Glaciers and ice sheets lose their mass by ablation (the output term of their surface mass balance) and discharging into a water body (dynamic loss). The latter is associated with multiple physical characteristics such as bed geometry, inland thinning, terminus stability, and basal conditions. Better assessing the dynamic loss, especially its spatiotemporal variability within a drainage basin, will help improve our understanding of the underlying processes and quantify the future contribution of sea level rise. We propose a new inverse model to decompose glacier elevation change and optimize the dynamic mass loss components for each pixel of the elevation data grid. The model unmixes the observed elevation change from remote sensing data using the modeled surface mass balance and the ice flux as constraints. We use two approaches to design the ice flux term; one is based on glacier surface velocity and the conservation of mass, and the other builds on the flow law and the Shallow Ice Approximation. We test the model for selected marine-terminating glacier outlets in the Greenland ice sheet. If the surface velocity can be decomposed into short-term (seasonal) and multi-year signals, our model may be able to further resolve the dynamic loss components of different physical processes.