Objectives  Satellite gravity field data can be used to monitor global terrestrial water storage changes and address common challenges such as climate change, disaster prevention and mitigation, etc. Several time-variable gravity models derived from gravity recovery and climate experiment (GRACE) mission have been released internationally. However, there are differences between those models provided by different institutions, and the accuracies are also uneven. The combination of terrestrial water storage change (TWSC) based on different satellite gravity field products would help to maximize the strengths and avoid weaknesses, so as to improve the accuracy of TWSC.

  Methods  Variance component estimation, entropy weight method (EWM) and coefficient of variation method are used to carry out the combination analysis on TWSC inverted from five GRACE satellite gravity field models.

  Results  The results show that all the three combined TWSC models can significantly reduce the discrepancies between time-variable gravity field inverted TWSC, and the signal-to-noise ratio has been improved by about 58% compared to Jet Propulsion Laboratory (JPL) model. The maximum difference in global TWSC trends before and after combination decreases from 0.011 cm/month to 0.001 cm/month, while the maximum amplitude difference decreases from 0.95 cm/month to 0.21 cm/month. At a spatial resolution of 1° × 1°, the difference in standard deviation of TWSC at latitude and longitude directions decreases from over 20 cm² to less than 3 cm². The correlation between the EWM combined TWSC model and hydrological model also improves by up to about 50% compared with that between the JPL independent model and hydrological model.

  Conclusions  The combined method can effectively suppress noise and significantly improve the accuracy of TWSC results, hence can perform as a new mode for high-precision time-variable gravity field applications to provide more reliable data support for terrestrial water storage inversions.