Bias-corrected sea ice thickness estimates through model-data fusion

Abundant observations of ice thickness are available for Arctic sea ice. However they are made with a variety of techniques and are sparse and variable in their spatial and temporal sampling. In contrast model simulations with coupled ice – ocean models such as the Panarctic Ice Ocean Assimilation and Modeling system (PIOMAS) give complete spatial and temporal coverage but contain errors and biases due to limitations in model physics or forcing data. This study seeks to characterize the errors in model-simulated thicknesses and obtain temporally and spatially varying corrections for the model estimates as well as uncertainty estimates for the corrected ice thickness. PIOMAS model estimates of the mean ice thickness from 1979 to 2012 are used for the initial estimates. Ice thickness observations from submarines, moored ULS instruments, airborne electromagnetic measurements, ICESat, and IceBridge are used to determine temporally and spatially varying corrections at the locations of the observations. These corrections are determined across the model grid using optimal interpolation to generate correction fields that are then applied to the initial model estimates. Revised estimates of the total ice volume and regional trends are then obtained