...
In the latest versions of E-OBS (from v18.0e onward), uncertainty is estimated using stochastic simulation to produce an ensemble of realizations of each daily field. A set of spatially correlated GRFs are generated for each day that are conditional on the residuals from the deterministic spatial trend model described above. The spatial structure of the random fields is defined through the calculation of an empirical variogram. An example of the uncertainty estimate is shown in Figure 8 for a day in a wet spell for central Europe. It shows, as an illustration, the first four ensemble members and a plot of the station locations, the ensemble-mean precipitation and the uncertainty in this estimate in the 24-hour accumulated precipitation. Further details of the method, comparisons against the older version of E-OBS (using the previous gridding method) and comparisons against national datasets are documented by Cornes et al. (2018). Below is a section providing guidance on the use of an ensemble dataset.
The mean across the ensemble members is calculated and is provided as the "best-guess" ensemble mean field. The spread is calculated as the difference between the 5th and 95th percentiles over the ensemble to provide a measure of the 90% uncertainty range. The global radiation dataset has a 10-member ensemble, while the other elements have a 100-member ensemble. 
Figure 8: Example of the estimate of the 24-hour precipitation accumulation for 1 June 2013. This is one day of a wet spell that caused severe flooding in central Europe. Shown are the first four ensemble members of E-OBS for precipitation for this day (left) and the right panel shows the station locations (black dots), the ensemble-mean precipitation estimate (coloured shading) and isolines of the ensemble spread which relates to the uncertainty in the precipitation estimate. This example shows that the largest uncertainty coincides with the area with a relatively low stations density and a high precipitation amount.
The E-OBS precipitation dataset is daily, meaning the observations cover 24 hours per time step. The exact 24-hour period can be different per region and element. The reason for this is that some data providers measure between midnight to midnight while others might measure from morning to morning. Since E-OBS is an observational dataset, no attempts have been made to adjust time series for this 24-hour offset. It is made sure, where known, that the largest part of the measured 24-hour period corresponds to the day attached to the time step in E-OBS (and ECA&D).
...