The coupled atmosphere/ocean/sea-ice models used for prediction on seasonal and decadal time scales typically exhibit a cold bias in North Atlantic sea surface temperature (SST) due to the misplacement of the Gulf Stream and the North Atlantic Current (see Figure 1). The cold bias is known to impact the overlying atmospheric circulation, leading to a bias in the atmospheric circulation over the North Atlantic as well as affecting the atmospheric variability in the Euro-Atlantic sector, including Germany.

Figure 1: A schematic showing the typical path of the Gulf Stream and North Atlantic Current in coupled forecast models. The contours follow the observed surface circulation. A large discrepancy is seen in the latitude band between Canada and Europe. This is the region occupied by the North Atlantic cold bias.

The first aim of ATMOS-MODINI is to alleviate the cold bias in the MiKlip prediction system by introducing a correction to the path of the Gulf Stream and the North Atlantic Current in the model and then to test the impact of this correction on the variability and forecast skill of the model, the latter by carrying out retrospective forecasts (also known as hindcasts).

The second aim of ATMOS-MODINI is to explore ways to improve the initialization of the MiKlip system in the tropics. It is well known that the tropics are a source of predictability for the extratropical atmosphere. However, initialisation of a forecast system in the tropics is more difficult than at other latitudes because of the vanishing at the equator of the Coriolis force due to the Earth’s rotation.

In the first phase of MiKlip we explored the MODINI initialization technique – a simple technique that uses the time series of observed wind stress applied to the ocean model component to initialize a coupled forecast model. We showed that hindcasts initialized using  MODINI have skill in the Pacific sector out to decadal time scales with implications for forecasting global mean surface air temperature as well as the atmospheric circulation.

Figure 2 shows a forecast initialized on January 1, 2015, using MODINI indicating warmer temperatures in the tropical Pacific that persist at least to 2024. The forecast also accurately predicted that 2015, globally, would be the warmest year ever recorded. ATMOS-MODINI will continue to the explore MODINI initialization as well as ways to combine MODINI initialization with other initialization procedures used in MiKlip. 

Figure 2: Forecasts for surface air temperature in Celsius compared to the 1990-2006 mean. Red indicates a forecast for warmer temperatures than for the period 1990-2006, blue colder temperatures (see Thoma et al., 2015, for the details).


Thoma, M., R.J. Greatbatch, C. Kadow and R. Gerdes, 2015b: Decadal hindcasts initialised using observed surface wind stress: Evaluation and Prediction out to 2024, Geophys. Res. Lett., 42 (15), 6454-6461. doi: 10.1002/2015GL064833.



GEOMAR Helmholtz Centre for Ocean Research Kiel
Prof. Dr. Richard J. Greatbatch

Remote control on North Atlantic Oscillation predictability via the stratosphere

2017 - Quart. J. R. Meteor. Soc., 143 (703B). pp. 706-719

Hansen, F., Greatbatch, R. J., Gollan, G., Jung, T. and Weisheimer, A. (2017), . DOI 10.1002/qj.2958.

Hansen, F. | R.J. Greatbatch, G. Gollan, T. Jung and A. Weisheimer

State-Dependence of Atmospheric Response to Extratropical North Pacific SST Anomalies

2017 - Journal of Climate, 30 (2). pp. 509-525

Zhou, G. | M. Latif, R. J. Greatbatch and W. Park

Interannual variability of tropical Pacific Sea level from 1993 to 2014

2017 - Journal of Geophysical Research: Oceans, 122 (1). pp. 602-616

Zhu, X. | R.J. Greatbatch and M. Claus

Initialization shock in decadal hindcasts due to errors in wind stress over the tropical Pacific

2016 - Clim. Dyn.

Pohlmann, H. | J. Kröger, R. J. Greatbatch, W. Müller

Austral Winter External and Internal Atmospheric Variability between 1980 and 2014

2016 - Geophysical Research Letters, 43 (5). pp. 2234-2239

Ding, H. | R.J. Greatbatch, H. Lin, F. Hansen, G. Gollan and T. Jung