Worrkshop on Op pen source Geo G spatial to ools 1st - 5thh May 2011
C CiSTUP Techhnical Repoort: 3
Fo oss 4G Worksho W op on Op pen sou urce Geo o spatia al tools Con nveners s T TV Ramachandra
T G Sitharam
M S Mo ohan Kumar
S Narendra a Prasad
C Chairman, OSG Geo India, Softw ware Developm ment committeee
Chaairman CiSTU UP, IISc
SSecretary K KSCST
C Chairman, OSG Geo India, Educaation and Train ning Committe ee
Geoinfformatics lab C Centre for Infrastruccture, Susttainable Transporta T ation and Urban Plaanning (C CiSTUP) Indian I Insstitute of S Science Bangalorre 560 012 , India http://ciistup.iisc.erneet.in http://wgbis.ces.iisc.ernett.in/foss
Foss 4G: Workshop on Open source Geo spatial tools
SI.No
Content
1 2 3 4 5
Workshop Overview Organisation Background Resource Persons Introduction to Gis and Remote sensing Foss4G – FOSS for Geoinformatics- Ramachandra T V and Uttam Kumar Introduction to Linux and Installing Ubuntu - Bharath H Aithal Working with GRASS - Anindita Dasgupta, Uttam Kumar QGIS Overview- Setturu Bharath GRDSS - Geographic Resources Decision Support System – an Open Source GIS - T. V. Ramachandra, Uttam Kumar, S.N. Prasad & Vaishnav B
6 7 8 9 10 11
12 13 14
Engaging Web for better administration Spatio-temporal landscape modelling for natural hazard vulnerability analysis in select watersheds of Central Western Ghats - T. V. Ramachandra, Anindita Dasgupta, Uttam Kumar, Bharath H Aithal, P. G. Diwakar and N. V. Joshi Land surface temperature with land cover dynamics: Multi resolution, spatio-temporal data analysis of Greater BangaloreRamachandra T V and Uttam Kumar Hybrid Bayesian Classifier for Improved Classification Accuracy Uttam Kumar, S. Kumar Raja, Chiranjit Mukhopadhyay, and T V Ramachandra Programme Schedule
Page Number X 1 2 3 13 20 27 87 105
115 125
135
146
155
Worksho W op Oveerview w owth of GIIS is hindeered by hig ghly pricedd proprietaary softwarres and unaavailability y The gro of ppublic geodata. Man ny of such licensed software rremain unuutilized in various ggovernmen nt officces becausse they req quire eitherr extensivee training oor easy avaailability oon multiplee locationss. All this can bee changed by the usee of Free/O Open sourcce softwaree tools. Thhis is signifficant sincee all ggovernmen nt data and d most educcational in nstitutions w work in local languaage media. FOSS GIS S or F Free/open source s Geo ospatial sofftwares aree more pow werful andd usable forr developinng systemss wheere more and a more public parrticipation is soughtt. In this rregard the proposedd workshop p duriing 1 - 5th h May 20 011 helps to understtand the ppotential annd limitatiions of Oppen sourcee geosspatial tools.
Venue: CiSTUP C Lecture L Halll, CiSTUP P, IISc, Banngalore-5660012 1st - 5th May M 2011 (5 days)
(CiSTU UP jointly with KSC CST, OSGEO, IIIT ((Hyderabad), IIRS and SACO ON)
X
1
Background CiSTUP,
Centre for infrastructure, Sustainable Transportation and Urban Planning was established at IISc (http://cistup.iisc.ernet.in) with the financial aid from the Government of Karnataka to develop a unique expertise having an interface in the areas of infrastructure, sustainable transportation and urban planning. CiSTUP is actively engaged in several activities such as basic and applied research and development, academic activities, training programmes, workshops and consultancy projects in the areas of infrastructure, sustainable transportation and urban planning. Geoinformatics lab has been set up at CiSTUP with the state of the art gadgets in spatial and temporal analysis.
KSCST,
the Karnataka State Council for Science and Technology (KSCST) is actively involved in Natural Resources Data Management System (NRDMS) program (http://kscst.org.in/nrdms.html) a multi-disciplinary and multi-institutional program aimed at developing methodologies for building and promoting the use of spatial data management and analysis technologies in local area planning.
OSGEO, the Open Source Geospatial Foundation (http://osgeo.org) created to support and builds the highest-quality open source geospatial software. The goal is to encourage the use and collaborative development of community-led projects. The OSGeo-India is actively involved in the capacity building in Open Source Geospatial software encouraging student and researchers by conducting lectures and workshops in collaboration with the academic institutions (http://wiki.osgeo.org/wiki/India_Chapter_Report_2007). The India OSGeo Chapter is aiding individuals and institutions interested in Open Source Geospatial Solutions and related issues, covering all activities associated with the application, development and promotion of Open Source Geospatial solutions in India. IIT (Hyderabad), the International Institute of Information Technology, Hyderabad (IIIT-H) is an autonomous university founded in 1998. It was set up as a not-for-profit public private partnership (NPPP) and is the first IIIT to be set up (under this model) in India. The Government of Andhra Pradesh lent support to the institute by grant of land and buildings. IIIT-H was set up as a research university focused on the core areas of Information Technology, such as Computer Science, Electronics and Communications, and their applications in other domains.
IIRS,
Indian Institute of Remote Sensing (IIRS) under National Remote Sensing Centre, Department of Space, Govt. of India is a premier training and educational institute set up for developing trained professional in the field of Remote Sensing, Geoinformatics and GPS Technology for Natural Resources, Environmental and Disaster Management. The main area of the function of the Institute is capacity building through technology transfer among user community, education at postgraduate level in the application of Remote Sensing and Geoinformatics for Natural Resources Management and promote research in Remote Sensing and Geoinformatics.
SACON, the Sálim Ali Centre for Ornithology and Natural History
was formally
inaugurated on 5th June 1990 and registered as a society under the Society Registration Act 1860. SACON, an autonomous organization is a national center for studies in Ornithology and Natural History. The center was named befittingly after Dr. Sálim Ali in appreciation of his lifelong services to India's bird life and conservation of natural resources.
2
Resource Persons 1st - 5th May 2011 (5 days) (CiSTUP jointly with KSCST, OSGEO, IIIT (Hyderabad), IIRS and SACON) SI.No Date 1/ 05/2011 1 1/ 05/2011 2
Resource Person T V Ramachandra Bharath H Aithal
3
1/ 05/2011
Uttam Kumar
4
2/ 05/2011
Bharath Setturu
5
2/ 05/2011
Anindita Dasgupta
6
2/ 05/2011
S. Narendra Prasad
7
2/ 05/2011
Sameer Saran
8
3/ 05/2011
Harish Karnatak
9
3/ 05/2011
Santosh
10
3/ 05/2011
Achary
11
4/ 05/2011
Uma Shama
12
4/ 05/2011
P M. Balamanikavelu
13
4/ 05/2011
Arulraj M
14
4/ 05/2011
Chiranjit Mukhopadhyay
Address Geoinformatics Lab, CiSTUP, IISc, Bangalore Energy & Wetlands Research Group, CES, IISc, Bangalore Energy & Wetlands Research Group, CES, IISc, Bangalore Energy & Wetlands Research Group, CES, IISc, Bangalore Energy & Wetlands Research Group, CES, IISc, Bangalore Senior Principal Scientist, Salim Ali Centre for Ornithology & Natural History, Hyderabad Geoinformatics Division, Indian Institute of Remote Sensing, Department of Space, Govt of India, 4 Kalidas Road, Dehradun Scientist, Geo Informatics Division, National Remote Sensing Centre, ISRO, Balanagar, Hyderabad Salim Ali Centre for Ornithology & Natural History, Hyderabad Salim Ali Centre for Ornithology & Natural History, Hyderabad GeoGraphics Laboratory, Bridgewater State College, USA Head, Bhuvan Cell, National Remote Sensing Centre, Hyderabad Lead Developer, Bhuvan, National Remote Sensing Centre, Hyderabad Management Studies, IISc, Bangalore
3
Introduction to GIS & Remote sensing Many of our decisions depend on the details of our immediate surroundings, and require information about specific places on the Earth’s surface. In this regard, recent developments in information technologies have opened a vast potential in communication, analysis of spatial and temporal data. Data representing the real world can be stored and processed so that they can be presented later in a simplified form to suite specific needs. Such information is called geographical because it helps us to distinguish one place from another and to make decisions for one place that are appropriate for that location. Geographical information allows us to apply general principles to the specific conditions of each location, allows us to track what is happening at any place, and helps us to understand how one place differs from another. Spatial information is essential for effective planning and decision-making at regional, national and global levels. The geographical information in the form of maps (based on field surveys), photos taken from aircraft (aerial photography), and images collected from the space borne platforms (satellite) can be represented in digital form, this opens an enormous range of possibilities for communication, analysis, modeling, and accurate decision making, but a degree of approximation. GIS can be defined as computerized information storage processing and retrieval system that has hardware, software specially designed to cope with geographically referenced spatial data. Collective name for such system is geographical information systems, (GISs). Processing geographical information includes:
Techniques to input geographical information, converting the information to digital form
Technique for sorting such information in a compact format on computer disks, and other digital storage media
Methods for automated analysis for geographical data, to search for the patterns, combine different kinds of data, make measurements find optimum sites or routes, and a host of other tasks
Methods to predict the outcome of various scenarios, such as the effects of climate change on vegetation
Techniques for display of data in the form of maps, images and other kinds of display
Capabilities for output of results in the form of numbers and tables.
4
Elements of GIS: Components of geographical data are Spatial and Attribute Database, Cartographic Display System, Map Digitizing System, Database Management System, Geographic Analysis System, Statistical analysis system and Decision support system. The linkages among these components are illustrated in Figure 1.1.
Figure 1.1: Components of GIS.
i)
Spatial and Attribute Database: Central to the system is the database – a collection of maps and associated information in digital form. Since the database is concerned with earth surface features, it is seen to comprise of two elements – a spatial database describing the geology (shape and position) of the earth surface features, and an attribute database describing the characteristics or quantities of these features. Thus, for example, we might have a property parcel defined in the spatial database and qualities such as its land use, owner, property valuation, etc. in the attribute database.
ii)
Cartographic Display System: Surrounding the central database, we have a series of software components. The most basic of these is the cartographic display system. The cartographic display system allows one to take selected elements of the database and produce map output on the screen or some hardcopy device such as printer or plotter.
5
iii)
Map Digitizing System: After cartographic display, the next most essential element is a Map Digitization System. With a map digitizing system, one can take existing paper maps and convert them into digital form, thus further developing the database. In the most common method of digitizing, one attaches the paper map to a digitizing tablet or board and then traces the features of interest with a stylus according to the procedures required for digitizing. Many maps digitizing system also allows for some editing of the digitized data. Scanners can also be used to digitize data such as aerial photographs. The results is a graphic image, rather than the outlines of features that are created with a variety of standard graphics file formats for export. These files are then imported into the GIS. Computer assisted design (CAD) and Coordinate Geometry (COGO) are two examples of software systems that provide the ability to add digitized map information to the database, in addition to providing cartographic display capabilities.
iv)
Database Management System: The next logical component in a GIS is Database Management System (DBMS), which is used to input, manage and analyze attribute information along with then spatial data. GIS thus typically incorporates a variety of utilities to manage the spatial and attribute components of the geographic data. DBMS aids to enter attribute data, such as tabular information and statistics, and subsequently extract specialized tabulations and statistical summaries to provide new tabular reports. The DBMS provides the ability to analyze attribute data. Many map analysis have no true spatial component, and for these a DBMS will often function quite well. For example, we might inquire of the system to find all property parcels where the head of the household is single but with one or more child dependents, and to produce a spatial map. Software that provides cartographic display, map digitizing, and database query capabilities are often referred to as Automated Mapping and Facilities Management (AM/FM) system.
v)
Geographic Analysis System: Up to this point, we have described a very powerful set of capabilities that the GIS offer, the ability to digitize spatial data and attach attribute to the features stored; to analyze these data based on those attribute; and to map to the result. But on inclusion geographic analysis system, we extend the capabilities of the traditional database query to include the ability to analyze data based on their location. Perhaps the simplest example of this is to consider what happens when we are concerned with the joint occurrence of features with different geographies. For example, suppose we want to find all areas of residential land on bedrock types associated with high levels of radon gas. A traditional DBMS cannot solve this problem because bedrock types and landuse divisions simply do not share the same geography. Traditional database query is fine as long as we are taking about attributes belonging to the same features. But when the features are different, it cannot cope. For this we need a GIS. In fact, it is this ability to compare different feature based on their common geographic occurrence that is the hallmark of GIS. This analysis is accomplished by the process of overlay, thus named because it is identical in character to overlaying transparent maps of the two entity groups on top of one another. Like the DBMS, the Geographic Analysis System as highlighted in Figure 1.1 has a two-way interaction with the database; the process is distinctly analytical in character. Thus while it may access data from the database, it may equally contribute the results of that analysis as a new addition to the database. For example we might look
6
for joint occurrence of lands on steep slopes with erodable soil under agriculture and call the results based on existing data and set of specific relations. Thus the analytic capabilities of the Geographic Analysis System and the DBMS play a vital role in extending the database through the addition of knowledge of relationships between features. vi)
Image Processing System: In addition to these essential GIS elements, remotely sensed image and specialized statistical analysis are also important. This we will discuss in the subsequent sections.
vii)
Statistical analysis system: GIS incorporates a series of specialized routines for analyzing the statistical description of spatial data and for inferences drawn from statistical procedures.
viii)
Decision support system (DSS): Decision support constitutes a vital function of a GIS. It helps in the construction of multi-criteria suitability maps, and address allocation decisions when there is multiple objectives involved while accounting for error in the process. Used in conjunction with the other components of the system, DSS provides a powerful tool in decision-making for resource allocation.
Map Data Representation A Geographic Information System stores two types of data that are found on a map—the geographic definitions of earth surface features and the attributes or qualities that those features possess. Most systems use nearly one or a combination of both the fundamental map representation techniques: vector and raster.
Vector: This refers to the spatial data represented in the form of point, line or polygon depending on the feature of interest (and scale). With vector representation, the boundaries or the course of the features are defined by a series of points that, when joined with straight lines, form the graphic representation of that feature. The points themselves are encoded with a pair of numbers giving the X and Y coordinates in systems such as latitude/ longitude, etc. The attributes of features are then stored in the database management system (DBMS). For example, a vector map of property parcels might be tied to an attribute database of information containing the address, owner’s name, property valuation and land use. The link between these two data files can be a simple identifier number that is given to each feature in the map (Figure: 1.2).
Raster:
In this case, the graphic representation of features and the attributes they possess are merged into unified data files. In fact, we typically do not define features at all. Rather, the study area is subdivided into a fine mesh of grid cells in which we record the condition or attribute of the earth’s surface at that point (Figure 1.2). Each cell has a numeric value (often
7
referred as digital number or spectral signature), representing a feature identifier, a qualitative attribute code or a quantitative attribute value. For example, a cell could have the value “6” to indicate that it belongs to District 6 (a feature identifier), or that it is covered by soil type 6 (a qualitative attribute), or that it is 6 meters above sea level (a quantitative value). Although the data we store in these grid cells do not necessarily refer to phenomena that can be seen in the environment, the data grids themselves can be thought of as images or layers, each depicting one type of information over the mapped region. This information can be made visible through the use of a raster display. In a raster display, such as the screen on your computer, there is also a grid of small cells called pixels (or picture elements). The word pixel is a contraction of the term picture element. Pixels vary in their color, shape or gray tone depending on features in the object. To make an image, the cell values in the data grid are used to regulate directly the graphic appearance of their corresponding pixels. Thus in a raster system, the data directly controls the visible form we see. Raster versus Vector: Raster systems are typically data intensive since they must record data at every cell location regardless of whether that cell holds information that is of interest or not. However, the advantage is that geographical space is uniformly defined in a simple and predictable fashion. As a result, raster systems have substantially more analytical power than their vector
Vector Raster Figure1.2
counterparts in the analysis of continuous space and are thus ideally suited to the study of data that are continuously changing over space such as terrain, vegetation biomass, rainfall and the like. The second advantage of raster is that its structure closely matches the architecture of digital computers.
8
As a result, raster systems tend to be very rapid in the evaluation of problems that involve various mathematical combinations of the data in multiple layers. Hence they are excellent for evaluating environmental models such as soil erosion potential and forest management suitability. In addition, since satellite imagery employs a raster structure, most raster systems can easily incorporate these data, and some provide full image processing capabilities. While raster systems are predominantly analysis oriented, vector systems tend to be more database management oriented. Vector systems are quite efficient in their storage of map data because they only store the boundaries of features and not that which is inside those boundaries. Because the graphic representation of features is directly linked to the attribute database, vector systems usually allow one to roam around the graphic display with a mouse and query the attributes associated with a displayed feature, such as the distance between points or along lines, the areas of regions defined on the screen, and so on. In addition, they can produce simple thematic maps of database queries. Compared to their raster counterparts, vector systems do not have as extensive a range of capabilities for analyses over continuous space. They do, however, excel at problems concerning movements over a network and can undertake the most fundamental of GIS operations that will be sketched out below. For many, it is simple database management functions and excellent mapping capabilities that make vector systems attractive. Because of the close affinity between the logic of vector representation and traditional map production, a pen plotter can be driven to produce a map that is in distinguishable from that produce by traditional means. As a result, vector systems are very popular in municipal applications where issues of engineering map production and database management predominate. Geographic database concepts: Regardless of the logic used for spatial representation, raster and vector, we begin to see that a geographic database as a Complete database for a given region and is organized in a fashion similar to a collection of maps. Vector systems come closest to this logic with what are known as coverages. Map like collection that contain the geographic definition of a set of features and their associated attributes tables. However, they differ from maps in two ways. First, each will typically contain information on only a single feature types, such property parcels, soil polygons, and the like. Second, they may contain a whole series of attributes that pertain to those features, such as a set of census information for city blocks. Raster system also uses this map like logic, but usually divides data sets into unitary layers. A layer contains all the data for a single attribute. Thus one might have a soil layer, a road layer and a land-use layer. There are subtle differences, for all intents and purposes, raster layer and vector coverage can be thought of as simply different manifestations of the same concepts as the organization of the database into elementary map-like themes. Layers and coverage differ from traditional paper maps, however, in an important way. When a map is digitized, scale differences are
9
removed. The digital data may be displayed or printed at any scale. More importantly, digital data layers that were derived from maps of different scale, but covering the same geographic area, may be combined. GIS provide utilities for changing the projection and reference system of digital layers. This allows multiple layers, digitized from maps having various projections and reference system, to be converted to a common system. With the ability to manage differences of scale, projection and reference system, layers can be merged with ease, elimination a problem that has traditionally hampered planning activities with maps. It is important to note, however, that the issue of resolution of the information in the data layers remains. Although features digitized from a poster sized world map could be combined in a GIS with features digitized from very large-scale local map, such as a city street map, this would normally not be done. The level of accuracy and detail of the digital data can be as good as that of the original maps. Georeferencing: All spatial data files in GIS are georeferenced. Georeferencing refers to the location of a layer or coverage in the space as a definition by a known coordinate referring system. With raster images, a common form of georeferencing is to indicate the reference system, the reference units and the coordinate positions of the left, right, top, and bottom edges of the image. The same is true of the vector data files, although the left, right, top and bottom edges now refer to what is commonly called the bounding rectangle of the coverage; rectangle which defines the limit of the mapped area (corners of a feature). This information is particularly important in an integrated GIS since it allows raster and vector files to be related to one another in a reliable and meaningful way. It is also vital for the referencing of the data values to actual positions on the ground. GIS Applicability: The society is so complex, and their activities so interwoven, that no problem can be considered in isolation or without regard for the full range of its interconnections. For example, a new housing development will affect the local school system. The volume of city traffic put constraints on the maintenance of buried pipe networks, affecting health. The action needed to solve such a problems are best taken on the basis of standardized information that can be combined in many ways to serve many users. GISs have this capability. Environmental and resource management: Decision making is becoming increasing complex as dwindling natural resources and more demanding economic priorities diminish the chances of today’s decision being right tomorrow. Furthermore, environmental awareness is constantly increasing among the general public, particularly among the younger generation. To help us map and monitor changes, and plan appropriate responses that take account of the
10
complex interactions of the Earth system, many countries now have comprehensive programs to capture and archive information on the existing natural resources and known sources of pollution, using technologies such as satellite remote sensing and GIS. The data may be used both to expose conflicts and to examine environmental impacts and even simulate the causes and the alternative will become possible. Planning and development The planning and development of new housing, roads, and industrial facilities require data on the terrain and other geographical information. Development often involves building on marginal terrain, increasing the density of the building in the areas already built up, or both. Yet the new structures must fit within the existing technical infrastructure; here computerization is a great aid. One of the benefits GIS holds for such projects is a mineralization of disruption to the existing infrastructure. Escalating construction costs have made the optimizing of building and road location extremely important. Minimizing blasting and earthmoving are significant aspect of minimizing costs. Flexibility is vital: plans should be amenable to rapid changes as decisions are made. The influence of special interest groups and individual citizens require that initial plans be presented effectively and in a manner that is easily understood. Simplified, visualized plans are instrumental in conveying both the content of the scheme and the nature of any likely impact on those concerned. Management and public services: In modern societies, decisions should be made quickly, using reliable data, even though there may be many differing viewpoints to consider and large amount of information to process. Today, the impact of development decisions is ever greater, involving conflicts between society and individuals, or between development and preservation. Information must therefore be readily available to decision makers; the majority of such information is likely to be geographical in nature, and best handled using GIS. Overviews of administrative units and properties are crucial in the development of both virgin terrain and built-up area, in both developed and developing nations. In many countries, property registration is extensive: even in smaller states, 2 to 3 million properties maybe involved. Moreover, property is also an economic factor in taxation and security for loans; so comprehensive overviews are essential to a well-ordered society. Computerized registers based on GIS technology are now well established in many countries. Land transportation: In many countries, the greater part of transportation has shifted from rail to road, at the same time, the use of private vehicle has greatly increased. These developments have created traffic problems, which cause loss of time and money. Large goods are now transported by road. In most countries the annual costs of traffic accidents have become extremely high. The automobile industry is now investing heavily in the development of driver information system, and several systems are now in the market. In principle, all of them involve simple GIS function with digital maps and supplementary information.
1 11
Remotee sensing Remote R sensiing brings up p the eventss of recordinng, observinng, and sensiing objects oor events in n faraway (rremote) placces. In remo ote sensing, the sensors are not in direct contaact etic radiatioon normally is used as thhe with the objects or events e being observed. Electromagn E informatiion carrier in remote seensing. The output of a remote senssing system is usually aan image rep presenting th he scene bein ng observed d. Remote sensing s can define d in two o formsActive Passive Passive remote r sensiing systems record the reflected ennergy of elecctromagneticc radiation oor the emitted energy frrom the earth h, such as caameras and thhermal infraared detectorrs. Active reemote sensin ng systems send s out theiir own energgy and recorrd the reflectted portion oof that energ gy from the earth’s surfaace, such as radar imaginng systems. Remote Sensing S involv ves four basicc inputs: 1. The T Target; 2 . The Platforrm; 3. The Seensor(s); and 4. The Signaal (usually eleectromagneticc radiation or acoustical waaves)
Principlees of Electro omagnetic Radiation R Electromagne E etic radiation n is a form of energy w with the propperties of a wave, and iits major so ource is the sun. Solar energy e traveeling in the form of waaves at the sspeed of lighht (denoted as c and equals e to 3* *108ms–1) is known ass the electroomagnetic sppectrum. Thhe gth is λ the distance d betw ween successsive crests of the wavees. The frequuency is thhe waveleng
1 12
number of o oscillation ns completeed per secon nd. Wavelenngth and freqquency are rrelated by thhe following g equation: C =λ*
The T electrom magnetic specctrum, despitte being seenn as a continnuum of wavvelengths annd frequencies, is divid ded into diffferent portiions by scieentific convvention (Fig. 1.1). Majoor divisionss of the elecctromagneticc spectrum, ranging froom short-waavelength, hiigh-frequenccy waves to o long-waveelength, low w-frequency waves, incllude gammaa rays, x-rayys, ultraviollet (UV) rad diation, visib ble light, inffrared (IR) radiation, r miicrowave raddiation, andd radio wavees. The visib ble spectrum m, commonlly known ass the rainboow of colorss we see ass visible lighht (sunlightt), is the porttion of the electromagne e etic spectrum m with waveelengths betw ween 400 annd 700 billio onths of a meter m (0.4–0.7 m). Althou ugh it is a nnarrow spectr trum, the vissible spectrum m has a greeat utility in satellite rem mote sensing g and for thee identificatioon of differeent objects bby their visible colors in photography. The IR spectrum m is the reegion of eleectromagnettic radiation n that extends from the visible v region n to about 1 mm (in wavvelength). Innfrared wavees can be further f partiitioned into the near-IR R, mid-IR, aand far-IR spectrum, w which includde thermal radiation. r IR R radiation can be meaasured by ussing electroonic detectorrs. IR imagees obtained by sensors can yield im mportant infformation onn the health of crops annd can help in visualizin ng forest fiires even when w they are a envelopeed in an oppaque curtain of smokke. Microwaave radiation n has a waavelength raanging from m approxim mately 1 mm m to 30 cm m. Microwaaves are emiitted from th he earth, from m objects suuch as cars and planes, and from thhe atmospheere. These microwavess can be detected d to provide innformation, such as thhe temperatu ure of the ob bject that em mitted the miicrowave. Because theirr wavelengthhs are so lonng, the energ gy available is quite smaall compared d with visiblle and IR waavelengths. T Therefore, thhe fields off view must be large en nough to deetect sufficieent energy to record a signal. Moost hus are chaaracterized by low sppatial resoluution. Activve passive microwave sensors th microwav ve sensing systems s (e.g., radar) pro ovide their own source oof microwavve radiation tto illuminatte the targetss on the grou und. Remote R sensiing, geograp phic informaation system (GIS) and gglobal positiioning system m (GPS) prrovides extreemely usefull tools for en nvironmentaal and naturaal resources managemennt. They arre widely recognized r as supportiing tools ffor the plaanning, monnitoring, annd managem ment of the appropriate utilization of o resourcess at the coun untry, regionnal and global levels.
1 13
FOS SS for Geo oinform maticss (FOS SS4G G) Ra amachandra T V and Uttam U Kum mar Geoinfo ormatics Lab boratory f Infrastructture, Sustain nable Transpo ort and Urba an Planning Centre for In ndian Institutte of Science e, Bangalore 560 012 E Mail: ce
[email protected]
Geoinformatic G cs constitute a vital component of info ormation scie ence for addre essing the problems of geography, g ge eosciences and a related b branches of e engineering. This domain ospatial analy ysis and mod deling, develo opment of ge eospatial dattabases, combines geo nformation sy ystems desig gn, human-co omputer interraction and b both wired an nd wireless in networking technologies. Geoinformati G ics uses geo ocomputation n for analysing on. One of the e major appliications of ge eoinformaticss in recent tim mes is the geoinformatio capes over multiple m spatia al and tempo oral scales encompassing ga study of variattion in landsc mains – land use and land d cover chang ge, climate cchange, wate er resources, variety of dom pment, natura al disaster mitigation, m etcc. Geoinforma atics include geographic urban develop nformation sy ystems, spatiial decision support s syste ems, global p positioning syystems (GPS S), in and remote se ensing. Geographic G In nformation Sy ystems (GIS) are increassingly being u used as the p principal tool fo or digital exploration of va ariation in lan ndscapes, ass they provide e the necesssary functionss fo or spatial datta collection, managemen nt, analysis a and representation (Turne er et al., 2001 1; Longley et al., 2005; Stein niger and We eibel, 2009; R Ramachandra a et al., 2004 4).These tools
14
provide new and critical ways of understanding our earth and its biogeochemical cycle. GIS software used for this kind of studies fulfill several GIS functionality including: i.) Ensure world wide development, advancement and application of solutions. ii.) Allow studying of data, methods and algorithm implementation. iii.) Furthermore, developed models and algorithms need not be reimplemented by others in order to continue research or validate previous results. Apart from these, researchers should have access to libraries of the original models for analysis, validation, development and implementation (Steiniger and Hay, 2009; Jolma et al., 2008b) for further improvement and customisation depending on the local requirement. Over the last years the paradigm of Free and Open Source Software (FOSS) development has taken root in the GIS community, resulting in the creation of several sophisticated GIS software projects whose aim is to develop free software for numerous purposes. GIS software fulfilling the specific requirements have been distributed with licenses that grant more freedoms of use and that support openness, such as licenses used by FOSS GIS projects (for example: http://grass.itc.it/; http://wgbis.ces.iisc.ernet.in/grass). FOSS have proved to be promising tools that allow us to see and change the software codeswritten in any programming language. FOSS is generally synonymous with free softwareand open source software, and describes similar development models, but with differing cultures and philosophies. Because of the way it is licensed, it has the potential to be legally given away for free or for very little cost and copied and shared with others. FOSS, F/OSS or FLOSS (for Free/Libre/Open Source Software) is liberally licensed to grant the right of users to study, change, and improve its design through the availability of its source code. It has more scope for being available in multiple languages and for being adapted or tweaked to particular needs. This can be very useful for students, researchers, teachers, scientists wanting to use legal software that is appropriate to their needs and fits within their modest budgets. This approach has gained both momentum and acceptance as the potential benefits have been increasingly recognised by many (Steiniger and Hay, 2009).This is proving to be the boon to researchers from economically disadvantaged countries.
The open source definition is used by the Open Source Initiative to determine whether or not a software license can be considered open source. Under the open source definition, licenses must meet the following ten conditions in order to be considered open source licenses: •Free
redistribution: the software can be freely given away or sold. (This was intended to expand sharing and use of the software on a legal basis.) •Source code: the source code must either be included or freely obtainable. (Without source code, making changes or modifications can be impossible.)
15
• • • •
•
• • •
Derived works: redistribution of modifications must be allowed. (To allow legal sharing and to permit new features or repairs.) Integrity of the author's source code: licenses may require that modifications are redistributed only as patches. No discrimination against persons or groups: no one can be locked out. No discrimination against fields of endeavor: commercial users cannot be excluded. Distribution of license: the rights attached to the program must apply to all to whom the program is redistributed without the need for execution of an additional license by those parties. License must not be specific to a product: the program cannot be licensed only as part of a larger distribution. License must not restrict other software: the license cannot insist that any other software it is distributed with must also be open source. License must be technology neutral: no click-wrap licenses or other mediumspecific ways of accepting the license must be required.
There is a distinction between open source software and free software. Open source software are those, for which the human-readable source code is made available under a copyright license (or arrangement such as the public domain) that meets the Open Source Definition. This permits users to use, change and improve the software, and to redistribute it in modified or unmodified form. It is often developed in a collaborative manner in a public domain. Public domain comprises the body of knowledge and innovation (especially creative works such as writing and inventions) in relation to which no person or other legal entity can establish or maintain proprietary interests within a particular legal jurisdiction. This body of information and creativity is considered to be part of a common cultural and intellectual heritage, which, in general, anyone may use or exploit, whether for commercial or non-commercial purposes. Public domain software is not protected by copyright and may be copied and used without payment (http://www.fsf.org/; wikipedia).
On the other hand, free software is software that can be used, studied, and modified without restriction, and which can be copied and redistributed in modified or unmodified form either without restriction, or with restrictions only to ensure that further recipients can also do these things. To make these acts possible, the human readable form of the program (called the source code) must be made available. The source code can be placed in the public domain, accompanied by a software license saying that the copyright holder permits these acts (a free software licence), etc. (http://www.fsf.org/). The first formal definition of free software states that software is free software if people who receive a copy of the software have the following four freedoms: 1.) Freedom 0: The freedom to run the program for any purpose. 2.) Freedom 1: The freedom to study and modify the program. 3.) Freedom 2: The freedom to copy the program so you can help your neighbor.
16
4.) Freedom 3: The freedom to improve the program, and release your improvements to the public, so that the whole community benefits. Freedoms 1 and 3 require source code to be available because studying and modifying software without its source code is highly impractical (http://www.fsf.org/). Proprietary software has restrictions on copying and modifying as enforced by the proprietor. Restrictions on modification and copying are sought by either legal or technical means or sometimes both. Technical means include releasing machine-readable binaries to users and withholding the human-readable source code. Legal means can involve software licensing, copyright, and patent law. Copyleft is a form of licensing and may be used to modify copyrights for works such as computer software, documents, etc. In general, copyright law allows an author to prohibit others from reproducing, adapting, or distributing copies of the author's work. In contrast, an author may, through a copyleft licensing scheme, give every person who receives a copy of a work permission to reproduce, adapt or distribute the work as long as any resulting copies or adaptations are also bound by the same copyleft licensing scheme. A widely used and originating copyleft license is the GNU General Public License. The GNU General Public License (GNU GPL or simply GPL) is a widely used free software license, originally written by Richard Stallman for the GNU project. It is the license used by the Linux kernel. The GPL is the most popular and well-known example of the type of strong copyleft license that requires derived works to be available under the samecopyleft. Under this philosophy, the GPL is said to grant the recipients of a computer program the rights of the free software definition and uses copyleft to ensure the freedoms are preserved, even when the work is changed or added to. This is in distinction to permissive free software licences, of which the BSD (Berkeley Software Distribution) licenses are the standard examples. The GNU Lesser General Public License (LGPL) is a modified, more permissive, version of the GPL, intended for some software libraries.
FOSS has become an essential component in geoinformatics research. Many free and open source software are available that facilitate customisation, provide good support via forums and email lists and have up-to-date documentation. Among the many GIS tools that are frequently used are Desktop GIS, Mobile GIS, Remote Sensing and Image Processing software, GIS extensions and libraries, Spatial Database Management Systems, Map Server and Geostatistical tools (Steiniger and Hay, 2009). Next, we present a non-comprehensive list of the FOSS commonly used in GIS applications along with their web address for further references.
1 17 Application
1 2 3 4 5 6 7 8 9 10 11 12 13 14
GIS
Software e
References
Sav GIS http://www..savgis.org Forestry GIS m/fgis.html http://www..forestpal.com GRASS es..isc.ernet..in/grass http://wgbis.ce QGIS http://qgis.o org ImageILW WIS http://ilwis.o org Processing /uDIG http://udig.rrefractions.ne et V SAGA A http://saga--gis.org Raster, Vector Analysis http://openjjump.org OpenJUM MP http://mapw window.org MapWind dow http://gvsig .gva.es gvGIS http://www.l vc.ele.puc-rio.br/projectss/ InterImag ge http://www..landserf.org Landserff http://www..ossim.org OSSIM http://grasss.itc.it/grass70/manuals/html70_user/rr.l Landscaper.li i.html (GRASS) Analysis mass.edu/lan ndeco/researcch/fragstats/ http://www.um Fragstats s fragstats.httml
15
16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37
Spatial S DBMS S Web MapServer M Statistical S software s
MySQL PostGIS S Postgre SQL ver GeoServ MapServ ver R Gstat Past GeoDa GeoVista STARS
Exploratory y GDAL/O OGR Data Analys sis Generic Mapping g Tool JAMA/GNU pology JTS Top Suite LUPOLib b OpenBug gs Sextane TerraLib MASON Libraries Rapast Simphon ny SWARM M Netlogo
http://www.myysql.org for http://posstgis.refractio ons.net http:// h www.ge eoserver.org g http:// h www.m mapserver.org g http://www.r-p h project.org/ http://www.gs h stat.org http://folk.uio. h .no/ohammer/past/ http://geodace h du/software enter.asu.ed http://www.ge h eovistastudio o.psu.edu http://regiona h lanalysislab.org/index.ph hp/Main/ST ARS A http://gdal.osg h geo.org http://gmt.soe h est.hawaii.ed du/ htttp://math.nisst.gov/javanu umerics/jama a/ htttp://tsusiatso oftware.net/jtts/main.html http://www.ufz h z.de/index.ph hp?en=4302 http://mathsta h at.helsinki.fi/o openbugs/ http://forge.os h sor.eu/projeccts/sextante/ http://www.ter h rralib.org/ http://cs.gmu. h .edu/~eclab/p projects/masson/ http://repast.s h sourceforge.n net/ http:// h www.sswarm.org http://ccl.nort h thwestern.ed du/netlogo/
1 18
38
Multi Agen nt
39 40 41 42 43 44 CMS 45 46 47 Modelling 48 49
50 51 52
Miscellaneous
(Op pen-)Start Log go OBEUS upal Dru Joo omla Atutor en Modeller Ope Des sktop GARP Max xent Lan ndisview PCRaster NET SAN S-D Distance
TAS S Harrvest Qru ule
http://ed ducation.mit.e edu/openstarrlogo/ w.tau.ac.il/~b bennya/resea arch1.html http://www http://drup pal.org/ http://www w.joomla.org g/ http://www w.atutor.ca/ enmodeller.so ourceforge.n net/ http://ope http://www w.cs.princeto on.edu/~scha apire/maxentt/ http://www w.nhm.ku.ed du/desktopga arp/index.htm ml http://sou urceforge.net/ t/projects/land disview aster.geo.uu.nl http://pcra http://ua.tt.u-tokyo.ac.jjp/okabelab/a atsu/ sanet/ http://www w.prd.uth.gr//res_labs/spa atial_analysiss/so ftware/Sd dHome_en.assp http://www w.uoguelph.cca/~hydrogeo/TAS/ http://www w.nrs.fs.fed.u us/tools/harvvest/
Proprietary so oftware licens ses impose several s restricctions on the e use of softw ware such ass ot allowing users to distrib bute the softw ware or to in nstall it on a ssecond comp puter or to no giive it to otherrs. The licens ses also proh hibit a reversse engineerin ng of the softw ware and modifying m the software. If the t source co ode is not avvailable, then it is not posssible to studyy ho ow algorithm ms are implem mented and itt is not possible to improvve the softwa are. FOSS lic censes, such h as the GPL and the LGP PL, explicitly allow users to study, mo odify and rediistribute softw ware. Consequently the following f thre ee benefits off FOSS have e been id dentified(Steiniger and Bo ocher, in pres ss): 1.FOSS 1 avoids ‘reinventin ng the wheel’, 2. in terms of the source code, c FOSS provides p the best ‘docum mentation’ ava ailable, and 3. users can adapt a the sofftware to their own needss without resttrictions.
All three points are essentia al for researc ch, when conssidering thatt research sh hould not be llimited ality that is prrovided by th he software, rresearch exp periments ne eed to be by the functiona peatable and reproducible e, and researrch can prog ress faster w when models can be analysed, rep validated, and improved dire ectly, i.e. bas sed on sourcce code, witho out the probllem of sinterpretatio on, as may be e the case when knowled dge is obtaine ed/interprete ed from article es mis (Stteiniger and Hay, H 2009). In addition to o these generral research advantages, the use of F FOSS lice enses have enhanced e education and knowledge trransfer, partiicularly in devveloping cou untries tha at don't have the (financia al) resources.. Students an nd researche ers can freelyy and legally dow wnload the software and study the alg gorithms. Fin nally it benefitts society in general, as tthe use e of free softw ware licenses can facilita ate the appliccation of new w technologiess and knowle edge tha atenables a sustainable s use of resourc ces (Jolma e et al., 2008b). If such unifiied software dev velopment an nd research efforts could be initiated tthen we see great potenttial to acc celerategeoin nformatics re esearch world d wide.
19 Acknowledgement: We thank Prof. T G Sitharam, Chairman, CiSTUP and Prof Mohan Kumar, Secretary, KSCST for agreeing to support the discussion meeting “Open Source GIS in India: Present Scenario” on 16th November, 2009. This article is written to provide the background information pertaining to FOSS4G.
References: • Jolma, A., Ames, D.P., Horning, N., et, al., 2008a. Free and open source geospatial tools for environmental modelling and management. In: Jakeman, A.J., Voinov, A.A., Rizzoli, A.E., Serena, H.C. (Eds.), Environmental Modelling, Software and Decision Support. Elsevier, Amsterdam, pp. 163–180. • Jolma, A., Ames, D.P., Horning, N., et, al., 2008b. Environmental modeling using open source tools. In: Shekhar, S., Xiong, H. (Eds.), Encyclopedia of GIS. Springer, New York, pp. 275–279. • Longley, P.A., Goodchild, M.F., Maguire, D.J., and Rhind, D.W., 2005. Geographic information systems and science, 2nd ed. Wiley, Chichester. • Steiniger, S., and Bocher, E., in press. An overview on current free and open source desktop GIS developments. International Journal of Geographical Information Science.doi:10.1080/13658810802634956. • Steiniger, S., and Hay, G. J., 2009. Free and open source geographic information tools for landscape ecology. Ecological Informatics, 4, 183-195. • Steiniger, S., and Weibel, R., 2009. GIS software — a description in 1000 words. Available from: http://www.geo.unizh.ch/publications/sstein/gissoftware_steiniger2008.pdf. • Turner, M.G., Gardner, R.H., and O’Neill, R.V., 2001. Landscape ecology in theory and practice pattern and processes. Springer, New-York. Useful links: http://wgbis.ces.iisc.ernet.in/foss http://www.opensourcegis.org http://www.spatialserver.net/osgis http://www.spatialanalysisonline.com http://www.ai-geostats.org http://ces.iisc.ernet.in/grass Mirror site for GRASS in India http://ces.iisc.ernet.in/biodiversity Geoinformatics Applications http://ces.iisc.ernet.in/energy
2 20
Inttroduction n to Linu ux and Insstalling U Ubuntu Linux is an operating system: a series of prrograms thatt let you inteeract with yyour computer and run other prograams.Linux is i modeled on the UNIIX operatingg system. Frrom the starrt, Linux waas designed to be a multti-tasking, multi-user m sysstem. These facts are ennough to makke Linux diifferent from m other well-known operating systtems. Howevver, Linux is even morre different than you miight imaginee. Linux has a capabilityy to be installled on any ttype computer hardwaree. Linux is a leading serv ver operating system, annd runs the 110 fastest suupercomputeers st in the wo orld (21 fasttest can be found f @SER RC, IISc). The deveelopment off Linux is on ne of the mo ost prominennt exampless of free andd open sourcce software collaboratio on; typically all the undeerlying sourcce code can be used, freeely modified, and redisstributed, bo oth commerccially and no on-commerccially, by annyone under licenses succh as the GNU G Generall Public Liccense. Typiccally Linux is packagedd in a formaat known as a Linux distribution fo or desktop an nd server usse. Some poppular mainsttream Linuxx distributionns D (and its derivativ ves such as Ubuntu), U Reed hat, Fedoora and opennSUSE. Linuux include Debian distributiions includee the Linux kernel and d supportingg utilities annd libraries to fulfill thhe distributiion's intendeed use. A distribution orienteed toward deesktop use may m include the X Winddow System,, the GNOM ME and KDE E desktop en nvironmentss. A distribu ution intendded to run as a server m may omit anny graphicall environmeent from the standard in nstall and innstead includde other softtware such aas the Apacche HTTP Server and d a SSH seerver like O OpenSSH. B Because Linnux is freely redistribu utable, it iss possible for f anyone to create a distributionn for any iintended usse. Common nly used app plications wiith desktop Linux systeems include the Mozillaa Firefox weeb browser, the OpenOfffice.org offiice applicatio on suite. n supporting g user space system toolss and librariees from the GNU Projecct (announceed The main in 1983 by b Richard Stallman) S aree the basis fo or the Free S Software Fouundation's prreferred nam me GNU/Lin nux.
Ubuntu U is a complete c Op pen source - desktop Linnux operatingg system, freeely availabble with w both com mmunity and d profession nal support. A And people should havee the freedom m to o customize and alter theeir software in whatever way they seee fit for theiir Use. Ubuntu U will always be free f of charge, and therre is no exttra fee for thhe “enterprisse ed dition”, we make m our very best work k available too everyone oon the same Free terms.
21
Ubuntu includes the very best in translations and accessibility infrastructure that the Free Software community has to offer, to make Ubuntu usable by as many people as possible. Ubuntu is shipped in stable and regular release cycles; a new release will be shipped every six months. You can use the current stable release or the current development release. A release will be supported for 18 months. Ubuntu is entirely committed to the principles of open source software development; we encourage people to use open source software, improve it and pass it on. Ubuntu is suitable for both desktop and server use. The current Ubuntu release supports Intel x86 (IBM-compatible PC), AMD64 (Hammer) and PowerPC (Apple iBook and Powerbook, G4 and G5) architectures.
Sponsorship by Canonical The Ubuntu Project is sponsored by Canonical Ltd. Canonical will not charge license fees for Ubuntu, now or at any stage in the future. Canonical's business model is to provide technical support and professional services related to Ubuntu. We encourage more companies also to offer support for Ubuntu, and will list those that do on the Support pages of this web site. Debian: Debian is an all-volunteer organization dedicated to developing free software and promoting the ideals of the Free Software community. Ubuntu and Debian Ubuntu and Debian are distinct but parallel and closely linked systems. The Ubuntu project seeks to complement the Debian project in the following areas: Package selection Ubuntu does not provide security updates and professional support for every package available in the open source world, but selects a complete set of packages making up a solid and comprehensive desktop system and provides support for that set of packages. For users that want access to every known package, Ubuntu provides a "universe" component (set of packages) where users of Ubuntu systems install the latest version of any package that is not in the supported set. Most of the packages in Ubuntu universe are also in Debian, although there are other sources for universe too. See the Ubuntu Components page for more detail on the structure of the Ubuntu web distribution. Development community Many Ubuntu developers are also recognized members of the Debian community. They continue to stay active in contributing to Debian both in the course of their work on Ubuntu and directly in Debian.
22
Getting Ubuntu Ubuntu can be easily downloaded based on the architecture of the system in use. Ubuntu parent website http://www.ubuntu.com/ support various mirrors located in various countries to download various version of Ubuntu CD/DVD. Burn the ISO image into the media. If your machine doesn't support CD booting, but you do have a CD, you can use an alternative strategy such as hard disk,usb stick, net boot, or manually loading the kernel from the CD to initially boot the system installer. The files you need for booting by another means are also on the CD; the Ubuntu network archive and CD folder organization are identical. So when archive file paths are given below for particular files you need for booting, look for those files in the same directories and subdirectories on your CD. Once the installer is booted, it will be able to obtain all the other files it needs from the CD.If you don't have a CD, then you will need to download the installer system files and place them on the hard disk or usb stick or a connected computer so they can be used to boot the installer. System Requirements Ubuntu does not impose hardware requirements beyond the requirements of the Linux kernel and the GNU tool-sets. Therefore, any architecture or platform to which the Linux kernel, libc, gcc, etc. have been ported, and for which an Ubuntu port exists, can run Ubuntu.
Supported Architectures Ubuntu 10.04 onwards supports three major architectures and several variations of each architecture known as “flavors”. Three other architectures (HP PA-RISC, Intel ia64, and IBM/Motorola PowerPC) have unofficial ports. Architecture Intel x86-based AMD64 & EM64T HP PA-RISC Intel IA-64 IBM/Motorola PowerPC sun4v
Ubuntu Designation i386 Intel amd64
Subarchitecture
Flavor
hppa
PA-RISC 1.1 PA-RISC 2.0
32 64
PowerMac Sun SPARC
pmac sparc
ia64 powerpc
sun4u sparc64
2 23
INSTAL LLING UBU UNTU 1. System must need a miniimum of 2.5 5gb Hard dissk space for installing L Linux, 256M MB raam. 2. This T can be checked by Pressing P F2 at a startup of the system aand checkingg the BIOS
3. Iff u want hav ve a dual bo oot(ex: Win ndows and L Linux, Installl windows first and theen liinux preferab bly in last haard drive/parrtition) 4. Set the boot parameters p in n the bios menu m while sttarting your computer.(nnormally F2)) 5. Go G to boot deevices prioritty and chang ge priority aas First priorrity as CD/D DVD drive annd laater 2nd priorrity as Hard disk d 3rd as Usb U and so onn.
6. Exit E saving Changes C 7. In nsert UBUN NTU cd in yo our CDROM M and wait uuntil it bootss and loads the necessarry boot files. 8. At A first UBUN NTU will assk for certain n informationn about Langguage, Placee and date.
2 24
9. Enter E the Key yboard layou ut in next steep(default : U US English)
10. Enter E little peersonal inforrmation such h as your nam me, User id aand passworrd in this stepp.
11. Now N Ubuntu will return to t the partitio oning screenn.
2 25
12. There T are two options o AUTO A paartition annd Manuual(Advancedd) partition(Recommended). 13. Auto A partition n will take partition p by itself i with uunfilled spacce. Manual P Partition is fo for th he user to paartition based d on his usag ge and requirrement. 14. Let L us consid der installing in system having h 100gbb of space w with 2gb Ram m 15. In I manual partition Roo ot (“/”) will be given soome space(aabout 20 gb)) Swap spacce will w be allottted based on n ram( in th his example its 2gb ram m, Hence alllot about 6ggb sp pace) 16. Allot A some space s to Home folder “/home” “ as 50gb so thaat the files can be easily acccessed (acts as sec. driv ve) 17. Allot A rest to /usr/local / so that any lib braries can innstalled withh ease( instaalling librariees in n /usr/local will w help in n access of libraries l by all the proggrams easy and access tto otther users eaasily) 18. Press next and installation n starts with h formatting tthe space thaat has been sspecified.
W this, yo ou have succcessfully in nstalled Linuux and is a time for nnew world oof 19. With Ubuntuing! U Happy H Time with Ubuntu u Linux.
2 26
Change the t root passsword at firstt use and maaintain the paassword This can be done by going to terrminal and ty yping SU PA ASSWD andd enter the nnew passworrd twice SUPPOR RT OPEN SOURCE SO OFTWARES S! Referencces http://ww ww.ubuntu.com
2 27
Hands-on Tutori T al for GRASS GIS Beginnerss
GRASS S GIS
Energy E an nd Wetlands Resea arch Group, Centre for f Ecolog gical Scien nces, Indian Institute of o Science e Bang galore – 560012 5 http p://wgbis.ces.iisc.e ernet.in/gra ass htttp://wgbis.ces.iisc.e ernet.in/fo oss
email:
[email protected] sc.ernet.in, energy@ce e es.iisc.erne et.in ___ __________ _________ __________ __________ __________ __________ _______
2 28
GRASS G Te eam at IIS Sc Dr. T. V. Ramachandra R a Uttam Kumar P S. N. Prasad N. V. Joshi H Aithal Bharath H. Anindita Dasgupta Setturu Bharath Vishnu Bajpei
[email protected] uttam
[email protected] ernet.in snaren ndra.prasad@ @gmail.com
[email protected] bhara
[email protected] anindita_ _dasgupta@ @ces.iisc.erne et.in settu
[email protected] ernet.in vishn
[email protected]
Usefful websites s for GRAS SS GIS refe erences 1)) http://gras ss.osgeo.org//wiki/Installattion_Guide 2)) http://grasss.osgeo.org//grass64/mannuals/html644_u ser/helptext.html 3)) http://grasss.osgeo.orgg/wiki/Importi ng_data 4)) http://grasss.osgeo.orgg/gdp/tutorialss.php 5)) http://ces.iisc.ernet.in/eenergy 6)) http://wgbis.ces.iisc.ernet.in/energyy/paper/reseaar html#rg chpaper.h
29 PART T-I
INSTA ALLATION OF O GRASS S 64
Log g on to http:///grass.osgeo o.org/ Click on the Do ownload Tab
Linux for free e download o of GRASS GIS Click on GNU/L
30
Click on to Dow wnload older GRASS verssions
This is what you get.
31
Scroll down to the t bottom of o the page. A And click on U Ubuntu.
This Screen ap ppears after clicking c the b button.
32
Click on karmic c option.
This is the wind dow what com mes next.
33
Scroll down and d click on the e i386 shown n below.
Then click to th he link mirror..pn.gov/ubun ntu/
3 34
Then save the file.
ownload is co omplete and file is saved. This appears on the screen when the do
3 35
Then double click on grass s 6.4.0-re5--2i386.deb. This will be downloaded in the ome/user/Dow wnloads dire ectory. /ho
Click on Installl Package.
Then installatio on starts.
36
Aftter the installation is finish hed click on C Close button n.
Co ongratulations s. Installation n is complete e.
37 PART – II
ST TARTING GRASS G 64
Click on the terrminal
Terminal opens s.
38
Type grass64 or type grass and d press Tab button b on you u key
Grass GIS S version 64 is ready for operation. o
3 39 PART – III I
DOWNLOA D ADING GRA ASS DATAS SETS
To download de efault data frrom Grass G IS Website: goto http://grrass.osgeo.o org/ Click on Download Tab
ample datase ets. This is Sp pearfish, Sou uth Dakota, U USA data sett. Click on the Sa
4 40
Click on Spearffish
at Data set de escription Click on Full tar gz (135MB). Additional information is available a d Quick usag ge tutorial. and
41
Sa ave file.
42
PART – IV
START GRASS G 64
These are the downloaded d GRASS data a sets.
Ex xtract here.
4 43
Start gra ass64 by typing grass 64 on terminal
et your grass directory herre and Click on nc_spm_ _08 and PERMANENT on n right Se sid de.
4 44
This is what ap ppears on scrreen as Grasss GIS displa ay is opened.
45
PAR RT – V
WO ORKING WITH W RASTE ER DATA Click on add a Raster la ayer.
Click on raster map to display.
46
This is what comes up next.
Go o to aspect off Select item--Raster map and click Okk.
4 47
Click on the ma ap display.
her utilities – zoom in, zoo om out, The image will be shown in the map dis play. The oth ve display (in n jpeg, tiff), pan, p measure ement of leng gth segment, printing are in the top sav me enu of the dis splayed imag ge.
4 48
No ow we will dis splay a classified land use e map. Display Landclasss96 rasterma ap. Then to add d legend, add raster lege end layer to g get the legend as shown b below.
49
PART – VI
WO ORKING WIITH 3D DAT TA
Click on the ele evation map to t see it on t he display sccreen.
un g.region fo or region settting. Ru
50
The region has s been set.
Click here on NVIZ. N
51
Aftter clicking on n NVIZ this is s what appea ars.
This is what we e get as output.
5 52
Click on draw and a change the directionss, exaggeration as you wish.
There is option to decrease e the light inte ensity. Here yyou can see the effect of lighting a from different angles.
5 53
Inc creasing lightt intensity fro om front.
We e can add ve ector layer fro om the map b rowser. Herre water and road layer iss added.
54
And change the e direction to o see a betterr view. .
55
PART T – VII
WORKING WITH VECTOR V DATA
Op pen a window w.
Click on add ve ector button.
56
Click on the vec ctor map to display. d
Se elect the vecto or file.
57
Click on display y active layerrs.
We e can see the e vector map p on the scree en.
5 58
To o query: Click k on the Querry button to obtain o additio onal informattion from data abase.
To measure line segment le ength. Click o on the Scale bar.
5 59
Drraw line to measure the distance betw ween one point to the othe er and see th he length on the e output file.
6 60 PA ART – VIII
WORKING WITH W OWN DATA: CR REATING A NEW LOC CATION
Aftter having a practice p on default d data, n now, we can start working with our ow wn data for our study area. Before that, we need to create a Loccation and Ma apset. Location can be cre eated in 1) X, Y coordinate syste em. c syystem. 2) Latitude, Longitude coordinate 3) UTM sys stem. We e start with creating X, Y location.
How H to create e Location in GRASS? Go o to the terminal and type grass64 -texxt
6 61
Here we ha ave to enter the t Location and Mapset name. We shall set LOC CATION: Bangalore and a MAPSET T: username e. We can cre eate a folder (directory) to o keep all the e data. Let us s name that folder f as “gra assdata”. The en set the de efault database directory (DATABAS SE: /home/user/grassdata a).
Aftter pressing < and < to ogether we ge et this on scrreen.
62
Aftter typing y when w we pres ss enter, we g get this.
Again after typing y we get this screen.
63
To create x, y location, type A on th he text box and a press y.
Aftter typing Y on o the text bo ox, write one line descripttion.
64
Type y after completing the description.
Pre ess tog gether to con ntinue.
6 65
Type y to accep pt the region.
c Prress Enter an nd Press Ctrrl + C to exit G GRASS. x, y location is created!!!!
6 66
PART – IX
IMPORT TING RAST TER / VECT TOR DATA A INTO GRA ASS
Type grass64 -tcltk to start GRASS in G GUI mode.
Now we e will import Landsat L data a (ETM+ 8 ba ands) to GRA ASS. These d data are downloa aded from GL LCF (Global Land Cover Facility). To import a file run r.in.gdal from terrminal or goto o File -> Impo ort-> Raster--> Multiple fo ormats using gdal.
Brows se the downlo oaded file fro om Download d directory orr wherever yo ou have keptt the files.
67
Giv ve the outputt file name an nd go to the menu option ns.
6 68
Click on overrid de projection and Run to import the im mage.
File is impo orted and we e can see it o on map displa ay by clicking g on it.
6 69
IMPOR RTING IRS 1C/1D/P6 LIS SS-III/IV Data a INTO GRAS SS
Us se the comma and r.in.gdal to import IRS S data or go to File, Impo ort raster map, multiple forrmat using gd dal from the selection. s
Give G the inpu ut and output file name an nd click on Run.
70
This appears on screen whe en the progra am has ran ccompletely.
Click on the ma ap display to see the impo orted bands.
7 71
You can import vecto or data from other o softwarres in many fformats (such h as dxf, mif, e, kml, etc.) See S the Impo ort options in File menu. W We use the ccommand shape file v.in.dxf to o import dxf vector v file.
Se ee the output dxf vector file by clicking g vector map display.
7 72
We use the command c v.iin.ogr to import kml (Obta ained from G Googleearth d digitization) vector file an nd then click on Run.
Display the km ml output vecttor file by cliccking vector m map display..
7 73
We use the command v..in.ogr to imp port shape ve ector file and then click on n Run.
e shape outp put vector file e. Clicking to vecttor map displlay to see the
7 74
PART T–X
IMAGE PROCESS SING – GENERATING G FALSE C COLOUR COMPO OSITE FRO OM LANDS SAT and IRS S LISS-III M MULTISPE ECTRAL MU ULTISPECT TRAL IMAG GERY
Type the comm mand i.landsa at.rgb on texttbox or goto Imagery Tab b on the main n menu of RASS to sele ect the same option from G olour compossite. GR GUI for creatting False co
This is what ap ppears.
75
In LANDSA AT red chann nel put ETMp plus_2000_ba and4, LANDS SAT green channel put ETMplus_2 2000_band3, LANDSAT blue b channel put ETMpluss_2000_ban d3 and click on Ru un.
The program has ran..
7 76
Type the comm mand r.compo osite in the te ext box or bro owse from G GRASS main Menu.
Name of o raster map p to be used ffor putt ETMplus_2 2000_band4. Name of o raster map p to be used ffor p put ETMpluss_2000_band d3. Name of o raster map p to be used ffor pu ut ETMplus_ _2000_band2 2. Give na ame to the ou utput file and click on Run n.
7 77
The prrogram has ran. r
Dis splay Landsa at ETM False e colour comp posite on the e map displayy.
78
Dis splay IRS Fa alse colour co omposite on t he map disp play.
7 79
Tutorial T for r.li program m in GRA ASS Also o see http p://grass.o osgeo.orgg/gdp/land dscape/r__le_manu ual5.pdf Creation o of configurattion file Type the ccommand on the terminall ‐ r.li.setup.
8 80 When r.li..setup window w is displayed d then click on NEW.
When the e new configu uration windo ow is displaye ed then enter the Configurration file nam me and namee of the rasster map.
8 81
Click on Setup samplin ng frame.
82 Then select Whole maplayer and cllick OK.
The whole e maplayer iss set as sampling frame.
83
Click on Setup samplin ng areas.
8 84 Select Mo oving window w and click OK K.
Click on U Use keyboard to enter movving window w dimension.
85 Select Recctangle and e enter 5 on height and widtth sizes.
Click on Save settings.
New conffiguration file my_conf is ccreated!!! The details of the metrrics are given in http://wggbis.ces.iisc.errnet.in/grass//grass64/man nuals/html644_user/r.li.htm ml
86 Type these commands on the GRASS terminal in order to get various spatial metrics such as MPS (Mean Patch Size), Patch density, Edge density, etc.
1. d.mon start=x0 (to start a monitor) 2. r.le.trace: Computes the number of patches along with their area and perimeter.
3. r.le.trace -pt map=urban_only out=urban_only_patch 4. r.li.mps: Calculates mean patch size index on a raster map, using a 4 neighbour algorithm r.li.mps map=urban_only conf=my_conf output=urban_only_mps A map of NULL values is considered to have zero patches. If you want to have null values instead run r.null setnull=0 map=urban_only_mps
5. r.li.patchdensity: Calculates patch density index on a raster map, using a 4 neighbour algorithm r.li.patchdensity map=urban_only conf=my_conf output=urban_only_patchden A map of NULL values is considered to have zero patches. If you want to have null values instead run r.null setnull=0 map=urban_only_patchden r.null setnull=‐1 map=urban_only_patchden (To remove negative values from the image).
6. r.li.padcv: Calculates coefficient of variation of patch area on a raster map r.li.padcv map=urban_only conf=Ani_conf output=urban_only_padcv r.null setnull=-1 map=urban_only_padcv 7. r.li.edgedensity: Calculates edge density index on a raster map, using a 4 neighbour algorithm r.li.edgedensity map=urban_only conf=my_conf output=urban_only_edgedens r.null setnull=-1 map=urban_only_edgedens
8 87
Qgis Q Ov vervie ew
Quantum Q gis is a GIS too ol for manipu ulating geoggraphical datta, 3-D analyysis, statisticcal analysis. QGIS is Open O Sourcee software and a it is freee of cost.It is an official project oof OSGEO (Open Sou urce Geosp patial Found dation).Avaiilable underr GNU Geeneral Publlic License.Q QGIS is mulltiplatformgiis that runs on o Linux, U Unix, Mac OS SX, and Winndows havinng supportfo or vector, raster, r and database d formats. It suupports manny commonn spatial daata formats (e.g. ESRIS Shape File, geotiff). QG GIS supportss Plug-ins tto do thingss like displaay tracks fro omthe GPS units. u Qgis is a modularr architecturee, which cann be extendeed by modulees for new data d sources and pluginss.
1. Histo ory: Gary G Sherman in Februaary 2002 starrted to devellop a giswithh a wide rannge of data stores.Th he first releasse came on July J 19, 2002 2. The first rrelease suppported only P Post GIS layers.Th he current veersion is Qgiis 1.6.0.
2. Gettiing Started 2.1. Insta allation Building QGIS can be possible in two way ys one is froom source aand other is binaries. Thhe Installation instructio ons are disttributed witth the QGIS S source coode and alsoo available at http://qgiis.org.Standaard installerr packages are a availablee for Windoows and Maac OS X. Foor GNU/Lin nux binary packages p are also availab ble. Get the llatest inform mation on binnary packagees
8 88
at the QG GIS website. For Ubun ntu follow th he steps show wn below Click on System Administratio A on Softwaaresources. Enter pass w word; a dialoogue box wiill be displaayed click on o Other So oftware; this will show w the packaage informattion which is included in the system m.
Click on Synaptic paackage Manaager to installl Qgis and oother sourcess.
89
Or
Code:
First Go to Sy ystem -> Ad dministratio on -> Softwaare Sourcess (eenter your paassword) "T Third Party y Software" Tab Click C "+Add"" When it prrompts you for f a APT Liine, enter: deb http://ppaa.launchpad.net/qgis/ubu untu/ jaunty main d Source" cllick on "Add Then T click "R Reload" to update u Then T type intto the termin nal:
~$ sudo apt-get a instaall qgis (It should d install) if it doesn n't then try this: t Code: ~$ sudo apt-get a updaate Code: ~$ sudo apt-get a instaall qgis Code:
9 90
~$ qgis It opens Qgis Q session n. Or you can n find it auto omatically addded to proggrams: Applicattions -> Edu ucation -> Quantum Q GIIS 2.2 Startting QGIS To start QGIS, Q choosse Qgis from m the menu, Applications A s EducatioonQgis. N Now the QGIIS map canv vas and an em mpty legend d will be app peared. Explorin ng the GUI interface: i GUI show ws 6 differen nt menu barss they are 1. Menu M bar 6. Sttatus bar
2.Tool barr
3.Legen nd
4.Ovverview mapp
5.Mapp canvas
Th he menu bar provides acccess to numeerous QGIS features usinng a standarrd hierarchiccal menu u. The most menu optio ons have a correspondinng tool and vvice-versa; tthe menus arre not organized o qu uite like the toolbars. Th he toolbar coontaining thhe tool is listted after eacch menu u option as a checkbox entry. Th he toolbars offers access to most of th he similar fuunctions as tthe menus, pplus additionnal tools for interactting with th he map. Eacch toolbar ittem has poppup help avvailable. Hold themouse over the t particulaar icon, a short s descripption of thee tool’s purrpose will bbe
9 91
displaayed. Every y menu bar can be mov ved around and can be switched off using righht mousse button con ntext. Th he map legeend area setts the visibility (used tto show or hide the layyer by checck box)aand z-orderiing of layerss. Z-ordering means thaat layers listed nearer tthe top of thhe legen nd are drawn n over layers listed lowerr down in thee legend. QG GIS - maps are displayed in map can nvas area.Thhe loaded veector and raster layers wiill be displayed in th his window.. The map view can be ppanned, zooomed in and out. The maap view and the legend are strongly bound to each otheer - the mapps in view reeflect changees you make m in the legend l area. Th he map overv view panel provides p a full f extent vview of layeers added to it, within thhe view is a red rectangle sh howing the current m map extent. This allow ws to quickly determ mining whicch area of the map is in current c view w. Th he status barr shows the current posiition in mapp coordinates (e.g. meteers or decim mal degreees) as the mouse m pointeer is moved across a the m map view. A progress baar in the statuus bar sh hows progreess of renderring as each layer is draawn to the m map view. Iff a new plugin or a plugin p updatte is availablle, a messag ge will be shhown in the sstatus bar. A At the far righht of thee status bar is i a projectorr icon; by cliicking it opeens the projeection properrties.
2.3 Working with h Vector Data: D QGIS Q uses th he GDAL/O OGR library y to read annd write vector data foormats. QGIIS suppo orts numero ous vector data formats,, including tthose suppoorted by the OGR librarry data provider p plu ugin, such ass ESRI shapeefiles,kml ettc.QGIS alsoo supports P PostGIS layers in a PostgreSQL L database using the Postgre P SQL L data provvider plugin.. Support fo for additional data ty ypes (ex. deliimited text) is provided by additionaal data proviider plugins. To lo oad a vector file click on n av vailable in toool bar or ctrrl+shift+v, a dialogue boox will be b displayed d allows to trraverse throu ugh the file system and load a shape file or other formaats.
The vector v file will w be loaded d and display yed in map ccanvas.
9 92
To im mprove the performance p of vector fiile, righ clickk on the layeer name seleect propertiees. The properties p dialogue d box x will be op pened with m multiple opttions i.e Syymbology taab, Lablees, Attributes, General, Metadata, M Acctions, Diaggram overlayy.
9 93
QGIS sup pports a num mber of sym mbology rendderers to conntrol how vector featurees are displaayed. Single symbol - a single s style iis applied too every objecct in the layeer. Graduated dsymbol - objects o within the layer are displayeed with diffeerent symbools classified by the valu ues of a partiicular field.C Continuous color - objects within thhe w a spreaad of colourrs classified by the num merical valuees layer are displayed with s fielld. Unique value v - objeccts are classiified by the uunique valuees within a specified within a specified s field with each valuehavingg a different symbol. The Labeels tab allows to enable labeling l feattures and conntrol a numbber of optionns related to fonts,placem ment, style, alignment a annd bufferingg. The Attrib butes tab alllows to man nipulate the attributes off the selecteed dataset.Thhe buttonsNeew Column and Delete Column cann be used, w when the dataaset is Editinng mode. Th he OGR libraary supportss toadd new columns, buut not to rem move them, if GDAL veersion >= 1.6 6 installed th hen the delettion is alloweed. The Gen neral tab peermits to ch hange the display nam me, set scaale dependennt rendering g options, vieew or changee the projecttion of the sppecific vectoor layer. The Metaadata tab co ontains geneeral informattion about tthe layer whhichis not yyet editable. It I includes th he type andlocation, num mber of featuures, featuree type, and thhe editing caapabilities. The T Extents section, prooviding layeerextent infoormation, annd the Layerr Spatial Reference Sy ystem sectioon, providinng informatiion about thhe projection n ofthe layerr. QGIS pro ovides the facility f to im mplement ann action bassed on the aattributes off a feature. Actions A are useful u to perform a searcch based on an attribute valueor vieew a web pag ge based on oneor o more values in veector layer. The Diagrram tab used d to overlaya graphicstoo a vector layyer.It allowss overlyingppie charts, baar charts.
2.3Worrking with h Raster Data D The T raster daata in GIS iss usually rem mote sensingg data such as aerial phhotography oor satellite imagery i and d modelled data d such as an elevationn matrix. In QGISthe GD DAL libraryyis used for raster r implem mentation.Q QGIS supportts the follow wing raster foormats, – Arc/Inffo ASCII Grrid – GRASS G Rastter – Geeotiff – JPEG – Erdaas Imagine – USGS ASCII DEM M Raster laayers are loaded either by y clicking on the Load R Raster icon oor by selectinng the View w Add Rasterr Layer menu u option. Mo ore than onee layer can bbe loaded at tthe same tim me by holdin ng down the Control or Shift S key and d clicking onn multiple ittems in the ddialog. Oncee a
9 94
raster lay yer is loaded d in the map p legend by clicking the right mousee buttonopenns a dialog tto set rasterr properties for f the layer.. Improvetthe performaance by right click on lay yer and channge properties by addingg the differennt options. The propertties include – Symbolog gy, Transparrency, Colorrmap, Generral, Metadatta, Pyramidss, and Histog gram.
Qgis Q renders the t raster lay yers in two different d form ms. Gray G scale (on ne band of th he image wiill be rendereed as gray orr in pseudoccolours) Three T band colour (thrree bands from f the im mage will bbe renderedd, each bannd reepresenting the red, greeen or blue component that will bee used to crreate a colouur im mage) Within W both render r types you can inv vert the colouur output using the Inveert colour maap check kbox.
9 95
QGIS Q has thee ability to display d each h raster layerr at varying transparenccy levels. Usse th he transparen ncy slider to o indicate to what extentt the underlyying layers (iif any) should be visible tho ough the currrent raster laayer. This is very useful, to overlay m more than onne raaster layer.E Ex: elevation n map overlaaid by a classsified rasterr map. This w will make thhe lo ook of the map m more threee dimension nal. The T Colormaap tab is only y available, when a singgle-band-renndering is seelected within th he tabSymbo ology Displays D basiic informatio on about thee selected raaster, includding the layeer source annd display namee in the legen nd (which can c be modiffied). The sppatial refereence system is prrinted here as a a PROJ.4--string. This can be moddified by hittting the channge button.
9 96
Displays D information abo out the rasterr layer, incluuding statistiics about eacch band in thhe cu urrent rasterr layer. Large L resolution raster layers can slow naviggation in Q QGIS. By crreating lower reesolution cop pies of the data d (pyramids), perform mance can bbe consideraably improveed ass QGIS sellects the most suitablee resolution to use deppending on the level oof zo oom.Disadvantage with this is it alteers the originnal data. The T Histogram m tab allow ws you to vieew the distriibution of thhe bands or ccolours in thhe raaster. The distribution off the histograam image caan be saved ffor further reeferences.
Raster R Callculator The T Raster Calculator C in the Layer menu m allowss performingg calculationns on basis oof ex xisting raster pixel valuees. The resullts are writteen to a new rraster layer w with a GDA AL su upported forrmat. Ex fro om the rasterr layer of Inndia classifieed image, too separate thhe Bangalore B lay yer multiply the Bangalo ore boundaryy raster layerr can results the classifieed ou utput file fro om the whole.
Geo G referencer: Geo G referenccing usually refers to th he method bby which loocations in tthe raster annd vector GIS fiiles are related to real eaarth-surface positions.Too start geo rreferencing aan nreferenced raster, we must m load it using the buutton you caan geo refereence an imagge un using already y georeferencced image or entering G GCPs manually. Here is the procedurre fo or image to image i georeeferencing. The T raster w will show up in the main working areea of the dialog. Once the raaster is loadeed, we can sttart to enter rreference pooints. Using U the Ad dd Point bu utton, add points to thee main workking area annd enter theeir co oordinates For this proceedure you haave two optio ons:
9 97
a)) Click on a point in the raster imagee and enter thhe X and Y coordinates manually b) Click on a point in thee raster imag ge and choosse the buttonn from map canvas to addd th he X and Y coordinates with the heelp of a georreferenced m map already loaded in thhe QGIS Q map caanvas. c)) With the button, you can c move thee GCPs in booth windowss, if they aree at the wronng place. Continue C enteering points.. You should d have at leaast 4 points, and the morre coordinatees you can provide, the bettter the resultt will be. Thhere are addiitional tools on the plugin dialog to zoo om and pan the working area in orrder to locat ate a relevannt set of GC CP points.
The T figure sh howing the geometricallly correctedd image of Bangalore overlaid witth Bangalore B reg gion vector file. f
9 98
Working W witth OGC Datta OGC services are increasingly y being usedd to exchangge geospatiall data betweeen diifferent GIS implementaations and daata stores. Q QGIS supportts WMS andd WFS as daata so ources. WM MS-support iss native; WF FS and WFS S-T is impleemented as a plugin. Thhe Open O Geospaatial Consorttium (OGC)) is an internnational orgganization with more thaan 30 00 commerrcial, govern nmental, no onprofit andd research oorganisationns worldwidde. Im mportant OG GC specificattions are: – WMS - Weeb Map Serviice – WFS - Web Featuure Service – W WCS - Weeb Coverage C Serrvice – CAT - Web Catallogue Servicce – SF FS - Simple Features for SQL S GML L- Geograph hy Mark-up L Language
Digitizing D QGIS supports dig gitizing from m a raster or vector layerr. It support editing in nclude shapee files, Post GIS, G and GR RASS data. T The attributee values in new table alllow real, intteger, and sttring data typ pes. Point, liine, and polyygon features are su upported. Atttributes are entered as features fe are ccreated. Editting Featuress include Moving veertices or poiints, inserting g new pointss where needded, deletingg unnecessarry points, deletin ng entire feaatures, cut orr copy featurres from onee layer and paste to an nother, multtiple featuress can be copiied/pasted inn one operatiion
9 99
GRASS G GIS Integraation The T GRASS plugin p proviides viewing g, editing, annd geo proceessing of GR RASS data. Itt provides access to GR RASS GIS databases d an nd functionallities. This inncludes visuualization of GRASS raster r and veector layers, digitizing vector layers,, editing vecctor attributes, creating new vecttor layers and d analyzing GRASS 2D and 3D dataa with more than 300 GR RASS modules.. GRASS data are stored d in a directo ory referred tto as GISDB BASE. This ddirectory often callled grass datta, must be created c beforre you start w working withh the GRAS SS plugin in QGIS. Within W this diirectory, the GRASS GIS S data are orrganized by pprojects storred in subdirecttories called LOCATION N. Each LOC CATION is defined by iits coordinatte system, map projection and geographical g l boundaries. Each LOCA ATION can have severaal MAPSETss. In order to t analyze vector and rasster layers with w GRASS modules, yoou must impport them intto a GRASS S LOCATIO ON. Creating g a new GRA ASS LOCA ATION Sttart QGIS an nd make suree the GRASS plugin is lloaded. Clickk on the Opeen Mapset icon in GRASS G toolb bar to open th he MAPSET T wizard or ppluginsGrrassopen M Mapset(if already created; c if no ot create a neew Mapset) .Select an exxisting GRA ASS databasee (GISDBA ASE) folder grassdata orr create one for the new LOCATION N using a filee manager on your com mputer then click c Next. Create C a new w MAPSET w within an existing LOCA ATION or too create a new n LOCAT TION altogetther by using g this wizardd to. Click oon the radio bbutton Creatte new locaation. All thee options in Grass tool bar b will be hiighlighted onnly after opeening the Mapset. . This is necessary, becaause new rastter or vectorr layers creatted during annalysis needd
10 00
to be wriitten to the cu urrently seleected LOCATION and M MAPSET. T The GRASS toolbox provides the function nalities to wo ork with Rasster/ Vector ddata.
To T add Grass raster layeer click on Add Grass raster layerr icon in thhe tool bar oor pluginsGraassAdd raaster layer a dialogue bbox will be displayed w with name oof Location, L Maapset, raster images. Clicck the namee of the rasteer file to be ddisplayed annd cllick OK.
10 01
The T Open GR RASS Tools box providees GRASS m module functtionalities to work with data insid de a selected d GRASS LO OCATION and a MAPSET T. Tool box shows the list of command ds with desccription and modules m avaailable. If useer want to w work with speecific command d, by clickin ng the comm mand will be executed byy showing soome options for input thee data.
QGIS Plugins P QGIS Q allows many new features/func f ctions to be eeasily addedd to the appliication with the help of o plugins(S Similar to Firrefox add-on ns). The Pluggin Managerr provides a rresource to load or unload u plugin ns. Many of the features in QGIS aree actually im mplemented aas either corre or extern nal plugins. – Core Plugiins are mainttained by thee QGIS Devvelopment Teeam and are automatically part off every QGIS S distribution n. They are w written in onne of two lannguages: C+ ++ or Python n. – Extern nal Plugins are a currently y all written in i Python. T They are storred in externnal repositorries and main ntained by th he individuall authors. Thhey can be aadded to QGIIS using the Plugin In nstaller.
10 02
GPS Pllugin GPS, G the Global Positioning System, is a satellitee-based systeem allows too find the exact possition anywh here in the world. w It is ussed as an aidd in navigatioon. It providdes the informatiion of latitud de, longitudee and elevatiion by usingg the signals from the sattellites. Waypoin nts, routes an nd tracks are the three baasic feature ttypes in GPS S data. Most receivers also havee the capabillity to store locations l (kn nown as wayypoints), seqquences of loocations that make up a planned ro oute and a trrack log or trrack of the reeceivers movvement overr time. Theree are differrent data form mats to storee GPS data; Qgis supporrts a standardd interchangge format (GPS eX Xchange form mat), that can n contain any y number off waypoints, routes and ttracks in the same filee. To load a GPX G file, thee plugin sho ould be loadeed. To load pplugins clickk on Plugins Plugin MaanagerGPS S Tools. When thiis plugin is loaded l a buttton with a sm mall handhelld GPS device will show w up in the toolbar.
OGR Converter C Plugin This T is very useful u plugin n; which prov vides the abiility to convvert vector daata from onee OGR-sup pported vecttor format to another (forr example too create KML L file from sshape file). The plug gin requires a few parameeters to be sp pecified befo fore running::
10 03
– Sourcee Format/Da ataset/Layer: Enter OG GR format annd path to thee vector file to be converted d – Targett Format/Da ataset/Layerr: Enter OGR format annd path to thee vector outpput file
Print Composer C The Qgis print composer o provides capability off printing mapps in differentt formats. It aallows addin ng elements such as the QGIS map canvas, c legeend, logo incclusion, northh arrow; imaage formats G/SVG/PDF support, adjustable draw wing scale, separate DPII settings, baasic shapes, like PNG and text labels. l Allow ws to specify y the size, grroup, alignm ment and posiition of eachh element annd adjusting g the propertiies. The layouut can be printed or exportted to imagee formats; Posstscript, PDF F and also provides p an option to savee the layout as a template aand load it aagain in anothher session.
10 04
Help p (Useful websites w fo or Quantu m GIS refferences) http://listts.osgeo.org//mailman/lisstinfo/qgis-user https://lissts.fossgis.dee/mailman/liistinfo/fossg gis-talk-liste http://listts.osgeo.org//mailman/lisstinfo/qgis-developer http://listts.osgeo.org//mailman/lisstinfo/qgis-co ommit http://listts.osgeo.org//mailman/lisstinfo/qgis-trrac http://listts.osgeo.org//mailman/lisstinfo/qgis-co ommunity-teeam http://listts.osgeo.org//mailman/lisstinfo/qgis-ed du
10 05
Geographic Resources Dec cision Support System S – an Op pen Source GIS
htttp://wgbis.ces.iiisc.ernet.in/ene ergy/paper/Geos spatial_to Abstracct Introduction dology Method Overvie ew and description of GRD DSS Databa ase connectivity Applica ations Accuracy Estimation Summa ary Referen nces PDF HOME
Geog graphic Resou urces Decision Support Syystem – an Op pen Source G GIS * Centre for Eco ological Sciences, Indian Institute off Science + Centre for Susta ainable Technolog gies, Indian Institu ute of Science # Salim Ali Centre e for Ornithology and a Natural Historry, Dehradun Geospatial Today y, September-Octo ober, 2004, Vol. 3 3(3), pp. 52-59, Hyyderabad, India
T. V. Ra amachandra (*,+ +), Uttam Kuma ar (*), S.N. Prasa ad (#) &Vaishna av B(*)
FIGURE ES Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6
TABLES Table 1 Table 2 Table 3 Table 4 Table 5 Table 6 Table 7
Address for Corresponde ence : Dr. T.V. Ram machandra Energy & We etlands Research Grroup Centre for Eccological Sciences Indian Institu ute of Science, Bangalore – 560 012, INDIA
1 of 10
Figure 1 : GRDSS The e process of deve eloping the GUI also involved underrtaking the task off identifying the co ommands and imp plementation of th he interfacing code e. The layout of th he interface had to o be altered to suiit the end use er's perspective. Figure F 2 shows the e structural diagra am of the GRDSS S graphical user in nterface.
Fully structured s and matured open source e spatial and temp poral analysis tecchnology seems to o be the officia al carrier of the future for planning of o the natural reso ources especially in the developing g nations. This te echnology has gained enormous mo omentum becausse of technical sup periority, affordability and ability y to join expertise e from all sections of the society. Su ustainable develop pment of a region n depends on the in ntegrated planning g approaches ado opted in decision m making which req uires timely and a accurate spatial data. With the increased develop pmental programm mes, the need for appropriate decission supportTel : Abstract A 91-80 0-23600985/2293 32506/ system has s increased in ord der to analyse and d visualise the deccisions associated d with spatial and temp poral aspects of na atural resources. In I this regard Geo ographic Informatiion System (GIS) along with229330 099 remo ote sensing data support s the applica ations that involve e spatial and temp poral analysis on d digital Fax :thematic maps an nd the remotely se ensed images. Op pen source GIS w would help in wide scale applications 91-80 0-23601428/2360 00085/ involving de ecisions at variou us hierarchical leve els (for example ffrom village panch hayat to planning 2360 00683[CES-TVR]c commission) on ec conomic viability, social acceptance e apart from techn nical feasibility. GRASS (Geographic Resources s Analysis Supporrt System, http://w wgbis.ces.iisc.erne et.in/grass) is an o open E-ma ail :source GIS tha at works on Linux platform (freewarre), but most of the applications are e in command line e cestv
[email protected], nece essitating a user frriendly and cost e effective graphical user interface (G GUI). cestv
[email protected]. Keeping the ese aspects in min nd, Geographic R Resources Decisio on Support System m (GRDSS) has been iisc.e ernet.indeveloped with functionality such as raster, to opological vector, image processing g, statistical analyysis, energ
[email protected] analysis, a graphics production, etc. T This operates thro ough a GUI develo oped in Tcltk (Too ol comm mand language / Tool T kit) under Lin nux as well as with h a shell in X-Win dows. GRDSS incclude options URL :such as Import /E Export of differentt data formats, Dissplay, Digital Imag ge processing, Ma ap editing, Rasterr http:///ces.iisc.ernet.inA Analysis, Vector Analysis, A Point Ana alysis, Spatial Qu ery, which are req quired for regiona al planning such /enerrgy/index.htmas watershed w Analysis, Landscape Ana alysis etc. This is customised to Ind dian context with a an option to extra act individual band d from the IRS (Ind dian Remote Sen sing Satellites) da ata, which is in BIL (Band Interlleaved by Lines) format. f The integrration of PostgreS SQL (a freeware) i n GRDSS aids ass an efficient
11/12
10 06 Geographic c Resources De ecision Support System S – an Op pen Source GIS S
htttp://wgbis.ces.iiisc.ernet.in/ene ergy/paper/Geos spatial_to
database d management system. In ntroduction GIS (G Geographic Inform mation System) is a system of hard ware, software an nd procedures to ffacilitate the management, maniipulation and anallyses of spatial-te mporal data. Its a application is wide e ranging from micro m level to mac cro level planning.. However this bo undless capabiliti es are limited by o ones ability tovisu ualise its implicatio ons. It has become an important co omponent in regio nal planning with a pivotal role in n major decisions related to earth re esources. This inccludes urban spra awl analysis involvving growth and migration m of popula ation, location of industries, employyment activities, b asic infrastructure e (water, electricity, etc.), transpo ortation (railway and a road networkss), drainage syste m (river basin qua ality, etc.), p etc. town planning, With the t growing incide ence of ecological and environmenttal impacts, GIS h has been used to analyse the impac ct of human activitties on the ecosys stems. This is posssible due to the o options to display and analys se changes in the e geo-referenced temporal t and spattial data. This help ps to solve complex problems regard ding planning and d management of resources and alsso biophysical mo odeling. The data input includ des spatial, statistiical and thematic data derived from m a combination off existing maps, a aerial photographs, interpreta ation of remotely sensed s images an nd secondary data a from the government agenc cies (census data,, land use data, ettc.). Digital image e processing techn nique aids in resto oration and rectific cation of data, seg gmentation of data and also in visu ualisation. The inte egration of spatiall and tempo oral technology he elp in addressing challenges c faced by the environme ent. The integratio on of GIS with re emote sensing da ata offers enhance ed capability for in nventorying, mapp ping, monitoring a and modeling to und derstand many en nvironmental proce esses. This integrration also helps i n using remotely ssensed image e to update the GIS and also maps.. The GIS thematiic data and attribu utes are used to g guide image classification. GRASS, a GIS based open source software e distributed unde er the GNU GPL (G General Public c License) integrates these two tech hnologies resultin g in numerous ad dvantages. GRAS SS was originally developed d by the U.S. Army Consttruction Engineeri ng Research Laboratories, a branc ch of the US Army Corp of Engineerrs, as a tool for lan nd management a and environmenta al planning by the e military. GRASS S is a raster/vectorr GIS, image proccessing system, an nd graphics produ uction system m (http://wgbis.ce es.iisc.ernet.in/grass). GRASS conta ains over 350 pro ograms and tools tto render maps and images on monitor m and paper; manipulate raste er, vector, and site es data; process m multi spectrral image data; an nd create, manage e, and store spatia al data. GRASS h has the options to interface with printers, p plotters, digitizers, d and data abases to develop p new data as we ll as manage exissting data. GRAS SS uses both intuiitive windows interface as well as ccommand line syn tax for ease of op perations and is s supported by developers and users worldwide. Num merous sub modu ules existed in GR RASS but most of them are in com mmand line argum ments. The comma and line syntax fo or GRASS was cu umbersome and time consuming for individuals, who were not expose d to computer pro ogramming and diid not posse ess any programm ming skills. In orde er to overcome thiss difficulty, it was the necessity for an econo omically viable, tec chnically superiorr and a more flexib ble spatial and tem mporal analyses ttools that led to the development of a user friendly gra aphical user interfa ace (GRDSS) for GRASS GIS alon ng with the database component as a depicted in figure 1.
e 2: GRDSS - GUII structural diagra m Figure Meth hodology
The current devellopment in GRDSS in the allied tecchnologies of remo ote sensing and G GIS is to build an integrated spatial decision support system (SDSS). R Recent advancess in the developme ent of graphical user interface bas sed programming have enabled rap pid prototyping, te esting, and applica ation development. Varrious GIS package es such as MapIn nfo (http://mapinfo o.com), Idrisi (http://www.idrisi.c clarku.edu), Geom media (http:// w ww w.intergraph.com) etc were compa aratively studied as per the set criteria a listed in table 1.
2 of 10
11/12
10 07 Geographic c Resources De ecision Support System S – an Op pen Source GIS S
htttp://wgbis.ces.iiisc.ernet.in/ene ergy/paper/Geos spatial_to
The T limitations of the t commercial G IS softwares were e also considered while implementiing the GRDSS to o offer optimal performance. The fo ollowing feature ch haracterises GRD DSS development. uirements - The G GRDSS GUI sourcces can be compiled on all PC's run nning LINUX. It i. System requ is a full system with X Window grraphical user interrface. Minimum P PC can be 486 (or better), with "LINUX". The LinuxL system with h a three-button m mouse, 32 MB RAM M (64 MB recomm mended) and a hard disk spa ace of 250MB (besside DOS/Window ws partition). GRD DSS binaries require only 45-50 MB an nd to compile add ditional 150 MB dissk space is requirred (to store the sources). The GRDSS sou urce code can be ccompiled using a GNU "C" compile er. ii. User friendly interface and me enu organisation - The user interfacce is designed to ssuit to the users ose who are not m much exposed to p programming. The e interface is custo omized and in India and tho simplified comp parable to propriettary softwares in tthe market. The m menu is organised sequentially in a user-friendly environment tha at satisfies both sspecific and comm mon users. iii. Functional ca apabilities - GRDS SS is equipped w with the functional ccapabilities such a as data import/export in n different formatss (including the IR S data format), m map analysis, data modeling, buffer generatio on, spatial and no on spatial statistica al calculation, digiital image processsing (spatial, spectral and tem mporal data analyysis), raster, vecto or and point analyysis and query, ma ap calculator, terrain modeling g, erosion modelin ng, report genera tion and can also handle spatial an nd non-spatial attribute informatiion), etc. database (both spatial data and a s – GRDSS exten nds all the help faccilities present in GRASS. All the commands with iv. Help facilities their functions and a options are prresent in help to g get started with the GRDSS. There is a mailing list and mailing list archive where e users can post ttheir doubts and ccan discuss throug gh e-mails worldwide. The help facilities and dthe documents a available on the G GRASS official web bsite provide enough informa ation to start with tthe software. v. Error explana ation and manual support - GRDSS S is equipped with h error explanation n, which automatically re enders error messsages through the e mail facility in Lin nux. Beside manyy sample data sets are availab ble on the GRASS S official website, including online tu utorials and manu uals.
3 of 10
11/12
10 08 Geographic c Resources De ecision Support System S – an Op pen Source GIS S
htttp://wgbis.ces.iiisc.ernet.in/ene ergy/paper/Geos spatial_to
Overview O and des scription of GRDS SS GRDSS G supports all a basic features required by any G GIS package. i. It supports a wide range of datta Import/Export tthat are also supp ported by other sta andard GIS including IRS (LIISS3 and PAN) da ata extraction and d DTEM/DTED (Digital Terrain Elevvation Model / Digital Terrain Elevation E Data) exxtraction. ii. It has capabilitties for displaying the raster/vector images. GRDSS can further intera act with commercial plottters and printers a and is capable of ccolor managemen nt. iii. GRDSS has the image processsing capabilities (rrectification, enha ancement, transforrmation, classification, da ata analysis, etc.). iv. GRDSS has the t capability of m manipulating raste r, vector and poin nt data. This includ des digitization, topology creation n, map overlay, arrithmetic operatio ns, DEM generatiion, decision supp port, distance, area an nd statistical repo ort, etc. v. It can perform wetland analysis , terrain modeling g with slope, aspect, relief, profile, ccost between a terrain da ata, runoff-modelin two locations to analyze ng, etc. vi. Thematic data a preparation, histtogram generation n, and charts of different formats. vii. Database (Po ostgreSQL) mana agement system. viii. Online tutoria als with help man uals and sample d datasets. Figure 4 depicts the GRD DSS hierarchical menu arrangeme ent.
4 of 10
11/12
10 09 Geographic c Resources De ecision Support System S – an Op pen Source GIS S
htttp://wgbis.ces.iiisc.ernet.in/ene ergy/paper/Geos spatial_to
Figure e 4: GRDSS hierarrchical menu arra ngement.Table 2 : Functions of GR RDSS modules
5 of 10
11/12
11 10 Geographic c Resources De ecision Support System – an Op pen Source GIS S
htttp://wgbis.ces.iiisc.ernet.in/ene ergy/paper/Geos spatial_to
Table 2: Function ns of GRDSS mod dules
6 of 10
11/12
11 11 Geographic c Resources De ecision Support System S – an Op pen Source GIS S
htttp://wgbis.ces.iiisc.ernet.in/ene ergy/paper/Geos spatial_to
Data abase connectivity y GRDSS uses PostgreSQL (a freew ware) as a databa ase for data storag ge, data retrieval, query and manipulation of both b spatial and atttribute information. It is the most ad dvanced open so urce database server used primarily for data stora age and retrieval, and complex data a analysis. PostgrreSQL is a sophisticated object-relational data abase manageme ent system (ORDB BMS). An ORDBM MS is an extension of the more m traditional re elational database e management syystems (RDBMS). For the advanced d developer, PostgreSQL even supp ports extensibility of o its data types a and procedural lan nguages. PostgreSQL has comprehensive SQL S support, referrential integrity, M MVCC or Multi- Ve rsion Concurrency Con ntrol to avoid unne ecessary locking that t increases the e reliability of the d database by logging changes before they are written w to the datab base. Applica ations Land L cover analysis: This has been done by computing both slope bassed and distance based vegetation v indices as the district had d varying extent of o vegetation cove ers. The result of la and cover analysis a is listed in n table 3. GRDS SS provide a supp port for spatial datta as well as attrib bute data by integrrating GIS, image e processing, GPS (globa al positioning systtem) and other tec chniques.
7 of 10
11/12
11 12 Geographic c Resources De ecision Support System S – an Op pen Source GIS S
htttp://wgbis.ces.iiisc.ernet.in/ene ergy/paper/Geos spatial_to
3: V Vegetation Indices sor data with Land use analysis: GRDSS through its capability c of analyzzing theTable statistica al, raster and vecto option ns for map overly, map algebra and d other operators a aid in analyzing th he landuse pattern n. Both superrvised and unsupe ervised classificatiion approaches w were tried to identiffy landuse catego ories in the distric ct using Gaussian maximum likeliho ood classifier (GM MLC). The level of accuracy in GML LC was 94.67% compared to unsupervised class sifier (78.07%). Th he composition off land use categorries (agric culture, forest, plan ntation, built-up an nd wasteland) are e listed in the table e 4.
Accuracy A Estimatiion
T Table 4: Land use details of Kolar district
Accuracy estimation in n terms of produce er's accuracy, use er's accuracy, ove rall accuracy and Kappa coeffiicient were calcula ated after generatting confusion ma atrix for supervised d classification (ta able 5) and unsup pervised classifica ation (table 6). Table 5: Error Matrix M Resulting fro om Classifying Trraining Set Pixels..
8 of 10
11/12
11 13 Geographic c Resources De ecision Support System S – an Op pen Source GIS S
htttp://wgbis.ces.iiisc.ernet.in/ene ergy/paper/Geos spatial_to
Table 6: Error E Matrix for Un nsupervised Classsification. The producer’s accuracy, a user’s ac ccuracy correspon nding to the vario us categories and d overall accuracyy results obtained are a summarized in n table 7.
Table T 7: Producerr’s accuracy, userr’s accuracy and o overall accuracy. A value was comp puted (0.931577) which w is as an ind dication that an ob bserved classification is 93 percent p better than one resulting fro om a chance. GRDSS G was also used to study the e watershed statuss in Kolar district a along with land usse and land coverr analyses. a The digital elevation mod del generated for tthe district is depiccted in figure 6. D Draping a land use e data d on the DEM enabled e to assess s the status of the e catchment of ma ajor water bodies. The temporal analysis a of Kolar district d provided a picture of land usse changes. The a analysis indicatess the unplanned developmental d activities without pro oper planning havve lead to the convversion of large sccale productive land to waste land d (43.75%). Watershed W misma anagement and co onversion of wate rbodies for variou us anthropogenic a activities (agriculture, ( buildings, etc.) were th he prime reasons ffor the depletion o of the water table along with increasing ecological problems suc ch as salinization, runoff, etc.
The land cover and land use an nalyses showed N NRVI, CTVI, PVI1 results are compa arable to GMLC in n entage area. The assessment a of the e watershed and h holistic approache es in landuse terms of perce planning consiidering the actual ground condition could help in enssuring the sustaina able developmentt. GRDSS a open sou urce GIS with cap pabilities such as rraster analysis, ve ector analysis, ima age processing, s is comparable to o any commerciall software available in the market map algebra and otther functionalities onomical (free wa fo or various applications including nattural resource ma nagement. An eco are) with user friendly GUI’s, GRD DSS is hoped to penetrate p in all secctions of decision making and contrribute to the Summary S su ustainable develop pment of India.
9 of 10
11/12
11 14 Geographicc Resources De ecision Support System – an Op pen Source GIS S
htttp://wgbis.ces.iisc.ernet.in/ene ergy/paper/Geosspatial_toda
R References : 1.GRASS official website :http:///grass.itc.it come.html 2. GRASS mirror site in India :http:://wgbis.ces.iisc.ernet.in/grass/welc 3. g list and mailing list archive :http:///wgbis.ces.iisc.ern net.in/grass/suppo ort.html GRASS mailing 4. GRASS softwa are download page :http://wgbis.cess.iisc.ernet.in/gras ss/download.html 5. Requirements SS GIS :http://wgbis.ces.iisc.ernet.in n/grass/grass5/source to compile GRAS /REQUIREMEN NTS.html 6. Mapinfo :http p://www.mapinfo.ccom/location/integ gration 7. Idrisi :http://w www.idrisi.clarku.e edu 8. Geomedia :h http://www.intergraph.com/dynamiccdefault.asp
E-mail E | Sahyadri| EN NVIS | Enerrgy | GRAS SS | CES | IISc | E-ma ail
http p://www.gisdevvelopment.net/m magazine/globa al/2008/decem mber/50.htm
11 15
Eng gaging Web for better adminisstration
D Directory A AudioCast Applications A C Careers C Companies D Downloads Education Events G Development GIS G Glossary G Guest Book H History In nterviews N News Policy Proceedings Professionals Publications Technology Thesis Tutorials
Engaging g Web for bettter administtration Home > GIS Development > The Global Geospatial Maga azine > 2008 > December H Home | About Us | Site Map
E Ezine I want to subscribe: GIS Weekly ( GIS Weekly Archive )
Dr S Narendra Pra asad Senior Principal Scie entist, Salim Ali Centre for Ornitholo ogy and Natural History, Deccan Reg gional Station, Hyderabad and Sec cretary, OSGeo snare
[email protected] om
Sub bscribe Newsletter | Search Site
Santosh Gaikwad S IT Officer , Salim Ali Centre f for O Ornithology and Natura al H History D Deccan Regional Station n, H Hyderabad santosh
[email protected]
Dr KS Rajan Associate Profess sor, International Institute of Inforrmation Technology Hyderabad and Treasurer, T OSGeo raja
[email protected]
Anno ouncement Page 1 of 1
India has been n witnessing trem mendous growth,, the like of which h has not been o observed since Independence e. The rapid pace e of developmentt has impacted p practically every a aspect of environ nment and the comm mon man. erative to put in place p effective go overnance system ms with major em mphasis on It has, therefore, become impe ansparency and at a the same time e being able to un nderstand the pro oblems and efficiency, acccessibility and tra issues faced by b the society. W Whether it is delive ery of proper hea alth care to the ccitizens, access to t safe drinking water, social in nfrastructure like e schools and oth her quality of life (QoL) concerns of the residents,, the knowledge and understan nding of geography or 'location' plays p an importan nt role in making the right decisions by the respective age encies or departm ments of the govvernment machin nery. The governance systems, therrefore, would req quire accurate an nd timely informa ation and data, which w is location ure. specific in natu A WEB GIS FOR RAJAHMUN NDRY o provide a spatia al data infrastruccture (SDI) for implementing e-go overnance. A GIS is an indisspensable tool to GIS has been developed for th he Rajahmundry Parliamentary Constituency, C Easst Godavari distrrict, Andhra d rural regions, in ntegrates data fro om multiple sourcces - remote Pradesh, whicch focuses on both the urban and sensing image ery, GPS surveyss and field studie es - and brings evverything togethe er on to an Open n Source GIS platform for the development of o a Spatial Decission Support Sysstem for civil and d public administrration from a eb enabled GIS. A Web GIS has been created bo oth in English as well as in the loccal language, desktop to We Telugu. nique in that multtiple stakeholderrs were involved at various stage es of developmen nt of the GIS. The work is un First, Memberr of Parliament V Aruna Kumar re eadily saw the po ossibilities and po otential of develo oping such a system for the e benefit of comm mon man and fun nded the project through t the MPL LADS scheme. The collector of o East Godavarii district, in turn, appreciated the merits of implem menting such a prroject and facilitated its administration a an nd implementation. The OSGeo In ndia chapter end dorsed the projec ct and the Salim Ali Centre for Ornithology and Natural History executed e the pro oject. M.Sc stude ents of Adi Kavi Nannaya N ajahmundry, University, Ra were trained to o carry out the project work in pa art fulfillment of th heir course requirements.
Search h
G
A PIGGYBACK RIDE ON GOO OGLE MAPS AP PIS nterface with embedded, detailed d street and Google Maps provides a highlyy responsive, intuitive mapping in M provides not only the map, ssatellite image orr a hybrid of both h but also a aerial imageryy data. Google Maps range of opera ations on the map including zoom ming, panning, information pop-up ps and overlays. Google Maps API provides an a interface into these operationss through JavaSccript objects. The e GIS application n for Rajahmundry parliamentary co onstituency has been b developed using Google Ma aps API. A spatial datab base was develo oped for it in PostgreSQL/ PostGIS database. The e advan- GIS DE EVELOPMENT 50 Open Sourrce Engaging We eb for better adm ministration DECE EMB E R 200 8 D Data flow using Google G Maps API tage of cre eating spatial database for Rajah hmundry was havving all the data in a central database including well defined privileges which makes m it possible e to extend the m more standard SQ QL queries with spatial s queries.
11 16 Engaging W Web for better administration
http:///www.gisdevellopment.net/ma agazine/global//2008/dece
DATAS SETS USED FOR R RAJAHMUNDR RY PARLIAMENTARY CONSTIT TUENCY The following shape file es were used whiile developing the e spatial databasse for Rajahmundry Parliamentarry uency. Constitu Point da ata Bank an nd ATM centres, cemeteries, buss stations, police stations, places of worship, cine ema halls, clubs, comme ercial complexes, community centtres, customer ca are centres, e-Se eva centres, edu ucational centres , electric sub-stations. n data Polygon Rajahm mundry parliamen ntary constituenccy boundary, slums in Rajahmun ndry parliamentarry constituency
. nvolved in develo oping the Web-G GIS application Steps in The following are prereq quisites that ensu ured the develop pment of Web-GIIS application forr Rajahmundry parliamentary constituency. pache Web serve er running PHP and a PostgreSQL//PostGIS • An Ap • Spatia al data in PostGIS S database for R Rajahmundry. • Popula ating the spatial data into Post- GIS G database • Outpu utting XML with PHP P • Generrating HTML pag ge for map visualisation The UR RL for Rajahmund dry parliamentary constituency W Web-GIS applicattion is http://www w.osgeo.in/google e /sample e. htm. The interfface which was developed d for Ra ajahmundry Parliamentary Constituency Web-GIS S applicattion allows users s to query againsst spatial data available in the PosstgreSQL/Post- G GIS database. FEATURES e spatial data • Intuitivve user interface for querying the Genera al public is familia ar with Google Maps M interface and its basic navigation. In this, we e added a simple query tool. In Figure 1, the ressults have been sshowed by the qu ng fair price (FP) shops within the e uery slums havin of 100 meters. A user can click on the FP shops marker m icons which will pop up an n info window to show radius o the info ormation about th he particular FP shop. s User can even e Advanta ages • Free ((except for developer time) • Quick development tim me - depending on o complexity of application eight application,, client side scrip pting • End product is light we n • Intuitivve user interface - general public already familiar with Google Maps interface and basic navigation • Fast, g good response tiime • Google provides solid background servvices (satellite da ata, roads, trafficc data, street view w, geocoding) g selected GIS da ata - not every mapping m applicatiion requires multtiple, complex ma ap • Effective for displaying layers es for learning, multitude m of samp ples and tutorials • Great on-line resource or focussed appllications • Best fo • Best fo or small municipalities, companie es (not huge data asets) Limitatio ons • Limite ed functionality co ompared to some e commercial pro oducts • Difficu ult to overlay morre complex GIS data d • Difficu ult to overlay multiple GIS layers
Engaging Web for better administration
http:///www.gisdevellopment.net/ma agazine/global//2008/dece
• Can nnot
read directlyy from GIS datab base, must conve ert to other forma ats (XML, KML) (but ( parts of proccess can be b automated with scripting)
get the info ormation about slums s by clicking on slums polygo on. The interface e allows the user to query slums n not having FP shops within the e radius of 100 meters m to know w which slums don'tt have the FP sho ops nearby. Figu ure 2 aving FP shops within w the radius of 100 meters. shows the result for the query slums not ha
• Light we eight Web-GIS application a This applic cation gets loade ed faster on the browser b and even n the querying tim me is less. Results too get displa ayed faster.
• Google provides p backgrround data such h as satellite datta, roads etc. • This app plication is usefu ul for common public, p policy ma akers, decision makers, m govern nment officials etc. A simple WebW GIS applica ation showing point of interests in n the Rajahmund dry parliamentaryy constituency wa as developed to get the inform mation about poin nt of interests such as parks and gardens, Banks and ATM centre es, e URL for it http:/// www.osgeo.in//google. overlay. htm. In this Police stattions etc (Figure 3). Following the application n, point of interessts can be overla aid by checking th he respective point of interest che eck boxes. The Rajahmundry parliame entary constituen ncy Web-GIS is also a available in the regional lang guage Telugu (Figure 4). Following the URL for Telugu T version off Rajahmundry parliamentary p con nstituency Web-G GIS application w.osgeo.in/google/sample_ te.htm m http://www.osg geo.in/ google /ovverlay_te.htm. http:// www The Indic IME I for Telugu software from ww ww.bhashaindia. ccom was used to o type the data in n Telugu. The PostgreSQ QL database wass encoded to UTF F-8 and then the e data was loaded d into PostgreSQ QL database.
THE WAY Y FORWARD We increas singly see the usse of Open Sourcce tools by a larg ge number of sta akeholders in virtu ually all thematicc areas of co oncern. Combine ed with the powe er of the use of In ndic languages as a preferred me edium, the Web GIS G will emerge e as one of the m most potent toolss for societal benefit.
ACKNOW WLEDGEMENTS S It is a pleasure to thank Vu undavalli Aruna Kumar, K MP for making this small sscale experimen nt a success, the nd particularly Drr PS Roy, V Ravvi Kumar, Dr Hanumantha Rao, Ramamurthy, R Sinha, Open Sourrce geospatial an Aneel Kum mar, Dr Sahu for their active involvement and help p.
Contactt US | Advertise with us Page 1 of 1 This site is bes st experienced with Inte ernet Explorer 4.0 and above, at default brows ser settings 1024 X 76 68 pixels B Broken links? Problems with site? Send email to t info@gisdevelopm ment.net ©GISdevelop pment.net. All rights s reserved.
Spatio‐temporal landscape modelling for natural hazard vulnerability analysis in select watersheds of Central Western Ghats T. V. Ramachandra, Senior Member, IEEE, AninditaDasgupta,Uttam Kumar, Student Member, IEEE, Bharath H Aithal, Student Member, IEEE, P. G. Diwakar and N. V. Joshi Abstract—Natural hazards such as landslides are triggered by numerous factors such as ground movements, rock falls, slope failure, debris flows, slope instability, etc. Changes in slope stability happen due to human intervention, anthropogenic activities, change in soil structure, loss or absence of vegetation (changes in land cover), etc. Loss of vegetation happens when the forest is fragmented due to anthropogenic activities. Hence land cover mapping with forest fragmentation can provide vital information for visualising the regions that require immediate attention from slope stability aspects. The main objective of this paper is to understand the rate of change in forest landscape from 1973 to 2004 through multi-sensor remote sensing data analysis. The forest fragmentation index presented here is based on temporal land use information and forest fragmentation model, in which the forest pixels are classified as patch, transitional, edge, perforated, and interior, that give a measure of forest continuity. The analysis carried out for five prominent watersheds of Uttara Kannada district– Aganashini, Bedthi, Kali, Sharavathi and Venkatpura revealed that interior forest is continuously decreasing while patch, transitional, edge and perforated forest show increasing trend. The effect of forest fragmentation on landslide occurrence was visualised by overlaying the landslide occurrence points on classified image and forest fragmentation map. The increasing patch and transitional forest on hill slopes are the areas prone to landslides, evident from the field verification, indicating that deforestation is a major triggering factor for landslides. This emphasises the need for immediate conservation measures for sustainable management of the landscape. Quantifying and describing land use - land cover change and fragmentation is crucial for assessing the effect of land management policies and environmental protection decisions. Index Terms—Forest fragmentation, landslide, SVM, MLC
INTRODUCTION Natural hazards such as landslides involving small to large ground movements are mainly triggered due to unstable slopes with scanty green cover. Unstable sloped are induced due to the removal of vegetation cover, rock falls, deep failure of slopes, shallow debris flows, ground water pressure, erosion, soil nutrients, soil structure, etc. The actual landslide often requires a trigger before being released and in most cases, a change in land cover (LC) due to the loss of vegetation is a primary factor that builds up specific sub-surface conditions for landslide to occur. The loss in forest cover due to LC change has increased rapidly in recent times due to increasing mankind needs. Forest cover is reduced to almost half of the ecologically desired amount [1] and about 72 percent of India’s forests have lost their viability for regeneration, with forest grazing being one of the most important causes [2]. Forest fragmentation apart from affecting the biodiversity and ecology of the region has a significant influence in the movement of soil (silt) and debris in undulating terrains with high intensity rainfall. Forest fragmentation analysis spatially aids in visualising the regions that require immediate attention to minimise natural calamities such as landslides. Spatial fragmentation map depicts the type and extent of fragmentation derived from land use (LU) data which are obtained from multi-source, multi-sensor, multi-temporal, multi-frequency or multi-polarization remote sensing (RS) data. The objectives of this paper are i.) Classification of multi-temporal RS data using Maximum Likelihood classifier to obtain LU map. ii.) Multi-temporal forest fragmentation analysis for five watersheds in Uttara Kannada to characterise the type and extent of fragmentation or loss of vegetation cover. iii.) Visualising the consequences of fragmentation for landslide susceptibility.
Methods Maximum Likelihood classifier (MLC) – Supervised classification of the image was performed using MLC. MLC has become popular and widespread in RS because of its robustness [3– 6]. It quantitatively evaluates both the variance and covariance of the category spectral response
pattern [7] assuming the distribution of data points to be Gaussian [8] which is described by the mean vector and the covariance matrix. The statistical probability of a given pixel value being a member of a particular class is computed and the pixel is assigned to the most likely class (highest probability value). If the training data pertaining to different classes contain n samples and the samples in each class are i.i.d. (independent and identically distributed) random variables and further if we assume that the spectral classes for an image is represented by ωn, n=1,…, N, where N is the total number of classes, then probability density p(ωn|x) gives the likelihood that the pixel x belongs to class ωn where x is a column vector of the observed digital number of the pixels. It describes the pixel as a point in multispectral space (M-dimensional space, where M is the number of spectral bands). The maximum likelihood (ML) parameters are estimated from representative i.i.d. samples. Classification is performed according to
x ω n if p(n | x) p( j | x) j n
(1)
i.e., the pixel x belongs to class ωnif p(ωn|x) is largest. The ML decision rule is based on a normalised estimate of the probability density function (p.d.f.) of each class. The discriminant function, gln(x) for ωn in MLC is expressed as
gln (x) = p(x|ωn )p(ωn ) (2) where p(ωn) is the prior probability of ωn, p(x|ωn) is the p.d.f. for pixel vector x conditioned on ωn[6]. Pixel vector x is assigned to the class for which gln(x) is greatest. In an operational context, the logarithm form of (2) is used, and after the constants are eliminated, the discriminant function for ωn is stated as G ln (x ) ( x n )T n1 ( x n ) ln | n | 2 ln P(n )
(3)
where n is the variance-covariance matrix of ωn, μn is the mean vector of ωn. A pixel is assigned to the class with the lowest G ln (x ) [6, 9-10].For each LU class (agricultural land, human settlement / residential / commercial areas, roads, forest, plantation, waste land / open land and water bodies) training samples were collected representing approximately 10% of the study area. With these 10% known pixel labels from training data, the aim was to assign labels to all the remaining pixels in the image.
Forest fragmentation: Forest fragmentation is the process whereby a large, continuous area of forest is both reduced in area and divided into two or more fragments. The decline in the size of the forest and the increasing isolation between the two remnant patches of the forest has been the major cause of declining biodiversity [11-15]. The primary concern is direct loss of forest area, and all disturbed forests are subject to “edge effects” of one kind or another. Forest fragmentation is of additional concern, insofar as the edge effect is mitigated by the residual spatial pattern [16-19].
LU map indicate only the location and type of forest, and further analysis is needed to quantify the forest fragmentation. Total extent of forest and its occurrence as adjacent pixels, fixed-area windows surrounding each forest pixel is used for calculating type of fragmentation. The result is stored at the location of the centre pixel. Thus, a pixel value in the derived map refers to between-pixel fragmentation around the corresponding forest location. As an example [20], if Pf is the proportion of pixels in the window that are forested and Pff is the proportion of all adjacent (cardinal directions only) pixel pairs that include at least one forest pixel, for which both pixels are forested. Pff estimates the conditional probability that, given a pixel of forest, its neighbour is also forest. The six fragmentation model that identifies six fragmentation categories are: (1) interior, for which Pf = 1.0; (2), patch, Pf < 0.4; (3) transtitional, 0.4 < Pf < 0.6; (4) edge, Pf > 0.6 and Pf-Pff> 0; (5) perforated, Pf
> 0.6 andd Pf-Pff< 0, and a (6) undettermined, Pf > 0.6 and Pff = Pff. Wheen Pff is larger than Pf, thhe implicatioon is that forrest is clumpeed; the probaability that ann immediate neighbour iss also forest is greater thaan the averagge probabilityy of forest within the windoow. Converseely, when Pff is smaller thaan Pf, the im mplication is that t whateverr is non-foresst is clumpedd. The differeence (Pf-Pff) characterises a gradient ffrom forest cllumping (edgge) to non-forrest clumpingg (perforated). When Pff = Pf, the moddel cannot disstinguish foreest or non-forrest clumping. The case off Pf = 1 (interior) representts a completely forested w window for which w Pff mustt be 1.
Suupport Vectoor Machine (S SVM): SVM are a supervisedd learning alggorithms baseed on statisticcal learning ttheory, whichh are consideered to be heuuristic algoritthms [21]. SV VM map input vectors to a higher dim mensional spaace where a m maximal sepaarating hyper plane is consstructed. Twoo parallel hyper planes aree constructed d on each sidee of the hypeer plane that separates the data. The seeparating hyper plane maxximises the distance d betweeen the two parallel p hyperr planes. An assumption a iss made that thhe larger the margin or distance betweeen these paraallel hyper plaanes, the betteer the generallisation error of T model prooduced by suppport vector classification c only dependss on a subset of o the classiffier will be. The the traininng data, becauuse the cost function f for building b the m model does noot take into acccount traininng points thaat lie beyond the margin [21]. When it i is not posssible to definee the hyper plane p by linear equations, the data cann be mapped into a higherr dimensionall space througgh some nonllinear mappinng functions.. A free and open source software – openModelleer [22] was used for prediccting the probbable landslidde areas usinng SVM. openModeller (hhttp://openmodeller.sourceforge.net/) inncludes facilitties for readinng landslide occurrence and a environm mental data, seelection of ennvironmental layers on wh hich the moddel should bee based, creatiing a fundam mental niche model m and proojecting the m model into an environmenttal scenario aas shown in Fig. F 1.
Study arrea and Daata The Uttaraa Kannada disttrict lies 74°9' to 75°10' E lonngitude and 133°55' to 15°31' N latitude, ex xtending over aan area of 10,,291 km2 in thee mid-western ppart of Karnataaka state (Fig. 2). It accountss for 5.37 % off the total area of the state w with a populatio on above 1.2 million m [23]. Thhis region has gentle undulatting hills, rising steeply from ma narrow coaastal strip bord dering the Arabbian sea to a plateau p at an altitude a of 500 m with occasiional hills risinng above 600––860m.
Fig. 1. Methhodology used for llandslide predictioon in openModelleer.
Fig. 2.Utttara Kannadaa district, Karrnataka, Indiaa.
– coastal landds This distrrict, with 11 taluks, can bbe broadly caategorised intoo three distinnct regions –– (Karwar, Ankola, Kum mta, Honnavaar and Bhatkkaltaluks), moostly forestedd Sahyadrian interior (Suppa, Yellapur, Sirsi and Siiddapurtalukss) and the eaastern marginn where the table land beegins (Haliyaal, Yellapur and Mundgoodtaluks). Clim matic conditiions range frrom arid to humid h due to physiographhic conditionss ranging from m plains, mouuntains to coaast.
Survey off India (SOI) toposheets off 1:50000 andd 1:250000 sccales were used to generatte base layers – district annd taluk bounndaries, waterr bodies, draainage networrk, etc. Field data were co ollected with a handheld GPS. RS dataa used in the study were Landsat L MSS (1973, ( spatiall resolution – 79m), Landssat TM (19899, spatial reso olution – 30m m), Landsat ETM+ E (2000,, spatial resollution – 30m m) [downloadeed from Gloobal Land Coover Facilityy, http://www w.landcover.org] and LISS-III Multi-sspectral (20004, spatial reesolution – 23.5m) proocured from m NRSC, H Hyderabad, Inndia. Googlle Earth daata (http://earrth.google.com m) served in pre p and post classification c process and vvalidation of the results.
Environm mental data suuch as precippitation of wettest w monthh were downnloaded from WorldClim – Global Cllimate Data [hhttp://www.w worldclim.org/bioclim]. Other environm mental layers (Aspect, ( DEM M, Flow acccumulation, Flow F directioon, Slope, Compound C Toopographic IIndex) used for modellinng landslide were obtaineed from USGS S Earth Resoources Observvation and Science (EROS S) Center baseed Hydro1Kddatabase [http:///eros.usgs.gov v/#/Find_Datta/ Products__and_Data_Available/gtopo30/hydro/assia]. The globbal LC change maps were obtained from Land Coveer Land Cover Changge Global Faacility, [http://glccf.umiacs.umdd.edu/servicess/landcoverchhange/landcovver.shtml; http://www w.landcover.oorg/services/llandcover chhange/landcovver.shtml]. Thhe spatial resolution of aall
the data w were 1 km. 1225 landslide occurrence o pooints of low, m medium and hhigh intensityy were recordeed using hanndheld GPS frrom the field and a publishedd reports.
Results an nd Discusssion a the imagess were resamplled to a commoon RS data were geometrrically correcteed on a pixel by pixel basis and L mapping (F Fig. 3) was donne using normaalised difference resolution of 30m with a dimension of 7562 x 6790. LC vegetation index (NDVI)) given as
NIR-Redd NIR+Reed (4) NDVI valuues range from m -1 to +1; incrreasing values from 0 indicaate presence off vegetation (aggriculture, foreest and plantattion) and negattive values inddicate absence of greenery (builtup, sand, faallow, water) as a given in Tabble I.
Fig. 3.NDVI and NDWI N for Uttaara Kannada. TABLE I: NDVII AND WDVI INDEX
OF UTTARA KANNA ADA.
NDVI
W WDVI
VEGETATION (%) (
NON-VEGETATION E (%)
WATER (%)
NON-WATER (%) (
1973
97.82
2.18
1.75
98.25
1989
96.13
3.87
2.20
97.80
2000
94.33
5.67
2.80
97.20
2004
94.00
6.00
2.50
97.50
d using norm malised differeence water indeex (NDWI) [244] given as Mapping oof water bodiess (Fig. 3) was done Green-NIR R Green+NIIR
(5) NDWI valuues above 0 indicate presencee of water bodies and values below 0 indicaate other classees.
Training pixels were collected froom the false colour compposite of the respective bands (for tim me period 19973, 1989, annd 2000) sincce historical data were unnavailable. For 2004 dataa, training daata uniformlyy distributed over the studdy area colleected with prre calibrated GPS were used. u The claass spectral chharacteristics for six LU categories (agrriculture, builltup / settlemeent, forest (ev vergreen, sem mievergreenn, deciduous),, plantation, waste w land / fallow / sandd, and water bodies / streaams) using RS R data weree obtained to assess a their innter-class sepparability. andd the images were classifieed using MLC C. Temporall classified im mages (into 6 LC classes – agriculturee, built-up, foorest, plantation, waste lannd and waterr bodies) are shown s in Figg. 4 and the sttatistics are giiven in Tablee II. This was validated with the repressentative fielld data (covering ~ 10% % of the studdy area) andd also using Google Earrth image.Prooducer’s, userr’s, overall acccuracy and Kappa K values ccomputed aree listed in Tabble III. wn in Fig. 5 annd Further, fivve watersheds were delineateed from the claassified images of four time period as show the statistics are listed inn Table IV. Forest fragmenttation model was w used to obbtain fragmenttation indices as Fig. 6 and the statistics s are prresented in Tabble V. Forest was w categorisedd into patch, trransitional, edgge, shown in F perforated,, and interior tyypes using proggrams in GRA ASS GIS (http://wgbis.ces.iiscc.ernet.in/foss)..
Fig. 4. Tempporal classifiedd image of Uttaara Kannada diistrict. TABLE II: I LU DETAILS OF O UTTARA KANN NADA DISTRICT
YEAR CLASS
(HA)
19733
1989
20000
2004
AREA A
AREA
AREA A
AREA
(%)
((Ha)
(%))
(Ha)
(%)
(H Ha)
(%)
AG GRICULTURE
25794
2.51
5 57465
5.59
97749
9.51
1154464
11.23
BUILTUP
4490
0.44
8 8671
0.84
13674
1.33
275577
2.68
FOREST
897420
87.29
8445665
82.255
749877
72.94
6274455
61.02
PLLANTATION
82439
8.02
8 85190
8.29
122160
11.88
184340
17.93
WASTELANDLAND
-
-
8 8483
0.83
15884
1.54
234456
2.28
WATER A BODIES
17995
1.75
2 22666
2.20
28796
2.80
499924
4.86
TOTAL
1028151
100
10028151
100
1028151
100
10288151
100
Fig. 5. Temporal LU U image of the five rive basinns of Uttara Kannnada district.
TABLE III: ACCURACY ASSESSMENT OF THE CLASSIFIED IMAGES YEAR
1973
1989
CATEGORY
AGRICULTURE
USER’S ACCURACY 74.12
PRODUCER’S ACCURACY 77.97
BUILTUP
68.45
65.42
FOREST
87.23
85.22
PLANTATION
81.27
77.23
WASTELAND
85.62
81.21
WATER BODIEE
78.50
74.29
AGRICULTURE
70.43
80.39
BUILTUP
78.78
81.11
FOREST
73.96
73.39
PLANTATION
79.62
77.19
WASTELAND
80.21
77.78
WATER BODIEE
31.02
64.82
AGRICULTURE
81.16
80.47
BUILTUP
84.29
87.50
FOREST
87.15
82.54
PLANTATION
86.51
87.23
WASTELAND
85.22
81.79
WATER BODIEE
86.40
97.56
AGRICULTURE
85.21
84.54
BUILTUP
86.47
83.11
FOREST
94.73
96.20
PLANTATION
92.27
91.73
WASTELAND
88.49
87.88
WATER BODIEE
83.13
81.33
2000
2004
OVERALL ACCURACY
KAPPA
76.89
71.92
75.78
69.82
83.21
79.21
87.83
83.11
It is to be noted that Supa dam was constructed in late 1970s. Hence this water body is absent in the 1973 Landsat multi-spectral classified image and prominent in the classified images of 1989, 2000 and 2004 (Fig. 4). Also, there was no waste land in 1973 as per the statistics in Table II. All the rivers are perennial since the catchment areas are mainly composed of interior forest. However, eventually there are drastic land cover changes consequent to unplanned developmental activities. In addition, the dams of major rivers in the district have inundated large vegetation areas. These areas have silt deposit at the river beds and have been classified as sand / waste land. Forest and plantation were considered as a single class - forest and all other classes were considered as nonforest as the extent of LC is a decisive factor in landslides. While the forest fragmentation map produced
valuable information, it also helped to visualise the state of forest for tracking the trends and to identify the areas where forest restoration might prove appropriate to reduce the impact of forest fragmentation. Forest fragmentation also depends on the scale of analysis (window size) and various consequences of increasing the window size are reported in [25]. The measurements are also sensitive to pixel size.Nepstad et al., (1999a, 1999b)[26 and 27]reported higher fragmentation when using finer grain maps over a fixed extent (window size) of tropical rain forest. Finer grain maps identify more non-forest area where forest cover is dominant but not exclusive.
Forest fragmentation (in %)
The criterion for interior forest is more difficult to satisfy over larger areas. Although knowledge of the feasible parameter space is not critical, there are geometric constraints [28]. For example, it is not possible to obtain a low value of Pff when Pf is large. Percolation theory applies strictly to maps resulting from random processes; hence, the critical values of Pf (0.4 and 0.6) are only approximate and may vary with actual pattern. As a practical matter, when Pf > 0.6, non-forest types generally appeared as “islands” on a forest background, and when Pf < 0.4, forests appeared as “islands” on a non-forest background. Fig. 6 shows the temporal change in forest patch type, revealing that patch, transitional and edge forest are increasing due to deforestation and interior forest is decreasing. The statistics in Table V shows the decrease in forest patch types from 1973 to 2004. Forest fragmentation is a vital indicator that is accountable for environmental changes. With the expansion of human settlement there is higher risk of forest degradation disturbing the biodiversity, affecting the water quality, endangering wildlife survival, habitat protection, etc. Uttara Kannada 100 80 60 40 20 0
Patch 1973 Transitional 1989Edge
2000 Peforated
2004 Interior
Forest fragmentation (in %)
Aganashini 100 80 60 40 20 0 Patch
Transitional
1973
Edge
1989
2000
Peforated
Interior
2004
Forest fragmentation (in %)
Bedthi 100 80 60 40 20 0 Patch
Transitional
1973
1989
Edge
2000
Peforated
2004
Interior
Forest fragmentation (in %)
Kali 100 80 60 40 20 0 Patch
Transitional
1973
Edge
1989
Peforated
2000
Interior
2004
Forest fragmentation (in %)
Sharavathi 100 80 60 40 20 0 Patch
Transitional
1973
Edge
1989
2000
Peforated
Interior
2004
Forest fragmentation (in %)
Venkatpura 100 80 60 40 20 0 Patch
Transitional
1973
1989
Edge
2000
Peforated
Interior
2004
Fig. 6.Temporal Forest fragmentation change of the five river basins of Uttara Kannada district.
Fig. 6.Forest fragm mentation in fivve river basins in Uttara Kannnada district. TABLE IV: LU DETAILS OF TH HE FIVE RIVER BAS SINS (IN HA AND %) % YEAR R
CATEGOR RY
1886.6
1.4
BUILTUPP
627.6
FOREST T
123500
PLANTATION WASTELAN ND
19899
20000
20044
BEDTHI E
KALI
3616.442
1.3
0.5
799.883
90
2513663
7930.9
5.8
-
-
WATER BOD DIES
2662.5
URE AGRICULTU
SH HARAVATHI
14294
3.8
13302
1.6
0.3
13188.7
0.4
1033.976
92
3201174
86
71536
163155.9
6
31593
8.4
-
-
-
2
1556..7
0.6
66966.8
4466
3.3
125500.9
4.6
BUILTUPP
883.9
0.7
856.99
FOREST T
118724
87
2398994
PLANTATION
8217.7
6.0
WASTELAN ND
784.9
WATER BOD DIES
3531.6
AGRICULTU URE
6179.3
4.5
BUILTUPP
969.7
0.7
FOREST T
11130
81
PLANTATION
12004.6
8.8
WASTELAN ND
1571.8
1.1
WATER BOD DIES
4580.5
3.4
AGRICULTU URE
13028
9.5
364144
BUILTUPP
2706.5
1.9
52211
URE AGRICULTU
19733
AGANA ASHINI
VENKATPURA 13 355.2
5.9
0.1
74.69 7
0.3
88
18282
81
29900
7.3
2289 2
10
-
-
-
-
1.8
23329
2.9
615.9 6
2.7
19667
5.3
41110
5.1
15 529.7
6.8
0.3
28866.6
0.8
784
1.0
404.4 4
1.8
87
3037758
81
63686
78
16725
74
17783
6.5
34152
9.1
88807
11
3165.8
14
0.6
1360..8
0.5
16966.5
0.5
729
1
508
2.3
2.6
1205..8
0.4
11916
3.2
30055
3.8
284
1.3
29413
11
334421
8.9
47781
5.9
1881.3
8.3
3838..4
1.4
43664
1.2
4114.8
0.5
717.6 7
3.2
2185441
80
2655508
71
59379
73
15363
68
163799
6
50542
14
12936
16
35 519.7
16
39788
1.5
27888.6
0.8
12258
1.6
82 22.65
3.7
1501..8
0.6
174453
4.7
24404
3
312.9 3
1.4
13
36575
9.8
37778.9
4.6
14 489.9
6.6
1.9
73644.2
1.9
17223.2
2.1
95 59.98
4.3 58
-
FOREST T
84127.8
61
1816774
66
2320097
62
47020
58
13048
PLANTATION
27836.2
20
35315
13
67093
18
23472
29
5896 5
26
WASTELAN ND
1847.4
1.4
54111
2
10332
2.8
17220.5
2.1
381.97
1.7
WATER BOD DIES
7064.4
5.2
96033
3.5
20549
5.5
34884.7
4.3
83 36.26
3.7
TABLE V: FORESTFRGMENTATION DETAILS OF THE FIVE RIVER BASINS (IN HA) YEAR
1973
2000
2004
CATEGORY
AGANASHINI
BEDTHI
KALI
SHARAVATHI
VENKATPURA
PATCH
0.59
0.16
1.49
-
0.11
TRANSITIONAL EDGE
189.17 1177.55
322.04 2398.25
1050.52 4442.65
94.99 729.81
85.40 347.26
PERFORATED INTERIOR
787.33 127895.85
1478.46 261524.81
3588.78 337778.54
457.62 74929.99
293.29 19506.20
PATCH TRANSITIONAL
83.69 631.03
217.57 1443.65
333.17 1995.02
94.67 670.22
38.33 215.23
EDGE PERFORATED INTERIOR
3638.08 857.74 120188.05
6198.87 1907.86 246239.30
6753.60 2526.22 322328.53
3155.90 839.13 66064.81
889.215 269.26 18178.51
PATCH TRANSITIONAL EDGE PERFORATED INTERIOR
213.07 1201.70
1326.61 4622.81
1637.77 5654.96
198.690 919.972
95.52 390.37
7263.70 1169.60 112006.30
18204.16 3506.29 205359.76
20319.45 3977.67 280116.92
4262.67 770.327 64838.71
1507.74 302.99 16249.80
PATCH TRANSITIONAL EDGE PERFORATED INTERIOR
363.82
947.32
1117.11
144.83
40.424
2106.73 11098.47 2199.95 92213.32
4207.65 17250.44 3942.96 186500.95
4414.24 17408.67 3120.97 267026.02
781.60 4189.86 717.20 62076.267
243.40 1274.87 275.79 16337.97
Precipitatioons of wettest month along with the seveen other layerss (as mentioned in section IIII) were used to predict lanndslides using SVM as shoown in Fig. 7. The landslidde occurrence points were overlaid on thhe probabilityy map to validaate the predictioon as shown inn Fig. 8.
Fig. 7..Probability distribbution of the landsslide prone areas.
obability distributioon of the landslidee prone areas by ovverlaying landslidee occurrence pointts. Fig. 8.Vaalidation of the pro
The SVM M map was 966% accurate with respect to the grounnd and Kappaa values 0.87733 and 0.90883 respectiveely [29]. Mostt of the prediccted landslidee areas as probable landslidde prone zonees (indicated in
red in Fig. 7) reside on terrain that is highly undulating with steep slopes and are frequently exposed to landslides induced by rainfall. Maximum number of landslide points occurring in the undulating terrain, collected from the ground fell in fragmented areas - patch, transitional and edge forest. Fig. 9 shows the landslide occurrence points obtained from ground using handheld GPS overlaid on the classified image of the district (A) and forest fragmentation map (B) based on 2004 RS data. This map indicates that hill slopes with undulating terrain and less vegetation cover (patch, transitional and edge forest) are more susceptible to landslides compared to north-eastern part of the district which has relatively flat terrain with large area utilised under agricultural practices.
The present study makes an effort to quantify forest cover by using fragmentation index for measuring the disturbances due to human impact giving a complete picture of the change that has occurred. The five prominent watersheds of Uttara Kannada district were assessed for quantifying forest fragmentation caused due to anthropogenic disturbances. Forest fragmentation model is useful to visualize state of forest fragmentation for an area and identifying areas where forest restoration might prove appropriate. If the vegetative areas have been cleared, the water retaining capacity of the soil has decreased, triggering landslides in those areas. Hence, there is a immediate need to restore those vegetative by forestation to ensure that the soil is retained on the hill slopes and do not activate any downward movement of the hill tops.
Natural disasters have drastically increased over the last decades. National, state and local government including NGOs are concerned with the loss of human life and damage to property caused by natural disasters. The trend of increasing incidences of landslides occurrence is expected to continue in the next decades due to urbanisation, continued anthropogenic activities, deforestation in the name of development and increased regional precipitation in landslide-prone areas due to changing climatic patterns [30].
Conclusion Landslides occur when masses of rock, earth or debris move down a slope. Mudslides, debris flows or mudflows, are common type of fast-moving landslides that tend to flow in channels. These are caused by disturbances in the natural stability of a slope, which are triggered by high intensity rains. The primary criteria that influence landslides are precipitation intensity, slope, soil type, elevation, vegetation cover and LC type. In this paper, LU analyses along with forest fragmentation were carried out for five perennial watersheds of Uttara Kannada district. SVM was used to generate probable landslide map. By overlaying the fragmentation map on the landslide probable map, past and future occurrences of landslides can be visualised. It is also evident that degradation of forest and LC change is an important factor that is not only responsible for triggering landslides, but also a major contributor to global warming, climate change, natural resource depletion and consequent detrimental effect on our environment.
Fig. 9. Landslide suscep ptibility points oveerlaid on the classified image of 20044 (A) and forest frragmentation map (B). Black points represent the t occurrence of llandslide points obbtained using handdheld GPS.
VI. ACKNO OWLEDGMENT This researrch was carriedd out with the financial assisstance from Kaarnataka Biodiiversity Board and ISRO -IIS Sc Space Tecchnology Celll (initial workk during 20077-2010). The environmentall layers were obtained froom WorldClim m - Global Clim mate Data. NR RSA, Hyderabaad provided thhe LISS IV data used for landd cover analysiis. We thank USGS Earth Resources Observation and Science (ERO OS) Center forr providing thee environmenttal layers andd Global Land Cover Facilityy (GLCF) for facilitating thhe Landsat imaages and Landd Cover Changge product.
REFE ERENCES [1] [2]
G. B. M Maithani, “Non-foreestry technology for fo forest based waasteland developmeent,” The Indian F Forester, vol. 116, pp. 931-937, 19900. S. N. Rai R and A. Saxena, “The extent of foorest fire, grazing and regeneration status in inventorried forest areas in n India,” The Indiian Foresterr, vol. 123, pp. 6899-702, 1997.
[3]
A. H. S Strahler, “The use of prior probabiilities in maximum m likelihood classif ification of remoteely sensed data,” Remote Sensing of Environnment, vol. 10, pp. 135-163, 1980.
[4]
C. Coneese and F. Masellli, “Use of error matrices to improove area estimatees with maximum likelihood classifi fication proceduress,” Remotee Sensing of Enviroonment, vol. 40, pp. 113-124, 1992.
[5]
L. Ediriiwickrema and S. Khorram, K “Hierarrchical maximum-likelihood classification for improveed accuracies,” IE EEE Transactions on Geoscieence and Remote Sensing, S vol. 35, pp. 810-816, 1997.
[6]
M. Zhenng, Q. Cai and Z. Wang, “In: Effectt of prior probabillities on maximum m likelihood classiffier,” Geoscience and a Remote Sensiing Sympossium, 2005, IGAR RS’05, Proceedingss 2005 IEEE Internnational, vol. 6, ppp. 3753-3756, 20055.
[7]
T. M. L Lillesand and R. W. W Kiefer, “Remote Sensing and Im mage Interpretatioon,” Fourth Edition, John Wiley andd Sons, 2002, ISB BN 9971-511-427-3.
[8]
U. Bayaarsaikhan, B. Bolddgiv, K-R. Kim, K-A. K Park and D. Lee, L “Change detection and classificcation of land coveer at HustaiNationnal Park in Mongolia,” Intern national Journal off Applied Earth Obbservation and Geooinformation, vol. 11, pp. 273-280, 2009. 2
[9]
A. R. Joohn and J. Xiuping g, “Remote Sensingg Digital image Annalysis: An Introduuction,” Springer-V Verlag Inc., New York, Y 1999.
[10] R. O. D Duda, P. E. Hart annd D, G. Stork, “P Pattern classificatiion,” New York, A Wiley-Intersciennce Publication, Seecond Edition, 20000, ISBN 9814-12-602-0. d and the nineteenth cenntury world economy,” Duke Univerrsity Press, Durhaam, [11] R. P. Tuucker, and J. F. Riichards, “Global deforestation North C Carolina, USA, 198 83.
[12] B. L. Turner, W. C. Clark, R. W. Kates, J. F. Richards, J. T. Mathews, and W. B. Meyer, “The Earth as transformed by human action,” Cambridge University Press, Cambridge, UK, 1990. [13] W. B. Meyer, and B. L. Turner, “Changes in land use and land cover: a global perspective,” Cambridge University Press, Cambridge, UK, 1994. [14] R. A. Houghton, “The worldwide extent of land-use change,” BioScience, vol. 44, pp. 305-313, 1994. [15] J. D. Hurd, E. H. Wilson, S. G. Lammey, and D. L. Civco, “Characterization of Forest Fragmentation and Urban Sprawl using time sequential Landsat Imagery,” In ASPRS 2001 Annual Convention, St. Louis, MO, 2001. [16] R. T. T. Forman, and M. Godron, “Landscape Ecology,” John Wiley, New York, 1986. [17] M. G. Turner, “Landscape ecology: the effect of pattern on process,” Annual Review of Ecology and Systematics, vol. 20, pp. 171197, 89 [18] S. A. Levin, “The problem of pattern and scale in ecology,” Ecology, vol. 73, pp. 1943-1967, 1992. [19] T. V. Ramachandra, U. Kumar, P. G. Diwakar and N. V. Joshi, “Land Cover Assessment using À Trous Wavelet fusion and K-Nearest Neighbour classification,” In proceedings of the 25th Annual In-House Symposium on Space Science and Technology, ISRO IISc Space Technology Cell, Indian Institute of Science, Bangalore, India, 29-30 January, 2009. [20] K., J. Riitters, R. O' Neill. Wickham, B. Jones, and E. Smith, “Global-scale patterns of forest fragmentation,” Conservation Ecology, vol. 4(2-3), 2000. [21] T. Kavzoglu, I. Colkesen, “A kernel functions analysis for support vector machines for land cover classification,” International Journal of Applied Earth Observation and Geoinformation, vol 11, pp. 352-359, 2009. [22] M. E. S. Munoz, R. Giovanni, M. F. Siqueira, T. Sutton, P. Brewer, R. S. Pereira, D. A. L. Canhos, and V. P. Canhos, “openModellera generic approach to species’ potential distribution modeling,”Geoinformatica. DOI: 10.1007/s10707-009-0090-7. [23] T. V. Ramachandra and A. V. Nagarathna, “Energetics in paddy cultivation in Uttara Kannada district,” Energy Conservation and Management, vol. 42( 2), pp. 131-155, 2001. [24] S. K. McFeeters, “The use of normalized difference water index (NDWI) in the delineation of open water features,”International Journal of Remote Sensing, vol. 17(7), pp. 1425–1432, 1996. [25] K. J. Riitters, R. O' Neill. Wickham, B. Jones, and E. Smith, “Global-scale patterns of forest fragmentation,” Conservation Ecology, vol. 4(2-3), 2000. [26] D. C. Nepstad, P. Lefebvre, and E. A. Davidson, “Positive feedbacks in the fire dynamic of closed canopy tropical forests,”Science, vol. 284, pp. 1832-1835, 1999a. [27] D. C. Nepstad, A. Verissimo, A. Alencar, C. Nobre, E. Lima, P. Lefebvre, P. Schlesinger, C. Potter, P. Moutinho, E. Mendoza, M. Cochrane, and V. Brooks, “Large-scale impoverishment of Amazonian forests by logging and fire,” Nature, vol. 98, pp. 505-508, 1999b. [28] R. V. O’Neill, C. T. Hunsaker, S. P. Timmins, B. L. Jackson, K. B. Jones, K. H. Ritters, and J. D. Wickham, “Scale problems in reporting landscape pattern at the regional scale,” Landscape Ecology, vol. 11, pp. 169-180, 1996. [29] T. V. Ramachandra, U. Kumar, Bharath H. Aithal, P. G. Diwakar and N. V. Joshi, “Landslide Susceptible Locations in Western Ghats: Prediction through open Modeller,” In proceedings of the 26th Annual In-House Symposium on Space Science and Technology, ISRO-IISc Space Technology Cell, Indian Institute of Science, Bangalore, India, 28-29 January, 2010. [30] T. V. Ramachandra, M. D. Subashchandran and AnantHegdeAshisar, “Landslides at Karwar, October 2009: Causes and Remedial Measures,” ENVIS Technical Report: 32,Environmental Information System, Centre for Ecological Sciences, Bangalore, 2009.
Hybrid Bayesian Classifier for Improved Classification Accuracy Uttam Kumar, S. Kumar Raja, C. Mukhopadhyay, and T. V. Ramachandra, Senior Member, IEEE Abstract— Widely used Bayesian classifier is based on the assumption of equal prior probabilities for all the classes. However, inclusion of equal prior probabilities may not guarantee high classification accuracy for the individual classes. Here, we propose a novel technique - Hybrid Bayesian Classifier (HBC), where the class prior probabilities are determined by unmixing a supplement low spatial-high spectral resolution multispectral (MS) data that are assigned to every pixel in a high spatial-low spectral resolution MS data in Bayesian classification. This is demonstrated with two separate experiments – firstly, class abundances are estimated per pixel by unmixing MODIS data to be used as prior probabilities while posterior probabilities are determined from the training data obtained from ground. These have been used for classifying the IRS LISS-III MS data through Bayesian classifier. In the second experiment, abundances obtained by unmixing Landsat ETM+ are used as priors and posterior are determined from the ground data to classify IKONOS MS images through Bayesian classifier. The results indicated that HBC systematically exploited the information from two image sources improving the overall accuracy of LISS-III MS classification by 6% and IKONOS MS classification by 9%. Inclusion of prior probabilities increased the average producer’s and user’s accuracy by 5.5% and 6.5% in case of LISSIII MS with 6 classes and 12.5% and 5.4% in IKONOS MS for the 5 classes considered.
Index Terms— prior probability, unmixing, Bayesian classifier
INTRODUCTION MAGE classification using conventional Bayesian classifier is a supervised method based on the prior Iprobabilities [1]. Prior probability resolves confusions among classes that are not well separable[2]and is therefore effective in improving classification accuracy. Bayesian classifier is generally used with the assumption that prior probabilities are equal, asreliable prior probabilities are not easily available. However, with the use of equal prior probabilities, the performance of the classifier is not optimal [3] which is evident from the misclassified pixels primarily due to spectral confusion between classes, apart from the increased computing and sampling requirements[4]. The theoretical analysis of the effect of prior probability has been discussed in detail by Z. Mingguo et al., (2009) [5]. Earlier works used prior probabilities based on previous year crop statistics[6], geographical data [7],elevation data[8] and spatial characteristics specified through a Markov random field model at reference resolution [9], improving the overall accuracy and Kappa coefficient. Therefore, it is desirable to obtain reliable prior probabilities for each spectral class and use them to classify the pixels that are likely to misclassify[1], even though they are difficult to determine within the same time period and for the same spectral classes. In this context, a new classification method - Hybrid Bayesian classifier (HBC)based on linear unmixing and Bayesian classifier is proposed to assign a class label to each pixel in a high spatial-low spectral resolution (HS-LSR) multispectral (MS) data. The prior probabilities of the different classes in the Bayesian classifier to classify each pixel in the HS-LSR data such as IRS LISS-III MS and IKONOS MS are obtained from abundance estimates by unmixing low spatial-high spectral resolution (LS-HSR) MS data such as MODIS and Landsat ETM+ respectively. The terms HS-LSR and LS-HSR have been used in a relative sense here, depending upon the spatial resolution of the images. For example, in the first experiment, MODIS bands are referred as LS-HSR and IRS LISS-III MS bands are HS-LSR data, while in the second experiment, Landsat ETM+ bands are LS-HSR and IKONOS MS bands are HS-LSR data.
The reaso on for selectin ng MODIS an nd Landsat ETM+ E imagess as LS-HSR R supplement data is becauuse of their econ nomic viabilitty and high temporal t reso olution that ennables their pprocurement ffor any part oof the globe, thrroughout the year y correspo onding to any HS-LSR dataa. However, tthe techniquee in general, ccan be applied on n any other LS-HSR L (such as hyperspectral bands)) and HS-LSR R (such as m multispectral bbands) data classsification. Th he novelty of this approacch lies in the fact that low w spatial-highh spectral andd low spectral-h high spatial reesolution MS data are com mbined to impprove the classsification results which ccan be thought of as a fusion process, p in thee sense, that information i fr from two
Fig. 1.Hybriid Bayesian Claassifier.
different sources (sensors) are com mbined to arrrive at a im mproved classsified image by systemattically exploiting g the relevantt information from both th he sources as shown in Figg. 1. Zhukov et al., (1999) [10] attempted d a similar multisensormu m ultiresolution n image fusioon based onn low spatial resolution ((LSR) unmixing and high spaatial resolutio on (HSR) classsification. Hoowever, the eend product w was not a classsified image, bu ut a set of MS fused imagess equivalent to t the numberr of LSR bandds at the HSR R image dimennsion. In the folllowing section, linear un nmixing using g Orthogonall Subspace P Projection (O OSP) and Bayyesian classifier are reviewed d. The proposed HBC is diiscussed in S ection III whhile section IV V demonstrates the experimen ntal results an nd validation followed by conclusion c inn Section V. Revieew of Method ds Linear Un nmixing:With h K spectral bands and M classes (C C1,..,CM), eacch pixel has an associateed K' dimension nal pixel vector y (y1 ,..., y K ) whose componentts are the graay values corrresponding tto the
x1 column vvector K spectra al bands. Let L E [e1 ,...,e M ] , wheere, for m 1,..., M , e m is a Kx representting the endm member specttral signaturee of the mth ttarget materiial. For a pixxel, let m ddenote the fractiion of the mth target ma aterial signa ature, and α ( 1 ,..., M ) ' denote th he M-dimenssional abundancce column vector. The lineear mixture model m for y is given by
y Eα η where,
η = (1 ,..., K )' ,
(1) aree i.i.d. N(0, 2 ) [11]. Eq quation (1) reepresents a sttandard signaal detection m model
where Eα OSP detects oone signal (tarrget) at a tim me, we α is a desiredd signal vector to be detected. Since, O divide a set s of M targ gets into desiired (C m ) an nd undesired (C1 ,...,C m 1 ,,C m 1 ,...C M ) targets. A loogical approach is to eliminatte the effects of undesired d targets that aare considereed as “interferrers” to C m bbefore the detecttion of C m taakes place. Now, N in order to find m , the desired ttarget materiaal is em . Thee term
Eα in (1) can be rewrritten to separate the desireed spectral siggnature em :
y em m Rr η
(2)
where r ( 1 ,... m1 , m 1 ,..., M ) and R [e1 ,...,e m 1 ,e m1 ,...,e M ] . Thus the interfering signatures in
R can be removed by the operator, P (I - R(R R)-1 R T )
(3) which is used to project y into a space orthogonal to the space spanned by the interfering spectral signatures, where I is the KxK identity matrix [12]. Operating on y with P , and noting that PR 0 ,
Py Pem m Pη .
(4)
After maximizing the SNR, an optimal estimate of m is
yT P T Py emT P T Pem
m =
(5)
Note that 1 m 0 and thus ( 1 ,..., M ) can be taken as proportional probabilities of the M classes i.e. P(C m ) m . Bayesian classifier: Associated with any pixel, there is an observation y . With M classes (C1 ,...,C M ) , ~
Bayesian classifier calculates the posterior probability of each class conditioned on y [13]: P(Cm |y )= ~
P( y|C m )P(Cm )
(6)
~
P( y ) ~
In (6), since P( y ) is constant for all classes, only P( y|C m )P(C m ) is considered. P(y | C m ) is computed ~
~
~
K
assuming class conditional independence, so, P(y | C m ) is given by P( y|C m )= P(y k |C m ) ~
~
(7)
k=1
New Hybrid Bayesian Classifier (HBC)
HBC uses the abundance of each class obtained from LS-HSR data by linear unmixing as prior probability while classifying the HS-LSR data using Bayesian classifier of the same geographical area and time frame. That is, given the observation vector y for a pixel, it is classified to fall in class lif ~
l = Arg Max P( y | Cm )P(C m ) , where P( y|C m ) is as in (7) calculated using the HS-LSR data and m
~
~
P(C m ) α m calculated using the LS-HSR data. The assumptions in this method are (i) if there are r HSLSR pixels contained in one LS-HSR pixel, i.e. resolution ratio is (r:1), the prior probabilities for all the r HS-LSR pixels are equal corresponding to the same LS-HSR pixel, and (ii) the two data types have a common origin or upper left corner, i.e. the edges of the r x r HS-LSR pixels overlaps exactly with the corresponding LS-HSR pixel. The limitations are (i) (K-1) should be ≥ M in LS-HSR data and (ii) M in HS-LSR should be ≤ M in LS-HSR data.
Experimental Results
Two separate experiments were carried out. In the first case, LISS-III MS data (3 bands of 23.5m x 23.5m spatial resolution resampled to 25m, acquired on December 25, 2002) of 5320 x 5460 size and MODIS 8day composite data (7 bands of 250m x 250m, acquired from 19-26 December, 2002) of 532 x 546 dimensionwere co-registered with known ground control points (RMSE - 0.11). Training data were collected from the ground representing approximately 10% of the study area covering the entire spectral gradient of the classes. Separate test data were collected for validation. LISS-III MS classified image using conventional Bayesian classifier is shown in Fig. 2 (a). Assuming that there are six (fixed) number of representative endmembers (pure pixels), the entire image was modeled in terms of those spectral components, extracted using N-FINDR algorithm [14] from MODIS images. In the absence of pure pixels, alternative algorithms [15] can be used for endmember extraction, which is a limitation of NFINDR. It may be noted that some objects (for example buildings with concrete roofs, tiled roofs, asphalt, etc.) exhibit high degrees of spectral heterogeneity representing variable endmembers. This intra-class spectral variation with variable endmembers can be addressed through techniques discussed in [16-18], and is beyond the scope of the current study.
Abundance values were estimated for each pixel through unmixing and used as prior probabilities in HBC to classify LISS-III MS data (Fig. 2 (b)). Table I is the class statistics and Table II indicates the producer’s and user’s accuracies. The overall accuracy and Kappa for HBC (93.54%, 0.91) is higher than Bayesian classifier (87.55%, 0.85). For any particular class, if the reference data has more pixels with correct label, the producer’s accuracy is higher and if the pixels with the incorrect label in classification result is less, its user’s accuracy is higher [5]. Bayesian classifier wrongly classified many pixels belonging to waste / barren / fallow as builtup. Forest class was over-estimated and plantation was underestimated by Bayesian classifier. The only minority class in the study area is water bodies. Classified image using conventional Bayesian classifier had 1.08% (8854 ha) of water bodies in the study area. After the ground visit, we found that there are not many water bodies and the extents of most individuals were < 2000m2. Therefore, only a few water bodies that had spatial extent ≥ 62500 m2 could be used as endmembers. The minimum detected water class was 5% using unmixing of LS-HSR data. It may have happened that the prior probability was unlikely for this class while classifying the LISS-III MS data using HBC. The classified image obtained from HBC showed 0.81% (66.20 ha) of water bodies. A few pixels were wrongly classified using HBC, therefore, the producer’s accuracy decreased from 90.91 (in conventional Bayesian classifier) to 88.18% (in HBC). However, given the same set of training pixels for classification, the user’s accuracy has increased from 88.89 to 97%. Producer’s accuracy increased for agriculture (2.6%), builtup (4.3%), forest (7%), plantation (11.5%) and waste land (10.6%) and user’s accuracy increased for agriculture (8%),
M classified images: i (a) B ayesian classifier, (b) HBC C. Fig. 2. LISS-III MS TABLE I CLASS STATISSTICS FROM BAYESIAN A AND H HBC FOR LISS--III DATA Classsifiers →
Bayesian cllassifier
HFC
Classs ↓
Haa
%
Ha
%
Agricculture
155 5, 451 1
19.04
1142931
17.511
Builttup
139 9, 759 9
17.12
779280
9.71
Foresst
93,, 241 1
11.42
555721
6.83
Planttation
89,49 93
10.96
1176132
21.588
Wastte land
329 9, 473 3
40.36
3355587
43.566
Wateer bodiees
8854 4
1.08
66.20
0.81
TABLE II ACCURACY C ASSESSMENT FOR L ISS-III DATA Classifiers →
HBC
Bayeesian classsifier
Class ↓
PA*
UA*
Agricultture
87.54
87.47
PA A* 90.155
UA** ↑
95.56
↑
Builtup
85.11
81.68
89.39
↑
98.33
↑
Forest
85.71
88.73
92.61
↑
96.36
↑
Plantation
84.44
91.73
95.95
↑
91.03
↓
Waste land
88.03
90.37
98.67
↑
89.66
↓
Water bodies
90.91
88.89
88.18
↓
97.00
↑
Average
86.96
88.15
92.49
↑
94.66
↑
* PA – Producer’s Accuracy; UA – User’s Accuracy.
builtup (16.6%), forest (7.6%) and water bodies (8%) in HBC output. On the other hand, producer’s accuracy decreased (2.7%) for water bodies and user’s accuracy decreased (~0.7%) for plantation and waste land classes in agreement with the similar observations reported in [5]. HBC was intended to improve classification accuracies by correctly classifying pixels which were likely to be misclassified by Bayesian classifier. Therefore a cross comparison of the two classified images located the pixels that were assigned different class labels at the same location. These wrongly classified pixels when validated with ground data revealed a 6% (~1742832 pixels) improvement in classification by HBC. In the second experiment, IKONOS MS data (4 bands of 4m, acquired on November 24, 2004) of 700 x 700 size and Landsat ETM+ data (6 bands excluding Thermal and Panchromatic, of 30m, acquired on November 22, 2004) of 100 x 100 dimension were co-registered (RMSE - 0.09). Landsat pixels were resampled to 28m so that 49 IKONOS pixels would fit in 1 Landsat pixel. The scenes correspond to Bangalore city, India near the central business district having race course, bus stand, railway lines, parks, builtup with concrete roofs, asbestos roof, blue plastic roofs, coal tarred roads with flyovers and a few open areas (play ground, walk ways, vacant land, etc.). Class proportions from Landsat image pixels were used as prior probabilities to classify the IKONOS MS images using HBC (Fig. 3 (b)). The overall accuracy and Kappa for HBC (89.53%, 0.87) is higher than Bayesian classifier (80.46%, 0.69). Producer’s accuracy increased by 17% for concrete, asbestos, blue plastic roof and open area and decreased by 7% for vegetation in HBC (Table IV). User’s accuracy increased by ~5.5% for all the classes in HBC. Bayesian classifier wrongly classified and overestimated many pixels belonging to open area as asbestos roof (Table III). Vegetation was overestimated by 9% using Bayesian classifier and open area was underestimated by ~15%. A cross comparison of the two classified images showed that 95733 pixels (0.33% of the study area) were differently classified by the two classifiers. Validation of these pixels showed an improvement of 9% by HBC, higher than accuracies reported in [6, 7 and 8].
Fig. 3. IKONOS S classified im mages: (a) Bayyesian classifi fier, (b) HBC. Conclusion C n of this techn nique lies in the t fact that aabundance esttimates from L LS-HSR dataa were The majorr contribution utilized ass prior probaabilities to claassify HS-LSR R data using a Bayesian cclassifier impproving the ooverall accuracy by 6% and 9% with IRS S LISS-III MS M and IKON NOS MS dataa respectivelyy, as comparred to conventio onal Bayesian n classifier, deemonstrating the t robustnesss of the approoach. REFERENCES [31] J. Ediriiwickrema and S. Khorram, “Hieraarchical Maximum m-Likelihood Classsification for Imp mproved Accuraciees,” IEEE Trans. Geosci. Remote Sens., vol. 35, no.. 4, pp. 810-816, 1997. M and M. A. Friedl, “Using priior probabilities in n decision-tree claassification of rem motely sensed data,,” Remote Sens. E Environ., [32] D. K. McIver vol. 81, pp. 253-261, 2002 2. a D. A. Landgreebe, “Fast likelihoo od classification,” IEEE Trans. Geossci. Remote Sens.,, vol. 29, no. 4,pp. 509–517, 1991. [33] C. Lee and [34] L. Pedro oni, “Improved claassification of Lan ndsat Thematic Maapper data using m modified prior probbabilities in large aand complex landsscapes,” Int. J. Remote R Sens., vol.2 24, no. 1, pp. 91-113, 2003. [35] Z. Ming gguo, C. Qiangguo and Q. Mingzho ou, “The Effect of Prior Probabilitiies in the Maximuum Likelihood Classification on Inddividual Classes: A Theoretical Reeasoning and Empirical Testing,” Ph hotogrammEng. Reemote Sens., vol. 775, no. 9, pp. 11099-1116, 2009. TABLE III CLASS STATISSTICS FOR IKONO OS DATA
Classiffiers → Class ↓
Baayesian classifier
HFC
Ha H
%
Ha
%
Concreete roofs
346 6.67
44.34
356.18
45.556
Asbesto os roofs
47.99
7.41
6.79
0.877
Blue pllastic roof
5.83
0.75
0.21
0.033
Vegetaation
329 9.72
42.18
260.01
33.226
Open area a
41.60
5.32
158.58
20.228
TABLE IV ACCURACY ASSE ESSMENT FOR IKO ONOS DATA Classifierrs →
Bay yesian claassifier
HFC
Class ↓
PA
UA
PA
UA
Concrete roofs
69.99
84.01
76.49
↑
93.89
↑
Asbestos roofs
84.77
87.77
91.89
↑
94.46
↑
Vegetation
94.21
87.55
87.24
↓
89.13
↑
Blue plastic roof
84.33
81.17
97.00
↑
85.60
↑
Open area
51.49
69.49
95.00
↑
74.22
↑
Average
76.96
81.99
89.52
↑
87.46
↑
[36] L. L. F. Janssen and H. Middelkoop, “Knowledge-based crop classification of a Landsat Thematic Mapper image,” Int. J. Remote Sensing, vol. 13, pp. 2827–2837, 1992. [37] H. Cetin, T. A. Warner and D. W. Levandowski, “Data classification, visualization, and enhancement using n-dimensional probability density functions (NPDF): AVIRIS, TIMS, TM, and geophysical applications,” Photogramm. Eng. Remote Sens., vol. 59, pp. 1755–1764, 1993. [38] A. H. Strahler, “The use of prior probabilities in maximum likelihood classification of remotely sensed data,” Remote Sens. Environ., vol. 10, pp. 135–163, 1980. [39] G. Storvik, R. Fjortoft, and A. H. S. Solberg, “A Bayesian Approach to Classification of Multiresolution Remote Sensing Data,” IEEE Trans. Geosci. Remote Sens., vol. 43, no. 3, pp. 539–547, 2005. [40] B. Zhukov, D. Oertel, F. Lanzl, and G. Reinhackel, “Unmixing-based MultisensorMultiresolution Image Fusion” IEEE Trans. Geosci. Remote Sens., vol. 37, no. 3, pp. 1212–1226, 1999. [41] U. Kumar, N. Kerle, and T. V. Ramachandra, “Constrained linear spectral unmixing technique for regional land cover mapping using MODIS data,” In: Innovations and advanced techniques in systems, computing sciences and software engineering/ed by KhaledElleithy, Berlin: Springer, pp. 87-95, 2008. [42] C-I. Chang, “Orthogonal Subspace Projection (OSP) Revisited: A Comprehensive Study and Analysis,” IEEE Trans. Geosci. Remote Sens., vol. 43, no. 3, pp. 502–518, 2005. [43] J. Han and M. Kamber, Data Mining: Concepts and Techniques, Morgan Kaufmann Publishers, San Francisco, CA, USA, 2003. [44] M. E. Winter, “N-Findr: an algorithm for fast autonomous spectral end-member determination in hyperspectral data,” In Proceedings of the SPIE: Imaging Spectrometry, vol. 3753, pp. 266-275, 1999. [45] A. Plaza, P. Martinez, R. Perez, and J. Plaza, “A Quantitative and Comparative Analysis of Endmember Extraction Algorithms FromHyperspectral Data” IEEE Trans. Geosci. Remote Sens., vol. 42, no. 3, pp. 650–663, 2004. [46] C. A. Bateson, G. P. Asner, and C. A. Wessman, “Endmember Bundles: A New Approach to Incorporating Endmember Variability into Spectral Mixture Analysis” IEEE Trans. Geosci. Remote Sens., vol. 38, no. 2, pp. 1083–1094, 2000. [47] C. Song, “Spectral mixture analysis for sub-pixle vegetation fractions in the urban environment: How to incorporate endmember variability?” Remote Sens. Environment, vol. 95, pp. 248–263, 2005. [48] G. M. Foody, and H. T. X. Doan, “Variability in Soft Classification Prediction and its Implications for Sub-pixel Scale Change Detection and Super Resolution Mapping” Photogrammetric Engineering & Remote Sensing, vol. 73, no. 8, pp. 923–933, 2007.
FOSS @ IISc
http://wgbis.ces.iiscernet.in/foss
Dr. T. V. Ramachandra, Geoinformatics lab, Centre for Infrastructure, Sustainable Transportation and Urban Planning (CiSTUP) Indian Institute of Science, Bangalore 560012 Tel: 91-080- 22933099 / 22932786 / 2293 2506 / 23600985 (Extn: 215/232) E mail:
[email protected] ,
[email protected],
