基于空间自回归和地理加权回归模型的佛山市中心城区河网水系演变驱动分析

为定量探究城市河流变化驱动成因,以2005～2010年佛山市中心城区为例,分析了河网水系演变及其驱动因子,根据河网水系变化与其驱动因子间一对多的映射关系,借助空间自回归和地理加权回归模型,分别从整体和局部分析了两者间的统计关系,结果表明:(1)末级河流长度减少量占各级河流总变化量约92.3%,变化最为显著,而城镇用地对水田、工业用地对水域的侵占以及农业活动则是影响末级河流的主要驱动因子;(2)整体来看,末级河流受建设用地扩张,尤其是工业用地扩张的负面影响程度最大;局部来看,各驱动因子的负面影响程度在不同空间位置上存在差异,以水田-城镇用地因子为例,其在罗村、老城区和桂城的交界区域以及南庄、罗村和老城区的交界区域负面影响程度较大;(3)空间自回归模型对区域河网水系变化与其驱动因子间的关系有整体、直观的把握,地理加权回归模型则能够描述驱动因子影响的空间非平稳性,有利于获取局部信息,结合两种模型的特点能够更加全面地刻画河网水系演变的驱动成因。

Analysis of River Network Changes Based on Spatial Auto-regressionand Geographic Weighted Regression Model

In order to quantitatively explore the causes of urban river changes, this paper took the central urban district of Foshanas an example to analyze the evolution of river network and its driving factors. According to the one-to-many mapping relationshipbetween the change of river network and its driving factors, the spatial auto-regression and geographic weighted regression modelswere used to analyze the statistical relationships wholly and partially. The results show that: (1) The IV-order rivers have the mostdramatic changes and its length reduction account for about 92.3% of the total-order rivers changes. And the encroachment of urbanland on paddy field and industrial land on waters, the agricultural activities are the major driving forces affecting the IV-orderrivers. (2) As a whole, the IV-order rivers are deeply affected by the expansion of construction land, especially the expansion ofindustrial land. In the local view, the negative effects of each driving factor vary in different spatial positions. Taking paddy fieldand urban land-use factors as an example, its negative effects are significant on the border areas among Luocun, old town andGuicheng, as well as on the border areas among Nanzhuang, Luocun and old town.(3) The spatial auto-regression has advantage onthe relationship between the changes of river network and its driving factors in a whole, while the geographic weighted regressionmodel can describe the spatial nonstationarity of the influence of driving factors, and it is beneficial for obtaining local information.The two combined models can more fully reveal the causes of the evolution of river network.