Improving the accuracy of reconstructed terrestrial water storage during the GRACE/GRACE-FO gap based on a new variable filter reconstruction model

ABSTRACT

Terrestrial water storage (TWS) is a fundamental component of the global hydrological cycle. Understanding changes in TWS is essential for assessing the characteristics of drought and flood events in a region. However, the discontinuity between Gravity Recovery and Climate Experiment (GRACE) and GRACE Follow-On (GRACE-FO) severely affects and limits their applicability in long-term hydrological studies. In this study, a new variable filter reconstruction model (NVFRM) is proposed to reconstruct the TWS anomaly (TWSA) by combining the Gamma test (GT) and Long short-term memory (LSTM) neural network. The research findings reveal that the TWSA reconstructed by the NVFRM is consistent with the GRACE-FO during the testing period (June 2018–December 2020). It shows that the Pearson correlation coefficient (PR), Nash-Sutcliffe efficiency coefficient (NSE), and root mean square error (RMSE) are 0.91, 0.81, and 4.90 cm, respectively, which demonstrates the reliability of the NVFRM. More importantly, four drought events are identified by the drought severity index (DSI) derived from the reconstructed TWSA, including August 2003–April 20 April 200420 April 200411–September 20 June 2011,20 June 201118–September 2018, and September 2019–December 2019, respectively. Flood events are monitored based on the flood potential index (FPI), which is calculated by using the reconstructed TWSA and precipitation. Flood events in the Poyang Lake Basin are characterized by a high FPI index of 0.58, 0.52 and 0.32 in summer (June–August), indicating a greater probability of flooding. The NVFRM model proposed in this study provides an effective measure for filling the missing TWSA in regional hydrological research, which is of great significance for monitoring extreme hydrological events and facilitating disaster prevention and mitigation.

Acknowledgements

Acknowledgment for the Poyang Lake Basin Vector Boundary Dataset support from “National Earth System Science Data Center, National Science & Technology Infrastructure of China. (http://www.geodata.cn)”. Thanks for the data of other institutions used in this study, including GRACE (http://www2.csr.utexas.edu), Noah (https://disc.gsfc.nasa.gov), Precipitation data (http://data.cma.cn), GEE data (https://earthengine.google.com), and scPDSI (https://www.uea.ac.uk), respectively. Supported by High-Performance Computing Platform of Xidian University. Yang Shuai, Zheng Wei and Yin Wenjie contributed equally to this article.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported in part by the National Natural Science Foundation of China (Grant No. 42274119), in part by the Liaoning Revitalization Talents Program (Grant No. XLYC2002082), in part by National Key Research and Development Plan Key Special Projects of Science and Technology Military Civil Integration (Grant No. 2022YFF1400500), in part by the Application Project of Innovative Achievements in the 'Wisdom Eye Action' of the Equipment Development Department of the Central Military Commission, in part by the Scientific Research Project of 'Double First-Class' Construction Project of Surveying and Mapping Science and Technology Discipline in Henan Province (Grant No. BZCG202303), and in part by the Key Project of Science and Technology Com-mission of the Central Military Commission.