Prediction of Snowmelt Days Using Binary Logistic Regression in the Umbria-Marche Apennines (Central Italy)

Abstract

Snow cover in a mountain area is a physical parameter that induces quite rapid changes in the landscape, from a geomorphological point of view. In particular, snowmelt plays a crucial role in the assessment of avalanche risk, so it is essential to know the days when snowmelt is expected, in order to prepare operational alert levels. Moreover, melting of the snow cover has a direct effect on the recharge of the water table, as well as on the regulation of the vegetative cycle of mountain plants. Therefore, a study on snowmelt, its persistence on the ground, and the height of the snow cover in the Umbria-Marche Apennines in central Italy is of great interest, since this is an area that is extremely poorly sampled and analysed. This study was conducted on the basis of four mountain weather stations equipped with a recently installed sonar-based snow depth gauge, so that a relatively short period, 2010–2020, was evaluated. A trend analysis revealed non-significant decreases in snow cover height and snow persistence time, in contrast to the significant increasing trend of mean temperature, while parameters such as relative humidity and wind speed did not appear to have a dominant trend. Further analysis showed relationships between snowmelt and the climatic parameters considered, leading to the definition of a mathematical model developed using the binary logistic regression technique, and having a predictive power of 82.6% in the case of days with snowmelt on the ground. The aim of this study was to be a first step towards models aimed at preventing avalanche risk, hydrological risk, and plant species adaptation, as well as providing a more complete definition of the climate of the study area.