AI4Glaciers: Introducing a multi-sensor photogrammetric system for calving monitoring at the Perito Moreno glacier

Understanding glacier dynamics is fundamental to predicting their response to a changing climate and mitigating associated risks. Glaciers, which hold approximately 70% of the world's freshwater, are losing mass at an unprecedented rate due to anthropogenic activities. Accurate predictions of glacier behavior under warming scenarios are crucial for short-term hazard mitigation and long-term water resource planning.

The project "AI4Glaciers: AI-Enabled Prediction of Glacial Calving based on 4D Real-Time Multi-Sensor Monitoring (AI4G)" aims to monitor, in near real-time, a section of the Perito Moreno Glacier's front using a multi-sensor photogrammetric system (RGB and thermal cameras) in 4D (3D + time). By correlating calving events with climatic conditions using artificial intelligence, the project seeks to understand the drivers of accelerated glacial calving processes and, consequently, glacier retreat.

In this contribution, we present the camera setup installed in January 2025 at the Perito Moreno Glacier as part of the AI4G project. The system comprises eight synchronized DSLR cameras capturing image pairs every 30 minutes during daylight hours. These images are transmitted twice daily to a central server via 4G connectivity, enabling near real-time analysis. Additionally, three thermal cameras (600x400 pixels), capturing data continuously 24 hours a day, were installed to generate photogrammetric reconstructions using temperature data.

With approximately 15,000 images collected monthly, photogrammetric models are generated using Structure-from-Motion Multi-View Stereo (SfM-MVS) techniques. These models are compared using change-detection algorithms to identify relative changes at the glacier front, including ice loss due to calving and pre-calving deformations.