SUMMARYSatellite geodetic technologies, such as the Global Navigation Satellite System (GNSS), Gravity Recovery and Climate Experiment (GRACE) and GRACE Follow-On (GFO), have complementary advantages in inferring terrestrial water storage (TWS) changes at regional and basin scales. We introduced a joint inversion strategy to infer TWS changes using GNSS- and GRACE/GFO-derived vertical displacements based on Green's function theory in the Yangtze River Basin (YRB) from January 2011 to December 2020. Additionally, we investigated the performance of variance component estimation (VCE) and Akaike's Bayesian Information Criterion (ABIC) to determine the optimal relative weights of different observation data. The performance of our joint inversion strategy was verified through a closed-loop simulation and multi-source hydrometeorological data [i.e. the time derivative of TWS changes (${\rm d}S/{\rm d}t$) from precipitation (P), evapotranspiration (ET) and run-off (R) based on the water balance equation, called P-ET-R]. The closed-loop simulation shows that the TWS changes from joint inversion have better consistencies with the synthetic signals than those of GNSS- and GRACE-only estimates, and the corresponding root mean square error (RMSE) decreased 1.43−6.28 mm and correlation coefficient (CC) increased 3−10 per cent. The ABIC was more suitable for the joint inversion of measured GRACE/GFO and GNSS data for TWS changes in the YRB. Analysis from the measured data shows that the spatial patterns and seasonal characteristics in TWS changes derived from GNSS, GRACE/GFO and their joint inversion are in good agreement in the YRB. The contribution of GNSS observations to the joint inversion in the upstream of the YRB is greater than that of GRACE/GFO due to the relatively densely distributed GNSS stations, but the opposite is true in the downstream. Furthermore, the joint inversion results have better agreements with P and P-ET-R compared to GNSS- and GRACE/GFO-only estimates in the upstream, and the corresponding CCs increased 5−7 per cent (for P) and 2−5 per cent (for P-ET-R), respectively, which further demonstrates the effectiveness of our joint inversion strategy. Our estimation strategy provides a new insight for joint inversion of GNSS and GRACE/GFO data to obtain more reliable TWS changes.