On the estimation of landslide intensity, hazard and density via data-driven models.

The geographic prediction of landslide occurrence is undertaken by assessing whether a slope may be stable or unstable. In other words, current practices treat slopes where a single landslide occurred in the same way as slopes where many landslides occurred. At the slope scale, this procedure inevitably underestimates the effect of multiple landslides.
Here we model the number of landslides per slope instead. Then, thanks to the close relation that the number of failures shows with respect to landslide size, we convert the estimated number of landslides into estimated landslide areas. Ultimately, we also estimate the expected proportion of a slope affected by landslides. This framework is more informative than the stable/unstable paradigm and may help landslide risk mitigation strategies.