RINA - Démonstrateur d’utilisation de l’intelligence artificielle pour la gestion opérationnelle des risques naturels gravitaires géologiques | FEREC 2020

1.4. CONCLUSION 11 Figure 1.2: Monthly time series for the number of rockfall (top) and for quantity of rainfall (bottom) before (left) and after (right) major protection works • point process of rockfall times is not an homogeneous Poisson Process, as time between event doesn’t follow exponential law, • strong seasonality in rainfall times series and quite strong in rockfall one, • significantly correlation between rockfall and rainfall from day-1 to round day-5 Some perspectives for future works could be: • rainfall episode in [2] belong to factors which can explain a part of rockfall hazard and should be take into account in future predictive model, • in [2] , consider other variable dimension like : intensity of rain in mm.h − 1 , rockfall frequency in h − 1 , • consider threshold of event size (> 0 . 1 m 3 ) because dependencies and dynamics for small or big rockfall seem to be not the same.