Glacier surges are dynamic instabilities that dramatically alter glacier flow and geometry. Their triggers remain poorly understood, and improved methods of monitoring to further constrain the phenomenon are therefore important. We present a novel method for detecting glacier surges automatically using surface elevation data from NASA’s ICESat-2 laser altimetry satellite. Elevation changes from 2018 to 2023 were computed relative to a high-resolution reference digital elevation model and analyzed using a hypsometric binning approach. We trained a Random Forest classifier on known surge events in Svalbard to identify spatial elevation change patterns indicative of surging. Our model detected 110 surges, of which 48 were false positives, 20 uncertain cases that may or may not be surges and 42 certain surges confirmed by external validation. Two of these are currently not part of any surge inventory. The classifier achieved an accuracy of 88.4% and highlighted features in the lower glacier region as most predictive. This study demonstrates that sparse altimetry data such as from ICESat-2 can effectively detect glacier surges and offers a promising, scalable approach to monitoring dynamic glacier instabilities.