目前关于土地需求量的预测方法往往忽视地理空间因素。本文提出了一种融合地理空间指标的土地需求量的多因素预测方法,通过在需求量预测计算中引入地理标准距离、标准差椭圆、Moran' s I指数等与土地利用类型的空间分布和空间形态相关的因素,探讨基于地理空间指标进行土地需求量预测的可行性和准确性。以佛山市南海区历年的土地利用调查数据为基础进行案例研究,结果表明:相对于地理经济指标,引入地理空间指标的土地需求量预测结果准确率更高。但由于数据量的限制,本文所得的结果需要收集更多数据采用多种统计分析方法进一步加以研究验证。

The paper puts forward a method to forecast land demand by fusing several geo-spatial indicators. Traditionally, economic and social factors were regarded as the main influencing factors in forecasting land demand, hence the effects of spatial factors were neglected by researchers. However, for each type of land use, its spatial distribution and spatial shape are bound up with the other type of land use by interacting each other, so spatial factor should be introduced to forecast land use demand as a key ingredient. Several computable factors called geo-spatial factors can be used to fulfill the requirement, such as geographic standard distance, standard deviation ellipse parameters and spatial autocorrelation coefficient. These geo-spatial factors can reflect the changing trend of each type of land use and disclose some interior rule of spatial distribution and spatial spreading. It is necessary to fuse these geo-spatial factors into forecasting land use demand. Considering the influence of geo-spatial factors, the paper gives a land use demand forecasting method by fusing economic factor and geo-spatial factor. The method adopts multiple linear regressions to create a linear relation between land use demand quantity and multi-factor value from multi-years data and finally figures out target year's forecasting result. To verify the method, the paper makes a case study based on land use investigation data (from 2003 to 2009) and economic factor data (from 2002 to 2009) of Nanhai in Foshan city. By computing and analyzing, the result shows that it is more veracious to fuse geo-spatial indicators and economic indicators into land use demand than to use economic indicators only. It can be concluded that geo-spatial indicators have closer correlation with land use status and can play a more important role in forecasting land use demand. But restricted by basis data, the result in this paper should be verified by collecting more integrated data and adopting other statistic computing methods. The further work will focus on analyzing geo-spatial indicators' internal mechanism related to land use changes and selecting more reasonable indicators as forecasting factors.