Debris flow hazard prediction and result explanation based on deep learning

Addressing the challenges of low accuracy, weak adaptability, and poor explainability in existing models for debris flow hazard prediction, this study introduces a novel forecasting approach. Utilizing a dataset from 159 disaster points within the Nujiang River basin in China and selecting 15 influential factors, the study employs a tripartite combination weighting method for the hazard assessment of debris flow hotspots. The hazard of debris flow is then predicted using a CNN-BiGRU-Attention model. Integrating literature review, and utilizing remote sensing explanation, field surveys, geological, and hydrological data with Geographic Information Systems and remote sensing technologies, the hydrological, and geological to the formation of debris flow disasters were identified., , , # Debris flow hazard prediction and result explanation based on deep learning

[https://doi.org/10.5061/dryad.7d7wm383n](https://doi.org/10.5061/dryad.7d7wm383n)

This study introduces a novel forecasting method. It uses a dataset of 159 hazard sites in the Nujiang River Basin, China, to select 15 influencing factors for disaster assessment of mudslide hotspots using a tripartite combination weighting method. All variables and units are in the file.