ISIMIP3a: climate input data update

Posted by Martin Park on Sept. 13, 2021

After a long summer break we would like to inform you about the latest developments regarding the ISIMIP3a climate input data. We had retracted all of that data in March due to discontinuities found in the GSWP3 dataset. We know that many of you are waiting for an associated data update.

Unfortunately, no removal of the discontinuities found in GSWP3 has materialized so far. Hence we are now working on our own observational climate input dataset for the 20th century to overcome the stillstand. We plan to build it based on the NOAA–CIRES–DOE Twentieth Century Reanalysis version 3 (20CRv3) using a methodology similar to that used to produce GSWP3-W5E5. This development will take a few months.

In order to no longer prevent you from doing ISIMIP3a simulations, model evaluation and impact attribution, we have now made available the latest GSWP3-W5E5 dataset despite its known caveats, alongside with the associated counterfactual and spin-up data. Those datasets are described in detail in Mengel et al. (2021). They differ from the GSWP3-W5E5 datasets that were taken offline in March as the new data are based on later versions of GSWP3 (v1.09 compared to v0.5b), W5E5 (v2.0 compared to v1.0), the harmonization method ISIMIP3BASD (v2.5.0 compared to v2.4.1) and the detrending method ATTRICI (v1.1 compared to v1.0). The 3a protocol is also up-to-date for you to start simulations.

The new 20CRv3-based ISIMIP3a climate input will be published as an additional dataset once it is ready. It will not replace but complement the existing GSWP3-W5E5 dataset. In principle, the methodology to be developed for 20CRv3 can also be applied to other 20th century reanalysis products, such as CERA-20C. We will explore this option in order to provide several different datasets. This shall facilitate the quantification of climate input data-related uncertainty in ISIMIP3a model evaluation and impact attribution studies.