Exploring the effect of kilometer-scale climate modeling on the representation of historical and future heat waves. A multi-model ensemble perspective

The latest generation of Convection Permitting Regional Climate Models (CPRCMs, <4 km resolution) provides a step change in our understanding of regional-to-local scale climate processes.

Recent studies highlight how km-scale modeling provides a more accurate representation of precipitation extremes compared to the driving convection-parameterized RCMs. Further, evidence suggests that changes in the soil moisture-precipitation feedback and regional precipitation recycling occur when moving to a km-scale. This generally translates into drier conditions in km-scale simulations, mainly during summer. Consequently, the different soil moisture content in explicit vs. parameterized simulations results in a different partitioning between heat fluxes, which in turn can modulate temperature extremes and heatwaves (HWs).

This study explores the representation of HWs and their future changes from an ensemble of twelve CPRCMs downscaled from CMIP5 GCM projections for historical and end-of-century periods over a greater alpine region. The two-step dynamical downscaling consists of downscaling GCMs to an intermediate 12–15 km resolution (convection-parameterized RCMs) and then using these fields to further downscaling to the kilometer scale.

Analyses are two-fold: (i) Exploring if the warmer/drier signal introduced by the km-scale points toward an improvement compared to the driving convection-parameterized simulations over the historical period. Here, distribution-based grid- and station-scale evaluation metrics are considered. (ii) Assessing if the km-scale signal is temporally stationary or if modulation of summer temperatures and HWs future changes can be expected. Key metrics are summer maximum temperature and relevant HW statistics (e.g., amplitude, persistence magnitude). HWs local-scale forcing, represented by the land-atmosphere coupling magnitude, is also analyzed.

Preliminary results show an added value from km-scale simulations. RCM cold biases are reduced and summer maximum temperature distribution is improved over a majority of reference stations. Concerning future changes both resolutions show a summer maximum temperature change signal ~ +6 °C characterized by a large spread among members (+4/+8 °C). Considering the ensemble mean, we do not observe strong modification of the climate change signal by the CPRCMs (±10%). However, this results from averaging change signal modifications from individual members that can be as much as up to ±25%, with no clear tendency toward an amplification/reduction of the driving RCM change signal.

Similar results are obtained considering only HW days. Driving the change signal alteration observed in some models is a difference between CPRCMs and RCMs in the partitioning of latent heat during HW days. In contrast to the CPRCMs, some RCMs produce positive future changes in latent heating during HW days, meaning there is sufficient soil moisture to allow latent heat to increase in response to an increased radiative forcing. 

To conclude, CPRCMs are warmer than RCMs during the historical period, resulting from improved and more realistic physics. This does not translate into an unambiguous modulation of the ensemble mean future change signal. However, those models that exhibit strong modulation could be driven by a different sign of HWs latent heat change signal. This aspect deserves further analysis since alterations of other relevant HW features, such as magnitude and persistence, have potentially large societal impacts.