CORDEX CORE Simulation Framework

A simulation framework in support of IPCC AR6

November 2018

CORDEX Scientific Advisory Team

This document presents the simulation framework for the CORDEX-CORE program. As discussed at the ICRC-CORDEX 2016 conference and in subsequent meetings, the main goal of the CORDEX-CORE framework is to provide a core set of comprehensive and homogeneous projections across all, or most, CORDEX domains. The immediate intent is to provide output that can support IPCC AR6 assessments. For this reason, the core framework is intended to be ambitious but sufficiently cost-effective to attract participation and produce timely output for the AR6. This program also stems from the results of a questionnaire send out to the CORDEX community through the CORDEX POCs in the fall of 2016, with feedback received in late 2016.

CORE has a specific focus on the IPCC AR6. It is not intended by itself to be the next phase of CORDEX. However, CORE simulations can provide a foundation in each region for further downscaling activities.

The fundamental goal underlying the CORDEX-CORE is to have a core set of RCMs downscaling a core set of GCM projections over all, or most, CORDEX domains. A further goal is to have a succinct set of simulations that has strong likelihood of being completed in a timely manner, so as to be available for IPCC assessments. The framework presented here to achieve these goals consists of the following components.

  1. Downscaling procedures

Ideally there should be a core set of RCMs downscaling GCMs over most CORDEX domains. Based on available personnel and computational resources, to date GERICS (with REMO) and ICTP (with RegCM) have committed to this while some other sets of RCM communities, such as CLM, have expressed interest in participating to this core group to the extent possible.

Other models/groups are also encouraged to complete simulations over specific domains of their interest, which could supplement the CORE experiments. We also welcome participation by groups performing Empirical Statistical Downscaling (ESD).

  1. Domains

The simulations should follow the size specifications for the CORDEX standard domains, giving particular attention to the specified minimum domain sizes.

The CORE set of domains includes:

  • Africa
  • South Asia
  • Central America
  • Southeast Asia
  • South America
  • East Asia
  • Europe
  • North America
  • Australia
  • Central Asia

Because the downscaling makes large demands on computing resources, some strategic planning is necessary to ensure early availability of multi-downscaling output for at least some of the domains. Therefore, the domains are listed in priority order. Africa is first, following the original CORDEX intent to give special attention to a continent that is especially vulnerable to climate change. The rest of the list gives priority to high population regions, especially those that may have relatively large vulnerability to climate change and lack local infrastructure to complete ensembles of projections.

It is noted that other CORDEX domains exist, such as the polar and MENA domains, for which, the production of simulations is left to the regional CORDEX activities. In particular, it is noted that some models are not set up for the polar domains, which undermines a goal of applying a common set of models to all of the targeted domains. The MENA domain overlaps several of the other CORDEX domains. While the MENA domain has some desirable properties, the overlap with several other domains undermines a goal stated above of having a succinct set of simulations.

  1. Grid spacing

In order to provide an advancement compared to previous CORDEX runs and to resolve additional features of regional climate, the grid spacing of the new CORE runs ideally should be of the order of a few tens km. However, a flexible choice of grid spacing between different domains and even within the same domain is allowed, to accommodate needs of different communities and groups.

The primary target is grid spacing in the range of 12.5 -25 km. For some domains, such as Europe, there are already plans to run at ~12.5 km grid spacing. For other domains, especially the larger ones, this grid spacing appears not feasible in light of our goals. Coarser resolutions are acceptable, but within certain limitations

  • To take advantage of completed simulations and recognizing that computing-resource constraints may limit new simulations for some regions or groups, existing simulations at 50 km grid spacing will be included in the archive, provided they meet all the criteria regarding driving GCMs, simulation period and scenarios.
  • New simulations from 2017 onward will use grids at least as fine as 25 km.
  1. Simulation period

All groups must carry out ERA-Interim simulations (1979-2015) for each domain they simulate.

Since not many people look at the 1950-1970 period, and since ERA-Interim does not cover that period, in order to reduce computation time, the minimum simulation period is 1970-2100, with a request for 1950-2100 if feasible.

  1. Driving GCMs and scenarios

With an eye toward IPCC deadlines, the simulations start with CMIP5 GCMs, and then eventually add some CMIP6 simulations as they become available.

The same GCMs cannot be downscaled over all domains, because sometimes the performance of some GCMs over specific domains can be very poor. In other words, only GCMs performing reasonably well over a given domain should be used.

To participate in CORDEX CORE, each RCM should downscale a minimum of 3 GCMs for 2 scenarios, RCP8.5 and RCP2.6. The two scenarios more or less cover the full IPCC range. The number of 3 for the GCMs is a minimum, ideally 5-6 GCMs would be better, if feasible.

The GCMs to be downscaled should:

  1. include both the RCP8.5 and RCP2.6 scenarios,
  2. cover the range of GCM climate model sensitivity (essentially low, medium, high sensitivity),
  3. have reasonably good quality of historical climate simulations in the regions for which they supply boundary conditions data (this is a must),
  4. have good quality of historical climate simulations for important large-scale features such as ENSO, NAO, etc. (this is probably necessary to satisfy criterion 3 for some regions).

Guided by these principles, the CORDEX CORE program gives highest priority to these GCMs: NorESM (low temperature sensitivity for RCP 2.6), MPI ESM (medium sensitivity), HADGEM2ES (high sensitivity). Further simulations may use GFDL ESM (low) and EC-EARTH (medium sensitivity), but only after completing simulations with the three primary GCMs , All RCMs should run the same GCMs over a given domain, so that we have a consistent and intercomparable set of runs.

  1. Archive

The format of output data should follow the standard CORDEX protocol. Given the large amount of data that will be produced only a minimum core set of variables is mandatory. See the Appendix to this document for two lists: mandatory and highly recommended variables. Output must be published on the Earth System Grid in order to satisfy the IPCC need to make this output available to as many researchers as possible.

Please find the framework document for download here and the variable list here