Search

By observing well‐known sea surface temperature (SST) indices and outputs from Climate Forecast System version 2 (CFSv2) hindcasts and predictions, in this study, we developed a hybrid statistical downscaling prediction model (HSDPM) based on a timescale decomposition approach, which effectively improved the forecasting ability of summer (June–August) precipitation over Hainan Island (HNSP) in China; this precipitation presented multiple obvious timescale variabilities. We found that the interannual variability of the HNSP is closely related to the observed North Atlantic tripole SST anomalies in the preceding winter and the summer sea level pressure over the South China Sea by CFSv2 predicted in March. On the other hand, the interdecadal variability in the HNSP is linked to the prewinter Pacific Interdecadal Oscillation (PDO) according to the observations, the CFSv2 predictions of the summer surface air temperature over the tropical central‐east Pacific. On this basis, both the interannual and interdecadal components of the HNSP are effectively predicted using the corresponding predictors via multiple regression; thus, the HSDPM is ultimately established by combining the above two components. Compared with the original CFSv2 model, the HSDPM model achieves a considerable improvement in performance in predicting the HNSP. Specifically, the temporal correlation coefficient, ratio of root‐mean‐square error and anomaly sign consistency rate between the observed and HSDPM‐predicted HNSPs are 0.70, 17.4% and 80.5%, respectively, which are significantly greater than the above three forecasting index values of 0.23, 38.3% and 48.8%, respectively, obtained from the original CFSv2 predictions. The application of the HSDPM may be beneficial for drought and flood prevention and mitigation in Hainan Island. [ABSTRACT FROM AUTHOR]

Northeast China (NEC) is located between the subtropical monsoon and temperate-frigid monsoon regions and exhibits two successive rainy seasons with different natures: the northeast cold vortex rainy season in early summer (May–June) and the monsoon rainy season in late summer (July–August). Summer rainfall over NEC (NECR) has a fundamental influence on society, yet its successful seasonal prediction remains a long-term scientific challenge to current dynamical models. The poor NECR prediction skill is partly attributed to the large NECR variability at both the interannual and interdecadal time scales. Here, we focus on the oceanic drivers of the late summer NECR variability and associated physical processes at interannual time scale. Then, we establish an empirical prediction model to predict the interannual variability of summer NECR at 1-month lead time (in June). The analysis of observations spanning 40 years (1963–2002) reveals three physically and synergistically influencing predictors of the late summer NECR interannual variability. Above-normal NECR is preceded in the previous spring by (a) warm sea surface temperature (SST) anomalies in the tropical northern Indian Ocean, (b) a positive thermal contrast tendency in the tropical West–East Pacific SST, and (c) a positive tendency of the North Atlantic tripolar SST. These precursors enhance the anomalous low-level anticyclone over the western North Pacific and southerly anomalies over NEC in late summer, which are beneficial to enhancing NECR. An empirical prediction model built on these three predictors achieves a forecast temporal correlation coefficient (TCC) skill of 0.72 for 1961–2019, and a 17-year (2003–2019) independent forecast shows a significant TCC skill of 0.70. The skill is substantially higher than that of five state-of-the-art dynamical models and their ensemble mean for 1979–2019 (TCC = 0.24). These results suggest that the proposed empirical model is a meaningful approach for the prediction of NECR, although the dynamical prediction of NECR has considerable room for improvement. [ABSTRACT FROM AUTHOR]