Coordinated climate simulations using the IFS

Chris Roberts
Sarah Keeley
Franco Molteni
Retish Senan (all ECMWF)
Torben Königk (SMHI)


Since 2016, researchers from ECMWF and the EC-Earth consortium have been collaborating with partners from 19 European institutes in the EU-funded PRIMAVERA project to develop a new generation of global climate models. The multi-decadal simulations performed for PRIMAVERA are providing constraints for model development and evaluation that are complementary to those available from short-term (re-)forecast datasets. These insights will be key to improving ECMWF forecasts on medium-range to seasonal timescales.

The core aim of PRIMAVERA is the development and process-based evaluation of a new generation of global climate models, with a focus on the impacts of increased horizontal resolution on simulations and predictions of regional climate. Ocean and atmospheric model resolution can affect many aspects of climate simulations, including climatological biases and the representation of key modes of climate variability, such as the El Niño–Southern Oscillation (ENSO). Understanding such sensitivities in ECMWF’s Integrated Forecasting System (IFS) and the EC-Earth model, and their relative importance at different timescales, is crucial for improving the fidelity of the analyses and forecasts produced by ECMWF and the EC-Earth consortium.

The experimental backbone of the PRIMAVERA project is a multi-model ensemble of climate model simulations covering the period 1950 to 2014. Within the project, ECMWF and EC-Earth are providing simulations based on the IFS atmosphere coupled to the NEMO/LIM ocean–sea ice model. The EC-Earth consortium is providing configurations based on IFS Cycle 36r4 (as used in the previous seasonal forecasting system, S4) and researchers at ECMWF are providing configurations based on IFS Cycle 43r1 (as used in the latest seasonal forecasting system, SEAS5). The EC-Earth and ECMWF submissions differ in several ways, including the type of atmospheric grid (linear vs cubic octahedral), coupling strategy (single executable vs dedicated coupler), tuning of model components, ocean and sea ice model versions, and the inclusion of ocean waves. The PRIMAVERA project provides a unique opportunity for systematic comparisons of different versions of the IFS with each other, and with other climate models.

Preliminary results concerning the impacts of increasing ocean and atmosphere resolution in climate experiments with IFS Cycle 43r1 can be summarised as follows:

  • All configurations of ECMWF-IFS successfully reproduce the observed long-term trends in global mean surface temperature.
  • Increasing the atmospheric resolution from 50 km to 25 km has little impact on climatological surface biases but increases the magnitude of a cold bias in the lower stratosphere.
  • Increasing the resolution of the NEMO ocean model from about 100 km to about 25 km substantially reduces biases in North Atlantic sea-surface temperature (SST) and northern hemisphere sea ice extent, but it increases the magnitude of a warm bias in the Southern Ocean.
  • Increasing the ocean resolution also improves the simulated magnitude and asymmetry of ENSO variability and improves the representation of associated non-linear SST–radiation feedbacks.
  • Ocean coupling and increased atmospheric resolution seem to improve the representation of teleconnections between tropical Pacific rainfall and geopotential height anomalies in the North Atlantic, but the significance of this result needs to be assessed.
  • Work is ongoing within the PRIMAVERA project to assess the impact of increased resolution on climate variability and extremes.

Multi-decadal coupled experiments are not currently performed routinely at ECMWF. However, the asymptotic behaviour of the coupled model is becoming important for numerical weather prediction with the development of coupled approaches to data assimilation and reanalysis. In systems using such coupled approaches, the climatological attractor of the model is important because of its influence as a background field for periods and/or regions with limited observational constraints. The scientific and technical developments required for the ECMWF-IFS configurations used in PRIMAVERA will thus help to evaluate the representation of the more slowly evolving components of the Earth system as part of the model development process.

Temperature and SST anomalies in different IFS configurations. Global mean anomalies of 2‑metre temperature over land using IFS Cycle 43r1 with a grid spacing of about 25 km (ECMWF-IFS-HR) and 50 km (ECMWF-IFS-LR) (top) and sea-surface temperature (SST) using an ocean model resolution of about 25 km (ECMWF-IFS-HR) and about 100 km (ECMWF-IFS-LR) (bottom), relative to the period 1981–2010. Model data are ensemble means from coupled configurations of ECMWF‑IFS. Yellow shading indicates periods corresponding to the Mount Pinatubo (1991), El Chichón (1982), and Mount Agung (1963) volcanic eruptions.

More details on the results presented here can be found in an article by Chris Roberts et al. in Geosci. Model Dev. doi:10.5194/gmd-11-3681-2018.