in the Deccan Plateau. For the present study, the catchment area immediately upstream of the Ukai. Reservoir was investigated to assess the relative ...
Hydrologicat Sciences—Journal-des Sciences Hydrologiques, 47( 1 ) February 2002
31
Assessing the vulnerability to soil erosion of the Ukai Dam catchments using remote sensing and GIS
SANJAY K. JAIN & M. K. GOEL National Institute of Hydrology. Roorkee 247667 (UP). India
sanjavfc/nih.emet.in Abstract The investigation of basins for planning soil conservation requires a selective approach to identify smaller hydrological units, which would be suitable for more efficient and targeted conservation management programmes. One criterion, generally used to determine the vulnerability of catchments to erosion, is the sediment yield of a basin. In India, sediment yield data are generally not collected for smaller sub-catchments and it becomes difficult to identify the most vulnerable areas for erosion that can be treated on a priority basis. An index-based approach, based on the surface factors mainly responsible for soil erosion, is suggested in this study. These factors include soil type, vegetation, slope and various catchment properties such as drainage density, form factor, etc. The method is illustrated with a case study of subcatchments immediately upstream of the Ukai Reservoir located on the River Tapi in Gujarat State, India. The area is divided into 16 watersheds and different soil, vegetation, topography and morphology-related parameters are estimated separately for each watershed. Satellite data are used to evaluate the soil and vegetation indices, while a G1S system is used to evaluate the topography and morphology-related indices. The integrated effect of all the parameters is evaluated to find different areas vulnerable to soil erosion. Two watersheds were identified as being most susceptible to soil erosion. Based on the integrated index, a priority rating of the watersheds for soil conservation planning is recommended. Key words soil erosion; soil conservation: sediment yield; G1S; remote sensing; NDVI; slope; India
Estimation de la vulnérabilité à l'érosion des sols des bassins du Barrage Ukai à l'aide de la télédétection et d'un SIG Résumé La planification de la conservation des sols nécessite une approche sélective pour identifier des unités hydrologiques plus petites que les bassins versants, dont l'échelle serait plus pertinente pour des programmes plus efficaces et ciblés de conservation. Un critère, généralement utilisé pour déterminer la vulnérabilité d'un bassin versant à l'érosion, est son apport solide. En Inde, les données sur l'apport solide ne sont généralement pas disponibles par sous-bassins si bien qu'il est difficile d'identifier les zones les plus vulnérables à l'érosion, à traiter en priorité. Nous proposons une approche basée sur un indice exprimé en fonction de facteurs caractérisant la surface du sol. Ces facteurs englobent le type de sol, la végétation, la pente et différentes propriétés du bassin comme la densité de drainage, un facteur de forme, etc. La méthode est illustrée avec un ensemble de sous-bassins à l'amont immédiat du Barrage Ukai, sur la Rivière Tapi, dans l'état de Gujarat en Inde. La zone est divisée en 16 sous-bassins, et différents paramètres caractérisant le sol, la végétation, la topographie et la morphologie sont estimés pour chacun d'entre eux. Des données satellitaires sont utilisées pour évaluer les indices liés au sol et à la végétation, tandis qu'un système d'information géographique est utilisé pour évaluer ceux qui sont liés à la topographie et à la morphologie. L'effet intégré de tous les paramètres est estimé afin d'identifier les différentes zones vulnérables à l'érosion. Deux sous-bassins particulièrement concernés ont ainsi été identifiés. Sur la base de l'indice intégré, un ordre de priorité pour la conservation des sols est recommandé. Mots clefs érosion des sols; conservation des sols; apport solide; SIG; télédétection; pente: Inde
Open for diseussion until I August 2002
32
Sanjay K. Jain & M. K. Goel
INTRODUCTION The development of land and water resources on a sustainable basis without deterioration and with a constant increase in productivity is the mainstay of mankind. Soil erosion is a complex dynamic process of land denudation by which productive surface soils are detached, transported and accumulated at a distant place. The detachment of soil particles occurs either by hydrological (fluvial) processes of sheet, rill or gully erosion, or through the action of wind. Soil erosion results in loss of precious soil resources for cultivation and causes siltation of reservoirs and natural streams (Kothyari, 1996; Biswas et al., 1999; Jain & Dolezal, 2000). In India, about 53% of the total land area is prone to erosion (Narayan Dhurva & Rambabu, 1983). The formulation of proper basin management programmes for sustainable development requires an inventory of the quantitative soil loss erosion and the priority classification of basins. A basin with a higher rate of erosion needs to be given higher priority for soil conservation measures to be adopted. The All India Soil and Land Use Survey (AISLUS), established in 1988, has been assigned the task of priority delineation. Initially, the AISLUS conducted soil surveys in the upper parts of catchments using Survey of India (SOI) topographic maps and village cadastral maps. Erosion control treatments were started in the upper parts of the catchments with a view that treatments taken up in the downstream catchments at later stages would not be adversely affected by unprotected upper reaches. During the early and mid 1960s, aerial photographs were used to identify severely eroded areas in different catchments. Subsequently, soil conservation works were initiated in close vicinity of the reservoirs and sediment carrier streams (Bhan, 1997). Sediment yield from a catchment is one of the main criteria for assessing the vulnerability of a watershed to soil erosion. However, this criterion requires continuous monitoring of sediment samples at the catchment outlet. Such data are hardly available in India for small watersheds. Although the sediment yield from large basins can be obtained from such observations, it is not possible to ascertain the vulnerability to soil erosion of small watersheds within a basin. A soil conservation programme is an expensive and cumbersome process, carried out in steps starting from the most vulnerable (highest sediment producing) region. Therefore, there is a need to assign relative priorities to different regions within a catchment. The major factors responsible for soil erosion include rainfall, soil type, and vegetation, topographic and morphological characteristics of the basin. (Kothyari, & Jain, 1997). Where there is a lack of data on rainfall and sediment yield, the relative vulnerability of watersheds can be assessed with respect to time-independent factors (soil type, topography and morphology). With the advancement of remote sensing techniques and data acquisition, it is now possible to generate and revise vegetation resource maps at the scale of even a few metres. The effect of vegetation can also be incorporated in such analysis. A geographic information system (GIS) is a computerbased system designed to store, process and analyse geo-referenced spatial data and their attributes. Using a GIS, the topographic and morphometric analysis may be carried out efficiently and different layers of information can be integrated, geographic information systems (using traditional and remotely sensed data) have already proved to play a very important role in analysing soil erosion and sediment yield, as evident from recent studies in the Indian Peninsula and southeast Asia (Jain & Kothyari, 2000; Baban & Yusof, 2000).
Assessing the vulnerability to soil erosion of the Ukai Dam catchments
33
In the present study, digital analysis of remotely sensed data has been carried out to assess the vegetation and soil related indices. Topographic and morphometric indices have been generated in a GIS using the topographic information (contour and drainage) from the SOI toposheets. Finally, all the indices have been combined to prioritize different watersheds in the area.
METHODOLOGY In assessing the relative vulnerability of different watersheds to soil erosion, the major factors responsible for soil erosion were considered using the Watershed Erosion Response Model (WERM). The determination of the different factors is briefly discussed below.
Rainfall The amount and intensity of rainfall affect the sediment yield from a basin. Rainfall is a random meteorological phenomenon. If a dense network of raingauge stations and long-term rainfall data are available, then the effect of rainfall characteristics on soil erosion may be taken into consideration. However, within small regions, rainfall characteristics do not vary to a large extent and can be assumed to be similar over a larger time span.
Vegetation Land use may have a major influence on erosion. Vegetation reduces the raindrop's capability to detach soil particles and significantly affects the erosion process. The effectiveness of vegetation depends on the height and continuity of canopy, density of ground cover and the root density. Generally, forests are most effective in reducing erosion because of their large canopies (Jain & Dolezal, 2000). The classification and mapping of vegetation are fundamental tools for obtaining knowledge about vegetation cover and its relationship to the environment. A number of methods have been used to identify different phenological stages of vegetation, including the application of the normalized difference vegetation index (NDVI), which is used as an indicator of vegetation condition (Ramamoorthi et al., 1991) and is expressed as: ND VI = (MR - VIS)I{NIR + VIS)
(1)
where, Mi? and VIS are reflectance in near-infrared and visible red bands, respectively. The normalization minimizes the effect of illumination geometry as well as surface topography.
Soil type The physical properties of soil affect its infiltration capacity and the extent to which the soil can be detached, dispersed and transported. The properties which most influence
34
San]ay K. Jain & M. K. Goel
erosion include soil structure and texture, organic matter content, moisture content, density (compactness), shear strength, as well as chemical and biological characteristics (Sharma et al., 1990). To study the effect of soil conditions in the watersheds, the soil brightness index (SBI) for each watershed was estimated. The variability in agriculture scene data can be captured in two dimensions by soil brightness index, as shown by Sharma et al. (1990), who gave coefficients for the calculation of brightness using the data of IRS-1B LISS-II. To compute these coefficients, Sharma et al. (1990) took about 60 soil samples with a wide range of physical properties. A principal components analysis was performed on the combined soil data to determine the overall distribution of the soil spectral data in four-dimensional space. A positive linear relationship was found among all the four bands (B1-B4), with maximum correlation between bands 2 and 3. The formula used for the calculation of soil brightness index (SBI) is as follows: SBI= 0.2623-51 + 0.6432-52 + 0.6302-53 + 0.3471-54
(2)
Topography One topographic feature that mostly influences the erosion process is the degree of slope: a higher slope resulting in higher erosion. With the advent of G1S techniques, it is now possible to prepare the digital elevation model (DEM) of an area. Using the DEM, the slope on all the grids in the area can be calculated and this information can be utilized for assessing the relative vulnerability to soil erosion.
Morphology The relationships between the morphology of streams and sediment yield have been considered important for many decades, especially when changes in morphology might somehow be linked to changes in sediment yield from the landscape. Important morphological characteristics of a watershed include drainage density, form factor, elongation ratio and circulatory ratio. Drainage density is defined as the quotient of the cumulative length of the streams to the total drainage area and is expressed in length per unit area (Choubey & Jain, 1992). A higher drainage density represents a relatively higher number of streams per unit area and thus a rapid storm response. It also represents conditions favourable for higher erosion from the catchment. The form factor is defined as the ratio of the basin area to the square of basin length. The circulatory ratio is defined as the ratio between the area of the basin and the area of the circle having the same perimeter as that of the basin (Choubey & Jain, 1992). A higher circulatory ratio induces lesser erosion. The elongation ratio is the ratio between the diameter of a circle having the same area (as that of the basin) and the maximum length of the basin. A higher elongation ratio induces lesser erosion. The vast size of areas, time constraints and yearly variations demand a fast inventory to be made of the situations. Land-use changes with time need continuous monitoring. In all these circumstances, remote sensing can be looked upon as an aid in land-use mapping. Land-use features can be identified, mapped and studied on the
Assessing the vulnerability to soil erosion of the Ukai Dam catchments
35
basis of their spectral characteristics: healthy green vegetation has considerably different characteristics in visible and near-infrared regions of the spectrum, whereas dry bare soil has a relatively stable reflectance in both regions of the spectrum. Thus, by using multispectral data appropriately, different ground features can be differentiated from each other and a thematic map depicting land use can be prepared. Using a GIS database, topographical and morphological characteristics of the watersheds can be estimated quite easily and all such information can be integrated for assessing the vulnerability of different watersheds to soil erosion. Using this methodology, the relative vulnerability of different watersheds has been assessed in the immediately contributing catchment area of Ukai Reservoir.
CASE STUDY OF UKAI CATCHMENT The study area The Ukai is the largest multipurpose project so far completed in Gujarat State, India. The Ukai Dam is located across the River Tapi near Ukai village in Surat District, Gujarat State. It is located between longitudes 73°32'25"-78°36'30"E and latitudes 20°5'0"~22°52'30"N. The total catchment area of the Ukai Reservoir (62 225 km2) lies in the Deccan Plateau. For the present study, the catchment area immediately upstream of the Ukai Reservoir was investigated to assess the relative vulnerability to soil erosion. The study area is shown in Fig. 1. The AISLUS is engaged in conducting rapid reconnaissance surveys for prioritization of smaller hydrological units within the catchment areas of river valley projects and flood prone rivers. It has developed a system for delineating and codifying the catchment areas into smaller hydrological units. The
£>
JF
€)
(5)
(i)
Fig. 1 The location of the study area.
w
1. Indus drainage 2. Ganges drainage 3. Brahmaputra drainage 4. All drainageflowinginto Bay of Bengal except those of 2 and 3 5. All drainage flowing into Arabian Sea except that at 1 6. WesternRajasthan mostly ephemeral drainage
36
Sanjay K. Jain & M. K. Goel
V
""Y
Scale 1:650.000
Fig. 2 The sub-basin map of the study area,
entire country has been divided into six major water resources regions. There are 35 basins, 112 catchments, 500 sub-catchments and 3237 watersheds, following a fivestage delineation approach. The study area considered herein falls under region 5 in the Tapti basin. The area has 86 watersheds and covers 65.95 million ha (A1SLUS, 1988). The total area is covered in two catchments, viz. left bank Ukai Dam to Puma confluence and right bank Ukai Dam to Purna confluence, with five and seven watersheds respectively in these two catchments. In the original codification scheme, the study area is covered in 12 watersheds, but, to handle the data, four watersheds were further subdivided into 5C3A1, 5C3A3, 5C2A1 and 5C2A6 making the total number of watersheds 16 (Fig. 2). The total study area up to the dam site is 5225 km",
Data used for the study Various watersheds in the area of interest were marked using the 1:1 000 000-scale Watershed Atlas of India maps (AISLUS, 1988). For preparation of vegetation and soil indexes, the remote sensing data of the LISS-II sensor of the 1RS-IB satellite, at spatial resolution of 36 m, were used. The data of 17 December 1993 were used in the present study, since the vegetation is more developed after the monsoon season and appears clearly in the remote sensing image. The remote sensing data of all the watersheds studied are covered in path/row 30-53 in quadrants Al and Bl of the satellite. For the preparation of drainage and contour maps at higher scales, Survey of India toposheets at scales of 1:250 000 and 1:50 000 were used. The area was covered by 19 toposheets. Interpretation and analysis The Watershed Erosion Response Model (WERM) used in the study considers four parameters, namely, morphological and topographical features, vegetation index and
Assessing the vulnerability to soil erosion of the Ukai Dam catchments
37
soil brightness index (CWC, 1999). The image processing was carried out on the ERDAS IMAGINE 8.3.1 system. The remote sensing image was geometrically rectified with respect to the Survey of India (SOI) 1:250 000 topographical maps. The geo-referenced images of different quadrants (Al and Bl) were stitched together using the MOSAIC function. A base map of the area of interest was prepared using the 1:250 000 toposheets. This base map contained the overall catchment boundary and the boundaries of various watersheds. This base map was used to extract the remote sensing data of the area of interest from the full scene of the satellite. Different watersheds were also separated from the full catchment area. The generation and calculation of all these parameters is explained below. Preparation of vegetation maps Vegetation is important in assessing the relative vulnerability of different watersheds to soil erosion. The presence of vegetation reduces both the detachment of sediments and their transportation. Remote sensing is a fundamental tool for the classification and mapping of vegetation. The normalized difference vegetation index (NDVI) as referenced to IRS-1B LISS-II data, is expressed as: ND VI = (54 - B3)I{B4 + B3)
(3)
where B4 and B3 are the reflectance in band 4 (0.77-0.86 \im) and band 3 (0.620.68 jam), respectively. From the geo-referenced raw image, the NDVI image was derived using the above equation. Using the base map, the NDVI images were separated for each basin, and then organized into five different classes using the technique of unsupervised classification. These five classes represented the extent of NDVI at each grid in the watershed. Since each basin was to be considered as a single identity, the area-weighted-vegetation {AWV) was calculated for each watershed as follows: „„„, A\-wV\ + A2-wV2 + A3-wV3 + A4-wV4 + A5-wV5 AWV = (4) AÏ + A2 + A3 + A4 + A5 where AI...A5 are the areas under each vegetation class, and wVl...wV5 are the weights for each vegetation class. The range of AWV was classified into five classes with weights varying from 1 to 5. High AWV h given the lowest weight, based on the reasoning that the watershed with a higher vegetation amount must suffer lesser erosion. Thus, a watershed with higher vegetation content is given low vulnerability. Preparation of soil brightness index maps Similar to the case of NDVI, the SBI images obtained for each watershed were divided into five different classes using unsupervised classification. These five classes represented the extent of SBI. In a similar method to that used for vegetation, area weighted soil (A WSo) values were calculated for all the watersheds. Then the range of A WSo values was divided
38
Sanjay K. Jain & M. K. Goel
into five classes with weights varying from 1 to 5. A high AWSo was given higher weight. This signifies that the higher the soil content (barren land), the higher is the vulnerability to erosion.
Estimation of morphological/topographical parameters For the estimation of morphological and topographical parameters, Survey of India toposheets were used. The drainage network was derived from the 1:50 000 Survey of India toposheets: the area is covered in part or in full by 19 topographical maps. All the maps were digitized and different thematic layers were generated for the contours and the drainage pattern. For drainage networking and creation of DEMs, the GIS software ILWIS 2.2 (Integrated Land and Water Information System) was used. The Strahler system (Strahler, 1964) was used for stream ordering. All the watersheds were found to be in the range of sixth or seventh order. After networking, the lengths of streams of each order were evaluated separately for each watershed. Using this GIS database, the physical characteristics of the watersheds, such as the drainage density, form factor, circulatory ratio and elongation ratio were estimated. Based on the length of different order streams, drainage density was calculated for each watershed separately and divided into five classes (weights varying from 1 to 5). The higher the drainage density, the higher will be the vulnerability to erosion and, hence, the greater the weight. Form factor, circulatory ratio and elongation ratio were calculated similarly for each watershed and divided into five classes. Higher values of the form factor, circulatory ratio and elongation ratio induce lesser erosion and higher values in these cases was assigned less weight. The different weights obtained for each morphological parameter were averaged out and again divided into five different classes. Thus, a single weight was assigned for all the morphological parameters taken together. Weights were assigned to each different range of average morphological weight, assuming that higher morphological weight induces higher erosion.
Slope layer In addition to the basin characteristics, slope is another prominent factor for soil erosion. The higher the slope, the greater will be the erosion. In the present case, the ILWIS system was used for evaluation of slope maps. Using the contour information for each watershed, a DEM was generated, and used for the estimation of slope factor for all the watersheds. Since a watershed may contain many slope categories, area weighted slope (AWS) was calculated for each watershed in the same way as for soil and vegetation.
Relative vulnerability of watersheds To account for the integrated effect of all the four parameters considered in this study, the individual weights of all the parameters were added together. This sum was further sub-divided into four different categories for the purpose of assessing the relative
39
Assessing the vulnerability to soil erosion of the Ukai Dam catchments
vulnerability, and codes were assigned to each category. Watersheds with higher final weights were considered to be most vulnerable to soil erosion. Thus, a watershed with a code of 1 is highly vulnerable to soil erosion and must be given the highest priority for the purpose of basin treatment and for the adoption of soil conservation measures. Table 1 presents the relative prioritization status of the watersheds, and the range of accumulated weights and the corresponding priority code is given in Table 2. Table 1 Priority code derived for all watersheds, using 17 December 1993 data. Watershed number
Weight: Morphology
Soil
Vegetation
Slope
Sum of weights
Priority code
5C3A1(1) 5C3A1(2) 5C3A2 5C3A3(1) 5C3A3(2) 5C3A4 5C3A5 5C2A1 5C2A2(1) 5C2A2(2) 5C2A3 5C2A4 5C2A5 5C2A6(1) 5C2A6(2) 5C2A7
4 4 4 4 1 2 2 4 2 2 1 5 2 5 3 2
1 4 5 3 4 4 4 5 2 3 4 4 5 2 2 2
3 3 4 4 4 4 5 2 2 5 5 5 1 3 4 3
5 5 2 2 3 4 2 1 4 4 5 3 1 3 4 4
13 16 15 13 12 14 13 12 10 14 15 17 9 13 13 11
2 1 1 2 3 2 2 3 4 2 1 1 4 2 2
->
.3
Tabic 2 Range of accumulated weights and the corresponding priority code. Range of accumulated weights
Vulnerability
Code
>14 13-14 11-12