资源环境科技发展态势分析平台

In the Alps, snow cover dynamics can be monitored using Earth observation (EO). However, low revisit frequency and cloud cover pose a challenge to long-term time series analysis using high spatial resolution EO images. In this study, we applied the random forest regression to model regional snowline elevations (RSEs). In this manner, daily snowline dynamics and their long-term trends can be derived, despite the aforementioned challenges. Of the six investigated Alpine catchments between 1984 and 2018, a significant increasing trend of RSEs is shown in four catchments in the early ablation seasons (between 5.38 2.64 and 11.29 4.79 m center dot a(-1)) and five catchments in the middle ablation seasons (between 4.17 2.62 and 8.76 4.42 m center dot a(-1)). On average, the random forest regression models can explain 75% of the RSE variations. Furthermore, air temperature was found influential in snow persistence especially during middle and late ablation seasons.

Plain Language Summary Snow cover in mountainous regions has been changing worldwide due to climate change in the past few decades. Most existing studies focused on snow cover areal variations in the latitudinal-longitudinal direction, while the understanding of snow cover dynamics in the altitudinal direction is limited. This study aims to provide a new method to derive snowline dynamics in the Alps using a machine learning technique. The results show that there has been a significantly hastened snowline recession during the period 1984-2018 within most of the investigated areas. These results could help to identify climate sensitivity areas at a local scale, where the snowline thereof are retreating increasingly faster. We found a high correlation between monthly regional snowline elevation anomalies and monthly air temperature anomalies especially during the middle and late ablation seasons. In the future, the cooperation between remote sensing scientists and environmental modelers is highly desired, not only to reduce the uncertainties in snowline modeling but also to enhance our knowledge of climate-snowline-runoff interactions at regional scales and hence develop effective adaptation strategies.