中文摘要：

针对普通线性回归分析以空间平稳数据为假设,而忽视空间数据局部变化特征的缺陷;该文以上海市西郊的青浦区为案例区,通过构建景观生态质量评价指标体系,采用综合指标评价法和等距离空间分类法,对该区184个行政村的景观生态质量进行评价和等级划分;引入局部线性地理加权回归模型用以评估压力类因素对景观生态质量的影响;对比线性回归模型、地理加权回归模型和局部线性地理加权回归3个模型在解释青浦区行政村景观生态质量空间差异的精度。结果表明：2014年青浦区各行政村的景观生态质量指数的变化幅度为0.03~5.49,呈现西部高,东部低的特征;景观生态质量与8种压力类影响因素的关系随着空间位置的改变表现出局部变化特征,即使同一因素对景观生态质量的影响在方向和大小上均表现不同;通过对比3个模型的解释结果可以判断,地理加权回归模型和局部线性地理加权回归模型在处理空间非平稳性和空间异质性上,其预测精度优于线性回归模型,而局部线性地理加权回归模型在处理由于＂边界模糊＂导致的＂边界效应＂问题上,其模拟的精度优于地理加权回归模型;村域景观生态质量变化具有空间差异性和尺度敏感性。与景观生态质量变化呈正相关的因素依次为：人均耕地面积、居民恩格尔系数和城镇化率;与景观生态质量变化呈负相关的因素依次为：人口抚养比、第二产业占GDP比例、人口密度、聚耕比、距最近城镇中心距离。研究结果可为都市郊区景观生态系统可持续发展和土地空间整治提供科学借鉴。

英文摘要：

The spatial stationary data are usually used in most ordinary linear regression analysis, which often has limitation of ignoring the local variations of spatial data. In this paper, in order to evaluation and classification of landscape ecological quality of 184 administrative villages in the western suburbs of Qingpu district in Shanghai City, we constructed evaluation index system of landscape ecological quality. The comprehensive index evaluation method and the equal distance space classification method were used. A new technique called local linear geographically weighted regression model（LLGWR model） was introduced to evaluate the impacts of landscape ecological quality. Comparison was made among ordinary linear regression model（OLS model）, geographical weighted regression model（GWR model）, and local linear geographically weighted regression model（LLGWR model） for the accuracy of the spatial differences of landscape ecological quality of administrative villages in Qingpu District. The results showed that： the variation range of landscape ecological quality index of administrative villages was 0.03- 5.49 in Qingpu District in the year of 2014. The characteristic of landscape ecological quality in that region was high in the west and low in the east. The relationship between the landscape ecological quality and the influence factors of the eight kinds of pressure types showed the characteristic of local change with the change of the spatial position even with the same factors acted on the landscape ecological quality. Compared among the results of the interpretation of the three models, we concluded that GWR model and LLGWR model were superior to the OLS model in the process of spatial heterogeneity and spatial heterogeneity; LLGWR model was better than the GWR model in dealing with the problem of ＂boundary fuzzy＂ which was caused by ＂boundary effect＂. In addition, village domain changes of landscape ecological quality had differences in space and scale sensitivity. Landscape and ecol