A study for the regional spatial pattern of agricultural land use by remote sensing and multiple choice model

This article focuses on developing a predictive method to simulate the distribution of agricultural land use patterns based on remote sensing and a multiple choice logit model. A multiple choice model is developed to determine land use options such as rice cropland, maize cropland, wheat cropland, soybean cropland and so on. The independent variables are grouped under two categories: basic environment variables and socioeconomic variables. The basic environment variables include natural factors and anthropogenic factors. Natural factors include basic conditions mainly to estimate crop productivity, such as temperature, precipitation, soil physical and chemical properties, and terrain variables. A GIS-based EPIC (Erosion Productivity Impact Calculation) model, which is developed to simulate land productivity for different crops in each grid at regional level, is embedded in multiple choice logit model. Remote sensing is used to collect the sample points of agricultural land use. Examinations show that the results in our test are acceptable.