| Summary: | The use of kilometre-scale ensembles in operational forecasting provides challenges for forecast evaluation and interpretation. New spatial methods for characterising and verifying convection permitting ensembles are developed, and tested on the 12 member Met Office 2.2km resolution UK ensemble. Each ensemble member is regarded as an equally plausible realisation of the true atmospheric state. A novel methodology is presented for spatial ensemble characterisation based on the Fractions Skill Score. Characterising the domain-wide ensemble behaviour, these methods identify useful spatial scales and spin-up times for the model, and demonstrate the upscale growth of errors and forecast differences. The ensemble spread is shown to be highly dependent on the spatial scales considered and the threshold applied to the field. Comparing differently-generated ensemble systems shows the utility of spatial ensemble evaluation techniques for assessing different ensemble perturbation strategies. It is also important to consider location-dependent ensemble behaviour. A new method for calculating the location-dependent spatial agreement of ensemble members is presented. Through comparing with radar observations, the location-dependent spatial skill of the ensemble is also quantified. These methods are verified using an idealised experiment. Six convective cases, and a summer season, are used to demonstrate the methods in an operational context, with links made to physical processes. Overall, the ensemble system is reasonably well-spread spatially. Poorer spread-skill is associated with a low fractional coverage of rain, and low synoptic-scale rain rates. Higher confidence in the location of precipitation is found to the northwest of the UK. To investigate coherent physical structures in the ensemble, the spatial approach was used to inform the calculation of multivariate correlations. Using the spatial approach, physically-meaningful correlations which demonstrate inter-variable relationships are obtained. Overall, the spatial approach is found to give useful information for forecasting, and for the interpretation and evaluation of convection-permitting ensembles.
|
|---|
