Estimating Earthquake Magnitude in Sentinel-1 Imagery via Ranking
Daniele Rege Cambrin, Isaac Corley, Paolo Garza, Peyman Najafirad
TL;DR
This work tackles the problem of estimating earthquake magnitudes globally from Sentinel-1 SAR imagery, addressing the low-data regime that hinders regression-based approaches. It introduces a metric-learning framework that jointly optimizes regression and pairwise ranking, using the margin ranking loss $L_{MR}$ together with $L_{MSE}$ to improve discrimination between magnitudes. Evaluated on the QuakeSet with a range of CNN and transformer architectures, the method achieves up to a 30%+ improvement in $MAE$ over regression-only baselines, with ConvNeXt and ViT-based models showing notable gains. The results demonstrate that incorporating ranking signals within mini-batches enhances generalization for earthquake magnitude estimation without requiring additional labels, offering a scalable tool for disaster response and global monitoring.
Abstract
Earthquakes are commonly estimated using physical seismic stations, however, due to the installation requirements and costs of these stations, global coverage quickly becomes impractical. An efficient and lower-cost alternative is to develop machine learning models to globally monitor earth observation data to pinpoint regions impacted by these natural disasters. However, due to the small amount of historically recorded earthquakes, this becomes a low-data regime problem requiring algorithmic improvements to achieve peak performance when learning to regress earthquake magnitude. In this paper, we propose to pose the estimation of earthquake magnitudes as a metric-learning problem, training models to not only estimate earthquake magnitude from Sentinel-1 satellite imagery but to additionally rank pairwise samples. Our experiments show at max a 30%+ improvement in MAE over prior regression-only based methods, particularly transformer-based architectures.