Impacts of convective treatment on tropical rainfall variability in realistic and idealized simulations

The prediction of precipitation in the tropics is a challenge for numerical weather prediction (NWP), meaning very low practical predictability there. However, previous studies indicated that intrinsic predictability in the tropics is up to a few weeks and thus longer than in the extratropics. Equatorial waves (EWs) from the linear shallow-water theory are considered the source of this long predictability. Most weather and climate models still struggle to accurately capture EWs, which is often attributed to parameterized convection. With advanced computing power, model development is moving toward high-resolution models with explicit convection. To evaluate the value of these high-resolution models, this thesis aims to provide important insights into the behavior of tropical precipitation due to the treatment of deep and shallow convection using the ICOsahedral Nonhydrostatic (ICON) model.

First, we examine the sensitivity of EWs to model configuration using realistic ICON simulations with varying horizontal grid spacings (80-2.5 km) and with different convectivetreatments between parameterized versus explicit deep and shallow convection. ... mehrTo robustly identify wave signals, we use two objective methods, one filtering rainfall using a fast Fourier transform and the other projecting two-dimensional wind and geopotential onto theoretical wave patterns. The results demonstrate that large-scale EWs are surprisingly consistent in terms of phase speed and wave amplitude with little sensitivity to model resolution, convective treatment and wave identification method. Rainfall signals of westward inertio-gravity waves (WIGs), however, show a large difference between parameterized and explicit convection with the latter showing marked rainfall signals but with no corresponding wind patterns. A composite analysis to link rainfall and wind fields of waves reveals that the identified signals in rainfall appear to be associated with mesoscale convective systems, the spatiotemporal scales of which overlap with those of WIGs, and thus are isolated as waves through space-time filtering.

Secondly, we analyze idealized ICON simulations in a tropical aquachannel configuration with zonally symmetric sea surface temperatures and with rigid walls at 30°N/S. The aquachannel simulations vary in the representation of deep and shallow convection but with the same horizontal grid spacing of 13 km. All aquachannel simulations have maximum rainfall at the equator, showing an intertropical convergence zone (ITCZ) there, but the rainfall amount increases by 35% with explicit deep convection. To physically understand this difference, we adapt a diagnostic based on a conceptual model by Emanuel (2019), assuming boundary-layer quasi-equilibrium (BLQE), the weak temperature gradient approximation, and mass and energy conservation. BLQE implies that moist entropy is in balance between surface enthalpy fluxes, which import high moist entropy to the BL, and convective downdrafts, which transport low moist entropy from the free troposphere into the BL. The results reveal that the rainfall differences are primarily associated with surface enthalpy fluxes through BLQE, while precipitation efficiency is surprisingly constant in the ITCZ. Further detailed analysis demonstrates that mean surface wind speed, which is closely related to the large-scale circulation, contributes most to the differences in surface enthalpy fluxes. Thus, the treatment of deep convection alters mean rainfall through tight links between surface winds, associated surface fluxes and convective mass flux.

Lastly, variability associated with EWs is examined in the aquachannel simulations by using the same wave identification methods used for the realistic simulations. All simulations show prominent signals of Kelvin waves (KWs) with large variations among them. Parameterized deep convection produces various eastward propagation with speeds of 5–27 m/s, while explicit deep convection exhibits a dominance of KWs with a zonal wavenumber of one and with a propagation speed of 24 m/s. Furthermore, explicit deep convection causes more pronounced structures of zonal wind and temperature in the lower stratosphere and a stronger link of wind-induced surface enthalpy flux exchange to the development of convection. Meanwhile, the treatment of shallow convection plays a role for temperature variation below 2.5 km. However, BL warming is in phase with maximum rainfall associated with KWs, which is opposite to observations. Parameterized deep convection generates a feature sharing similarities with the Madden Julian Oscillation, which is not found in the other aquachannel simulations.

The novelty of this thesis lies in understanding the behavior of tropical rainfall in both realistic and idealized simulations by using diagnostics adapted for systematic comparisons between different simulations, mainly due to different convective treatments. This allows us to obtain valuable insights into the sensitivity of tropical rainfall and its variability to model configuration, ultimately paving the way for developing more accurate weather and climate predictions in the tropics.