CR 78:165-178 (2019)  -  DOI: https://doi.org/10.3354/cr01566

ABSTRACT: The sensitivity of tropical rainfall to sea surface temperature (SST) fluctuations was investigated using a multi-scale analysis framework, covering time scales between a few days and 1 decade. The analysis includes an intercomparison between a satellite-based observational dataset and two twentieth century general circulation model (GCM)-based products: the ERA-20CM ensemble of atmospheric simulations and the ERA-20C reanalysis. The results provide robust evidence for the physical linear response of regional rainfall to SST fluctuations at multi-year time scales. All datasets showed tight correlations between rainfall and local SST, the Niño 3.4 index and local surface downwelling longwave radiation (DLR) over wide tropical regions. Additionally, the rainfall-DLR correlation held under a clear sky approximation. These 3 correlations are tightly related since the Niño 3.4 index and clear sky DLR can be computed from SST alone. The latter requires a robust correlation between local precipitable water vapor (W) and SST (i.e. Clausius-Clapeyron relationship), which was confirmed at multi-year scales over large portions of the tropics. All datasets also showed a robust transition in all these correlations at faster time scales (below ~1 yr), suggesting a change in the physical mechanisms controlling rainfall (and W) variability. Based on the linear sensitivities, 3 simplified models were proposed to reconstruct 3-yearly tropical rainfall time series at 5° resolution, forced only by ERA-20CM’s prescribed SST fields. The simplified linear response models were shown to reproduce observations with similar accuracy to GCM-based products, despite having a (easily correctable) systematic bias. The good agreement between all datasets, including the simplified model, provides confidence to the representation of the tropical rainfall response to SST fluctuations in GCM-based and satellite-based products, at least at multi-year time scales.

KEY WORDS: Tropical Rainfall · Multiscale Analysis · Rainfall sensitivity · Statistical modelling