Computer-aided method for land classification based on relief morphometry analysis

Abstract

There are suggested the computer-aided methods for classification of placor lands using Geographical Information Systems (GIS) and a neural expert system. Quantitative relief characteristics are adopted as a basis for recording. In the GIS ArcGIS 10 system, an electronic map of placor lands located in Maslyanino District of Novosibirsk Region was created, a database consisting of topographical and soil maps was formed. By the method of ANUDEM interpolation was developed a topologically correct digital relief model containing maps as follows: hypsometric; angle and exposure of slopes; plan, profile and total curvature of the Earth’s surface; cumulative runoff. The boundaries of elementary surfaces, being homogeneous morphological formations, were established. Parameters reflecting the intensity of erosion processes (SPI index) were taken into consideration. The main point of the classification is to attribute the elementary surfaces to a certain group of lands as to a complex of traits, for which particular scales containing characteristics of relief, soil cover, drainage of a territory, a degree of the erosion development have been created. A knowledge base was formed to train multi-layer neural network using GIS databases and particular scales; a neural network was trained. By means of the neural expert system, classification and topology of lands were conducted. Placor lands are located on flat and weakly expressive sites, and characterized by the following traits: plan curvature of the Earth’s surface of 0 to 0.03, profile of 0 to 0.15, and total of 0 to 0.22; angles of relief incline less than 1.5^o; horizontal dismemberment of relief less than 0.5 km/km²; vertical dismemberment of relief less 5 m; SPI index changes from - 13.80 to - 6.47.