Global Mean

Main concepts

The global_mean command is based on the global_mean.py script which computes the global averages for many dynamical and physical fields. It compares the output against a set of pre-computed climatological values defined in the reference climatology (see below) and produces a table with the comparison. This includes the most important dynamical and physical fields for both the atmosphere and the ocean (e.g. land temperature, salinity, etc.). Different datasets are taken in consideration, providing an estimate of the interannual variability in the form of standard deviation.

Most importantly, it provides estimates for the radiative budget (including clouds radiative forcing) and for the hydrological cycle (including integrals over land and ocean) and other quantities useful for fast model assessment and for model tuning.

Usage

Running the global mean evaluation is rather simple:

global_mean EXP Y1 Y2

Alternatively, you can also run the python script in ecmean library:

./global_mean.py EXP Y1 Y2

Positional Arguments

EXP

Experiment identification.

Y1

Starting year of analysis.

Y2

Final year of analysis.

Optional Arguments

-h, --help

Show this help message and exit.

-s, --silent

Do not print anything to standard output.

--version

Show ECmean’s version number and exit.

-l LOGLEVEL, --loglevel LOGLEVEL

Define the level of logging. The default is ‘warning’.

-j NUMPROC

Specify the number of processors to use.

-c CONFIG, --config CONFIG

Set up a specific configuration file. The default is config.yml.

-i INTERFACE, --interface INTERFACE

Set up a specific interface file, overriding the configuration specified in config.yml.

-m MODEL, --model MODEL

Specify the model name, overriding the configuration specified in config.yml.

-e ENSEMBLE, --ensemble ENSEMBLE

Specify the variant label (ripf number for cmor).

-o DIR, --outputdir DIR

Specify the path of the output directory. This will create YAML and PDF folders for tables and figures.

--reference REFERENCE

Specify the reference dataset to use for comparison. The default is EC23. Other options include EC26-HIST, EC26-PDAY, and EC26-CMIP.

--trend

Compute trends on multiple years. This option is only available in table format.

--line

Append a single line to the table.

--addnan

Activate to plot also in the heatmap also fields which does not have a comparison against observations. Default is False.

Example

Usage example for CMIP5:

global_mean historical 1990 1999 -j 12 -m EC-EARTH -e r1i1p1 -i CMIP5

This will compute the global mean for member r1i1p1 of the EC-EARTH model in the CMIP5 historical experiment.

Output

Data are stored in machine-readable format in a YAML file, which includes much more details such as the global and regional mean over different seasons. In addition, ECmean produces also a figure including a more detailed comparison for different seasons and regions. This is available only for the datasets for which we have access to a gridded dataset.

Global mean figure for EC-Earth3

An example for a single year of the EC-Earth3 historical r1i1p1f1 simulation. Colors indicate the model bias as standard deviation of the interannual variability from observations. Blue implies negative bias, red positive bias. In each of the tiles the larger number shows the model value, while the smaller one is the reference value.

Finally, a txt table including the comparison with some predefined dataset, for the global mean yearly averages.

References available

Currently, two main different references for climatological values are available, EC26 and EC23, covering different observation periods.

EC26

EC26 is the updated reference framework for global mean observational datasets, which provide a flexible framework for evaluating climate models. It provides temporally consistent baselines tailored to different model forcing configurations, declined in three configurations (CMIP, HIST, PDAY) to match the intended timeframes. It also includes a global near-surface air temperature target, which was not available in the older reference. The variables are derived from a combination of observational and reanalysis products (e.g. CRU v4.09, Berkeley Earth, ERA5, MSWEP v2.80, CERES EBAF v4.2.1, ESA-CCI), depending on the physical quantity considered.

All configurations share the same variable definitions and masking strategy (global, land, ocean), but differ in their temporal averaging window to ensure consistency with the intended model experiments. All metadata (datasets, masks, periods and other properties) are defined in the corresponding YAML configuration file.

EC26-CMIP (1985–2014)

This reference dataset is designed for the evaluation of CMIP6 historical simulations against a consistent observational baseline. It is aligned with the CMIP6 historical period (1985–2014), ensuring temporal consistency between model climatologies and observational targets. Radiative fluxes are restricted to the satellite era (2000–2014), while ocean salinity follows its specific observational availability window.

Date used in GM Reference - EC26-CMIP

Long Name

Dataset

Mask

Period

2m Temperature (land-only)

CRU (v4.09)

land-no-antarctica

1985-2014

2m Temperature

BERKELEY-EARTH

global

1985-2014

Mean Sea Level Pressure

ERA5

global

1985-2014

Precipitation

MSWEP (v2.80)

global

1985-2014

Evaporation

ERA5

global

1985-2014

Precip. minus Evap.

ERA5

global

1985-2014

Total Cloud Cover

EBAF_v4.2.1 (v4.2.1)

global

2000-2014

Low Cloud Cover

ERA5

global

1985-2014

Medium Cloud Cover

ERA5

global

1985-2014

High Cloud Cover

ERA5

global

1985-2014

Precipitation (ocean)

MSWEP (v2.80)

ocean

1985-2014

Precip. minus Evap. (ocean)

ERA5

ocean

1985-2014

Precipitation (land)

MSWEP (v2.80)

land

1985-2014

Precip. minus Evap. (land)

ERA5

land

1985-2014

TOA Net

CERES EBAF (v4.2.1)

global

2000-2014

TOA SW Net

CERES EBAF (v4.2.1)

global

2000-2014

TOA LW Net

CERES EBAF (v4.2.1)

global

2000-2014

TOA SW Net (clear sky)

CERES EBAF (v4.2.1)

global

2000-2014

TOA LW Net (clear sky)

CERES EBAF (v4.2.1)

global

2000-2014

SW Cloud Forcing

CERES EBAF (v4.2.1)

global

2000-2014

LW Cloud Forcing

CERES EBAF (v4.2.1)

global

2000-2014

Surface Net SW

Wild 2020

global

Surface Net LW

Wild 2020

global

Surface Sensible Heat Flux

Wild 2020

global

Surface Latent Heat Flux

Wild 2020

global

Net Surface (no snowfall)

Wild 2020

global

Net Surface

Wild 2020

global

TOA - Sfc Net Radiation (no snow)

None

global

TOA - Sfc Net Radiation

None

global

Sea Surface Temperature

ESA-CCI-L4 (v3.0.1)

ocean

1985-2014

Sea Surface Salinity

Copernicus Global Analysed Sea Surface Salinity

ocean

1993-2014

SSH

None

global

Net Water Flux into ocean

None

global

Sea Ice Area

ESA-CCI-L4 (v3.0.1)

ocean

1985-2014

EC26-HIST (1981–2010)

This configuration is designed for comparison with model simulations using historical forcing or present-day forcing fixed around 1990. The reference period spans 1981–2010 where observational coverage allows. Radiative fluxes follow the CERES satellite era (starting in 2000), while other variables use the 1981–2010 window.

Date used in GM Reference - EC26-HIST

Long Name

Dataset

Mask

Period

2m Temperature (land-only)

CRU (v4.09)

land-no-antarctica

1981-2010

2m Temperature

BERKELEY-EARTH

global

1981-2010

Mean Sea Level Pressure

ERA5

global

1981-2010

Precipitation

MSWEP (v2.80)

global

1981-2010

Evaporation

ERA5

global

1981-2010

Precip. minus Evap.

ERA5

global

1981-2010

Total Cloud Cover

EBAF_v4.2.1 (v4.2.1)

global

2000-2010

Low Cloud Cover

ERA5

global

1981-2010

Medium Cloud Cover

ERA5

global

1981-2010

High Cloud Cover

ERA5

global

1981-2010

Precipitation (ocean)

MSWEP (v2.80)

ocean

1981-2010

Precip. minus Evap. (ocean)

ERA5

ocean

1981-2010

Precipitation (land)

MSWEP (v2.80)

land

1981-2010

Precip. minus Evap. (land)

ERA5

land

1981-2010

TOA Net

CERES EBAF (v4.2.1)

global

2000-2010

TOA SW Net

CERES EBAF (v4.2.1)

global

2000-2010

TOA LW Net

CERES EBAF (v4.2.1)

global

2000-2010

TOA SW Net (clear sky)

CERES EBAF (v4.2.1)

global

2000-2010

TOA LW Net (clear sky)

CERES EBAF (v4.2.1)

global

2000-2010

SW Cloud Forcing

CERES EBAF (v4.2.1)

global

2000-2010

LW Cloud Forcing

CERES EBAF (v4.2.1)

global

2000-2010

Surface Net SW

Wild 2020

global

Surface Net LW

Wild 2020

global

Surface Sensible Heat Flux

Wild 2020

global

Surface Latent Heat Flux

Wild 2020

global

Net Surface (no snowfall)

Wild 2020

global

Net Surface

Wild 2020

global

TOA - Sfc Net Radiation (no snow)

None

global

TOA - Sfc Net Radiation

None

global

Sea Surface Temperature

ESA-CCI-L4 (v3.0.1)

ocean

1981-2010

Sea Surface Salinity

Copernicus Global Analysed Sea Surface Salinity

ocean

1993-2010

SSH

None

global

Net Water Flux into ocean

None

global

Sea Ice Area

ESA-CCI-L4 (v3.0.1)

ocean

1981-2010

EC26-PDAY (2000–2024)

This configuration is intended for evaluation of model simulations using present-day forcing conditions representative of the 2010–2012 period. The reference period spans 2000–2024 (or the maximum available year depending on dataset availability; e.g. 2023 for ESA-CCI-L4 products). By using a more recent averaging window, EC26-PDAY reflects contemporary radiative balance and hydrological cycle conditions, making it suitable for fixed present-day forcing experiments.

Date used in GM Reference - EC26-PDAY

Long Name

Dataset

Mask

Period

2m Temperature (land-only)

CRU (v4.09)

land-no-antarctica

2000-2024

2m Temperature

BERKELEY-EARTH

global

2000-2024

Mean Sea Level Pressure

ERA5

global

2000-2024

Precipitation

MSWEP (v2.80)

global

2000-2024

Evaporation

ERA5

global

2000-2024

Precip. minus Evap.

ERA5

global

2000-2024

Total Cloud Cover

EBAF_v4.2.1 (v4.2.1)

global

2000-2024

Low Cloud Cover

ERA5

global

2000-2024

Medium Cloud Cover

ERA5

global

2000-2024

High Cloud Cover

ERA5

global

2000-2024

Precipitation (ocean)

MSWEP (v2.80)

ocean

2000-2024

Precip. minus Evap. (ocean)

ERA5

ocean

2000-2024

Precipitation (land)

MSWEP (v2.80)

land

2000-2024

Precip. minus Evap. (land)

ERA5

land

2000-2024

TOA Net

CERES EBAF (v4.2.1)

global

2000-2024

TOA SW Net

CERES EBAF (v4.2.1)

global

2000-2024

TOA LW Net

CERES EBAF (v4.2.1)

global

2000-2024

TOA SW Net (clear sky)

CERES EBAF (v4.2.1)

global

2000-2024

TOA LW Net (clear sky)

CERES EBAF (v4.2.1)

global

2000-2024

SW Cloud Forcing

CERES EBAF (v4.2.1)

global

2000-2024

LW Cloud Forcing

CERES EBAF (v4.2.1)

global

2000-2024

Surface Net SW

Wild 2020

global

Surface Net LW

Wild 2020

global

Surface Sensible Heat Flux

Wild 2020

global

Surface Latent Heat Flux

Wild 2020

global

Net Surface (no snowfall)

Wild 2020

global

Net Surface

Wild 2020

global

TOA - Sfc Net Radiation (no snow)

None

global

TOA - Sfc Net Radiation

None

global

Sea Surface Temperature

ESA-CCI-L4 (v3.0.1)

ocean

2000-2023

Sea Surface Salinity

Copernicus Global Analysed Sea Surface Salinity

ocean

2000-2024

SSH

None

global

Net Water Flux into ocean

None

global

Sea Ice Area

ESA-CCI-L4 (v3.0.1)

ocean

2000-2023

EC23

This is from the old version of ECmean and does not include values for global tas. This reference dataset collects global mean observational targets used by the global_mean.py script to compute global averages. The variables are derived from a combination of observational and reanalysis products (e.g. CRU, ERA5, MSWEP, CERES-EBAF, ESA-CCI, Wild 2020), depending on the physical quantity considered. Most fields are defined over the 1991–2020 period, while other variables use shorter observational windows due to data availability. All metadata (datasets, masks, periods and other properties) are defined in the corresponding YAML configuration file.

Date used in GM Reference - EC23

Long Name

Dataset

Mask

Period

2m Temperature (land-only)

CRU

land-no-antarctica

1991-2020

Mean Sea Level Pressure

ERA5

global

1991-2020

Precipitation

MSWEP

global

1991-2020

Evaporation

ERA5

global

1991-2020

Precip. minus Evap.

ERA5

global

1991-2020

Total Cloud Cover

CERES-EBAF

global

2000-2020

Low Cloud Cover

ERA5

global

1991-2020

Medium Cloud Cover

ERA5

global

1991-2020

High Cloud Cover

ERA5

global

1991-2020

Precipitation (ocean)

MSWEP

ocean

1991-2020

Precip. minus Evap. (ocean)

ERA5

ocean

1991-2020

Precipitation (land)

MSWEP

land

1991-2020

Precip. minus Evap. (land)

ERA5

land

1991-2020

TOA Net

CERES-EBAF

global

2000-2020

TOA SW Net

CERES-EBAF

global

2000-2020

TOA LW Net

CERES-EBAF

global

2000-2020

TOA SW Net (clear sky)

CERES-EBAF

global

2000-2020

TOA LW Net (clear sky)

CERES-EBAF

global

2000-2020

SW Cloud Forcing

CERES-EBAF

global

2000-2020

LW Cloud Forcing

CERES-EBAF

global

2000-2020

Surface Net SW

Wild 2020

global

Surface Net LW

Wild 2020

global

Surface SH

Wild 2020

global

Surface LH

Wild 2020

global

Net Surface (no snowfall)

Wild 2020

global

Net Surface

Wild 2020

global

Sea Surface Temperature

ESA-CCI-L4

ocean

1991-2020

Sea Surface Salinity

ESA-CCI

ocean

2010-2020

SSH

None

global

Net Water Flux into ocean

None

global

TOA - Sfc Net Radiation (no snow)

None

global

TOA - Sfc Net Radiation

None

global

Sea Ice Area

ESA-CCI-L4

ocean

1991-2020

Sea Ice Area (Northern Hemisphere)

ESA-CCI-L4

north

Sea Ice Area (Southern Hemisphere)

ESA-CCI-L4

south

Reference climatology computation

Reference climatology are computed by the ecmean/utils/reference-create.py script, which is included in the repository for documentation. It is based on a YML file which is tells the script where to retrieve the data, identifying all the required data folder and names. Of course, in the remote case you would like to develop a new climatology, you can create your own YML file and run the script to produce the reference climatology Examples are the create-reference-wilma-EC26.yml and create-reference-wilma-EC23.yml files, which are used to produce the EC26 and EC23 reference climatology, respectively. The results are produced into a YML file for in ecmean/reference/gm_reference_EC**.yml which includes the global and regional mean over different seasons as well the interannual standard deviation. Full details on the datasets used are found there.