Quality Evaluation of the Weekly Vertical Loading Effects Induced from Continental Water Storage Models

To remove continental water storage (CWS) signals from the GPS data, CWS mass
models are needed to obtain predicted surface displacements. We compared weekly GPS
height time series with five CWS models: (1) the monthly and (2) three-hourly Global
Land Data Assimilation System (GLDAS); (3) the monthly and (4) one-hourly Modern-
Era Retrospective Analysis for Research and Applications (MERRA); (5) the six-hourly
National Centers for Environmental Prediction-Department of Energy (NCEP-DOE) global
reanalysis products (NCEP-R-2). We find that of the 344 selected global IGS stations,
more than 77% of stations have their weighted root mean square (WRMS) reduced in the
weekly GPS height by using both the GLDAS and MERRA CWS products to model the
surface displacement, and the best improvement concentrate mainly in North America and
Eurasia.We find that the one-hourly MERRA-Land dataset is the most appropriate product
for modeling weekly vertical surface displacement caused by CWS variations. The threehourly
GLDAS data ranks the second, while the GLDAS and MERRA monthly products
rank the third. The higher spatial resolution MERRA product improves the performance of
the CWS model in reducing the scatter of the GPS height by about 2–6% compared with
the GLDAS. Under the same spatial resolution, the higher temporal resolution could also
improve the performance by almost the same magnitude. We also confirm that removing
the ATML and NTOL effects from the weekly GPS height would remarkably improve the
performance of CWS model in correcting the GPS height by at least 10%, especially for
coastal and island stations. Since the GLDAS product has a much greater latency than the
MERRA product, MERRA would be a better choice to model surface displacements from
CWS. Finally, we find that the NCEP-R-2 data is not sufficiently precise to be used for this
application. Further work is still required to determine the reason.