EGHH

Rather late due to work commitments, but nethertheless attached is my very simplistic look at the last set of models (0z and 6z) and the coming [potentially] snow. Models for Friday.pdf

See here, Steve: http://forum.netweather.tv/topic/75477-winter-model-discussion-cold-hunting-080113-00z-onwards/page__view__findpost__p__2469633

I doubt it to be honest. The control run is simply the same set of initialization data as the operational run but run at the same (generally lower) resolution as the rest of the ensemble. Essentially you'll see cases where the operational and the control agree (so the higher resolution of the operational isn't crucial in finding the evolution) and cases where the control goes off on it's own - indicating a situation that's sensitive to resolution, thus error (or smaller scale features). The remainder of the ensemble runs are created from initialization data that is tweaked in a special, statistically significant way. The results of these run show how sensitive the evolution is to errors in the initialization data. It is done thus because it's impossible to sample the whole atmosphere at the same time. So, the observation data that is received is assimilated into yet another model, which produces an approximation of the whole atmosphere at the initialization point. You see, this is an approximation of an approximation of... so the ensembles show how sensitive the situation is to these errors in the start point and, in a sense, how sensitive the current base state of the atmosphere is to change. That's before you even get into the computational approximations (parameterisations) to the thermodynamic equations that the models need to solve. I think most forecasters will use graphs derived from the ensemble prediction systems to show the variablity in possible outcomes and use this data as a guide as to how certain one can be about a certain forecast parameter - such as temperature, precipitation etc. I think looking at them synoptically (other than characterising them by GWL type, for instance) is little more than of passing interest. A more thorough look at how to use ensemble products can be found here: http://www.ecmwf.int/products/forecasts/guide/user_guide.pdf

This presentation gives a contextual description of how Shannon Entropy is used: http://www.ecmwf.int/newsevents/meetings/forecast_products_user/Presentations2011/PDFs/Mylne_Parker.pdf This also contains intersting information about how the Grosswetterlagen (GWL) types are presented (don't think I've seen much if any references to GWL on here before)