基于最大熵模型的华蓥山林地浅表层滑坡风险析因

Attribution of Superficial Landslide Risk of Forestland in Huaying Mountains Based on MaxEnt Model

【目的】确定环境变量对林地浅表层滑坡风险预测的相对贡献率,明确影响浅表层滑坡风险的关键植被因素及其减灾区间,探明植被因素与非植被因素对浅表层滑坡风险的耦合效应,为林地浅表层滑坡风险评价和减灾决策制定提供科学依据。【方法】以华蓥山林地为研究对象,选取17个浅表层滑坡影响因子,采用最大熵模型进行林地浅表层滑坡风险模拟,输出各因子对林地浅表层滑坡风险预测的相对贡献率,对比分析考虑或不考虑植被因素条件下林地浅表层滑坡风险对各因子的响应变化。【结果】1)模型精度受试者工作特征曲线检验结果显示,不考虑植被因素的情况下,模型模拟精度为0.887,达到很准确的精度水平;考虑植被因素的情况下,模型模拟精度提升3.1%,为0.915,达到极准确的精度水平。2)工程地质岩组、蓄积量、距断层距离、地形起伏度、高程、绿色比值植被指数、平面曲率和林分类型8个因子对浅表层滑坡风险预测的累计贡献率达80%,其中植被因素对林地浅表层滑坡风险预测具有重要作用,主要体现在蓄积量、植被覆盖度和林分类型3方面。3)植被因素的存在造成浅表层滑坡风险对平面曲率、坡向、高程变异系数、坡度变率和坡面曲率5个变量的响应发生变化:对平面曲率、高程变异系数和剖面曲率所产生的浅表层滑坡风险起削弱作用,削减幅度分别为4.9%、5.9%和8.1%;对坡度变率所产生的浅表层滑坡风险起加剧作用,加剧幅度为10.9%;对坡向所产生的浅表层滑坡风险具有正反2方面作用,加剧和削减幅度分别为12.8%和6.4%。【结论】MaxEnt模型用于林地浅表层滑坡风险模拟具有较高模拟精度,能够直观表达浅表层滑坡风险对各影响因子的响应;利用MaxEnt模型预测林地浅表层滑坡风险时,除了地质、地形、地貌、土壤等常规影响因素外,植被因素也是关键环境变量,其对模拟精度具有重要贡献;植被因素的存在整体上不改变浅表层滑坡风险对其他影响因子的响应趋势,但对于某些非植被因素的极端值所产生的浅表层滑坡风险具有重要影响,呈现出耦合效应,既可能加剧也可能削弱浅表层滑坡风险。

【Objective】The aim of this study was to determine the relative contribution rate of environmental variables to forestland superficial landslide risk,to clarify the key vegetation factors affecting the superficial landslide risk and their disaster reduction range,and to reveal the coupling effect of vegetation and non-vegetation factors on the superficial landslide risk.This research can provide important theoretical support for landslide risk assessment and disaster mitigation decision.【Method】Taking the forest land in Huaying Mountains as the research object,17 environmental variables were selected,and the MaxEnt model was used to determine the relative contribution rate of environmental variables to the prediction of superficial landslide risk of the forest land,the response changes of superficial landslide risk to various factors were analyzed based on the existence of vegetation factors.【Result】1)The ROC test results of the model accuracy showed that the simulation accuracy of the model was 0.887 without considering the vegetation factor,which reached a very accurate level of accuracy,while in the case of considering the vegetation factor,the simulation accuracy of the model was 0.915,which reached an extremely accurate level of accuracy,and the accuracy was improved by 3.1%.2)The cumulative contribution rate of engineering geological rock group,forest volume,distance to fault,terrain relief,elevation,green-red vegetation index,plane curvature,and stand type to the prediction of superficial landslide risk reached 80%.Vegetation factors played an important role in the superficial landslide risk prediction,mainly reflected in forest volume,vegetation coverage,and stand type.3)The existence of vegetation factors changed the response of superficial landslide risk to five variables including plane curvature,slope aspect,elevation variation coefficient,slope variable,and profile curvature.Vegetation factors weakened the superficial landslide risk caused by plane curvature,elevation variation coefficient,and profile curvature,and the reduction rates were 4.9%,5.9%,and 8.1%,respectively.Vegetation factors aggravated the superficial landslide risk caused by slope variability,with an aggravating rate of 10.9%,and it had both positive and negative effects on the superficial landslide risk generated by the slope aspect,with the aggravation and reduction rates of 12.8%and 6.4%,respectively.【Conclusion】Our results demonstrated that the MaxEnt model had high simulation accuracy for the simulation of superficial landslide risk of forest land,and it expressed the response of the superficial landslide risk to various influencing factors intuitively.When using this model to predict the superficial landslide risk of forest land,in addition to the conventional influencing factors such as geology,topography,landform,soil,etc.,the vegetation factor was also a key environmental variable in the model simulation,which had an important contribution to the simulation accuracy.The existence of vegetation factors generally did not change the response trend of the superficial landslide risk to other influencing factors,but would have an important impact on the superficial landslide risk caused by the extreme values of some non-vegetation factors,showing a coupling effect,which might aggravate or weaken the superficial landslide risk.