基于空间自回归CA模型的城市土地利用变化模拟与预测

Simulation and Prediction of Urban Land Use Change with Spatial Autoregressive Model Based Cellular Automata

该文构建了一种基于空间自回归的地理元胞自动机(CA)模型——SARCA模型,该模型能够较好融合地理系统模拟中的空间自相关特性,且获取的CA参数具有明确的物理意义。以1995-2015年上海城市土地利用为案例,验证了该模型的有效性。CA参数和城市土地转化概率表明,上海市外环线对于CA参数的贡献相比其他空间变量具有压倒性优势,到外环线距离越近则土地发展为城市的概率就越大。将基于Logistic回归的CA模型(LogCA)作为比较对象,模拟同期上海全域城市土地利用变化过程。CA规则表明,SAR在赤池信息量准则(AIC)、残差的描述性统计量和空间自相关指标等方面均优于Logistic回归。同时,SARCA模型在2005年和2015年的土地利用模拟结果总体精度分别为86.3%和82.0%,均优于LogCA模型的模拟结果(总体精度分别为79.8%和76.3%)。

This paper presents a spatial autoregressive （SAR） model based cellular automata model （ SARCA） to simulate complex urban land use change by incorporating spatial autoct rrelation. The CA paraimeters retrieved by the SAR model have clear physical meanings closely associated with urban land use. The proposed CA model has been used to simulate urban land use diange of Shanghai from 1995 to 2015. CA parameters and land conversion probability maps show that the outer ring expressw Tay of Shanghai has the most contribution to CA transition rules. This indicates that the closer a land parcel to the outer ring expressway the higher its probability being converted from non - urban to urban. The Logistic regression based CA （ LogCA） model as a comparative model has also been implemented to the same study area. The fitting performance of CA transition rules shows that SAR is better than logistic regression as reflected by the Akaike Information Criterion （ AIC ） and thestat is tics and Morans Ifor residuals. T he simulation results demonstrate that the overall accuracy of SARCA is 86 3% in 2005 and 82 0% in 201 5, indicating that the proposed CA model has a better performance in simulating urban land use change than the Log CA model （79 8 % in 2005 and 76 3% in 2015）.