2016年香港大屿山天然山坡滑坡易发性数据

本数据以栅格数据格式展示了2016年香港特别行政区大屿山的天然山坡滑坡易发性。本数据基于天然山坡滑坡编录以及包括高程、坡度、坡向、平面曲率、剖面曲率、平面曲率、土地覆盖、植被覆盖度、岩性、距断层线距离和地形湿度指数在内的滑坡孕灾环境因子，使用结合通道拓展化预训练CNN与随机森林的滑坡易发性评价集成模型（Ensemble Model Combining Channel-expanded Pre-trained Convolutional Neural Networks and Random Forest，CPCNN-RF）计算得到。天然山坡滑坡编录所记录发生于1992年至2007年的历史滑坡用于模型训练，2008年至2016年的历史滑坡用于易发性评价结果的精度评价，AUC精度和总体精度分别为0.95和88.78%。本数据可直接使用GIS软件开并进行后续处理，并可作为滑坡风险管理研究的基础数据。

This dataset presents the natural hillslope landslide susceptibility of Lantau Island, Hong Kong Special Administrative Region in 2016 in raster data format. It was calculated using an ensemble model for landslide susceptibility assessment, namely the Ensemble Model Combining Channel-expanded Pre-trained Convolutional Neural Networks and Random Forest (CPCNN-RF), based on natural hillslope landslide inventory data and landslide-conducive environmental factors including elevation, slope, aspect, plan curvature, profile curvature, plan curvature, land cover, vegetation coverage, lithology, distance to fault lines, and topographic wetness index. Historical landslides recorded in the inventory from 1992 to 2007 were used for model training, while those from 2008 to 2016 were employed for accuracy evaluation of the susceptibility assessment results. The Area Under the Receiver Operating Characteristic Curve (AUC) score and overall accuracy reached 0.95 and 88.78%, respectively. This dataset can be directly used with GIS software for further processing, and can serve as fundamental data for landslide risk management research.

搜集汇总

背景与挑战

背景概述

该数据集是2016年香港大屿山天然山坡滑坡易发性栅格数据，空间分辨率为30米，使用集成模型CPCNN-RF基于历史滑坡编录（1992-2007年用于训练，2008-2016年用于评估）和多种诱发因子（如高程、坡度、土地利用等）生成。数据精度较高，AUC为0.95，总体准确率达88.78%，适用于滑坡风险管理的基础数据应用。