What the development of global climate models at the km-scale teaches us about tropical convection

Accurately capturing location, diurnal as well as seasonal variability of the tropical rainbelt, over land and over ocean, has remained beyond reach for climate models relying on statistical representations of convection. In this context, simulations explicitly resolving convection over regional domains with fixed sea-surface temperatures demonstrated advantages in reproducing several aspects of precipitation from diurnal to sub-seasonal scales.

Credit: Daniel Althoff (distributed via imaggeo.egu.eu), CC BY-NC-SA 3.0

In a study published in Geophysical Research Letters, Hans Segura and his collaborators Cathy Hohenegger, Bjorn Stevens and Christian Wengel show the seasonal and diurnal aspects of tropical precipitation that are (or not) captured in a simulation in which convection and mesoscale ocean eddies are explicitly represented on a 5km mesh global domain, the first simulation of its kind.

Previous studies have focused on regional domains, and with specified sea surface temperatures (SSTs), making it hard to identify the reason for the increase in fidelity in the representation of precipitation. By showing what the model gets right and finding an explanation for the remaining biases when it is set free from the constraints imposed at lower and lateral boundaries, the authors try to disentangle how our climate system works on their way to exploring setups that would allow global coupled simulations at even higher resolution — learning by doing.

Using the Sapphire configuration of the ICOsahedral Nonhydrostatic (ICON) model, the authors demonstrate that the diurnal cycle presents similar characteristics to the simulated over regional domains, well represented over the ocean but over land, the amplitude is too strong. This bias is not an obstacle to representing the tropical rainbelt seasonality over land with high fidelity to observations, e.g., the poleward migration and expansion of the rainbelt during the monsoon season as well as its zonal movement. Over ocean, the rainbelt seasonality is also reproduced in the eastern Pacific and the Atlantic but not in the Indo-Pacific region. The authors track this bias to an incorrect representation of the SST structure with too cold and hence too little precipitation at the equator.

Regions that are well reproduced are also regions showing little interannual variability in summer precipitation in observations, something that was not apparent in past work using parameterized convection. They infer that explicitly resolving convection, even imperfectly, is sufficient to represent important characteristics of convection in regions where there is strong surface forcing, for instance, over land. Over land, a negative soil moisture feedback could be the mechanism for this robustness in the representation of the tropical rainbelt despite the presence of SST biases. Over ocean, the biases in the SST structure suggest that air-sea interaction in regions of relatively modest SST gradients may be quite delicate and merits more study.

Original publication

Segura, H., Hohenegger, C., Wengel, C., & Stevens, B. (2022). Learning by doing: seasonal and diurnal features of tropical precipitation in a global-coupled storm-resolving model. Geophysical Research Letters, 49: e2022GL101796. doi:10.1029/2022GL101796.


Dr. Hans Segura
Max Planck Institute for Meteorology
Email: hans.segura@we dont want spammpimet.mpg.de