Data-driven prediction of present-day glacial isostatic adjustment in North America

Geodetic measurements are assimilated into an a priori model of present-day glacial isostatic adjustment (GIA) via least-squares inversion. The result is an updated GIA model wherein the final predicted signal is informed by both observational data with realistic errors, and prior knowledge (or intuition) of GIA inferred from models. This method has recently been implemented within a limited but growing number of GIA studies (e.g., Hill et al. 2010). The combination method allows calculation of the uncertainties of predicted GIA fields, and thus offers a significant advantage over predictions from purely forward GIA models. Here, we show the results of using the combination approach to predict present-day rates of GIA in North America. Both GPS-measured vertical land motion rates and GRACE-measured gravity observations are assimilated into the prior model. In order to assess the influence that each dataset has on the final GIA prediction, the vertical motion and gravimetry datasets are incorporated into the model first independently (i.e., one dataset only), then simultaneously. Similar to other studies, we find that the predicted model uncertainty decreases as the number of data incorporated increases. Because the a priori GIA model and its associated covariance are developed by averaging predictions from a suite of forward models that varies aspects of the Earth rheology and ice sheet history, the final GIA model is not independent of forward model predictions. However, we examine in detail the effect of using different representations of the input model covariance in order to determine the sensitivity of the final model result to the prior GIA model information. Along parts of the North American coastline, improved predictions of the long-term (kyr-scale) GIA response and its uncertainty at present-day allows better constraint of both the magnitude and uncertainty of the component of measured present-day sea-level change that is attributable to shorter-term forcing.