Results from ice flow models are inherently sensitive to input parameters (such as bed geometry or ice viscosity), boundary conditions (such as basal stress or surface temperature), initial conditions (such as present-day geometry and flow), and intrinsic mechanical assumptions (such as the shallow-ice approximation, or hydrostatic balance). In order to increase confidence in model projections of the state of glaciers, ice shelves and ice sheets on decadal to centennial time scales, a systematic analysis of uncertainties and the way these propagate forward is necessary. In addition, data assimilation of heterogeneous datasets is needed to constrain the transient evolution of such models. This involves methods such as adjoint-based initialization or optimization, control methods, or ensemble-based approaches such as Markov Chain Monte-Carlo methods. In this session, we solicit contributions that improve our knowledge of uncertainties in mass flux at the grounding line, calving rate, ice volume, and other key model outputs, and that reveal the dependence of these quantities on model inputs, boundary conditions and model formulation. Contributions that describe improvements in data assimilation for transient ice-flow simulations are also encouraged. Any presentations that enhance our understanding of sensitivities and biases in models of ice sheets, glaciers and ice shelves are welcomed.