Impact-based probabilistic modeling of hydro-morphological processes in China (1985-2015)

Area of Science:

• Earth and Environmental Sciences
• Geosciences
• Natural Hazards

Background:

• Hydro-morphological processes (HMP), including debris flows and flash floods, pose significant threats to infrastructure and lives.
• Climate change is expected to exacerbate the frequency and intensity of HMP-driven events.
• Assessing HMP risk requires not only hazard probability but also loss estimation for effective territorial management.

Purpose of the Study:

• To model the impact level of HMPs across China from 1985 to 2015.
• To estimate spatial probabilities of specific HMP impact levels, combining financial and life losses.
• To develop a predictive tool for informing authorities on locations vulnerable to HMP-induced losses.

Main Methods:

• Utilized a comprehensive HMP catalogue for China spanning thirty years (1985-2015).
• Implemented the Light Gradient Boosting (LGB) classifier to model six distinct impact levels.
• Treated each impact level (financial and life losses) as separate target variables for classification.

Main Results:

• Achieved excellent to outstanding classification performance for all six impact categories.
• Reported mean AUC values ranging from 0.862 to 0.915 across the different impact levels.
• Demonstrated the model's capability to estimate spatial probabilities of HMP impacts over a large domain.

Conclusions:

• The developed LGB model shows strong predictive performance for HMP impacts.
• The cartographic outputs can effectively inform authorities about areas susceptible to specific magnitudes of human and infrastructural losses.
• This approach offers a novel method for assessing natural hazard impacts and enhancing disaster risk reduction strategies.