Representing Model Error in the AWFA Joint Mesoscale Ensemble by a Stochastic Kinetic Backscatter Scheme

A common problem in probabilistic forecasting is that most ensemble

systems tend to be underdispersive, thus leading to overconfident

uncertainty estimates and an underestimation of extreme weather

events. This underdispersivness might arise in part from a

misrepresentation of unresolved subgrid-scale processes and can thus

be remedied by stochastic parameterizations.

Here we present results from a stochastic kinetic backscatter scheme

implemented into the AWFA Joint Mesoscale Ensemble (short-range WRF

ensemble forecasting system).

In comparison to the ensemble system without stochastic parameterization,

the backscatter scheme improves the spread-error-relationship and leads to

better probabilistic skill scores, especially for extreme events.

It is also discussed how the ensemble with backscatter scheme performs

in relation to a multi-physics ensemble, where each ensemble member

uses a combination of different physical parameterizations.