The Holocene Temperature Conundrum

By Andy May

This post has been translated in German by Christian Freuer here.

As my regular readers know, I’ve been writing about the Holocene, especially Holocene climate, for over ten years. My colleagues, Javier Vinós and Renee Hannon, have also written extensively on this topic. This fills a void left by the IPCC, which ignores the Holocene, a fact made very clear in my last two posts, here and here. Here I would like to review a very comprehensive and informative review article of what the authors, Darrell S. Kaufman and Ellie Broadman, call “The Holocene global temperature conundrum.” The paper was published in Nature February 15, just a little over a week ago. I was made aware of it by a reader who identifies himself as “Mike G,” he was also kind enough to send me a pdf of the article.

Kaufman and Broadman define the Holocene Temperature Conundrum in this way:

“A landmark reconstruction of GMST showed peak warmth during the mid-Holocene, when GMST reached about 0.8 °C higher than that of the preindustrial period. By contrast, early transient climate modelling showed that GMST was around −0.5 °C colder during the mid-Holocene compared with preindustrial [aka Little Ice Age] temperature, followed by continued warming. This discrepancy between the late Holocene global cooling trend inferred from proxy evidence versus the warming trend simulated by climate models is known as the ‘Holocene temperature conundrum’”

(Kaufman & Broadman, 2023)

Thus, the proxy evidence we have suggests that the early Holocene global mean surface temperatures (GMST) were warmer than the Little Ice Age (aka the preindustrial period) by about 0.8°C, but the IPCC/CMIP6/PMIP4 climate models suggest that the early Holocene was cooler, by 0.3 to 0.5°C than the Little Ice Age.

After exhaustively reviewing the evidence for temperature changes in the Holocene, Kaufman and Broadman conclude that global mean surface temperatures (GMST) were higher during the Holocene Climatic Optimum, and that subsequent global cooling took place over the following period often called the Neoglacial. Figure 1 is from my previous post, it provides reasonable regional temperature reconstructions for the Arctic, Northern Hemisphere, tropics, Southern Hemisphere, and Antarctica.

Figure 1. The regional temperature reconstructions shown are in 30° latitude slices, that is the Arctic (green line) is from 90°N to 60°N, the Northern Hemisphere (heavy black line) is from 60°N to 30°N, the tropics are from 30°N to 30°S, the Southern Hemisphere is from 30°S to 60°S and the Antarctic from 60°S to 90°S. The Holocene Climatic Optimum warm period, the Mid-Holocene Transition (MHT), and the Neoglacial Period are identified. Sources: (May, 2018, Chapter 4) and here.

As Figure 1 makes clear it is difficult to accurately and honestly represent Holocene temperature changes with a “global average.” Early in the Holocene the different latitude slices go in different directions and the Neoglacial Period cooling is not truly global, but mostly in the Northern Hemisphere. None-the-less, Kaufman and Broadman conclude:

“GMST probably peaked sometime late during the first half of the Holocene, at roughly 6.5 ka [4500BC], as evidenced by extensive proxy data and supported by theory and models. Proxy evidence reported in several studies indicates that GMST was roughly 0.5 °C higher during this millennial-scale period compared with 1850–1900, with most of the warming occurring at middle to high latitudes in the Northern Hemisphere.”

(Kaufman & Broadman, 2023)

Thus, the data runs opposite the models. When I was a computer modeler, this was the end of it, sorry buddy, you blew it. But we live in a different world today. The authors go on to say:

“On the modelling side, none of the 16 latest-generation global climate models that participated in the recent CMIP6-PMIP4, 6-ka experiment simulated GMST that exceeded their preindustrial control runs. GMST at 6 ka in these models averaged 0.3 °C cooler than the preindustrial period (Fig. 1b), arguing against a global HTM.”

(Kaufman & Broadman, 2023)

They provide a plot of the CMIP6 modeled temperature difference between the Holocene Climatic Optimum (HCO) and the Little Ice Age (LIA). It is our Figure 2.

Figure 2. The difference between CMIP6/PMIP4 climate model temperatures during the Holocene Climatic Optimum and the Little Ice Age. The models hindcast that the Little Ice Age was warmer by 0.3°C, on average, than the Holocene Climatic Optimum. This is against all proxy evidence and all historical records.

The map of climate model results in figure 2 is damning enough, but the histogram to the right suggests that the middle Northern Hemisphere latitudes were nearly the same in the HCO and the LIA! We know that cannot be correct. Numerous historical, proxy, and archaeological records from around the world show that the LIA was the coldest and most miserable time in human history in the Northern Hemisphere.

Kaufman and Broadman conclude:

“If our preferred interpretation [the HCO was warmer than the LIA] is correct and recent global warming was proceeded by a multimillennial global cooling trend, then this points to the need to improve our understanding of natural climate forcings and feedbacks, along with their representation in climate models. The very large seasonal and latitudinal changes in insolation caused by orbital forcing is the most likely instigator of climate feedbacks over the second half of the Holocene.”

(Kaufman & Broadman, 2023)

All I have to say is “hear! hear!” Climate models are clearly not modeling natural climate change accurately, especially not orbital forcing. If you cannot model natural climate change, you have no idea what the human influence on climate is. If you cannot model the Holocene correctly, you don’t understand climate change.

Works Cited

Kaufman, D., & Broadman, E. (2023, February 15). Revisiting the Holocene global temperature conundrum. Nature, 614, 425-435. Retrieved from

4.8 44 votes
Article Rating
Newest Most Voted
Inline Feedbacks
View all comments
February 25, 2023 2:32 am

There is evidence that the world was hotter during the Holocene Climate Optimum. The Sahara was wet as evidenced by all the ancient human wall drawings depicting animals, trees, people swimming etc. as well as the geological record, including all the sub surface aquafers.
The Historian Barry Cunliffe wrote “Britain Begins”. In this book he mentions how the British Isles were much warmer over 6,000 years ago. The evidence comes from remains of Flora showing that the tree line was several hundred metres higher than today – The Brecon Beacons had trees at the highest point. Also, in the north of Scotland including the Islands there are many abandoned Neolithic settlements. They were all abandoned around the time that the Holocene Climate Optimum came to an end.
In addition – about 30 years ago I watched a TV documentary about the demise of the Vikings in Greenland. The scientists involved used not only standard archaeological skills but also modern criminal forensics to re-examine. They could see that the flora had changed as well as the insects, and their conclusion was that the temperature dropped to the point where the Viking settlements were untenable. I can no longer find this documentary.

Mark Luhman
Reply to  sskinner
February 25, 2023 8:22 am

 “I can no longer find this documentary.” you can bet they deep sixed that, it will never see the light of day for as long as I am alive.

Reply to  Mark Luhman
February 25, 2023 9:50 am

It is somewhere in some library archive like microfiche that never got transferred to digital by choices of climate change advocates like in google.

Scarecrow Repair
Reply to  sskinner
February 25, 2023 8:50 am

I remember learning long ago, maybe even back in K-12 time, that Greenlanders used to raise cattle until it got too cold, and don’t today. Whether they don’t now because it’s too cold, there’s not enough land, easier shipping makes it unprofitable, or there’s some other reason, I do not know.

Reply to  Scarecrow Repair
February 25, 2023 12:22 pm

The IPCC declines the idea of a “green” Greenland during the early Middle Age’s Climatic Optimum with permanent Viking settlements thus showing a milder climate than today. However there were historically documented such settlements on the western coast with several villages and even a bishopric, before the climate progressively cooled and the glaciers began growing until they fully destroyed and erased them.

Steve Keohane
Reply to  Jack
February 26, 2023 5:55 am

In grade school back in the 50s, we were told the Vikings confused the placement of Iceland and Greenland, and gave them each other’s name.

Reply to  Steve Keohane
February 26, 2023 9:45 am

When Eirik (“the Red”) Thorvaldson was banished from Iceland he had the best part of 3 years to explore the land glimpsed many years earlier by a sailor driven west of iceland by storms. Eirik found a coast on the SW corner of Greenland that had a relatively mild climate , plenty of grazing for caribou (and therefore potentially for cattle) and an abundance of marine fish and mammals. Hence “the Green Land ” compared to the rocky , infertle part of Iceland to which he and his father had been exiled for the family trait of settling arguements by violent , sometimes fatal, methods. The settlements that he initiated eventually grew to a population of about 2000, with a monastry , nunnery , cathedral and about 280 farms.
As the Little Age developed from about 1200 , pastoral farming became more difficult and the analysis of their diet shows a greater dependence on a marine diet . They could have stayed and copied the Thule ( pre- Inuit ) way of life , but that would have meant abandoning the European , hierarchical society that they had been used to (one of the last pieces of documentary evidence from the East Settlement, in ca 1408 , is the execution of a slave for alleged rape). They simply left , to return to Iceland or norway or possible the Scottish Islands dominated by Norse immigration.

Ron Long
February 25, 2023 2:37 am

Thanks, Kip, good comments about the pesky Holocene maximum and the modelling efforts. Both because of the ENSO figure from your yesterday post and the obvious disconnect from the NH and ANT temperature between 7,000 and 4,000 BC in todays post, it sure looks like ocean currents are the major influences of world-wide climate conditions. The ability of ENSO to influence the weather everywhere in the world is impressive, and the current Bad Girl conditions have produced the worst wine grape harvest in recorded Argentina history.

Reply to  Ron Long
February 25, 2023 9:14 am

“Thanks, Kip, good comments about the pesky Holocene maximum and the modelling efforts.:

Who is Kip?

Ron Long
Reply to  Richard Greene
February 25, 2023 11:40 am

Sorry, Andy. Senior moment.

Hans Erren
February 25, 2023 2:56 am

During the holocene climate optimum there was a sea level high stand in The Netherlands and Suriname as indicated by synchronous largest extent of chenier formation.

George T
February 25, 2023 3:04 am

My complaint from the start. These GCM cannot model the past climate much less model the future. This challenge is simply ignored. The entire climate change narrative driven by CO2 is a fraud and I am seeing the results each day.

Reply to  George T
February 25, 2023 6:20 am

There’s a booster for that.

Michael S. Kelly
Reply to  George T
February 25, 2023 9:59 pm

It may be that GCMs cannot model the past because it is impossible for them to do so, even if they’re right. Even IPCC admits that the GCM’s system of coupled, non-linear partial differential equations is chaotic, and that one feature of a chaotic system is extreme sensitivity to initial conditions. Change them by the tiniest amount between two simulations, and you’ll get two entirely different results down the line. I saw one article on Judith Curry’s site that compared two runs of the same GCM with a trillionth of a degree difference in the temperature initial conditions, and they eventually diverged into two unrelated solutions.

We don’t know the current temperature at each of the hundreds of thousands of grid points in a modern GCM to any accuracy whatever, let alone the other properties such as wind velocity, humidity, particulate content, etc. In other words, we can’t specify the actual initial conditions today for a climate model. Whether the model is run forward or backward is irrelevant. Even if it “correctly” models everything, the run cannot give results that one would actually measure (if that were possible) 100 years in the future or 5,000 years in the past. Add to that the fact that we are trying to use them to figure out if they do model everything correctly, and you can see that this is a hopeless effort – and that’s true no matter how many improvements to climate models we make.

As an aside, I doubt that a computer code as complex as a GCM could be run “backwards” by doing something as simple as putting a minus sign in front of the time step. I have never read a description of how it’s done, but I know first-hand that the codes implementing numerical algorithms (of all types) are generally intolerant of “simple” changes. I’d be interested in learning how it is supposedly done.

February 25, 2023 3:48 am

I’m almost entertained by the ones freaking out about how it’s now two degrees in Colorado Springs and not one degree Fahrenheit. Oh, and your millimeter sea level rise is an “existential threat “. What’s an existential threat are Russia and Mainland China, but good luck telling them that….

Clyde Spencer
Reply to  johnesm
February 26, 2023 3:43 pm

Oh, and your millimeter sea level rise is an “existential threat “.

To put that into context, annual lateral motion of crustal plates is about an order of magnitude faster than sea level rise.

February 25, 2023 3:48 am

Holocene is here to stay for another millennium or two.
The fashionable and designer (so called) Anthropocene was so 2021.
I won’t cry for you Anthropocene
The truth is, you were never here
All through my wild days, my mad existence
I kept looking for you everywhere
But you vanished into past distance.

Last edited 1 month ago by vuk
Ron Long
Reply to  vuk
February 25, 2023 4:01 am

Don’t Cry For Me, Argentina?

Reply to  Ron Long
February 25, 2023 4:14 am

Evidente, from Caca-verde musical.

Ron Long
Reply to  vuk
February 25, 2023 6:19 am

vuk, see my comment above about poor wine grape harvest, that’s right, I’m Crying In Argentina.

Frank from NoVA
Reply to  Ron Long
February 25, 2023 7:29 am

‘…poor wine grape harvest…’

That’ll put a damper on future asados.

Reply to  Ron Long
February 25, 2023 8:01 am

No sympathy for you from here in the UK.
We are vegetating on turnips

comment image

Reply to  vuk
February 26, 2023 12:06 am

Baldrick will be upset

February 25, 2023 3:49 am

Instead of this “….then this points to the need to improve our understanding of natural climate forcings and feedbacks,…” why don’t we all agree to this “we need to improve our understanding.”

That should apply to all the sciences.

Further, instead of using the words “is” and “was, just be honest and say “maybe is” and “maybe was”.

This paper should remind us that science is a process and knowledge evolves, regardless of what some activists and politicians want us to believe.

Thank you, Andy, for another informative article. Bookmarked.

February 25, 2023 3:59 am

If only the pay of the modellers was performance related to accurate results rather than what they churn out from wild assumptions and guesstimate parameters.

At least on this occasion the decline is not hidden.

Reply to  strativarius
February 25, 2023 1:35 pm

They are long retired with healthy government pensions before any of their projections are ever tested.

I took up the projections from the CSIRO’s CMIP3 work that went to 2300 with absurd ocean surface temperature – by 2019 they were already far above what was being measured. CSIRO actually responded stating they no longer make projections to 2300. They did not even try to defend the predictions for 2019 made 20 years earlier. I suspect the CMIP3 vintage modellers are now retired living off fat government pensions.

One thing current climate modellers avoid is agreeing that the old models were wildly wrong.

Despite all their efforts to condemn CO2 it is not impacting the big manufacturers like China and India. So their folly will be clear by 2100 but anyone who pointed out the folly will not be around to point out the absurdity.

That said, the Northern Hemisphere will continue its long natural warming trend that was set in train about 1000 years ago. The black NH curve in the Figure 1 has reached the bottom of the cycle. It will rise until snowfall in the NH overtakes snow melt as ii has done 4 times in the last 400k years.

Clyde Spencer
Reply to  RickWill
February 26, 2023 3:48 pm

One thing current climate modellers avoid is agreeing that the old models were wildly wrong.

They were only wrong because they didn’t follow the lead of Exxon, don’tcha know!

Bob Weber
February 25, 2023 4:54 am

“If you cannot model natural climate change, you have no idea what the human influence on climate is.”

There you go!

Ulric Lyons
February 25, 2023 5:59 am

Misunderstanding Greenland temperature change leads to rewriting history.

1365-1195 BC, grand solar minima caused strong warming in Greenland, and led to the collapse of nine civilisations including the Minoans.

The Roman warm period began in the 1st century BC and continued until the Early Antique Little Ice Age from 350 AD, with Greenland then warming again. Another grand solar minimum, the start of the real dark ages.

The 8th century AD according to Esper 2014, had the warmest northern European summer temperatures of the MWP, and strong cooling in Greenland.

The warm spike in GISP2 around 1000 AD was the Oort solar minimum.

The Richard Alley series is questionable during the LIA, there should be warmer spikes during the centennial solar minima.

comment image

Ulric Lyons
Reply to  Ulric Lyons
February 26, 2023 1:20 pm

Here is a relevant temperature proxy for the Oort solar minimum, and warming during the LIA centennial solar minima:

comment image

February 25, 2023 6:48 am

the need to improve our understanding of natural climate forcings and feedback
We cannot predict the earth’s tides except by reference to past observations. The first principle tidal forcings from the sun and moon predict global tides will not exceed 1 foot.

Yet we think we can predict climate from first principles. And like the tides, first principles tell us that natural climate variability is low. Thus GHG effects must be large.

Like the tides, the only true measure of climate variability is past climate. You cannot predict variability from first principle forcings.

February 25, 2023 8:44 am

I still wonder if the LIA is the actual beginning of the Glaciation phase when I looked at Figure 1 on the chart showing that while all the other lines show relative stability which should be expected the heavy black line that covers 60N-30N region shows a long running dip to the bottom of the LIA then a big rebound which should slow down in the decades ahead then a new drop once again.

The LIA started dissipating around 1700 long before the slow CO2 increase could have significantly impacted the worlds climate as vividly shown in the famous GLACIER BAY rapid melt back from the mid 1700’s onward.

Reply to  Andy May
February 25, 2023 9:31 am

Oh, thank you for the link, I have posted it at my climate section of the forum with full mention of his copyright and Publisher of his work.

I will read it but there have been too many long running signs that the Interglacial period is at least in late fall part of the Earth Climate cycle which means close to glaciation phase will begin (winter phase) if it has not yet started as you state which will be devastating to us which is why we should be digging into this part of science research a lot more than the endless future warming modeling scenarios of which I quit reading in detail years ago as they are a waste of time and quite overblown worry since it is expected to be small anyway as CO2 warming role in it gets smaller over time.

Glaciation research is badly neglected a fertile area of research begs to be increased as it is undeniably getting close if it hasn’t already started which definitely means we should be getting prepared for it now!

Javier Vinós
Reply to  Sunsettommy
February 25, 2023 12:23 pm

The consensus has reached glaciation research and declared we are safe for tens of thousands of years due to CO2 present levels. I wish the people saying that could be revived in 5000 years and see their faces when they see by themselves how foolish they were.

What awaits humankind in a few thousand years is a test modern humans have never experienced. The descent into a new glaciation is something nobody alive then will like. The human population is going to decrease big time.

Reply to  Sunsettommy
February 25, 2023 2:10 pm

Warming of the oceans in the NH is the precursor to glaciation. So far only Greenland and Iceland are gaining permanent ice extent because they are surrounded by open ocean water that is warming up.

Warm surface water results in more atmospheric moisture available for increasing snowfall. NH snow coverage is increasing in autumn and winter but melt is still exceeding snowfall. The NH oceans will get a lot warmer yet and snowfall records will be a feature of weather reporting for next 9000 years.
comment image

By the end of next century, the permafrost will be advancing southward again.

The vast majority of people associate snow with cold. However creating snow is energy intensive. Every tonne of snow on land required ocean heat input equivalent to burning 100kg of coal. Snowfall actually makes the air temperature warmer in fall and winter because the radiating temperature is lower so less heat is lost. Same as clouds reducing nighttime heat loss from land.

What we are observing now has occurred 4 times in the past 400k years in similar circumstances of Earth’s orbit – perihelion is moving toward the boreal summer solstice. That increases NH summer solar intensity while reducing winter solar intensity. The combination increases winter advection of moist ocean air to land. These changes are all observable now but very few people understand what they are observing. They have this silly notion that CO2 somehow affect Earth’s energy balance.

Ulric Lyons
Reply to  Andy May
February 26, 2023 2:09 pm

“the LIA is just the coincidence of a Bray and Eddy solar low occurring at the same time.”

The LIA was a regular GSM series, which occur every 863 years on average. I have a concise explanation for the occurrence of the GSM series. Every fourth GSM series at 3453 years is evident in GISP2, and which can be confirmed with discrete mapping of each centennial solar minimum to determine the length. The best analogue for the LIA is the 4.2kyr event, and the best analogue for the next GSM series is the late Neolithic collapse period, 1365-1195 BC, which typically gets labelled as the Minoan Warm Period in the GISP2 series.

There are no Eddy or Bray cycles in solar activity.

Last edited 1 month ago by Ulric Lyons
Dave Andrews
Reply to  Sunsettommy
February 25, 2023 9:10 am

And Alpine glaciers began retreating around 1850 long before any effect of CO2 could be felt.

February 25, 2023 9:11 am

Computerized climate models are human-created virtual reality, which is why they differ from observed real reality. That’s where the conundrum lies.

Humans hypothesize (and hyperventilate) and their biases nurture their theories. It is in the crucible of careful observation that those theories are proved or disproved.

So far the theory of human CO2 emissions causing catastrophic warming—um—”climate change” remains disproved, and more so every year that temperatures and sea level rise fail to approach the trends prophesied by RCP projections; the improbable RCP8.5 being the one referenced in the vast majority of climate-related “studies”.

Projected temperature and sea level rise above 1986–2005 averages by 2100 from RCP8.5 is 3.7 °C (range of 2.7 to 4.8) and 0.6 meters (range of 0.45 to 0.82).

Projected temperature and sea level rise above 1986–2005 averages by 2100 from observations is about 1.3 °C and 0.3 meters, which is closest to the “no big deal” RCP2.6 scenario.

Javier Vinós
Reply to  stinkerp
February 25, 2023 12:25 pm

Computerized climate models are human-created virtual reality, which is why they differ from observed real reality.

I never understood the faith people put in climate models. They are no different than “Call of duty” in terms of representing reality.

Smart Rock
February 25, 2023 9:38 am

GMST probably peaked sometime late during the first half of the Holocene, at roughly 6.5 ka [4500BC], as evidenced by extensive proxy data and supported by theory and models

GMST at 6 ka in these models averaged 0.3 °C cooler than the preindustrial period

Kaufman & Broadman appear to be saying that the models were both right and wrong. In the world of 1984, this would be an excellent example of doublethink. We’re well on the way to achieving Orwell’s dystopian utopia, but we’re not quite there yet, so we might call it “cognitive dissonance”

Or perhaps it’s just sloppy writing and even sloppier proof-reading – editing – peer review?

Last edited 1 month ago by Smart Rock
February 25, 2023 10:26 am

Cue Stokes et al.

February 25, 2023 11:45 am

The GHG hot spot predicted by theory and all the climate models is a failed prediction that in any other branch of science would have killed the GHG radiative theory on the spot.

Javier Vinós
Reply to  ferdberple
February 25, 2023 12:30 pm

Scientists abhor a vacuum. Scientific theories are not killed until there’s a replacement. The biblical flood was a theory until it could be substituted. Static continents theory could not be replaced despite lots of evidence against it until plate tectonics arrived.

Frank from NoVA
Reply to  Javier Vinós
February 25, 2023 5:24 pm

‘Scientific theories are not killed until there’s a replacement.’

Which presents a huge problem given the politicization of science.

John Shewchuk
February 25, 2023 4:48 pm

Exactly right Andy. This is what I say each time I give a talk about the failing climate models … they are all useless because none can replicate past climate changes — like the Medieval Warm Period and Little Ice Age — and prior peiods.

Pat from Kerbob
February 26, 2023 2:18 pm

Models are holy items to the faithful.
It’s disturbing.

February 26, 2023 9:56 pm

See Climateaudit what these ‘landmark reconstructions’ are worth.

Rosenthal et al 2013 instead suggests, that peak Holocene temperatures were at least 3 deg warmer than the Little Ice Age:

And the reason is easy to understand:

1) Milankovich cycle was at peak 10 K years ago, with summer insolation about 35 W/m2 higher at 60 deg North than today. (versus 2 W/m2 from man made CO2)

2) That means back then the Arctic ice melted MUCH faster in spring/summer than in the 20th century. But that’s not all.

3) Total incoming radiation (top of atmosphere) is over 500 W/m2 in Arctic summer.
This spring/summer ice melt is an albedo flipper, from white ice/snow, reflecting >90% energy back to space, to black/brown water/land, absorbing almost everything into the Arctic region.

4) While Milankovich cycles in essence only redistribute incoming energy around the globe with little change the total, this albedo flipping tips the balance, Arctic summer insolation is the game changer, relentlessly pushing vast amounts of additionally solar energy into lands and seas at higher northern latitudes.

5) Over Antarctica, there was a summer insolation minimum 10K years ago & a maximum at present. But Antarctic albedo does not flip, because
a) too cold for ice melt
b) region cloudy most of the year
But the increase increase of energy there lacks elsewhere on the globe, which also supports the 9000 years temp decline.

February 27, 2023 6:25 am

The CMIP6 models have problems. Despite global temperature having been falling a little behind the medians of the CMIP3 & CMIP5 models, and tropical upper troposphere temperature having been greatly falling behind median of the CMIP3 & CMIP5 models (indicating perhaps above-actual water vapor positive feedback), the CMIP6 models are projecting more warming than the CMIP3 & CMIP5 models did. However, this does not mean that increasing greenhouse gases is not making the world warmer, or even that these climate models are (on average) less than 50% accurate overall at modeling effects of increase of greenhouse gases.

%d bloggers like this:
Verified by MonsterInsights