CIDR interpolation: an enhanced SSA-based temporal filling framework for restoring continuity in downscaled GRACE(-FO) TWSA products

Abstract

Global total water storage anomaly (TWSA) products derived from the Gravity Recovery and Climate Experiment (GRACE) and its Follow-On mission (GRACE-FO) are critical for hydrological research and water resource management. However, persistent data gaps hinder their application in long-term water cycle monitoring. We propose a Correlation-based Iterative and Decompose-Restore (CIDR) interpolation method to address the dual limitations of conventional Singular Spectrum Analysis (SSA) methods: instability caused by flawed iterative strategy and accuracy degradation due to inadequate denoising. CIDR solves the above issues with two key ways: (i) a correlation-driven iteration stopping rule ensuring the method's overall consistency, and (ii) component-specific interpolation with optimized denoising through a decompose-restore framework. The 0.5{\deg} downscaled GRACE(-FO) TWSA dataset was employed with artificial gaps replicating intra-mission and inter-mission scenarios to evaluate the method's performance. At the grid scale, the global average relative Root Mean Square (rRMS) and Nash-Sutcliffe Efficiency (NSE) improved by 17.5% and 42.6% compared to the SSA method for intra-mission gaps, and by 18.6% and 27.6% for inter-mission gap interpolation, respectively. In addition, the CIDR method shows a significant improvement over the SSA method within the global basins, with the global basin area-weighted average rRMS improved by 20.3% and 20% for both gaps, respectively. This research is expected to provide an advanced gap-filling method to meet the dual requirements of hydrological and climate research for high spatial resolution and long-term continuous TWSA products, which is universally applicable for other hydrological and geodetic time series.