Close this search box.

UN IPCC’s New ‘Hockey Stick’ Temperature Graph

IPCC’s New “Hockey Stick” Temperature Graph


The United Nations’ Intergovernmental Panel on Climate Change (IPCC) published their latest assessment report (AR6) in 2021. In 2023, the Clintel Foundation published a report which criticizes AR6.


Clintel is short for Climate Intelligence, and the Clintel Foundation doesn’t think there’s a climate emergency. Overall, Clintel’s main criticism is that the IPCC hasn’t reviewed the scientific literature in an objective way, as is their stated mission.1)

This article focuses on the topic of chapter 2 of the Clintel report: The Resurrection of the Hockey Stick.

What is a “hockey stick” temperature graph?

The image below shows a “hockey stick” temperature graph for the past 1000 years. This particular version is from the Summary for Policymakers of the IPCC’s third assessment report from 2001, and it applies only to the Northern Hemisphere.

The graph is called a hockey stick graph because the shape somewhat resembles an ice hockey stick with a long, flat “shaft” and a big “blade” shooting up at the end. If there had been more variability in the temperature before the “blade” – meaning that the “shaft” wasn’t flat – then it wouldn’t be a hockey stick graph.

Short summary

A “hockey stick” temperature graph has made a comeback in the latest IPCC report from 2021 (AR6), after being absent in the fourth and fifth reports.

Unlike the previous two reports which included multiple temperature estimates, AR6 relies on just one temperature estimate for the past 1000-2000 years, sparking criticism.

Moreover, this single estimate has itself been criticized. The perhaps most important issue relates to a method that incorrectly generates hockey stick temperature graphs from non-hockey stick data.

In general, the quality and amount of data that can tell us something about past temperatures quickly decline when we go further back in time. Based on the currently available data, it might therefore be impossible to know whether the global temperature of the past 1000-2000 years has a hockey stick shape or not.

Why is the hockey stick important?

A hockey stick temperature graph is convincing evidence that human emissions of greenhouse gases have a big impact on the climate, and according to IPCC, humans are responsible for roughly 100% of all warming over the past 150 years.

For Clintel, who doesn’t think there’s a climate emergency, it would make their argumentation easier if past temperature variability were higher, as more of the recent warming could then be attributed to natural causes. They write that if past temperature variability has been high, then “the temperature rise of the past 170 years would have to be shared [between] anthropogenic and natural causes.” 2)

However, this way of thinking is a fundamental mistake, according to climate scientist Ulf Büntgen. He argues that a high pre-industrial variability would mean that the atmosphere is more sensitive to changes in e.g. greenhouse gases, than if pre-industrial temperature variability has been low.3)

In any case, what’s most important is, of course, to get a correct estimate (or “reconstruction”) of past temperatures — whose shape may or may not resemble a hockey stick. Let’s explore!

From Chapter 2 of the Clintel report:

One of the big surprises of the IPCC’s AR6 report was the comeback of the so-called “hockey stick“. This term refers to the northern hemispheric and global temperature development of the past 1000-2000 years. More than two decades ago, Mann et al. (1999) published a reconstruction in which the temperatures of the pre-industrial period 1000-1850 AD appear rather flat and uneventful (the “shaft” of the ice hockey stick), followed by a fast and allegedly unprecedented warming since 1850 (the “blade”). The hockey stick became world famous because it was featured prominently in the Summary for Policymakers (SPM) in the IPCC’s 3rd Assessment report [see image at the top of this article].

The hockey stick graph received a fair amount of criticism, and Clintel explains that the hockey stick graph was subsequently corrected — resulting in a temperature reconstruction with more pre-industrial variability, and thus without a hockey stick shape. IPCC’s fourth and fifth assessment reports did not include a hockey stick temperature graph, but in the most recent assessment report (AR6), the IPCC once again presents a temperature reconstruction with low pre-industrial variability – a new hockey stick:

Image from the Summary for Policymakers (page 6) in IPCC’s latest assessment report.

The new hockey stick temperature reconstruction is based on the work of PAGES 2k, the flagship of the PAGES (Past Global Changes) project. Clintel writes:

The PAGES 2k group is specialised in climate reconstructions and back in 2013 was comprised of the majority of all active paleoclimatologists. In 2019, PAGES 2k published a new version of the temperature development of the past 2000 years (PAGES 2k Consortium, 2019 4) [hereafter “PAGES 2k 2019”]). Surprisingly, it differed greatly from the predecessor version. Even though the database had only mildly changed, the pre-industrial part was now suddenly nearly flat again. The hockey stick was reborn[.]


Evidence suggests that a significant part of the original PAGES 2k researchers could not technically support the new hockey stick and seem to have left the group in dispute.

It would be very interesting to read more about this. Unfortunately, Clintel doesn’t provide a reference.5) Anyway, Clintel continues:

Meanwhile, the dropouts published a competing temperature curve with significant pre-industrial temperature variability (Büntgen et al., 2020)[.] On the basis of thoroughly verified tree rings, the specialists were able to prove that summer temperatures had already reached today’s levels several times in the pre-industrial past. However, the work of Ulf Büntgen and colleagues was not included in the IPCC report[.]. [Emphasis added]

Did they really prove it? Probably not.

This is figure 4 from Büntgen et al. 2020, where they compare their own results with other temperature reconstructions covering the past 1000-2000 years, including PAGES 2k 2019. The PAGES 2k temperature reconstruction is the white line in the middle, the gray shading being the uncertainty range. Büntgen et al.’s results are the lines denoted EA and EA+. EA is Eurasia (above 30 degrees north), and EA+ additionally includes the North Atlantic region. The black line on the right side is thermometer data for June, July and August (JJA) in the 30-70 degrees north latitude band.

PAGES 2k 2019 is a global multi-proxy temperature reconstruction. Büntgen et al. 2020, on the other hand, applies only to the extra-tropical Northern Hemisphere, uses only one proxy-type – tree rings, and only looks at summer temperatures, not yearly temperatures. Still, it may actually make sense to compare Büntgen et al.’s reconstruction with PAGES 2k 2019, but that’s not immediately obvious, and a discussion of this by Clintel could have been beneficial.

What’s a multi-proxy temperature reconstruction?

In the previous paragraph, the term multi-proxy temperature reconstruction was used. A proxy — in this case a proxy for temperature — is an alternative to a direct temperature measurement with a thermometer. We’ve only had thermometers for a few hundred years, so we have no direct measurements of temperatures before that. To be able to estimate past global temperatures, scientists thus have to use less accurate alternatives instead of direct temperature measurements. These alternatives are called proxies. One proxy that can be used for temperature is tree ring width. Trees will typically grow more in a warm year than in a cold year, so for certain temperature sensitive trees, bigger tree rings could mean higher temperature.

multi-proxy temperature reconstruction is an estimate of past temperatures that’s based on multiple proxy types. Tree ring width is typically one of several proxy types used in a multi-proxy temperature reconstruction.

The reason it may still make sense to compare Büntgen et al. 2020 and PAGES 2k 2019 is that PAGES 2k 2019 is also biased towards the Northern Hemisphere, towards summer temperatures, and towards the use of tree ring proxies:

PAGES 2k 2019 used 257 6) proxies out of a total of 692 proxies in the PAGES 2k database. 210 of these are high-resolution proxies with yearly resolution. In the below image, which is parts A, B and C from Figure 2 of Anchukaitis and Smerdon 2022, we can see
A) that most of the proxies with yearly resolution are based on trees (tree rings),
B) that the average latitude is not the equator, but 47 degrees north, and
C) that most of the proxies capture summer temperature (the growing season for trees in the extra-tropical Northern Hemisphere), not yearly temperature.

Criticism of PAGES 2k 2019 and IPCC

Anchukaitis and Smerdon 2022, where the above image is taken from, is an article that reviews “the strengths and limitations of existing global and hemispheric paleoclimate temperature reconstructions and highlight[s] likely sources of [existing] uncertainties, all in the context of [IPCC AR6].” They’re quite critical of the IPCC’s new hockey stick graph.

One thing they criticize IPCC for, is that the IPCC did not consider a variety of multi-proxy temperature reconstructions, which they had done in their previous two assessment reports (AR4 and AR5):

Despite the attempts in AR4 and AR5 to reflect uncertainties across multiple reconstruction efforts and to represent time-dependent uncertainties as they expanded back in time, these efforts were surprisingly abandoned in the most recent AR6 Working Group I (WG1) report in favor of a single ensemble-based reconstruction of global temperature with relatively static uncertainty bounds over the [past 2000 years.] […] The most recent assessment […] is thus a turn away from the attempts in previous reports to provide a full accounting of uncertainty in reconstruction efforts […], an incomplete representation of forward progress in both understanding and quantifying disagreement in temperature reconstructions of the [past 2000 years], and is an unnecessary return to a singular representation of large-scale temperature estimates that span all or part of the last several millennia.

Another point made by Anchukaitis and Smerdon, which they also touched upon in the above quote, is that the uncertainty range in the new hockey stick graph is far too narrow, especially when we get further back in time:

[T]he estimated uncertainties for this reconstruction used in AR6 only reflect the methodological differences as applied to the PAGES2k dataset at decadal and longer time scales. Much larger latent uncertainties are almost certainly present due to the change through time in proxy availability, sensitivity, and spatial distribution[.] Because the consequences of these uncertainties are poorly represented in AR6, which is intended to reflect an assessment of the current state of the science, the report fell short in its representation of what we know and what we have learned about Common Era [=past 2000 years] temperatures over the last two decades. [Emphasis added]

Clintel, of course, also criticizes PAGES 2k 2019:

Like its predecessor, the new hockey stick by PAGES 2k 2019 is based on a large variety of proxy types and includes a large number of poorly documented tree ring data. In many cases, the tree rings‘ temperature sensitivity is uncertain. For example, both PAGES 2k Consortium (2013) and PAGES 2k Consortium (2019) used tree ring series from the French Maritime Alps, even though tree ring specialists had previously cautioned that they are too complex to be used as overall temperature proxies (Büntgen et al. 2012; Seim et al., 2012).

In contrast, Büntgen et al. (2020) were more selective, relied on one type of proxy (in this case tree rings) and validated every tree ring data set individually. Their temperature composite for the extra-tropical northern hemisphere differs greatly from the studies that use bulk tree ring input.

In some cases, PAGES 2k composites have erroneously included proxies that later turned out to reflect hydroclimate and not temperature. In other cases, outlier studies have been selected in which the proxies exhibit an anomalous evolution that cannot be reproduced in neighbouring sites (e.g. [Medieval Warm Period] data from Pyrenees and Alboran Sea in [PAGES 2k 2013]) (Lüning et al., 2019b). Outliers can have several reasons, e.g. a different local development, invalid or unstable temperature proxies, or sample contamination.

However, starting from the third sentence, this text is virtually copy-pasted from a study by Lüning and Lengsfeld (2022). Clintel could have at least added a reference to that study. Also, it’s unclear whether the last paragraph applies to PAGES 2k 2019 or just to earlier PAGES 2k versions. If it does indeed apply to the 2019 version, then the Clintel report would have benefited from documenting it.

A new temperature reconstruction that includes short-term trends

As noted, PAGES 2k 2019 is a multi-proxy temperature reconstruction (although it may be biased towards tree-rings), while Büntgen et al. 2020 only uses tree rings. All things equal, multi-proxy reconstructions should be preferred over single-proxy reconstructions; according to Büntgen et al. 2022, there is “community-wide agreement that multi-proxy compilations are the most appropriate methodology to climate reconstructions“.

One reason that multi-proxy reconstructions are better is that some proxy types can reveal short-term temperature variability, but not long-term variability, while others can reveal long-term variability, but not short-term variability.

And while “tree rings are excellent at capturing short frequency variability, they are not very good at capturing long-term variability.” (Fundamentals of Tree-Ring Research, James H. Speer, 2009) This is also pointed out by Büntgen et al. 2020 itself, the study that used tree-rings only:

[Tree ring width] is particularly limited when reconstructing the amplitude and duration of climatic extremes[.]


While we have no doubt about the timing of past summer cooling in our reconstructions, accuracy of the temperature amplitude remains somewhat challenging.

Since some proxy types are good for capturing short-term trends, while others are good for capturing longer-term trends, the best way to make a temperature reconstruction should be to use proxy-types that capture long-term variability for the long-term trends, and then use e.g. tree rings to show the short-term variations on top of the long-term trends.

Titled “A frequency-optimised temperature record for the Holocene“, Helen Essell et al. 2023 is a study that combines short-term and long-term proxies in this way. It’s the first study that attempts to “present […] temperature variability on interannual timescales over the past 12 000 years.” Previous studies have had a much lower (worse) time resolution, and hence the temperature graphs from those studies have been smoother and with less variability.

However, while “interannual signals ([less than] 10 years) are best captured by tree-ring chronologies (wood)“, only 3 tree-ring proxies were included in the study. Unfortunately, the study doesn’t discuss uncertainties related to the low number of short-term proxies.

The study performs a global temperature reconstruction, but like most other “global” reconstructions, there is a bias towards the Northern Hemisphere and summer temperatures.

Also note that the study looks at the time period 12,000 – 0 BP, where BP means “before present” and “present” is defined as the year 1950. This is a longer time period than we normally talk about for hockey stick temperature graphs, which is 1000-2000 years. Hockey stick temperature graphs also normally include the time period after 1950, which is often represented by thermometer data in addition to proxy data. Helen Essell et al. 2023 does not use thermometer data.

Although the time period is different, it’s interesting to compare IPCC’s hockey stick graphs with the new temperature reconstruction from Helen Essell et al. 2023, which looks like this (where light green shows the uncertainty range):

If the results of this study are broadly correct, then pre-industrial temperature variability has been high, which would mean that hockey stick temperature graphs are not correct.

Since Helen Essell et al. 2023 is a new and novel study, it may or may not contain important errors, but their criticism of the IPCC is very relevant regardless; the IPCC should not compare long-term average temperatures with recent yearly (high) temperatures:

[W]e remain critical of the interpretation of the smooth trajectories of existing Holocene temperature reconstructions, which have influenced policy debate. For instance, the Sixth Assessment Report of the Intergovernmental Panel on Climate Change (IPCC), and its ‘Summary for Policy Makers’ and ‘Technical Summary’ compared recent annual extremes against past centennial averages. Such unequal comparison has the potential to mislead portrayal of the threat of global warming as low- and high-frequency records reflect different aspects and amplitudes of the Earth’s climate system. [References removed]