滑坡危险性等级区划方法研究

以四川省地理环境条件为研究背景，根据相关单位提供的资料，对四川省滑坡危险性等级进行了区划研究。将地形地貌、地层岩性、断裂构造、河流水系、降雨量、地震烈度等6项作为滑坡的主要影响因素，采用模式识别方法进行滑坡危险性等级区划，对算法中所涉及到的危险系数计算公式、因子权重分析、阈值选取等进行了一系列控制试验，验证了算法的可行性以及降低算法中存在的不确定性。采用“识别率”和“改变率”2个准则来判断“分类结果作为新的训练集”，即RTS试验的收敛性，从而给出识别率高、改变率稳定的分类结果，以及能合理反映识别结果的最佳参数。通过3次逐级识别分类，将四川省滑坡危险度划分为7个等级，区划结果与实际滑坡发生情况吻合。本文方法同样适用于其它地区的滑坡危险性等级区划。

Study on the Method of Landslide Hazard Rank Division

On the basis of a case study regarding the geographical conditions of Sichuan, according to mate- rials provided by relevant units, this paper studies the landslide hazard rank division in Sichuan. Topogra- phy, stratum lithologic character, faulted structure, average annual rainfall, water system, seismic inten- sity are taken as landslide major effect factors. Pattern recognition is used to carry out the landslide hazard rank division. A series of control tests, such as weight of factor and selection of threshold, are conducted for danger coefficient formula. Feasibility of method is validated and uncertainty is reduced. Concept of“i- dentifying rate” and “change rate” are proposed and RTS test is used for the method, which takes the high identifying rate and steady change rate as classifying standard. The optimal parameters are selected which can reflect the recognition results reasonably. Altogether three times recognition are taken to realize seven rank divisions in landslide hazard of Sichuan. The results from pattern recognition method coincide well with landslide occurrence. The method also can be used for other region＇s landslide hazard rank division.