The Choco low-level jet under different sea surface temperature conditions: present and future climate

Abstract

ABSTRACT : The Choco low-level jet (CJ) is an important atmospheric feature that strongly modulates the climate in Mesoamerica and the Caribbean (MAC) at the intra and interannual time scales. However, the representation of this jet in General Circulation Models (GCMs) has been previously found to exhibit large biases associated with the misrepresentation of local Sea Surface Temperature (SST) and Sea Level Pressure (SLP) gradients. Here, we evaluate the ability of the Weather Research and Forecasting (WRF) dynamical downscaling simulations to represent the CJ and its relationship with precipitation patterns over the MAC region, for historical and projected (RCP8.5) scenarios. To accomplish our target, we select a GCM and the best configuration of WRF to run dynamical downscaling simulations at a grid size of 36 km over the MAC region. These simulations are performed for 11 different years associated with the 4 coldest (cold events), 4 warmest (warm events) and 3 near-average (neutral events) SST anomalous conditions over the region Niño 1+2 region for each scenario (historical and RCP8.5).

Our results show that CMIP6 models maintain systematic biases previously identified for CMIP3 and CMIP5, like the double Intertropical Convergence Zone (ITCZ) bias, the warm Equatorial Pacific Cold Tongue bias, and the representation of drier regions like northern South America and Central America, with mean dry precipitation biases of 30%. On average, the Taylor Diagrams suggest that CESM1-BC, NorESM2-MM and CESM2 have the best performance for representing the analyzed climate features in the MAC region, with CESM1-BC exhibiting the best performance for all the variables (temperature, zonal wind at 925 hPa and SST), except precipitation. In terms of the WRF configuration, simulations suggest that the highest sensitivity is associated to the cumulus scheme and to a lesser extent to the Planetary Boundary Layer (PBL) scheme. Only little differences are observed when changing the Land Surface Model physics. Experiments using Kain-Fritsch and Grell-Devenyi (New Tiedtke) cumulus parameterizations present the wettest (driest) precipitation biases, with biases above (below) 100% (-50%) over northern South America. These rainfall overestimations are greater for simulations that also used Mellor-Yamada-Janjic PBL scheme, reaching precipitation biases above 120% over northern South America.

From the downscaling of CESM1-BC historical simulations, WRF largely improves the representation of the vertical structure of the CJ and the ITCZ seasonal latitudinal migration. However, these improvements are mainly observed for warm and neutral events. During cold events, WRF exhibits a large overestimation of the CJ characteristics (intensity, position and altitude). These variations in WRF’s performance are related to its simulation of the SLP differences between the regions Niño (Niño 1+2 region), Ocean (eastern tropical Pacific) and Land (northeastern South America landmass), which are similar to ERA5 for warm and neutral events but larger for cold events. Lastly, in terms of the overall mean state, WRF simulates a stronger and northward shifted CJ, and it notoriously improves the precipitation mean bias over northern South America (in around 4 mm/day) and Central America (in around 3 mm/day). Regarding the simulations of future conditions, a weaker and southward shifted CJ with respect to present-day simulation is observed. These changes in the CJ dynamics are associated with variations in the SLP differences, with warm and cold (neutral) events accounting for important reduction (intensifications) during the mid-century. It is important to note that these changes are different between the analyzed cases, showing that neutral events exhibit the strongest intensification of the CJ, followed by warm and cold events, respectively.

Finally, our WRF simulations present important improvements for representing the CJ features and the mean precipitation in the MAC region, especially during warm and neutral events. Despite this, it is also evident that the 36 km grid size does not seem to be enough to account for substantial improvements in the representation of the precipitation interannual variability over northern South America, compared to results from CESM1-BC, especially during cold events. These results are maybe associated with the use of a cumulus scheme and the misrepresentation of the complex topography of the region by WRF, suggesting the necessity of simulations at finer resolutions that can better represent the complexity of processes in this region of the tropical Andes.