The Yangtze River Basin is an important economic area in China. Monitoring TWS change in this basin is strongly meaningful for exploitation of water resources and economic development. Filtering is necessary to derive the TWS from GRACE time-variable gravity field models, but the signal is distorted. Therefore, we subtract the leakage error from the filtered TWS change to acquire the attenuation signal. We use a scale factor derived from the basin characteristic function to recover the attenuation signal. We propose a so-called third-filter method to gain the temporal distribution of the leakage error. In addition, we compare the TWS obtained using the third-filter method with that from other approaches, e.g., the scaling-factor approach, addictive correction approach, and multiplicative correction approach. In the spatial distribution, we compare the results from these three approaches with that from the second-filter method. Furthermore, we validate the effectiveness of our method in simulation studies. By using the GRACE data, we further analyze the effect of leakage error on TWS change in the Yangtze River Basin, and we compare the restored signal after correcting the leakage error with the WGHM and in-situ measurements of TGR impoundment. The results indicate: (1) From the third-filter method, the annual amplitude of TWS change in the Yangtze River Basin corresponds to a 4.4 ± 0.5 cm equivalent water thickness, while the semi-annual amplitude is up to an equivalent water thickness of 0.7 ± 0.5 cm. These results are more consistent with the WGHM model after leakage error is corrected. (2) Additionally, the GRACE and model results show coincident annual fluctuation of the TWS in the basin, with the largest increase in August every year. The whole trend of third-filter method result in TGR area is closer to yearly change rate of TGR volume. (3) The third-filter method effectively reconstructs the temporal distribution of the TWS from the GRACE filtering data. And the multiplicative correction method may achieve a higher precision of spatial TWS change in the basin.