Climate Model Metrics
PCMDI Simulation Summaries (v1.2.0) [Go to latest]
The PCMDI Metrics Package (PMP) is a capability that can be used to produce a diverse suite of objective summaries of Earth System Model (ESM) agreement with observations. At PCMDI, we are routinely applying the PMP to multiple generations of CMIP, including the most recent results from CMIP6 as they become available. The summaries available below will be regularly updated with new and an expanding set of results.
Mean Climate (results)
- Using well-established statistics, we provide large-scale seasonal and mean state climatology comparisons between CMIP simulations and observationally-based data. These include traditional measures (e.g. bias, pattern correlation and root-mean-square error) for global, hemispheric, tropical, extra-tropical, and other selected domains using satellite data and atmospheric reanalysis as references. These statistics are routinely computed as part of model evaluation. We use summary diagrams developed by PCMDI scientists (Taylor 2001; Gleckler et al. 2008) to objectively compare the consistency between the observed and simulated climate.
Back to List
- Based on the work of Lee et al. (2019), we present skill metrics for the Northern Annular Model (NAM), the North Atlantic Oscillation (NAO), the Southern Annular Mode (SAM), the Pacific North American pattern (PNA), the North Pacific Oscillation (NPO), the Pacific Decadal Oscillation (PDO), and the North Pacific Gyre Oscillation (NPGO). For NAM, NAO, SAM, PNA, and NPO the results are based on sea-level pressure, while the results for PDO and NPGO are based on sea surface temperature. Our approach distinguishes itself from other studies that analyze modes of variability in that we use the Common Basis Function approach (CBF), in which model anomalies are projected onto the observed modes of variability. Using the Historical simulations, the skill of the spatial patterns is given by the Root-Mean-Squared-Error (RMSE), and the Amplitude gives the standard deviation of the Principal Component time series. The skill metrics are calculated with respect to a primary and secondary sets of observations denoted by the triangles in each cell of the Portrait Plots.
Back to List
Monsoon Precipitation: Fractional Accumulation (results)
- Based on the work of Sperber and Annamalai (2014), we present skill metrics that indicate how well models simulate the onset, decay, and duration of monsoon based on the analysis of climatological pentads of precipitation. Using Historical simulations, the results are based on area-averaged data for All-India Rainfall (AIR), Sahel, Gulf of Guinea (GoG), North American Monsoon (NAM), South American Monsoon (SAM), and Northern Australia (AUS).
Back to List
Madden-Julian Oscillation (results)
- Based on the work of Ahn et al. (2017), we present skill metrics that indicate how well models simulate eastward propagation of the MJO. We apply frequency-wavenumber decomposition to precipitation from observations (GPCP-based; 1997-2010) and the CMIP5 and CMIP6 Historical simulations (1985-2004).
Back to List
Update History
- v1.2.0 (29 August 2019): Mean climate metrics added
- v1.1.0 (18 July 2019): MJO metrics added
- v1.0.0 (20 June 2019): Initial public release
- v1.0.0-beta (18 June 2019): Addition of monsoon precipitation onset, decay, and duration in CMIP5.
- v1.0.0-alpha (31 May 2019): Release of summary statistics for extratropical modes of variability in CMIP5 and CMIP6.
Back to List
Questions or comments about the PCMDI Simulation Summaries should be sent to the PMP team.