Séminaire
Flooding from intense rainfall, resulting from convection, causes millions of euros of damage each year. However, convection has limited predictability often resulting in short lead times for warnings of such events. I present work from my PhD and subsequent post-doc which aimed to determine the spatial scale of perturbation growth across the UK from simple stochastic perturbations, and the impact of perturbation growth from a physically-based stochastic boundary layer scheme compared to initial and boundary condition perturbations. These themes are examined using convective-scale configurations of the Met Office Unified Model. The spatial perturbation growth is shown to be influenced by the environment in which it occurs, with environments leading to more organised convection having more localised growth compared to those resulting in scattered showers. Furthermore, through the use of a super-ensemble it is shown that growth from a stochastic boundary layer scheme can equal that of initial and boundary conditions. Evidence is also presented that shows neighbourhood post-processing techniques can artificially increase the ensemble size, allowing the production of more reliable probabilities of rainfall events being forecast. These results have implications both within operational forecasting centres and the research community in the interpretation of convective-scale ensembles for highly localised precipitating events.
bruno.deremble@ens.fr