Remote sensing and GIS-based study analysis on alpine glacier environments

The work analyses various remote sensing techniques utilized to monitor alpine glacier environments, which are crucial for agriculture, tourism and CO2 capture. Due to current climate change, these glacier environments are undergoing subtle changes and require constant monitoring. Using aircraft acquisition methods, DIGISKY s.r.l captured high-resolution RGB imagery (25 cm) of the Monte Bianco range. However, processing these large images posed challenges in terms of computational requirements. This study develops a workflow to minimize computational demands, with the entire analysis conducted using QGIS, an open-source software. The results highlight effective methodologies for efficient processing and monitoring of alpine glacier environments. This study focused on applying machine learning (ML) techniques to a rock debris glacier in the Monte Bianco range, specifically the Miage Glacier, one of the widest rock debris glaciers in Europe. The workflow utilized ML techniques to achieve a multi-class supervised classification using 25 cm resolution RGB orthophoto imagery. A QGIS plugin, Orfeo Toolbox, was employed for this analysis, which is designed for satellite imagery processing using ML techniques. The study compared pixel-based and object-based classifications, evaluating their accuracy and precision for a multi-class classification on the Miage Glacier, with assigning 18 classes in total. Additionally, the same workflow was applied to Sentinel satellite imagery with a 10-meter resolution, where 20-meter resolution additional bands were resampled to 10 meters. Both classifications were performed on the Miage Glacier for individual years, with the month of acquisition coinciding with the month of acquisition of aircraft imagery. The majority of Individual classes achieved 80-90% accuracy and precision. Various ML algorithms were tested, with statistics provided for each to identify the top two performers. Integrating these two top performing algorithms resulted in a significant improvement in overall multi-class classification accuracy and precision by 2-3%. A scientific study was conducted on the Miage Glacier using existing research to showcase downwasting and backwasting trends where the mean annual downwasting rate from previous studies aligned with our results, indicating significant changes. This study also explored the role of vegetation on the rock glacier, providing vital information on glacier recession at the moraines. Additionally, it examined ice-contact lakes, such as Miage Lake, and their water level variations over time. Ice-contact lakes like Miage Lake reveal how internal thawing of the rock glacier ice cliffs has caused significant fluctuations over time. Using a 50 cm resolution Digital Surface Model provided by DIGISKY s.r.l and a Digital Terrain Model, resampled to 50 cm resolution, from Val d’Aosta, we visually demonstrated differences in surface volume and moraine recession. The results indicate a slowdown in debris flow, likely due to the backwasting of the parent glaciers feeding the rock glacier. With reduced debris cover and increased exposure to solar radiation, the ice cliffs have begun to thaw at critical points reaching an extent of 30-40 meters in depth, highlighting the impacts of climate change on rock debris glacier.