Convective-Scale Predictability and Ensembles

Advances in both high-resolution modelling and ensemble prediction have brought about a significant improvement in the Met Office's forecasting capability. Further improvements will be gained by combining these two approaches.

The recently-introduced 1.5 km mesh 'storm-permitting' model for the UK is able to represent hazardous weather such as thunderstorms, fog and damaging winds more realistically than ever before.
The current operational ensemble system (18-km mesh) produces twenty-four different, yet plausible, forecast scenarios that reveal the uncertainty associated with the weather pattern on a particular day.

We are now examining the potential benefit of running an ensemble of high-resolution forecasts. The objective is a forecast system that will accurately represent the local weather and provide useful probabilities about each event. From a practical perspective, this means that decisions to take mitigating action against the impact of severe weather will be based on a better assessment of the threat.

This requires innovative research to improve our understanding of the nature of the predictability of hazardous weather in high-resolution models, and provide a scientific foundation for new systems.

All forecasts have errors that grow bigger as the forecasts become longer. This degradation in performance happens most quickly for the small features we may care about and can now represent, such as thunderstorms. Beyond a few hours it becomes very difficult to predict that a storm will occur at a particular place or time. The forecast should then move away from specifying whether it will happen or not and instead convey an indication of the likelihood. The differences between each forecast in an ensemble can help to identify the uncertainty.

Practical issues to be tackled include; how to present probabilistic forecasts effectively; how to use a small ensemble; and how to evaluate the forecast skill.

The work involves collaboration between the Mesoscale Modelling Group in Reading and the Ensembles groups in Exeter, and between the Met Office and the Department of Meteorology at Reading University.

Key Aims:

  • To improve our understanding of the processes that govern the predictability of local weather.
  • To investigate the performance of a high-resolution model ensemble (that runs 24 forecasts) using several recent significant weather events.
  • To decide how best to design a smaller ensemble, with just a few forecasts, that gives similar results to the 24-forecast ensemble.
  • To develop new methods for assessing the performance of a high-resolution ensemble.
  • To develop new practical ways of presenting forecasts using probabilities.  

Current Projects

  • Testing a research ensemble system that runs 24 forecasts of the Met Office 1.5km model over the UK on several interesting weather events, such as the Ottery-St-Mary hail storm.
  • Comparison of Methodologies for comparing different forecasts within an ensemble. These can be used to pick a subset of dissimilar forecasts from the existing 24-km model ensemble to initiate a smaller 1.5-km model ensemble.
  • Development of new probabilistic products.
  • Research into the impact of uncertainty in the 'mesoscale' weather pattern (e.g. fronts, small cyclones) on the predictability of local weather events.
  • Research into the impact of uncertainty in the representation of physical processes (e.g. cloud and rain formation) on the predictability of local weather events.

Last updated: 3 September 2013