Sensitivity of resolved convection to ocean and land surfaces in the Tropical Atlantic and Amazon Basin

Abstract

Los modelos climáticos tienden a representar incorrectamente la lluvia tropical, así como sus interacciones con la superficie debido a su baja resolución espacial que imposibilita resolver procesos de convección. En esta tesis doctoral se investiga la sensibilidad de la convección resuelta hacia su superficie para escalas de tiempo estacionales sobre el océano y diurnas sobre el continente. Sobre el océano, se analiza la interacción de la convección resuelta y la temperatura de la superficie del mar (TSM) relacionada con el Modo Meridional Atlántico (MMA). Se demuestra que la respuesta de la precipitación al MMA en simulaciones desacopladas con convección explícita (E-CON) y parametrizada (P-CON) puede interpretarse como un desplazamiento meridional de la precipitación de aproximadamente 1 grado. Sobre el continente, se investiga si las mejoras en la representación de la precipitación en E-CON pueden atribuirse a la representación de sistemas convectivos organizados (SCO) en la cuenca amazónica. Identificamos que la distribución de la intensidad de la precipitación y el ciclo diurno son las mejoras más importantes en la representación de la precipitación. La precipitación asociada a SCO muestra picos de precipitación nocturna entre 1h-6h, lo que permite representar una distribución espacial heterogénea en el ciclo diurno como en las observaciones.

Surface interactions modulate precipitating convection in the tropics. However, convective parameterizations fail in representing tropical precipitation and its interactions with the surface. In recent years, the increased use of “storm-resolving“ simulations has shown promising improvements on the simulation of precipitation. In this dissertation, we investigate the sensitivity of resolved convection to its underlying surface from seasonal timescales over ocean to diurnal variations over land in the tropical Atlantic sector. Over ocean, we tackle the interaction of resolved convection and the sea surface temperature (SST) related to the Atlantic Meridional Mode (AMM) with uncoupled and coupled simulations. We investigate whether the explicit representation of convection leads to i) a robust precipitation response to the AMM and ii) a weaker coupling with SST in contrast to simulations that parameterize convection. We show that the precipitation response to the AMM can be interpreted as a meridional shift of the mean-state precipitation towards the warmer hemisphere. Simulations with explicit (E-CON) and parameterized (P-CON) convection exhibit a similar shift, of about 1 , despite of their distinct mean-state precipitation. In contrast, E-CON exhibits stronger mean-state surface winds which translates into greater wind-driven latent heat flux, and can potentially produce stronger changes in the SST anomalies. We test this hypothesis with coupled simulations. Both the precipitation and SST anomalies respond differently to the AMM during its decay from May to July. In May, the shift of the mean-state precipitation in E-CON is consistent with the uncoupled simulations demonstrating a robust precipitation response to the AMM; whereas the P-CON simulations exhibit a displacement of 2 . Moreover, the cooling of SST is stronger in E-CON than in P-CON. This is influenced by wind-driven latent heat flux anomalies, which are larger in E-CON and lead to a stronger cooling by 0.5 to 1.5K. The wind-driven latent heat flux explains a significant part of the total SST cooling in E-CON (67%) as compared to P-CON (48%), which agrees with the proposed hypothesis. Over land, we investigate whether improvements in the representation of precipitation with explicit convection can be attributed to the representation of organized convective systems in the Amazon. We identify that the distribution of precipitation intensity and the diurnal cycle are the precipitation features with major improvements by the E-CON simulations. Light and high intensity precipitation rates are particularly well reproduced by E-CON, whereas they remain biased in P-CON. The E-CON simulations, unlike P-CON, also reproduce the heterogeneous times of maximum precipitation, with most regions featuring their maximum rain in the afternoon (18h-20h), but others depicting a rain peak overnight. The precipitation associated with organized convective systems display overnight precipitation peaks between 1h-6h in E-CON, which enables the representation of heterogeneous times of maximum precipitation as in observations. Moreover, the simulated diurnal evolution of the size and intensity of organized convective systems is consistent with their observed life cycle. We show that E-CON simulates a realistic diurnal cycle of organized convective systems, which helps to improve the overall representation of the precipitation diurnal cycle.