Climate Bulletin - About the data and analysis

The following describes the data and analysis of the monthly Climate Bulletins on temperature, sea ice and hydrological variables. It applies to the bulletins covering April 2019 onwards, which mainly build on the ERA5 reanalysis.

From November 2022 onwards, following a major change in hardware at ECMWF, the software used for the climate bulletin data processing chain has been updated. Prior to this date, time-series were derived from ERA5 on its native reduced Gaussian grid. For better traceability and for consistency with the post-processed ERA5 dataset on the 0.25 regular grid which is used for the maps and made publicly available in the CDS, time series are now also calculated based on this dataset. This means that all adjustments made in that dataset are also reflected in the time-series. Further updates have also been implemented for the calculation of spatial averages. The overall impact of these changes on the time series used for the monthly Climate Bulletins is as follows:

• Temperature: a difference of the order of 10^-4 for global and 10^-3 for European temperatures. This affects the non-significant digits in the time-series, but does not affect statements of any previous or current bulletin. To mark this change, the graphs produced from November 2022 onward are given a different version number (v2.1 for maps, v1.1 for time series).
• Sea ice: time series are unaffected by this change, as they are produced by OSISAF on separate hardware.
• Hydrological variables: This change affects the non-significant digits in the time-series, but does not affect statements of any previous or current bulletin. To mark this change, the graphs produced from December 2022 onward are given a different version number (v2.1 for maps, v1.1 for time series).

From June 2021 onwards, the time series of Arctic and Antarctic sea ice extent used in the sea ice bulletin are based on the OSI SAF Sea Ice Index (instead of ERA5 for months up until May 2021). The OSI SAF product provides a more homogeneous climate data record back in time than ERA5 and is available with the same temporal latency.

The bulletins prior to April 2019 were based on the ERA-Interim reanalysis. The about-the-data sections for these bulletins, as well as comparisons between ERA-Interim and other datasets can be found here: temperature, sea ice and hydrological variables.

The main differences between ERA5 and ERA-Interim are outlined here.

Region definitions

European regions are defined as follows.

• Europe: 25W-40E, 34N-72N
• Southwest Europe: 25W-15E, 34N-45N
• Northwest Europe:25W-15E, 45N-72N
• Southeast Europe: 15E-40E, 34N-45N
• Northeast Europe: 15E-40E, 45N-72N
• Central Europe: 2E - 24E, 45N - 55N
• European Arctic: 25W-60E, 66N-90N

Area averages

Area averages for temperature over the European regions are for land only, unless otherwise stated. All other area averages are for the domains where the averaged variable is defined. Land sea masks are defined according to the native grid of the used data set.

Seasonal definition

Seasons are defined according to the boreal definition.

For maps showing seasonal values for 2018: winter, December 2017 - February 2018 (DJF); spring, March 2018 - May 2018 (MAM); summer, June 2018 - August 2018 (JJA); autumn, September 2018 - November 2018 (SON). For time-series graphs, the same analogy is used for previous years.

Reference periods and anomalies

From January 2021 onward, the main reference period used is 1991-2020. What this means is described in a web article. For a transition period, the graphics and data relating to the previous reference period (1981-2010) can be found by using the tab-function above each graphic. For key indicators, the 1981-2010 value is included when hovering over the 1991-2020 value. The anomaly for a particular variable and month is the difference between the value of the variable for that month and the average value of the variable from 1991 to 2020 for that month of the year.

From the October 2021 temperature summary onward, a new approach has been adopted to relate recent global temperature to 1850-1900, a period taken to represent the pre-industrial level. The previous approach used the estimate provided by the IPCC Special Report on “Global Warming of 1.5°C”. Since the publication of this report, new and updated global temperature datasets have been published, which has resulted in a new estimate, as outlined in the IPCC AR6 WGI report “Climate Change 2021: The Physical Science Basis”. Based on this estimate a new approach for monitoring global temperature change since the 1850-1900 period is being used in the WMO statements on “The state of the global climate” from the Preliminary Statement for 2021 onwards (see details there, under “Datasets and methods – Global temperature data”). This approach results in a best estimate of 0.69°C with an uncertainty range (0.54 to 0.78 °C) to relate the standard WMO reference period 1981-2010 to 1850-1900. Extending this approach to the WMO reference period 1991-2010 gives a best estimate of 0.88°C with an uncertainty range (0.72-0.99 °C).

This method is also adopted for the C3S monthly climate bulletin, to align with global monitoring efforts. It should be noted that the above estimates are for multi-annual averages of global mean temperature and, as such, they provide a general framework that can also be used for annual averages. The approach adopted for daily and monthly anomalies is outlined in the following section.

The offset of 0.88°C used since 2021 to convert ERA5’s annual global temperature anomalies with respect to 1991-2020 to temperature differences from the 1850-1900 reference level is sufficient for assessing twelve-month averages, but for shorter periods (daily and monthly) additional assumptions or estimates are needed.

The Table below shows the increases in global-mean surface temperature from 1850-1900 to 1991-2020 for each month of the year from the Berkeley Earth, HadCRUT5 ensemble-mean and NOAAGlobalTemp datasets, using the latest versions available in early December 2023. These are updated versions of three of the four datasets used in Chapter 2 (section 2.11.1.3 and Table 2.3) of the Working Group I contribution to the Sixth Assessment Report (AR6) from the United Nations’ Intergovernmental Panel on Climate Change (IPCC) to estimate annual-mean change. For consistency with AR6, global averages are calculated by averaging northern and southern hemispheric means that in turn are calculated by averaging over all grid squares providing data values. The average of the values from the three datasets over all months of the year is 0.88°C, consistent with the value used to relate ERA5 annual averages to the 1850-1900 reference level (see ‘Reference periods and anomalies’ section above).

Notwithstanding their uncertainties, all three datasets indicate a similar annual variation in temperature change. Change is largest in March and smallest in July, which is normally the warmest month of the year globally. The dataset average has an annual range of about 0.2°C. Monthly averages of the actual offsets applied for ERA5 are also shown in the Table. The offsets have been determined objectively using Fourier fitting of a single harmonic, rounding coefficients to two decimal places. The temperature offset (in °C) for day n of a 365-day year is calculated using the following equation (for n = 1, 2, …, 365):

 

Its monthly values, included in the table, differ by no more than 0.03°C from the three-dataset averages. The spread among the three datasets is considerably larger: it reaches 0.19°C in February and is 0.14°C for the annual average. These offsets differ slightly, by up to 0.02°C, from those used in the preliminary calculations reported in a June 2023 article.

Application of the daily offset requires a daily climate for 1991-2020. This is determined from 30-day averages for dates from 1 January to 31 December, omitting 29 February. As the raw values are subject to noise due to the limited sampling, they are smoothed by Fourier filtering with a wavenumber eight truncation. This truncation gives a reasonably smooth annual evolution of the climate whilst ensuring that the difference between the monthly climates computed from monthly mean ERA5 data and from averaging the smoothed daily values is smaller in magnitude than 0.005°C for each month of the year.

For 29 February, the offset of 1850-1900 from 1991-2020 and the 1991-2020 climate are determined by averaging values for 28 February and 1 March.

ERA5

Data (monthly averages of original fields) | Documentation

ERA5 is a global atmospheric reanalysis from 1979 onwards. The resolution is hourly, but for this report the 'monthly means of daily means' are used. The native horizontal grid is ~31 km (reduced Gaussian grid N320), but data can be downloaded on a 0.25 deg regular lat/lon grid. ERA5 data are used for surface air temperature, precipitation, soil moisture and sea ice.

Until October 2022: The time series in the bulletin are calculated from the native grid, maps are produced from the 0.25 deg regular lat/lon grid, based on the post-processed ERA5 dataset in the Climate Data Store (CDS).
From November 2022: Time series and maps in the bulletin are based on the post-processed ERA5 dataset in the Climate Data Store (CDS) on the 0.25 deg regular lat/lon grid. 

Currently it is possible to download hourly and monthly averaged fields for ERA5. Gridded monthly anomalies and climatologies (including possible adjustments) used for the Climate Bulletin are also available for download from the Climate Data Store (CDS).

ERA5 surface air temperature

ERA5 surface air temperature is defined on all of the global domain and over all surfaces, all values are shown.

For time series the ERA surface air temperature is based on the original version of the data.

For maps and gridded fields ERA surface air temperature is based on an adjusted version of the data. In this version surface air temperatures have been adjusted for the period 1979-2013 to compensate for an inadvertent failure in production to utilize observationally-based analyses of the water temperatures of the Great Lakes. As such information was used in ECMWF’s earlier ERA-Interim reanalysis, the monthly average surface air temperatures over the Great Lakes from ERA5 are adjusted by the 1981-2010 average of the differences between the ERA-Interim (adjusted as below) and ERA5 temperatures for the month in question.  These adjusted analyses are used to compute the monthly climatological fields needed to define anomalies. The adjustment is applied only over the Great Lakes; elsewhere the monthly average temperatures and corresponding climatologies are derived entirely from ERA5 analyses.

The  summaries published for months prior to March 2019 were based on ERA-Interim data that were adjusted to compensate for two production issues:

1. Values over sea were taken from the background forecast model not the analysis, to avoid a detrimental effect of analysing biased air-temperature observations from ships.
2. Values over sea prior to 2002 were further adjusted by subtracting 0.1°C. This accounted for a change in bias that arose from changing the source of sea-surface temperature analysis.

Used for temperature, sea ice and hydrological summaries.

ERA5 precipitation

Values of precipitation come from a sequence of 12-hour background forecasts. These forecasts owe their skill to many types of data, of which those most directly related to precipitation are a set of all sky microwave radiances assimilated from 1987 onwards and composite radar rain rates from ground based radar assimilated from 2009. Observations from rain gauges are not assimilated, but gridded estimates from them, which may suffer from gaps in coverage and measurement biases, do provide independent datasets for evaluation.

Used for the hydrological summaries.

ERA5 relative humidity of surface air

Over land, values of the relative humidity of surface air are determined quite directly from observational records for regions where plentiful observations of surface air humidity were made. Elsewhere, the background forecast model plays a stronger role, enabling values of surface relative humidity to be derived less directly from other types of assimilated observation. As illustrated below, values are least reliable where surface observations are sparse and the background model forecasts of related variables such as precipitation are biased.

Used for the hydrological summaries.

ERA5 soil moisture

Soil-moisture values are for the uppermost 7 cm of soil, as modelled by the Hydrology Tiled ECMWF Scheme for Surface Exchanges over Land (H-TESSEL). HTESSEL uses a variety of soil texture classes with specific properties, as infiltration capacity and wilting point.

The time-series calculations for ERA5 soil moisture take into account fractional values in the land sea mask. For all ERA5 soil moisture calculations a mask is applied for regions designated to have permanent ice cover (including Antarctica and much of Greenland) or no vegetation, or that otherwise have a climatological mean annual precipitation rate of less than 0.3 mm/day. These regions appear in grey on the maps. In addition, for soil moisture maps, where some land-points are undefined due to the mismatch between dataset and map-coastline resolution, these land-points are reset using interpolation with the neighbouring sea values and subsequently superimposing the land sea mask.

Soil moisture from ERA5 is included in the European State of the Climate 2018 to provide a qualitative picture of the major anomalies and to show its consistency with the other variables.

Used for the hydrological summaries.

ERA5 sea ice

ERA5 does not analyse sea-ice observations directly. Instead, it incorporates analyses of fractional sea-ice cover (or concentration) produced elsewhere, with additional processing steps to ensure consistency across the whole time period. More details can be found in this report (section 3.6.2 Sea-surface boundary conditions).

Monthly-mean values of the ERA5 sea-ice cover are interpolated to a regular 1°x1° grid that is used for plotting maps. The ERA5 data are defined only for model grid points that are designated as sea or lake points. A much higher resolution is used to define the shorelines used in plotting the maps. Coastal values that are undefined on the 1°x1° grid are reset using immediately neighbouring values of ice cover. Other undefined values are set to zero ice cover. The maps of average sea-ice cover for a particular month include a climatological ice edge for that month. This ice edge is defined by the location of the 15% value in the average cover for 1981-2010.

Time series for sea ice are given as monthly values for either sea ice area or sea ice extent for the Arctic and Antarctic, which are calculated from the original ERA5 sea ice cover variable as follows:

• For both sea ice area and sea ice extent a cutoff is applied, with sea ice concentrations being set to 0 for all values smaller than 15%.
• Lake ice is not taken into account in the calculations, but also set to 0.
• Monthly sea ice area is derived from the monthly mean of sea ice cover.
• Daily sea ice extent is derived from the daily sea ice cover, by setting all values greater than or equal to 15% to 1.
• Monthly sea ice extent is then calculated from the daily sea ice extent.

The "Arctic" and "Antarctic" sea-ice area and extent shown in the graphs are the sums of the resulting fractional ice cover of each grid box multiplied by the area of the grid box, taken over all grid boxes north of 20°N and south 20°S.
 
Used for the sea ice summaries (until May 2021 for time series of sea ice extent and area).

 

EUMETSAT OSI SAF Sea Ice Index

Since June 2021, the time series of sea ice extent anomalies for the Arctic and Antarctic regions are based on the Sea Ice Index (OSI-420) produced by EUMETSAT Ocean and Sea Ice Satellite Application Facility (OSI SAF). This product provides time series of sea ice extent and sea ice area at daily and monthly frequencies. It is derived from daily gridded sea ice concentration data also produced by EUMETSAT OSI SAF with R&D input from the ESA Climate Change Initiative project. For the C3S Climate Bulletin, additional processing is done by C3S to derive the climatologies for the 1981-2010 and 1991-2020 reference periods, as well as monthly anomalies with respect to these two periods.

Two versions of the product have been used so far for the C3S sea ice summaries:

• OSI SAF Sea Ice Index v2.1 used from June 2021 to June 2023: This version of the product is derived from the daily gridded OSI SAF Sea Ice Concentration climate data record (CDR) and interim CDR v2.0 products (OSI-450 and OSI-430-b) and covers the period from 1979 onwards.
• OSI SAF Sea Ice Index v2.2 used from July 2023 onward: This version of the product is derived from the daily gridded OSI SAF Sea Ice Concentration climate data record (CDR) and interim CDR v3.0 products (OSI-450-a and OSI-430-a) and covers the period from October 1978 onwards.