Avalanche Hazard Forecasting and Monitoring based on Meteorological Data and Remote Sensing Technologies using Google Earth Engine Platform

This research presents an automated multi-factor model on the GEE platform, integrating ERA5-Land meteorological data, topographic information (SRTM), and vegetation cover characteristics (Copernicus) to forecast and monitor avalanche hazards across an area of 97,800 km² encompassing 497 avalanche-prone sites. The analysis of five winter seasons (2019-2024) is based on the following threshold criteria: cumulative 3-day snowfall ≥ 10 cm, slope steepness ≥ 30°, snow depth ≥ 60 cm, and temperature variation ≥ 28°C; the contribution of wind (≥ 8 m/s) was assessed separately. The resulting binary risk masks showed that the average area of high-risk zones was 2776,06 km² (2.83% of the territory). Validation using adjusted ground-based data demonstrated high overall accuracy (Accuracy = 0.92) and good precision (Precision = 0.73), with moderate recall (Recall=0.57; F1-score=0.64), indicating a minimization of false alarms at the cost of potentially missing some events. Sensitivity analysis confirmed the dominant influence of slope steepness: adjusting the threshold by ±10% changed the extent of hazardous zones from 1.38% to 3.72%. The wind factor had limited significance at the regional scale during the study period. The developed interactive GEE-based web application enables reproducible generation of risk maps and can support timely planning.