• Luke Boxall

Forecast Skill

September/October are months known for their very low forecast skill, often hated by those trying to forecast medium-range weather, every year around this time of year there is normally some kind of big forecast bust.

One of the reasons for this is the fact that it is a transition season - the global jet-streams/systems are on the move from their Summer locations and heading into their winter locations, in this state of movement - Teleconnection correlations often break down and often ends up as a variable/oscillatory period switching between Summer/Winter-style regimes.

The other reason is due to a peak in global tropical activity at this time of year and this pumps a lot of chaotic energy into the atmospheric which the computer models just can not handle. They have got much better in general over the past few years but still some very obvious flaws occur with the systems themselves but also the hidden element of how these systems interact with the mean flow which is much harder to observe and quantify.

This issue has been brought to light recently due to the main forecasts only calling for Hurricane Michael to strengthen just about to CAT 3 even just 48 hour out (originally 60-72 hours out it was a very weak CAT 1 in the outlook - 60 hour forecast shown below) - however it smashed into Florida just 1mph shy of being a CAT 5. To be fair on this occasion the track was well forecasted so the weather centres will focus on this bit as a success but the models failed on strength and this has knock-on effect with forecast skill downwind if you can't get forecasts right 48-60 hours ahead.

Another forecast bust caused from tropical storm/hurricane occurred over Europe last Fri (12th).

Hurricane Leslie was a Tropical Storm/Hurricane that was very poorly handled by the weather models - often taking a different path to what was forecasted - it ended up circling the Atlantic for a good 10 days and then drifted slowly North-Eastwards for another 10 days.

Thursday 11th October the forecast for Leslie had the storm re-curving south-wards towards the Canary Islands - here is Thursday 11th October ECMWF Operational 12Z run for forecast hour 60 and you can see the Storm over the Canary Island circled:

The remains of Hurricane Michael skirts above the circle and the forecast track was important for both these storms - with Leslie staying south- ex-Michael was on course to hit UK early next week and help to push in a strong Westerly flow over UK into North Germany by Weds - South-Westerly regimes are most often followed by Westerly regimes.

Below is the percentage of Thursday's ECMWF ENS 0z run that went for the various weather regimes which shows a South-westerly to start the week (SWA) - moving more into a Westerly style flow by Weds-Fri (WA/WZ/TRW) with a 10-30% risk of Conti-high (BM) - models had been struggling with over-forecasting high pressure for a large number of weeks in the 6-10day window - so 10-30% risk of BM was fairly low and the trend was towards WA/WZ at this point in time from:

Then by Friday's 0z Run this has happened:

The Westerly support for Weds-Fri dropped drastically and it become much more apparent that Weds was doing to be a Conti-high - with strong support for Conti-high to tick over until the weekend. This meant warmer but much calmer forecasts for DE - what caused this?

The forecast from Hour 60 (like for Michael over Florida) was completely wrong but this time it was the track that was wrong for Leslie - strength was actually correct. Previously only around 10-30% wanted to hit Portugal - now most runs wanted to hit S Portugal - shown below is the ECMWF Operational Forecast from Friday 12Z run - where it hits near Lisbon:

The knock-on effect of this was the the remains of Michael in the Mid-Atlantic was then drawn into the wake of Leslie and pushed into Iberia as well - which formed a high pressure bulge to the North of Portugal and meant the in-coming lows during that following week were now further North - before we had 35-50kts of wind shown over UK - shown below the forecast now was for 15-35kts:

Butterfly effect - slight change in storm track brings the potential for big changes downstream - to have the forecast flip between hour 48-60 is far from ideal - it just raises the point that for something we rely upon so much - there is still a lot of uncertainty when it comes to weather and there are often surprises.

My rule of thumb when tropical storms are about is to drop confidence to zero or at least very low because more often than not - model show hints of wind in the outlook but the out-turn is very random - as this case proves. This is also why forecast skill is often very low this time of year with multiple tropical storms around the globe often occurring concurrently.


The Weather Perspective Ltd