Page History

Introduction

What are decadal climate predictictions?

Decadal climate predictions (not quite a forecast not quite a climate scenario but rather a middle ground).

The Decadal Climate Prediction Project (DCPP)

The Decadal Climate Prediction Project addresses a range of scientific issues involving  (Boer et al., 2016) addresses the ability of the climate system to be predicted on annual to , multi-annual and decadal timescales, the skill that is currently and potentially available, the mechanisms involved in long timescale variability, and the production of forecasts of benefit to both science and society.

DCPP, part of CMIP6

The DCPP as part of CMIP6

The CMIP6 data archive is distributed through the Earth System Grid Federation (ESGF) though many national centres have either a full or partial copy of the data. A quality-controlled subset of CMIP6 data are made available through the Climate Data Store (CDS) for the users of the Copernicus Climate Change Service (C3S).

Decadal Climate Prediction Project Data in the CDS

The decadal climate prediction project (DCPP) data in the Climate Data Store (CDS) are a quality-controlled subset of the wider DCPP data. The CDS provides data for two DCPP experiments dcppA-hindcast and dcppB-forecast from four modelling centres.

The CDS subset of CMIP6 data has been through a quality control procedure which ensures a high standard of dependability of the data. It may be for example, that similar data can be found in the main CMIP6 ESGF archive however these data come with very limited quality assurance and may have metadata errors or omissions.

DCPP Experiments

The two DCPP experiments that are included in the C3S Climate Data Store, dcppA-hindcast and dcppB-forecast, are described in Table 2.1 below. The experiment descriptions are harvested from ES-DOC. We use content from the ES-DOC experiment extended_description attribute.

...

Experiment Name

...

Experiment Long Name

...

Extended Description

...

dcppA-hindcast

...

hindcasts initialized from observations with historical forcing

dcppA-hindcast is a set of retrospective decadal forecasts (known as hindcasts) that are initialised every year from and performed with a coupled atmosphere-ocean general circulation model (AOGCM).

Prior to the year 2020 the models running these simulations are forced with historical conditions that are consistent with observations, these conditions include atmospheric composition, land use, volcanic aerosols and solar forcing. For simulations that extend b (e.g. 2015-2025) the models are forced with future conditions from the ssp245 scenario for the period from 2020 to the end of the simulation.

There are 10 ensemble members for each start date and simulations run for 10 years.

. The information generated by the DCPP and archived on the Earth System Grid Federation (ESGF) nodes that is made accessible in the CDS can provide a basis for socially relevant operational climate predictions on annual to decadal timescales.

DCPP: Part of CMIP6

The Decadal Climate Prediction Project (DCPP) is a contributing MIP (Model Intercomparison Project) to the 6th Coupled Model Intercomparison Project (CMIP6 – Eyring et al., 2016), which is running as part of the World Climate Research Programme (WCRP).  DCPP addresses a range of scientific issues involving the ability of the climate system to be predicted on annual to decadal timescales, the skill that is currently and potentially available, the mechanisms involved in long timescale variability, and the production of forecasts of benefit to both science and society.

The CMIP6 data archive is distributed through the ESGF. A quality-controlled subset of CMIP6 global climate projection data are made available through the Climate Data Store (CDS) for the users of the Copernicus Climate Change Service (C3S). Dedicated ESGF data nodes are used for C3S in France (at IPSL) and in Germany (DKRZ).  Similarly, the decadal climate prediction project (DCPP) data in the Climate Data Store (CDS) are a targeted, quality-controlled subset of the DCPP commissioned by C3S.

The published datasets are the ones which took part on the C3S sectoral demonstrator service. This demonstrator provided decadal prediction products tailored to specific users from the agriculture, energy, infrastructure and insurance sectors (see details at https://climate.copernicus.eu/sectoral-applications-decadal-predictions). The data were used in these demonstrators following processing procedures necessary to extract valid information (e.g., bias adjustment); details on this processing are available in the technical appendix at https://climate.copernicus.eu/sites/default/files/2021-09/Technical_appendix_2020.pdf. Any application - similar to or different from these examples - needs to consider and apply the required data processing with care.

Decadal Climate Prediction Project Data in the CDS

DCPP Experiments

The CDS provides data access to two DCPP experiments: dcppA-hindcast which consists of retrospective decadal forecasts that can be used to assess historical decadal prediction skill, and dcppB-forecast which are experimental quasi-real-time decadal forecasts that form a basis for potential operational forecast production. For these DCPP experiments, each model performs multiple overlapping simulations that are initialised annually throughout the experiment. The dcppA-hindcast and dcppB-forecast experiments are further described in the table below. The DCPP experiment descriptions presented here are based on information harvested from Earth System Documentation (ES-DOC).

Models

Data for the dcppA-hindcast and dcppB-forecast experiments published in the CDS are generated from simulations run by the models described in the table below. The model descriptions presented here are harvested from the dataset DOI pages held at the World Data Centre for Climate (WDCC), further model details can be found on the ES-DOC. The EC-Earth3, MPI-ESM1-2-HR, MPI-ESM1-2-LR and HadGEM3-GC31-MM models were configured with 360-day years (where every month has 30 days), whereas the CMCC-CM2-SR5 model was configured with a 365 day year (with an irregular number of days in each month). 

DCPP hindcasts can be used to assess and understand the historical decadal prediction skill of climate models.

...

dcppB-forecast

...

forecast initialized from observations

...

dcppB-forecast is a set of real-time decadal forecasts that are initialised every year from 2019 in real time and ongoing. The forecasts are performed with the same coupled atmosphere-ocean general circulation model (AOGCM), which was used to generate the hindcast data.

The models running these simulations are initialised using observed data, thereafter they are forced with future conditions from the ssp245 scenario that include atmospheric composition, land use, and solar forcing.

There are 10 ensemble members for each start date and simulations run for 10 years.

Models

Data for the dcppA-hindcast and dcppB-forecast experiments are generated from simulations run by models described in the table below. The model descriptions are harvested from the dataset DOIs held at the World Data Centre for Climate (WDCC)., further details can be found on the Earth System Documentation site.

Grids

CMIP6 data is reported either on the model’s native grid or re-gridded to one or more target grids with data variables generally provided near the centre of each grid cell (rather than at the boundaries).  For CMIP6 there is a requirement to record both the native grid of the model and the grid of its output (archived in the CMIP6 repository) as a “nominal_resolution”.  The "nominal_resolution” enables users to identify which models are relatively high resolution and have data that might be challenging to download and store locally. Information about the grids can be found in the model table above, under 'Model Details' and within the NetCDF file metadata.

Ensembles

Each modelling centre typically run the same experiment using the same model with slightly different settings 10 times to confirm the robustness of results and inform sensitivity studies through the generation of statistical information. A model and its collection of runs is referred to as an ensemble. Within these ensembles, four different categories of sensitivity studies are done, and the resulting individual model runs are labelled by four integers indexing the experiments in each category

e.g. r<W>i<X>p<Y>f<Z>, where W, X, Y and Z are positive integers as defined below:

• The first category, labelled realization_index (referred to with letter r), performs experiments which differ only in random perturbations of the initial conditions of the experiment. Comparing different realizations allow estimation of the internal variability of the model climate.
• The second category, labelled initialization_index (referred to with letter i), refers to variation in initialisation parameters. Comparing differently initialised output provides an estimate of how sensitive the model is to initial conditions.
• The third category, labelled physics_index (referred to with letter p), refers to variations in the way in which sub-grid scale processes are represented. Comparing different simulations in this category provides an estimate of the structural uncertainty associated with choices in the model design.
• The fourth category labelled forcing_index (referred to with letter f) is used to distinguish runs of a single CMIP6 experiment, but with different forcings applied.

Parameter listings

Data for dcppA-hindcast experiment and the dcppB-forecast experiments will include parameters at monthly and daily resolution as described in the tables below. The parameter descriptions are harvested from the CMIP6 Data Request via the CLIPC variable browser.

Monthly Parameters

...

ESGF variable id

...

units

...

Standard name

...

Long name

...

Description

...

tas

...

K

...

air_temperature

...

Near-Surface Air Temperature

...

Near-surface (2 meter) air temperature

...

pr

...

kg m^-2 s^-1

...

precipitation_flux

...

Precipitation

...

Includes both liquid and solid phases

...

psl

...

Pa

...

air_pressure_at_sea_level

...

Sea Level Pressure

...

Sea level pressure

Daily Parameters

...

Variable name

...

units

...

Standard name

...

Long name

...

Description

...

tasmin

...

K

...

air_temperature

...

Daily Minimum Near-Surface Air Temperature

...

Daily-minimum near-surface (2 meter) air temperature

...

tasmax

...

K

...

air_temperature

...

Daily Maximum Near-Surface Air Temperature

...

Daily-maximum near-surface (2 meter) air temperature

...

pr

...

kg m^-2 s^-1

...

precipitation_flux

...

Precipitation

...

Includes both liquid and solid phases

...

zg500

...

m

...

geopotential_height

...

Geopotential Height at 500hPa

...

Geopotential height of the 500 hPa surface

Start-Date Ensembles

The DCPP experiments published in the CDS, are a suite of overlapping simulations that are initialised every year throughout the duration of the start-date range specified by the experiment. The simulations begin in November to allow for DJF (December, January, February) seasonal averages to be calculated. There are 10 simulations (ensemble members) for each start-date (called "Base year" in the CDS form), except for the MPI-ESM1-2-LR model which has 16 ensemble members.

The start-date ensemble is reflected in the DCPP data naming convention with the addition of a s<yyyy> start-date ensemble identifier. Please note that the conventional CMIP6 ripf ensemble identifiers are omitted for this particular dataset since all the ensemble members are concatenated into one file.

See some more more details in the File naming conventions and In-file metadata modifications sections below.

Practical details of the published data

In the table below some practical details of the data is shown including the base year (or start year) period covered and the number of ensemble members. For each start year there are (at least) 10 years of corresponding hindcast or forecast data available. Hindcast and forecast start years are not distinguished in the CDS form. Please note that the ensemble members are not available individually, but they are concatenated into one file while the data is downloaded, and generally users are encouraged to use all members instead of selecting one member of the predictions. 

*Note: Since hindcast and forecast data begins in November, the actual period the data covers includes only November and December for each start year, however the last year includes November and December. For example, for the 1960 start year, 1960 includes November and December and 1961 - 1970 have full coverage. 

Parameter listings

Data for the dcppA-hindcast experiments and the dcppB-forecast experiments will include parameters at monthly and daily resolution as described in the tables below. The parameter descriptions presented here are harvested from the CMIP6 Data Request via the CLIPC variable browser.

Grids

DCPP data like the rest of CMIP6 is reported either on the model’s native grid or re-gridded to one or more target grids with data variables generally provided near the centre of each grid cell (rather than at the boundaries). A grid_label (found in the file name following the ensemble identifier and also in the file's global metadata attributes) indicates whether the data is provided on the model's native grid (gn) or has been re-gridded (gr) to a target grid. For DCPP data in the CDS, only data from the EC-Earth3 model has been re-gridded to a target grid, data from the other models are provided on each model's native grid.  The file's "nominal_resolution" global metadata attribute gives an indication of the resolution of the data, for the DCPP data in the CDS the nominal resolution of the models is 100km (except for the MPI-ESM1-2-LR model, which is 250 km). 

 Data Format

The CDS subset of DCPP data are provided as NetCDF files. NetCDF (Network Common Data Form) is a file format that is freely available and commonly used in the climate modelling community. See more details:  What are NetCDF files and how can I read them

 A CMIP6 NetCDF file in the CDS contains:

• global metadata: these fields can describe many different aspects of the file such as
  • when the file was created
  • the name of the institution and model used to generate the file
  • Information on the horizontal grid and regridding procedure
  • links to peer-reviewed papers and technical documentation describing the climate model,
  • links to supporting documentation on the climate model used to generate the file,
  • software used in post-processing.
• variable dimensions: such as time, latitude, longitude and height
• variable data: the gridded data
• variable metadata: e.g. the variable units, averaging period (if relevant) and additional descriptive data

File naming conventions

When you download a DCPP file from the CDS it will have a naming convention that is as follows:

<variable_id>_<table_id>_<source_id>_<variant_label>.nc

 Where:

• variable_id: variable is a short variable name, e.g. “tas” for “temperature at the surface”.
• table_id: this refers to the MIP table being used. The MIP tables are used to organise the variables. For example, Amon refers to monthly atmospheric variables and Oday contains daily ocean data.
• source_id: this refers to the model used that produced the data.
• variant_label: is a label constructed from the start year of the simulation as  s<yyyy>, where yyyy is the start year

Data Format

The CDS subset of DCPP data are provided as NetCDF files. NetCDF (Network Common Data Form) is a file format that is freely available and commonly used in the climate modelling community. See more details:  What are NetCDF files and how can I read them

 A CMIP6 NetCDF file in the CDS contains:

• Global metadata: these fields can describe many different aspects of the file such as
  • when the file was created
  • the name of the institution and model used to generate the file
  • Information on the horizontal grid and regridding procedure
  • links to peer-reviewed papers and technical documentation describing the climate model,
  • links to supporting documentation on the climate model used to generate the file,
  • software used in post-processing.
• variable dimensions: such as time, latitude, longitude and height
• variable data: the gridded data
• variable metadata: e.g. the variable units, averaging period (if relevant) and additional descriptive data

File naming conventions

When you download a DCPP file from the CDS it will have a naming convention that is as follows:

<variable_id>_<table_id>_<source_id>_<experiment_id>_<variant_label>_<grid_label>_<time_range>.nc

 Where:

• variable_id: variable is a short variable name, e.g. “tas” for “temperature at the surface”.
• table_id: this refers to the MIP table being used. The MIP tables are used to organise the variables. For example, Amon refers to monthly atmospheric variables and Oday contains daily ocean data.
• source_id: this refers to the model used that produced the data.
• experiment_id: refers to the set of experiments being run for CMIP6. For example, PiControl, historical and 1pctCO2 (1 percent per year increase in CO2).
• variant_label: is a label constructed from 4 indices (ensemble identifiers) r<W>i<X>p<Y>f<Z>, where W, K, Y and Z are integers.
• grid_label: this describes the model grid used. For example, global mean data (gm), data reported on a model's native grid (gn) or regridded data reported on a grid other than the native grid and other than the preferred target grid (gr1).
• time_range: the temporal range is in the form YYYYMM[DDHH]-YYYY[MMDDHH], where Y is year, M is the month, D is day and H is hour. Note that day and hour are optional (indicated by the square brackets) and are only used if needed by the frequency of the data. For example, daily data from the 1st of January 1980 to the 31st of December 2010 would be written 19800101-20101231.

Quality control of the CDS-CMIP6-DCPP subset

The CDS subset of the DCPP data have been through a set of quality control checks before being made available through the CDS. The objective of the quality control process is to ensure that all files in the CDS meet a minimum standard. Data files were required to pass all stages of the quality control process before being made available through the CDS. Data files that fail the quality control process are excluded from the CDS-CMIP6-DCPP subset, data providers are contacted and if they are able to release a new version of the data with the error corrected then providing this data passes all remaining QC steps may be available for inclusion in the next CMIP6 DCPP data release.

 The main aim of the quality control procedure is to check for metadata and gross data errors in the CMIP6 files and datasets. A brief description of each of the QC checks is provided here:

1. CF-Checks: The CF-checker tool checks that each NetCDF4 file in a given dataset is compliant with the Climate and Forecast (CF) conventions, compliance ensures that the files are interoperable across a range of software tools.
2. PrePARE: The PrePARE software tool is provided by PCMDI (Program for Climate Model Diagnosis and Intercomparison) to verify that CMIP6 files conform to the CMIP6 data protocol. All CMIP6 data should meet this required standard however this check is included to ensure that all data supplied to the CDS have passed this QC test.
3. nctime: The nctime checker checks the temporal axis of the NetCDF files. For each NetCDF file the temporal element of the file is compared with the time axis data within the file to ensure consistency. For a time-series of data comprised of several NetCDF files nctime ensures that the entire timeseries is complete, that there are no temporal gaps or overlaps in either the filename or in the time axes within the files.
4. Errata: The dataset is checked to ensure that no outstanding Errata record exists.
5. Data Ranges: A set of tests on the extreme values of the variables are performed, this is used to ensure that the values of the variables fall into physically realistic ranges.
6. Handle record consistency checks: This check ensures that the version of the dataset used is the most recently published dataset by the modelling centre, it also checks for any inconsistency in the ESGF publication and excludes any datasets that may have an inconsistent ESGF publication metadata.
7. Exists at all both partner sites: It is asserted that each dataset exists at all three partner sites CEDA, DKRZ and IPSLboth partner ESGF data nodes at IPSL and DKRZ.

It is important to note that passing these quality control tests should not be confused with validity: for example, it will be possible for a file to pass all QC steps but contain errors in the data that have not been identified by either data providers or data users.

...

Some updates have been applied to the DCPP netCDF files in the CDS. These conform with the CF Metadata Conventions and improve the usability of the time dimension when multiple overlapping decadal experiments are used together. This will make the calculation of multi-model ensembles more straightforward. Conventions and improve the usability of the time dimension when multiple overlapping decadal experiments are used together with different start dates (adding additional time coordinates facilitates to use multiple datasets in parallel and enables unambiguous selection of time). The specific details of the updates include the following modifications:

• A “realization” variable is added, to represent the ensemble member
• The “sub_experiment_id” global attribute is adjusted to include the start year and month of the simulation
• A “reftime” variable is added, representing the start time of the simulation
• A “leadtime” coordinate variable is added, which is the prediction range of the forecasts: this is calculated from the “reftime” plus and the valid times from the existing time variable

...

• The "long_name" attribute of the "time" coordinate is updated to "valid_time". 

Citation information

The CMIP6 data Citation Service provides information for data users on how to cite CMIP6 DCPP data and on the data license. The long-term availability and long-term accessibility are granted by the use of DOIs for the landing page e.g . http://doi.org/10.22033/ESGF/CMIP6.1317. Available CMIP6 data citations are discoverable in the ESGF or in the Citation Search at: http://bit.ly/CMIP6_Citation_Search (search for DCPP at the top of the page).

Known issues

CDS users will be directed to the CMIP6 ES-DOC Errata Service ES-DOC Errata Service (see dcppA-hindcast and dcppB-forecast for experiment ID) for known issues with the wider CMIP6 data pool. Data that is provided to the CDS should not contain any errors or be listed in the Errata service, however this will still be a useful resource for CDS users as data they may be looking for but cannot access may have been withheld from the CDS for justifiable reasons.have been withheld from the CDS for justifiable reasons.

Particularly, the "daily maximum near-surface air temperature" variable is missing for the HadGEM3-GC31-MM model due to the fact that in this model "grid point single time step spikes leading to excessively large daily maximum temperature value" were found.  Details of the problem can be found at https://errata.es-doc.org/static/view.html?uid=76b3f818-d65f-c76b-bfd8-cae5bc27825c

Subsetting and downloading data

CDS users will now be are able to apply subsetting operations to CMIP6 decadal datasets. This mechanism (the "roocs" WPS framework)  that runs at each of the partner sites: CEDA, IPSL and DKRZ and IPSL. The WPS can receive requests for processing based on dataset identifiers, a temporal range, a bounding box and a range of vertical levels. Each request is converted to a job that is run asynchronously on the processing servers at the partner sites. NetCDF files are generated and the response contains download links to each of the files. Users of the CDS will be able to make subsetting selections using the web forms provided by the CDS catalogue web-interface. More advanced users will be able to define their own API requests in the CDS Toolbox that will call the WPS. Output files will be automatically retrieved so that users can access them directly within the CDS.

...

Eyring, V. et al. (2016) ‘Overview of the Coupled Model Intercomparison Project Phase 6 (CMIP6) experimental design and organization’, Geoscientific Model Development, 9(5), pp. 1937–1958. doi: 10.5194/gmd-9-1937-2016.

Climate Change 2021: The Physical Science Basis, the Working Group I contribution to the Sixth Assessment Report. Available at: https://www.ipcc.ch/report/sixth-assessment-report-working-group-i/ (Accessed: 14 September 2021)

World Climate Research Programme (2020) CMIP Phase 6 (CMIP6): Overview CMIP6 Experimental Design and Organization. Available at: https://www.wcrp-climate.org/wgcm-cmip/wgcm-cmip6 (Accessed: 2 November 2020).

Related articles

...