西苕溪流域径流对土地利用变化的空间响应分析

针对流域土地利用变化所引起的径流空间变异问题,论文以太湖上游西苕溪流域为例,基于SWAT模型(Soil and Water Assessment Tool)模拟的不同土地利用情景下月尺度径流过程,通过GWR模型(Geographically weighted regression)在空间上定量评估了土地利用/覆被变化对流域径流过程的影响。结果表明:径流变化在流域空间分布上存在一定非平稳性,其与子流域内面积变化较大的土地利用类型相关性显著,其中城镇用地影响最大,林草地和耕地影响次之。径流变化对城镇用地的空间响应关系表现为由上游到下游逐渐增强,而对林草地和耕地的响应关系表现为从流域上游到下游逐渐减弱。对比发现,多因子GWR模型相对于单因子GWR模型更适合综合分析径流对土地利用/覆被变化的空间响应关系。

Runoff Spatial Responses to Land Use Change in Xitiaoxi River Basin

This study aims to evaluate the impact of land use/cover change (LUCC), especially rapid urbanization, on annual runoff change in Taihu Basin, where monsoonal flood is prone to happen. Xitiaoxi River Basin, one of the main sub-basins of Taihu Lake Basin, was chosen as study area to study the spatial variation of runoff under different land use conditions. A new spatial analysis method called geographically weighted regression model (GWR) was employed to quantitatively evaluate the impact of LUCC on runoff variation. The monthly runoff process was generated by SWAT model (Soil and Water Assessment Tool). The results showed that: 1) All R² and Nash-Sutcliffe efficiency E_NS were above 0.85, and the relative errors |Re| were all less than 15% in the calibration and validation period, suggesting SWAT model perform well. 2) The runoff change was spatially nonstationary, and was significantly correlated with the mostly changed type of land use in the sub-basin, among which the influence of urban land-use had the greatest influence, followed by the forest-grass land and cultivated land. Urban land-use expansion could increase local regional runoff depth by 37.6%-45.2%, while the forest-grass land and cultivated land shrinking could increase local regional runoff depth by 16%-26.2% and 9.2%-15.4% respectively. 3) Spatially, the influence of urban land-use change on runoff depth increased gradually from upstream to downstream in the basin. On the contrary, the impacts of the forest-grass land and cultivated land on runoff process presented decline trend from upstream to downstream. 4) Compared with single-factor GWR model, the multifactorial GWR model had better prediction accuracy and was more suitable to analyze the spatial relationship between runoff and LUCC.