基于人工神经网络模型的福建南平市滑坡危险性评价

Landslide risk assessment in Nanping City based on artificial neural networks model

滑坡灾害持续影响着人民生命财产安全和地区社会经济可持续发展,滑坡危险性评价能够为防灾减灾和区域规划提供有效的理论依据。以福建省南平市为研究区,区内1 711个历史滑坡灾害点,选择高程、坡度、坡向、曲率、地质岩性、土壤类型、降雨、水系、土地利用类型、公路和铁路共11个影响因子构成基本评价体系。使用Spearman相关系数对各因子进行共线性分析。基于1 711个滑坡样本和1 711个随机选取的非滑坡样本数据,利用人工神经网络模型对研究区进行了滑坡危险性评价,并利用混淆矩阵和接收者操作特征曲线(ROC)对模型进行验证。结果表明:混淆矩阵精度84.91%,ROC曲线下面积AUC值0.93,说明模型具有较高精度和预测率。使用自然间断法将滑坡危险性分为5个等级,结果表明研究区内危险性最高地区位于延平区和浦城县,顺昌县和松溪县次之,其余地区多为低危险区和较低危险区。研究结果可为当地区域规划和防灾减灾工程提供一定的理论依据和科学指导。

Landslide hazards continuous sequence the safety of people’s lives and property and the sustainable development of regional society and economy,and landslide risk assessment can provide an effective theoretical basis for disaster mitigation and regional planning.A total of 1 711 historical landslide hazard sites around Nanping City were obtained,and 11 impact factors,including elevation,slope,aspect,curvature,geological lithology,soil type,rainfall,water system,land use,road and railway etc.were selected.The covariance analysis of each factor was carried out using the Spearman correlation coefficient.Based on the data of 1 711 landslides and 1 711 non-landslides,an artificial neural network(ANN) model was used to evaluate the landslide risk in the study area,and the model was validated using a confusion matrix and receiver operating characteristic(ROC) curve.The results show that the confusion matrix accuracy was 84.91% and the area under the ROC curve(AUC) was 0.93,indicating that the model has high accuracy and prediction rate.The landslide risk index was classified into five classes by natural break method,and the results show that the highest risk areas in the study area locate in Yanping District and Pucheng County,followed by Shunchang County and Songxi County,and the rest of the areas were mostly low-risk areas and lower-risk areas.The results of the study can provide theoretical basis and scientific guidance for local regional planning and disaster mitigation.