Last modified on Apr 22, 2024 10:09

Issued by: CSIC

Date: 15/12/2023

Service contract: 2022/C3S2_381_Lot1_Contractor/SC1

1. Introduction 

This document describes the "Copernicus Interactive Climate Atlas gridded monthly dataset" which constitutes an evolution of the IPCC Atlas gridded monthly dataset recently published in the CDS¹. This dataset has been entirely produced using a workflow ensuring total reproducibility within the CDS. The climate information used to construct the dataset is fully available in C3S CDS. The dataset has a corresponding visualisation tool, which is the Copernicus Interactive Climate Atlas, available at https://atlas.climate.copernicus.eu/atlas.

In addition to the description of the dataset, we are summarising the fit-for-purpose quality control applied to the datasets, documenting and curating the dataset based on the results from basic products (maps and climate stripes), and highlighting and fixing any issue encountered.

2. CDS catalogues, variables and harmonization

2.1. CDS catalogues and datasets used

The Copernicus Interactive Climate Atlas gridded monthly dataset integrates information from several climatic observations, reanalysis and projection datasets. Table 1 shows the datasets and the specific CDS catalogues used. Note that the CORDEX catalogue includes two different datasets of different horizontal spatial resolution²: 1) CORDEX-CORE includes the homogeneous subset of CORE³ simulations at 0.22° and 3) CORDEX-EUROPE covering Europe at 0.11° resolution. Note that in order to harmonize the grids used for the different datasets in the Atlas dataset, a set of common nested regular sub-grids of varying resolutions (2°, 1°, 0.5°, 0.25°, 0.125°) were used to (conservatively) interpolate some of the original datasets, particularly multi-model ensembles (note that some products, e.g. ERA5-Land and ERA5⁴ are maintained in their original regular grids provided in the CDS to preserve their nominal resolution). The reference grids used for the different products included in the datasets are available in the project’s GitLab repository⁵ (access can be granted upon ECMWF request) together with other auxiliary information.

Some of the datasets available in the CDS are distributed under several catalogues according to different temporal resolutions (hourly, daily, or monthly) or subsets of variables (e.g. single-level or multi-level). To produce the dataset, we used the most convenient catalogue depending on the variable or index (see Table 2) following the main principle that groups of main variables and derived indices (e.g. maximum temperature and number of days with maximum temperature over 35°C) should be based on the same catalogue. For instance, maximum temperature (TX) is required at daily temporal resolution to compute derived indices; therefore, for ERA5, the catalogue with hourly data (Reanalysis-era5-single-levels) was  selected for this variable (and others), whereas the catalogue with monthly data (Reanalysis-era5-land-monthly-means) was selected for other variables, such as RLDS (surface_downwelling_longwave_radiation), which do not require daily data. Color codes in Table 1 indicate different temporal resolutions of the raw datasets available on the CDS catalogues (hourly: green, daily: blue and monthly: purple).

Table 1: Descriptions of the datasets and CDS catalogue entries used. Color codes indicate the different temporal resolutions (hourly: green, daily: blue and monthly: purple) of the datasets available in the CDS catalogues (one or several for each catalogue). The column “horizontal resolution” indicates the grids used for the different datasets (products regridded to the common regular sub-grids: 2°, 1°, 0.5°, 0.25°, 0.125°, are indicated with an asterisk).

Data harmonization across the different datasets and catalogues is crucial in this contract. For instance, ERA5 and ERA5-Land do not provide directly daily data, so this have to be computed aggregating the hourly values. However, both datasets follow different conventions for coding hourly data (ERA5 provides hourly accumulated values whereas ERA5-Land provides accumulated values since the beginning of the forecast, at 00 for a period of 24 hours) and require different post-processing to obtain daily values https://confluence.ecmwf.int/pages/viewpage.action?pageId=197702790 . This needs to be taken into account when harmonizing the different datasets in order to obtain a standard common definition and units for each of the variables (Sec. 2.3).

2.2. Variables and indices

 Table 2 shows the set of 30 variables/indices included in the dataset. This dataset includes the 21 climate variables and indices included in the original IPCC Atlas gridded monthly dataset, as well as 9 additional variables (the new variables are indicated with "*" after the sequence number). Color codes are included for variables/indices characterizing different Climatic Impact Driver (CID) categories: heat and cold (red), wet and dry (blue), wind and radiation (black), snow and ice (violet), coastal and open ocean (blue), circulation (grey). Italics indicates the indices derived from other variables (see Table 4).

Table 2: Description of the 21 climate variables and indices included in the IPCC Interactive Atlas, as well as the additional 9 new ones for the first version of the C3S Interactive Atlas dataset (indicated with "*" after the sequence number). Color codes are included for variables/indices characterizing different Climatic Impact Driver (CID) categories: heat and cold (red), wet and dry (green), wind and radiation (black), snow and ice (violet), coastal and open ocean (blue), circulation (grey). Italics indicates the indices derived from other variables (Table 4).

Table 3 shows the variables from Table 2 which are calculated for the different datasets (see Table 1). Note that bias adjusted indices are not computed for reanalysis/observational datasets, but only for climate projections (CMIP6 and CORDEX-EUR-11 by now). In the case of CMIP5, the only variables/indices considered are those included in the IPCC Atlas gridded monthly dataset due to the complex harmonization required for some of the new variables, which makes comparison across datasets difficult and potentially misleading.

Table 3. Description of variables/indices calculated for the different datasets. Data not available is indicated by 'N/A' and data not considered is indicated by 'N/C'. (*) Indicates variables not considered due to lack of data (only available for a reduced subset).

2.3. Harmonization of variables across catalogues

Table 4 shows the CDS particular variables from the different catalogues displayed in Table 1 used to construct the variables/indices displayed in Table 2. Note that Table 4 describes only the raw climate variables and not the derived indices (denoted in italics in Table 2). These raw variables are defined from different variables, temporal resolutions and units available in the different CDS catalogues. This table also describes the harmonization/conversion work carried out to produce the dataset using a standard definition and units for the variables/indices. The first and second columns display the code and the standard name and units. The third column describes the CDS name of the variables and the harmonization work done for the different datasets to obtain the standard variables and units. The colors used correspond to the particular temporal aggregation of the specific catalogue used for the different datasets. N/C and N/A indicate "Not Considered" or "Not Available" variables. 

Besides the name and unit standardization described in Table 2, the harmonization pre-processing implements calendar conversion (standard calendar⁷) and standard coordinate system (using -180 to 180), as described in D381.1.3.1 (Section 2.3). Moreover, the variables and indices calculated from these datasets are interpolated to common grids to form the final dataset.

Table 4: Harmonization work across the different catalogues carried out for the variables of the dataset. Color codes indicate the temporal resolutions used for the different datasets available in the CDS catalogues, which correspond to the colors indicated in Table 1 (green for hourly, blue for daily and purple for monthly). Unit harmonization (when needed) is indicated with "==>".

2.4. Available ensembles for climate projection products

The climate projection datasets listed in Table 1 are based on multi-model ensembles including simulations from the historical and future scenarios (RCPs or SSPs). In order to compute climate change information, simulations from both historical and future scenarios are required. Therefore, in the Atlas dataset we consider only those models providing simulations for historical period and, at least, for one future scenario. Figures 1, 2 and 3 show the final ensembles computed for the different variables (in columns) for the CMIP6, CORDEX-EUR and CMIP5 datasets. 

Figure 1: Final ensemble used in the "Copernicus Interactive Climate Atlas gridded monthly dataset" with the different models (in rows) for the different variables (in columns) for the CMIP6 dataset. Colors indicate the different scenarios available in the CDS.

 

Figure 2: Final ensemble used in the "Copernicus Interactive Climate Atlas gridded monthly dataset" with the different models (in rows) for the different variables (in columns) for the CORDEX-EUR-11 dataset. Colors indicate the different scenarios available in the CDS.

Figure 3: Final ensemble used in the "Copernicus Interactive Climate Atlas gridded monthly dataset" with the different models (in rows) for the different variables (in columns) for the CMIP5 dataset. Colors indicate the different scenarios available in the CDS.

Note that the CORDEX CDS catalogue was already used in the IPCC Atlas gridded monthly dataset (not part of the present version) and that ensemble is described in the documentation of the CDS catalogue entry⁸

2.5. Mosaic approach for CORDEX-CORE

One of the main novelties of the dataset is related to the CORDEX-CORE project⁹ . CORDEX-CORE is a highly ambitious initiative aimed at providing regional climate projections with global coverage (see Figure 4) and high resolution (0.25°x0.25°) by combining of 2 RCMs nested to 6 GCMs selected to span the widest range of uncertainty (see Table 6). The dataset includes, for the first time, the CORDEX-CORE simulations spatially blended using the mosaic approach described in Diez-Sierra et al. 2022¹⁰ . This dataset addresses two main issues related to the CORDEX-CORE simulations: 1) avoiding domain selection in overlapping areas by choosing the domain that best fits each one of the IPCC AR6 reference regions and 2) avoiding the multiple native projections of the different domains by providing the data in a regular global mesh. Due to its global coverage and higher resolution, this dataset constitutes the main source of information to analyze climate change for megacities globally, so it is a strategic dataset for the Atlas.

The two RCMs used in CORDEX-CORE are REMO2015 and RegCM4; the three GCMs used are MOHC_HadGEM2-ES, MPI-M_MPI-ESM-LR, and NCC_NorESM1-M (sorted in high to low climate sensitivity); note that MIROC-MIROC5 and NOAA-GFDL-GFDL-ESM2M are used as backups for high and low sensitivity, respectively. The list of simulations is given in Table 5, which describes the matrix of precise model versions used in the different domains. Figure 5 shows the final ensembles computed for the different variables (in columns) for the CORDEX-CORE dataset.

Figure 4: Spatial coverage of the CORDEX-CORE simulations (displaying near-term temperature change for RCP8.5 relative to 1961-1990),

Table 5: CORDEX simulations for the different domains used to build the CORDEX-CORE dataset (note that EUR-11 simulations are upscaled to the global 0.22 grid). Empty cells indicate that the default versions for the two RCMs (RegCM4_7 and REMO2015) are used. Different versions and/or backup GCMs are indicated in the table.

Figure 5: Final ensemble used in the "Copernicus Interactive Climate Atlas gridded monthly dataset" with the different models (in rows) for the different variables (in columns) for the CORDEX-CORE dataset. Colors indicate the different scenarios available in the CDS (see Table 5 for the specific configuration of the global mosaic).

2.6. Software used for index calculation

Most of the variables and indices included in the dataset are fully defined in Table 2 and require the application of simple temporal aggregations or threshold computations; therefore, they have been implemented in Python as part of the Atlas workflow to reduce the required computational cost. The exceptions are detailed below.

The Standardised Precipitation Index (SPI-6) is a monthly index that compares accumulated precipitation for 6 months with the long-term distribution (reference period: 1971-2010) for the same location and accumulation period, as the number of standard deviations from the median. This index has been calculated using using Python xclim software. 

The Standardised Precipitation Evapotranspiration Index (SPEI-6) is a monthly index that compares accumulated precipitation minus potential evapotranspiration (PET, Thornthwaite definition) for 6 months with the long-term distribution (reference period: 1971-2010) for the same location and accumulation period, as the number of standard deviations from the median. This index has been calculated using using Python xclim software applying the Thornthwaite (TW) version for the calculation of PET (using tasmin and tasmax as input).

Cooling degree-days (CD) and Heating degree-days (HD) have been implemented in Python as part of the Atlas workflow, following the definition used in the IPCC AR6 Atlas (IPCC, 2021: Annex VI: Climatic Impact-driver and Extreme Indices [Gutiérrez J.M., R. Ranasinghe, A.C. Ruane, R. Vautard (eds.)]. Cambridge University Press, Cambridge, pp. 2205–2214, doi:10.1017/9781009157896.020).

2.7. Bias adjustment

In this first version of the dataset we build on recent work on an intercomparison of different bias adjustment methods¹¹ which shows that a simple linear scaling method is a good choice for threshold-based temperature indices; the adjustment of the mean performed by these methods produces overall good results, as compared with more complex methods such as the ISIMIP3 method used in the IPCC Interactive Atlas. As an example, Figure 6 shows the results for the variable tasmax bias adjusted (tasmaxba) using both the ISIMIP3 and linear scaling methods for a particular model. From top to bottom, the panels display the results for linear scaling, ISIMIP3b and the difference between both methods.

All calculations are done using the ibicus package¹² which implements a number of bias adjustment methods (including linear scaling and ISIMIP3 among others). In principle, climate indices including absolute thresholds are most sensitive to biases. This is more pronounced as the threshold values become more and more extreme. This is the case of the extreme maximum temperature indices as also reported by Iturbide et al. 2022 (On the need of bias adjustment for more plausible climate change projections of extreme heat, https://doi.org/10.1002/asl.1072). We use the simple linear scaling method for bias adjustment. In this version the two most extreme climate indices were bias adjusted for the CMIP6 and CORDEX-EUR-11 datasets: monthly count of days with maximum temperature above 35°C and monthly count of days with maximum temperature above 40°C. The proposal for future versions of the Atlas dataset (if computing resources permit) is expanding the results including an additional bias adjustment method (the quantile mapping ISIMIP3 method), allowing users to explore this source of uncertainty. Note that this is aligned with the recommendation of the IPCC AR6 Report¹³.

Figure 6: Illustrative comparison between two commonly used bias adjustment methods of different complexity: ISIMIP3b (upper left) and linear scaling (upper right) for an illustrative model (the ACCESS-CM2 model). The panel in the bottom shows the difference between them.

2.8. File format and archiving

Files have been generated using netcdf-c verion 4.4.1.1 and hdf5 version 1.10.1 libraries using NETCDF4 data model. The resulting files are NetCFD format and metadata is CF1.9¹⁴ compliant allowing for string type NetCDF variables (used to define some attributes in the files, such as members). The attribute convention for data discovery is ACDD-1.3¹⁵ compliant (including reference, geospatial, etc.).
Data is stored in different files for different experiments, scenarios and variables/indices (using the naming convention experiment scenario_index.nc{_}), including all members in the same file using the member attribute and the full time periods mentioned in Table 1. The original data files were compressed aggressively, which means that although the file size looks small, but the necessary memory to read the data from these files might be large.

2.9. Spatial subsetting

The CDS form (and API) offers the possibility to spatially subset the data to be downloaded. The users can select any latitude-longitude area. If the selected area is larger than the region covered by the dataset then the full dataset will be provided. If the selected area includes any data points of the data then those data points will be provided. Requests, which don't include any point in the region of the dataset will fail. 

The ERA5, ERA5-Land, ORAS5, CMIP5 and CMIP6 datasets are covering the entire Globe, while E-OBS and CORDEX-EUR-11 are for Europe (see Figure 7) and CORDEX-CORE is a mosaic covering mostly the continental parts of the World (see Figure 4).

Figure 7: Illustrations of the extent of the E-OBS (top) and CORDEX-EUR-11 (bottom) European domains

In the longer term, we intend to introduce temporal subsetting too easing the selection of the time period and allowing to choose shorter time periods. 

3. Fit-for-purpose quality control 

Some basic quality control procedures were implemented to check the consistency of the data produced for the different indices and variables across datasets. This section describes the tests implemented to analyze and fix issues in the dataset. 

3.1. Tests implemented

An exhaustive quality control of the different variables and indices has been conducted with the aim of identifying potential issues in the original data (units, coordinates, etc.) as well as during the workflow for index generation. This procedure allowed us to detect data issues and to detect and fix metadata issues in the original data, and to identify problems and bugs during data harmonization. The quality control has been carried out based on four different qualitative tests:

• Spatial maps: Spatial maps were used to identify spatial displacements with respect to a reference, such as for example the coastline (see Figure 8). This analysis helped to identify problems during interpolation and issues with the coordinates of the raw data. Only a few problems were detected in the original datasets (in particular the IPSL-CM5A2-INCA_r1i1p1f1 simulations presented strange native lon and lat coordinates for sst and siconc, so this simulation was not used in the dataset).

Figure 8: Spatial map for near surface air temperature (t) for Euro-CORDEX.

• Climatologies: Mosaics of spatial maps for the different ensemble members are used to characterise the ensemble dispersion (in particular, considering climatological means for the period 1950-1990). This analysis helped to identify problems within the spread of the ensembles of climate projections. In particular, the simulation EUR11_MPI-M_MPI-ESM-LR_r1i1p1_IPSL_WRF381P_v1 are an outlier of the ensemble (with physically inconsistent values) for most of the variables, so this simulation was not used in the dataset.
• Climate Stripe-plots: Stripe-plots are used to identify various types of issues, such as identifying simulations that significantly differ from the ensemble range or identifying periods with missing values. The variables are annually aggregated for the entire domain (globally for CMIP and regionally for CORDEX) to generate these plots. Figure 8 shows a stripe plot for near surface air temperature (t) for CMIP6 for the SSP5-8.5 scenario. The different models forming the ensemble (as detailed in Sec. 2.4) are included in rows, and the annual mean values (spatially aggregated over the full domain) are included in columns. This analysis helped to identify models with long missing periods, which were not used in the dataset. In particular, CNRM-CM6-1_r1i1p1f2 for variable sst and scenario ssp370, and IITM-ESM_r1i1p1f1 for variable sst all scenarios and all variables for scenario ssp370.

Note that in many cases, the ensemble shows large uncertainty, with a large spread with low- and high-end models. The qualitative inspection of the stripe plots was used to detect outliers (models with values which are significantly lower/higher than the ensemble spread, indicating wrong units, or similar problems). Although the spread was high in some cases, the outliers were physically consistent (with the exception of EUR11_MPI-M_MPI-ESM-LR_r1i1p1_IPSL_WRF381P_v1, already mentioned) and, therefore, they were not removed from the dataset.

Figure 9: Example of climate stripe-plot for CMIP6. Note that the figure shows that different GCMs forming the ensemble exhibit different temperature magnitudes (i.e. systematic biases if they were compared to an observational dataset).

• Regional mean climate box-plots: Box-plots are calculated for each IPCC AR6 region and for the entire domain, comparing the regional mean results of the "Gridded monthly climate projection dataset underpinning the IPCC AR6 Interactive Atlas" with the Copernicus Atlas dataset. This is done to check the consistency between the two datasets and also as a double check for the workflow followed in the calculation of the dataset (comparing the results with those of the IPCC workflow). Figure 10 shows a box-plot for near surface air temperature (t) for six illustrative IPCC AR6 regions (in the different panels). Each panel shows multiple lines, each connecting the mean value of IPCC (left) and Copernicus Atlas (right) datasets for a particular model; horizontal black lines indicate coincident results for a particular GCM for the IPCC and Copernicus datasets. Note that there are some IPCC (or Copernicus) models with no counterpart (individual points, not joined by lines). This analysis allowed to double check the sensitivity of the climate change results to the changing GCMs in the C3S and the AR6-Atlas datasets; overall, no major (larger than the ensemble interquartile range) differences were found between the regional climate change signals from both datasets.

Figure 10: Example of box-plots for CMIP6 near surface air temperature (t) for different IPCC AR6 regions (in different panels) for the first version of the Copernicus Atlas (left, in blue) and the IPCC version (right, in green). The gray dashed lines join results from the same model in both datasets; small red squares indicate the ensemble means and the red lines join the medians of both ensembles.

4. Versions of the document 

The following table details the changes (and dates, in chronological order) that this document has undergone.

Related articles

• Page:
  Consolidation of the IPCC Interactive Atlas Dataset in the CDS
• Page:
  Copernicus Interactive Climate Atlas: User Guide
• Page:
  Gridded data underpinning the Copernicus Interactive Climate Atlas: Description of the datasets and variables
• Page:
  Copernicus Interactive Climate Atlas (C3S Atlas)
• Page:
  CORDEX: Regional climate projections