Impact of surface temperature biases on climate change projections of the South Pacific Convergence Zone

Abstract

The South Pacific Convergence Zone (SPCZ) is poorly represented in global coupled simulations from the Coupled Model Intercomparison Project Phase 5 (CMIP5), with trademark biases such as the tendency to form a "double Intertropical convergence zone" and an equatorial cold tongue that extends too far westward. Such biases limit our confidence in projections of the future climate change for this region. In this study, we use a downscaling strategy based on a regional atmospheric general circulation model that accurately captures the SPCZ present-day climatology and interannual variability. More specifically, we investigate the sensitivity of the projected rainfall response to either just correcting present-day CMIP5 Sea Surface Temperature (SST) biases or correcting projected SST changes using an emergent constraint approach. While the equatorial western Pacific projected rainfall increase is robust in our experiments and CMIP5, correcting the projected CMIP5 SST changes yields a considerably larger reduction (~ 25%) than in CMIP5 simulations (~ + 3%) in the southwestern Pacific. Indeed, correcting the projected CMIP5 warming pattern yields stronger projected SST gradients, and more humidity convergence reduction under the SPCZ. Finally, our bias-corrected set of experiments yields an increase in equatorial rainfall and SPCZ variability in the future, but does not support the future increase in the frequency of zonal SPCZ events simulated by CMIP5 models. This study hence suggests that atmospheric downscaling studies should not only correct CMIP5 present-day SST biases but also projected SST changes to improve the reliability of their projections. Additional simulations with different physical parameterizations yield robust results.

Impact of projected sea surface temperature biases on tropical cyclones projections in the South Pacific

“…1b shows the scatterplot between the historical rainfall bias and the projected SST change (ΔSST), averaged over the equatorial western Pacific across the 31 CMIP5 models (this analysis is similar to the analysis shown in Fig. 2 of Dutheil et al 26 ). The significant correlation between these two variables (−0.56, p-value < 0.001) indicates that the stronger the present-day dry bias ( Fig.…”

“…5d). These temperature gradient changes responsible for upper tropospheric jets changes may be explained as follows: correcting the projected SST yields a larger drying of the SPCZ western portion as a consequence of near-surface south-easterlies penetrating further west and north in response to SST gradient changes, as already discussed in Dutheil et al 26 and Fig. 1b).…”

“…1c vs 1d), which indicates that the bias correction method improves the reliability of the projected SST changes. As detailed in Dutheil et al 26 , forcing our model with these uncorrected and corrected CMIP5 ensemble mean SST warming patterns results in a large sensitivity of the future SP rainfall pattern to the projected SST pattern with a considerably larger southwestern Pacific rainfall reduction when correcting projected SST changes (contours in Fig. 1e vs 1f).…”

“…Li et al 25 for instance identified a strong relation between the present-day western equatorial Pacific dry bias and SST projections in CMIP5 models that can be used to correct and reduce the uncertainty in those SST projections. Using the PGW downscaling approach, Dutheil et al 26 further compared regional South Pacific (SP) climate change simulations forced by two different SST surface boundary conditions: the uncorrected CMIP5 ensemble mean SST and its "corrected" counterpart based on the Li et al 25 emergent constraint (see Methods). Their results indicate a large sensitivity of the future SP rainfall pattern to the projected SST pattern, with a considerably larger future SPCZ drying in the corrected simulations than in the uncorrected ones, due to altered SST gradients changing the circulation and humidity convergence.…”

“…To that end, we use a similar WRF model configuration than the one used in Dutheil et al 26 , except that it includes a two-way nested domain at 21 km located over the Southwest Pacific that allows a realistic representation of TCs. We then perform three simulations: one present-day (labelled PD) simulation, and two climate-change simulations forced with the same boundary forcing as the one used in Dutheil et al 26 : a multi-model uncorrected CMIP5 SST warming pattern for the first climate-change simulation (labelled CC), and an emergent-constraint-corrected CMIP5 SST warming pattern for the second climate-change simulation (labelled COR). All simulations use the PGW approach: the projected SST change is added on top of present-day SST conditions to avoid issues associated with present-day SST biases in CMIP5.…”