Drivers of global glacial erosion rates

Glacial erosion must be quantified to better constrain numerous geomorphic and orogenic processes; however, accurate models of glacial erosion have been limited by sparse data. Here we use machine learning tools to develop equations that integrate glacial erosion and glaciological, topoclimatic and geological variables based on a global-scale synthesis of 181 contemporary glacier-derived erosion rates. The results reveal environment-specific erosion rate equations for surge-type, marine- and land-terminating glacial settings. We demonstrate that glacial velocity is not the most statistically important predictor of glacial erosion in any environment. Instead, an improved prediction of glacial erosion is attained when velocity is considered with additional glaciological, topoclimatic and geological variables, with the most dominant influences exhibited by precipitation, glacial elevation, length, latitude and the underlying geology. Using these equations, we estimate erosion rates for 85% of contemporary glaciers, with 99% eroding between 0.02 and 2.68 mm yr−1. Our results suggest a need to adjust how we predict or hindcast glacial erosion rates and highlight their sensitivity not only to changes in glacial sliding velocity but also to additional glaciological, topoclimatic and geological influences. Considering glaciological, topoclimatic and geological variables in addition to glacial velocity improves the prediction of glacial erosion rates according to a machine learning-based global analysis.