Land-use regionalization based on landscape pattern indices using rough set theory and catastrophe progression method

Abstract

This study presents the rough set theory and catastrophe progression method to regionalize the land-use change and to analyze the land ecological process. It uses the land-use database of Yixing City of Jiangsu Province, an urbanized and industrialized city in Yangtze River Delta of China, as an exemplification. The study area is divided into six kinds of land-use types according to the national standard of land-use classification. It includes arable land, garden, woodland, urban–rural construction land, water, and unused land. The six kinds of land-use types are formed into their corresponding landscape types in the scale of 1:10,000 by the aid of ArcGIS9.3 software of ESRI. In ArcGIS9.3, the landscape pattern indices are calculated by using Fragstats (raster version 3.3) software. Based on these landscape pattern indices, an integrated indicator system of landscape regionalization of land use in Yixing was established, and land-use regionalization models are set up using the catastrophe theory. Rough set theory is introduced to avoid the subjectivity in the indicator’s importance in catastrophe models. The hidden rule among the raw data is acquired by knowledge reduction of the data mining in the rough set theory. In the process, indicators needed to be arranged according to the computed importance of an attribute without considering the determination of weight function. This greatly avoids the subjectivity in the process of weight factor determination. The zoning of land use based on landscape indices finally is made by the multi-indicator integrated catastrophe progression method. According to these indices, Yixing is divided into four grading land-use zones when the rough set and catastrophe progression methods are combined. The zones include high-, medium-, low-, and weak-intensity zones, indicating that land use primarily varies the landscape pattern. With the increase of water and forest area proportion, the human disturbance to land system wanes; patch fragmentation reduces; patch shape complexity enhances; and landscape diversity decreases. Moreover, it can mostly avoid the subjective evaluation in artificially determining factor weights by using rough set theory. It makes the zoning results more objective and exact.