Researchers used machine learning to predict erosion rates for over 180,000 glaciers, highlighting the influence of precipitation, elevation, and geology.


A new study in Nature Geoscience has revealed that 99 per cent of the world’s glaciers erode at rates between 0.02 and 2.68 millimetre per year, offering the most precise global estimates to date.

The research, led by scientists from the University of Victoria, used machine learning models to predict erosion rates for over 180,000 glaciers worldwide. The analysis identified precipitation, glacial elevation, length, latitude, underlying geology, and glacial velocity as key drivers of erosion, with precipitation emerging as the most influential factor across all environments.

The highest erosion rates were observed in Alaska, Central and South Asia, the Caucasus and the Middle East, and New Zealand. Regions with widespread modern ice cover such as Alaska, the Canadian Arctic, Greenland, Scandinavia, and the southern Andes were found to experience the majority of glacial erosion.

Using data from the Randolph Glacier Inventory version 6, researchers estimated erosion rates for 185,081 glaciers, representing 85 per cent of global glaciers. Collectively, these glaciers erode about 23 gigatonnes of bedrock annually. For comparison, fluvial sediment flux from exorheic drainage areas contributes 18.5 — 20 gigatonne per year.

The study also found that glacial velocity alone is not the most significant predictor of erosion; rather, the most accurate forecasts arise when velocity is combined with topoclimatic, geological, and glaciological variables. Factors such as seismic activity, rock type, and geothermal heat flux play a substantial role in shaping erosion patterns.

The findings have broad implications for understanding landscape transformation, sediment transport, and nutrient cycles, as well as for planning in sectors such as long-term infrastructure stability, landscape management, and nuclear waste storage.