|
Title:
Landslide susceptibilityprediction and mapping in tropicaland subtropicalareas |
Abstract:
Predicting and mapping susceptibility of landslide geo-disastersare critical in subtropi-cal and tropical areas as strong rainfall often precipitates the occurrence of landslides, debris-flows or mudflows, causing damages and property losses to human society. This paper is to present our recent county-level and province-level studies on landslides in subtropical and tropical areas taking Jiangxi and Guangdong as examples. Remote sensing data including digital elevation model (DEM) and multiyear autumn satellite images, geological maps, monthly rainfall, roads and settlements data were employed for the research. After data processing including vectorization and digitization of the geological maps (e.g., stratigraphic lithology, faults) and buffering of faults, rivers and roads, etc., integrated datasets consisting of stratigraphic lithology, elevation, slope, aspect, interpolated rainfall layers, buffering zones of faults, roads and rivers, were prepared for the related counties. Field observation data, i.e., landslide events and samples, were utilized to constitute randomly training sets (70%) and validation sets (30%). Different single machine learning algorithms, e.g., logistic regression (LR), random forests (RF), support vector machines (SVM), gradient-boosting decision tree (GBDT), Light Gradient Boosting Machine (LGBM), and hybrid models such as weight of evidence (WoE)-RF, WoE-SVM, WoE-LR, information value (IV)-based LR, RF-LGBM, etc., were applied to predict the landslide susceptibility after training using the training sets in the counties Poyang, Guixi, Chongren, Ruijin in Jiangxi and Wuhua in Guangdong, and the whole province of Jiangxi, and all the produced models were validated by the validation sets. The results show that these different machine learning models are able to produce reliable prediction of landslide susceptibility with an AUC (area under the Receiver Operating Characteristic (ROC) curve) of 0.89-0.97 and an overall accuracy (OA) of 82%-92%, and the hybrid models may derive prediction results with higher reliability. Hence, the research may provide an operational reference for predicting the landslide hazard in the tropical and subtropical areas and serve for disaster reduction and prevention action of the local governments in South China.
|
Biography:
Weicheng Wu, with a PhD in Environmental Geography from the University of Paris I (2003), won the title of “Top Ten Chinese Leading Talents in Science and Technology in Europe” in 2018 and the title of Long-term Innovative Leading Scientist of the “Thousand Talents Plan” of Jiangxi Province in 2018. Professor Wu has gained more than 30 years of professional experience, of which 22 years were spent in European universities such as Ecole Partique des Hautes Etudes (EPHE) in France, University of Louvain in Belgium and University of Sassari in Italy, and international organization, namely, the International Center for Agricultural Research in the Dry Areas (ICARDA). He has been engaged in the application of Geo-information Technology in agriculture-forest-pasture resources assessment, land degradation monitoring, natural disaster risk identification and modeling. He has managed and/or participated in more than 20 international cooperative R&D projects supported by CNRS of France, the European Space Agency (ESA), and the world consortia such as AusAID, CGIAR, IFAD, UNDP, USAID, etc. His projects are distributed in Central, Western and Southern Asia, Eastern, Northern and Western Africa, European Mediterranean Region, and Northwestern and South China. His R&D activities provide technical support for poverty alleviation in the poor developing countries, especially, those that have been suffering from wars, and a scientific basis for sorting out their food security issue.
Dr. Wu was the chairman of the International Conference on Geo-Information Technology and Its Applications (ICGITA 2019), and acts as a member of the Council of the International Association for Water, Environment, Energy and Society (IAWEES) and a scientific committee member or advisor of several international conferences. At the same time, Dr. Wu is the Associate Editor of the International Journal of Remote Sensing, and the Guest Editor of Remote Sensing, International Journal of Geo-Information, and Land. He has published more than 100 scientific papers and is the reviewer of more than 20 world’s top SCI journals in the fields of remote sensing, environment, soil and geography. In 2018, Dr. Wu joined East China University of Technology.
|