Impact of including CMIP6 ‘hot’ models in hydrological impact studies.

Climate change is already impacting different aspects of our lives, creating new risks and exacerbating existing ones. Developing effective adaptation and mitigation strategies requires a robust understanding of the magnitude and uncertainty of climate change impacts. A top-down approach is generally used to study climate change impacts on hydrology, forcing the hydrological models with the projections of multiple climate models and studying the impacts. To this end, typically, the impact researchers have given equal weight to climate models considering them independent and equally plausible, giving rise to the notion of “model democracy”. However, model democracy has been criticized fundamentally, and in model ensembles in which the justifiability of some models is challenged, such as CMIP6, model democracy is not a viable option anymore. Some of the CMIP6 models project a warmer future than those predicted by CMIP5 previously.  The climate sensitivity, a measure of the temperature rise in case of increased atmospheric carbon dioxide concentration, of these “hot models” is higher than the range that is expected to be plausible based on observations and our knowledge of planetary physics. The use of hot models in Climate change impact studies biases and overestimates the severity of the impacts. In this study, the impact of the inclusion (or exclusion) of hot models in a multi-model ensemble on the findings of large-sample hydrological climate change impact studies is evaluated. For 3107 North American catchments, we quantify this impact in terms of the magnitude and uncertainty of multiple streamflow metrics, such as mean annual streamflow and the hydrological extremes. The results exhibit a distinct spatial pattern in which the hot models' removal results in reduced streamflow metrics variability in northern regions (Canada and Alaska), southeast US, and along the US pacific coast. The reduced variability means that the hot models contribute to the extremes of the distributions in these regions. The variability reduction is highly dependent on the location of the catchments. Our findings emphasize the importance of the appropriate selection of climate models and display some of the dangers of including ill-advised models in climate change impact studies.

Keywords: Climate change, GCMs, CMIP6, Impact study,  Hydrology, hot models, climate model selection, Uncertainty