'Ensemble mean' in CDS data

In the documentation of ERA5 (ERA5: data documentation), section "Date and time specification", it says that:

"In the CDS: analyses are provided rather than forecasts, unless the parameter is only available from the forecasts."

However, I understand that if in the CDS webpage (https://cds.climate.copernicus.eu/cdsapp#!/dataset/reanalysis-era5-single-levels?tab=form), product type "Ensemble mean" is selected, it is always forecasts that are provided for all variables selected. Would that be correct?

(I guess this is obvious, but just to be sure, the wording is a bit confusing to me).

Thanks, Ricardo.

Hi Ricardo,

for the ERA5 CDS datasets (including the ensemble mean), it is always the case that accumulated variables (such as total precipitation) are taken from the forecasts, while instantaneous variables (such as 2m temperature) can also come from the analysis.

Hope that helps,


Hello Michela,

Thanks for the help! In order to be sure, I downloaded those two variables (t2m and tp) from CDS, both the 'reanalysis' and 'ensemble_mean' product types. And it happens that their values at the same 'validity times' (0h, 3h, 6h, ...), are always different. So I guess that 'ensemble_mean' always comes from forecasts, whereas 'reanalysis' always comes from reanalysis.

By the way, the CDS webpage itself (this one: https://cds.climate.copernicus.eu/cdsapp#!/dataset/reanalysis-era5-single-levels?tab=form), deactivates some variables when 'ensemble_mean' is selected. For instance, 'Vertical integral of northward cloud liquid water flux' is only available when 'reanalysis' is selected, but not when 'ensemble_mean' is selected.

Best regards, Ricardo.

Hi Ricardo, 

The reanalyses are from reanalyses ERA5 but the ensemble mean is the mean of the 10 ensemble members which is different. The 10 ensemble members are composed of  a member control unperturbed and 9 perturbed members. Also between reanalysis and ensemble members (and so ensemble mean) resolution is different (time and spatial). 

Reanalyses : every hour, 31 km (TL639), HRES

Ensemble members : every 3hours, 63 km (TL319)

So when you select ensemble-mean (or mambers or spread) some of the time series or variable ar not available. 

Best regards