Added values and uncertainties of convection permitting regional climate model simulations for urban impact studies over Europe

Extreme weather and climate events such as heat waves, droughts or heavy precipitation are already impacting urban areas worldwide, and such extremes are expected to become more frequent and/or severe with climate change. For vulnerability assessment and climate resilient urban planning, local decision-makers and stakeholders need high-resolution climate information that is tailored to their needs according to different geographical contexts (e.g. through the representation of mountainous areas, coastal lines or city characteristics). They also need information on the appropriate time scale, from specific events of a few days to decade-long statistics. Today, regional climate information often comes from global climate models that are downscaled to the local scale using statistical tools or regional climate models (RCMs) such as those used in the CORDEX initiative.

Longterm RCM simulations achieve horizontal resolutions of the order of ten kilometers and offer added value in certain respects compared with their global counterparts, but remain insufficient in certain specific geographical contexts such as the representation of cities, highly heterogeneous mountainous areas or along coastlines. The latest generation of RCMs, known as Convection Permitting Regional Climate Models (CPRCMs), now reach a spatial resolution of a few kilometers and can better represent heterogeneous land surfaces, with the potential offering a new quality of climate information better suited to local applications.

Here, we compare some evaluation simulations (e.g. driven by reanalysis) carried out as part of the EURO-CORDEX initiative (12,5 km RCM) and the CORDEX Flagship Pilot Study on Convection (3 km CPRCM) over the period 2000-2009. We analyze their ability to represent the urban climate of different European cities and the differences resulting from choices in urban parameterizations, land cover representation approaches (dominant coverage or fractional approaches) and land cover databases. We show that:

• For most European cities, RCM simulations have a too coarse resolution; for example, for all coastal cities, the points that should be considered urban are mainly covered by water.

• CPRCM simulations enable these areas to be better represented thanks to the increased resolution, but there are significant differences depending on how the different land covers are represented in a grid cell and how urban areas are simulated.

• Depending on the meteorological variables of interest, some of the simpler urban parameterizations (altered slab) give results that are relatively close to the more sophisticated ones (multi-layer urban canyon).

• While the increased complexity of CPRCM simulations enables urban climate to be better represented, it also increases the differences between simulations and makes it more difficult to quantify uncertainties and synthesize results into a general assessment (which is often needed by decision-makers) underlining the growing need to use ensembles of climate models for impact assessment.