Mapping Ice Stream Sensitivity in the Amundsen Sector to Uncertainty in Ice Velocity Observations

Abstract Satellite‐derived ice‐velocity products are essential for initializing ice‐sheet models and forecasting ice evolution in a warming climate. Understanding where sea‐level rise (SLR) projections are sensitive to observational errors is key to improve model accuracy and reduce uncertainty. Using Automatic Differentiation in a time‐dependent ice sheet model, we generate sensitivity maps linking forecast changes to uncertainties in model inputs (parameters and observations). Our analysis shows that uncertainty in satellite‐derived velocities significantly affects Volume Above Floatation (VAF) projections. In Thwaites catchment, uncertainty in velocities can shift 40‐year projections by 25%, and differences in velocity products by 45% (with local differences 100 m/yr). Forecast sensitivity is highest near the grounding line, but for some ice streams (e.g., Thwaites), upstream velocities become increasingly influential over time. Sensitivity maps for Thwaites highlight the need for targeted, high‐precision velocity observations in tide‐affected regions and subglacial lakes to constrain future SLR projections.