Enhancing SWOT discharge assimilation through spatiotemporal correlations

Abstract

The highly intermittent and sparse nature of the Surface Water and Ocean Topography (SWOT) river observations in both space and time poses a major challenge to users for most applications. While interpolation/assimilation efforts exist to assemble discrete SWOT river observations into a spatially complete and temporally continuous record, the efficacy of such efforts is severely compromised by large observation gaps. In this study, we introduced spatiotemporal error correlations in runoff fields into the Inverse Streamflow Routing (ISR) model to improve the gap-filling capability for creating continuous daily discharge. This enhanced ISR model was first investigated in the Ohio River basin, which showed very significant improvements with better reproduced spatial and temporal dynamics of the runoff fields and the discharge time series. Results using 25 validation gauges showed a more concentrated Kling-Gupta Efficiency distribution, where the median was improved from 0.494 to 0.750. The enhanced ISR model was further tested over 16 representative global river basins to verify the effectiveness and robustness of the proposed method. The best fitted e-fold correlation length and time leading to the greatest improvements varied largely among the 16 basins. While all basins benefitted from the spatiotemporal correlations in the SWOT gap-filling, basins suffering from poorer initial guess and fewer SWOT observations, such as the Dnieper, Kolyma and Murray basins, showed larger performance gains than others. Although the experiments were idealized and did not reflect all potential errors that may be present in a real-world application, we demonstrate that the utility of SWOT river products and their value in hydrologic sciences may be boosted by expanding a local/intermittent observation platform into a far-reaching source of hydrologic information through both river hydrodynamics in channel networks and spatiotemporal coherence of runoff errors.