Answers are grounded in this paper and cite their sources.
Post-processing numerical weather prediction (NWP) model forecasts while preserving spatiotemporal dependence structures remains challenging. The spatial-mode-based calibration (SMoC) method was developed to post-process forecasts with inbuilt spatial dependence structures. However, the temporal dependence structures were not considered in the original form of the SMoC. To address this limitation, we develop a Gaussian copula-embedded SMoC method (SMoC-GCA) to account for both spatial and temporal dependencies. The Empirical Orthogonal Function (EOF) decomposition is first applied to reduce the spatial dimensions as in the original SMoC. The Gaussian Copula Approach (GCA) is then used to model the temporal dependence structures among corrected principal components. This two-step formulation reduces the parameter space compared to conventional parametric spatiotemporal post-processing methods. To evaluate the performance of the SMoC-GCA method, we apply it to post-process daily sub-seasonal precipitation forecasts over the Gan River Basin. The results are compared with those of the original SMoC, and those of conventional grid-by-grid post-processing method with non-parametric reordering approaches. The results demonstrate that the SMoC-GCA improves ECMWF raw sub-seasonal precipitation forecasts in terms of accuracy and reliability. It is also shown that the post-processed precipitation forecasts exhibit better spatiotemporal correlation structures compared to both the original SMoC and the grid-by-grid post-processing method. These findings demonstrate that the SMoC-GCA is effective for post-processing numerical forecasts with well-preserved spatiotemporal structure across ensemble members.
No takes yet. Share an insight, caveat, or question.
Li et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: