Statistical uncertainty analysis-based precipitation merging (SUPER): A new framework for improved global precipitation estimation

Authors: DONG, J. - Massachusetts Institute Of Technology; Crow, Wade; XI, C. - Tianjin University; TANGDAMRONGSUB, N. - National Aeronautics And Space Administration (NASA); GAO, M. - Tianjin University; SUN, S. - Collaborator; QIU, J. - Sun Yat-Sen University; WEI, L. - Nanjing University Of Information Science And Technology (NUIST); GAO, H. - Collaborator; DUAN, Z. - Lund University

Submitted to: Remote Sensing of Environment
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 9/28/2022
Publication Date: 10/22/2022
Citation: Dong, J., Crow, W.T., Xi, C., Tangdamrongsub, N., Gao, M., Sun, S., Qiu, J., Wei, L., Gao, H., Duan, Z. 2022. Statistical uncertainty analysis-based precipitation merging (SUPER): A new framework for improved global precipitation estimation. Remote Sensing of Environment. 283. Article 113299. https://doi.org/10.1016/j.rse.2022.113299.
DOI: https://doi.org/10.1016/j.rse.2022.113299

Interpretive Summary: The accurate estimation of daily rainfall accumulation is a key part of efforts to track agricultural drought, optimize irrigation scheduling and manage water resources in agricultural basins. A large variety of methods are currently employed to derive such estimates - including ground-based rain gauges, atmospheric forecasting models and a range of remote sensing techniques. However, considered on their own, each of these approaches are limited by notable shortcomings. This paper describes a novel statistical strategy for merging daily precipitation accumulation estimates derived from multiple sources. In this way, we derive a unified, best-available estimate of rainfall accumulation that improves upon each of the original products used to generate it. The approach is especially effective in data-poor areas of the globe lacking ground-based rain gauges and will eventually be used to improve USDA’s ability to globally monitor agricultural drought and inter-annual variations in crop productivity.

Technical Abstract: Multi-source merging is a viable tool for improving large-scale terrestrial precipitation estimates. Existing merging frameworks typically use gauge-based precipitation error statistics and neglect the inter-dependency of different precipitation products. However, gauge observation uncertainties can bias merging weights and yield a sub-optimal precipitation analysis, particularly over data-sparse regions. Likewise, frameworks ignoring inter-product error cross-correlation will overfit precipitation observation noise. Here, a Statistical Uncertainty analysis based Precipitation mERging framework (SUPER) is proposed for addressing these challenges. Specifically, a quadruple collocation analysis is employed to estimate product error variances and covariances for commonly used precipitation products. These estimates are subsequently used to merge all products using a least-squares minimization approach. In addition, false precipitation events are removed via a reference rain/no-rain time series estimated by a newly developed categorical variable merging method. As such, SUPER does not require any rain gauge observations to reduce random and rain/no-rain classification errors. In addition, by considering precipitation product inter-dependency, SUPER avoids overfitting measurement noise present in multi-source precipitation products. These are major theoretical advantages over traditional gauge-based merging approaches. Results show that the overall RMSE of SUPER-based precipitation is 3.35 mm/day and the daily correlation with gauge observations is 0.71 [-] – metrics that are superior to either a recent precipitation reanalysis or remotely sensed precipitation products. Finally, we also demonstrate that SUPER outperforms the state-of-the-art multi-source merged product over most conditions. In this way, we propose a new framework for robustly generating global precipitation datasets that can improve land surface and hydrological modeling skill in data-sparse terrestrial regions