The West Antarctic Ice Sheet contains enough ice to cause 3.3 meters of sea level rise. The ice streams of its Amundsen Sea sector, which alone could contribute up to 1.2 meters of sea level rise, are thinning faster than in any other region on earth, and have the potential for rapid collapse due to inland-deepening bedrock. Using a combination of novel inverse modelling, a comprehensive ice-sheet model, and remote sensing we will: 1) Estimate the present state of the critical Amundsen sector 2) Predict its future behaviour 3) Quantify the uncertainty of these estimates and predictions The physics of ice-sheet retreat is qualitatively understood, but the detailed behaviour is dependent upon a very large number of parameters that cannot be measured directly (e.g, spatially-varying basal traction and ice stiffness). However, numerical ice sheet models have now evolved to the point where a number of relevant physical processes, such as grounding line movement and ice-sheet response to ocean forcing, can be represented accurately. Moreover, the satellite-observational record continues to grow, creating opportunities for assimilation of this new data into models. Such a model-data synthesis can allow key underlying and hidden physical parameters to be determined, facilitating data-driven prediction of future ice-sheet contribution to sea levels. However, techniques for the assimilation of data using ice sheet models remain at an early stage. A considerable amount of data remains unused and fundamental questions, such as the specific information required for reliable predictions, remain unanswered. Moreover, model simulations of future behaviour of ice sheets generally do not account for the uncertainty inherent in estimates of hidden parameters, which can potentially grow with forecast horizons. Accounting for these uncertainties is vital so that informed risk and cost-benefit analyses of sea-level rise protection and adaptation can be carried out. In the proposed project we will develop a model-based framework which will efficiently assimilate the data record for the Amundsen sector (Fig. 1), providing estimates of key physical quantities, and predictions of future behaviour. Crucially, measures of uncertainty will be provided for the estimate and predictions. We will further study the impact that different observations have on our model predictions and uncertainty therein, providing information that will be of value to future observational campaigns. While the Amundsen region is chosen as a focus in the interest of critical relevance and timeliness, the methodology can be applied more generally in other regions of Antarctica, or Greenland.