CLUE-S模型对村镇土地利用变化的模拟与精度评价

利用航片、IKONOS和Quickbird影像得到研究区3期（1991、2001和2009年）土地利用历史数据，运用CLUE S模型，基于1991、2001年的土地利用数据对2009年土地利用格局进行模拟预测，并将模拟结果与2009年真实土地利用数据进行比较。类型水平上，选择ROC曲线统计和偏离度指数分别对各地类的Logistic回归拟合精度和CLUE S模型模拟精度进行评价；景观水平上，采用景观指数和Kappa指数系列方法，从综合预测能力、景观格局、数量和空间位置等方面对CLUE S模型的模拟精度进行全面评估。结果表明:①CLUE S模型对各土地利用类型的模拟精度均较高，各地类的Logistic回归的拟合精度随着模拟分辨率的提高而逐渐增加；②随着模拟时间的缩短，CLUE S模型对整体景观格局的模拟精度提高；③该模型对土地利用数量的模拟精度明显优于其对空间位置的模拟精度。总体而言，CLUE S模型在村镇尺度的模拟效果良好，预测精度较高，但在空间位置和景观格局精度方面还有待提高

SIMULATION AND ACCURACY ASSESSMENT OF VILLAGE LAND USE CHANGE BASED ON CLUE-S MODEL

Based on land use historical data of three periods extracted from high-resolution aerial photos,IKONOS and Quickbird images,we applied CLUE-S model to forecast the land use pattern of 2009 based on the 1991 and 2001 land use data,and compared the simulated results with the real land use of 2009.On class level,we chose ROC curve and deviation degree index to evaluate the accuracy of logistic regression and CLUE-S model,respectively.On landscape level,we selected landscape metrics and Kappa index series to comprehensively assess the accuracy of CLUE-S model with respect to integrated predictive ability,landscape pattern,quantity and spatial location.The results showed that the CLUE-S model had satisfactory precision for each land use type,and showed higher fitting precision of logistic regression of each land use type with the spatial resolution increasing.With simulating time decreasing,the precision of CLUE-S model for landscape patterns enhanced.The predicted precision of quantity was obviously better than spatial location.As a whole,the comprehensive simulation effect of CLUE-S model was favorable,but the predicted precision of spatial location and landscape pattern should be improved.