EU Heatwave Study: Models Don’t Fit, Data Not As Expected—Must Be Climate Change!

The long hot summer of climate change/heatwave propaganda continues with the release of yet another ‘study‘ of European temperature extremes, this time by the Swiss Federal Institute of Technology.

As usual, the story is splashed all over the media that Europe is warming even faster than models predicted.

So it’s just the usual ‘it’s worse than we thought’ rubbish.

However, in order to understand the uncertainties and the limitations inherent in this supposedly ‘confident’ scientific assessment of European extreme temperature trends you have to actually read the paper, not rely upon the low-info alarmist hogwash splashed across the usual mainstream media outlets.

This will give you a flavor of what’s being reported:

Over the past seven decades, the number of extreme heat days in Europe has steadily increased, while the number of extreme cold days has decreased, according to new research. Alarmingly, this trend is happening at rates faster than those proposed by climate models.

For most Europeans, this new study will hardly come as a surprise. This summer, for example, temperatures in southern France reached a record 46 degrees Celsius (114.8 degrees Fahrenheit), with similar temperature extremes happening at other locations on the continent.

Indeed, Europe is getting progressively hotter, and the data bears this out. What’s disturbing, however, and as new research published today in Geophysical Research Letters points out, this warming trend is occurring faster than the projections churned out by most European climate models.

And as the new paper also notes, the observed increases in temperatures “cannot be explained by internal variability.” In other words, this warming trend is the result of human-caused climate change.

Take home message: OMG Europe’s burning up much faster than we expected! The models can’t explain it. Natural variability can’t explain it. Therefore it must be a man-made climate crisis! PANIC!!

Lead author of the study, Ruth Lorenz, is quoted as saying:

“Even at this regional scale over Europe, we can see that these trends are much larger than what we would expect from natural variability. That’s really a signal from climate change.”

Really? Much larger than she would expect from natural variability? The obvious thing to do then is to go look at the basis of that claim.

Before we delve into the study itself though, I’ll remind you of what I’ve already said in a previous post on European heatwaves:

So, what the authors are saying here is that there is no trend in wave 7 phase-locking events which would give rise to particularly hot summers in western Europe over the period of satellite observation, but that there is a very significant increase in phase-locking events after 1999, which has resulted in a number of hot summers in Europe, including the very hot summer of 2003 and the memorable hot dry early summer of 2018 in northwestern Europe.

What they say is that natural multidecadal variability may be the cause behind this change but that it’s plausible it may be due to Arctic warming, which implies anthropogenic causes (though note that this is only the case if such warming is attributable largely to anthropogenic causes).

So what we have here is the high likelihood that multidecadal variability is dictating the probability of circulation patterns occurring which would result in extreme heatwaves in Europe like 1947, 2003, and 2018. But dynamics are only one side of the equation. What about absolute temperatures?

The conclusion is plain: summers in Europe have got warmer due to the secular trend in warming experienced by most of the globe, but the probability of extreme summer heatwaves occurring due to particular circulation patterns shows no overall trend in the 20th century but is most likely controlled by a multidecadal variability (the AMO).

So, looking at what happened during 1947, and comparing that to what’s been happening in the 21st century, internal variability influencing atmospheric circulation patterns is a very likely direct cause of the extreme temperatures we have seen recently, but Lorenz states that internal variability can’t explain the observed trebling of extreme heat days since 1950 . . . . there’s a clue there!

Land-use changes and Urban Heat Island (UHI) may also have contributed to ‘record’ hot temperatures.

The secular global warming trend has contributed of course, but to suggest that it is the only thing which can explain the “disturbing” increase in hot weather in Europe is misleading at best, fraudulent at worst.

In another recent post, I pointed out what’s been said by Geert van Oldenborough re. the observed increase in temperature extremes in Europe:

As part of the analysis, the authors also looked at how extreme the temperatures seen during the July heatwave were in comparison to those seen in the past.

The authors find that the temperatures seen during this heatwave were around 3C higher than they would have been in 1900.

This is double the heatwave temperature increase expected by climate models – which are used to make projections about future climate change, van Oldernborgh says:

“The models only predict that heatwaves get warmer at about 1.5C per degree of global warming. So for every degree of global warming, they predict that heatwaves get 1.5C hotter – a little bit faster but not really exceptional.”

The world has seen around 1°C of global warming so far – meaning that the models would expect heatwaves to be around 1.5C hotter today than in pre-industrial times. However, temperatures during this heatwave were actually around 3C warmer, he says:

“We really need to do a lot more serious research than we can do within one week to look at why there is such a big discrepancy between the observed trends and the modeled trends.

Well, it seems Lorenz has solved it: models are wrong, observations are ‘way too hot’, internal variability can’t explain it, so man-made climate change must be much worse than we thought! Simples.

Let’s take a look at her and her colleagues stunning new insight into European heatwaves, shall we? The first sentence of the introduction and we are already into the sales patter:

Human activities are estimated to have caused approximately 1.0 °C of global warming above preindustrial levels (Masson‐Delmotte et al.

The reference is IPCC SR15 SPM.

I’ve already examined this attribution claim here; it directly conflicts with what is said in the last IPCC assessment report, AR5. Not an auspicious start.

The authors are already making claims which are not supported by even the anthropogenic climate change consensus.

Heat and Humidity

After stating that,

We here test for a large network of stations whether changes observed in extreme temperature are only due to a shift in mean temperatures toward a globally warmer climate or if there is also a change in the variability of the temperature distribution.

Trend detection of extremes has mostly been done at global to hemispheric scale. At regional to local scale, internal variability can strongly offset or amplify local to regional trends in extremes over several decades (Fischer & Knutti, 2013; Perkins & Fischer, 2013).

We here test whether we can detect a climate change signal at stations across Europe. In order to minimize the effect of internal variability, it is essential to aggregate across large regions and analyze as long periods as possible.

The authors then explain that they are using a dataset which goes back only as far as 1950:

We use daily mean (TG), daily maximum (TX), daily minimum temperature (TN), and daily relative humidity (HU) from the European Climate Assessment & Dataset (ECA&D, Klein Tank, et al., 2002).

This dataset provides quality‐controlled station data for around 4,000 stations in Europe, the exact number depends on the variable. Data are available from 1950 to present, we use data until October 2018.

The bright sparks among you will note that this is three years after 1947 when Europe experienced a major heatwave analogous to and in many cases even worse than 2003, which, as we have seen, is likely attributable to a repetitive dynamic circulation pattern (multidecadal variability).

The initially promise-sounding 4,000 stations are also reduced to just 1,000, covering the entire continent of Europe:

These constraints reduce the data set to around 1,000 stations for temperatures and even less for humidity (∼440).

So, the aspiration of robustly analyzing large regions over sufficiently long periods in order to eliminate the influence of natural variability turns out to be just that – an aspiration.

This is the basis of the headline claim in all those media stories:

Days with extreme heat (TX > 99th percentile), as well as extreme heat stress (WBT > 99th percentile), have at least tripled over the period 1950–2018. On average across Europe (EUR region), they increased from around 2 days/year in 1950 to about 6 days/year in 2018 as estimated from a linear trend (Figures 1a and 1c).

TX is derived from 1000 stations. ‘Heat stress’ (WBT=wet bulb temperature) is however derived from a much smaller number of stations mainly in western Europe, so can hardly be considered to be representative of the continent as a whole:

Note that the WBTx trends are limited to a much smaller network of stations mostly in western Europe that provided the necessary humidity measurements.

Testing for Natural Internal Variability

The influence of natural variability upon the data was not assessed with reference to any physical/meteorological parameters – as was more convincingly done by Otto et al using reanalysis data encompassing 1947 – but by using statistical analysis only, namely the block bootstrapping method.

This, apparently, allowed the authors to estimate the changes in the mean and variance of the data and assign a significance beyond the mere chance of the magnitude of the change in variance (i.e. extremes), which they then attributed to climate change in the absence of being able to explain it otherwise.

Science!

Being no expert in statistical analysis techniques, I’m not in a position to professionally critique the methods they used to quantify internal variability but the obvious points are: their data is truncated at 1950, after the very significant Central European heatwave of 1947, and their analysis contains no reference whatsoever to atmospheric dynamics and multi-decadal variability in atmospheric circulation patterns, beyond the brief mention that such variability “can strongly offset or amplify local to regional trends in extremes over several decades.”

They then go on to dismiss this possibility using limited data and probably a rather limited statistical analysis of that data. Again, to quote Otto et al:

The science of extreme event attribution has rapidly expanded in recent years, with numerous studies dedicated to determining whether and to what extent anthropogenic climate change has increased the likelihood of specific extreme weather events occurring.

However, the majority of such studies have focussed on extreme events which have occurred in the recent past (usually within the past 10 years) while minimal research efforts have considered the multitude of high-impact extreme climatic events which occurred throughout the instrumental record.

It very much looks like this latest paper from Lorenz et al is another of the variety explicitly focused on more recent extremes, which ignores high impact extreme events (and their dynamical causes) before 1950.

It’s Not ‘Worse Than We Thought’

Even their statistical analysis of the impact of internal variability, limited as it is, appears to be less than it is cracked up to be in the press and by Lorenz herself.

As far as I can tell, the extreme heat findings were significant only for Central Europe (not Western Europe or the Med) and even in the case for central Europe, all they demonstrated is the increase in extremely hot days already identified by van Oldenborough above when he comments that “the models only predict that heatwaves get warmer at about 1.5C per degree of global warming.”

I quote from Lorenz et al:

We find that in CEU heat extremes warmed more than the summer mean at about 85% of the stations and with a median warming difference of 0.14 °C per decade (Figure 3a).

This implies that across CEU [Central Europe] heat extremes have warmed by about 50% more than the mean. On the other hand, the warming of heat extremes and mean is about the same in NEU and MED (difference in trends between TXx − TG_JJA not significantly different from zero, p-value >0.05, Figures 3c and 3e).

Hence, it is only in CEU that the warming in the extremes was larger than in the mean in this data set. This is remarkably consistent with projections of heat extremes both in many global climate models (GCMs) and RCMs that are projected to be amplified due to enhanced variability across central Europe but not necessarily over southern and northern Europe (Cattiaux et al., 2013; Fischer & Schär, 2009; Fischer et al., 2014; Orlowsky & Seneviratne, 2012; Seneviratne et al., 2006).

It has been suggested that the increase in variability results from (a) land surface feedbacks, which are particularly relevant for central Europe, a transition region between a wet regime in the north and a dry regime in the south (Cattiaux et al., 2013; Fischer & Schär, 2009; Seneviratne et al., 2006), (b) the fact that the typical source regions of warming air advection in southern and continental eastern Europe warm more than the source regions of cold‐air advection in northern Europe and in the Atlantic (Holmes et al., 2016), and (c) that the warming is amplified due to fewer clouds and higher incoming shortwave radiation (Tang et al., 2012).

So in other words, it’s not really ‘worse than we thought’, it’s just as models predicted and it’s not really a big deal and it’s only central Europe which is really affected and this because of the possible interplay of factors other than climate change.

So much for the magnitude of the increase in extremes (though note that the Lorenz et al data ends in 2018, so doesn’t include summer 2019).

The authors and the media make a big deal about the frequency of extremely hot days, which they say have increased from 2 annually to 6 annually, in particular, the frequency of extreme heat stress days (humidity included, which latter result they derive only from a limited number of stations in western Europe).

Is that a big deal? An increase from 2/365 (0.5%) to 6/365 (1.6%), i.e. from tiny to very small? Hardly a ‘crisis’.

The Lorenz paper concludes:

We detect a clear signal from climate change in the trends in extreme temperature and heat stress based on observational data that cannot be explained by internal variability. We demonstrate that on average across Europe the number of days with extreme heat and heat stress has more than tripled from 1950–2018 from less than 2 days to more than 6 days per year.

Changes are consistent across subregions, daytime and nighttime temperatures, and across different percentile thresholds. Likewise, the intensity of daily (TXx) to weekly hot extremes has increased by about 2.3 °C (median across Europe) from 1950–2018.

The median rate of change of 0.33 °C per decade is larger than the global average temperature warming of about 0.2 °C per decade today (Masson‐Delmotte et al., 2018).

In CEU, the subregion with the strongest intensification, hot extremes have warmed about 50% more than the corresponding summer mean temperatures while in NEU and MED hot extreme and mean trends are similar.

This contrasts strongly with the media hype about this study. For instance, Newsweek says:

Parts of Europe are warming faster than climate models have predicted, with the continent experiencing a spike in the number of days with extreme heat, scientists have said.

The Fail says:

Climate change is increasing the number heatwaves in Europe faster than computer modeling has previously predicted, new research suggests.

It’s not true. All the study demonstrates is that the increase in the magnitude of extremely hot days is consistent with model projections (in just one part of Europe) and that the frequency of extremely hot days and heat stress days has trebled from 2 per year in 1950 to 6 per year in 2018.

Furthermore, using a limited statistical analysis of limited data, the authors ‘demonstrate’ that these changes cannot be attributed to natural internal variability – whereas we have seen that they can, in large measure, indeed be attributed to multidecadal changes in atmospheric circulation patterns.

The media, in cahoots with the lead author of the study, give the impression that this result is some sort of revelation, a wake-up call, is “disturbing” and indicative of a crisis, whereas in actual fact it is just bog-standard ‘science’ involving statistical analysis of temperature data.

Nothing new here: same old, same old, but never let it be said that the media, climate alarmists and climate scientists miss an opportunity to turn routine research into a crisis and thence turn a crisis into a full-blown drama.

Read rest at CliScep