Nan Wang, Weiming Cheng , Hongyan Zhang, Cees J. van Westen , Junnan Xiong , Changjun Liu, Luigi Lombardo 

Hydro-morphological processes (HMP, any natural phenomenon contained within the spectrum defined between debris flows and flash floods) pose a relevant threat to infrastructure, urban and rural settlements and to lives in general. This has been widely observed in recent years and will likely become worse as climate change will influence the spatio-temporal pattern of precipitation events. The modelling of where HMP-driven hazards may occur can help define the appropriate course of actions before and during a crisis, reducing the potential losses that HMPs cause in their wake. However, the probabilistic information on locations prone to experience a given hazard is not sufficient to depict the risk our society may incur. To cover this aspect, modeling the loss information could open up to better territorial management strategies.

In this work, we made use of the HMP catalogue of China. This catalogue reports reliable records from 1985 to 2015 across the whole Chinese territory. Specifically, we implemented the Light Gradient Boosting (LGB) classifier to model the impact level that locations across China have suffered from HMPs over the thirty-year record. In doing so, we estimated spatial probabilities of certain HMP impact, something that has yet to be tested in the natural hazard community, especially over such a large spatio-temporal domain.

This experiment follows a project launched by the Chinese government with the aim of improving national efforts against climate change and improving societal resilience to disastrous events. In this context, the good predictive performance our model produced suggest that the cartographic output could be useful to inform authorities of locations prone to human and infrastructural losses of specific magnitudes.