From data to application: Harnessing big spatial data and spatially explicit machine learning model for landslide susceptibility prediction and mapping

• Main Authors: KHANT, Min Naing, ANN, Mei Yi Victoria Grace, KAM, Tin Seong
• Format: text
• Language: English
• Published: Institutional Knowledge at Singapore Management University 2024
• Subjects: Big Data; Geospatial Machine Learning; Landslide Susceptibility; Random Forest; Spatial Nonstationarity; Geographic Information Sciences; Numerical Analysis and Scientific Computing
• Online Access: https://ink.library.smu.edu.sg/sis_research/9835; https://ink.library.smu.edu.sg/context/sis_research/article/10835/viewcontent/3681763.3698477_pvoa_cc_by.pdf
• Institution: Singapore Management University

Recent advancements in information and communication technology have significantly enhanced access to extensive geospatial data, presenting a valuable opportunity to leverage big spatial data for improved modeling and predictive capabilities in natural disaster risk assessment. This paper explores the integration of a comprehensive dataset comprising historical landslide events and various geo-environmental variables within a spatially explicit machine learning framework. The study empirically demonstrates that incorporating big spatial data allows a more nuanced understanding of local variations and spatial dependencies. Ultimately, this empirical assessment produces more accurate landslide risk predictions than traditional baseline models. Using Italy’s expansive Valtellina Valley as a case study covering over 3,308 km2, the study illustrates the potential of this integrated approach to enhance predictive outcomes and improve the granularity of the produced landslide susceptibility risk map. The study findings underscore the transformative potential of big spatial data in improving landslide susceptibility assessment and supporting informed decisions in disaster risk management and preparedness.