Jakobshavn Isbræ, one of Greenland’s fastest-flowing outlet glaciers, experiences extensive surface melt that alters its near-surface enthalpy state through both sensible heating and cryo-hydrologic warming. This near-surface enthalpy is a function of both ice temperature and liquid water content, and acts as a boundary condition for evolution of the glacial thermal regime, especially for heavily crevassed ice, yet it remains poorly constrained by existing satellite products. We develop a framework to derive spatially explicit proxies for near-surface enthalpy over Jakobshavn by combining remote sensing techniques with machine learning. High-resolution C-band SAR (Sentinel-1) are used to detect melt onset and “snow wetness”, while ICESat-2 photon-counting LiDAR (ATLAS) processed with the DDA-ice-2 algorithm provides along-track estimates of primary and secondary surface heights (snow, firn, water) and associated scattering metrics. In this study we (i) quantify how enthalpy proxies derived from high-resolution SAR compare with apparent liquid water signatures inferred from ICESat-2 DDA-ice-2 outputs, and (ii) assess how geostatistical metrics (vario-based roughness) and GEOCLASS-image convolutional neural network classifications compare with backscatter-threshold melt detection.