资源环境科技发展态势分析平台

Climate change is expected to have severe impacts on natural systems as well as various socio-economic aspects of human life. This has urged scientific communities to improve the understanding of future climate and reduce the uncertainties associated with projections. In the present study, ten statistically downscaled CMIP5 GCMs at 1/16th deg. spatial resolution from two different downscaling procedures are utilized over the Columbia River Basin (CRB) to assess the changes in climate variables and characterize the associated uncertainties. Three climate variables, i.e. precipitation, maximum temperature, and minimum temperature, are studied for the historical period of 1970-2000 as well as future period of 2010-2099, simulated with representative concentration pathways of RCP4.5 and RCP8.5. Bayesian Model Averaging (BMA) is employed to reduce the model uncertainty and develop a probabilistic projection for each variable in each scenario. Historical comparison of long-term attributes of GCMs and observation suggests a more accurate representation for BMA than individual models. Furthermore, BMA projections are used to investigate future seasonal to annual changes of climate variables. Projections indicate significant increase in annual precipitation and temperature, with varied degree of change across different sub-basins of CRB. We then characterized uncertainty of future projections for each season over CRB. Results reveal that model uncertainty is the main source of uncertainty, among others. However, downscaling uncertainty considerably contributes to the total uncertainty of future projections, especially in summer. On the contrary, downscaling uncertainty appears to be higher than scenario uncertainty for precipitation.