Using ensemble data assimilation to diagnose flow-dependent forecast reliability
Rodwell, Mark J.
Weather forecasting is fundamentally a probabilistic task due to the growth of unavoidable initial-state uncertainty. Moreover, the growth rates of these uncertainties can depend on the atmospheric flow so that predictability may vary from day to day. The established approach to representing uncertainty in probabilistic forecasting is to make an ensemble of forecasts, each starting from a slightly different initial state and including a different realisation of model uncertainty. A key question is how to assess the ensemble’s ability to represent the flow-dependent growth of uncertainty. Results suggest that such assessments are not easy to make at the medium range due to complications associated with error propagation and non-linear interactions. Using a specially developed ensemble reliability budget, appropriate for shorter-range assessments within the data assimilation window, these issues can be minimised and flow-dependent deficiencies in representing uncertainty can be identified. An analysis of the reliability budget can also help identify the causes of deficiencies in representing uncertainty. Results are illustrated for a flow situation where mesoscale convection is likely to occur over North America and which often results in reduced predictive skill for Europe several days later.
Accès à la notice sur le site du portail documentaire de Météo-France