Application of Remote Sensing and GIS for Modeling ... - IIC Academy

8MB Size 2 Downloads 27 Views

The established methodology of DIP, Spatial database techniques and remote ... using hybrid classification approach NDVI techniques, Masking techniques and ...
Application of Remote Sensing and GIS for Modeling and Assessment of Land Use/Land Cover Changes in Krishna Delta

V.S.S.KIRAN*, G Jai Sankar and M. Jagannadha Rao

1. Lead Faculty, IIC Academy at IIC Technologies Ltd, Visakhapatnam – 530041. 2. Professor, Dept of Geo Engineering, Andhra University, Visakhapatnam- 530017 2. Professor, Dept of Geology, Andhra University, Visakhapatnam- 530017. *Email: [email protected]

ABSTRACT The fluvial activity of a river system has been a dynamic process by which the land surface area will undergo a continuous transformation both spatially and temporarily. The study of these changes of any river system will provide valuable scientific inputs to understand the nature of changes that are occurring by natural process as well as the impact of anthropogenic activities. Especially Land use and Land cover is an important component in understanding the interaction of the human activities with the environment and thus it is necessary to simulate environmental changes from periodically. In addition another one factor of changes in deltaic regions is climate conditions because present time the climatic change is one of the global environmental challenges for the significant impact on deltaic region. Remote Sensing has been established to be an important tool to study such dynamic changes of any natural process. Since it gives us reasonable pictures of entire process in spatial and temporal terms. In this study the Krishna Delta (16°7'14.07"N to 15°47'24.29"N and 80°43'29.29"E to 81°17'3.04"E) of Andhra Pradesh has been taken up to study for “Change Detection” using the remotely sensed data and GIS tools. In this present study the Landsat TM & ETM data for a period of twenty seven years from 1973 to 2000 (three images) and LISS data for a period of three year from 2009 to 2011(two images) will be used. Besides the Landsat and LISS satellite images data the elevation data from ASTER & SRTM are also be used to understand changes in slopes. The established methodology of DIP, Spatial database techniques and remote sensing and GIS models (Erdas Imagine 2013 and ArcGIS10.2.1) was carried out using hybrid classification approach NDVI techniques, Masking techniques and classifications techniques i.e. supervised and unsupervised are used to generate data on land use/land cover changes. The data generated in the form of various output maps, comparative graphs and DEMS have been presented in this paper. In addition the Statistics generated from satellite remote sensing data helped in understanding the physical process and changes in land use/land cover in space and time variation. The most important geological considerations and factors controlling the fluvial system resulted various changes have been brought out. The paper established that the right combination of RS & GIS techniques could be perfectly used for the change detection studies. Keywords: Remote Sensing, GIS, DIP, Landsat TM & ETM, ASTER/SRTM.

INTRODUCTION: We aware of every day the change is coming in environment. Faster, slower, or going in circles, it is the biggest constant in our world. An attained a couple things that are already in Remote sensing environment, that makes understanding imagery and change much more palatable. In Remote sensing the change detection is an important application to gain knowledge about the geographical changes. It is a method to find out the changes of specific features within a certain time period. During this period, the change detection application provides spatial distribution of features and detailed information of its changes, which carried out from the different periods of remote sensing satellite data. ABOUT THE STUDY AREA: The study area relates to the deltaic region located in Krishna district and Andhra Pradesh state. Study area is covered at 16°7'14.07"North to 15°47'24.29"North latitudes and 80°43'29.29"East to 81°17'3.04"East longitudes. The mouth of Krishna River is Hamsaladeevi, Krishna district, Andhra Pradesh. The delta of this river is one of the most fertile regions in India and was the home to ancient Satavahana and Ikshvaku Sun Dynasty kings. METHODOLOGY: The methodology is derived into the different parts i.e. Data collection, Geo-referencing, projection, Data preparation, Land Use/Land Cover classification, Data Analysis, LU/LC Area calculation, Land use/ Land cover change detection using GIS statistical models & Image Subtraction method, Surface Analysis etc. Data Collection: For this present study the several datasets were required to reach objective and aim. Data themes including slope, aspect, land use/land cover classes like Lake, Water Body, Heavy Dense Forest, Medium Dense Forest, Open Forest, Sand Deposition, Wet land, Agriculture Land, Aquaculture Land and Settlement etc. The 1973, 1990 and 2000 Tm & Etm plus landsat satellite datasets were downloaded from the Global Land Coverage Facility (GLCF) website and 2009 & 2011 liss-3 resourcesat satellite datasets were downloaded from the Bhuvan Data Archive website. The remaining data sets i.e. Ground Elevation Data was obtained from the SRTM & ASTER website and Base maps SOI Reference Maps was obtained from Department of SOI. Combined, these sources provided the datasets used for the comprehensive analysis of this project. Erdas Imagine 2013 was used to create land use/ land cover maps of all years and ArcCatalog 10.2.1 was used to create GeoDatabases to store datasets and the results of analysis. Geo Referencing & Projection: All collected data is geometrically corrected using some ground control GPS points and the SOI base points in the software of ArcGIS 10.2.1. Resourcesat and Landsat satellite data georeferenced by the 2nd order polynomial method and double imagery rectification process. The data used for the analysis were all projected using the Global Coordinate System "GCS WGS 1984" and Projected Coordinate system “Universal Transverse Mercator” projection system with 44 North zone and spheroid is WGS 1984. Data Preparation: The data were geocorrected and projected in ArcGIS and it’s managed in ArcCatalog 10.2.1. As a vector shape file using defines the study area boundary and saved as a database file. The satellite datasets was subsetted from its original size to represent the study location from the Krishna Deltaic region of the Andhra Pradesh state in India using the raster "clip‟ tool in "Data Management Tools‟ within ArcToolbox. Figure 1 shows the study area of Krishna Delta. Figure-1: Study Area

Land Use Land Cover Classification: For this present study, the classification scheme is derived in different parts because the input satellite data is having some of atmospheric error and a few areas are covered by the clouds. For this purpose we applying the TNDVI technique (equation -1) and Image clear interpretation purpose a few of Digital image processing techniques applied i.e. histogram equalization, linear stretch, filtering techniques etc. The Land use/Land cover classification gave a rather broad classification where the land use land cover was identified by specific digits. The used classification methods is supervised classification with minimum parallelepiped & maximum likelihood classification were assigned in Erdas Imagine 2013.

TNDVI= SQRT((NIR-RED)/(NIR+RED)+0.5) -------(1)

The LU/LC maps prepared for all of the five years 1973,1990,2000,2009 & 2011. All Land Use and Land Cover datasets were divided into 10 different pre-determined categories by the USGS (Figure-2). Area Calculation: This aspect of analysis examined the area and percentage change for each year (1973, 1990, 2000, 2009 and 2011) for each land use land cover type. On the display attribute table we can add one area column and the count that attribute field, which makes up the number of cells in a particular raster class, was used to compute the area in square miles for each individual land use/land cover category. The following results in percentage of Land Use / Land Cover for all five year samples 1973, 1990, 2000, 2009 and 2011 is represented in figure 3. Land Use Land Cover Change Detection Using Image Subtraction and GIS Statistical Model: In the change detection, multi-band remote sensing image is usually used due to it provides enough information. The change detection methods of multi-band remote sensing image can be divided into image subtraction method and the method of change detection after land use/ land cover classification, apart from this change detection based on elevation change information of digital surface model. Some of the change detection methods are Image subtraction method, changing vector analysis method, Principal component analysis method, Pixel based classification method, and cell statistical method and spectral features variance method are usually used in the former change detection method. In this present research we processed the multi band remote sensing data to create an Image segmentation model in GIS and also using the change detection model image subtraction method in Erdas Imagine and GIS cell Statistics function in Spatial Analyst tool in ArcGIS, changes that occurred between LULC 1973 and 2011 were identified from where ever areas is changed and also identified unchanged area and minimum changed area. Specially in Image Subtraction method is depend on the gained from the subtraction of the gray values of corresponding pixels of images after image registration. The gray values of the subtraction image is to show the extent of changes of two images. But the present generated model is first we classified the all land use/land cover classes and we are using these output from image subtracted method in the base of predefined values of each pixel in particular class(flowchart-1 and equation 2). The resultant Output Images is represented in figure 4. Results & Discussion: The resulted most significant changes noted were within the agricultural land, medium dense forest, open forest and aquaculture area. The percentage difference shows a substantial increase in Wet land dense forest to agriculture land between 1973 and 1990, especially both agricultural land and forest cover area both increases in aquaculture land at the time period from 1990 to 2000, 2009 and 2011. There was a decrease in Wet land from 1973 (146.661 sq. miles) to 2011 (8,631 sq. miles) and an increases in Aquaculture land from 1990 (85.764 sq miles) to 2011 (152.698). This change may be somewhat related to the decrease of wet land and forest area within the study area. Wet and forest land decreased significantly by 25-30 % from 1973 to 2011. Figures 4& 5 illustrate the changes that occurred over the 38 years of time period and also its gives a tabulated illustration of the percentage change.

Flow Chart -1: Present the Change detection Model of Image Subtraction and GIS Statistics method of the present study

Figure-2: Land Use/ Land Cover Classification Map of All Years

Figure-3: Land Use/ Land Cover Area Percentage Table & Bar Graph

Figure 4: Final Change Detection Resulting Image ----------------------- Output-5











Geo Database

Image Subtraction Method

Geo Statistics for Area Calculation


Image Segmentation based Subtraction Method: Base on the above flow chart model we are procession the all five image in one by one process i.e. (output 2-output 1), (output 3- output 2), (output 4- output-3) (output 5- output 4) and finally overlay of all outputs in GIS and generate the final out put with the help of layer stacking method. The formuNRSqr}~”¢£Š ‹ ¯ ° ¹ º çҽ竚Œš{šjŒjYG5#hà}+hò*5?CJOJ[?]QJ[?]^J[?]aJ#hà}+hº ë5?CJOJ[?]QJ[?]^J[?]aJ hà}+hº

ëCJOJ[?]QJ[?]^J[?]aJ hà}+hö\¦CJOJ[?]QJ[?]^J[?]aJ hDøh<(CJOJ[?]QJ[?]^J[?]aJhDøCJOJ[?]QJ[?]^J[?]aJ hDøhö\¦CJOJ[?]QJ[?]^J[?]aJ#hà}+hö\¦5?CJOJ[?]QJ[?]^J[?]aJ)hh|5?B*[pic]CJOJ[?] QJ[?]\?^J[?]aJph)hð~Œ5?B*[pic]CJOJla which used for Image Subtraction method:

Dxk ij= xk ij(t2)- xk ij(t1)+ C ------------------------------------- - equation 2 Where i, j as pixel coordinates, k for the classified image, xk ij(t1)for the pixel (i, j) value of k-class of image, t1, t2 for the time of the first and the second image, C is constant value 255 because 8 bit satellite image value is 0 to 255.