On the Limitations of Uncertainty Schemes (Invited Presentation)

With clear evidence that initial-condition uncertainties are not
sufficient to entirely explain forecast uncertainty, the role of model
uncertainty is receiving increasing attention.  Operational weather
centers now routinely employ stochastic or other perturbation methods to
increase the reliability of ensemble forecasts.  While the performance of
stochastic parameterization schemes such as the stochastic kinetic-energy
backscatter or the stochastically perturbed physical tendency scheme --
is undisputed, they have been criticized as being added a posteriori to
models that have been independently developed and tuned.

In contrast, schemes that aim at representing uncertainty at its source,
for example the stochastically perturbed parameter approach, are by
themselves not able to generate enough spread to produce reliable
ensemble forecasts as forecast lead time increases.  These claims will be
supported with probabilistic forecasts produced by the Weather and
Research Forecast (WRF) model.

The argument will be laid out that current weather models have
insufficient upscale propagation of errors, so that the introduction of
realistic uncertainties in the physical processes cannot produce reliable
ensemble systems.