Met Office Experiences with Convection Permitting Models

Humphrey Lean, Met Office@Reading, Reading, UK

A prerequisite of an NWP based nowcasting system is a model that can accurately represent the phenomena that are required to be forecast. The Met Office has been carrying out research into km scale models for a number of years and has been routinely running a UK 4km model and, more recently, a UK 1.5km model (the "UKV") for forecasting out to T+36. I will draw on these experiences to present a summary of the benefits and deficiencies of these "convection permitting" models relating, in particular, to the representation of convection. When suitable (fuzzy) precipitation verification techniques are used the 4km model performs better than a 12km model and, in turn, the 1.5km model performs better than the 4km. These results are also borne out by subjective forecaster verification. Despite these good signs, it is clear that convection is still under-resolved with models at these gridlengths. I will describe ongoing work to understand and improve its representation including research at higher horizontal resolutions and projects to compare in more detail the representation of convection with observational data. I will also mention the well known predictability issues with models at these resolutions which mean that we are moving towards probabalistic products and running convection permitting ensembles.