...

Acronyms

...

In this section we will briefly mention the main features of the system. In general, CERRA consists of three components:  CERRA   (i) CERRA and CERRA-EDA, three-dimensional reanalysis systems, and (ii) CERRA-Land, a two-dimensional surface reanalysis system.

...

The CERRA system is based on the HARMONIE-ALADIN data assimilation system which is has been developed and used within the HIRLAM and ALADIN consortia. The CERRA system ACCORD consortium. It is implemented and optimized for the entire European area with surrounding sea areas (see Fig. 2) with a horizontal resolution of 5.5 km and 106 vertical levels. The system uses lateral boundaries conditions obtained from the ERA5 global reanalysis (Fig. 5). Also, the large scales in the regional system are constrained by data from the global reanalysis. The increase of resolution from the global reanalysis (RA) as well as the precursor regional reanalysis (RRA) is depicted in Figure 6.

...

Figure 5: Three different stages of RA: the global reanalyses from ERA5 are used as lateral boundary conditions for CERRA and CERRA-EDA reanalyses. Subsequently, short-forecasts forecast data from CERRA are used as background fields for the CERRA-Land surface reanalysis. As indicated by the vertical arrows, the amount of assimilated observations per area unit increases, in principle, from the global to the regional reanalysis as indicated by the arrows.

The CERRA system employs the 3D variational analysis (3D-VAR) method depicted schematically in Figure 4. At fixed points in time the model state is adjusted based on the observed state, taking into account the error statistics of both model and observations. The CERRA high-resolution system is has been running with eight assimilation cycles per day performing analyses at 00 UTC, 03 UTC, 06 UTC, 09 UTC, 12 UTC, 15 UTC, 18 UTC and 21 UTC. The forecasts lengths vary between 6 and 30 hours (see section 24.1.3 for more information) depending on the starting hour.
A flow-dependent background error covariance matrix (B) matrix is estimated using a 10-member ensemble of data assimilation (EDA) system, which is briefly described in the next section.
 

...

The main purpose of the CERRA-EDA system is to create an 'online', continuously serially updated flow-dependent B background error covariance matrix for use in the CERRA high-resolution CERRA system. CERRA-EDA is set up for the same geographical region as CERRA but at high-resolution, has the same number of vertical levels,  but a horizontal resolution of 11 km. The EDA system utilises lateral boundary conditions from ERA5 and the same type of observations, including satellite observations , as CERRA . high-resolution. CERRA-EDA is a ten member EDA system, where nine members differ through the perturbation of observations. But, one of the members is running with unperturbed observations forming the so called control member.

...

More information on the CERRA-EDA system and especially regarding the construction of the B background error covariance matrix can be found in El-Said et al. (2021).

...

The CERRA-Land analysis system uses the 2D-analysis system MESCAN and the land surface platform SURFEX to generate a coherent surface and soil analysis. The system combines CERRA forecast fields and additional surface observation (e.g. precipitation), to generate high-resolution (5.5 km) 2-dimensional analyses over Europe. MESCAN is a surface analysis system which uses an optimal interpolation algorithm for the analysis of 2-m 2m temperature and relative humidity and 24-h total accumulated precipitation (Soci et al., 2016). SURFEX is a land surface platform, which is driven by temperature, humidity, precipitation, wind and radiative fluxes (Bazile et al., 2017).

...

As depicted in Figure 3, all parameters are computed in grid points. Having a horizontal grid spacing of 5.5 km, as for the CERRA system, implies that each grid point characterizes values for an area of roughly 30 km² (5.5km*5.5km). This needs to be considered when for instance CERRA data are compared with observations. The CERRA-EDA system has a horizontal resolution of 11km, meaning that each grid point characterizes values for an area of 121 km² (11km*11km).

Vertical resolution

As already mentioned in Section 13.3.1, the CERRA system has 106 levels in the vertical levels (also called model levels) from the surface up to 1 hPa. However, only a very restricted number of parameters is stored on model levels. The main reason for that is the amount of needed storage space, when all parameters would be stored for all levels. Moreover, the vertical model grid is on hybrid-sigma coordinates, which do not match with any standard pressure level. Hence, this makes it quiet complex to use. Information about the model levels is here.

Therefore, the major part of the data is post-processed and stored on 29 selected pressure levels between 1000 - 1 hPa with a higher number of levels at lower altitudes. In addition, some parameters are also stored on 11 height levels between 15-500m. One reason to provide atmospheric variables on height levels is for applications in the wind energy sector. The exact levels both for pressure and height levels are given in Section 5 in the corresponding tables.

The CERRA system also contains a soil model which has 3 layers in the vertical levels. The three levels layers represent approximately the surface, the soil at root depth and the deep soil. Due to the used force-restore scheme in the soil model it is not possible to relate the levels layers with a certain depth in metre. Users interested in soil parameters should consider to use the data from the CERRA-Land system.

The CERRA-Land soil model has 14 layers in the vertical levels, which range from the surface to a depth of 12m. The edges between different levels are at 0.01m, 0.04m, 0.1m, 0.2m, 0.4m, 0.6m, 0.8m, 1.0m, 1.5m, 2m, 3m, 5m, 8m , and 12m. Values for a certain level reflect the mean value over the level layer thickness.

Time resolution

In general, data are stored with hourly resolution for the CERRA dataset. However, for all time steps, users have different options to select from and this is not always an easy choice. The preferred selection might vary for different parameters and the application of the user, respectively.
Figure 7 gives an overview on available analysis and forecast times. First, there are the eight analyses at 00 UTC, 03 UTC, 06 UTC, 09 UTC, 12 UTC, 15 UTC, 18 UTC and 21 UTC highlighted in red. Analysis data at these hours are assumed to be of higher quality than the forecasts valid at the same hours as they are in general closer to observations. They are available only every third hour and not all parameters are available for the analyses (the available analysis parameters are listed in Section 35). The forecast model is then started from the analysis and the output is saved hourly for the first six hours as indicated in dark blue in Fig. 7. Whereas the forecasts initiated at 00 UTC and 12 UTC continue until the forecast hour 30, all forecasts initiated at other hours (e.g. 15 UTC) stop after 6 hours. However, note that the output frequency of the forecasts initiated at 00 and 12 UTC is three hourly after the first six forecast hours (see blue boxes in Fig. 7) are completed. Due to the forecast lengths, the forecasts are overlapping and for every hour of the day data might be chosen from forecasts initiated at different forecasts hours and eventually from the analysis, respectively. For instance, at 12 UTC, the users can choose between the analysis and four different forecasts.
 

Figure 7: This table illustrates how different forecasts overlap and which options users have at a certain hour of the day. The availability of data is illustrated for the example date 2009/12/10. The color colour coding reflects analysis (red) and forecasts (blue). Moreover, different shades of blue correspond to frequency of the saved forecasts – hourly in dark blue and 3 hourly forecasts in blue.

...

Similar to above, results in complex terrain, such as mountainous regions or coastal areas, are generally less reliable than results over a more homogeneous terrain. The models cannot can hardly represent the strong local gradients that sometimes are caused by the variable (e.g. temperature gradients) that are usually driven by the complex terrain.

Figure 8 illustrates this behaviorbehaviour. Here, we show locations in Sweden having the best (blue) and the worst (red) correlations between the UERRA 2m - temperature and observational sites. A total of 853 measurement sites have been investigated. The Figure shows 50 locations each of highest and lowest correlation. Clearly, correlations are the lowest in the Swedish mountains and along the (east) coast.
Users need to be aware that the reanalysis provides gridded data where each grid point value describes an entire grid box area. That is in contrast to observations, which are usually point measurements. In case users need information with a higher horizontal resolution than provided by the CERRA systems, further downscaling (statistically or dynamically) needs to be considered. For instance, the correlations indicated in Fig. 8 increased when a linear interpolation from the four closest grid points to the observational site was applied than purely taken the values from the closest grid point. 

...

Figure 8: A validation of the UERRA 2m - temperature data with Swedish observations. 50 places each with highest (blue) and lowest (red) correlation are shown out of 853 measurement sites included in the investigation.

...

The production of the CERRA data is delayed by 2-3 months with respect to real time. The delay is directly dependent on the availability of some reprocessed datasets (e.g. GNSS-RO). Note: at the time when the dataset was released, a contract for the near-real time update of CERRA dataset had not been concluded, and therefore the production was suspended; it will be resumed once the contract is signed. One should be aware that it takes several months to produce data that will bridge the gap between June 2021 (last available month) and the near-real time, and that only after filling the gap, the dataset will be updated with a delay of 2-3 months behind the real time. For instance, the release of data for January 2023 2024 can be expected in April 20232024.

• Can we use reanalysis data for local applications?

...

Parameters on single level

10m wind speed

The 10-metre (10m) wind speed is the wind speed valid for the grid area determined for a height of 10m above the surface. The parameter is given in m/s. It is computed from both the zonal (u) and the meridional (v) wind components by wind speed=u²+v². The 10m wind speed is available for the analysis and the forecast time steps. For the forecast, the value is instantaneous meaning that it is valid for the last time step of the integration at the issued time step.

...

The land-sea mask is a field that contains, for every grid, the proportion of land in the grid box. The parameter is dimensionless and the values are between 0 (sea) and 1 (land). The land-sea mask is constant in time and the field is available for every analysis.

Orography

The orography is the height of the terrain with respect to the model defined globe. Each grid point has one value representing the mean over the grid point domain. The orography is given as geopotential height in metre [m²/s²]. The orography is constant in time and the field is available for every analysis.

Surface roughness

The surface roughness describes the aerodynamic roughness length (over land). Each grid point has one value representing the mean over the grid point. The surface roughness is given in metre [m]. The effective surface roughness is depending on the orographic component (constant part), the snow depth, the evolution of the Leaf Area Index and the fraction of vegetation, which is different for each month. Surface roughness is available for the analysis and the forecast time steps. 

...

Parameters on height levels

Wind speed

Wind speed is the wind speed valid for the grid area determined for a certain height (15m-500m) above the surface. The parameter is given in m/s. It is computed from both the zonal (u) and the meridional (v) wind components by wind speed=u²+v². The wind speed is available for the analysis and the forecast time steps. The value is instantaneous meaning that it is valid for the last time step of the integration at the issued time step.

...

Parameters on pressure levels

Cloud cover

Cloud cover is the percentage of sky covert with clouds. It is valid for the grid at the corresponding height. The parameter is given in %. Total cloud cover is only available for the forecast time steps. The value is instantaneous meaning that it is valid for the last time step of the integration at the issued time step.

...

Table 4: Overview of parameters on model levels

Specific humidity

The specific humidity is the mass of water vapour per unit mass of air valid for the grid area at the corresponding model level. Only analyses are stored for parameters on model levels. The parameter is given in kg/kg.

...

Below are the essential parameters describing the grid and the used Lambert Conformal Conic projection. More information about the grid and coordinates can be found in the FAQ.
Number of points along x-axis: 1069
Number of points along y-axis: 1069
X-direction grid length: 5500 m
Y-direction grid length: 5500 m
Projection: Lambert Conformal Conic
Central meridian: 8
Standard parallel 1: 50
Standard parallel 2: 50
Latitude of origin: 50
Earth assumed spherical with radius: 6371229 m

CERRA-EDA

CERRA-EDA comprises the same domain as CERRA and has exactly the same set of parameters as the high-resolution CERRA dataset. It differs only in the horizontal resolution, which is 11km as well as in the number of available time steps. CERRA-EDA has four analyses per day, at 00, 06, 12 and 18 UTC. Starting from the analyses, forecasts are run for six hours. Forecast fields are saved with hourly resolution.

The CERRA-EDA grid description

Below are the essential parameters describing the CERRA-EDA grid and the used Lambert Conformal Conic projection. More information about the grid and coordinates can be found in the FAQ in section 2.3.
Number of points along x-axis: 565
Number of points along y-axis: 565
X-direction grid length: 11000 m
Y-direction grid length: 11000 m
Projection: Lambert Conformal Conic
Central meridian: 8
Standard parallel 1: 48
Standard parallel 2: 48
Latitude of origin: 48
Earth assumed spherical with radius: 6371229 m

Known issues

Wrong metadata for the 2m maximum and minimum temperature

The metadata for the forecast step range is incorrect in the GRIB2 files for the maximum 2m temperature since previous post-processing and the minimum 2m temperature since previous post-processing. The correct step range is given in the table below.

 Minor data assimilation issues (issues when some observations were not available)

...

For completeness, a list of the occurrences of missing observational data and affected periods are listed below. 

References

Missing forecast parameters

References

• Ridal, M., Bazile, E., Le Moigne, P., Randriamampianina, R., Schimanke, S., Andrae, U. et al. (2024) CERRA, the Copernicus European Regional Reanalysis system. Quarterly Journal of the Royal Meteorological Society, 1–27. Available from: https://doi.org/10.1002/qj.4764
• Bazile E, R. Bazile E, R. Abida, A. Verelle, P. Le Moigne and C. Szczypta (2017): MESCAN-SURFEX surface analysis, deliverable D2.8 of the UERRA project, http://www.uerra.eu/publications/deliverable-reports.html
• El-Said A., P. Brousseau, M. Ridal and R. Randriamampianina (2021): A new temporally flow-dependent EDA estimating background errors in the new Copernicus European Regional Re-Analysis (CERRA), Earth and Space Science Open Archive, pp. 28, doi 10.1002/essoar.10507207.1, https://doi.org/10.1002/essoar.10507207.1
• Niermann D. et al. (2017): Scientific report on assessment of regional analysis against independent data sets, deliverable D3.6 of the UERRA project, http://www.uerra.eu/publications/deliverable-reports.html
• Ridal M., S. Schimanke and S. Hopsch (2018): Documentation of the RRA system: UERRA (C3S deliverable D322_Lot1.1.1.2, Documenting the UERRA system)
• Soci C., E. Bazile, F. Besson and T. Landelius (2016). High-resolution precipitation re-analysis system for climatological purposes. Tellus A, Dynamic Meteorology and Oceanography, 68:1, DOI: 10.3402/tellusa.v68.29879
• Verver Gé (2017): User Guidance, deliverable D8.4 of the UERRA project, http://www.uerra.eu/publications/deliverable-reports.html

...