Diagnostic Metrics for Evaluating Sub-Seasonal Predictions Based on Arctic-Midlatitude Teleconnection

Recently, a distinct midlatitude climate closely interconnected with the Arctic climate system has emerged as a robust testbed for assessing climate prediction models. Here, a comprehensive set of diagnostic metrics has been specifically developed to focus on Arctic-midlatitude teleconnections. This set comprises 10 items grouped into three sections: performance, teleconnection, and processes, encompassing 4 key variables. These quantitatively intuitive metrics concentrate on the fundamental aspects of the Arctic-midlatitude climate linkage. They encompass fields depicting climatology, teleconnection patterns, major variability modes, and unique characteristics of Arctic-midlatitude teleconnections. The metrics are applied to prediction performances derived from 3 models including the National Centers for Environmental Prediction Systems (NCEP CFS), Global Seasonal Forecasting System version 5 (GloSea5), along with the newly devised prediction systems configured by the Community Earth System Model 2 (CESM2). The evaluation of metrics employs Japanese 55-year reanalysis (JRA-55) data for three sub-seasons (October-November, December-January, February-March) spanning the boreal winters of 1996 to 2016. GloSea5 model stands out by notably excelling in 6 out of 12 metric items, particularly concerning simulated Arctic Oscillation (AO) variability, upper-level circulation patterns, and anomaly thickness related to boreal-winter temperatures. NCEP CFS exhibits reasonable performance across the metrics throughout the sub-seasons, with an exception being its deficiency in simulating upper-level atmospheric circulation patterns. CESM2 reveals clear shortcomings in capturing major teleconnection patterns such as Warm Arctic Cold Eurasia (WACE) compared to other models. Nevertheless, CESM2 shows a progressive enhancement of skills across the three sub-seasons. Possible origins that affect models’ performance are also discussed. The metrics provide a tool for evaluating the performance of climate prediction systems and facilitating the assessment of past and projected future changes in the boreal winter climate associated with the Arctic-midlatitude teleconnection.