The monthly changes in total water storage (△TWS) can be employed for drought and flood monitoring and early warning and can be obtained from the total water storage anomalies (TWSA) of the Gravity Recovery and Climate Experiment (GRACE). However, the relatively short GRACE time series limits its further wide application. To this end, a combined prediction (CP) model including Support Vector Machine (SVM) and Artificial Neural Network (ANN) was proposed in this study for the reconstruction and extension of monthly TWSA from 1960 to 2012. Moreover, an innovative input selection strategy is proposed to build a monthly TWSA prediction model, in which the partial correlation algorithm is used to select the best input variables from candidate input variables. These candidate input variables include streamflow, precipitation, evaporation, and soil moisture storage (SMS). The Yunnan province, a typical humid area in China, was selected as a case study. The results show that: (1) The innovative input selection strategy effectively improves the simulation ability of the model, especially when the candidate input variables influence each other; (2) The performance of the CP model using the innovative input selection strategy is best; (3) The monthly △TWS obtained from the extension of TWSA recorded five of the seven extreme meteorological drought events in Yunnan Province from 1961 to 2001, therefore, the reliability of the expanded TWSA is better than GLDAS TWSA. Generally, the findings of this study showed that the CP model using an innovative input selection strategy is a useful and powerful tool for monthly TWSA prediction.