August 11, 2011

Models go astray, all of them including the EC and GFS. By the time you get to the end of the run, errors amplify.  Most forecasters are wary of week 2 detail.

Earlier in the spring and summer,  model analyses are important for the determining the extent of the precipitation, severe weather and heat. Now it is becoming critical for the tropical threat. There is a small difference between a flow that whips a storm to the maritimes or out to sea and one that draw it inland over the mainland, even in a year like this when the mean flow is more conducive for landfall.

We have to be careful with puttting too much faith in an operational deterministic run. Occasionally the operational run is an outlier in the ensemble of runs for that time period.

That is why there are ensembles. These are parallel runs of the model with perturbations maade to the initial state. When most all agree, we have a higher confidence in the solution, because we know that even if the model doesn't get the initial state exactly right, there is enough stability in the pattern to get a similar result.

The member start out virtually identical.

A high degree of variance shows up in the ensembles as a flat flow. You can't use means in that case because they erroneously project a flat (zonal) flow. In actual fact, most members can be amplified with ridges and troughs but they put the positions in different locations which average out flat. See for example yesterday morning's EC ensembles left versus the operational right.

The example of an outlier came with the 00z GFS last night (white lines for select height lines in the spaghetti chart of all ensemble member right). You can see most members in red and green were much more amplified and here the ensemble mean was a  better tool (left). and WHAT WE ARE RELYING ON CURRENTLY.

Here is the same member by member 384 hour forecast view. See the operational run top left varies from the other solutions.

The EC ensembles run 51 members and we often want to see that ensembles set before we make decisions in weeks 2 and 3. When the MJO is active we use that as a a check and sometimes as the chosen tool. We look at the 27 day solaar cycle, recurving typhoons, and other proprietary tools, etc.

Remember that the CME on the sun that caused the eruption suggested a brief flat flow with above normal heights in southern Canada and northern US and around the hemisphere in northern Siberia followed by a developing PNA pattern with a trough off the west coast, a western ridge and eastern trough. Whether or not the sun can take a bow in the fact this sequence is shown in the models we will never know for sure. We will examine the 500 mb anomalies early next week and again in 15 days to get a better idea if this was a factor.