Projected changes in future tropical cyclone precipitation in the HighResMIP ensemble

Tropical cyclone precipitation (TCP) is expected to increase at a rate close to that predicted by the Clausius-Clapeyron relation (7%/K). Some studies have projected "super Clausius-Clapeyron" rates for TCP, arguing that there could be a mutual enhancement of latent heat release and vertical velocity. The intercomparison project HighResMIP/PRIMAVERA offers an unprecedented opportunity to assess the response of TCP to a future warming and to evaluate the uncertainty of tropical cyclone precipitation arising from models' formulation, and horizontal resolution in particular. This ensemble comprises both forced and coupled experiments with horizontal resolutions covering the range 1^o to 0.25^o. We first assess TCP in the frame of reference of the storm. We find that TCP response depends both on model formulation and resolution : while some models simulate a large increase of precipitation at the core of the TC, some others simulate a decrease at the core and an increase in the outer region. In all cases, TCP response tends to be amplified at higher resolution. Second, we assessed the frequency distribution of TCP averaged in a 5 and 2^o caps, corresponding to the total TCP and precipitation of the inner core region, respectively. For tropical vortices of low intensity, the precipitation increase is moderate, from 2 to 3%/K, and robust across models, as they feel the global energetic constraint. However, the spread of the response is much larger when a 2^o cap is considered and when more intense precipitation rates are considered. We find, however, that a large part of the model spread is explained by how they simulate TCP changes in the frame of reference of the storm : models with a large increase of precipitation at the centre of the storm can simulate an increase of TCP beyond the Clausius-Clapeyron rate in the inner core region. Finally, we explore several possible emergent constraints for TCP, based on the TCP response to the present-day variations of several potential drivers. We find that SST and potential intensity (PI framework) represent good candidates for an emergent constraint. We discuss how such an emergent constraint could help us reduce the uncertainty in TCP future projections but also understand more precisely which aspects of model formulation are responsible for this uncertainty.