The study of avalanche events is particularly important to assess and predict the degree of risk involved in a given area and time. Currently, the danger level is calculated on a scale of five values (from 1 to 5) and is determined on the basis of stratigraphy of snow cover and meteorological variables. In this project we propose an alternative methodology based on a space-time marked point process. The intensity function of this process indicates the limiting expected rate of occurrence of snow avalanches of a certain size occurring on day t at location (x, y), conditioned on the historical information available prior to time t. Also, we use a self-exciting model to deal with unobserved random space-time effects. The model depends also on a number of meteorological (temperature, humidity, snowfall in the days before, etc.) and environmental (degree of slope, exposure, altitude, etc.) variables which may be considered as covariates in the model. To show the ability of the model in estimating and forecasting the avalanche hazard we consider data of the main Alpin regions in Italy, acquired from the “Catasto Regionale delle Valanghe”, Bollettino Valanghe and meteorological data collected from weather stations installed in different sites of the Alpine zone.

Keywords: Snow avalanche events; Marked point process; Non-homogeneous poisson process; Conditional intensity function

Biography: Harte D. (2010). PtProcess: An R Package for Modelling Marked Point Processes Indexed by Time. Journal of Statistical Software, 35, 8.

Ogata Y. and Zhuang J. (2005). Space–time ETAS models and an improved extension. Tectonophysics, 413, 1-2, 2, 13-23.

Peng R. D., Paik F., Schoenberg J. W. (2005). A space-time conditional intensity model for evaluating a wildfire hazard index. Journal of the American Statistical Association, 100(469): 26–35.

Schoenberg, F.P., Chang, C., Pompa, J., Woods, J., and Xu, H. (2010). A Critical Assessment of the Burning Index in Los Angeles County, California. (in review)