Can big data and random forests improve avalanche runout estimation compared to simple linear regression?

Accurate prediction of snow avalanche runout-distances in a deterministic sense remains challenge due to the complexity all physical properties involved. Therefore, many locations including Norway, it has been common practice define runout distance using angle from starting point end zone (α-angle). We use large dataset events Switzerland (N = 18,737) acquired optical satellites calculate α-angle for each avalanche. The α-angles our are normally distributed with mean 33° and standard deviation 6.1°, which provides additional understanding insights into distribution. Using feature importance module Random Forest framework, we found most important topographic parameter predicting be average gradient release area β-point. Despite modern machine learning (ML) method, simple linear regression model yield higher performance than ML attempts. This means that is better an operational context.