ISIMIP3a: Discontinuities found in climate input data

Posted by Iliusi Vega del Valle on March 9, 2021

Unfortunately, the bias-adjustment problem identified by Jerry Nelson does not only affect ISIMIP3b but also the observational climate forcings provided in ISIMIP3a. There the bias adjustment method is applied to adjust GSWP3 ensuring a smooth transition to W5E5 in the combined forcing dataset GWSP3-W5E5. For the period 1901-1978, this adjustment introduces similar jumps in the seasonal cycle as found in the bias-adjusted GCM data, see the multi-year daily mean value plots for all Giorgi regions.

In the course of this analysis, we realized that the same problem exists in the original GSWP3 data. As this dataset is provided externally, we cannot fix it ourselves. Hence we are currently exploring different options of how to backward-extend the W5E5 dataset, for the case that the discontinuities in GSWP3 cannot be removed.

The bottom line of all this is that we will update not only the ISIMIP3b bias-adjusted climate input data but also the ISIMIP3a observational and counterfactual climate input datasets (both GSWP3-W5E5 and GSWP3). We are not taking this decision lightheartedly, and we will try to gauge the impact of the discontinuities on downstream impact variables, based on already submitted ISIMIP3a simulations, to help decide whether those simulations need to be repeated.

We are really sorry for this. The good news is that this gives us the chance to improve our ISIMIP3 climate forcings on several levels. Next to fixing the discontinuity issue we intend to switch to the latest version of GSWP3 (from v0.5b to v1.09) and update W5E5 (from v1.0 to v2.0, based on WFDE5 v2.0). The latter update would involve:

  • the inclusion of the most recent years 2017-2019
  • the correction of a shortwave radiation processing bug
  • the correction of an ERA5 spatial interpolation bug
  • a better diurnal temperature range correction

We are in contact with Copernicus to include W5E5 v2.0. The update may hopefully take about one month but we currently do not know yet. We will keep you posted about this.

So at the moment, please do not start any new ISIMIP3a simulations. We will discuss with the sector coordinators how to handle already completed runs. If you have already done ISIMIP3a/3b simulations, please feel free to get in touch with them or us directly to look into the issue together.