Impact of sea surface temperature biases on extended-range forecasts

The impact of sea surface temperature (SST) biases on extended-range forecast skill scores are investigated using a series of coupled forecast experiments in which the SST biases are removed by adding a correction term during the coupled integrations. The bias correction depends on the forecast lead time and on the starting calendar date, and it is estimated from a set of coupled re-forecasts. Results show that this methodology is efficient in removing the SST biases in the Extratropics but is less successful in the Tropics. Therefore additional experiments have been performed where the bias correction has been applied only to the Extratropics or to the North Atlantic.  Removing the SST biases in the Extratropics results in a modest reduction of the coupled model atmospheric biases over the North Atlantic, and increased extended range forecast skill over Europe. Most of these improvements are reproduced if the bias correction is applied only over the North Atlantic region. This impact of the SST bias correction is particularly strong when there is an MJO in the initial conditions. The MJO teleconnections following an active MJO over the Western Pacific are significantly stronger when the SST bias corrections have been applied, while the teleconnections following an active phase of the MJO over the Indian Ocean are unaffected.