The optimal choice of hydrodynamic models to assess the impact of climate change on the cryosphere

Abstract

This study is targeted at narrowing the range of uncertainties in predictive cryospheric modeling associated with climatic projections. We used the output from 36 CMIP5 GCM runs for the period 1976–2005 and calculated trends of several climatic characteristics that largely govern the state of the cryosphere, i.e. seasonal and mean annual air temperature, thawing degree-day sums, annual and winter precipitation sums. Data from 744 weather stations were used to identify and delineate 17 regions, which demonstrate coherent temperature changes in the past decades. Results from GCMs and observations were averaged over the «coherent regions» and compared with each other. Ultimately, we evaluated the skills of individual CMIP5 GCMs, ranked them in the specific context of predictive cryospheric modeling, identified top-end models in each of the 17 regions and eliminated the outliers. Selected top-end GCMs were used to compose optimal regional ensembles that were compared with the ensemble consisting of all available models. An optimal ensemble was also constructed for the area underlain by permafrost in Russia. Results indicate that the all-model ensemble in most regions underestimates the projected temperature changes compared to the optimal ensemble. Elimination of the outliers narrows the range of uncertainty in regional climate projection by 5–20%.

Keywords

• climate change
• ensemble climatic projections
• modeling