Luswishi Farmblock Land Cassifciation

  • 246 views
Uploaded on

 

More in: Education
  • Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
    Be the first to comment
    Be the first to like this
No Downloads

Views

Total Views
246
On Slideshare
0
From Embeds
0
Number of Embeds
0

Actions

Shares
Downloads
16
Comments
0
Likes
0

Embeds 0

No embeds

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
    No notes for slide

Transcript

  • 1. Free Open Source Software (FOSS) Geographic Resource AnalysisSupport System (GRASS GIS 6.4) for Mapping Land use, land cover dynamics, Luswishi Farm Block, Lufwanyama District By Charles Bwalya Chisanga July 2012
  • 2. Table of ContentsList of Figures ...................................................................................................................................... iiList of Tables .......................................................................................................................................iiiAbstract ............................................................................................................................................... iv1.0 Introduction ................................................................................................................................ 1 1.1 Goal ........................................................................................................................................ 2 1.2 Purpose................................................................................................................................... 2 1.3 Objectives .............................................................................................................................. 22.0 Description of the Study Site ...................................................................................................... 23.0 Data Collection ........................................................................................................................... 43.1 Satellite images data acquisition................................................................................................. 44.0 Methodology............................................................................................................................... 45.1 RESULTS AND DISCUSSIONS ............................................................................................... 55.1 Change detection techniques ...................................................................................................... 55.2 Image Processing and Classification .......................................................................................... 65.3 Unsupervised classification ........................................................................................................ 7 5.3.1 Set region setting to red image: ......................................................................................... 7 5.3.2 Creation of group and subgroup ........................................................................................ 7 5.3.3 Clustering of image data and Generation of unsupervised statistics................................. 7 5.3.3 Unsupervised classification ............................................................................................... 8 5.4.1 Normalized Difference Vegetative Index (NDVI) .......................................................... 10 5.4.2 Image Ratioing ................................................................................................................ 11 5.4.3 Ratio transformations ...................................................................................................... 125.5 Tasseled Cap transformation (TC) ............................................................................................ 185.7 Principal Component Analysis (PCA) ...................................................................................... 205.8 Image Fusion ............................................................................................................................ 24 5.8.1 Image fusion with HSI transformation ............................................................................ 24 5.8.2 Brovey transformation methodology .............................................................................. 266.0 Conclusions .............................................................................................................................. 27References ............................................................................................................................................. i i
  • 3. List of FiguresFigure 1: Location of the Luswishi Farm Block .................................................................................. 3Figure 2: Copperbelt Province Boundaries .......................................................................................... 3Figure 3: False Colour Composite for 2001 and 2005 ......................................................................... 9Figure 4: 2001 land cover classification for 2001 (left) and 2005 (right)............................................ 9Figure 5: NDVI map for 2001 and 2005 ............................................................................................ 11Figure 6: Spectral Ratioing for NIR and Red for 2001 (Left) and 2005 (Right) ............................... 12Figure 7: Image Ratio TM3/TM4 for 2001 and 2005 ........................................................................ 13Figure 8: Image Ratio TM4/TM3 for 2001 and 2005 ........................................................................ 13Figure 9: Image Ratio Histograms for 2001 and 2005....................................................................... 14Figure 12: Image Ratio TM2/TM3 for 2001 and 2005 ...................................................................... 14Figure 10: Image Ratio TM3/TM2 for 2001 and 2005 ...................................................................... 15Figure 11: Image Ratio TM4/TM5 for 2001 and 2005 ...................................................................... 15Figure 12: Image Ratio TM5/TM4 for 2001 and 2005 ...................................................................... 16Figure 13: Image Ratio TM5/TM7 for 2001 and 2005 ...................................................................... 16Figure 14: Image Ratio TM3/TM5 for 2001 and 2005 ...................................................................... 17Figure 15: Image Ratio TM7/TM2 for 2001 and 2005 ...................................................................... 17Figure 16: Image Ratio Histograms for 2001 and 2005..................................................................... 18Figure 17: Tasseled Cap transformation for 2001 .............................................................................. 19Figure 18: Tasseled Cap transformation for 2001 .............................................................................. 20Figure 19: Change in Luswishi Farm Block from 2001 to 2005 overlaid with roads and rivers ...... 23Figure 20: Major changes (2001 to 2005) overlaid with roads and rivers ......................................... 24Figure 21: Image fusion with HSI transformation for Luswishi Farm Block 2001 (left); 2005 (right)at 14.25m ............................................................................................................................................ 26Figure 22: Image fusion with Brovey transform 2001 (left) and 2005 (right) ................................... 27 ii
  • 4. List of TablesTable 1: TM Band and their spanned wavelength ................................................................................ 4Table 2: RASTER MAP CATEGORY REPORT for 2001 and 2005 ................................................ 10Table 3: NDVI for typical cover types ............................................................................................... 10Table 4: NDVI for 2001 and 2005 ..................................................................................................... 11 iii
  • 5. AbstractThe world of GIS is always in development, and this fact is particularly true for Open Source GIS.Among the Geographic Information System (GIS) software, in the last years Geographic ResourceAnalysis Support System (GRASS) GIS has known wide popularity as it is well documented.GRASS GIS was developed by the United States Army Construction Research EngineeringLaboratories (USA CERL) as a tool for military land management and environmental planning. Inthe early nineties its licenses switched to GPL and by 1998 a GNU/Linux release was freelyavailable. GRASS has evolved into one of the most comprehensive, general purpose open sourcegeoinformation systems since its original design as a land management software tool for militaryinstallations. Support for environmental applications has been an integral part of its 20+ years ofdevelopment. Current stable release includes, among other features, a 2D/3D vectorial engine, SQLbased database systems and a rich variety of raster and vector graphic formats, allowingapplications in the areas such as geospatial data management and analysis, image processing,graphics/maps production, spatial modeling and visualization, both in academic and commercialfields as well as by many governmental agencies and environmental consulting companies.Change detection is the measure of the distinct data framework and thematic change informationthat can guide to more tangible insights into underlying process involving land cover and land usechanges than the information obtained from continuous change. Digital change detection is theprocess that helps in determining the changes associated with land use and land cover propertieswith reference to geo-registered multitemporal remote sensing data. It helps in identifying changebetween two or more dates that is uncharacterized of normal variation. Change detection is useful inmany applications such as land use changes, habitat fragmentation, rate of deforestation, urbansprawl, and other cumulative changes through spatial and temporal analysis techniques such as GISand Remote Sensing along with digital image processing techniques. Vegetation indices (VIs) areamong the oldest tools in remote sensing studies. Although many variations exist most of them ratiothe reflection of light in the red and NIR sections of the spectrum to separate the landscape intowater, soil, and vegetation.Normalized Difference Vegetation Index (NDVI), Principal Component Analysis (PCA), TasseledCap Transformation (TCT) images were derived from the satellite data analysis. As for land usechange detection technique, PCA and unsupervised were used. Land use change detection wasbased on the difference of multi-date of NDVI, PCA and panchromatic images.Key words: GRASS GIS, GIS, NDVI, PCA, Remote Sensing, classification, Landsat7 EMT, Imagedifferencing, change detection and Image Ratios. iv
  • 6. 1.0 IntroductionA Geographic Information System (GIS) is software for gathering, storing, managing, andpresenting data that are geographically referenced. The characteristic of a GIS is the capability ofexamining different sets of geo-referenced data and reaching a conclusion about the existingrelationship among the data. In other words, a GIS can link information that is difficult to associateby other means (Harmon and Shapiro, 2008).According to Harmon and Shapiro (2008) and Buckley (1997) a GIS has four components: a datainput subsystem, a data storage and retrieval subsystem, a data manipulation and analysissubsystem, and a data reporting subsystem (or a data output and display subsystem). FurthermoreHarmon and Shapiro (2008) indicated that the data inputs include spatial and thematic data derivedfrom a combination of existing maps, aerial photographs, and manual interpretations of remotelysensed imagery. Remotely sensed imagery is imagery acquired at some distance using a camera or asensor. With a GIS an analyst can define spatial procedures to generate new information, such as thebest location to build a road, preserve a wildlife habitat, or harvest timber. Remote sensing data arean essential and cost-effective means to update the spatial and thematic information in a GIS.Remote sensing products are also valuable in themselves, as a cost effective source of geographicinformation such as vegetation type, vegetation productivity, water quality, and land use change.Harmon and Shapiro (2008) reported that an image-processing system has at least five elements:image input, image storage, image analysis, accuracy assessment, and information reporting. Thetwo systems (GIS and Image-processing) are similar. Only the input and analysis elements requiredifferent functions for each system. If the GIS is a raster based system, the storage and reportingfunctions are identical. Some of the analysis functions may overlap. For example, the imageaccuracy assessment programmes are actually standard GIS analysis routines. It seems natural then,that these two systems should be integrated. When GIS and image-processing systems areintegrated, not only can the remote sensing data be used to update the GIS, but the GIS thematicdata and attributes can be used to guide image classification. By integrating GIS and image-processing capabilities, GRASS GIS offers the analyst these advantages.A GIS can be divided into five physical components: People (GIS specialists and end users), Data(vector and raster), Hardware (computer servers, desktops, Local Area Network), Software (GRASSGIS, Saga GIS, ArcGIS, Idrisi Andes, ERDAS Imagine), and Procedures (well-designedimplementation plan and business rules). All of these components need to be in balance for thesystem to be successful. No one part can run without the other.Mahmoodzadeh (2007) define changes detection as a technique used in remote sensing to determinethe changes in a particular object of study between two or more time periods. According toRamachandra and Kumar (2004) change detection is the measure of the distinct data framework andthematic change information that can guide to more tangible insights into underlying processinvolving land cover and land use changes than the information obtained from continuous change.Digital change detection is the process that helps in determining the changes associated with landuse and land cover (LULC) properties with reference to geo-registered multitemporal remotesensing data. It helps in identifying change between two or more dates that is uncharacterized ofnormal variation. Change detection is useful in many applications such as LULC changes, habitatfragmentation, and rate of deforestation, coastal change, urban sprawl, and other cumulativechanges through spatial and temporal analysis techniques such as GIS and Remote Sensing along 1
  • 7. with digital image processing techniques. Change detection is an important process for monitoringand managing natural resources and urban development because it provides quantitative analysis ofspatial distribution in the area of interest (Mahmoodzadeh, 2007).In this regard Open Source GIS such as GRASS GIS helps in LULC analysis in a cost-effectiveway. GRASS GIS is being developed for the GNU/Linux, Mac OS/X, and Windows operatingsystems. It is a GIS used for geospatial data management and analysis, image processing,graphics/maps production, spatial modeling, and visualization. Most of the commands in GRASSGIS are command line arguments and requires a user friendly and cost-effective graphical userinterface (GUI). This paper discusses different type of satellite image analysis such as imageunsupervised classification, image ratioing, Principal Component Analysis (PCA), NormalizedDifference Vegetation Index (NDVI) and Tasseled Cap Transformation (TCT).1.1 GoalThe goal of this paper was to undertake change detection, construct a Landsat7 ETM+ ImageDatabase and LULC Maps of the Luswishi Farm Block using GRASS GIS.1.2 PurposeThe purpose of the analysis is to undertake change detection of the Luswishi Farm Block.1.3 Objectives  To carry out the LULC and temporal change analysis for Luswishi Farm Block, Lufwanyama  To construct a TM Image Database and LULC Maps of the Luswishi Farm Block2.0 Description of the Study SiteLuswishi Farm Block area is located on the Copperbelt Province of Zambia in Lufwanyama districtbetween 100 kilometers to 132 kilometers west of Kalulushi. An M18 Kalengwa road runs centrallythrough the Farm Block. The area is between Luswishi River in the east and Mushingashi River tothe west. In the north it lies between Mushingashi-Kambilombilo Resettlement Scheme andMirumbi river at latitudes 12o 45’ S and 13o 05’ S and longitudes 27o 05’ E and 27o 15’ E.Mushingashi River also marks the boundary between the Kambilombilo Resettlement and the FarmBlock. The northern part of the farm block is not easily accessible as it is served by tracks and so isthe southern half. Present settlements are randomly scattered within the area. Few farming familiesown livestock. Active or sizable settlements are along the Kalengwa road and the state of this roadis poor to very poor. 2
  • 8. Figure 1: Location of the Luswishi Farm BlockFigure 2: Luswishi Farm Block Map overlayed with Roads and Rivers 3
  • 9. 3.0 Data Collection3.1 Satellite images data acquisitionSatellite images of LANDSAT 7 ETM+ (Enhanced Thematic Mapper plus) were downloaded fromthe Global Land Cover Change Facility [http://glcf.umiacs.umd.edu/data/] through the webinterface and imported to GRASS GIS database using the GEOTIFF/TIFF (r.in.gdal) conversionmodule. The images were acquired in one scene (with eight bands) (path 173, row 069) andcovering the whole area of the Luswishi Farm Block. The pixel sizes are 30.0m x 30.0m for bands 1(blue), 2 (green), 3 (red), 4 (Near Infrared), 5 (mid infrared) and 7 (mid infrared), 57m x 57m forthermal band 6 and 14.25m x 14.25m for panchromatic band 8. The ETM+ data included one scene(path 173, row 069) from 1st May 2001 and 12th May 2005. A list of the bands with their spannedwavelength (in micrometers) is given in Table 1 below.Table 1: TM Band and their spanned wavelengthLandsat 5 (TM sensor) Wavelength Resolution Application (micrometers) (meters)Band 1 (Blue) 0.45 - 0.52 30 Soil/vegetation discrimination; bathymetry/coastal mapping; cultural/urban feature identification; forest type mappingBand 2 (Green) 0.52 - 0.60 30 Green vegetation mapping (measures reflectance peak); cultural/urban feature extractionBand 3 (Red) 0.63 - 0.69 30 Vegetation vs. non-vegetated and aiding in plant species discrimination (plant chlorophyll absorption); cultural/urban feature identificationBand 4 (Near Infrared) 0.76 - 0.90 30 Identification of plant/vegetation types, health, and biomass content; water body delineation; soil moisture discriminationBand 5 (Mid Infrared) 1.55 - 1.75 30 Sensitive to moisture in soil and vegetation; discriminating snow and cloud covered areasBand 6 (Thermal 10.40 - 12.50 120 Useful in vegetation stress analysis, soil moistureInfrared) discrimination related to thermal radiation; thermal mapping (urban, water)Band 7 (Mid Infrared) 2.08 - 2.35 30 Discrimination of mineral and rock types; sensitive to vegetation moisture content4.0 MethodologyBasically, all satellite image-processing/analysis activities/operations can be grouped into threecategories: Image Rectification and Restoration (atmospheric effects, radiometric or geometricdistortion etc); Image Enhancement (contrast, edge enhancement, colour composites or ratios); andInformation Extraction (supervised and unsupervised classification). The former deals with initialprocessing of raw image data to correct for geometric distortion, to calibrate the dataradiometrically and to eliminate noise present in the data. The enhancement procedures are appliedto image data in order to effectively display the data for subsequent visual interpretation. It involvestechniques for increasing the visual distinction between features in a scene.Data from Landsat7 ETM+ imageries were processed using GRASS GIS software and spatialanalysis, interpolation and other calculations. There are many change detection techniques fromvisual comparison to detailed quantitative approaches (Wickware and Howarth, 1981). In this papertechniques evaluated are: PCA, NDVI, image ratios, unsupervised classification and image fusion. 4
  • 10. The methodology of the study involved - 1. Reviewing of topological map sheets and literature 2. Creation of base layers like district boundary, farm block boundaries, road network, mapping of water bodies, etc. from the topographic sheets of scale 1:250,000 and 1:50,000. 3. Extraction of bands from the data respectively downloaded from Global Land Cover Change Facility 4. Identification of ground control points (GCP’s) and geo-correction of bands through resampling. 5. Image restoration (geometric correction (Geometric correction is the process of assigning real-world co-ordinates to an image) and Radiometric restoration), Cropping (subsetting or clipping) of data corresponding to the study area. 6. Fusion of data using RGB (Red, Green, Blue) to HIS (Hue, Intensity, Saturation) and HIS to RGB conversion technique. 7. Histogram generation, Bi-spectral plots, Regression analysis, Principal Component Analysis (PCA), accuracy assessment, computation of the Root Mean Square Error (RMSE) for the images. 8. Computation and analysis of various vegetation indices (NDVI). 9. Generation of TCC (True Colour Composites) and FCC (False Colour Composite) and identification of training sites on FCC. 10. Collection of attribute information from field corresponding to the chosen training sites using Geographical Positioning System (GPS). 11. Classification of remote sensing data. LULC analyses in the farm block. 12. Change detection analysis using different techniques (Image differencing, Image ratioing, etc.). 13. Detection, visualization and assessment of change analysis. 14. Statistical analysis and report generation.5.1 RESULTS AND DISCUSSIONS5.1 Change detection techniquesDifferent change detection techniques such as composite analysis, normalized vegetation index,image ratioing, Principal Component Analysis and image fusion were used to attempt to assess theamount of change in the study area. Before the analysis could be performed the raster images wereclipped by masking. Raster MASK allows the user to block out certain areas of a map from analysisby hiding them from other GRASS raster modules. MASK is a raster map which contains the values1 and NULL. Those cells where the MASK map shows value 1 are available for display andprocessing while those assigned NULL are hidden. The map name MASK is a reserved filename forraster maps. The presence of a MASK map in a MAPSET entails that it will be used as for all rasteroperations when reading raster data. Any raster data falling outside of the MASK are treated as if itsvalue were NULL. To create a MASK, a base map should be available that is used to select whichvalues will represent the hidden and the active areas (Neteler and Mitasova, 2008). The procedureundertaken to mask the boundary of Luswishi Farm Block are as follows: v.to.rast – convert vectorto rast; r.mask – create a mask from the raster; and r.resample – to create a new raster clipped by 5
  • 11. the mask. Both images were resampled using the nearest neighbor algorithm. All the satellite bands for2001 and 2005 time periods were clipped within the GRASS GIS database.5.2 Image Processing and ClassificationIn general there are two methods for classification of remotely-sensed data; unsupervised andsupervised classification. For both methods the classification process required two major steps. Thefirst step in an unsupervised image classification is performed by i.cluster; the first step in asupervised classification is executed by the GRASS programme i.class. In both cases, the secondstep in the image classification procedure is performed by i.maxlik. The data have to be analyzed forsimilarities in their spectral responses, and then the pixels have to be assigned to classes. Theunsupervised method is fully automated based on the image statistics but it delivers only abstractclass numbers. Then the main task is to find the reasonable number of clusters/classes and assignground truth information to these classes. The supervised classification on the other hand requiresuser interaction and here training areas covering known land use have to be digitized. Imagestatistics are automatically derived from these training areas and used for classification.Common classification methods such as the Maximum Likelihood Classifier (MLC) are pixelbased. GRASS additionally provides a different method called Sequential Maximum A Posteriori(SMAP) classifier which takes into account that neighborhood pixels may be similar. The fact that aneighborhood of similar pixels will lead to spatial auto-correlation is used to improve the results.Figure 3: Unsupervised and Supervised classification procedure for multispectral data (Eastman,2006) (Adapted from Neteler and Mitasova, 2008) 6
  • 12. 5.3 Unsupervised classificationIn an unsupervised classification, the maximum-likelihood classifier uses the cluster means andcovariance matrices from the i.cluster signature file to determine to which category (spectral class)each cell in the image has the highest probability of belonging. In a supervised image classification,the maximum-likelihood classifier uses the region means and covariance matrices from the spectralsignature file generated by i.class, based on regions (groups of image pixels) chosen by the user, todetermine to which category each cell in the image has the highest probability of belonging. Ineither case, the raster map layer output by i.maxlik is a classified image in which each cell has beenassigned to a spectral class (i.e., a category). The spectral classes (categories) can be related tospecific land cover types on the ground.The programme will run non-interactively if the user specifies the names of raster map layers, i.e.,group and subgroup names, seed signature file name, result classification file name, and anycombination of non-required options in the command line, using the form.5.3.1 Set region setting to red image:g.region rast=2001_lsat7_2001.3 -p5.3.2 Creation of group and subgroupi.group allows the user to collect raster map layers in an imagery group by assigning them to user-named subgroups or other groups. This enables the user to run analysis on any combination of theraster map layers in a group. The user creates the groups and subgroups and selects the raster maplayers that are to reside in them. Imagery analysis programmes like i.point, i.rectify, i.ortho.photoand others ask the user for the name of an imagery group whose data are to be analyzed. Imageryanalysis programmes like i.cluster and i.maxlik ask the user for the imagery group and imagerysubgroup whose data are to be analyzed.Command: i.group [-rlg] group=name [subgroup=string] [input=name[,name,...]] [--verbose] [--quiet]i.group group=lsat7_20010 subgroup=lsat7_20010input=lsat7_2001.1,lsat7_2001.2,lsat7_2001.3,lsat7_2001.4,lsat7_2001.5,lsat7_2001.7i.group group=lsat7_20050 subgroup=lsat7_20050input=lsat7_2005.1,lsat7_2005.2,lsat7_2005.3,lsat7_2005.4,lsat7_2005.5,lsat7_2005.75.3.3 Clustering of image data and Generation of unsupervised statisticsi.cluster performs the first pass in the GRASS two-pass unsupervised classification of imagery,while the GRASS programme i.maxlik executes the second pass. Both programmes must be run tocomplete the unsupervised classification.i.cluster is a clustering algorithm that reads through the (raster) imagery data and builds pixelclusters based on the spectral reflectance of the pixels. The pixel clusters are imagery categories thatcan be related to land cover types on the ground. The spectral distributions of the clusters (which 7
  • 13. will be the land cover spectral signatures) are influenced by six parameters set by the user. The firstparameter set by the user is the initial number of clusters to be discriminated.i.cluster starts by generating spectral signatures for this number of clusters and "attempts" to end upwith this number of clusters during the clustering process. The resulting number of clusters andtheir spectral distributions, however, are also influenced by the range of the spectral values(category values) in the image files and the other parameters set by the user. These parameters are:the minimum cluster size, minimum cluster separation, the percent convergence, the maximumnumber of iterations, and the row and column sampling intervals.Command: i.cluster [-q] group=name subgroup=name sigfile=name classes=integer [seed=name][sample=row_interval,col_interval] [iterations=integer] [convergence=float] [separation=float][min_size=integer] [reportfile=name] [--verbose] [--quiet]i.cluster group=lsat7_20010 subgroup=lsat7_20010 sigfile=lsat7_20010_sig classes=8reportfile=lsat7_20010.txti.cluster group=lsat7_20050 subgroup=lsat7_20050 sigfile=lsat7_20050_sig classes=8reportfile=lsat7_20050.txt# look at report filegedit lsat72001.txtgedit lsat72005.txt5.3.3 Unsupervised classificationi.maxlik is a maximum-likelihood discriminate analysis classifier. It can be used to perform thesecond step in either an unsupervised or a supervised image classification.Command: i.maxlik [-q] group=name subgroup=name sigfile=name class=name [reject=name] [--overwrite] [--verbose] [--quiet]i.maxlik group=lsat7_2001 subgroup=lsat7_2001 sigfile=lsat7_2001_sig class=lsat7_2001_classreject=lsat7_2001_rejecti.maxlik group=lsat7_2005 subgroup=lsat7_2005 sigfile=lsat7_2005_sig class=lsat7_2005_classreject=lsat7_2005_reject# false color composited.mon x0 ; rgb b=2001_lsat7.7 g=2001_lsat7.5 r=2001_lsat7.1d.rgb b=2005_lsat7.7 g=2005_lsat7.5 r=2005_lsat7.1 8
  • 14. Figure 3: False Colour Composite for 2001 and 2005#classification resultd.mon x1d.rast.leg lsat7_2001_classd.rast.leg lsat7_2005_class Figure 4: 2001 land cover classification for 2001 (left) and 2005 (right) 9
  • 15. Table 2: RASTER MAP CATEGORY REPORT for 2001 and 2005 Categories (unsupervised classification) 2001 20051 Mixed vegetation 23,323 25,3932 Health vegetation 21,577 19,9623 Sparse vegetation 24,432 21,7064 Water 16,587 15,6375 Bare land/agricultural land 9,769 10,8506 Human settlements 4,391 6,530 Total Hectare 100,079 100,0795.4 Vegetation Indices5.4.1 Normalized Difference Vegetative Index (NDVI)Landsat7 ETM+ data can be used to assess the type, extent, and condition of vegetation over aregion, or vegetative change over time. These studies require that a vegetative index be calculatedfrom sets of remotely-sensed data. Vegetation indices are commonly used to monitor vegetation atvarious spatial and temporal scales. Vegetation indices are based on vegetation spectral properties inthe red and the near-infra red part of the spectrum. The most widely used index is the NormalizedDifference Vegetative Index (NDVI). The NDVI is calculated as a ratio between measuredreflectivity in the red (band 3) and near-infrared (band 4) portions of the electromagnetic spectrum.These two spectral bands are chosen because they are affected by the absorption of chlorophyll inleafy green vegetation and by the density of the green vegetation on the surface. Moreover, in redand near-infrared bands, the contrast between vegetation and soil is at a maximum.To study the vegetation status with NDVI, the Red and the Near Infrared channels (NIR) are takenand used as input for simple map algebra in the GRASS GIS command r.mapcalc (ndvi = 1.0* (nir - red)/(nir + red). With r.colors an optimized "ndvi" color table can be assignedafterward. By normalizing the difference in this way, the values can be scaled between a value of -1to +1. This also reduces the influence of atmospheric absorption. The table below shows typicalreflectance values in the red and infrared channels, and the NDVI for typical cover types (Holben,1986). Water typically has an NDVI value less than 0, bare soils between 0 and 0.1 and vegetationover 0.1. Negative values of NDVI (values approaching -1) correspond to water. Values close tozero (-0.1 to 0.1) generally correspond to barren areas of rock, sand, or snow. Lastly, low, positivevalues represent shrub and grassland (approximately 0.2 to 0.4), while high values indicatetemperate and tropical rainforests (values approaching 1). Figure 4 and table 4 shows the NDVImap and computed statistics for 2001 and 2005 respectively.Table 3: NDVI for typical cover types COVER TYPE RED NIR NDVI Dense vegetation 0.1 0.5 0.7 Dry Bare soil 0.269 0.283 0.025 Clouds 0.227 0.228 0.002 Snow and ice 0.375 0.342 -0.046 Water 0.022 0.013 -0.257 10
  • 16. Figure 5: NDVI map for 2001 and 2005Table 4: NDVI for 2001 and 2005 2001 Ndvi ranges Hectares 2005 Ndvi ranges Hectares -0.470588--0.300916 5 -0.301587--0.163228 59 -0.300916--0.131244 16 -0.163228--0.024868 1,038 -0.131244-0.038429 722 -0.024868-0.113492 20,713 0.038429-0.208101 49,295 0.113492-0.251852 74,952 0.208101-0.377773 49,947 0.251852-0.390212 3,310 0.377773-0.547445 95 0.390212-0.528571 7 Total hectare 100,079 Total hectare 100,079NDVI is usually used for observing the spatial distribution of vegetated area along with the healthstatus. It is the most commonly used VI as it retains the ability to minimize topological effects whileproducing a linear measurement scale. In additional, divisions by zero errors are significantlyreduced. Furthermore, the measurement scale has the desirable property of ranging from -1 to 1with 0 representing the approximate value of no vegetation. Thus negative values represent non-vegetated surfaces (Eastman, 2006).5.4.2 Image RatioingSpectral Ratioing is used to highlight differences between features and often reduces the effect ofdifferent light reflectance intensities. Spectral ratioing can be used to detect stressed or diseasedcrops. One common ratio, called Normalized Difference Vegetation Index (NDVI) is: (NIR-RED) /(NIR+RED) or simplified as below to (NIR /RED). To perform this spectral ratio:STEPS 1. Open the GRASS shell and type in: r.mapcalc etm_ratio="etm40/etm30" 11
  • 17. Confirm with Enter 2. Go to the Browser and click the Refresh button 3. Load etm_ratio to the map canvas. 4. Change raster properties to Pseudo color and activate the option “Invert Color map”.Figure 6: Spectral Ratioing for NIR and Red for 2001 (Left) and 2005 (Right)The result clearly captures the contrast between the red and infrared bands for vegetated pixels, withhigh index values being produced by combination of low red and high infrared reflectance. Theindex is susceptible to division by zero errors and the resulting measurement scale is not linear. As aresult, RATIO Vegetative Indices (VI) images do not have normal distribution making it difficult toapply some statistical procedures (Eastman, 2006).5.4.3 Ratio transformationsRatio transformations of the remotely sensed data can be applied to reduce the effects of theenvironment. Ratios also provide unique information and subtle spectral-reflectance or colordifferences between surface materials that are often difficult to detect in a standard image. It is alsouseful for discriminating between soils and vegetation. The number of possible ratio combinationsfor a multispectral sensor with P bands is n= P(P-1). Thus for the TMs six reflectance bands thereare thirty different ratio combinations - 15 original and 15 reciprocal. For the purpose of this studyten band ratios were examined using GRASS GIS to identify the Land Use – Land Cover (LULC)features.TM3/TM4: This ratio has defined barren lands and urban area uniquely. But it could not definewater body, forests and croplands. 12
  • 18. Figure 7: Image Ratio TM3/TM4 for 2001 and 2005TM4/TM3: This ratio distinguished vegetation, water and croplands. It has enhanced forests,barren lands. Because forests or vegetation exhibits higher reflectance in near IR region (0.76 -0.90u m) and strong absorption in red region (0.63-0.69u m) region. This ratio uniquely defines thedistribution of vegetation. The lighter the tone, the greater the amount of vegetation present. Figure 8: Image Ratio TM4/TM3 for 2001 and 2005 13
  • 19. Figure 9: Image Ratio Histograms for 2001 and 2005TM2/TM3: this ratio has distinguished croplands, barren lands sharply. But it hasnt separatedcroplands, forests and water body. Both forests and water body has appeared as lighter tone andbarren land appeared has dark tone. It did not enhance urban area. Chlorophyll has strongreflectance in the band 2 (0.52 -0.60u m) region and strong absorption in the band 3(0.63 -0.69u m)region, vegetation has appeared as higher tone. Figure 12: Image Ratio TM2/TM3 for 2001 and 2005TM3/TM2: This ratio has separated forests and croplands. Because band 3 (0.63-0.69m m) is thered chlorophyll absorption band of healthy green vegetation and band 2 (0.52-0.69m m) is thereflectance band from leaf surfaces. This ratio can be useful to discriminate broad classes ofvegetation. Croplands have appeared as lighter (brighter) tone and forests appeared as dark tone. 14
  • 20. Figure 10: Image Ratio TM3/TM2 for 2001 and 2005TM4/TM5: It enhances the water body, vegetation and presence of moisture content in thecroplands. Water body has appeared as dark tone and vegetation as lighter tone. Because water is astrong absorber in near IR region( band4) and higher reflectance in band 5 region. It can be usefulfor discriminating water bodies from land. Figure 11: Image Ratio TM4/TM5 for 2001 and 2005TM5/TM4: It has separated water body from forest, barren lands and vegetation. In this ratio waterhas appeared as dark tone and forest, barren lands, bare croplands all have exhibited brighter tone. 15
  • 21. Figure 12: Image Ratio TM5/TM4 for 2001 and 2005TM5/TM7: It has separated water body from lands (soils). It has also enhanced presence ofmoisture in croplands. All water bodies appeared as dark tone. Both band 5 and band 7 are sensitiveto moisture content variation in soils and vegetation. This ratio is useful for crop-drought studiesand plant vigor investigations. This ratio separates land and water uniquely. Since soils exhibitstrong absorption in the band 7 (2.08 -2.35u m) and high reflectance in band 5 (1.55 - 1.75u m), soilhas been enhanced in this ratio. Land has appeared as lighter tone and water appeared as dark tone. Figure 13: Image Ratio TM5/TM7 for 2001 and 2005TM3/TM5: This ratio enhances barren lands, highways, street patterns within the urban areas andurban built-up or cemented areas. It could not enhance the clear water but it enhanced turbid water.This ratio is useful for observing differences in water turbidity. Barren lands, highways, urban andbuilt-up areas have appeared as lighter tone and forests, water body and croplands appeared as darktone. 16
  • 22. Figure 14: Image Ratio TM3/TM5 for 2001 and 2005TM7/TM2: This ratio has separated forests and croplands. But it could not separated forests fromwater body; both features have appeared as dark tone. It enhances highways, urban and built-upareas and croplands and all of them have appeared as lighter tone. Figure 15: Image Ratio TM7/TM2 for 2001 and 2005Ramachandra and Kumar (2004) indicated that geocorrected images (Green, Red and Near InfraRed bands) of different dates were ratioed pixel by pixel (band by band) basis. Areas of change arerepresented by pixel values that differ from 1. The ratio value greater than 1 or less than 1represents a change depending upon the nature of changes occurred between the two dates. Figure16 below shows the histogram obtained for the near infrared band after performing the imageratioing. Image ratios are the basis of a simple algebraic method used for feature extraction,reduction of terrain illumination effects, image enhancement, computation of vegetation indices andmore. Image ratios are the basis of a simple algebraic method used for feature extraction, reductionof terrain illumination effects, image enhancement, computation of vegetation indices and more. Ingeneral, the ratio result for pixels with very different values for the input channels is larger(brighter) than for pixels with similar values. The image ratio equations can be computed withr.mapcalc. It is important to include a multiplier of 1.0 at the beginning of the map algebra 17
  • 23. expression because integer values are being divided. Otherwise the results would be zero and notthe expected floating point numbers (Neteler and Mitasova, 2008).Band 7 (Medium infrared) and 4 (Near infrared)g.region rast=2005_lsat7.1 -pr.mapcalc "ratio7_4_2005=1.0 * 2005_lsat7.7/2005_lsat7.4"d.erased.rast ratio7_4 Figure 16: Image Ratio Histograms for 2001 and 20055.5 Tasseled Cap transformation (TC)The TC transformation optimizes data viewing for vegetation studies as one of the availablemethods for enhancing spectral information content of Landsat7 TM. Four bands for TCcomponents are generated:  tasscap.1: corresponds to brightness, measure of soil  tasscap.2: corresponds to greenness, measure of vegetation  tasscap.3: corresponds to wetness, interrelationship of soil and canopy moisture  tasscap.4: corresponds to atmospheric haze.The Tasseled Cap Transformation in remote sensing is the conversion of the readings in a set ofchannels into composite values; i.e., the weighted sums of separate channel readings. One of theseweighted sums measures roughly the brightness of each pixel in the scene. The other compositevalues are linear combinations of the values of the separate channels, but some of the weights arenegative and others positive. One of these other composite values represents the degree ofgreenness of the pixels and another might represent the degree of yellowness of vegetation orperhaps the wetness of the soil. Usually there are just three composite variables (Watkins, 2006).i.tasscap calculates Tasseled Cap (Kauth Thomas, TC) transformation for LANDSAT-TM data(TM4, TM5, TM7). The TC transformation maps for 2001 and 2005 are shown in figures 6 and 7below. 18
  • 24. Tasseled Cap transformation 2001: Brightness Tasseled Cap transformation 2001: greennessTasseled Cap transformation 2001: Wetness Tasseled Cap transformation 2001: Atmospheric haze Figure 17: Tasseled Cap transformation for 2001 19
  • 25. Tasseled Cap transformation 2005: Brightness Tasseled Cap transformation 2005: greenness Tasseled Cap transformation 2005: Wetness Tasseled Cap transformation 2005: Atmospheric haze Figure 18: Tasseled Cap transformation for 20015.7 Principal Component Analysis (PCA)Principal Component Analysis (PCA) is a linear transformation technique related to FactorAnalysis. Given a set of image bands, PCA produces new set of images, known as components thatare uncorrelated with one another and are ordered in terms of the amount of variance they explainfrom the original band set (Eastman, 2006). Furthermore, Basith et al., (2010) indicated that PCA isa mathematical transformation that converts original data into new data channels that areuncorrelated and minimize data redundancy. This technique is usually used to reduce the number ofspectral components (spectral bands) to fewer principal components accounting for the mostvariance in the original multispectral images. Image spectral bands of two or more dates are treatedas a single data set. The current implementation of PCA in GRASS GIS uses an eigenvaluedecomposition (EVD) based algorithm (Richards et al., 2006). This algorithm features covariancematrix computation and matrix multiplication based data re-projection whose performancedegradation becomes prohibitive for large data sets. 20
  • 26. A characteristic of PCA is that the piece of information common to all input bands (high correlationbetween bands) is mapped to the first Principal Component (PC) whilst subsequent PC accounts forprogressively less of the total scene variance. This principle can be applied to multitemporaldatasets. If two images covering the same ground area but taken at different times of the year aresubjected to PCA, the first PC will contain all of the information that has not changed between thetwo dates whilst the second PC will contain all the changed information. According to Basith et al.,(2010) when PCA is applied to data embracing several spectral bands, data is concentrated in thefirst two or three components. The other components generally contain only noise. The areas ofgreatest change are found in the tails of the image histogram (Mahmoodzadeh, 2007).In GRASS, the Principal Component Transformation is implemented in i.pca. The module requiresthe input channel names (at least two images) and a prefix for the transformed PC image files,which will be enumerated incrementally. Optionally, the data can be rescaled to a range differentfrom the default range of 0-255. i.pca is an image processing programme based on the algorithmprovided by Vali (1990), that processes n (2 >= n) input raster map layers and produces n outputraster map layers containing the PC of the input data in decreasing order of variance ("contrast").The output raster map layers are assigned names with .1, .2, ... .n suffixes. The current geographicregion definition and mask settings are respected when reading the input raster map layers. Whenthe rescale option is used, the output files are rescaled to fit the min-max range.Step 1g.region rast=lsat7_2001.10 -pGRASS 6.4.0+42329 (luswishifb):~ > i.pcain=lsat7_2005.10,lsat7_2005.20,lsat7_2005.30,lsat7_2005.10,lsat7_2005.20,lsat7_2005.30out=pca_01_05Eigen values, (vectors), and [percent importance]:PC1 1940.11 (-0.4877,-0.3785,-0.3448,-0.4877,-0.3785,-0.3448) [99.16%]PC2 15.58 (-0.3641,-0.0784, 0.6011,-0.3641,-0.0784, 0.6011) [0.80%]PC3 0.85 (-0.3600, 0.5921,-0.1408,-0.3600, 0.5921,-0.1408) [0.04%]PC4 0.00 (0.1640,-0.6878,-0.0002,-0.1640, 0.6878, 0.0002) [0.00%]PC5 0.00 (-0.6176,-0.1474, 0.3113, 0.6176, 0.1474,-0.3113) [0.00%]PC6 -0.00 (0.3028, 0.0721, 0.6349,-0.3028,-0.0721,-0.6349) [-0.00%]GRASS 6.4.0+42329 (luswishifb):~ > r.univar -e pca_01_05.3100%total null and non-null cells: 3085120total null cells: 1852714Of the non-null cells:----------------------n: 1232406minimum: 196maximum: 216range: 20 21
  • 27. mean: 204.607mean of absolute values: 204.607standard deviation: 1.29921variance: 1.68793variation coefficient: 0.634976 %sum: 2521588131st quartile: 204median (even number of cells): 2053rd quartile: 20590th percentile: 206GRASS 6.4.0+42329 (luswishifb):~ >Step 3: PCA can be used to perform a simple change detection analysisd.rgb b=pca_01_05.1 g=pca_01_05.2 r=pca_01_05.3d.rgb b=lsat7_2001.10 g=lsat7_2001.20 r=lsat7_2001.30d.rgb b=lsat7_2005.10 g=lsat7_2005.20 r=lsat7_2005.30r.univar -e pca_01_05.3# Simple change detection:# consider high PCA values in high PCA component as changer.mapcalc "changes_01_05=if(pca_01_05.3 > 205,1,null())"# vectorize and remove small areas# (3 pixel * 28.5m)^2 = 7310.2m^2r.to.vect -s changes_01_05 out=changes_01_05 feat=areav.clean changes_01_05 out=major_changes_01_05 tool=rmarea thresh=7300# overlay to Oct/1987 mapd.rgb b=lsat7_2001.10 g=lsat7_2001.20 r=lsat7_2001.30d.vect major_changes type=boundary col=redd.vect lakes fcol=blue type=area# overlay to May/2002 mapd.rgb b=lsat7_2005.10 g=lsat7_2005.20 r=lsat7_2005.30d.vect major_changes type=boundary col=redd.vect rivers fcol=blue type=areaThe resulting vector polygons captured significant land use changes, especially conversion fromvegetated area to developed areas. Changes (Figure 19) and major changes (Figure 20) whereforests were being converted to agricultural land are 22, 978.39 ha and 9, 204.5 ha respectively. 22
  • 28. Figure 19: Change in Luswishi Farm Block from 2001 to 2005 overlaid with roads and rivers 23
  • 29. Figure 20: Major changes (2001 to 2005) overlaid with roads and rivers5.8 Image FusionAccording to Neteler and Mitasova (2008) satellite data sets with high radiometric resolution(multispectral channels) lack high geometric resolution and vice versa. However, for an accurateimage interpretation, both radiometric and geometric resolution should be high. Image fusion is amethod used to geometrically enhance images with high radiometric resolution by merging themultispectral channels with a panchromatic image. GRASS provides the HIS (i.rgb.his, ihis.grb)and the Brovey transform methods. GRASS provides two color conversion modules, the i.rgb.his toconvert an image from RGB to HSI and i.his.rgb to convert back from HSI to RGB. The Broveytransform method is implemented in GRASS as i.fusion.brovey. The formula was originallydeveloped for LANDSAT-TM5 and SPOT, but it also works well with LANDSAT7 TM.i.fusion.brovey - Brovey transform to merge multispectral and high-resolution panchromaticchannels.5.8.1 Image fusion with HSI transformationTwo data sets were used whose acquisition time was 2001 and 2005 to avoid land use change whichmay modify the result. For an HSI fusion the three RGB channels were transformed to HSI color 24
  • 30. model. The general idea of HSI is replacement of the intensity channel with a high resolutionpanchromatic channel for the back transformation from HSI to RGB color model. The effect is thatthe color information in lower resolution is merged with the high spatial resolution of thepanchromatic channel. In term of GIS a resolution change is required before back- transforming theimage to archive the highest spatial resolution in the output.The Landsat7 TM images for Luswishi Farm Block were enhanced each at 28.5m resolution withthe panchromatic channel of 14.25m resolution of the same satellite acquired at the same time.Before starting the procedure as outlined below, the input channels contrast were enhanced withr.colors. Then the input channels were converted to HSI color model with i.rgb.his at 28.5mresolution. The higher resolution was set with g.region as defined by the panchromatic channel. Toimprove the geometrical resolution, the original intensity image which resulted from the RGB toHIS transformation was replaced by the panchromatic channel for the back- transformation to theRGB color model. Finally three new RGB channels at 14.25m resolution containing the multi-spectral information from the input channels were generated.STEP 1: Applying a contrast stretch (histogram equal.)g.region rast=lsat7_2001.10 -pr.colors lsat7_2001.10 color=grey.eqr.colors lsat7_2001.20 color=grey.eqr.colors lsat7_2001.30 color=grey.eqr.colors lsat7_2001.40 color=grey.eqSTEP2: RGB image fusion at 28.5m resolution of RGB channelsd.erased.rgb b=lsat7_2001.10 g=lsat7_2001.20 r=lsat7_2001.30STEP 3: RGB/IHS conversioni.rgb.his blue=lsat7_2001.10 green=lsat7_2001.20 red=lsat7_2001.30 hue=hue intensity=int saturation=satSTEP 4: IHS/RGB back conversion. With ETMPAN replacing old intensity Imageg.region rast=lsat7_2001.80 -pi.his.rgb hue=hue intensity=lsat7_2001.80 saturation=sat blue=etm2001.10_15 green=etm2001.20_15 red=etm2001.30_15STEP 5: color contrast enhancementr.colors etm2001.10_15 color=grey.eqr.colors etm2001.20_15 color=grey.eqr.colors etm2001.30_15 color=grey.eqSTEP 6: Visualization of higher resolution color composited.erased.rgb b=etm2001.10_15 g=etm2001.20_15 r=etm2001.30_15 25
  • 31. Figure 21: Image fusion with HSI transformation for Luswishi Farm Block 2001 (left); 2005(right) at 14.25m5.8.2 Brovey transformation methodologyAn alternative method for image fusion is Brovey transformation. This method can easily beimplemented in GRASS GIS. The formula for Brovey transformation is shown as below.i.fusion.brovey performs a Brovey transformation using three multispectral and the panchromaticsatellite image scene channels. Three new channels are calculated according to the formula andsteps below: DN_b1 DN_fused = ------------------------------- * DN_pan DN_b1 + DN_b2 + DN_b3STEP 1:if not done yet, apply a contrast stretch (histogram equal.)g.region rast=lsat7_2001.1 -pd.erasei.fusion.brovey -l ms1=lsat7_2001.1 ms2=lsat7_2001.2 ms3=lsat7_2001.3 pan=lsat7_2001.8 out=brov_2001STEP 2: it temporarily sets raster resolution to PAN resolution: 14.25# Original RGBd.erased.rgb b=lsat7_2001.1 g=lsat7_2001.2 r=lsat7_2001.3STEP 3: Brovey transformed RGBg.region -p rast=brov_2001.redd.erased.rgb b=brov_2001.blue g=brov_2001.green r=brov_2001.redd.vect luswishifarmb col=red 26
  • 32. i.landsat.rgb b=brov_2001.blue g=brov_2001.green r=brov_2001.redSTEP 4: Reversed B and G channel assignmentd.rgb g=brov_2001.blue b=brov_2001.green r=brov_2001.redCalculation of Brovey fusion map from Luswisi Farm Block Landsat scene:g.region rast=lsat7_2001.1 -pSTEP 5: R, G, B composite at 28.5md.rgb b=lsat7_2001.1_sub g=lsat7_2001.2_sub r=lsat7_2002_1.3_subSTEP 6: Brovey fusioni.fusion.brovey -l ms1=lsat7_2001.2_sub ms2=lsat7_2001.4_sub ms3=lsat7_2001.5_sub pan=lsat7_2001.8_sub outputprefix=brovey_2001_subSTEP 7: Display at 14.25mg.region rast=brovey_2001_sub.blue -pd.rgb b=brovey_2001_sub.blue g=brovey_2001_sub.green r=brovey_2001_sub.red Figure 22: Image fusion with Brovey transform 2001 (left) and 2005 (right)6.0 ConclusionsHolistic decisions and scientific approaches are required for sustainable development of the farmblock. Change detection techniques using temporal remote sensing data provide detailedinformation for detecting and assessing land cover and land use dynamics. Different changedetection techniques were applied to monitor the changes. The change analysis based on two dates,spanning over a period of four years using unsupervised classification, showed an increasing trendin human settlement and bareland and decline in spatial extent of health vegetated areas. Depletion 27
  • 33. of water bodies and large extent of bareland in the farm block is mainly due to lack of integratedland use planning and watershed approaches and mismanagement of natural resources.Image Ratio transformations of the remotely sensed data can be applied to reduce the effects of theenvironment. Ratios also provide unique information and subtle spectral-reflectance or colordifferences between surface materials that are often difficult to detect in a standard image. It is alsouseful for discriminating between soils and vegetation.NDVI calculations are based on the principle that actively growing green plants strongly absorbradiation in the visible region of the spectrum while strongly reflecting radiation in the NearInfrared regionThe NDVI has been most commonly used to map spatial and temporal variation invegetation and the calculation of the NDVI value turns out to be sensitive to a number of perturbingfactors including: atmospheric effects, clouds, soil effects, anisotropic effects and spectral effects.The limitation of the study is that no identification of ground control points (GCP’s) and collectionof attribute information from field corresponding to the chosen training sites using GeographicalPositioning System (GPS) was done. Bi-spectral plots, Regression analysis, accuracy assessmentand computation of the Root Mean Square Error (RMSE) for the images were not undertaken andno supervised classification was performed on the satellite imagery. 28
  • 34. ReferencesBasith A, Matori A. N., Harahap I. S. H. and Talib J. A, Application of land use change detectionfor identification of land slide risk areas in Pulau Penang using a decade of landsat7 ETM+images. MRSS, PWTC, Malaysia, April 28-29, 2010Buckley D. J. (1997), The GIS Primer – An Introduction to Geographic Information Systems,.Corporate GIS Solutions Manager Pacific Meridian Resources, Inc.Carl F. Jordan, Derivation of Leaf-Area Index from Quality of Light on the Forest Floor. EcologyVol. 50, No. 4 (Jul., 1969), pp. 663-666Dawelbait M. A. A. and Morari F., LANDSAT, Spectral Mixture Analysis and Change VectorAnalysis to Monitor Land Cover Degradation in a Savanna Region in Sudan (1987-1999-2008),Intl. J. Water Resources & Arid Environ., 1(5): 366-377, 2011Deng, JS and Wang, K and Deng, YH and Qi, GJ, "PCA-based land-use change detection andanalysis using multitemporal and multisensor satellite data", International Journal of RemoteSensing, vol. 29, no. 16, pp. 4823, 2008.Eastman J. R. (2006), Guide to GIS and Image Processing, IDRISI Andes. Clark UniversityGlenn E. P. Huete A. R., Nagler P. L. and Nelson S. G. (2008), Relationship Between Remotely-sensed Vegetation Indices, Canopy Attributes and Plant Physiological Processes: WhatVegetation Indices Can and Cannot Tell Us About the LandscapeGRASS GIS 6.4.1 Reference ManualHarmon V. and Shapiro M. (2008), GRASS Tutorial: Image Processing. U. S. Army ConstructionEngineering Research LaboratoryLu, D and Mausel, P and Brondizio, E and Moran, E, Change detection techniques, InternationalJournal of Remote Sensing, vol. 25, no. 12, pp. 2365, 2003.Koutsias, N and Mallinis, G and Karteris, M, A forward/backward principal component analysisof Landsat-7 ETM+ data to enhance the spectral signal of burnt surfaces, ISPRS Journal ofPhotogrammetry and Remote Sensing, vol. 64, no. 1, pp. 37, 2009.Mahmoodzadeh H., Digital Change Detection Using Remotely Sensed Data for Monitoring GreenSpace Destruction in Tabriz. Int. J. Environ. Res. 1 (1): 35-41, 2007, Winter 2007 ISSN:1735-6865Graduate Faculty of Environment University of TehranNeteler M. and Mitasova H. (2008), Open Source GIS, A GRASS GIS Approach 3rd Edition.Springer Science+Business Media, LLC.Ramachandra T. V. and Kumar U. (2004), Geographical Resources Decision Support System for i
  • 35. land use, land cover dynamic analysis. Proceedings of the FOSS/GRASS User Conference –Bangkok, Thailand, 12-14 September 2004Richards J. A. and Jia X. (2006), Remote Sensing Digital Image Analysis: An Introduction,Springer VerlagRyan L. (1997), Creating a Normalized Difference Vegetation Index (NDVI) image UsingMultiSpecWatkins, Thayer (2006), The Tasseled Cap Transformation in Remote SensingWickware G.M., and Howarth P. J., (1981). Change Detection in the Peace – Athabasca DeltaUsing Digital Landsat Data. Remote Sens. Environ., 11, 9-25. ii