中文摘要：

基于南江县境内244个典型土质滑坡统计样本,利用BP神经网络模型,采用3种不同的方案（基于不同的评价参数）对滑坡体积进行预测。方案一选取坡高、坡度、坡向、高程、植被覆盖率、岩层倾向、岩层倾角等7项评价参数;方案二选取坡高、坡度、坡向、岩层倾向、岩层倾角等5参数;方案三选取坡高、坡度、坡向等3参数。研究结果表明：3种方案建立的BP神经网络模型都具有较高的可靠性,其预测结果都可以较好地逼近真实滑坡体积值,BP神经网络能有效应用到滑坡体积预测中;3种方案预测值与实际值基本吻合,且两者间的相关系数分别为0.87083,0.90826,0.86119,评价参数的合理选择对滑坡体积预测的准确性有着重要的影响;方案二的相关系数最高,其预测准确性最好,这表明坡高、坡度、坡向、岩层倾向、岩层倾角是影响滑坡体积的重要因素,植被覆盖率和高程为其次要影响因素。

英文摘要：

Thousands of soil landslides induced by heavy rainfall destroy a large number of villages and farmland, causing serious loss of life and property, so it is of important practical significance to forecast volume of landslide group for ensuring safety of lives and properties of local residents. Based on 244 typical soil landslides in Nanjiang, BP neural network is used to forecast landslide volume by employing three different plans （based on evaluation of different parameters）. In scheme 1, slope height, slope, aspect, elevation, vegetation coverage, strata inclination and strata dip angle are selected as the evaluation parameters to forecast landslide volume. In scheme 2, slope height, slope, aspect, strata inclination and strata dip angle are selected as the evaluation parameters to forecast landslide volume; In scheme 3, slope height, slope, aspect are selected as the evaluation parameters to forecast landslide volume. The results show that all BP neural network models have highly reliable, and the prediction results are in good approximation with the true values. BP neural network can be effectively applied to predict landslide volumes. The correlation （R） between predictive values and actual values is respectively 0. 87083,0.90826 and 0.86119. Reasonable choice of evaluation parameters has an important influence on accurate prediction of landslide volume. Scheme 2 has the highest correlation and the best predictive accuracy, means that slope, aspect, strata inclination and strata dip angle are important factors affecting landslide volume, vegetation coverage and elevation are the secondary factors.