Contributed by Anders Persson, a HEPEX guest columnist for 2014
My previous column featured the severe July 2004 flooding in southern Sweden and how I was brought in to help my hydrological colleagues inform the press. A common question was if the heavy rain had been forecast. My answer was that there had been clear indications several days ahead and a formal warning had been issued a few days before the event.
But a post-mortem investigation complicated the picture: SMHI had issued forecasts along two rather contradictory lines. The meteorologists had more or less verbatim accepted the ECMWF deterministic high-resolution forecasts (HRES). It correctly predicted the rain (both the intensity and the position in the interior of southern Sweden) four days ahead. But the HRES was “jumpy” and next day indicated heavy rainfall far to the northwest close to the Norwegian border and next day far to the southeast near the Baltic Sea (see fig 1). Only the day before did the HRES return to its forecast of heavy rain in the interior of southern Sweden.
The meteorologists on duty followed the HRES computer output which four days before the event (9 July, not shown) positioned the heavy rain more or less were it occurred. But the next day’s 3-day forecast the HRES had it far to the NW, the next day far to the SE and only on the day before was the excessive rain put in the right place (Figure 1, left column).
The hydrologists decided to “paint with a broad brush” and initially issued warnings that within 3-4 days there would be heavy rain somewhere in Southern Sweden (Figure 1, right column). Two days before the event they narrowed the rain area to the interior of Southern Sweden and the day before their forecast was more or less identical to the meteorologists’.
Formally this means that half of the meteorologists’ forecasts were “correct”. But due to the forecast “jumpiness” it is difficult to see how anybody could trust these forecasts and, more importantly, base any decisions on them.
The hydrological forecast service took a different approach. Realising, perhaps from discussions with the meteorologists and/or the spread of the ensemble forecasts that the situation was not very predictable, their forecasts four days in advance warned about heavy rain somewhere in Southern Sweden. This forecast was only marginally altered the next day. Two days in advance they shrunk the possible target area to central parts of Southern Sweden and then, the day before the rain started, they issued more or less the same forecast as the meteorologists.
The hydrologists’ first three forecasts were not as detailed as their meteorological colleagues’. They took the line to only state what they were certain about and then, as the certainty grew, slowly “zoomed” the forecast into the correct target area.
We can argue whether their forecast was “better” than the meteorologists or not. However, their forecasts gave an impression of consistency which the meteorological forecasts surely lacked. I can imagine that the responsible water authorities therefore felt they could trust these forecasts more than those which rather “jumped around”.
If this had been an isolated case of misjudgement in Sweden in July 2004, we could leave it at that. But in my experience from around the world, or at least Europe, is that this culture of overconfident detailed forecasting has been a general meteorological forecast practice for more than 30 years, more or less since the start of ECMWF – which might not be a coincidence…
On numerous training courses I have recommended meteorological forecasters not to blindly accept the less predictable details in the deterministic output from the HRES.
I have showed them theoretically and practically that ignoring or smoothing out small and therefore less predictable scales in the HRES would not only highly improve their performance in relation to the computer generated output, their forecasts would also, as “an extra bonus”, be much less “jumpy” and appear more reliable.
But the meteorological forecasters were not keen to follow my “less is better” advice. I have never understood why, in particular since “beating the computer” was high on the agenda for many of them. Some claimed that they wanted to maintain details to make their forecast appear “meteorologically realistic”, others seemed be happy to follow the HRES and take the credit when it was right and blame it when it was wrong .
But shouldn’t I take comfort in the fact that at least the SMHI hydrologists had followed my “less is better” advice?
-But I never told them! The hydrologists never attended the meteorological courses. They came up with this clever solution on their own!
Next post: 23 May 2014.
Anders will be contributing to this blog over the year. Follow his columns here.
 The legendary Danish forecaster Steffen Hartby is one of the few who has publicly criticised this attitude which he considered undermined the standing of the weather forecast community. For those who can read Scandinavian languages, see his article “Vil du virkelig være meteorolog?” (Do you really want to become a meteorologist?) in “Vejret” published by the Danish Meteorological Society, 2004:4 pages 15-17.