A global climate model based, Bayesian climate projection for northern extra–tropical land areas

Projections with contemporary global climate models (GCMs) still markedly deviate from each other on magnitude of climate changes, in particular, in middle to subpolar latitudes. In this work, a climate projection based on the ensemble of 18 CMIP3 GCM models forced by SRES A1B scenario is performed for the northern extra–tropical land. To assess the change of soil state, off–line simulations are performed with the Deep Soil Simulator (DSS) developed at the A.M.Obukhov Institute of Atmospheric Physics, Russian Academy of Sciences (IAP RAS). This model is forced by output of the above–mentioned GCM simulations. Ensemble mean and ensemble standard deviation for any variable are calculated by using Bayesian averaging which allows to enhance a contribution from more realistic models and diminish that from less realistic models. As a result, uncertainty for soil and permafrost variables become substantially narrower. The Bayesian weights for each model are calculated based on their performance for the present–day surface air temperature (SAT) and permafrost distributions, and for SAT trend during the 20th century. The results, except for intra–ensemble standard deviations, are not very sensitive to particular choice of Bayesian traits. Averaged over the northern extra–tropical land, annual mean surface air temperature in the ensemble increases by 3.1 ± 1.4 K (ensemble mean±intra–ensemble standard deviation) during the 21st century. Precipitation robustly increases in the pan–Arctic and decreases in the Mediterranean/Black Sea region. The models agree on near–surface permafrost degradation during the 21st century. The area underlain by near–surface permafrost decreases from the contemporary value 20 ± 3 mln sq. km to 14 ± 3 mln sq. km in the late 21st century. This leads to risk for geocryological hazard due to soil subsidence. This risk is classified as moderate to high in the southern and western parts of Siberia and Tibet in Eurasia, and in the region from Alaska to the Labrador Peninsula in North America.

► Climate of the northern extratropical land is projected for the 21st century under SRES A1B scenario.
► Statistics for the GCM ensemble output is calculated by using Bayesian averaging.
► Averaged over the northern extra-tropical land, annual mean SAT grows by 3.1 ± 1.4 K.
► Near-surface permafrost area decreases to 14 ± 3 mln km2 leading to risk of soil subsidence.