Occurrence probability assessment of earthquake-triggered landslides with Newmark displacement values and logistic regression: The Wenchuan earthquake, China
Occurrence probability assessment of earthquake-triggered landslides with Newmark displacement values and logistic regression: The Wenchuan earthquake, China
复制标题
DOI：
10.1016/j.geomorph.2016.01.004
复制
发表时间：
2016-04
期刊：
影响因子：
3.9
通讯作者：
Ying Wang;Chongzhen Song;Q. Lin;Juan Li
中科院分区：
文献类型：
--
作者：
Ying Wang;Chongzhen Song;Q. Lin;Juan Li
纽马克位移模型已被用于预测地震引发的滑坡。逻辑回归（LR）也是一种常用的滑坡危险性评价方法。将Newmark位移模型与LR模型相结合，应用于2008年5月12日汶川8. 0级地震后的汶川县和北川县，建立了基于机理的滑坡发生概率模型，提高了预测精度。选取汶川县1904个滑坡点和3800个随机非滑坡点作为训练数据集。采用Newmark模型，得到了30 × 30 m网格的永久位移分布。四个因素（DN，地形起伏，排水沟和道路的距离）被用作LR的自变量。然后，得到一个组合模型，其中汶川县的AUC值为0.797。以北川县617个滑坡点和非滑坡点作为验证数据集，AUC = 0.753。该方法也可应用于其他地区的地震滑坡。
The Newmark displacement model has been used to predict earthquake-triggered landslides. Logistic regression (LR) is also a common landslide hazard assessment method. We combined the Newmark displacement model and LR and applied them to Wenchuan County and Beichuan County in China, which were affected by theMs.8.0 Wenchuan earthquake on May 12th, 2008, to develop a mechanism-based landslide occurrence probability model and improve the predictive accuracy. A total of 1904 landslide sites in Wenchuan County and 3800 random non-landslide sites were selected as the training dataset. We applied the Newmark model and obtained the distribution of permanent displacement (Dn) for a 30 × 30 m grid. Four factors (Dn, topographic relief, and distances to drainages and roads) were used as independent variables for LR. Then, a combined model was obtained, with anAUC(area under the curve) value of 0.797 for Wenchuan County. A total of 617 landslide sites and non-landslide sites in Beichuan County were used as a validation dataset withAUC= 0.753. The proposed method may also be applied to earthquake-induced landslides in other regions.