THE USE OF A KNOWLEDGE-BASED DECISION RULE COMPUTER PROGRAM IN THE INTER-ANNUAL LAND COVER AND LAND USE CHANGE ANALYSIS OF THE UPPER MAGAT WATERSHED

摘要

Radiometric correction was difficult to perform on imageries of mountainous areas where atmospheric and climatic data were difficult to obtain, because of the absence of weather stations. It was a general knowledge that the classification of the non-radiometrically corrected imagery resulted to misclassifications, low accuracy and inconsistencies. From these problems, the objectives of this study were drawn, namely: to utilize a knowledge-based decision rule program to improve the accuracy and assist in the pre-analysis of non-radiometrically corrected land cover and land use data. The study site was the Upper Magat Watershed, Province of Nueva Vizcaya, Philippines. Eight sets of Landsat TM data taken from 1988 to 1998 were used as inputs to the project. The raw imageries were geometrically rectified, and classified image mosaics were produced. A Turbo Pascal Program was created with a set of knowledge-based decision rule criteria. The classified imageries were used as input to the program and the result was a new set of classified imageries. Cross-tabulation of the classified imageries and those refined by the computer assisted program was performed. Confusion matrices of the 1998 classified imageries were generated. Results of the cross-tabulation showed that the classified imageries refined by the computer program had no incidence of invalid change results unlike the classified imageries produced without the assistance of the knowledge-based computer program. Also, its classification accuracy was higher than that of the imageries produced without the assistance of the computer program. We therefore conclude that the use of knowledge-based decision rule computer programs to assist the standard classification procedures improved the accuracy of the land cover and land use data and the consistency of the land cover and land use change results.