There was previously an error affecting all of the bias-corrected data (all mbcn-gridMET and mbcn-Daymet files) in the NA-CORDEX archive. A description of the error can be found below.
As of 2021-11-19, this error has been fixed. We re-ran the bias-correction for all data published through the NCAR Climate Data Gateway and verified that the error was no longer present.
If you downloaded bias-corrected data before 2021-11-19, you should replace it with the new version. Our apologies for the inconvenience.
We have identified a problem in all of the bias-corrected data (all mbcn-gridMET and mbcn-Daymet files) in the NA-CORDEX archive.
There is an offset of approximately 1 month in the data values relative to the time coordinate. Essentially, January model data was bias-corrected to match February observations, February was corrected to look like March, and so on.
We do not yet know the cause of this problem. We are working to
find it, and will then redo the bias-correction and republish the
data.
The error occurred because the program used to perform the bias-correction combines code from two different sources that use very different approaches to handling the time dimension in the data. There was a subtle bug in bringing them together that created a misalignment of the observational data and the model data. We rewote the code in question to fix the bug.
This problem went unnoticed for a time because it only shows up clearly in the monthly distribution of values averaged over a medium-sized region. At shorter time scales and over smaller regions, the problem is hidden by variability, and at longer timescales and over larger regions, it is hidden by the effects of the seasonal cycle and topography, respectively.
For more details, contact Seth McGinnis (mcginnis@ucar.edu).
Our apologies for any inconvenience that has resulted.