A look at the ocean in the EC-Earth climate model

Abstract

EC-Earth is a newly developed global climate system model. Its core components are the Integrated Forecast System (IFS) of the European Centre for Medium Range Weather Forecasts (ECMWF) as the atmosphere component and the Nucleus for European Modelling of the Ocean (NEMO) developed by Institute Pierre Simon Laplace (IPSL) as the ocean component. Both components are used with a horizontal resolution of roughly one degree. In this paper we describe the performance of NEMO in the coupled system by comparing model output with ocean observations. We concentrate on the surface ocean and mass transports. It appears that in general the model has a cold and fresh bias, but a much too warm Southern Ocean. While sea ice concentration and extent have realistic values, the ice tends to be too thick along the Siberian coast. Transports through important straits have realistic values, but generally are at the lower end of the range of observational estimates. Exceptions are very narrow straits (Gibraltar, Bering) which are too wide due to the limited resolution. Consequently the modelled transports through them are too high. The strength of the Atlantic meridional overturning circulation is also at the lower end of observational estimates. The interannual variability of key variables and correlations between them are realistic in size and pattern. This is especially true for the variability of surface temperature in the tropical Pacific (El Niño). Overall the ocean component of EC-Earth performs well and helps making EC-Earth a reliable climate model.

A Flux coupling in EC-Earth

Global conservative regridding of a flux F requires that

$$ F = F^{\ast}, $$

(1)

where the asterisk indicates the flux after regridding. The fluxes can be written as the sum over all grid cells in the source and target grids, respectively,

$$ F = \sum_i f_i A_i, $$

(2)

$$ F^{\ast} = \sum_j f^{\ast}_j A^{\ast}_j. $$

(3)

where f is the grid cell mean flux and A is the grid cell area. Each grid cell can consist of a number of tiles (surface types) with fractional area α _t (α ^* _t ), so that

$$ F = \sum_i \sum_t \alpha_{i,t} f_{i,t} A_i, $$

(4)

$$ F^{\ast} = \sum_j \sum_t \alpha^{\ast}_{j,t} f^{\ast}_{j,t} A^{\ast}_j. $$

(5)

Since both summations have a finite range, they can be swapped

$$ F = \sum_t \sum_i \alpha_{i,t} f_{i,t} A_i, $$

(6)

$$ F^{\ast} = \sum_t \sum_j \alpha^{\ast}_{j,t} f^{\ast}_{j,t} A^{\ast}_j. $$

(7)

In the coupling, flux conservation should be guaranteed for each tile type,

$$ \sum_i \alpha_{i,t} f_{i,t} A_i = \sum_j \alpha^{\ast}_{j,t} f^{\ast}_{j,t} A^{\ast}_j. $$

(8)

The atmosphere model computes the fluxes f _i,t for each tile and each grid cell of the atmospheric grid. The tile fractions for land are prescribed and those of the open ocean and sea ice are received from NEMO. The latter are given for the binary ocean mask used by NEMO. The fractions are adjusted to match the land distribution so that the total of land, ocean and sea ice is everywhere equal to one in IFS.

IFS sends the tile fractions α _i,t and the tile fraction weighted fluxes α _i,t f _i,t to NEMO using the first order conservative regridding method of OASIS. This regridding method takes care of the area weights and is locally conservative. This latter property implies that the last equation above is also valid for a single grid cell in the target grid and can be simplified to

$$ \sum_i \alpha_{i,t} f_{i,t} A_i w_{i,j} = \alpha^{\ast}_{j,t} f^{\ast}_{j,t} A^{\ast}_j, $$

(9)

where w _i,j is the overlap of the grid cell i in the source grid and the grid cell j in the target grid.

The coupling now works as follows: IFS computes the fluxes for each grid cell and tile type and sends the fields α _i,t and α _i,t f _i,t to NEMO,

$$ \alpha_{i,t} \rightarrow \alpha^{\ast}_{j,t}, $$

(10)

$$ \alpha_{i,t} f_{i,t} \rightarrow (\alpha_{j,t} f_{j,t})^{\ast} \equiv \alpha^{\ast}_{j,t} f^{\ast}_{j,t}, $$

(11)

and NEMO computes the flux for each grid cell and tile type as

$$ f^{\ast}_{j,t} = (\alpha_{j,t} f_{j,t})^{\ast} / \alpha^{\ast}_{j,t}. $$

(12)