【摘要】 修正的通用土壤侵蚀方程（revised universal soil loss equation,RUSLE）模型中的植被覆盖管理措施因子（C）可用来表示植被覆盖对土壤的防蚀作用,是评价土壤侵蚀的关键参数之一,而区域尺度高质量时间序列C因子的合理估算及空间分布特征是预测区域土壤侵蚀动态变化的重要环节。因此,及时准确地掌握区域尺度长时间序列的C因子,对研究土壤侵蚀动态变化与植被的关系至关重要。选择南京市1988—2013年10期遥感影像,基于后向反射（back propagation,BP）神经网络和遥感数据反演植被结构因子叶面积指数（leaf area index,LAI）,构建LAI与C因子的量化耦合模型,通过¹³⁷Cs同位素示踪技术获取C因子的实测值,验证并探讨反演模型的精度。结果表明:（1）基于归一化植被指数（normalized difference vegetation index,NDVI）计算的传统植被覆盖度在整体上比基于LAI计算的植被方向性覆盖度偏大,得到的C值与实测值相比整体上偏小,均方根误差（root mean square error,RMSE）为87. 829%;而基于LAI计算的C值与实测值接近,RMSE为30. 017%,能更好地反映实际的植被结构信息;（2）C值大于0. 3的区域主要分布于建筑物较为密集、植被稀疏且植被结构简单甚至无植被覆盖的市区;C值小于0. 05的区域主要分布在植被密集且植被结构复杂的丘陵山区,南京市C值的分布与植被覆盖和土地利用类型关系密切。（3）从全市整体来看,1988—2013年C值小于0. 05的抵抗土壤侵蚀能力较强的区域面积先由南京市总面积的15. 66%（1988年）减小到9. 43%（2006年以前）,后逐渐增大到12. 07%;C值大于0. 3的抵抗土壤侵蚀能力较弱的区域面积先由南京市总面积的7. 29%缓慢增大到9. 22%（2002年）,后迅速增大到12. 31%（2002—2006年）,然后缓慢减小至11. 77%。所提出的基于BP神经网络和LAI反演的长时间序列C因子估算方法是可靠的,可为区域尺度土壤侵蚀定量遥感监测提供新途径。更多还原

【Abstract】 The vegetation cover and management factor（C）in the revised universal soil loss equation（RUSLE） model is used to indicate the effect of vegetation cover on soil erosion,and is one of the key parameters for evaluating soil erosion.The precise estimation and the spatial distribution characteristics of the long-term sequence of C factor at the regional scale are important for the prediction of dynamics of regional soil erosion. Therefore,timely and accurate grasping of the long-term sequence of C factor at the regional scale is very important for studying the dynamic relationship between soil erosion and vegetation. The remote sensing image in Nanjing City from 1988 to 2013 was selected and used. The quantification coupling model between leaf area index（LAI） and C factor was constructed based on BP neural network and the vegetation structure factor LAI inversed by remote sensing. The field measurement of C factor was obtained by ¹³⁷Cs isotope tracer technique,and the accuracies of the inversion models were verified. The results showed that:（1）The traditional vegetation coverage based on normalized difference vegetation index（NDVI）was generally larger than that based on LAI. The C value was generally smaller than the measured value,and the RMSE was 87.829%. While the C value calculated by LAI was close to the measured value,and the RMSE was 30.017%. Therefore,the C value calculated by LAI can reflect the actual vegetation structure information better than by NDVI.（2）The area with C value greater than0.3 was mainly distributed in the urban areas with dense buildings,sparse vegetation and simple vegetation structure or even no vegetation cover. The area with C value less than 0.05 was mainly distributed in the hills and mountainous regions with dense vegetation structure. The distribution of C value in Nanjing City was closely related to vegetation cover and land use types.（3）Over the whole city where the C values were less than 0.05,during the period from 1988 to 2013,the area with strong resistance to soil erosion firstly decreased from 15.66%（1988）to 9.43%（before 2006）and then gradually increased to 12.07%. For the region where C values were greater than 0.3,the area with weaker resistance to soil erosion firstly slowly increased from 7.29% to 9.22%（2002）,then it quickly increased to 12.31%（2002-2006）and then slowly decreased to 11.77%. Therefore,it is feasible to retrieve the long-term sequence C factor based on BP neural network and LAI proposed in this paper,which provides a new way for quantitative remote sensing monitoring of soil erosion on regional scale.更多还原