ISIMIP2a-extended protocol released

Posted by Martin Park on Feb. 28, 2020

Dear ISIMIP community,

with this email we announce the publication of the extended ISIMIP2a simulation protocol! (If the link does not work for you, please navigate to > Protocol.)

This protocol is very similar to the ISIMIP3a protocol circulated on 21-2-2020, and we suggest to prioritize simulations for ISIMIP3a and 3b over ISIMIP2a-extended. However, we still encourage all impact modeling teams to consider performing simulations according to this ISIMIP2a-extended protocol – even if you have already contributed simulations to the original ISIMIP2a round. These simulations will enhance the basis for evaluating the models that were used for the future projection runs in ISIMIP2b, which have been widely used and will continue to be of great value.

What’s the difference between 2a and 2a-extended?

Compared to the original ISIMIP2a simulation protocol, the forcing data have been extended to cover more recent years up to 2016; new atmospheric forcing datasets based on ERA5 and EWEMBI have been added; further sectors have been included; and many conventions and definitions have been updated and harmonized[1]. Simulation data can - and should, if feasible - be submitted for the entire simulation period; there is no more specific "reporting period" as in the original ISIMIP2a. Experiment definitions have been standardized across sectors; e.g. the definition of "nosoc" has slightly changed compared to the original ISIMIP2a; see section 4.7 in the protocol.

Why should I run ISIMIP2a-extended simulations?

- If you have performed ISIMIP2b simulations but no ISIMIP2a simulations yet, please use the same model version used for ISIMIP2b, and run ISIMIP2a-extended simulations. These can then serve to evaluate the historical performance of that model version.

- The same applies if you have already contributed ISIMIP2a simulations but with a different model version than used in ISIMIP2b.

- If you have performed ISIMIP2b and ISIMIP2a simulations using the same model version, you may still be interested in extending the latter, by

--- adding simulations for the years between 2010 and 2016 should enable better comparison to observational data, for which quality and/or availability tends to increase over time.
--- adding simulations with some of the new or improved forcing datasets; e.g. GSWP3-W5E5, an atmospheric forcing product based on the new ERA5 reanalysis, bias-corrected and extended backwards to 1901.

Please note:
- existing ISIMIP2a simulations will remain consistent with this extended protocol; new simulations will be added to the existing ISIMIP2a data repository.
- if you run simulations with a different model version compared to your older ISIMIP2a simulations, then use a different model name (e.g. add a version tag). See section 5 in the protocol. Most likely you WILL need a new name.
- Preliminary versions of GSWP3-W5E5 and GSWP3-EWEMBI datasets have temporarily been available on the server. If you downloaded any of those, please make sure to update with the datasets that are on the server now. [2]

We hope that all of this will be useful and support many interesting analyses. Please do not hesitate to get in touch with us if you have any questions.

We'd like to express our thanks to all the sectoral coordinators and the data developers who supported this important update.

Best regards from Potsdam,

the ISIMIP team

[1] The extension of ISIMIP2a follows the decision taken at the 2018 Strategy Group meeting, see
[2] These and other changes in the input data are also documented in the Input Data changelog on the ISIMIP website: