net radiation and the daily mean value varied between a minimum of 37.2 and maximum of 131.6 Wm-2. The average annual value for a 24-hour period of H was 79.5 Wm-2. When the soil is moist, a significant part of the energy is dissipated into evaporation. λE showed two peaks during its annual cycle (Figure 6-2). The seasonal peaks for the entire Indus Basin correspond to the two agricultural seasons, once in rabi and once in kharif. The λE varied in the range from 10.9 (during winter with more cloud covers and lower temperatures) to 57 Wm-2 (during the periods of more canopy cover and higher temperature), with an annual average of 32.4 Wm-2. The average λE for the entire Indus Basin coincided with an ET of 1.2 mm d-1, but large variability among LULC classes occurred. The basin-wide evaporative fraction (Λ) is calculated as 0.28, equivalent to a Bowen ratio of 2.5. Hence, the amount of heat released into the atmosphere is 2.5 times more than for water vapor, if both are expressed in energy terms. Soil heat flux (G) is normally ignored when seasonal averages are considered because of its small scale. However, G can account for a significant portion (3 – 5 %) of the total
104
energy during summer (DOY 113 – 171); indicating that G is transferred from shallow to deep soil while for the rest of the year, the reverse process occurs. 6.4.2
Actual evapotranspiration estimates
The total transpiration and evaporation in the basin was estimated at 233 km3 yr-1 and 263 km3 yr-1, respectively. The major portion of water was consumed as non-beneficial evaporation (E), mainly from water-logged soils, dry soils and open water bodies. High annual ET values occurred on the alluvial plains as depicted in Figure 6-3. Irrigated agriculture is the major land use class (22.6%) in the basin and is a major consumer of water. It accounts for the annual ET rates of between 700 – 1200 mm and represents the middle part of the frequency distribution in Figure 6-4. The highest values (1200 – 1550 mm yr-1) were found in the tail end of the basin: in particular in the right bank of the Indus River, and southern parts towards the Indian Ocean, in the Sindh Province of Pakistan. Water-logged soils, rice paddies with shallow phreatic surfaces, and flooded areas normally occur in these parts of the basin, especially during kharif. Besides higher soil water content, factors such as higher solar radiation, higher air temperatures, more rainfall, and cultivation of higher consumptive use crops are the reasons for the higher ET.
Figure 6-3 ETLook estimated cumulative actual evapotranspiration for the hydrological year 2007 (January to December). The canal command areas for irrigated cropland are superimposed on the ET map.
105
Figure 6-4 provides the frequency distribution of annual ET. The average ET for all land use classes was 426 mm yr-1 during 2007. The 2% lowest value was 60 mm yr-1 and the 2% highest value was 1550 mm yr-1.
Figure 6-4 Frequency distribution of the ETLook estimated annual ET in the Indus Basin at spatial resolution of 1 km×1km for 2007. A sensitivity analysis was performed to understand the role of topsoil moisture data in the ET estimation procedure. The results are provided in Figure 6-5(a) and 6-5(b). Two land use classes and the average climatological condition of the year were used.
Figure 6-5 The response of evaporation, transpiration and evapotranspiration rates to surface soil moisture (n=100) for two representative land uses (a) Bare soil (71°22'54.123"E, 28°38'50.042"N) and (b) Irrigated rice-wheat rotation (75°23'53.59"E, 30°40'37.719"N).NDVI values of 0.05 and 0.67 were selected for bare soil and full grown irrigated rice – wheat land use, respectively.
106
The curves in the Figure 6-5(a) and Figure 6-5(b) display the response of model outputs (E and T) to variation in surface soil moisture. It is evident from the figure that the ET responds to surface soil moisture variability. Bare soil shows a fast response in E to surface soil moisture, while T remains negligibly small by absence of leaves (NDVI=0.05). E is the dominant flux in the overall ET process of bare soil. The response is curvilinear with the highest sensitivity occurring between 0.05 and 0.25 cm3 cm-3. The effects are lower when θ>0.25 cm3 cm-3 prevails. The parameter rsoil and the non-linearity of Equation (6.11) is one reason for this result. Another explanation is the non-linear relationship between the resistance and the latent heat fluxes that generally exists (not shown in this paper). The combined effect yields the S-type curve that is portrayed in Figure 6-5(a), and to a lesser extent in Figure 6-5(b). Figure 6-5(b) reveals that, in closed canopies, T dominates E. The net radiation is absorbed partially by the canopy, and the bare soil surface receives less energy for evaporation. At an NDVI of 0.67, E increases with increasing topsoil moisture, up to 0.18 cm3 cm-3. Apparently there is always soil evaporation in rice-wheat rotation systems, which is confirmed by many other agro-hydrological studies (e.g. Sarwar and Bastiaanssen, 2001; Ahmad et al., 2002). Canopy transpiration depends entirely upon root zone soil moisture rather than on the surface soil moisture. Therefore, T shows less sensitivity to surface soil moisture. The same can be concluded on the ET response to surface soil moisture changes. The effect of other input parameters on ET is summarized in Table 6.2. The lower and higher ranges of model input parameters are given, together with ET estimates for the average climate in the Indus Basin. The values of the input parameters were changed with specific increments. The sensitivity index (SI) was determined and the parameters were ranked based on the absolute values. The surface soil moisture appears to be the most important parameter for describing ET variability, with ET values ranging from 2.3 to 6.3 mm d-1, followed by the coefficient c in rsoil with a range of 2.5 to 6.2 mm d-1. The measurements of AMSR-E are thus essential for achieving proper ET modeling results, and form the key input parameter as was suggested. Table 6.2 Sensitivity of estimates of ET to model parameter values for irrigated rice-wheat land use. The last column depicts the sensitivity in terms of slope. ∆ is change, and M is mean. Parameter θAMSRE (cm3cm-3) rsoil, c (s m-1) rs,min (s m-1) NDVI (-) rsoil, b (s m-1)
Min 0.05
Input value Base Max 0.15 0.35
∆out
Mout
SC
SI
0.2
Resulted ET (mm) Min Base Max 2.3 5.1 6.3
4.0
4.3
13.3
0.62
∆in
Min
0.30
−10.0
−3.0
5.0
15
7.5
2.5
5.1
6.2
3.7
4.3
0.23
0.40
40.0
80.0
500
460
270
5.6
5.1
3.4
−2.2
4.5
−0.005
−0.3
0.05
0.45
0.67
0.62
0.36
3.9
5.1
5.9
2.0
4.9
3.2
0.24
10.0
30.0
70.0
60
40
5.8
5.1
4.3
−1.5
5.1
−0.025
−0.2
Base=Definition of fixed reference values during sensitivity test.
107
Model parameter sensitivity was investigated using a Monte Carlo simulation experiment with 1000 pairs of randomly generated input parameters. Based on this experiment the mean ET for “irrigated rice-wheat rotation” was 3.2 mm d-1 with an SD of 1.7 mm. The standard error for this distribution was 0.05 mm. A 95% confidence interval was used to determine the 2.5th and 97.5th percentiles, which ranged between 3.1 and 3.3 mm d-1. This level of uncertainty reflects that the model generates results with a potential error of 3.4%. 6.4.3
Validation
6.4.3.1 Field measurements Several field methods to measure ET fluxes can be used to validate the results. AsiaFlux has erected flux towers in China and India, but not in Pakistan (Mizoguchi et al., 2009). Therefore, to evaluate performance, ET estimates by ETLook were compared with the measured values given by PARC (1982) and Ahmad (2002) for 1975-80 and 2000-01 respectively (onwards referred to as “measured” values). Figure 6-6 shows the results of irrigated crops.
Figure 6-6 A comparison of evapotranspiration in rice, wheat rotation measured by previous studies, and those estimated by ETLook for 2007 in the Indus Basin. The correlations were good with an R2 of 0.70, and an RMSE of 163 mm (0.45 mm d-1). The RE between ETLook and measured ET values ranged from –1.9% to –28% with an average of –11.5%. The negative RE means ET figures from ETLook were lower than the field measurements. The regression line fitted through the origin has a slope of 0.89. This implies that ETLook estimates for 2007 were 11% lower than ET from previous studies. This difference of 11% is acceptable, considering the climatic differences between the years, the scale difference between in-situ measurements, and the 1 km remote sensing pixel size, as well as the uncertainty embedded in field measurements. Figure 6-7 shows the comparison of annual and seasonal ET from ETLook, from previous remote sensing and modeling studies (year 1995-96, 2001-02), and from other models 108
(year 2000, 1999-2006) (onwards referred to as “modeled” values). There is a reasonably good agreement at annual scale, with an R2 of 0.76 and an RMSE of 108 mm yr-1 (or 0.29 mm d-1). The values for the rabi season are reasonable (R2 of 0.60 and RMSE of 47.9 mm). However, the kharif season shows a relatively low R2 (0.54) and a high RMSE of 70.7 mm (or 0.39 mm d-1). The RE ranges between –13% and 18%, with an average of 6.5%. The RE for rabi ranges from –13% to 32%, with an average of 8%, while for kharif, the range is between –20% and 13%, with an average of –3%. This confirms the difficulty of modeling ET under cloudy conditions, and supports the inclusion of microwave data in computational processes, such as in ETLook. Note that there is no bias towards the lower or higher end of the ET data, and that the average slope is 1.05. Since Figure 6-6 suggests an underestimation of ET, and Figure 6-7 an overestimation, we believe that the ETLook performance is satisfactory for this data scarce basin.
Figure 6-7 Comparison of evapotranspiration modeled/estimated by previous studies conducted during the years 1995-96 (Bastiaanssen et al., 1999), 1995-2000 (Sarwar and Bill, 2007), 2001-02 (Shakoor et al., 2006; Ahmad et al., 2009) and 1999-2006 (Shakir et al., 2010) and ET estimated by ETLook for the year 2007 in the Indus Basin. ETLook has also been validated in regions other than the Indus, e.g. Australia and China. Some of these unpublished results are presented as a demonstration of the model performance under different climates and landscapes of ETLook. The National Water Commission of the Australian Government has provided Australian Water Resources (AWR) data for the year 2005. The water use data of eight states and 23 jurisdictional areas are publically available through http://www.water.gov.au/. The ET is computed as the difference between rainfall and runoff; storage changes and ground water are not considered. ET values from the water balance were compared against ETLook (Figure 6-8). Considering that the annual values were averaged over a large area, correlation was reasonable with an R2 of 0.70 and an RMSE of 112 mm (0.31 mm d-1). RE between the ETLook and AWR ET values ranged from –40% to 36% with an average of – 2.8%.
109
Figure 6-8 Comparison of evapotranspiration estimated by ETLook and estimates provided by Australian water commission for the year 2005. In China, ETLook estimated latent heat flux in the year 2009 was compared with flux tower measurements obtained from the eddy covariance flux measurement station at Heibei, Qinghai, China (37°36′ N, 101°20′ E) (Figure 6-9). Annual values correlated well with an R2 of 0.92 and an RMSE of 11mm (0.04 mm d-1). The RE of 9.5% between the two datasets is satisfactory since there is always a mismatch of scales between 1 km ETLook pixel estimates and ground measurements (flux tower).
Figure 6-9 Comparison of latent heat flux estimated by ETLook and measured by flux tower at Heibei, Qinghai, China (37°36′ N, 101°20′ E), for the year 2009. Each point represents 8-day average value. ETLook compared with other results and relevant statistics are summarized in the Table 6.3. The mean R2 of comparisons was 0.77 and the RMSE 0.28 mm d-1, while the absolute
110
mean of RE was 6.3%. This level of uncertainty needs to be considered in the presentation of ET mapping results using ETLook. Table 6.3 Validation statistics for ETLook results compared with previous studies’ at various spatial scales. Annual scale Study Areas Spatial Scale 2 R RMSE (mmd-1) RE (%) * Field 0.70 0.45 –11.50 Indus Basin Regional† 0.76 0.29 6.50 Basin – – 1.00 Field – – – Australia Regional – – – Basin¶ 0.70 0.31 –2.80 Field‡ 0.92 0.04 9.50 China Regional – – – Basin – – – Absolute Mean 0.77 0.28 6.3 Sources: * PARC (1982); Ahmad (2002). † Bastiaanssen et al., 1999; Shakoor et al. (2006); Sarwar and Bill (2007); Ahmad et al. (2009); Shakir et al. (2010). ¶AWR, ‡ Flux tower Heibei, China. 6.4.3.2 Water balance A map depicting differences in rainfall and evapotranspiration (R–ET) was prepared using TRMM rainfall data, calibrated by Cheema and Bastiaanssen (2012), and the ET results from this study (Figure 6-10). It shows areas with net water production (R>ET) and areas with net water consumption (ET>R). This indicates the value of spatial data to describe hydrological processes and withdrawals. The pixels that produce water (R>ET) are discharge areas responsible for streamflow and ground water recharge. These areas are in the upstream parts of the basin, and are the source of the rivers Indus, Jhelum and Sutlej that feed the large reservoirs Tarbela, Chashma, Mangla and Bhakra, respectively. Areas with sparse vegetation and low ET also have higher rainfall than ET and are water producing areas. Large parts of the Tibetian Plateau comprise such areas. The Rajasthan Desert between India and Pakistan also exhibits positive values of R–ET, which suggests groundwater recharge. Net water consumption areas are generally the irrigated areas, lakes and reservoirs. Irrigation increases crop ET far beyond the level of rainfed crops. In the Indus Basin, 30.3 % of the total land area is composed of net consumer areas, and 22.6% (26.02 mha) is irrigated land. The mountain valleys are net water consumers; the valleys receive both seepage water through the groundwater system and surface water from the higher elevated mountains, which generally results in shallow water table areas in the vicinity of streams.
111
Figure 6-10 Rainfall-Evapotranspiration (R – ET) difference map of the Indus Basin for the hydrological year 2007. The water balance of the irrigated areas covering 26.02 mha was computed to validate ET results on a large scale (Table 6.4). Total annual groundwater abstraction in Pakistan’s part of the Indus Basin is given by Qureshi et al. as 51 km3. Chadha (2008) estimated that for the Indian part of the Indus Basin 18.5 km3 is being abstracted from the groundwater system. This totals to 69.5 km3 yr-1. The surface water releases into the main canals add up to 122 and 36 km3 yr-1 in Pakistan and India, respectively. These data on releases from Tarbela, Mangla, Chashma, Thein, Pong and Bhakra reservoirs, as well as flows into the main irrigation canals were obtained from Punjab Irrigation Department and Indus Water Commission, Pakistan. If we assume a conveyance efficiency of 80% that is locally checked and verified (Habib, 2004; Jeevandas et al., 2008), then 126.4 km3yr-1 will arrive at the farm gate through the network of canals. Adding the 69.5 km3 of groundwater from locally operating tube wells, the total amount of water used is about 196 km3. If we take an on-farm irrigation efficiency of 80% to describe losses of water that is not properly stored in the root zone, the total ET from irrigation will be 156.8 km3 yr-1. Note that a regional scale on-farm efficiency for the total irrigation system includes recycling of non-consumed irrigation water (Perry, 2007). A total irrigation efficiency of 64% (0.8×0.8) for one contiguously irrigated alluvial plain is realistic. It can however also be 60% or 70%. The rainfall over the irrigated area is 117 km3 yr-1. The net rainfall infiltrated into the soil – 112
after runoff and percolation losses - and available for uptake by roots is 94 km3 yr-1 (assuming 80% efficiency). The total ET for the irrigated land on the basis of water balance is 94 + 156.8 = 250.8 km3 yr-1, or 964 mm yr-1. ETLook results provided an estimate on the basis of the energy balance as being 254 km3 yr-1, or 974 mm yr-1. While this is a difference of 1 % only, we conclude that the results are congruent and within acceptable ranges that are usually related to water balances. Table 6.4 Water balance for the irrigated areas in the Indus Basin during the hydrological year 2007. Annual rainfall R km3 mm
6.5
117 451
From surface water 158 607
Irrigation (IRR) At From farm ground gate water 126.4 69.5 486 267
Total (column 4+5) 196 753
Evapotranspiration (ET) IRR R Total
156.8 603
94 361
250.8 964
ET ETLook
254 974
Summary and conclusions
The first requirement for an operational ET monitoring system is that the satellite data must be available at all times. Microwave satellite data are operationally provided – even under all weather conditions – and their growing number of standard databases form an attractive source for developing ET models. ETLook can assess the spatial and temporal (daily, 8-day, or monthly) patterns of the surface energy balance and actual evapotranspiration. Computing E and T separately, on the basis of the energy balance, has the advantage that complex transient moisture flow computations in the unsaturated topsoil can be circumvented. The novelty of this paper is a doable computational method for non-beneficial E and beneficial T that can be applied under conditions of persistent overcast skies, and in data scarce environments. The sensitivity analysis revealed that the surface soil moisture is the most important parameter for describing ET variability. Variability of surface soil moisture revealed that the ET values for rice-wheat rotation system on an average day ranged between 2.3 to 6.3 mm d-1, followed by the coefficient c of soil resistance, with a range of 2.5 to 6.2 mm d-1. Good agreement was attained between ETLook and previously conducted field measurements and remote sensing studies. R2 varied between 0.70 and 0.76 at annual time scale (RMSE: 0.45 and 0.29 mm d-1 respectively). Tests in Australia and China provided similar agreements based on watershed measurements. The water balance of 26.02 mha of irrigated land is congruent and matches generic data on surface water and groundwater supply. There are discrepancies in timescales shorter than a year. However, no bias was evident towards the lower or higher end of the ET values. The observed errors could be due to the meteorological differences between the years of study. The determination of wind speed and air humidity needs more attention in future studies. Better quality soil maps will also improve the quality of the ET results. The average value for latent heat flux in the Indus Basin is 32 Wm-2, which corresponds with an ET of 1.2 mm d-1 (426 ± 14.5 mm yr-1). The average value for rainfall is 383 mm
113
yr-1. Over-exploitation and negative storage changes of water occur at the basin scale (ET >R). The negative change in storage can be ascribed to reduced volumes of water stored in reservoirs and aquifers. Retirement of glaciers also contribute to water storage changes. The power of having access to daily soil moisture data from passive microwave measurements onboard satellites is at the same time limited by the low resolution of AMSR-E surface soil moisture pixels (25 km). Several methods exist to deal with the downscaling of soil moisture, but the best method that is doable under a wide range of conditions still needs to be found. More sophisticated solutions on downscaling can be gleaned from topographic information (e.g. height above drain, distance to drain, accumulated upstream drainage area) and soil properties (infiltration capacity, water holding capacity, drainage capacity). It is expected that satellites with synthetic aperture radar will provide high-resolution soil moisture values in an operational context in the near future, in addition to the thermal data that are already used for routine mapping of soil moisture under clear sky conditions. The analytical relationships between topsoil and subsoil moisture need improvement and more testing. This analysis was conducted for a one year cycle only, to raise confidence in using the first version of ETLook algorithm (ETlook 1.0). Future analysis with longer time series is recommended, since shorter time series may be of low significance. Despite the limitations mentioned, the current paper has demonstrated that the ET results show potential for determining water depletion in ungauged basins.
114
7 Spatial quantification of groundwater abstraction for irrigation in the Indus Basin using pixel information, GIS and the SWAT model Chapter based on: Cheema, M.J.M., Immerzeel, W.W. and Bastiaanssen, W.G.M. 2012. Spatial quantification of groundwater abstraction for irrigation in the Indus Basin using pixel information, GIS and the SWAT model. Groundwater, (under review).
7.1
Introduction
Quantification of groundwater abstraction, especially in arid regions where recharge is genuinely small, is of prime importance for sustainable basin scale water resources. Rapid population growth and increased irrigation development for food security has resulted in exhaustive groundwater abstractions in many alluvial plains (e.g. Foster and Chilton, 2003; Shah et al., 2007). Wada et al. (2010) created a global map of groundwater abstractions in 2010, which indicates several areas with abstractions exceeding 100 mm yr1 . Siebert et al. (2010) developed a global inventory on groundwater which estimates that 43% of the total consumptive irrigation water use is met through groundwater. Groundwater abstractions are temporally episodic and spatially variable and depend upon the crop irrigation needs, surface water availability and water quality. The spatial variability in groundwater availability and water requirement by crops complicate the quantification of abstractions. The Indus Basin is a typical example showing high variability in land use, climate, canal water availability, soil types and irrigation practices without any regulation in place to measure the groundwater abstraction. Irrigated agriculture is common land use in the trans-boundary Indus Basin covering 23% of the total area. Groundwater is utilized solely or in conjunction with surface water to augment the insufficient and unreliable surface water supplies. Different studies (e.g. Scott and Shah, 2004; Sarwar and Eggers, 2006) have estimated that about 40 to 50% of irrigation needs of the basin are met through groundwater abstraction. Arshad et al. (2008) has estimated that the groundwater abstraction is up to 60% in the areas having rice – wheat rotation land use. The continuous abstractions, in high quantities, can adversely affect the overall water balance when the average value consistently exceeds the recharge over a long period. Therefore, accurate information on spatial groundwater abstraction and depletion is urgently required to support development of management plans. The use of tabular values per district or province is no longer sufficient for achieving progress in controlling over-exploitation of the aquifers. Policy makers can take more effective actions if groundwater activities are expressed by means of pixels. The benefits of pixel-based information include the availability of information in terms of geographical coordinates, coverage of a discrete land area, quantified abstraction rate, and the identification of land owners. This paper employs new methods to obtain groundwater information from pixels, using satellite measurements, GIS systems and hydrological models. The groundwater use estimation is normally carried out using the tubewell utilization factor technique or water table fluctuation methods (Maupin, 1999; Healy and Cook, 2002; Qureshi et al., 2003). These methods become less suitable when applied at basin scale due
115
to the poor spatial density of the point measurements. Alternatively, abstraction data on groundwater use can be derived from hydrological models. The success of these models depends primarily on availability of comprehensive input data and how well the models are calibrated (Zhang et al., 2008). Long-term time series data sets with high spatial detail are difficult to obtain in spatially heterogeneous basins with a limited gauging network (Sivapalan et al., 2003). Extreme spatio-temporal variability in precipitation, evapotranspiration in combination with low density surface and groundwater point measurements makes models prone to errors. Measurements at few stations and their spatial extrapolation for the entire basin may yield unreliable estimates of water use. The uncertainties associated with the measured input data may also lead to biases in the model estimations (Srinivasan et al., 2010). The use of remote sensing techniques in combination with spatially distributed hydrological models has shown great potential to overcome these difficulties. Many researchers, including Houser et al. (1998), Boegh et al.(2004), Bastiaanssen et al. (2007), and Immerzeel et al. (2008a) have successfully used remote sensing to parameterize hydrological models. Remotely sensed evapotranspiration has also been used successfully in calibrating hydrological models (e.g. Droogers and Bastiaanssen, 2002; Immerzeel and Droogers, 2008; Jhorar et al., 2011). In the present study we develop, for the first time, a detailed Soil Water Assessment Tool (SWAT) model application that encompasses the entire transboundary Indus Basin. The SWAT model is forced and parameterized, using remote sensing derived datasets of elevation, land use, temperature, and precipitation. We then calibrate the SWAT, at the highest possible spatial detail, using estimates of actual evapotranspiration (ET) based on the ETLook algorithm as described in Cheema et al. (2012). We then use the calibrated model to spatially estimate the total irrigation water supply at the farm gate of irrigated areas following the principles similar to Droogers et al. (2010b). The surface water diverted at the head of the canal command areas has been integrated with this dataset to isolate the spatial distribution of irrigation by 1 km pixels of gross groundwater abstractions. The main objective of this paper is to explain, demonstrate and validate the methodology to determine groundwater abstractions using 1 km pixels. It identifies the hotspot areas with a discretization of 100 ha and consequently groundwater-pumping activities are no longer a hidden piece of information.
7.2 7.2.1
Material and methods Study area
The Indus Basin lies between latitude 24°38′ to 37°03′ N and longitude 66°18′ to 82°28′ E located in four countries (Figure 7-1). The lifeline of the Indus Basin is the Indus River that traverses China, Afghanistan, India and Pakistan, moving from upstream to the downstream end of the basin. The total size of the basin is 116.2 million ha (mha). In total 53% of the area of the basin is located in Pakistan. The area in India is 33% followed by China and Afghanistan with 8% and 6%, respectively. The elevations range from 0–8600 m above mean sea level (a.m.s.l). The basin exhibits complex hydrological processes due to variability in topography, rainfall, land use, and water use. The average annual rainfall 116
varies from less than 200 mm in the desert area to more than 1500 mm in the north and north-east of the basin. The thirty year (1961 – 90) average reference crop evapotranspiration (ETo) varies between 650 mm in the northern parts and 2000 mm in the southern desert areas of the basin. Water is diverted from the Indus River and its major tributaries (Jhelum, Chenab, Ravi, Beas and Sutlej) through a network of canals to irrigate the agricultural lands. The main reason for this diversion is that the rainfall is inadequate to fulfill crop water requirements. However, the availability of canal water is unreliable and that has motivated farmers to augment shortages in surface water by groundwater resources (Shah et al., 2000). Two agricultural seasons kharif (May to October) and rabi (November to April) are in practice. The main crops cultivated are wheat, cotton, rice, fodder, sugar cane and fruit orchards. Vegetable are also raised in some areas. Perennial sugarcane and seasonal fodder is also grown in tracts. Wheat is the major crop grown in rabi. Rice and cotton are the major crops of kharif season.
Figure 7-1 Location of the Indus Basin and land use scheme used in the SWAT model. The codes are explained in Table 7.1. 7.2.2
Soil and Water Assessment Tool
The Soil and Water Assessment Tool (SWAT) is a process based distributed hydrological model which provides spatial coverage of the integral hydrological cycle including atmosphere, plants, unsaturated zone, surface water, and groundwater. A comprehensive description of the model can be found in literature (e.g. Arnold et al., 1998; Srinivasan et al., 1998; Neitsch et al., 2005), however, for the convenience of our reader, we summarize the SWAT model in the following paragraphs.
117
SWAT provides continuous simulation of evapotranspiration, percolation, return flow, storage change, surface runoff, channel routing, transmission losses, crop growth and sediment transport (Kannan et al., 2011). We selected SWAT because it represents a simple groundwater reservoir that acts as an interface between soil moisture in the unsaturated zone, groundwater storage in the saturated zone and surface water systems. The latter is essential for water exchanges and for understanding recycling mechanisms of non-consumed water. The spatial water balance of the unsaturated zone reads as: 7.1 where ΔSus is the change in storage of the unsaturated zone (mm), RSWAT is the amount of precipitation (mm), IRRSWAT is the amount of total irrigation applied (mm), Qsurf is the amount of surface runoff (mm), ETSWAT is the actual evapotranspiration (mm), Qlat is the amount of lateral flow through the unsaturated zone (mm), Qperc is the amount of percolation (mm) and Cr is the capillary rise (mm). Net Groundwater Use (NGU) is introduced by Ahmad et al. (2005) as the difference between vertical exchanges (abstraction, recharge, capillary rise), and NGU can be computed from this data. SWAT computes daily ETo and potential plant transpiration (Tp) according to meteorological input data and crop coefficients based on Penman-Monteith method (Monteith, 1965). Daily crop height and leaf area index (LAI) are controlling aerodynamic and canopy resistances and are used in calculating Tp. Potential soil evaporation is an exponential function of ETo and the soil cover, which is reduced during periods with high plant water use. Actual soil evaporation is limited by the soil water content (θ) and is reduced exponentially when θ drops below field capacity. The potential plant water uptake can be defined as follows for calculating the actual plant transpiration: 7.2 where wup,z (mm) is the potential plant water uptake from the soil surface to a specified depth from the soil surface on a particular day, Tp(mm) is the maximum plant transpiration on a given day, βw (-) is the water use distribution parameter, z is the depth from the soil surface (mm), and zroot is the depth of root development in the soil (mm). Actual plant water uptake equals actual plant transpiration and shows exponential reduction as θ drops below field capacity. Hence, soil moisture regulates actual transpiration fluxes, and deviations of soil moisture will be propagated into deviation in actual transpiration. Actual evapotranspiration modeled by SWAT (ETSWAT) is the sum of interception, actual soil evaporation, and actual plant transpiration. Soil moisture in the root zone is a function of irrigation water supply at the farm gate. Irrigation has been modified consecutively until ETSWAT on the water balance of Eq (7.1) was matching ET from ETLook. Some major principles of ETLook will be explained in the latter section. This technical approach provides a vehicle to infer the actual irrigation applications in a realistic way, without the need to construct large databases on irrigation water distribution. Santos et al. (2010) used a similar approach for irrigation systems in Spain.
118
The water balance of the saturated zone, i.e. -shallow aquifer- can also be computed using SWAT. The net groundwater depletion (DEPgw: amount of water leaving the shallow aquifer) can be estimated for the irrigated areas using information of canal water losses (LOSScw) as: 7.3 where Qgw is the return flow from the shallow aquifer towards the river (mm). Note that the capillary rise is considered as a component of the irrigation water supply. IRRgw is the gross groundwater that is abstracted from the shallow aquifer. The SWAT model subdivides the Indus Basin into sub-basins, which are further divided conceptually into hydrological response units (HRU). HRUs are based on unique and homogenous combination of land use and soil type. We delineated 132 sub-basins with average area of 8000 km2 and 2459 HRUs in the Indus Basin. We identified 489 HRU as irrigated where conjunctive use is practiced. We used the average data on radiation, wind speed, relative humidity, air temperature, and rainfall for each sub-basin. We fed the above mentioned data into SWAT for the computation of Tp and wup,z. We used a model spin up period of two years to initialize the model, in particular soil moisture. 7.2.3
Data
We obtained the rainfall data in SWAT as 25 km grids from the Tropical Rainfall Measurement Mission (TRMM) as described by Huffman et al. (2007). The products 3B42 (daily) and 3B43 (monthly) were collected for the year 2007. Daily 3B42 data were aggregated per month and a correction factor for each month was established using calibrated 3B43 monthly data (Cheema and Bastiaanssen, 2012). Using these correction factors, daily grids of rainfall (RTRMM) were generated from January 1, 2007 to December 31, 2007. The rainfall that was deposited below a threshold temperature was classified as snow. Solar radiation has been computed from the extra-terrestrial radiation in association with an atmospheric transmissivity. The atmospheric transmissivity was inferred from optical depth information obtained at 1 km pixel resolution Moderate Resolution Imaging Spectro radiometer (MODIS) cloud product (MYD06_L2) downloaded from https://wist.echo.nasa.gov/~wist/ api/imswelcome/. The spatially distributed meteorological data for maximum (Tmax) and minimum (Tmin) air temperature, relative humidity (RH) and wind speed (U2) was prepared using meteorological distribution model (Daymet) described by Thornton et al.(1997). The meteorological station data was obtained from 65 meteorological stations under the aegis of the Pakistan Meteorological Department (PMD). Weather station data for India, China and Afghanistan were obtained from the National Oceanic and Atmospheric Administration (NOAA) National Climatic Data Center (NCDC). The NCDC collects meteorological data from real time reporting stations worldwide in agreement with World Meteorological Organization (WMO) regulations (ftp://ftp.ncdc.noaa.gov/pub/data/gsod/). Data from 16 stations with complete datasets were downloaded. Air temperature, relative humidity and wind speed data from 81 stations, collected at standard height of 2 m, were
119
obtained. The gridded daily datasets were aggregated per sub-basin, which provided 132 hypothetical meteorological stations with uniform distribution. A detailed land use and land cover (LULC) map developed by Cheema and Bastiaanssen (2010) was used to infer information on different LULC classes in the Indus basin. Twenty seven LULC classes were classified. Those classes were clustered into 21 land use classes based on SWAT land use library. The details of seven irrigated land use classes including growing season and irrigation depths are provided in the Table 7.1. The actual dates vary spatially and temporally. For example, wheat crop is sown between 1–30 November and irrigation depths may vary from 45 mm to 105 mm, while number of irrigations may vary from 3 to 5. These irrigation practices, provided by Pakistan Agricultural Research Council, PARC (1982) and Ahmad (2009), were adopted for initiation of the SWAT model. The FAO digital soil map of the world (FAO, 1995) was used to derive soil properties with the aid of pedo-transfer functions (Droogers, 2006). Forty eight different soil units were found in the basin and the alluvial plains are predominantly characterized by vertisols and the steeper slopes by fluvisols. A GTOPO30 – DEM was used in watershed delineation and defining streams. A stream network developed by International Water Management Institute (IWMI) was used in stream delineation process where topography was relatively flat. The surface water supplies at canal heads for various CCA were obtained from Punjab Irrigation department (PID), Water and Power Development Authority (WAPDA) and Indus Water Commission (IWC), Lahore, Pakistan. Table 7.1 Land use and land cover classes in the Indus Basin. The cropping period and number of irrigation with depths are also provided for irrigated land use classes. SWAT Land use class Area Growing Cropping Irrigation No Code (Depth,mm) (%) season period 15 May – 15 Nov 5 (120) Irrigated mixed AGI1 3.6 kharif 15 Nov – 30 Apr cotton,wheat rabi 4 (75) rotation/orchards 15 May – 15 Nov 5 (120) Irrigated mixed AGI2 4.3 kharif 15 Nov – 30 Apr cotton,wheat rabi 4 (75) rotation/sugarcane 15 Jun – 30 Oct Irrigated rice,wheat AGI3 8.3 kharif 15 (100) 15 Nov – 30 Apr rotation rabi 4 (75) 15 Jun – 30 Oct Irrigated mixed rice,wheat AGI4 2.2 kharif 15 (100) 15 Nov – 30 Apr rotation/cotton rabi 4 (75) 01 Aug – 30 Oct Irrigated fodder,wheat AGI5 2.3 kharif 6 (75) 15 Nov – 30 Apr rotation rabi 4 (75) 15 Jun – 30 Oct Irrigated rice,fodder AGI6 1.7 kharif 15 (100) 15 Jan – 30 Apr rotation rabi 6 (75) 15 Jun – 30 Oct Irrigated mixed rice,wheat AGI7 0.2 kharif 15 (100) 15 Nov – 30 Apr rotation/sugarcane rabi 4 (75) Forests/cropland alpine AGR1 2.5 Rainfed crops wheat/grams AGR2 1.2 rabi Rainfed crops mixed AGR3 1.5 kharif cotton,wheat rabi 120
rotation/fodder Rainfed crops general Rainfed crops and woods Bare soil Forests deciduous alpine Forests evergreen broadleaf Forests evergreen needleleaf Pastures deciduous alpine Pastures deciduous lowland Pastures evergreen Savanna deciduous Savanna evergreen closed Savanna evergreen open Very sparse vegetation Snow and ice permanent Snow and ice temporary Urban and industrial settlements Water bodies 7.2.4
AGR4 AGR5 BARE FRSD FRSD FRSE
10.1 3.3 6.3 3.2 0.5 3.8
PAST PAST PAST RNGB RNGB RNGB RNGE SAIP SAIT URBN
6.7 6.5 3.0 11.1 1.9 2.9 1.9 3.6 4.8 1.6
WATR
1.0
ETLook
The ETLook algorithm uses a two layer Penman–Monteith equation (Monteith, 1965) by dividing each pixel of the image into bare soil and canopy to infer evaporation (E) and transpiration (T), respectively(Pelgrum et al., 2010) . The Penman–Monteith equation for E and T can be written as: 7.4
7.5 where E represents evaporation and T represents transpiration in Wm-2. Δ (mbar K-1) is the slope of the saturation vapor pressure curve which is a function of air temperature (Tair, °C) and saturation vapor pressure (es, mbar); Δe(mbar) is vapor pressure deficit, which is the difference between the saturation vapor content and the actual vapor content; ρ (kg m3 ) is the air density and cp is specific heat of dry air =1004 J kg-1 K-1; γ (mbar K-1) is the psychometric constant; Rn,soil and Rn,canopy are the net radiations at soil and canopy respectively; rsoil and rcanopy are resistances of soil and canopy, while ra,soil and ra,canopy are aerodynamic resistances for soil and canopy. All resistances are in s m-1. The resistance for E is based on top soil moisture measurements from satellite data. The resistance to T is based on root zone soil moisture that is estimated from topsoil moisture and the presence of photosynthetically active vegetation. Total ET is the sum of T and E and the units can 121
be converted to mm day-1. Details of the ETLook algorithm can be found in Pelgrum et al. and Cheema et al. (2012). In the current study, the ET data is available with an eight day interval for the period of one calendar year from January 1st to December 31st, 2007. 7.2.5
Model calibration procedure
The calibration of the SWAT model was performed by comparing SWAT modeled ET (ETSWAT) with spatially observed ET (ETETLook) for all HRUs. As in complex distributed hydrological model with numerous parameters with a high spatial and temporal heterogeneity, conventional stream flow calibration has a large risk of equifinality problems (Beven, 2006). Moreover, it becomes ineffective in basins, such as the Indus, where stream flow is under human control (Immerzeel and Droogers, 2008). A stepwise heuristic iterative approach was therefore adopted to perform calibration by adjusting the key soil and groundwater parameters. A number of important model parameters, which have a large influence on ET, were used in the model calibration. The sensitive parameters that have control on E and T fluxes were identified from the previous literature (e.g. Immerzeel and Droogers, 2008; Immerzeel et al., 2008a; Githui et al., 2011; Kannan et al., 2011). The soil water holding capacity (Ф), capillary rise (Cr), depth of the evaporation front (Ψ) and the relative water uptake by plant roots as a function of soil moisture (wup,z) have been calibrated. Their allowable ranges were bound to 0.06 to 0.60 mm/mm for Ф, 0.02 to 0.9(Pelgrum et al., 2010) for Cr, 0.01 to 1.0 for Ψ and 0.01 to 1.0 for wup,z. Forty eight soil types with two layers resulted in 96 different parameters for Ф to be optimized. Ψ, Crand wup,z coefficients were optimized for each HRU where irrigation was absent. Most parameters were optimized per HRU of each land use class to capture the spatial heterogeneity because land use information is available at relatively detailed level compared with the soil type information. Default values of these parameters were adopted for base run and the implemented adjustments were constrained by the ranges of parameters suggested by Neitsch et al.(2005). Three common statistical indicators, as described by Hoffmann et al.(2004), were used to quantify the achieved level of calibration and to evaluate the SWAT model’s overall performance .The Nash-Sutcliffe model efficiency (NSE) (Nash and Sutcliffe, 1970), Pearson’s correlation coefficient (r) and percent bias (Pbias) between modeled and observed ET were determined, which are given as: 7.6 7.7
7.8 where, O represents the observed (ETETLook) and M represents the modeled (ETSWAT) ET. Ō is mean observed and n is the total number of observations. NSE ranges between −∞ and 1, with NSE = 1 being the optimal value. The Pbias reveals to which degree the modeled
122
value is smaller or larger than the observed values given in percentage. Small values of Pbias are preferred. 7.2.6
Pixel based groundwater abstraction data
The pixel information on actual evapotranspiration using ETLook (ETETLook) and rainfall (RTRMM) from TRMM is valuable additional spatial information. This information can be used to infer the total irrigation water supply at the farm gate for each pixel (IRRRS), when being integrated with HRU fluxes obtained from SWAT calculations of Eq. (7.3). Assuming that capillary rise and storage changes to be part of applied water, the analytical expression becomes: 7.9 Irrigation water, diverted to main canals irrigating a specific canal command area (CCA), was aggregated to monthly and annual irrigation volumes. The resulting vector maps of canal water supplies for each CCA were prepared. The supplies were then converted into depths by dividing over the area of each CCA. The map of irrigated land use was overlaid with the map of CCA to identify the irrigation command areas matching with each HRU. The result is a canal irrigation vector map (IRRcw) that has a spatial refinement as compared to the HRU. The overlay helped to partition the total irrigation water supply (IRRRS) into canal water use (IRRcw) and finally gross groundwater abstraction from shallow aquifer (IRRgw) (Figure 7-2): 7.10 The annual depths of canal water vary from 200 mm to 1700 mm per diversion. Conveyance efficiency of 70% was considered for canals in Pakistani part of the Indus Basin (Habib, 2004; Arshad et al., 2005; Kreutzmann, 2011) and 80% for canals in Indian part of the Indus Basin (Bastiaanssen et al., 1999; Kroes et al., 2003; Jeevandas et al., 2008).
Figure 7-2 Schematic diagram showing data sources to infer groundwater abstraction information at 1 km pixel.
123
7.3 7.3.1
Results and discussion Model calibration
The model performance was evaluated using three statistical indicators namely NSE, Pearson’s correlation r and Pbias. The performance was assessed at sub-basin and HRU levels. At sub-basin level, NSE of the calibrated model was 0.93, while under uncalibrated conditions the NSE was 0.52. Improvement in Pearson’s correlation was observed from 0.78 to 0.97, suggesting a strongly improved correlation between ETSWAT and ETETLook when model parameters where adjusted according to the existing ET layers from the energy balance. The Pbias resulted in -0.4, which is very low as compared to 17.3 (base run) indicating no systematic under or over prediction of ET is observed at subbasin level. Figure 7-3 shows the correspondence between the modeled and observed ET at sub-basin scale.
Figure 7-3 Comparison between modeled (ETSWAT) and observed (ETETLook) actual evapotranspiration for 132 sub-basins in the Indus Basin. Figure 7-4 shows the results for 489 HRUs that contain irrigated land only. The NSE, Pbias and Pearson’s r were 0.93, -2.3 and 0.97. This level of agreement shows that SWAT can produce ET values from the soil water balance that is very similar to the ET of irrigated crops as interpreted from satellite images. Figure 7-5 shows the spatial patterns of the annual sum of ET. The spatial pattern of ET modeled by SWAT was in good agreement with ETETLook. However, some local differences were observed and the reason is that ETSWAT results show a larger within land use variation.
124
Figure 7-4 Comparison between modeled (ETSWAT) and observed (ETETLook) actual evapotranspiration for 489 irrigated HRUs.
Figure 7-5 Cumulative ET from 1st January 2007 to 31st December 2007 derived with ETLook (left) and modeled with SWAT (right). Both figures display aggregated data per HRU for the sake of compatibility. Figure 7-6 provides more insight in the temporal ET patterns. The monthly ET shows good agreement between the modeled monthly ETSWAT and ETETLook with a correlation coefficient 125
(R2) of 0.87. November and December show a low ET rate due to lower solar altitudes and low ambient temperature. The warm atmosphere and large rainfall amounts due to monsoon system are the reason for peak ET rates during July. The strong reduction in ET during the month of May –when land is prepared for the summer crop– is picked up well by SWAT.
Figure 7-6 Comparison of monthly ETETLook against ETSWAT for all 489 irrigated HRU’s during 2007. It is notable that in the months of September and October ETETLook is higher than the modeled values while the reverse is observed for the months of June and July. One of the main reasons is that most of the fields became fallow due to the harvest of kharif crops. Normally harvesting starts at different dates at different locations depending upon the crop maturity. However, in SWAT parameterization each land use was assigned with a single date of harvesting. The moisture retained especially in paddy fields, contributed to evaporation thus causing higher ETETLook than the ETSWAT. In the contrary, during the months of June and July, ETSWAT has shown higher values. These two months corresponds to the monsoon months with higher rainfall and considerable irrigation is supplied to the crops especially to rice and cotton. It suggests that ETSWAT over estimates ET during these months. Moreover, during model simulations, the specified irrigation dates for a particular crop were the same for the HRUs representing that crop. It was assumed that the total area under that particular crop is irrigated on the specified date and specified depths. In reality, the irrigation of a particular crop is completed over a period of time depending on the farmer’s rotation for canal supplies. Another source of temporal discrepancy could be that during the simulation periods canal water supply to a certain CCA is taken as constant for
126
the entire command area that in reality varied based on the distance from the canal head. All these factors can be the cause of the deviations. Overall it is concluded that the calibration on actual ET is highly satisfactory given the high correlation, NSE and low biases. The Indus SWAT model is calibrated at the highest possible spatial detail (HRU level) and the temporal ET patterns are also simulated with reasonable accuracy. The adopted calibration strategy is effective and outperforms earlier work in this field (Immerzeel and Droogers, 2008). 7.3.2
Spatial patterns of water supply and consumption
The spatial distribution of total irrigation estimated by applying pixel information (IRRRS), surface water supplied at farm gate (IRRcw), percolation to aquifer (Qperc), gross irrigation from groundwater (IRRgw) and related groundwater depletion (DEPgw) are presented in Figure 7-7 (a – e). The total canal water available at the farm gate for each canal command is estimated at 113 km3 (or 434 mm) (Figure 7-7(a)). This amount is computed from the reservoir releases and reported conveyance losses. Canal water available at farm gates varies from 200 to 900 mm yr-1. This spatial variability in canal supplies is due to the nonperennial system and variability in water released from the reservoirs. The highest rate of IRRcw is observed in lower Indus especially in Sindh province. These areas are suffering from salinity and a deteriorated groundwater quality (Qureshi et al., 2010b) owing to the high intensity of rice cultivation on this low laying river plain areas. The total irrigation estimated by pixel information (IRRRS) using Equation 7.9 is 181 km3 (696 mm). Total applied water varies between 200 to 1400 mm yr-1 in the irrigated areas across the basin (Figure 7-7(b)). The irrigation application is found to be largest (700 – 1400 mm yr-1) in Jacobabad, Shikarpur and Larkana districts of northern Sindh. Higher values are also found in Noshehroferoz and Nawabshah districts in southern Sindh. Narowal and Gujranwala districts in northern Punjab, Jhang, Toba Tek Singh and Pakpattan districts in central Punjab and Khanewal, Multan, Lodhran and Vehari districts of southern Punjab have received higher total irrigation. Districts of Amritsar, Ludhiana, Jalandhar, and Ferozpur in Indian Punjab and Patiala district of Haryana also have received higher irrigation rate. The reason of this higher irrigation is the large-scale cultivation of high water consumptive crops like rice, sugarcane, cotton etc. Aquifer recharge (Qperc) ranges between 10 – 600 mm during the year as depicted in Figure 7-7(c). A percolation of 71 mm from irrigated fields (especially in irrigated rice – wheat land use) and high rainfall during monsoon are the sources of this recharge. The losses from canals (LOSScw = 144 mm) also contributes to the aquifer. This LOSScw does not indicate that water is lost permanently from the system but it represents non-consumed water that can be potentially recycled.
127
(a)
(b)
(d) 128
(c)
(e)
Figure 7-7 Spatial maps of (a) Canal water supplies (b) Irrigation estimated using remote sensing products (c) Percolation (d) Gross groundwater abstraction and (e) Net groundwater depletion The canal water supplies are not sufficient to meet the crop water requirements and hence do not match with IRRRS. The deficit is met through groundwater irrigation and Figure 7-7(d) shows gross groundwater abstraction rates (IRRgw) for each pixel estimated by using Eq 7.10. The data shows that, on annual basis, an amount of 300 to 900 mm is abstracted from the aquifers for irrigating crops. The highest values for IRRgw are observed in middle and northeastern parts of the basin. These areas contain relatively good quality groundwater resources (Arshad et al., 2007) and are located in the Punjab province of Pakistan and the Indian state of Haryana.
129
The largest groundwater abstractions in Pakistan occur in the province of Punjab. The IRRgw in the districts of Multan, Khanewal, Pakpattan, Vehari and Lodhran covering lower Bari doab (area between Ravi and Sutlej) ranges between 300 – 700 mm. The average abstraction in lower Bari doab is 400 mm. The middle and lower Rechna doab - area between Chenab and Ravi rivers - also shows similar trends and IRRgw ranges between 200 – 600 mm. In the upper Rechna doab, fragmented pockets of high IRRgw (500 -700 mm) are observed especially in the districts of Narowal, Gujranwala and Sialkot. The reason of this high groundwater abstraction is that the rice is extensively cultivated but canal water do not suffice. IRRgw is also observed in pockets of the Sargodha district in Chaj doab (area between Jhelum and Chenab rivers). Groundwater is mostly saline here but fragmented lenses of fresh water are available. Conjunctive use of groundwater with surface water is a normal practice in these areas. In Sindh province, groundwater abstraction is fragmented with significant groundwater abstractions occurring in district of Larkana, Jacobabad and Shikarpur. These areas have higher annual ET because of rice cultivation or high cropping intensities. Groundwater recharge by percolation from fields and canals can be recycled which results in conjunctive use of groundwater and surface water in these northern parts of Sindh province (Siebert et al., 2010). The districts of Jalandhar, Phagwara, Ludhiana and Bathinda of Indian Punjab are vulnerable to extensive groundwater pumping. The value ranges between 400 to 900 mm. Irrigated rice,wheat rotation is the dominant land use that requires extensive irrigation to meet crop water demand. The surface water supplies are not sufficient to meet the needs therefore the deficit is met through groundwater. Large number of small capacity tubewells is installed to pump groundwater. According to Shankar et al.(2011), tubewells density in the Punjab and Haryana state is 27 and 14.1 tubewells per km2 in 2001.and the number is increasing. The information on groundwater abstraction in different LULC is important to identify the opportunities for saving water or relocation of water. The analysis for irrigated LULC is provided in the Table 7.2 which shows that the maximum groundwater was abstracted (64% of total groundwater abstraction) in the “irrigated rice,wheat rotation” land use. It is 53% of the total irrigation water supplied to this land use. The groundwater supply in “irrigated mixed cotton,wheat rotation/orchards” is 43% of the total irrigation water supply. Groundwater contribution to the total irrigation supplies in the “irrigated mixed cotton,wheat rotation/sugarcane” and “irrigated rice,fodder rotation” land uses is 23%. The results show that the total irrigation supplies in the irrigated areas of the Indus Basin is 181 km3, and this extra water should originate from irrigation practices. An amount of 68 km3 originates from groundwater, while the surface water contribution is 113 km3. This diagnosis suggests that groundwater supplies 68/181 or 38% of the total water applied at the farm gate. The result is in agreement with the 40 to 50% groundwater contribution reported by Sarwar and Eggers (2006).
130
Table 7.2 Amount of water supplied through various sources in irrigated land uses of the Indus Basin during 2007. Irrigated land uses Area ET R Total Surface Ground Irrigation water water supply supply supply mm mm mm mm mm (mha) (km3) (km3) (km3) (km3) (km3) Irrigated mixed 3.99 1130 474 783 448 334 cotton, wheat (45.1) (18.9) (31.3) (17.9) (13.4) rotation/orchards Irrigated mixed 5.78 805 523 489 376 111 cotton, wheat (46.5) (30.2) (28.2) (21.7) (6.4) rotation/sugarcane Irrigated rice, 9.97 1102 462 830 396 441 wheat rotation (110) (46.1) (82.8) (39.5) (43.9) Irrigated mixed 2.77 785 316 601 538 64 rice, wheat (21.7) (8.7) (16.6) (14.9) (1.8) rotation/cotton Irrigated wheat, 2.27 748 493 409 427 -18 fodder rotation (16.9) (11.2) (9.3) (9.7) (-0.4) Irrigated rice, fodder rotation
1.25
1073 (13.4)
178 (2.2)
963 (12.1)
745 (9.3)
218 (2.7)
Total
26.02
974 (254)
451 (117.4)
696 (181)
434 (113)
262 (68)
The gross groundwater abstraction can be explored further to quantify the aquifer depletion (Figure 7-7(e)). The total depletion of 31 km3 (121 mm yr-1) in the aquifer has been computed from IRRgw and the return flow Qgw. The return flow, e.g. base flow from the groundwater to the surface water system, of non-consumed water that is fed back into the river network of 20 km3 yr-1 is included in the analysis. The net groundwater abstracted (gross abstraction 68 km3 yr-1 minus recharge from leaking fields and canals 57 km3 yr-1) became 11 km3 or 42 mmyr-1. The average net groundwater use (NGU) in Rechna doab is 101 mm which is within 20% of the NGU estimated by Ahmad et al (2005). They estimated NGU of 82 mm in the Rechna doab using geo information techniques for the year 1993-94. The largest net groundwater depletion (DEPgw) occurs in Punjab province of India (200 to 800 mm yr-1). Jeevandas et al. (2008) has estimated a net deficit of 260 mm between crop consumptive use and surface supplies in Indian Punjab. The Haryana state of India is also vulnerable to serious groundwater depletion developments (400 to 600 mm yr-1). A recent assessment of groundwater abstractions by NASA showed that the three states of India (i.e. Punjab, Haryana and Rajasthan) lost about 109 km3 of water during 2002 to 2008 leading to decline in water table of 330 mm per year (Rodell et al., 2009). The groundwater overdraft at this alarming rate could potentially change the transboundary groundwater flow between India and Pakistan as also documented in IUCN (2010).
131
Another independent validation of the present results was performed against the GRACE terrestrial water storage change estimates for the year 2007. The GRACE provides change in terrestrial water storage observed at monthly time scale. Figure 7-8 shows the change for the whole year. The change in storage in the irrigated areas of the basin is from 0 to 1000 mm yr-1(Figure 7-8). The spatial patterns of high groundwater abstraction coincide with a broad region of intensive groundwater extraction and water table decline estimated by GRACE. The differences between our estimates are, in combination with the errors associated with GRACE groundwater change estimates (Schrama et al., 2007; Huang et al., 2008; Duan et al., 2009), acceptable.
Figure 7-8 Terrestrial water storage changes observed from GRACE satellite for the irrigated areas in the Indus Basin during the year 2007. 7.3.3
Accuracy assessment
The usability of the IRRgw information for carrying out water management plans depends on the accuracy of the estimates. The influence of uncertainty in evapotranspiration, rainfall, SWAT outputs and canal supplies on IRRgw computations based on Eq (7.9) and (7.10) were tested. Seven pixels representing “irrigated cotton,wheat rotation/sugarcane”, “irrigated cotton,wheat rotation/orchards”, “irrigated rice, wheat rotation” and “irrigated rice,fodder rotation” were randomly selected. The locations of these land uses are provided in the Table 7.3. Cheema et al. (2012) estimated ±4% uncertainty in ETETLook estimates while RTRMM have deviation of ±6% at annual scale (Cheema and Bastiaanssen, 2012). The errors in IRRcw are taken as ±15% (Habib, 2004; Ahmad et al., 2005). The error in SWAT model output parameters (e.g. Qsurf, Qperc, Qlat ) are taken as ±15% (Harmel et al., 2006).
132
One thousand pairs of data series were randomly generated to estimate IRRgw using the uncertainty range of ±4, ±6, ±15 and ±15% for ETETLook, RTRMM, IRRcw and SWAT model outputs, respectively. For all seven locations the absolute deviation was plotted against its probability of occurrence. The maximum absolute error (100% probability), ranges between 122 to 213 mm with an average of 151 mm (Table 7.3). There is 70% probability that the absolute error in IRRgw will be within 62 mm yr-1. The mean error at 50% probability of exceedance is 41 mm yr-1. In the areas with high groundwater abstraction rate (> 400 mm yr-1), this error can be considered within acceptable range (Ahmad et al., 2005). Table 7.3 Annual water balance components and computed gross groundwater abstraction at seven selected locations. The absolute deviation of IRRgw at 100%, 70% and 50% exceedance probability is also provided. Pixel Location
32°25' 48.12"N 73°24' 39.23"E 31°17' 34.54"N 73°04' 37.52"E 31°07' 05.83"N 75°42' 18.35"E 30°08' 49.02"N 75°28' 58.89"E 32°06' 42.44"N 73°49' 32.15"E 28°07' 17.29"N 67°59' 16.69"E 28°27' 13.85"N 68°28' 35.80"E Average
Land use
ET
R
ETLook
TRMM
Qsurf
Qperc
Qlat
IRRcw
IRRgw
(mm) AGI1
Absolute deviation of IRRgw(mm) at probability of exceedance 100 70 50 (%) (%) (%)
(mm)
(mm)
(mm)
(mm)
(mm)
(mm)
1311
767
131
70
0
722
23
213
84
54
AGI2
1174
409
50
48
0
365
498
122
47
30
AGI3
1202
580
237
169
0
230
798
142
51
33
AGI3
1265
429
59
53
3
359
592
124
50
33
AGI3
1098
579
163
105
0
348
439
148
52
35
AGI6
1316
148
24
37
0
534
695
131
62
42
AGI6
1361
183
16
23
0
800
417
175
86
60
1246
442
97
72
0.4
480
494
151
62
41
133
7.3.4
Water balance
The combination of pixel information on ET and rainfall, GIS information on canal water supplies and losses and SWAT model outputs make it practically possible to analyze the entire water balance of unsaturated and saturated zones in the irrigated areas of the Indus Basin separately. The annual water balance components in unsaturated and saturated zones are summarized in Figure 7-9 (a) and (b), respectively.
Figure 7-9 Annual water balance in (a) unsaturated and (b) saturated zones over the entire irrigated areas of the Indus Basin for the year 2007. The total area is 26.02 mha It is evident from the Figure 7-9(a), that the total inflows to the unsaturated zone is estimated at 1147 mm. This inflow includes rainfall (451 mm), irrigation from surface (434 mm) and groundwater (262 mm). The major outflow component from the unsaturated zone is evapotranspiration (974 mm). A considerable amount (97 mm) leaves the system as surface runoff and 75 mm percolates down to the shallow aquifer. The water balance is closed in case of unsaturated zone (zero storage change). Figure 7-9(b) shows the inflow and outflow components of the saturated zone. The inflows to saturated zone include seepage and leakage from fields and canals that are equivalent to 219mm. An amount of 262mm gross amount of groundwater is abstracted from the aquifer, while 79mm leaves the system as return flow. This outflow causes a depletion of 121mm in the saturated zone. For an area of 26.02mha, this is a net depletion of 31km3yr-1. The water balance in the irrigated areas of the Indus Basin is further explored by separating the basin into Pakistan and Indian parts. The detailed water balance components
134
are provided in Table 7.4. The part of Indus in Pakistan and India are written as Indus Basin Pakistan (IB-PK) and Indus Basin India (IB-IN), respectively. Table 7.4 Water balance components for the irrigated areas in Pakistani and Indian parts of the Indus Basin for year 2007. The values in bracket represent water volume in km3. Region
IB-PK IB-IN Indus Basin
ET
R
ETLook
TRMM
(mm) (km3) 978 (154) 963 (99) 974 (254)
(mm) (km3) 381 (60) 558 (57) 451 (117)
Qsurf
Qperc
IRRRS
IRRcw
IRRgw
LOSScw
Qgw
DEPgw
(mm) (km3) 85 (13) 117 (12) 97 (25)
(mm) (km3) 56 (8) 103 (11) 75 (19)
(mm) (km3) 740 (117) 629 (64) 696 (181)
(mm) (km3) 485 (77) 357 (36) 434 (113)
(mm) (km3) 255 (40) 271 (28) 262 (68)
(mm) (km3) 195 (31) 67 (7) 144 (38)
(mm) (km3) 60 (9) 107 (11) 79 (20)
(mm) (km3) 64 (10) 208 (21) 121 (31)
Table 7.4 shows that intense groundwater irrigation is carried out in the Indian part of the Indus Basin to meet consumptive use of crops especially paddy. The groundwater proportion to the total irrigation is 44% in IB-IN while in IB-PK; it is 34% that is in agreement with Foster and Chilton (2003). Foster and Chilton (2003) estimated groundwater proportion of 34% in Pakistan based on information extracted from FAO Aquastat database system. The total consumptive use of water in IB-PK is 117km3, 77km3is supplied through surface water and remaining 40 km3 is from groundwater. These values are in close agreement with the estimates of Siebert et al. (2010). They made a global inventory on groundwater use for irrigation and computed surface and groundwater contribution of 78 km3 and 39 km3, respectively in Pakistan. Annual gross groundwater abstraction in IB-PK is given by Qureshi et al. (2010a) as 51 km3. Groundwater abstraction in 2007 estimated by Government of Pakistan (GOP) for entire Pakistan is 60 km3(GOP, 2010). All these estimates are based on the estimated tubewell density and their approximated working hours. Taking into account the fact that the tubewell withdrawal estimates have higher deviations as reported by Ahmad et al, (2005), the groundwater abstraction estimated at 1 km pixel resolution in this study can be considered reasonable. The groundwater depletion in IB-IN is estimated at 21 km3 as shown in Table 7.4. These results are also consistent with the findings of Rodell et al. (2009). They estimated average groundwater depletion of 17.7±4.5 km3 yr-1 over the Indian states of Punjab, Haryana and Rajasthan. They used terrestrial water storage change observations from GRACE satellite combined with the hydrological models for the years 2002 to 2008.
7.4
Conclusions
The Indus Basin is facing water shortage in time and space domains due to intensive development in irrigation, domestic and industrial water needs. Irrigation is the largest consumer that is using both surface (113 km3 or 434 mm) and groundwater (68 km3or 262 mm) to meet the crop water requirements. Uncontrolled groundwater abstraction, consistently exceeding recharge, is threatening the groundwater reserves in the basin. Groundwater is the hydrological component with the largest uncertainty. By using remote sensing information in combination with GIS data on canal flows and SWAT model
135
outputs, a spatial estimate of the groundwater abstractions and depletions over the entire irrigated area of the Indus Basin is obtained at a resolution of 1 km. Certain areas in Pakistani and Indian provinces of Punjab and Haryana have experienced extensive groundwater abstractions to augment surface supplies at unsustainable levels. The gross groundwater abstractions within the irrigated areas accumulate to 68 km3 (262 mm) in the year 2007 and the corresponding groundwater depletion equaled 31 km3 (121 mm) due to the recharge of leaking canals and irrigation fields also considering return flow to rivers. Sustainable groundwater management is under threat as groundwater abstraction exceeds recharge consistently in these areas. The spatial maps of groundwater abstraction and depletion identify the hot spots that need special attention of water management experts. The districts of Jalandhar, Phagwara, Ludhiana, Bathinda in Indian part of the basin and Narowal, Sialkot, Khanewal, Jhang and Larkana in Pakistani part of the basin are most vulnerable. The only solution to safeguard access to water for food and environment is to reduce groundwater abstractions. Net depletion should be virtually neutral averaged over longer time period. This can be achieved by negotiating groundwater abstractions using the maps provided in Figure 7-7 (d) and (e). Monitoring of groundwater abstraction can be implemented using the same methodology and procedures as outlined in this paper. The technological procedures are outlined and validated. Recharge by constructing wells or delay action dams should be facilitated. The role of trans-boundary aquifers should be given equal importance as the attention that goes to surface water exchanges between administrative boundaries. The analysis is based on only one year needs more years to consider. There is also a potential problem of limited validation as no information on spatial groundwater abstraction is available. The use of single conveyance efficiency is cautious. Conveyance efficiency based on per unit length of the canal should be tested in the future studies.
136
8 Summary and conclusions 8.1
Rationale
This PhD research investigates the use of multi-sensor satellite information to map complex hydrological processes and water management practices in data scarce river basins. The internationally shared Indus Basin is taken as an example. The international character of the basin has made it difficult to study water flows and resource management using traditional point based data sets. The Indus Water Commission has virtually no access to any data, and consequently little power to monitor the developments and changes of the water system. The limited data sharing and lack of trust between the riparian states with political conflicts make it even worse. A first pre-requisite for regional scale cooperation on water issues and comparison between riparian states is a standardized description of water flows, not only in streams, but also in aquifers and withdrawals to irrigation systems. It is recognized that reliable data on water resources conditions is insufficient in the Indus Basin. The lack of a dense network of hydro-meteorological observatories is the plausible cause. Pixel based information obtained from satellites can therefore be an attractive alternative solution that is worth investigating, especially because the entire basin can be encompassed and measured in the same standard manner. Transparent procedures to convert raw satellite data into water management information need to be developed. This thesis provides the scientific grounds for developing transparent data collection procedures based on earth observations. A knowledge base using intelligent pixels is developed and validated. The major goal of the research was to develop scientific methodologies to efficiently utilize satellite measurements for quantifying conjunctive water use in data scarce river basins. The first step to achieve this prime objective was to establish a reliable water balance that can be used for further hydrological analysis. The water balance structure includes an interactive link with water use, which makes it possible for users to appraise the overall water management situation. If this is feasible for the Indus Basin, then it can be used also for other basins in the world. Four knowledge gaps were identified which impede a successful basin scale hydrological analysis: (i) lack of fundamental data (e.g. hydrological, water management, meteorological),(ii) lack of knowledge on land use, (iii) missing information on groundwater resources for the entire basin, and (iv) lack of analytical tools to study alternative solutions to combat over-exploitation of water resources and become more climate resilient. The specific contribution of the study and innovative aspects are highlighted in each of the chapters which include, but are not limited to the development of methodologies to determine the spatial variability of land use (chapter 3), rainfall (chapter 4), soil moisture (chapter 5), evapotranspiration (chapter 6), and groundwater abstraction (chapter 7). Groundwater abstraction is one of the biggest problems in sustainable water management, but information on abstractions cannot be obtained from thousands of individual piezometer readings. It is clear that there are major water balance constituents of which
137
temporal patterns can never be accurately simulated by traditional point based hydrometeorological monitoring programs. Spatial data can be used for preparing water accounts, prohibiting new upstream water resources developments, groundwater restoration plans, developing fair irrigation management practices (adequate, reliable, uniform), providing access to water during droughts, estimating impact of retiring glaciers, reducing non-beneficial water use, enhancing recycling of drainage water, introducing green water credits in upstream catchments, etc. Some tangible recommendations are provided at the end of this chapter. This study can be considered as a first in its kind on the water balance of the entire Indus Basin including Pakistan, India, China and Afghanistan. Making available reliable information on the data scarce and politically divided Indus Basin is fundamental for reviewing and implementing international agreements. The study was carried out using freely available public domain satellite data from the World Wide Web. The original source of satellite data is freely accessible and transparency is guaranteed. Such information enables water managers and decision makers of developing countries to efficiently manage water resources in an era of population growth, diminishing per capita water resources, and threats of climate change. This thesis describes the research results of preparing digital data layers on land use, rainfall, soil moisture, evapotranspiration and groundwater abstractions, without undertaking sophisticated field measurement campaigns. The only field data used originates from weather stations and measurements of water released from large reservoirs into the main irrigation canals.
8.2
Pixel land use
A land use database is essential to provide information on the type of water consumers and the returns in terms of food production, wood production, hydropower, environmental services, economic benefits, etc. For judicious allocation of water, the crops grown in the area have to be identified. In chapter 3, an innovative way of discerning land uses is presented. Spatially consistent 1 km × 1 km pixel information of various land uses was generated using NDVI data freely available from vegetation sensors on board the SPOT satellite for the year 2007.The temporal profiles of NDVI were used to describe the phenological cycle of all agro-ecosystems. There were considerable differences in NDVI patterns between the land uses. It is the first land use map with 27 classes that covers the entire Pakistan. The irrigated land uses showed two peaks in a year while evergreen forests have consistently higher NDVI throughout the year. The rainfed land uses have peaks in phase with rainfall and soil moisture conditions. The phenological cycle thus obtained, supported by expert knowledge and ground information, was used to identify specific crops and crop rotations. Overall accuracy of 77% (Kappa = 0.73) was attained, which is deemed sufficient for supporting water management analyses of large heterogeneous landscapes.
138
8.3
Pixel rainfall
Any water resources analysis starts with quantification of rainfall. Aerial rainfall for catchments and basins is normally interpolated from rain gauge networks. A density of less than four gauges per 10,000 km2, is however insufficient to capture the spatial heterogeneity of rainfall processes. Any small error of a large water balance component such as rainfall will produce significant errors in the smaller components such as runoff. The spatial interpolation of point measurements in heterogeneous landscapes and mountains result in erroneous estimates. Dense networks are needed that are difficult to establish and maintain in developing countries. Rainfall estimation by means of satellites is therefore essential. Satellite based sensors provide an integrated measurement of rainfall in time and space. They are an excellent alternative for rainfall measurement. Sensors on board the TRMM satellite provide three-hourly global rainfall estimates that can be freely downloaded. In chapter 4, details of these sensors are provided, as well as results from accuracy tests of the TRMM in the Indus Basin. It was concluded that these satellite derived rainfall products still show relatively large uncertainties and are difficult to validate. The influence of geography on satellite estimates was apparent with an exponential increase in deviations between TRMM and rain gauge measurements with increasing elevations. It demonstrates that standard statistical relationships with topography cannot be applied. Two methodologies were therefore used to calibrate the monthly estimates of the TRMM rainfall (product 3B43) with low density rain gauge measurements. One method was regression analysis and the other geographical differential analysis. Both techniques improved the results with Nash Sutcliffe efficiency by 81% and 86%, respectively. However, the standard error of estimate of the geographical differential analysis was 41 mm lower than the regression analysis. The advantage of the geographical differential analysis is that the differences in rainfall are used for spatial interpolation and not the absolute values of rainfall. The values for the differences reflect certain underlying terrain features, and the corrections at specific locations were thus bigger than in other locations. The deviations in measured and estimated rainfall can be reduced considerably to 6% This is equivalent to an amount of water of 26 km3 yr-1, which is 17 % of the withdrawals to the irrigation sector, being approximately 150 km3yr-1.Any small error in rainfall measurement thus affects the calculation of water availability for diversions.
8.4
Pixel surface soil moisture
Soil moisture is considered as another state variable that is informative of the water balance. Under arid conditions soil moisture patterns are a direct indication of the presence of water. Wet top soils reflect irrigation water distribution or shallow water table areas. It is practically impossible to attain soil moisture information at large spatial scales with in situ sensors, and for this reason it is not common to include soil moisture in management decision making. Passive microwave sensors on board satellites such as AMSR-E, SMOS and Feng Yung provide global scale estimates of daily surface soil moisture for free. These sensors provide continuous soil moisture estimates without being affected by
139
weather conditions. That is also the main reason for applying a new evapotranspiration (ET) algorithm based on soil moisture measurements. Surface soil moisture is estimated by applying inversion techniques to the brightness temperature measured by satellites. The technique is not error free; therefore it was necessary to validate the satellite soil moisture prior to its use in estimating other hydrological processes. Due to the non-availability of in situ soil moisture measurements in vast river basins, classical validation techniques are not technically feasible. Therefore, alternative validation approaches were needed to build confidence in using satellite soil moisture products. The response of vegetation to soil moisture and soil moisture to rainfall was studied to explain the soil moisture behavior (chapter 6). Strong relationships between TRMM rainfall and AMSR-E surface soil moisture in the land use classes “rainfed”, “very sparse vegetation”, and “bare soil” were observed. The rainfall events in these land uses have a high association with the soil moisture measured by AMSR-E. For irrigated land, this association was lower due to extra supplies from irrigation – and thus perturbations of the rainfall – soil moisture relationship. At annual time scale a stronger correlation exists between the AMSR-E mean soil moisture and TRMM accumulated rainfall (Spearman’s rank correlation coefficient rs=0.74) than between TRMM accumulated rainfall and NDVI (rs=0.70), which is explicable on the basis of soil physical processes. Mean soil moisture and NDVI have stronger correlation (rs=0.85) compared to TRMM rainfall and NDVI (rs=0.70) which is also according to expectations. A time lag between soil moisture and NDVI time series was observed. Such a lag was expected due to delayed response of vegetation against moisture in the root zone. The lag time varied between zero to 60 days, and was generally longer for the wet kharif season. For the dry rabi season, a Pearson’s r> 0.60 was found for 75% of the cases with zero to 40 days lag. For the wet kharif season, it was found for 81% cases but with a lag of 20 to 60 days. The maximum surface soil moisture value of 0.35 to 0.45 cm3 cm-3for a pixel was similar to the top layer saturated moisture content expected on the basis of soil texture maps and pedo-transfer functions. Higher values occurred in flooded lands and paddy fields. This suggests that absolute AMSR-E values properly describe soil moisture under wet land surface conditions.
8.5
Pixel evapotranspiration
Evapotranspiration (ET) accounts for the dominant part of the outgoing fluxes of vegetated land in semi-arid climates. Because ET can be managed partially, it provides a vehicle to control the available water resources for agriculture, forests, swamps, wetlands, and other types of land cover. The actual rate of evapotranspiration cannot be inferred from routine weather data. Rainfall can be measured easily with cheap gauges, but not ET from land surfaces (except at sites equipped for scientific energy balance equipment). Even though water institutions and international river basin commissions have the duty to manage the water resources of river basins, they do not use basin scale evapotranspiration information at all. The financial constraints in the developing countries restrict installation of permanent flux observation towers. There is no single flux tower present in the Indus
140
Basin, while one seventh of the world population lives here and a volume of 496 km3of water evaporate every year. Therefore, a new methodology (ETLook) has been tested that provides spatial estimates of evapotranspiration from satellite measurements and a surface energy balance. There are several ET algorithms available in the literature, but the advantage of this particular algorithm is that it provides continuous estimates of ET throughout the year without being affected by weather. ETLook is a two-source model that can infer information on nonbeneficial evaporation and beneficial transpiration separately using microwave derived surface soil moisture. Microwave radiometry is least affected by cloud cover and can thus provide continuous surface soil moisture information even in monsoon periods and for high altitude regions with persistent clouds. This study was a first attempt to use microwave technologies to accurately estimate evapotranspiration over vast areas of the Indus Basin, using public domain datasets. Due to technical problems, the data from AMSR-E is not available since October 4, 2011. However, passive microwave radiometers like SMOS (European Space Agency) of Feng Yung (Chinese Space Agency) also provide surface soil moisture estimates. The estimated evapotranspiration at 1 km pixel resolution correlated well with some historic lysimeter measurements, Bowen ratio measurements, and remote sensing studies (R2of 0.70 to 0.76 at annual time scale; RMSE of 0.29 and 0.45 mm d-1).It was demonstrated that the pixel scale ET fluxes at daily, 8-day, or monthly time scales can be estimated. The total basin wide ET in 2007 was 496 km3yr-1 while rainfall was 443 km3yr1 . This revealed that more water is evaporated from the land surface than what was received through rainfall. Multiple line agencies from the region and the gravity mission from NASA also suggested a net storage change related to net groundwater withdrawals, declining groundwater tables, snowmelt, and retiring glaciers. The only solution to make the environment of the Indus Basin more sustainable is to reduce ET, and in particular the non-beneficial ET. ET information can greatly assist water managers and policy makers with water allocations. It is also important to gain knowledge of net water producing (R>ET) and water consuming areas (ET>R). The stream flow resulting from net water producing areas can be managed by land use change and adjusted cultivation practices. The concept of green water credits is based on certain agreed upstream ecosystem services to generate sufficient stream flow. Net consuming areas–such as irrigation systems - can be controlled by regulating the water diversions and withdrawals. The irrigation sector is often criticized for inefficient use of water resources. By comparing water diversions to consumptive use, estimates of non-consumed water flows can be made. Irrigation water supply in Pakistan is based on equal access to water for all farmers. Without doubt, such aspiration is difficult to achieve anywhere in the world and especially in areas where water resources are limited. Farmer communities often claim that they are not receiving the volume of water they are entitled to. Pixel values of ET will provide a new source to verify these conditions; not only to certify that they receive water, but at the same time also to verify that they are not using water non-beneficially and non-productively.
141
8.6
Pixel groundwater abstraction
Unmetered and large groundwater withdrawals are two of the biggest water resources problems in the Indus Basin. An adequate assessment of the groundwater withdrawals can be made from the spatial tools presented in this thesis. With spatial estimation of rainfall and ET, it is possible to determine the amount of irrigation water supplies that match ET rates. A hydrological model was used for this purpose. Pixel information on topography, land use, soils, rainfall, and evapotranspiration was used to provide input data into the SWAT hydrological model. SWAT is a well-known and accepted distributed hydrological model that describes many relevant hydrological, soil physical and bio-physical processes. It was used because of its free availability, global adoptability, and the ability to compute atmospheric, land surface, soil, groundwater, and stream flow processes. The SWAT model parameterization and calibration was supported by 8-day ET layers from ETLook. A calibration procedure with variable irrigation water supply and soil physical properties was carried out for each hydrological response unit. The calibrated model was then applied to generate output maps of total irrigation water supply at the farm gate, surface runoff, and combined drainage and percolation, which provide fundamental insights in the breakdown of irrigation water flows into consumed and non-consumed outflows of all irrigated fields present in the Indus basin in an uniform manner. The release of surface water from the major reservoirs to the canal command areas was 151 km3yr-1. By integrating irrigation from canal water that arrives at the farm gate (113 km3yr-1) with the total irrigation water supply estimated from the SWAT model (181 km3yr-1); it was possible to isolate the total groundwater abstractions for irrigation (68 km3yr-1). Pakistan supplies 40 km3and India 28 km3 every year. Table 8.1 provides an overview of the bulk water balance of the irrigated areas of the Indus Basin. Table 8.1The water balance of all irrigated areas combined and considered as being one single“ big field” of 26.02 million ha Inflow km3 yr-1 Outflow km3yr-1 Rainfall 117 Surface runoff 25 Canal water supply India 36 Interception evaporation 5 Canal water supply Pakistan 77 ET from rainfall 99 Groundwater supply India 28 ET from irrigation 154 Groundwater supply Pakistan 40 Drainage and percolation 19 from fields Total 298 298 The vector maps of canal water supplies for every canal command area were superimposed on the vector maps of surface runoff and drainage / percolation and the raster maps of rainfall and ET. This yielded a map with 1 km pixels of gross groundwater abstractions for the entire Indus Basin (Figure 8-1). This is the first map with this level of detail. It shows the hotspot areas at a 100 ha resolution. Information on groundwater pumping activities can thus be made public knowledge.
142
Figure 8-1 Schematic diagram showing data sources to infer groundwater abstraction information at 1 km pixel. The groundwater abstraction can be explored further to quantify the groundwater depletion. The groundwater depletion (gross abstraction 68km3yr-1 minus recharge from leaking fields, amounting to19km3 yr-1,and canals amounting to 38km3 yr-1) was 31km3 yr1 . The groundwater return flow of 20km3 yr-1 in the irrigated areas of the Indus Basin was included in the analysis. The actual groundwater depletion is larger due to outflow of groundwater into the river system. The largest groundwater depletion occurs in the Punjab province of India (350 – 800 mm yr-1). The Haryana state of India is also threatened by serious groundwater depletion. The only solution to safeguard access to water for food and environment is to reduce net groundwater usage to virtually nothing over a longer time period. Figure 8-2 provides information on various water balance components for irrigated areas of Indus Basin lying in Pakistan and India separately.
Figure 8-2 Water balance components in the irrigated areas of Pakistani and Indian part of the Indus Basin
143
8.7
New data sources
Classical data (e.g. data from governmental organizations) are generally point measurements and its accessibility is generally poor. Remotely sensed information is an effective alternative source. It has a public domain status, and everybody can have access to raw satellite information. Data from an international fleet of sensors can be found in the Earth Observing System Data and Information System (EOSDIS). Eight Data Active Archive Centers (DAAC’s), representing a wide range of earth science disciplines, are operational under NASA to process, archive, and distribute EOSDIS data. The Earth Resources Observation System (EROS) Data Center of the USGS provides, in addition, access to land processes data from both satellite and aircraft platforms. Most of the satellite data at relative coarse resolution, which serves the purpose at basin scale, is freely available. The fine resolution data is also available at a fee. NASA and ESA have new policies to keep prices of images low, so that satellite information becomes everybody’s business. Some examples of internationally opened satellite databases can be found at http:// www. daac. gsfc. nasa. gov,https:// www. wist. echo. nasa. Gov /~wist /api/imswelcome/orhttp://www.nsidc.org/data/ae_land3.html. The classical sources of data and associated problems with their alternative solutions are summarized in Table 8.2. The use of satellite data as an alternative for getting firsthand knowledge on hydrology, agriculture, environment and geography, was explored in this thesis. The climatic data available through meteorological departments was measured using routine weather stations. Space borne measurements made it possible to get these datasets at higher temporal and spatial scales. Table 8.2 Applicability of pixel information for database generation in vast and data scarce international Indus Basin Database Classical data Associated problems Alternative Solution and acquisition applicability sources Land use - Global - Generalised land NDVI time series at 1 databases* cover classes km pixels with intervals - International - No information on of 8 to 10 days organizations§ specific crop - Governmental rotations organizations - Outdated Cropped - Governmental - No real time Land use map area organizations information - International - Late dissemination organizations - Tabular data Biomass - Global net - No data Advanced algorithms production primary turning raw data into production quantified information maps Crop yield - Governmental - Late dissemination The pixel information on organizations’ - Administrative unit land use and biomass statistics wise information production can provide 144
Rainfall
Snow cover Soil moisture
- Meteorological department - World meteorological organization - Global databases‡ - Global datasets
- Specific Projects - Field experiments Evapo - Field transpiration experiments Solar - Meteorological radiation department
- Absence of spatial data - Point measurements - Sparse raingauge networks ( 0.60 gevonden voor 75% van de gevallen met een tijdsverschil van nul tot 40 dagen. Tijdens het natte kharif seizoen was dit 81% voor de gevallen met een verschil van 20 tot 60 dagen. De maximale waarde voor het bovenste bodemvocht van 0.35 tot 0.45 cm3 cm-3 voor een pixel was gelijk aan de verwachte toplaag verzadigingsbodemvochtwaarde op basis van bodemtextuurkaarten en pedo-transfer functies. Hogere waarden kwamen in de overstroomde gebieden en in de rijstvelden voor. Dit suggereert dat absolute AMSR-E waarden bodemvocht juist beschrijven tijdens natte bodemcondities.
154
9.5
Pixel verdamping
Verdamping is de meest dominante uitgaande term van de waterbalans van begroeid land in semi-aride klimaten. Omdat verdamping gedeeltelijk bestuurd kan worden, heeft het de mogelijkheid om de hoeveelheid beschikbaar water voor landbouw, bossen, moerassen, en andere landtypes te controleren. De werkelijke verdamping kan niet worden bepaald uit standaard meteorologische reeksen. Neerslag kan direct worden gemeten met goedkope regenmeters, maar verdamping niet (behalve op specifieke locaties die zijn uitgerust met geavanceerde instrumenten). Ondanks dat waterinstellingen en internationale stroomgebiedscommissies de verplichting hebben om de waterhuishouding te regelen, gebruiken zij geen verdampingsinformatie op stroomgebiedsschaal. De financiële beperkingen in ontwikkelingslanden staan de installatie van permanente verdampingsflux observatietorens in de weg. In de Indus Basin is geen enkele fluxtoren aanwezig, terwijl een zevende van de wereldbevolking hier leeft en een volume van 496 km3 water elk jaar verdampt. Daarom is een nieuwe methode (ETLook) getest, die ruimtelijke schattingen van verdamping geeft op basis van satellietmetingen en de energiebalans van het landoppervlak. In de literatuur zijn er verschillende ET-algoritmes beschikbaar, maar het specifieke voordeel van dit algoritme is dat het continue schattingen van ET gedurende het gehele jaar geeft, zonder dat het wordt beïnvloed door het weer. ETLook maakt gebruik van twee bronnen, waaruit informatie over onnuttige evaporatie (E) en nuttige transpiratie (T) kan worden verkregen door gebruik te maken van bodemvocht op basis van microgolfmetingen. Microgolf radiometers wordt het minst beïnvloed door bewolking en kan daarom dus continue bodemvocht informatie geven, zelfs tijdens de moesonperiode en voor hooggelegen gebieden met aanhoudend bewolking. Dit onderzoek was een eerste poging om microgolftechnieken te gebruiken om de verdamping nauwkeurig te schatten over grote gebieden van de Indus Basin door gebruik te maken van openbare datasets. Wegens technische problemen, is de AMSR-E data niet beschikbaar sinds 4 oktober 2011. Echter passieve microgolf radiometers, zoals SMOS (Europese ruimtevaart organisatie) of Feng Yung (Chinese ruimtevaart organisateie) leveren ook vergelijkbare bodemvochtschattingen. De geschatte verdamping op 1 km pixel resolutie correleerde goed met enkele historische lysimeter metingen, Bowen ratio metingen en eerder uitgevoerd remote sensing onderzoeken (R2 van 0.70 tot 0.76 op jaarschaal; RMSE van 0.29 en 0.45 mm d-1). De vergelijking laat zien dat de verdampingflux op pixelschaal per dag, 8-daags, of maandelijkse tijdsschaal geschat kan worden. De totale stroomgebiedsverdamping in 2007 was 496 km3jr-1 terwijl de regen 443 km3jr-1 was. Dit laat zien dat de er meer water verdampt dan dat er ontvangen wordt door regenval. Meerdere regionale overheidsinstanties en de zwaartekrachtmetingen van de Grace satelliet suggereren ook een netto bergingsverschil dat gerelateerd is aan netto grondwateronttrekkingen, dalende grondwaterniveaus, sneeuwsmelt en terugtrekkende gletsjers. De enige oplossing om het Indus stroomgebied duurzamer te maken is het verminderen van verdamping en voornamelijk de onnuttige verdamping (E). Vlakdekkende informatie over verdamping kunnen watermanagers en besluitvormers enorm helpen bij de allocatie van het water. Het is ook belangrijk om kennis over de netto 155
waterproductie (R>ET) en waterconsumptie (ET>R) gebieden te vergaren. De afvoer van de netto water producerende gebieden kan worden beheerd door veranderingen in het landgebruik en aangepaste landbouwmethodes in de hand te houden. Het concept van ‘green water credits’ is gebaseerd op bepaalde overeengekomen bovenstroomse ecosysteem diensten te hanteren waar er voldoende rivierafvoer wordt gegenereerd. Netto water consumptiegebieden, zoals irrigatiesystemen, kunnen worden gecontroleerd door het regelen van de waterverdeling en onttrekkingen. De irrigatiesector wordt vaak bekritiseerd over inefficiënt watergebruik. Door het vergelijken van de wateronttrekking met de waterconsumptie, kunnen schattingen van de niet-geconsumeerde waterstromen gemaakt worden. De toevoer van irrigatiewater in Pakistan is gefundeerd op gelijkwaardige toegang tot water voor alle boeren. Zonder twijfel is zo’n aspiratie overal in de wereld moeilijk te bereiken en met name in gebieden waar water beperkt is. Boerengemeenschappen beweren vaak dat zij niet het volume water krijgen waar zij recht op hebben. Verdampingswaardes op pixelniveau is een nieuwe informatiebron, die gebruikt kan worden om dit te controleren. Niet alleen om te bevestigen dat zij water ontvangen, maar tegelijkertijd ook om te verifiëren dat zij water niet-nuttig en niet-productief gebruiken.
9.6
Pixel grondwateronttrekking
Ongecontroleerde en forse grondwateronttrekkingen is een groot waterproblemen in de Indus Basin. Een passende schatting van de grondwateronttrekkingen kan worden gemaakt met de ruimtelijke tools die in dit proefschrift worden gepresenteerd. Met ruimtelijk schattingen van neerslag en verdamping, blijkt het mogelijk te zijn om de hoeveelheid irrigatiewater te bepalen dat nodig is om bepaalde verdampingswaarden te bereiken. Een hydrologisch model is gebruikt om dit doel te bereiken. Pixel informatie over topografie, landgebruik, bodems, neerslag en verdamping is gebruikt als input data voor het hydrologische SWAT model. SWAT is een bekend en algemeen aanvaard gedistribueerd hydrologisch model dat vele relevante hydrologische, bodemfysische en bio-fysische processen beschrijft. Dit model is gekozen omdat het vrij beschikbaar is, mondiaal toepasbaar is en het de mogelijkheid heeft om atmosferische, landoppervlakte, bodem, grondwater en stromingsprocessen te berekenen. De parametrisatie en kalibratie van het SWAT model werd ondersteund door kaartlagen van 8-daagse verdamping verkregen uit het ETLook model. Voor elke hydrologische response eenheid werd een kalibratie procedure uitgevoerd met variabele irrigatiewater toevoer en variabele bodemfysische eigenschappen. Het gekalibreerde model werd vervolgens toegepast voor het genereren van kaarten van de irrigatiewatergift per perceel, de afvoer en de gecombineerde drainage en percolatie. Dit verschaft essentiële inzichten in de verdeling van het irrigatiewater naar geconsumeerd en niet-geconsumeerde watergebruik van alle geïrrigeerde velden in de Indus Basin op een uniforme manier. De uitstroming vanuit de belangrijkste reservoirs naar de hoofdirrigatiekanalen was 151 km3jr-1. Door het koppelen van de hoeveelheid kanaalwater dat aankomt bij een irrigatieveld (113 km3 jr-1) met de totale irrigatiegift geschat op basis van het SWAT model (181 km3 jr-1), is het mogelijk om de totale grondwateronttrekking te kwantificeren (68 km3 jr-1). De verdeling over de landen is dat Pakistan elk jaar 40 km3 levert en India 28 km3. Tabel 9.1 geeft een overzicht van de bulk waterbalans van de geïrrigeerde gebieden van de Indus Basin. 156
Tabel 9.1: De waterbalans van alle geïrrigeerde gebieden gecombineerd en beschouwd als een enkel ‘groot veld’ met een omvang van 26.02 miljoen/ ha Instroming km3 jr-1 Uitstroming km3jr-1 Neerslag 117 Afvoer 25 Watertoevoer kanalen India 36 Interceptie verdamping 5 Watertoevoer kanalen 77 Verdamping uit neerslag 99 Pakistan Grondwaterlevering India 28 Verdamping uit irrigatie 154 Grondwaterlevering 40 Drainage en percolatie 19 Pakistan van velden Totaal 298 298 De vectorkaarten van het kanaalwater voor elk irrigatieservicegebied zijn gesupponeerd op de vectorkaarten van de afvoer en drainage/percolatie en de rasterkaarten van neerslag en verdamping. Dit levert een kaart met 1 km pixels op van grondwateronttrekkingen voor de gehele Indus Basin (Figuur 9-1). Dit is de eerste kaart met zo’n gedetailleerd niveau. Het laat de hotspot gebieden zien met een resolutie van 100 ha. Informatie over grondwateronttrekkingen kan dus nu openbare kennis worden.
Figuur 9-1 Schematisch diagram van de databronnen gebruikt voor het verkrijgen van pixelinformatie over grondwateronttrekkingen met een resolutie van 1km. De grondwateronttrekkingen kunnen nog verder worden onderzocht voor het bepalen van de grondwateruitputting. Grondwateruitputting (bruto onttrekking 68 km3 jr-1 minus aanvoer vanuit percolerence velden 19 km3 jr-1,en kanalen 38 km3 jr-1) was 31 km3 jr-1. De terugstroming van grondwater van 20 km3 jr-1 in de geïrrigeerde gebieden van de Indus Basin is in de analyse meegenomen. De werkelijke grondwateruitputting is groter door de uitstroming van grondwater naar het riviersysteem. De grootste grondwateruitputting vindt plaats in de Punjab provincie van India (350-800 mm jr-1). De staat Haryana in India wordt ook bedreigt door grondwateruitputting. De enige oplossing om toegang tot water voor voedsel en de omgeving te kunnen waarborgen is door het reduceren van het netto grondwatergebruik tot vrijwel niets over een langere periode. Figuur 9-2 geeft een beeld van de verschillende waterbalans componenten apart voor de geïrrigeerde gebieden van de Indus Basin gelegen in Pakistan en India.
157
Figuur 9-2 Componenten van de waterbalans van de Pakistaanse en Indiaase geïrrigeerde delen van de Indus Basin.
9.7
Nieuwe gegevensbronnen
Klassieke gegevens (bijvoorbeeld data van overheidsinstanties) zijn over het algemeen puntmetingen en zijn vaak moeilijk toegangbaar. Remote sensing informatie is een goed alternatief. Het is openbaar en iedereen heeft toegang tot de ruwe satelliet informatie. Gegevens van een internationale vloot van sensoren kan gevonden worden in de Earth Observing System Data and Information System (EOSDIS). Acht Data Active Archive Centers (DAAC’s), die een ruime verscheidenheid aan aardwetenschappelijke disciplines vertegenwoordigen, zijn operationeel onder de NASA om te verwerken, archiveren en distribueren van EOSDIS gegevens. De Earth Resources Observation System (EROS) Data Center van de USGS biedt tevens toegang tot gegevens over land fysische processen verkregen vanuit zowel satellieten als vliegtuigen. De meeste satellietgegevens met een relatief grove resolutie, wat voldoende is voor de stroomgebiedsschaal, is vrij beschikbaar. Gegevens met een fijnere resolutie is beschikbaar tegen betaling. De NASA en ESA voeren het beleid om de kosten voor beelden laag te houden, zodat iedereen gebruik kan maken van deze gegevens. Een paar voorbeelden van internationaal open databases kan gevonden worden ot http:// www. daac. gsfc. nasa. gov,https:// www. wist. echo. nasa. Gov /~wist /api/imswelcome/ of http://www.nsidc.org/data/ae_land3.html. In Tabel 9.2 zijn de klassieke bronnen met bijbehorende problemen en alternatieve oplossing van gegevens samengevat. Het gebruik van satellietgegevens als een alternatief om uit de eerste hand kennis over hydrologie, landbouw, milieu en geografie te verkrijgen is onderzocht in dit proefschrift. Klimaatgegevens worden gemeten met standaard
158
meteorologische stations. Vlakdekkende metingen maken het mogelijk deze datasets op een hogere tijd en ruimteschaal te verkrijgen. Tabel 9.2 Toepasbaarheid van pixel informatie voor de generatie van geografische bestanden in het grote en data schaarse stroomgebied van de Indus. Onderwerp Landgebruik
Akkerbouw gebieden
Biomassa productie Gewas opbrengst
Regenval
Sneeuwbedek king Bodemvocht
Verdamping
Zonnestraling
Klassieke bronnen voor de gegevens - Globale databases* - Internationale organisaties§ - Overheidsinstelli ngen - Overheids instellingen - Internationale organisaties
Gerelateerde problemen - Zeer algemen klasses van landgebruik - Geen informative over specifieke gewasrotaties - Verouderd - Geen real time informatie - Late verspreiding van gegevens - Data in tabelvorm - Geen directe metingen voorhanden
Alternatieve oplossing en toepasbaarheid NDVI tijdseries op 1 km pixels met intervallen van 8 tot 10 dagen
Landgebruikskaart
- Globale netto primaire productie kaarten - Statistieken van - Late verspreiding van overheidsinstellin gegevens - Informatie alleen per gen administratieve eenheid - Afwezigheid van vlakdekkende gegevens - Meteorologische - Punt metingen afdeling - Schaarse regenmeters - World netwerken (