Felkner, John 
  
3 SPECIFICS OF DIGITAL ANALYSIS ENVIRONMENT 
3.1 Raster Environment 
The analysis completed for this research takes place in a digital context — i.e. it is a computer analysis and a computer 
predictive land use change model. All input economic and environmental factors and all land use and land use change 
maps derived from satellite imagery were converted to digital raster grids — that is, a "map" that consists of a regular 
grid of cells, or pixels, each of which is square in shape, has an x,y position, and a z-value assigned to it (reflecting 
whatever factor the map represents, such as topography, income, population, etc.). 
The creation of the models in the GIS computer systems makes it easy to change the spatial resolution of the cells 
through mathematical resampling processes. However, in the case of this research, the minimum spatial analysis unit 
size is a pixel approximately 60 by 60 meters, because data derived from Landsat MSS satellite imagery contains no 
details smaller than a 60 by 60 meter spatial resolution (1989 Landsat TM data has a 30 by 30 meter spatial resolution, 
but the 1979 Landsat MSS data has only 60 by 60 meters — thus, any resampling of pixel size below 60 x 60 meters is 
redundant). 
3.2 Specific Digital Outputs 
Digital map products from the research include the following: 
a. maps of land use change in Sisaket and Chachoengsao between 1979 and 1989 using processed Landsat 
satellite imagery, depicting the following land use changes: 
i. conversion of forest to agriculture; 
ii. conversion of agriculture to urban; 
iii. conversion of forest to urban; 
iv. no change. 
b. creation of continuous digital maps of each Province (Sisaket and Chachoengsao) representing sub-Provincial 
spatial variability of four socio-economic factors — 
i. population potential, 1980s; 
ii. income distribution, 1980s; 
iii. proximity/accessibility to markets, 1980s; 
iv. proximity to infrastructure networks, 1980s. 
c. creation of continuous digital maps of each Province representing sub-Provincial spatial variability of four 
environmental/physical factors — 
i proximity to timber resources, late 1970s; 
ii. proximity to water resources, late 1970s; 
iii. soil moisture index, 1980s; 
iv. topography. 
d. creation of predicted land use change maps for 1999 for both Provinces using a statistical tree classification 
with maps of land use change 1979-1989 as the dependent categorical factor and the economic and 
environmental input factors as the independent variables. 
Three separate tree classifications were run: 
i a tree classification predicting future land use change using the four economic input factors; 
ii. a tree classification predicting future land use change using the four environmental input 
factors; 
iii. a tree classification predicting future land use change using all eight economic and 
environmental input factors. 
(A major hypothesis of the research is that using both the economic and the environmental input factors will 
produce more accurate predicted maps of land use change than will either the economic or the environmental 
factors independently.) 
  
436 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B7. Amsterdam 2000.