Intercomparison of Dynamically and Statistically Downscaled Climate Change Projections over the Midwest and Great Lakes Region

Downscaling of global climate model (GCMs) simulations is a key element of regional-to-local-scale climate change projections that can inform impact assessments, long-term planning, and resource management in different sectors. We conduct an intercomparison between statistically and dynamically downscaled GCMs simulations using the hybrid delta (HD) and the Weather Research and Forecast (WRF) Model, respectively, over the Midwest and Great Lakes region to 1) validate their performance in reproducing extreme daily precipitation (P) and daily maximum temperature (T-max) for summer and winter and 2) evaluate projections of extremes in the future. Our results show the HD statistical downscaling approach, which includes large-scale bias correction of GCM inputs, can reproduce observed extreme P and T-max reasonably well for both summer and winter. However, raw historical WRF simulations show significant bias in both extreme P and T-max for both seasons. Interestingly, the convection-permitting WRF simulation at 4-km grid spacing does not produce better results for seasonal extremes than the WRF simulation at 12 km using a parameterized convection scheme. Despite a broad similarity for winter extreme P projections, the projected changes in the future summer storms are quite different between downscaling methods; WRF simulations show substantial increases in summer extreme precipitation, while the changes projected by the HD approach exhibit moderate decreases overall. The WRF simulations at 4 km also show a pronounced decoupling effect between seasonal totals and extreme daily P for summer, which suggests that there could be more intense summer extremes at two different time scales, with more severe individual convective storms combined with longer summer droughts at the end of the twenty-first century.

Automated summary

Comparing the downscaled simulations from different methods, it is found that the HD statistical downscaling approach performs reasonably well in reproducing extreme precipitation and maximum temperature, while raw historical WRF simulations show significant bias in extremes. The projected changes in future summer extreme precipitation differ between downscaling methods, with the WRF simulations showing substantial increases and the HD approach exhibiting moderate decreases overall. The WRF simulations at 4 km also suggest a decoupling effect between seasonal totals and extreme daily precipitation for summer, indicating the possibility of more intense summer extremes at different time scales in the late twenty-first century.