基于GIS和ANN的中国区域贫困化空间模拟分析

Spatial Simulation Using GIS and ANN for Regional Pauperization in China

中国区域贫困化产生的主导因素经历了制度因素、政策因素到自然因素的变化。本文在定量分析中国区域贫困化与自然要素关系的基础上,利用GIS和ANN(人工神经网络)技术模拟了1999年中国区域自然贫困化的空间分布。研究结果表明:地形因素如地形高程、地形破碎度、平均坡度与区域贫困化有显著的负相关关系。中国区域自然贫困化空间分布格局具有明显的空间集聚特性,自然致贫指数较高的区域集中分布在西部干旱和高寒地区、西南喀斯特地区、中部的燕山、太行山、秦巴山地。ANN模拟结果与现在中国主要贫困县分布相比较,其空间构型大体一致。

This paper analyzes the causes of regional poverty and anti-poverty process in China. The causes that lead to regional poverty includes regime factors, policy factors and natural environmental factors, and nowadays natural environmental factors become the main causes of regional poverty in China. Using regression model, the relationship between regional poverty and natural environmental factors is established and results show that relief features such as altitude, surface fragmentation are negative correlation, while some socio-economic factors are positive correlation with dependent variable-per captia gross domestic product (GDP). Based on the regression results, we develop the BP model with 6 neurons in input layer and 1 neuron in output layer. Have trained and tested BP model with training data, the natural impoverishing index is evaluated. Compared with the distribution of poor counties, ANN modeling results are more explicit and more fine, which are consistent with the spatial distribution of poverty counties in China at general. According to the geo-spatial analysis, pattern of natural poverty exhibits the spatial aggregation. The higher natural impoverishing index is mainly distributed in the western arid and cold area, southwestern karst area, Yanshan Mountainous area, Taihang Mountainous area and Qin-Ba Mountainous area. ANN model simulation is a valid approach to study pauperization.