Guoqiang Jia, Qiuhong Tang, Stefano Luigi Gariano, Massimo Melillo, Ximeng Xu, Guoyong Leng, Xu Li, 2022, Broadening and Deepening the Rainfall-Induced Landslide Detection_ Practices and Perspectives at a Global Scale, Climate Risk and Sustainable Water Management, edited by Q. Tang & G. Leng, pp. 267–288. Cambridge_ Cambridge university press, 2022,
URL: http://www.cnr.it/prodotto/i/466011

A better detection of landslide occurrence is critical for disaster prevention and mitigation. Over the past four decades, great achievements have been made, ranging from inventories to mapping, susceptibility analysis to triggering threshold identification. Here, we proposed a model to establish global distributed rainfall thresholds, by linking triggering rainfall with geo-environmental causes related to landslide events. The model was based on multiple linear regression method, to define rainfall thresholds as a function of diverse geo-environmental variables, fitted and validated by a combined and relatively accurate landslide dataset. Results show primarily feasible performances for training and testing datasets, with low mean absolute error (0.22 log(mm)) and a high coefficient of determination (0.67) totally. We further prepared global distributed threshold maps for sub- and multi-daily rainfall durations. They share similar spatial distributions in line with previous research. The normalized rainfall index, defined as the ratio of precipitation amount over distributed rainfall thresholds, can be an index of possible landslide occurrence, that is, regions with a normalized index over 1.0 correspond to high probability. We argue that distributed rainfall threshold models are an improvement of empirical threshold models and susceptibility assessments by considering the interaction between triggering rainfall and geo-environmental causes, and promising for better hazard assessment.

Data from https://intranet.cnr.it/people/