Felkner, John 
  
The eight factors are all in the form of continuous raster grid maps of each province. All have grid cells of identical 
spatial resolution (60 x 60 meters). 
The environmental models are the following: 
i. topography (based on a Digital Elevation Model (DEM)) 
ji. soil moisture index — using a computer GIS model that has as its inputs precipitation, soil hydrologic group 
maps and slope to derive an index representing relative soil moisture of each pixel; 
iii. proximity to timber resources — a GIS raster grid in which each cell in the model is assigned a value reflecting 
its relative proximity to existing timber resources (derived from Landsat satellite imagery); 
iv. proximity to water resources — a GIS raster grid in which each cell in the model is assigned a value reflecting 
its relative proximity to existing water resources (from satellite imagery). 
The socio-economic models are: 
i. population potential — a continuous raster grid attempting to model sub-provincial spatial population 
variability on average for the 1980s, based on population village values from the Community Development 
Department (CDD) Survey of the Thai Government data (conducted every two years). The model assumes 
that population distribution will occur along infrastructure networks as a function of travel time along those 
networks; 
ii. proximity to markets — a raster grid in which each cell has a value representing its relative proximity to 
markets (defined as villages, towns, cities, and infrastructure linkages to national and international markets 
external to the Province), assuming that access to markets will occur along infrastructure networks as a 
function of travel time along those networks; 
iii. proximity to infrastructure — a raster grid in which each cell has a value representing its relative proximity to 
infrastructure networks. The model takes into account steepness of slope as an inhibitor to infrastructure 
access. It also assumes that areas of land adjacent to major infrastructure segments (major highways) have a 
higher proximity value that land equally close to less important infrastructure elements (dirt roads); 
iv. poverty measure — a raster grid in which each cell has a value representing a model of sub-Provincial spatial 
variation in wealth. the model uses data pertaining to income or other indicators of wealth from the CDD data 
at the CDD village level, and then assumes that wealth will be spatially distributed along infrastructure 
networks as a function of access/travel time along those networks. 
The population and proximity to markets models distribute data values derived from the CDD village values spatially 
along the road network by using a negative exponential gravity model. Once these values are distributed throughout the 
extent of the road network, a continuous surface is achieved by using an Inverse Distance Weighted (IDW) spatial 
interpolation. 
22 Land Use Change Detection 
Land use change detection was obtained by processing 1979 and 1989 Landsat Multi-Spectral Scanner (MSS) and 
Thematic Mapper satellite imagery for Sisaket and Chachoengsao Provinces. 1979 and 1989 images were first 
classified using and unsupervised classification to derive maps of landcover for each Province for both 1979 and 1989. 
Land use change detection from 1979 to 1989 was then obtained by performing an unsupervised classification on a 
single dataset containing data from both the 1979 and 1989 images. The resulting clusters were classified with 
reference to the 1979 and 1989 landcover maps. The assumption was that certain clusters would represent areas of 
change, as described in (Jensen 1996). 
  
434 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B7. Amsterdam 2000.