The ECMWF Integrated Forecasting System (IFS) includes a time-evolving representation of the 3D ocean state. This coupling to a dynamic ocean model improves the representation of air–sea interactions and has been shown to improve forecasts at the sub-seasonal to seasonal (S2S) time range (from about 2 weeks to about 2 months). However, the inclusion of additional ocean processes also introduces the potential for systematic errors in sea-surface temperature (SST), which can have a negative impact on forecast quality. In particular, ocean models with a grid spacing of about 25 km, as used in the IFS, struggle to accurately simulate the location and structure of the Gulf Stream and its associated sharp gradients in SST.
Here we describe recent work at ECMWF to evaluate the impact of such SST errors on S2S forecasts using an online SST-bias correction methodology. The reference S2S experiment (CTRL) is a 15-member ensemble initialised on the 1st and 15th of each month of an extended winter period (November–March) from 1989 to 2015 (i.e. 270 start dates). The bias-corrected experiment (BCFC) is the same as CTRL, but the SSTs seen by the atmosphere in the North Atlantic region are adjusted by the model SST bias derived from CTRL, which varies as a function of location, calendar start date, and forecast lead time.
The applied bias correction effectively reduces SST biases in the North Atlantic region. The resulting southward shift of the Gulf Stream drives changes in convective precipitation and vertical motion (not shown), which has consequences for atmospheric predictability that extend beyond the North Atlantic. In particular, we find that reducing North Atlantic SST biases leads to significantly improved S2S forecasts of atmospheric circulation anomalies over Europe. Furthermore, the impacts extend beyond the North Atlantic and Europe and circumnavigate the globe along the northern hemisphere subtropical waveguide (the figure shows an example). This response is typical of the propagation of stationary Rossby wave activity along the jet stream.
Interestingly, this impact of SST biases on forecast skill is modulated by the Madden–Julian Oscillation (MJO). Forecasts with an active MJO in the initial conditions exhibit a stronger impact over Europe and along the northern hemisphere waveguide, whereas forecasts without an active MJO have a stronger response over the Gulf Stream and North Atlantic. This sensitivity is unrelated to changes in MJO forecast skill, which is not sensitive to North Atlantic SST biases. Instead, we speculate that this effect is a consequence of the impact of the MJO on the background state and its associated teleconnections that steer or obstruct planetary wave activity that is initiated in the North Atlantic region.
In conclusion, the results from these sensitivity experiments provide important evidence for the potential benefits to ECMWF forecasts of higher-resolution ocean models (i.e. with a grid spacing of less than 10 km) that can better resolve the position of the Gulf Stream. Such models will be investigated at ECMWF during the next few years as part of the recently funded Horizon 2020 NextGEMS project.
Further information on the impact of SST biases on ECMWF forecasts can be found in an article by the authors entitled ‘Hemispheric impact of North Atlantic SSTs in sub‐seasonal forecasts’ in Geophysical Research Letters, https://doi.org/10.1029/2020GL091446.