Contributors:  A. Troccoli (ICS), M. Borga (ICS/UNIPD), M. Zaramella (ICS), G. Aldrigo (ICS), R. Bortolami (ICS), R. Ciceri (ICS) S. Cordeddu (ICS), L. Lusito (ICS), E. Restivo (ICS), S. Strada (ICS), C. Zanetti (ICS), Y-M. Saint-Drenan (ARMINES), R. Amaro e Silva (ARMINES), S. Parey (EDF), S. Claudel (EDF), H. Upton (EDF), K. Nielsen (WEMC)

History of modifications 

List of datasets covered by this document 

Acronyms and abbreviations

Introduction

This documentation describes the Global climate and energy indicators from 1950 to present derived from reanalysis CDS dataset. It is based on ERA5, a global atmospheric reanalysis produced by the European Centre for Medium-Range Weather Forecasts (ECMWF), and it covers the historical period from 1950 to near present. The dataset has been developed within the Copernicus Climate Change Service (C3S) Energy service to support climate and energy applications.

The dataset includes both gridded (NetCDF) and spatially aggregated (CSV) indicators, computed using reproducible methods and updated regularly as new ERA5 data become available. Climate indicators such as temperature, precipitation, wind speed, and solar radiation form the basis for estimating energy indicators related to wind power, solar photovoltaic, hydropower, and electricity demand.

This documentation follows the complete data processing workflow, from data retrieval and pre-processing to bias adjustment, indicator computation, and spatial and temporal aggregation. Additional technical details on the energy conversion models and climate data processing tools are available in dedicated external pages, which have been linked along this document in the relevant sections:

• Tools for Climate Data Processing: describes the wind profile scaling using the power law, the computation of exclusion areas, and the spatial and temporal aggregation procedures.
• Energy Conversion Models: explains the energy data used to calibrate the models, and provides detailed descriptions of the methodologies for wind power, solar photovoltaic power, hydropower, and electricity and energy demand estimation.
  
  

Overview of the Workflow

The generation of climate and energy indicators follows a structured workflow that transforms raw reanalysis data into processed, user-ready outputs. The processing chain includes the following main steps:

• Data retrieval and pre-processing: ERA5 reanalysis data are retrieved at hourly resolution and converted from GRIB to NetCDF format. Wind components are combined to derive wind speed at both 10 m and 100 m heights, while solar radiation is converted from accumulated to instantaneous values. Known issues in ERA5 wind datasets are also addressed. See Section 2.1 and Section 2.2.

• Bias adjustment and computation of the wind speed at 100 m: ERA5 wind speed data are bias-adjusted using long-term climatologies from the Global Wind Atlas (GWA2), separately for 10 m and 100 m heights. This step corrects systematic errors in ERA5, especially over complex terrain (see Section 2.3). Alpha coefficients for the wind power law are then computed from the bias-adjusted 10 m and 100 m wind speeds, and are used to derive an alternative estimate of wind speed at 100 m via power-law scaling (see Section 2.4).

• Computation of climate indicators: Key climate variables—temperature, precipitation, wind speed, and solar radiation—are processed into indicators at their native 0.25°x0.25° grid resolution. These indicators are then spatially aggregated to national and sub-national levels (ADMIN 0 and ADMIN 1) and temporally aggregated to daily, monthly, seasonal, and annual scales. See Section 2.5.

• Computation of energy indicators: Climate indicators, both in gridded and aggregated form, are used as input to estimate potential generation and demand, together with input energy data employed for calibration and validation (see Section 3.1). Separate models are used for wind power, solar photovoltaic power, hydropower, and electricity demand (see Section 3.2). These models vary by region and data availability. All the energy indicators are provided as aggregated files, while only some of them are also available in gridded form. For more details, please refer to Section 3.3.

A schematic representation of the overall workflow is shown in Figure 1.1.

Image Added

Figure 1.1: Workflow for the historical data stream, illustrating the processing chain from ERA5 climate data to gridded and aggregated climate and energy indicators.

Computation of the Climate Indicators

Data Retrieval and Pre-processing Steps

Anchor
Section2_1 Section2_1

The workflow for processing the historical data stream is outlined in Figure 1.1. 

ERA5 data are retrieved in GRIB format from the Climate Data Store (CDS) using the CDS Application Programming Interface (API). Following retrieval, the GRIB files are converted into NetCDF format, and the resulting files are saved as monthly datasets using 16-bit precision to reduce file size while maintaining adequate numerical accuracy.

A unit conversion is applied to thesurface solar radiation downwards variable: originally provided as hourly accumulated energy in joules per square metre (J m⁻²), it is converted to instantaneous power in watts per square metre (W m⁻²) by dividing by 3600 (i.e., the number of seconds in an hour). 

Total precipitation is used as provided by ERA5, representing hourly accumulated values and expressed in metres (m). Air temperature at 2 m height is likewise processed in its native unit, Kelvin (K).

Next, a spatial mask is applied to all climate variables to restrict the domain to the area defined by the Global Wind Atlas

version 3 (GWA3). This includes all land areas and extends 300 km offshore, excluding polar regions. In addition, the domain has been extended to include the grid points required to cover the ENTSO-E’s Pan-European Climate Database (PECD)

. This ensures consistency with other related datasets and explains the presence of a rectangular extension over Europe in the domain mask. The resulting spatial coverage is shown in Figure 2.1.

Following spatial masking, wind speed

(ws)

at 10 m and 100 m is derived from the corresponding zonal

(u)

and meridional

(v)

components using the standard formula

, applied at each grid point and time step:

{ws}_{10} = \sqrt{u_{10}^2 + v_{10}^2}, \; \; {ws}_{100} = \sqrt{u_{100}^2 + v_{100}^2}

The full set of climate variables retrieved and processed includes:

• 2 m air temperature (TA)

• Total precipitation (TP)

• Surface solar radiation downwards (GHI)

• 10 m wind speed (WS10)

• 100 m wind speed (WS100)

These variables serve as the input to the subsequent computation of climate and energy indicators.

Image Added

Figure 2.1: Spatial domain mask used for the climate and energy indicators dataset, based on the GWA3. 
The mask includes land areas and offshore regions up to 300 km, with an additional rectangular area covering the ENTSO-E PECD domain to ensure consistency with related datasets.

Addressing Reanalysis Known Issues

Anchor
Section2_2 Section2_2

Before proceeding with the computation of climate indicators, specific issues identified in the original ERA5 wind dataset at 10 m are addressed through targeted corrections. These issues are summarised below.

Wind speed drop at 10:00 UTC:
A known issue in the ERA5 dataset (see ERA5: data documentation#Knownissues, point 8) results in a systematic drop in wind speed at 10:00 UTC. To correct this artefact, the wind speed value at 10:00 UTC is recalculated using linear interpolation between the values at 09:00 and 11:00 UTC. This is equivalent to a temporal average and is implemented using the interp function (with method set to linear) from the Xarray Python library.

To validate the effectiveness of the correction, the results were evaluated in four control domains located in different regions of the world. These control boxes are shown in Figure 2.2. Figure 2.3 illustrates the impact of the correction on the diurnal wind speed cycle (based on a 2009–2018 subset) for each of these boxes, comparing the original and corrected datasets.

Unrealistic extreme wind speeds:
Another issue affecting the ERA5 10 m wind dataset is the occasional occurrence of unrealistically high wind speed values. These extreme values, sometimes reaching up to 300 m s⁻¹, occur a few times per year at specific locations and are not physically plausible.

To mitigate this, a threshold of 60 m s⁻¹ is applied to the computed wind speed. Any values exceeding this cut-off are discarded. This threshold is slightly higher than the 50 m s⁻¹ limit recommended by ECMWF for the individual wind components (u and v), allowing for a conservative yet practical filter.

Image AddedFigure 2.2: Geographical control boxes used for validating the correction of the 10:00 UTC drop. 

Image Added

Figure 2.3: Impact of the 10:00 UTC correction on the diurnal cycle of 10 m wind speed in the four control boxes. Blue lines show the original dataset; orange lines show the corrected dataset. 

Bias Adjustment of the Reanalysis Dataset

Anchor
Section2_3 Section2_3

Although ERA5 is widely recognised for its improved accuracy compared to previous reanalysis products, certain systematic biases remain, especially in complex terrains (Olauson, 2018; Urraca et al., 2018; Hersbach et al., 2020; Hassler and Lauer, 2021; Keller and Wahl, 2021). A comprehensive evaluation of ERA5's performance was conducted to assess the need for bias adjustment across several key variables: 2 m temperature, total precipitation, surface solar radiation downwards, and wind speed.

The findings indicate that ERA5 reproduces temperature, precipitation, and radiation fields with generally high reliability at the global scale and over Europe (Hersbach et al., 2020; Lavers et al., 2022). 

Minor warm and wet biases were identified, particularly in mountainous regions, but these were not considered significant enough to impact the energy indicator modelling used to obtain this dataset. Specifically, Urraca et al. (2018) showed that ERA5's solar radiation estimates were comparable in bias to satellite-based products over inland areas. For temperature, ERA5 aligns closely with E-OBS observations, with larger biases (1–3 °C) found in complex terrain such as the Alps (Kotlarski et al., 2014; Velikou et al., 2022). Regarding precipitation, ERA5 shows a wet bias compared to E-OBS as already found by Bandhauer et al. (2021) and Lavers et al. (2022). Overall, the mean bias is lower than 1 mm day^-1, with high spatial correlation (R² > 0.80).

In contrast, wind speed showed more significant limitations due to terrain effects and spatial smoothing at ERA5’s 0.25° native resolution, which fails to capture local wind variability critical for wind energy modelling (Murcia et al., 2022). Since wind speed directly influences the estimation of wind power indicators, bias adjustment was deemed necessary for both the 10 m and 100 m wind speed datasets. The bias adjustment was performed using the Delta methodology, with reference data from the Global Wind Atlas version 2 (GWA2, Davis et al., 2023).

Bias Adjustment for Wind Speed

 Wind Speed Reference Dataset

As anticipated, the bias adjustment of wind speed relies on the Global Wind Atlas (GWA), a publicly available high-resolution dataset developed using mesoscale and microscale modelling to account for terrain-induced wind variability (Lledó et al., 2019; Davis et al., 2023). Among other height levels, GWA provides mean wind speeds at 10 m and 100 m over land and shoreline areas (up to 20 km offshore) at 250 m resolution.

For this study, we used GWA version 2 (GWA2), which is based on ERA-Interim as the underlying reanalysis, subsequently downscaled with the WAsP microscale model. GWA2 is appropriate for correcting reanalysis products such as ERA5, which tend to under-represent wind speeds in regions with complex terrain (Murcia et al., 2022).

Although GWA3 exists and was used to define the broader modelling domain (up to 300 km offshore), GWA2 was selected for bias adjustment because it provides long-term mean wind speed fields needed for the Delta method over the reference period 2006–2018.

Due to the coastal limitation of GWA2, grid cells between 20 km and 300 km offshore (within the GWA3 domain) remain unadjusted (see Figure 2.4).

Image Added

Figure 2.4: Map highlighting the regions where wind speed at 10 m and 100 m has been bias-adjusted (GWA2 domain) and where no correction has been applied (within the extended GWA3 domain beyond 20 km offshore).

Bias Adjustment Methodology

The chosen method for bias adjustment is the Deltamethodology (Murcia et al., 2022), which is well-suited for use with datasets like GWA that only provide long-term mean wind speeds. The adjustment is applied by scaling ERA5 wind speeds based on the ratio of mean wind speeds in GWA2 and ERA5 for the reference period 2006–2018.

For each grid cell and height (10 m and 100 m), a scaling factor is computed as:

\Delta = \frac{(\text{Mean wind speed})_{\text{GWA}}}{(\text{Mean wind speed})_{\text{ERA5}}}

​This scaling factor is applied uniformly to the full ERA5 time series at both 10 m and 100 m heights. This method maintains ERA5’s spatial and temporal coherence while correcting for under- or over-estimations caused by coarse resolution and orographic smoothing.

Unlike more complex approaches such as quantile mapping or the Cumulative Distribution Function transform​ (CDF-t), which require full wind speed distributions, the Delta method is computationally efficient and appropriate for global-scale applications where only mean values (as in GWA) are available.

The resulting bias-adjusted datasets (WS10 and WS100) retain ERA5’s hourly resolution and are delivered as monthly NetCDF files from 1950 to near present, consistent with other datasets in the stream.

Wind Profile Scaling for High Heights

Anchor
Section2_4 Section2_4

 

In contrast to ERA5, many datasets, such as climate projections and seasonal forecasts, do not provide wind speed at 100 m height. In these cases, 100 m wind speed is reconstructed using a power-law formulation, which requires a location-dependent alpha coefficient (α). This coefficient represents the vertical wind shear between 10 m and 100 m and is computed for each grid point, hour of the day, and month of the year to capture both diurnal and seasonal variability. The resulting α fields are distributed as a dedicated NetCDF file on the CDS.

For the historical dataset, two versions of the 100 m wind speed are available:

• Bias-adjusted 100 m wind speed, obtained by directly adjusting the ERA5 100 m field as explained in the previous section.

• Power-law-based 100 m wind speed, derived by applying the α coefficients to the bias-adjusted ERA5 10 m wind speed.

Both versions are provided to users because they support different analytical requirements. The direct GWA-corrected WS100 preserves the native vertical structure of ERA5 where available, while the power-law-based WS100 ensures consistency with datasets that do not include a 100 m level, such as climate projections and seasonal forecasts.

These two alternatives allow users to choose the approach that best fits their application and facilitate cross-stream comparability when analysing historical, seasonal, and projected wind conditions.

Further details on the wind profile scaling methodology can be found on the page Power Law for Wind Profile Scaling.

Gridded and Aggregated Climate Indicators

Anchor
Section2_5 Section2_5

Spatial and Temporal Aggregation Procedures

Anchor
Section2_5_1 Section2_5_1

Some crucial steps of the processing workflow involve aggregating the gridded climate indicators derived from ERA5 into more accessible formats, both spatially and temporally. These processes are the spatial and temporal aggregations.

Spatial Aggregation

• Input: Gridded (NetCDF) indicators.

• Output: Aggregated indicators at ADM0 (ADMIN0, national level) and ADM1 (ADMIN1, first sub-national level) regions

  Footnote
  https://www.naturalearthdata.com

  .

The output is delivered in CSV format, where each column corresponds to a region and each row to a time step. Spatial aggregation is computed using a weighted average over land grid cells, taking into account land-sea masks and the fraction of each cell lying within the administrative boundaries. The temporal resolution of the output files is the same as the input.

Temporal Aggregation

• Input: Hourly gridded climate indicators in NetCDF format.

• Output: Daily, monthly, seasonal, and annual averages or totals, depending on the variable. Aggregated values are computed independently for each grid cell and stored in NetCDF format for the gridded version, and in CSV format for the spatially aggregated (ADM0/ADM1) version.

More detailed descriptions of both procedures can be found at the following pages:

• Spatial Aggregation Procedure

• Temporal Aggregation Procedure

Overview of Climate Indicators

Anchor
Climate_Indicators Climate_Indicators

The historical data stream provides a suite of climate indicators, delivered in both gridded NetCDF format and regionally aggregated CSV format. These indicators serve as the foundation for the derivation of energy indicators and support a broad range of climate-related analyses and applications.

All indicators are consistently derived from ERA5 data, covering the period from 1950 to near present. Table 2.1 below summarises the full set of climate indicators included in the dataset, specifying their units, source variables, aggregation and output formats.

Table 2.1: Gridded and aggregated climate indicators provided in the historical data stream.

*WS100 is provided in two versions, as explained in Section 2.4: bias-adjusted 100 m wind speed, obtained by directly correcting the ERA5 wind speed at 100 m, and power-law-based 100 m wind speed, derived by applying the α coefficients to the bias-adjusted ERA5 10 m wind speed.

Computation of the Energy Indicators

The energy indicators provided in this dataset are derived from the climate indicators described in the previous sections, using dedicated models that convert meteorological variables into estimates of electricity generation or demand. These models are designed to reflect the physical and operational characteristics of different energy technologies. The energy indicators include:

• Wind power energy indicators

• Solar photovoltaic energy indicator

• Hydropower energy indicators

• Electricity and energy demand indicators

In addition to climate inputs, the models require reference energy data—such as installed capacity, historical electricity generation, and system configurations—for calibration and validation. These reference datasets ensure that the resulting indicators are regionally representative and consistent with observed energy system behaviour.

The following sections summarise the input and output data associated with each energy conversion model implemented in the historical stream, along with the reference energy datasets used in their development. For full details on the methodologies and assumptions behind each model, please refer to the dedicated page: Energy Conversion Models.

Reference Energy Datasets

Anchor
Section3_1 Section3_1

To accurately simulate electricity generation and demand from climate indicators, the energy conversion models rely not only on meteorological inputs but also on high-quality reference energy datasets. These datasets include observed or reported information on:

• Installed capacity (e.g., wind, solar, hydroelectric)

• Actual electricity generation by technology and country

• Load profiles and total electricity demand

• Energy system characteristics (e.g., wind turbine characteristics, hydropower reservoir capacities, seasonal patterns)

The reference energy data serve two primary purposes:

1. Calibration – ensuring that model outputs reflect realistic levels of energy production and demand.

2. Validation – comparing simulated indicators against historical records to assess model performance and regional representativeness.

These datasets are collected from a mix of publicly available sources, commercial providers, international statistics, national energy agencies, and, in some cases, directly from Transmission System Operators (TSOs). When necessary, data gaps are addressed using proxy information or scaling techniques.

A full description of the datasets and sources is available on the dedicated page: Reference Energy Datasets.

Energy Indicators Modelling

Anchor
Section3_2 Section3_2

The conversion of climate indicators into energy indicators is performed using a set of energy conversion models, each designed to represent the physical and operational characteristics of a specific energy system. These models take climate variables—such as wind speed, solar radiation, precipitation, and temperature—as input and output energy-related variables like electricity generation or demand.

This section provides a summary of the input and output data used in each model as implemented in the historical stream. Full methodological details, including calibration procedures and underlying assumptions, are described in the dedicated external page: Energy Conversion Models.

Wind Power Energy Indicator

Anchor
Section3_2_1 Section3_2_1

Input data:

• Bias-adjusted ERA5 historical wind speed at 10 and 100  m (two approaches available for 100 m: GWA-corrected and power-law-derived)

• Wind turbine characteristics (e.g. power curve, hub height, rotor diameter)

• Wind farm location and installed capacity (from The Wind Power, Global Energy Monitor, and other sources)

Model overview:
The wind power model estimates capacity factor by applying turbine-specific power curves to wind speeds extrapolated to the turbine hub height, using either 10 m or 100 m wind speed as the reference, depending on the turbine configuration. For turbines with hub heights ≥ 100 m, both WS100 versions (directly bias-adjusted and power-law-based) are used to compute capacity factors.

Output data:

• Gridded wind power capacity factor at hourly resolution (NetCDF, ERA5 native grid)

• Spatially aggregated indicators at ADM0 and ADM1 levels (CSV format) with hourly, daily, monthly, seasonal, and annual resolution

• Both gridded and aggregated data are provided at different hub heights and turbine types.

Figure 3.1 below illustrates the mean and interannual variability of simulated offshore wind power capacity factor (Vestas V164/8000 turbine) over the 1991–2020 period, both annually and for the month of November.

See Wind Power Conversion Model for more details.

Image AddedImage Added

Image AddedImage Added

Figure 3.1: Mean and standard deviation of annual (upper panels) and November’s (bottom panels) WOF capacity factor (specifically: Vestas V164/8000 turbine) over 1991-2020.

Solar Photovoltaic Energy Indicator

Anchor
Section3_2_2 Section3_2_2

Input data:

• ERA5 surface solar radiation downwards (GHI)

• ERA5 2 m temperature (TA)

• Tilt and azimuth configuration (derived from private PV plant metadata and global rules)

Model overview:
This model estimates photovoltaic electricity production using a physics-based conversion model that considers solar irradiance and air temperature to estimate capacity factors which consider the installed DC capacity as the normalisation factor. It incorporates representative module tilt and azimuth for utility-scale fixed installations (tilt and azimuth), and system losses, and is applied spatially across the entire domain.

Output data:

• Gridded solar photovoltaic capacity factor at hourly resolution (NetCDF, ERA5 native grid)

• Spatially aggregated indicators (CSV format at ADM0 and ADM1 levels) at hourly,daily, monthly, seasonal, and annual temporal resolutions.

Figure 3.2 below shows the average and interannual variability of SPV capacity factor output across the domain for the period 1991–2020, with a focus on both annual values and those specific to November.

See Solar Photovoltaic Energy Conversion Model for more details.

Image AddedImage Added

Image AddedImage Added

Figure 3.2: Mean and standard deviation of annual (upper panels) and November’s (bottom panels) SPV capacity factor over 1991-2020.

Hydropower Energy Indicators

Anchor
Section3_2_3 Section3_2_3

Random Forest (RF) Regression Model (European Domain)

Input data:

• ERA5 precipitation and 2 m temperature, aggregated to weekly resolution and at ADM0 level

• Hydropower generation, installed capacity, and reservoir filling rates from the ENTSO-E Transparency Platform

Model overview:

This statistical machine learning model estimates hydropower generation in Europe using a Random Forest (RF) regression approach trained on observed high-resolution generation data. The model is developed separately for each country and is used to compute the following indicators:

• Generation from reservoirs (HRG)
• Generation from run-of-river and pondage (HRO)
• Inflows to reservoirs (HRI)

The model uses lagged weekly climate variables (precipitation and temperature) to account for delayed hydrological responses and is validated using historical records through a Leave-One-Year-Out cross-validation strategy. Only countries with at least two years of reliable data are included. Model performance is evaluated using metrics such as the Nash-Sutcliffe Efficiency (NSE).

Output data:

Time series of aggregated hydropower indicators (HRG, HRO, and HRI) as CSV files at ADM0 level and at weekly, monthly, seasonal and annual temporal resolution.

Figure 3.3 below illustrates the average and interannual variability of simulated HRO over the period 1991–2020, focusing on both annual patterns and seasonal characteristics in November.

See Hydropower Conversion Models, Section 1 "Random Forest Regression Model (European Domain)" for more details.

Image AddedImage Added

Image AddedImage Added

Figure 3.3: Mean and standard deviation of annual (upper panels) and November’s (bottom panels) HRO over 1991-2020 (European domain).

Installed Capacity Weighted Precipitation (IWP) Proxy (Global Domain)

Input data:

• ERA5 precipitation aggregated at monthly resolution and at subnational level (NUTS-2 for European countries and ADM1 for the rest of the globe)

• Hydropower plant locations and installed capacities from the Global Energy Monitor

Model overview:

A proxy approach that computes hydropower potential as a weighted sum of n-month cumulative precipitation over regions hosting hydropower plants. Weights are based on installed capacity.

Output data:

Aggregated IWP time series as CSV files at ADM0 level at monthly, seasonal and annual temporal resolution.

Figure 3.4 below shows the long-term mean and interannual variability of the IWP proxy, highlighting both annual patterns and seasonal characteristics for November over the 1991–2020 period.

See Hydropower Conversion Models, Section 2 "Installed Capacity Weighted Precipitation (IWP) Proxy (Global Domain)" for more details.

Image AddedImage Added

Image AddedImage Added

Figure 3.4: Mean and standard deviation of annual (upper panels) and November’s (bottom panels) IWP over 1991-2020.

Electricity and Energy Demand Indicators

Anchor
Section3_2_4 Section3_2_4

Electricity Demand Model (European Domain)

Input data:

• ERA5 climate variables aggregated at ADM0 level and at daily temporal resolution:

  • 2 m temperature (TA)

  • Surface solar radiation downwards (GHI)

  • Wind speed at 10 m (WS10)

• Electricity load data from ENTSO-E:

  • ENTSO-E Power Statistics: historical data from 2006 to 2014

  • ENTSO-E Transparency Platform: operational data from 2015 onward

Model overview:

A Generalised Additive Model (GAM) is used to estimate daily electricity demand across 34 European countries. The model captures the nonlinear relationship between climate variables (e.g., temperature, solar radiation, wind) and electricity load, accounting for seasonal cycles and day types. Input data are preprocessed (including detrending) to isolate climate-driven variability. Country-specific models are trained and validated using cleaned historical datasets.

Output data:

• Electricity demand time series (ADM0 level, CSV format) at daily, monthly, seasonal and annual temporal resolution

• Temporal coverage: 1950 - near present

Figure 3.5 below illustrates the spatial patterns and interannual variability of electricity demand, showing both annual and November means and standard deviations over the 1991–2020 period.

See Electricity and Energy Demand Models, Section 1 "Electricity Demand Model (European Domain)", for more details.

Image AddedImage Added

Image AddedImage Added

Figure 3.5: Mean and standard deviation of annual (upper panels) and November’s (bottom panels) Electricity Demand (EDM) over 1991-2020.

Energy Demand Proxy (Global Domain)

Input data:

• ERA5 2 m air temperature (daily resolution)

• Gridded population data (for country-level aggregation)

Model overview:

The model estimates energy demand using a proxy metric called Energy Degree Days (EDD), calculated as the sum of Heating Degree Days (HDD) and Cooling Degree Days (CDD). These are derived from daily mean temperatures following definitions aligned with IEA standards (HDD: T<15°C, reference 18°C; CDD: T>24°C, reference 21°C). HDD and CDD are first computed at the grid level using ERA5 data and then aggregated to country level using population-weighted averaging. The resulting EDD index serves as a proxy for energy demand and is especially valuable in data-scarce regions, though it is also applicable in data-rich countries.

Output data:

• HDD, CDD and EDD time series (ADM0 level, CSV format) at monthly, seasonal and annual temporal resolution

• Temporal coverage: 1950 - near present

Figure 3.6 below presents the spatial distribution and interannual variability of the EDD proxy, showing both annual and November means and standard deviations for the 1991–2020 period.

See Electricity and Energy Demand Models, Section 2 "Energy Demand Model (Global Domain)" for more details.

Image AddedImage Added

Image AddedImage Added

Figure 3.6: Mean and standard deviation of annual (upper panels) and November’s (bottom panels) EDD proxy over 1991-2020.

Gridded and Aggregated Energy Indicators

Anchor
Section3_3 Section3_3

Spatial and Temporal Aggregation Procedures

The spatial and temporal aggregation procedures applied to energy indicators follow the same principles used for climate indicators (see Section 2.5.1). Gridded data are aggregated over standard administrative regions, ADM0 (national) and ADM1 (sub-national), and resampled to multiple time scales, including daily, monthly, seasonal, and annual resolutions, depending on the nature of each indicator.

More detailed explanations are provided in the dedicated documentation pages:

• Spatial Aggregation Procedure

• Temporal Aggregation Procedure

Exclusion Areas

Some energy indicators, particularly those related to renewable generation, such as solar and wind, may be affected by geographic constraints that limit the suitability of certain locations for energy production. To address this, the Service provides a set of exclusion area layers that define geographic zones where energy production is restricted or unlikely, such as protected areas, urban zones, steep slopes, and polar regions.

While these exclusion layers were not used in the standard modelling of energy indicators in this dataset, they are made available to users to enhance post-processing, filtering, or custom analyses.

For a full description of available exclusion criteria, NetCDF files, and variable names, see the dedicated documentation page: Exclusion Areas Computation.

Overview of Energy Indicators

The energy indicators included in this dataset are derived from the Climate Indicators and the Energy Data previously described, and through the application of specific Energy Conversion Models. Table 3.1 summarises the complete list of energy indicators, detailing the associated units, temporal and spatial resolution, domain of application, and original data sources.

Table 3.1: Gridded and aggregated energy indicators in the historical data stream.

Known issues

There are no known issues.

Appendix

Anchor
Appendix Appendix

Filenames Convention and Characteristics

This paragraph aims to explain the filename convention used for the dataset described in this Product User Guide. Table 5.1 details the structure and possible fields of the filenames. Specifically, the last column indicates the corresponding section of the CDS catalogue where users can personalise their choice. If "Not applicable" is indicated, it means that the user cannot modify this field, and the data are downloaded with fixed characteristics that are not customizable. Table 5.2 details the structure and filenames of the ancillary NetCDF files (described in Table 5.3) that are available in the CDS under the widget "Weights and masks".

# General

## Title

### 2 metre temperature - C3S Energy Lot1 file version v1.00

## Abstract

### ERA5

## Date

### 2026-02-06

## Date type

### Publication: Date identifies when the data was issued

## Unit

### K

## URL

### https://cds.climate.copernicus.eu/

## Data format

### CSV

## keywords

### ERA5 : Reanalysis : Copernicus : C3S : C3S Energy : ICS

## Comment

### Produced by Inside Climate Service

## Point of contact

### ECMWF User Support, https://confluence.ecmwf.int/site/support

# Usage

## Access constraints

### Intellectual property rights: The IP of these data belongs to the EU Copernicus programme

## Use constraints

### CC-BY 4.0, https://creativecommons.org/licenses/by/4.0/

## Temporal extent

## Begin date

### 1950-01-01 00:00

## End date

### 1950-01-31 23:00

## Temporal resolution

### 01d

## Geographic bounding box

### westBoundLongitude 0.5

### eastBoundLongitude 359.5

### southBoundLatitude -89.5

### northBoundLatitude 89.5

## Spatial resolution

### ADMIN0

# Lineage Statement

## Original Data Source

## Statement

### The original data sources are ECMWF ERA5 Reanalysis (available at: https://cds.climate.copernicus.eu)

#