Considerations on the prediction of hazards (mainly landslides) and their consequences

The general assumptions and the most popular methods used to assess landslide hazard and for landslide risk evaluation have not changed significantly in recent decades. Some of these assumptions have conceptual weaknesses, and the methods have revealed weackneses and limitations. After an introduction on what we need to predict in order to assess landslide hazard and risk, I introduce the strategies and main methods currently used to detect and map landslides, to predict landslide populations in space and time, and to anticipate the number and size characteristics of expected landslides. For landslide detection and mapping, I consider traditional methods based on visual interpretation of aerial photography, and modern approaches that exploit visual, semi-automatic or automatic analysis of remotely sensed imagery. For spatial landslide prediction, I discuss the results of a review of classification-based statistical methods for evaluating landslide susceptibility. For temporal forecasting, drawing on a review of geographical landslide forecasting and early warning systems, I discuss short-term forecasting capabilities and their limitations. Then, I discuss the long-term landslide projections considering the impact of climate variations on landslide projections. Regarding the numerosity and size of landslides, I discuss existing statistics on the length, width, area, and volume of landslides obtained from populations of event-triggered landslides. This is followed by an analysis of the consequences of landslides. I conclude by offering recommendations on what I imagine we should do to make significant progress in our collective ability to predict the risk posed by landslide populations and to mitigate their risk. My understanding, but also my feeling and hope, is that some - perhaps many - of the recommendations are general, and may be applicable to other hazards as well.