Evaluation of Leading Modes of Climate Variability in the CMIP Archives
The Climate Variability and Diagnostics Package is applied to simulations in the CMIP3/5/6 archives. Metrics of fidelity are computed using best-available observations both within and across CMIP generations. Bias patterns are decomposed using Principle Component Analysis and models are scored using pattern correlations. While contrasts within and across CMIP generations are found to be substantial, progressive improvement is also detectible and characterized by improved teleconnections and time-space variability.
The improved skill of climate models in representing internal modes is documented, suggesting enhanced capabilities for performing attribution, prediction, and projection.
Over time, climate models are shown to have improved in simulating the strength, pattern, and transient evolution of ENSO and the PDO. Progress in simulating the spectra of these modes and the patterns of extratropical modes is found to be more mixed and also varies considerably within CMIP generation.