Uncertainty Tracer
Motivation:
- Suppose that we have dry adiabatic reasonably linear motion evolving over a region of Oklahoma, a tracer that tracks modeling error should treat this as reasonably certain
- e.g. smoothed orography and other idealizations should be a "source" of variance and this should propagate with finite speed (i.e. approximately hyperbolically with the wind).
Notes
-
Assume that we have a quantitative estimate of modeling uncertainty from associated physics parameterizations.
-
Assume that non-local contributions (e.g. diffusion and parabolicity) are negligible over the whole domain (Possibly better to do this in non-hydrostatic)
-
Assume that physics parameterizations have an associated "modeling uncertainty" which
-
Tractability assumption: all sources of uncertainty are the same?
-
This is probably most useful for short-term forecasts?