Detection of Spatial Biases in GCM-based Predictions and Projections

Global forecasting models are used for assessing the impacts of climatic and environmental changes that affect different regions over different time scales. The spatial extent and regional connectivity of these changes raise questions regarding model performance and the impact of forecast uncertainty on decision making. Objective climate regionalization offers a tool to address these questions and to detect the spatial bias and regional shifts of GCM-based predictions and projections. In this work, we use climate regionalization to evaluate the forecasting models and to capture relevant similarities/differences between model outputs and observations. These regionalizations highlight the seasonal and even month-to-month specificity of regional climate associations, emphasizing the need to consider time of year as well as research question when defining a coherent region for climate analysis. Results show that some GCMs capture the climatic coherence of the regions and associated teleconnections, while other models can break a region into uncorrelated subregions or produce a similar variability that is spatially displaced from observations. Shifts in climate regions under projected twenty-first-century climate change for different GCMs and emissions pathways are examined. This regionalization approach was tested for different models at different time scales, to provide a way for using objective regionalization to detect spatial biases in global forecasting systems. It can also be extended to test whether dynamical downscaling can be used for spatial bias correction.