A global perspective on the spatial representation of climate extremes from km-scale models

Abstract Weather and climate extremes are rare manifestations of climate variability that can severely impact society and the environment. To investigate their properties and changes on a global scale, observational records are often complemented by climate models such as those from the latest Coupled Model Intercomparison Project (CMIP6). However, typical CMIP6 models have a grid spacing of about 100 km and, therefore, do not allow the representation of extremes at local scales important for impacts. Here, we provide a global view on the information lost at such resolutions, focusing on temperature and precipitation extremes. We draw on two next-generation, km-scale global climate models, run with a grid spacing of about 10 km, and regrid them to a range of coarser resolutions. From the regridded data, we then investigate the spatial sub-grid variability hidden at the CMIP6-like 100 km grid spacing to quantify the effect resolution has on the representation of extremes globally. We find clear patterns of such a resolution effect on a diverse set of temperature extremes, particularly in mountainous areas, at coastlines, and along large rivers. For the example of annual maximum temperature, the difference between the 10 km and 100 km grid spacings can exceed 10 ∘ C. For precipitation, globally aggregated low and high extremes are shown to be underestimated at coarse resolution, with the strongest spatial signals emerging in regions with complex topography and in the tropics. Our results quantify existing knowledge and demonstrate the importance of spatial resolution for the representation of climate extremes, in particular for hotspot regions such as coastlines, which often coincide with densely populated areas. The advent of ever higher resolved global models, hence, allows improved estimates of local climate impacts and related risk assessments with global coverage.