Impact of a Stochastic Kinetic Energy Backscatter scheme across time‐scales and resolutions. (9th April 2014)

• Record Type: Journal Article
• Authors: Sanchez, Claudio; Williams, Keith D.; Shutts, Glenn; Collins, Matthew
• Is Part Of: Quarterly journal of the Royal Meteorological Society. Volume 140:Number 685(2014:Oct.)
• Page Start: 2625
• Page End: 2637
• Publication Date: 2014-04-09
• Subjects: Meteorology -- Periodicals; 551.5
• DOI: 10.1002/qj.2328
• Languages: English
• ISSNs: 0035-9009

Abstract

<abstract abstract-type="main" xml:lang="en"> <title> <x xml:space="preserve">Abstract</x> </title> <p>Stochastic physics is one of the preferred methods to represent model uncertainty in ensemble prediction systems of medium‐range weather prediction and seasonal forecasting. These schemes increase the ensemble spread and improve probabilistic skill scores. However, little is known about how the stochastic perturbations interact with different atmospheric processes. In order to provide deeper insight into the impacts of stochastic physics on the representation of the atmosphere the stochastic kinetic energy backscatter (SKEB2) scheme has been used in the Met Office Unified Model (MetUM) across different time‐scales. We use 'classic' verification techniques such as the Root Mean Error Square (RMSE) index in combination with novel 'object‐oriented' verification metrics such as the Reading University Tracking system (RUTRACK) for extratropical cyclones. We find that the SKEB2 degrades the RMSE and Anomaly Correlation Coefficient (ACC) of individual short‐range deterministic forecasts. On average the kinetic energy backscatter by the SKEB2 counteracts the excessive dissipation of extratropical cyclones, improving the model, but its forcing does not scale well across resolutions. Over the Tropics the SKEB2 improvements of the mean climatology and temporal variability are noteworthy, but driven by spurious Rossby waves. There are aspects of the SKEB2 that could be improved to create a more realistic stochastic representation of model uncertainty.</p> </abstract>