Atmospheric blocking and upper‐level Rossby‐wave forecast skill dependence on model configuration

Weather models differ in their ability to forecast, at medium range, atmospheric blocking and the associated structure of upper‐level Rossby waves. Here, we evaluate the effect of a model's dynamical core on such forecasts. Operational forecasts from the ensemble prediction systems (EPSs) of the European Centre for Medium‐Range Weather Forecasts (ECMWF), the Met Office (MO) and the Korean Meteorological Administration (KMA) are used. Northern Hemisphere model output is analysed from the winters before and after a major upgrade to the dynamical core of the MO‐EPS (called MOGREPS). The KMA‐EPS acts as a control as it uses the same model as MOGREPS, but uses the older dynamical core throughout. The confounding factor of resolution differences between MOGREPS and the KMA‐EPS is assessed using a MO forecast model hindcast experiment with the more recent dynamical core, but with the operational resolution of the KMA‐EPS. The introduction of the new dynamical core in MOGREPS has led to increased forecast blocking frequency, at lead times of 5 and 7 days, counteracting the typically observed reduction in blocking frequency with lead time. Hit rates of blocking activity, onset and decay are also increased in the main blocking regions (without a corresponding increase in the false positive rate). The previously found reduction of the upper‐level ridge area and tropopause sharpness (measured by an isentropic potential vorticity gradient) with lead time is also reduced with the new dynamical core. This dynamical core improvement (associated with a reduction in implicit damping) is thus demonstrated to be at least as effective as operational resolution improvements in improving the forecasts of upper‐level Rossby waves and associated blocking.