MOGREPS - The Met Office Short-range Ensemble Prediction System
Ken MYLNE
Met Office, United Kingdom
Abstract
The Met Office has developed a regional ensemble prediction system designed specifically for short-range prediction, to assess the uncertainty in the next-day weather forecast. MOGREPS (Met Office Global and Regional EPS) uses the North Atlantic and European (NAE) version of the Unified Model to provide ensemble forecasts at 24km resolution out to 36h ahead. A global ensemble is also run to 72h, primarily to provide boundary conditions for the regional model. Both ensembles have been running in the operational suite since September 2005 as part of a year-long trial.
Initial condition perturbations for MOGREPS are provided by the ensemble transform Kalman filter (ETKF). The ETKF is a version of the ensemble Kalman filter which allows computationally efficient update of the ensemble perturbations without performing an analysis. In our system the transform matrix is calculated using the whole set of observations that are available to the 4D-Var data assimilation system. Model error is accounted for by a stochastic physics scheme which perturbs key values in the parameterisation schemes. This plays an important role in the uncertainty related to forecasts of surface weather, but does not significantly feedback onto the dynamics of the model. The regional ensemble is driven with initial condition perturbations and lateral boundary conditions derived from the global ensemble.
The performance of the ensemble systems has been encouraging. For 500hPa geopotential height forecasts the spread of the ensemble matches the root-mean-square error of the ensemble mean forecast at around 3 days, and rank histograms are near-flat at these lead times. Subjective feedback on the system's performance from forecasters has been mostly positive. Further verification results will be presented.