Based on multidimensional equivalent water height (EWH) time series in the Qinghai-Tibetan Plateau recovered from GRACE data, rotated multi-channel singular spectrum analysis (RMSSA) was employed to separate and reconstruct its more accurate local mode of inter-annual oscillations of terrestrial water storage (TWS). The results show that RMSSA could effectively suppress the mode mixture of MSSA, and improve the physical interpretation of the inter-annual oscillations of TWS. Three significant inter-annual oscillations with periods of 6.1a, 3.4a, and 2.5a have been found in the multidimensional EWH series in the Qinghai-Tibetan Plateau (QTP), which account for 38.5 %, 23.5 %, and 16.7 % of the total variance, respectively (after the seasonal and long term have been deducted). The spatial patterns and propagation paths of these three inter-annual oscillations are different and exhibit their own independent local characteristics. Based on the analysis of multi-source GRACE GSM data, the results show that the data solution errors have little influence on the extraction of inter-annual oscillations of TWS. The significant 6.4a, 3.5a, and 2.5a inter-annual oscillations are also found in CPC hydrologic model in the QTP using RMSSA, which account for 22.9, 29.9, and 19.3 % of the total variance, respectively. Three inter-annual oscillations separated from GRACE and CPC show similar spatial patterns and significant cross-correlations, respectively. The maximum cross-correlation coefficients are above 0.5 at the 95 % confidence level over 42, 71, and 75 % of the grids in the QTP, respectively. The results indicate that the soil moisture and terrestrial water storage from GRACE have common inter-annual oscillations and corresponding driving factors in the QTP. We conclude that these three inter-annual oscillations of TWS can be explained by the influence of the Arctic oscillation, oceanic Niña, and Indian Ocean dipole.