On the operational predictability of blocking

The entire 7-year archive of ECMWF operational analysis and forecast data is used to assess the skill of the Centre’s model in short- and medium-range forecasting of atmospheric blocking. The assessment covers 7100-day periods, from 1 December to 10 March of all winters from 1980-81 to 1986-87, inclusive. A slightly modified version of the Legenäs and Økland objective zonal index is used to quantify both observed and forecast occurrence of blocking. The study is performed on 500 hPa geopotential height and on Euro-Atlantic and Pacific blocking separately. It is found that blocking frequency is severely underestimated in medium-range forecasts; the model is, on average, reasonably skilful if the initial conditions are blocked, but blocking onset is poorly represented if it occurs more than a few days into the forecast. This inability in entering the blocking regime has a substantial impact on the systematic error of the model.