European Heat Waves

Those of us who follow such things, know that Europe has been hit with two killer heat waves since the year 2000. Super-killer heat waves. The first, in 2003, covered much of western Europe and led to about 70,000 deaths due to heat stress. The second, in 2010, hit Russia and killed another 50,000 while destroying a third of Russia’s wheat crop (Russia is one of the largest wheat exporters in the world).

But even those of us who do follow such things, may have missed the fact that Europe has actually had three super-killer heat waves since the year 2000.

I decided to explore sustained high heat in data for daily high temperature from the European Climate Assessment and Dataset network, and I did so by exponentially smoothing the temperatures on a 20-day time scale. That way, very high values indicate high heat over a sustained period of time; perhaps that this is a good way to identify the super-killer heat waves.

Then I scanned each year from 1950 through 2019 for the annual maximum 20-day exponentially smoothed temperature. I included only years which had data for at least 150 of the 153 days in the months May through September, and only included stations including at least 65 of the 70 years from 1950 through 2019.

The super-killer heat waves tend to stand out in plots of these values. Typical for the 2003 heat wave is Toulouse-Blagnac, France; the extremity of the 2003 heat wave is apparent:

The 2010 Russian heat wave is well captured by the result from Moscow:

A striking example is from Kursk:

Last year brought new record highs to many locations in northern Europe, like Gotska in Sweden:

All told, 810 station records met my selection criteria. For each I recorded which year had the highest value of all, and found that the three years with the most peaks have all been this century. The biggest one of all, with the most stations reporting their highest 20-day exponentially smoothed temperature, wasn’t the super-killer Russian heat wave of 2010 or the super-killer heat wave of 2003. It was last year.

There’s a lot to do before drawing firm conclusions, but the preliminary indication is that heat waves in Europe aren’t just getting hotter and lasting longer, they’re covering a larger area too.

Thanks to Colin Letham, Tim Baumann, Aart Shakel, James Vogan, Marianne Soby, and Robert Day for kind donations to the blog. If you’d like to help, please visit the donation like below.

This blog is made possible by readers like you; join others by donating at My Wee Dragon.


17 responses to “European Heat Waves

  1. Last year was very hot in the Netherlands. Records of 75 years ago were finally broken. It was also a killer heatwave if you look at the statistics of death.

  2. Can you also track down mortality data?
    A powerful way to emphasize to the public that AGW is not some amorphous, vague threat in the future is that it is killing people here and now.

  3. “For each I recorded which year had the highest value of all”
    Is this really a valid way to identify heat waves? Sure, it identifies the hottest individual (smoothed) value for any given station, but many stations will not have been affected by any heat waves within the historical record. Since the long-term trend is upwards, then all of the non-heat-wave stations will inevitably record their highest values in the last few years.

    A more reasonable definition of heat “wave” would use some kind of outlier detection function to find discrete peaks that rise above the local background.

    • @PIIE2,

      Definitions, definitions …

      Heat wave.

    • pjie2

      “A more reasonable definition of heat ‘wave’ would use some kind of outlier detection function to find discrete peaks that rise above the local background.”

      I agree.

      If you look at a descending sort of the absolute temperatures in Europe since 1880, you lots and lots of measurements provided by Spanish stations, and you see at position 9:

      SP000008410 CORDOBA_AEROPUERTO ; 2003 08 01 ; 46.2 °C

      This was not necessarily a consequence of the heat wave which occurred in France and Italy at that time: the average of the top 50 temperatures for Cordoba namely is 44.6 °C.

      Maybe we should restrict our investigations to those measurements meeting the two requirements:
      – it is a high temperature, above a common minimum of x °C;
      – it shows a departure of y °C over the common mean.

      The difficulty is to determine a meaningful value for this ‘y’…

  4. Nous sommes tous les francais.

  5. The heat wave this year was really hot. Temperature records in Germany, The Netherlands, Belgium, France and the UK were shattered, not just broken, but shattered.

  6. Language nitpick: “Gotska Sandön” means “The sandy island near Gotland”. Just calling it “Gotska” is very wierd. It’s like calling “French Antilles” just “French”.

    • Thomas P

      !ndeed, you are right. In the GHCN daily station file, we can see:

      SWM00002584 58.4000 19.2000 11.0 GOTSKA SANDON
      SWM00002589 58.3942 19.1975 16.0 GOTSKA SANDON A

  7. Lovely work, thank you!

    Questions for all: To what extent were heat waves expected due to global warming? and are heat waves more intense and happening sooner than expected?


  8. One reason why last year got less attention was that the most abnormal heat was focused on Scandinavia, much of it in the more sparsely populated northern parts. (This year’s western European heatwaves were intense but relatively short, so probably won’t show up strikingly on a 20-day average).

    Some countries are better than others at tracking excess mortality from extreme heat. I’ve seen estimates from France from this year’s heatwave (around 500 in June and 1000 in July), but not yet from any other countries.

  9. 1. To conclude that heat waves are affecting larger area would require accounting for station density, not just the number of stations. Is station density or area per station included in the analysis?

    2. The criterion seems to allow only one heat wave per location, thus under-representing locations with multiple heat waves. A measure that related the high temperature anomaly over the 20 day period relative to the location average and counting a positive anomaly of some threshold, e.g. +4C might be more informative. +4C does not sound like mucn, but the values being compared are 20-day averages.

    3. 20-day average seems overly long. A heat wave of 10 days could get washed out by lower than normal temps. on the other days. I realize you need a sequence of days to filter out noise, but similar analysis using 10 or 14 day sequence could reveal more heat waves that were hidden in the 20-day average.

    As always, interesting and insightful work Tamino.
    FYI – 2019 average NASA GISSTemp is running just a tiny bit below 2016 record. This is especially notable because the 2016 record was the result of a monster El Nino. But ONI3.4 temperatures, i.e. El Nino influence, have only been tepidly warmer than average in 2019. Thus 2019 is near a new record global average annual land and sea surface temperature even without a strong El Nino influence. Not good. If this keeps up, once the solar cycle bottoms out ca. 2022, every year thereafter may have a >50% chance of being a new record (subject to ENSO influence varying above or below zero).

  10. I’m wondering about Tamino’s stat concerning heat waves in Europe.

    Simply because it is hard to understand, for a person living in Germany, how Europe should have experienced stronger heat waves in 2018 and 2019 than in 2003, when e.g. France had for these years the following mortality rates for excessive heat:
    – 2003: 15,000
    – 2018: 66 (!!!)
    – 2019: 1,500

    I’m a perfect nobody in statistics, a field where Tamino has often enough shown his outstanding knowledge and experience. So it is very difficult for me to contradict him.

    Nevertheless, I read in his head post above:
    For each I recorded which year had the highest value of all…

    Do I read right? Does this mean that all highest temperatures are considered, even if they are at say + 15 °C? If I have read right, so may I ask: what does these highest values have to do with heat waves?

    Should not we look at those maxima instead which have exceeded a fixed minimum?

    I collected for Europe (35 states plus Russia’s European part till the Ural mountains) all valid GHCN daily reports above 30 °C from all available stations, and plotted, for 2003, 2018 and 2019, from May, 1 till September, 30
    – the number of stations having reported 30 °C or more:

    – the day’s average temperature of these reports:

    The graphs look nice but aren’t very informative, as usual for dot plots. Here, numbers will do a better job. We see below, for the three years:
    – the total number of reports over 30 °C;
    – the average number of stations per calendar day;
    – the average temperature over the five months.

    Here is ‘my’ podium:
    (1) – 2003: 30,931 / 252 / 40 °C;
    (2) – 2018: 26,539 / 216 / 39 °C;
    (3) – 2019: 20,189 / 164 / 39 °C.

    While the average temperature over the May/Sep period is nearly the same for the three years, it seems that in the sum, 2003’s heat lasted longer than the other years, with 2018 ahead of 2019.

    This however cannot explain the mortality factor for 2003 being 10 times higher than in 2019, let alone why it was over 200 times higher than in 2018.

    There must have been, near the heat, some other relevant factors, e.g.
    – dry, windless weather patterns accentuating the heat effect;
    – but also the fact that of the 15,000 mostly aged persons who died 2003 in France due to the heat wave, a great number of these died lonely, without any help, neither from their family nor from the staff of many small retirement homes.


  11. One complication when you try and assess extreme events by impact, rather than by physical measures such as temperature or precipitation, is that comparison is obscured by the actions of society in recognizing and mitigating risk. There is a natural tendency to assume that everything is getting ‘worse’, but our ability to forecast, and plan pre-event to minimize impacts – for instance identifying vulnerability in advance – means that modern society, especially in rich countries, is much better insulated from shocks than we were in the past. None of which is justification for complacency.

    • @Astringent,

      … [O]ur ability to forecast, and plan pre-event to minimize impacts – for instance identifying vulnerability in advance – means that modern society, especially in rich countries, is much better insulated from shocks than we were in the past. None of which is justification for complacency.

      Sure, some, but our ability to plan is overplayed, since we don’t do it rationally in most countries, to wit, placing petroleum refineries in floodable hurricane prone areas. Also, our economy and our food supply is presently dependent upon extended supply chains. The aphorism is that most of the food eaten today by a farm family comes from more than 1000 miles away. And while we can shield ourselves with things like emergency backup generators — as we have — there’s no guarantee that the propane supply for running them in an extended regional disaster will be provided.

      Accordingly, I would say although we are wealthier, our risk cross section is much bigger than poorer countries with more self-sufficient people. Moreover, if infrastructure is compromised, sources of wealth, namely work, might disappear or be furloughed or be difficult to get to.

      • All good points. Yet I don’t think that everything ‘we’ve’ done has been counterproductive. My first suspicion that the reduced heatwave mortality may be due to increased access to cooled spaces, either as a function of public health initiatives which I’ve seen reported–‘cool shelters’ run by municipalities, etc.–or as a function of increased deployment of air-conditioning.