Evaluation of High Resolution Unified Model Configurations at the Hazardous Weather Testbed

Many operational centres are now realising significant improvements to forecasts of rainfall from kilometre-scale models that are capable of permitting convection explicitly. Such models already provide big benefits compared to the previous generation models with grid length of order 10 kilometres with parameterised convection. However at kilometre scales convection is under resolved, which may result in adverse effects such as excessive peak precipitation intensities, too large convective cells, too few cells and delayed convection initiation. At the Met Office a 1.5-kilometre convection permitting version of the Unified Model (UM) is run routinely for the UK and utilised operationally by meteorologists. Both objective verification metrics and feedback from operational meteorologists have shown that, for the representation of convection, running the UM at 1.5 kilometres provides significant benefits over the previous 4-kilometre convection permitting model. Conversely, in the context of the United States, there appears to have been something of a consensus that 3 kilometres is high enough resolution – certainly for larger storms on the Great Plains. At the Met Office there is ongoing work to better understand the resolution dependence of explicit convection in models down to much higher resolutions than 1 kilometre. The aim of this work is to more clearly understand model convective behaviour as a function of gridlength and, as a consequence, to enable informed decisions about future model resolution. In the Met Office, operational meteorologists provide feedback on operational model performance and are often involved in the assessment of new model configurations prior to operational acceptance. This provides a closed R2O-O2R loop that is of great benefit to the future development of the UM for operations. Over recent years we have trialled a number of UM versions as part of the annual forecasting experiments in the NOAA Hazardous Weather Testbed (HWT). The Met Office became a contributing partner in these experiments, both in terms of model-forecast data and personnel (forecasters and researchers), beginning in 2012. The same paradigm for NWP development is applied to the HWT collaboration, the planning and assessment of the UM contributions are a joint effort between Science and Operations aimed at not only supporting the goals of the HWT, but also to provide maximum benefit to UM development. Case studies post-HWT show, at least for the UM, that gridlengths below 4 kilometres give benefits in terms of initiation time, storm structure etc. In practice an important consideration is whether these benefits are worth the large increase in supercomputer resource required to deliver them compared to other model enhancement options such as larger domains, longer/more frequent runs, ensembles etc. This depends on the impact of the benefits on the real forecast, i.e. how these benefits pull through into operations. The poster will provide insight into some of the case studies performed and how the assessment of these by researchers, forecasters and at the HWT will help inform future development and configurations of the UM.