Dependence of ensemble spread on model uncertainties for extratropical cyclone simulations

In order to examine the dependence of the ensemble forecast spread on model uncertainty, we use the model grid‐point spacing and the index of shallow‐convection top as surrogates for model uncertainty, to construct two groups of ensemble forecasts (‘experiments’). In each ensemble, we use several members that differ only in the size of their model uncertainties. An initial baroclinic‐jet basic state is constructed, with a balanced three‐dimensional perturbation, to trigger development of a baroclinic wave. In terms of the minimum central surface pressure, the spread associated with each ensemble increases with increasing model uncertainty. Similar sensitivity is observed for the maximum ensemble spread and the domain‐averaged spread in terms of surface pressure. In both experiments, the domain‐averaged spread is linearly proportional to the size of the model uncertainty, and varies in time with a power‐law dependence. The ensemble spread is mainly collocated with the primary cyclone during the cyclone development stage, when amplitude error dominates, and is elongated along the regions with strong gradients during the mature stage, when the phase error dominates. Forecast errors also grow in association with the secondary cyclones that develop in the simulation. The maximum ensemble spread associated with secondary cyclones increases with increasing model uncertainty, as for the primary cyclones. It is found that model uncertainties have a larger effect at the surface on the spread associated with the upstream cyclone, compared to the downstream cyclone (at the same integration time), because of the different stages of development of the upstream and downstream cyclones. Copyright © 2007 Royal Meteorological Society