The CMIP6 multi-model ensembles technical documentation

This is the technical documentation for the multi-model ensembles of Coupled Model Intercomparison Project Phase 6 (CMIP6) global climate models (GCMs) provided on the Canadian Climate Data and Scenarios (CCDS) website. The documentation provides a description of the CMIP6 multi-model ensemble datasets, ensemble methodology, and a list of models included in the ensembles.

On this page

• Overview
• Data and processing
• Data download and interpolation
• Programs and languages
• Calculation of anomalies
• Ensemble statistics
• Thresholds for sea ice variables
• Global gridded datasets
• Tabular datasets
• Map datasets and plots
• Time series datasets and plots
• List of models in ensembles
• Dataset licence

Overview

Climate projections vary across GCMs due to differences in the representation and approximation of earth systems and processes, and natural variability and uncertainty regarding future climate drivers. As a result, it is common practice to use a multi-model ensemble, or average across many GCMs, as opposed to using a single GCM in isolation. For this purpose, the Canadian Climate Data and Scenarios (CCDS) site has global and national multi-model ensembles for a suite of variables, scenarios, future time periods, and percentiles.

Provided on CCDS are four types of products based on the CMIP6 multi-model GCM ensembles regridded to a common 1x1 degree grid resolution: time series datasets and plots, maps and associated datasets, tabular datasets, and global gridded datasets. Results are available for four future periods, and up to five percentiles and six Shared Socioeconomic Pathways (SSPs) of the CMIP6 ensemble distribution (see Table 1 for more details). Projected changes are expressed as anomalies according to a historical reference period of 1995-2014. Please see Table 1 for a description of the main characteristics of each product.

Data and processing

CMIP6 multimodel ensembles were produced in phases. The first phase included temperature and precipitation variables. The second phase added near-surface wind speed and snow depth, and the third phase added sea ice concentration and thickness. Processing techniques varied among phases and/or variables (see below for more details). Additional variables may be added in the future.

Data download and interpolation

Monthly CMIP6 GCM datasets were downloaded from the Earth System Grid Federation (ESGF) online database in NetCDF format.

Each model simulation was interpolated to a common 1x1 degree global grid prior to any analysis or processing. Temperature and precipitation were regridded using the Climate Data Operators (CDO) version 1.9.3 bilinear interpolation function (‘remapbil’). Near-surface wind speed and snow depth were regridded using the Iris Python package version 3.1.0 and the nearest-neighbour interpolation regridding scheme (scheme: ‘nearest’). The more complex native grids of the sea ice variables necessitated using the nearest neighbour remapping function (‘remapnn’) from CDO (version 2.0.5) for interpolation. See Table 2 for a list of included models.

Programs and languages

Temperature, precipitation, sea ice concentration, and sea ice thickness variables were processed in CDO to calculate the CMIP6 multi-model ensemble statistics (functions: ‘ensmean’ or ‘enspctl’), time periods (functions: ‘timmean’, ‘selyear’, ‘yearmean’), and anomalies. Temperature and precipitation were processed in R version 4.1.0. to produce the time series and map plots as well as the tabular data, while sea ice variables were processed in Python version 3.10.4.

ESMValTool version 2.5.0 was used to regrid near-surface wind speed and snow depth (preprocessors: ‘regrid’) and calculate multi-model ensemble statistics (preprocessors: ‘multi_model_statistics’), time periods (preprocessor: ‘seasonal statistics’ or ‘annual_statistics’), and anomalies. Python version 3.10.4 was used to produce the time series and map plots as well as the tabular datasets. Note that expansions to the original temperature and precipitation dataset (tabular datasets, and gridded data ensemble percentiles) utilized ESMValTool.

The number of models in each ensemble differ according to model availability for each SSP and variable. See Table 2 for additional details. All scripts were executed on a Unix system in a Bourne-again shell (bash) environment. Only one realization member from each model was included in the multi-model ensembles, giving each model equal weight.

Calculation of anomalies

Projected changes in climate are provided as anomalies according to a historical reference period of 1995-2014 for all anomaly datasets and products. Anomalies (n) are calculated by subtracting the mean over the historical reference period (µ) from original values at the time of interest (x) using the following expression:

$n_i=x_i-{\mu }_{\mathrm{1995-2014}}$

Projected anomalies (n) for a future time period are calculated by subtracting the mean over the historical reference period (μ) from the mean of the projected values of the future time period (x) using the following expression:

$n_{\mathrm{2081-2100}}=x_{\mathrm{2081-2100}}-{\mu }_{\mathrm{1995-2014}}$

Ensemble statistics

Ensemble statistics, specifically the 5th, 25th, 50th (median), 75th, and 95th percentiles, are provided as a statistical measure to demonstrate some of the range of uncertainty across model projections.

Thresholds for sea ice variables

The CMIP6 sea ice variables have set thresholds for the historical time periods of anomaly calculations, below which the ocean is considered ice-free. This was done so that projected changes in sea ice would not be skewed by grid cells with historically low values. Grid cells with low values may contain ice that has not fully formed (sea ice thickness), or that is not connected to a larger mass of ice, more accurately described as ice floe (sea ice concentration).

For sea ice concentration anomalies, if in the historical period the concentration of sea ice was equal to 15 percent or less, then the value was converted to missing, as such low coverage constitutes open ocean. For sea ice thickness anomalies, values between 0 and 0.1 m in the historical period were converted to missing, as values under 10 cm are considered ‘new ice’.

Global gridded datasets

Global gridded datasets for temperature, precipitation, near-surface wind speed, sea ice concentration, sea ice thickness, and snow depth are available at a monthly temporal time scale as raw (actual) values for the multi-model ensembles and for the individual CMIP6 models used to create the ensembles. Historical simulations (1850-2014) and model projections for the six SSPs (2015-2100) (when available) are available for download on the common 1x1 degree global grid. See Table 1 for additional product details and Table 2 for the ensemble model list.

Tabular datasets

Tabular datasets of future projected changes in temperature, precipitation and snow depth are available for Canada and Canadian provinces and territories at annual and seasonal time scales. Tabular data are calculated with areal means based on a common 1x1 degree grid. See Table 1 for additional product details and Table 2 for the ensemble model list.

Map datasets and plots

National maps of future projected changes in temperature, precipitation, near-surface wind speed, sea ice concentration, sea ice thickness, and snow depth are available at annual and seasonal time scales. The associated anomaly datasets (global) are also available for download. See Table 1 for additional product details and Table 2 for the ensemble model list.

Time series datasets and plots

Time series plots are available for Canada at annual and seasonal time scales. The areal mean of the time series was computed on the common 1x1 degree grid. The time series plots display the range of model simulations (i.e., model spread) extending from the historical time period (1900-2014) to future projected changes (2015-2100) for the six SSPs. See Table 1 for additional product details. The time series plots also incorporate boxplots that display the 25th, 50th (median), and 75th percentiles of model distribution for 2081-2100 for each SSP. Global time series datasets of the multi-model ensemble mean are also available for download. See Table 2 for the ensemble model list.

List of models in ensembles

Dataset licence

Open Government Licence - Canada (http://open.canada.ca/en/open-government-licence-canada)

Individual model datasets and all related derived products, including the multi-model ensembles, are subject to the terms of use of the source organization.