Hi,
We are using the 10-member ERA5 ensemble to quantify meteorological uncertainty in our NEE flux predictions. In the process, we compared the ERA5 ensemble mean with ERA5 over Europe and identified a bias in the computed VPD (derived from t2m, q, and sp). This bias appears to be largely driven by differences in specific humidity at 1000 hPa (q).
The aggregated annual mean time series over Europe is shown below:
The spatial distribution of the VPD differences reveals a persistent and coherent bias
And the time series of the linear regression slope suggests that this bias is more pronounced during winter:
In addition, we found that ssrd (surface shortwave downward radiation) slightly exceeds the ensemble uncertainty range after 2020.
I am therefore wondering whether this is a known issue in ERA5/ERA5-ensemble products, or whether I may be overlooking something.