Landcover change detection using digital analysis of remotely sensed satellite data: A methodological study
Charat Mongkolsawat, Phanee thirangoon
Khin Kaon University, khon Kaen 4002 Thailand
Abstract
The objectives of this study were to develop a land cover change detection methodology through digital analysis of remote sensing satellite data. Two geocoded TM subscences of yasothon province northwest Thailand acquired during the wet season and the dry season were selected for analysis a supervised classification based on land cover and terrain was performed on each scène which yielded a September image with 11 classes and an April image with 14 classes. The two classified images were mathematically combined resulting in a unique linear combination of all 154 possible grey level image through a process of regrouping these grey levels in to classes the resultant image province meaningful land cover dynamics with respect to terrain type.
Introduction
Satellite remote sensing is widely accepted as a technique to study land use land cover change Dynamics .The use of satellite data for compiling land use change area is becoming substitute for data derived from time consuming aerial photo interpretation .In Thailand land use change is rapidly increased encroachment on forest reserve for agriculture is at alarming rates with significant impacts on soil erosion soil Stalinization rural communities and climate patterns. Better assessment of land use land cover change using digital analysis of remotely sensed satellite data can help decision maker to develop effective plans for the management of land.
Two main approaches to digital change detection have been reported. Both involve autitemporal images and can be categorized as single data set or separate data set analysis, separate data set analysis involves classification of each-date imagery separately into landcover classes. The results were subsequently compared 9gordon, 1980; willighton et al, 1986). Single data set analysis involves coregistering and resembling multitemporal images into the same dataset matamatical transformation, aminly image differencing and /or rationing, is then applied to the raw coregistered images to produce a residual image indicating the relative change in reflctance between the two dates. This technique is reported to give slightly more accurate result . (see , for example, Nelson, 1982; jenson and toll, 1982 : woodwell et al., 1983; Singh, 1986; Quarmby et al., 1987 ) Image differencing and image rationing have been used to detect abrupt changes in canopy reflectance from forest harvest and road construction (woodwell0 et al., 1983; tucker et al., 1984: Pelletier and Sader, 1985, ) as well as urban encroachment of agricultural land (Quarmby et al., 1987).
West of the digital change detection reported are two dates investigation. Sader (1988) ufllized Landsat imagery of three dates, supplemented by aerial photography, to study forest change dynamics in a troprical area The technique analyses the three date imageries as a simple data set and nvolves computation of normalized difference vegetation index of each data as data compression procedure. Modified parallelepiped classifier was employed to generate a multitemporal greenness image. Thus this technique is suitable for detection of forest change.
The objective of this study was to develop a landcover/landuse change detection methodology through digital analysis of satellite data.
Study Area
To study change detection methodology, substances of yasothon, Northeast Thailand, corresponding to central part of landsat tM path/row 127 /49 was selected.
Geologically, the area is underlain by a thick sequence of Mesozoic rock of the Maha sarakam formation gently undulating terrain with sparse trees and isolated patches of forest remants is commonly found. The loamy paleaquults and loamy paleustults are found extensively in lower terraces and upper terrace respectively, clayey Tropaquepts are present narrow strips in the flood plain. The soils are mainly derived from alluvium of sandstone origin. Aeolian deposits are found sporadically.
