Quantifying Long-Term Atmospheric and Hydrospheric Mass Variations by Employing Numerical Weather Prediction Models and Space Geodetic Observations

Meteorological and geodetic systems monitor the variability of the total mass of the atmosphere and terrestrial water storage on time scales from hours to decades and on spatial scales from small turbulent eddies to planetary-scale waves. These systems feature, in most cases, at least one satellite component which allows monitoring on a global scale. Even though state-of-the-art numerical weather prediction models assimilate many of these observations to achieve more accurate weather forecasts, components of the model state can significantly deviate from independent measurements, owing to the quality of observations assimilated, the assimilation procedure, as well as the atmospheric dynamics parameterization in the model. Therefore, there is a need to validate numerical weather models to mitigate erroneous signals that could be interpreted as false climate signatures. This work focuses on long-term (2000-2022) variations induced by atmospheric moisture transport. We employ atmospheric state vectors from ECMWF’s ERA5 reanalysis to force LISFLOOD, a global hydrological model. First, total mass variations from LISFLOOD are validated against mass anomalies from the satellite gravimetry missions GRACE and GRACE-FO. Second, since Earth’s crust slightly deforms under mass loading, we utilize data from a global network of ground based GNSS stations to validate the model’s long-term displacement trends. Third, to assess the long-term stability of the humidity component of the data set forcing the hydrological models, we quantify long-term integrated water vapor variability from ERA5 and compare it with GNSS-derived integrated water vapor. We calculate this integrated water vapor employing the associated zenith wet delays, a by-product estimated during the space geodetic data analysis. Since neither mass anomalies from GRACE and GRACE-FO nor atmospheric delays and displacements from GNSS are assimilated into ERA5, we interpret the long-term discrepancies between models and observations as an accuracy and stability metric for the former.