Forecasting Global Land Water Storage using GRACE data

Existing approaches for seasonal forecasts of land water storage via land surface models use meteorological forecast products as forcing data. Yet, such meteorological forecast data contain large uncertainties, which inevitably map into highly uncertain land water storage predictions. As a result, current seasonal forecasting of land water storage contains extensive uncertainties. The Gravity Recovery and Climate Experiment (GRACE) satellite mission greatly contributed to monitoring historical land water storage change (TWSC). But many applications like seasonal forecasting of land water storage using GRACE data remains underexplored. Here we analyze the lag relationship between hydrometeorological variables - e.g., precipitation, sea surface temperature, or runoff - and GRACE-derived total water storage change (TWSC). We find that TWSC detected by GRACE lags behind all considered hydrometeorological variables by a few months after removing seasonal effects. By using this lag relationship we forecast the nonseasonal TWSC fields up to one year lead. The prediction approach developed here is based on mostly observational inputs and the validation by GRACE-FO observations suggests it can provide more reliable prediction of global land water storage as compared to model simulations.