Are there any good reasons to believe in a seasonal forecast, or should we look elsewhere to find a signal for our positions in the energy market?
We know that the weather outlook for the winter will have a big impact on the production and demand of energy. Mild winter normally gives a high amount of precipitation for the Scandinavian hydropower as well as a windy season, with high production in Germany. While cold winters, on the other hand, give low wind production and higher demand for energy.
Thus, if we can predict what this year’s winter season might bring; whether we will have a mild, normal (what is normal?), or cold winter, then traders and energy producers will have a much better opportunity to take the right positions on the energy market.
A long-term forecast is either based on your statistical data or on a coupled ocean & atmospheric model that computes a forecast 1-9 months in advance. Just as we see it for a normal weather forecast, the further out in the future we are looking, the less value we see in the forecast. The value of a long-term forecast more than 4 months ahead is statically close to zero.
Who is operating the global long-term models? Typically, national institutes with large processor power. Here in Europe, it is ECMWF, DWD, and Met Office the industry is looking at.
The more information, the better, so if we combine several seasonal forecasts with a moderate confidence level, it will make us a bit more certain.
If we use statistics to predict winter weather – should we then use the last 50, 20 or 10 years? Is the more-the-better approach still valid here? We know today that science that since our climate is changing and we’ve experienced the warmest winters during the last 20 years, there is not much value in looking 50 years back.
Even for as short block of time as 20 years, there are significant changes in both directions. For instance, we see that for Hamburg, during the last 10 years had 0.5 degrees colder January than the first decade of 2000. But also 0.3 degrees warmer March and 0.6 degrees warmer October than the last 10 years. In the statistics, we look at also shorter winters, which is the new normal.
How skilled are the long-term models for the winter season? If we look at the ECMWF seasonal models for winter (December, January, and February), based on a forecast made here in October, the correlation coefficient for the temperature at 2 meters for the continent and UK is 0.2 or lower, which is a low correlation and thus not much of a value in a seasonal temperature forecast. As it comes to Scandinavia, most of the region shows a better correlation coefficient with the seasonal models – up to 0.6 – and thus, on average, somewhat better value in the seasonal forecast.
The quality of a seasonal forecast is also dependent on the season. Summer months are e.g. easier to predict in Southern Europe than in Scandinavia.
At the moment, we see colder than normal weather across Europe, and this is most likely to continue 10 days longer before milder airmass covers the continent. For Scandinavia, it can continue till Christmas. Looking at winter 2022/2023 with both a statistical approach and seasonal forecast predictions at hand, it seems like yet another mild winter for Scandinavia, with 0.5 to 1.5 degrees above normal (1993-2016).
Source: ConWX PPA Analyzer and www.ECMWF.int