by Pascal Richet
Institut de Physique du Globe de Paris, 1 Rue Jussieu, 75005 Paris, France
"Abstract
As simply based on fundamental logic and on the concepts of cause and effect, an epistemological examination of the geochemical analyses performed on the Vostok ice cores
invalidates the marked greenhouse effect on past climate usually assigned to CO2 and CH4.
In agreement with the determining role assigned to Milankovitch cycles, temperature has, instead, constantly remained the long-term controlling parameter during the past 423 kyr,
which, in turn, determined both CO2 and CH4 concentrations, whose variations exerted, at most, a minor feedback on temperature itself.
If not refuted, the demonstration indicates that the greenhouse effect of CO2 on 20th century and today's climate remains to be documented, as already concluded from other evidence.
The epistemological weakness of current simulations originates from the fact that they do not rely on any independent evidence for the influence of greenhouse gases on climate over long enough periods of time.
The validity of models will, in particular, not be demonstrated as long as at least the most important features of climate changes,
namely the glacial–interglacial transitions and the differing durations of interglacial periods, remain unaccounted for.
Similarly, the constant 7 kyr time lag between temperature and CO2 decreases following deglaciation is another important feature that needs to be understood.
Considered in this light, the current climate debate should be considered as being the latest of the great controversies that have punctuated the march of the Earth sciences,
although its markedly differs from the preceding ones by its most varied social, environmental, economical and political ramifications.
5.2 The threat of circular reasoning
Petit (2013) asserted that the amplifying role of CO2 on climate change was first demonstrated by the Vostok analyses
and added that these results were then “very rapidly taken into account by IPCC,
which recently concluded that human activities are responsible for the current climate warming.”
The importance of the Vostok results has, thus, been central in the current debate,
even though it is now commonly emphasized that global warming is demonstrated instead from a quantitative understanding of the physical mechanisms
through which temperatures and CO2 concentrations are related in advanced climate models.
Regarding their Vostok analyses, Petit et al. (1999) stated that
“results from various climate simulations make it reasonable to assume that greenhouse gases have, at a global scale,
contributed significantly (possibly about half, that is 2–3 ∘C) to the globally averaged glacial–interglacial temperature change.”
That this statement is clearly contradicted by the present analysis in turn invalidates those climate simulations from which it derived.
The CO2 feedback supposedly shown by the ice-core results thus appears to exemplify a rather common situation
whereby a preconceived notion of causality has led to the misinterpretation of the data
– perhaps also because these results were not plotted as a function of time but of depth,
which may carry the illusion that CO2 peaks systematically precede temperature peaks.
This situation illustrates the peril of transposing theoretical concepts to a very complex system when the observational support is incomplete
or when an independent, rigorous assessment of the validity of the procedure is lacking.
In other words, interpreting the CO2 and temperature records of ice cores in the light of climate models has represented an incorrect methodological leap.
Ironically, any claim that models accurately reproduce the reported climate evolution since the late 20th century would rather illustrate their spurious nature,
and not prove their validity,
if the temperature rises of this period are not caused by increases in CO2 concentrations.
There is, additionally, a great epistemological weakness in climate models because the timescales of 150 years at most they consider with direct or indirect observations
are tremendously short with respect to those of even the shortest fluctuations exhibited by the climate record (see Figure 1 at end of article).
The situation is analogous to that one would face in attempts made at understanding the basic physics of tides through focusing on a single ripple at the water surface
and not on entire ebbing and waning cycles of variable amplitudes.
The reliability of climate models should thus be ascertained on the basis of their ability to match at least the main features of the latest glacial cycles,
beginning with the sharp glacial–interglacial transitions.
Given the fundamental role assigned to greenhouse gases,
any specific model cannot be considered valid as long as the width differences between temperature and CO2 peaks are not accounted for quantitatively.
As a matter of fact, current models suffer from the circular nature of the reasoning behind their assumed feedback mechanism whereby,
in the last analysis, the predicted influence of CO2 simply conforms to the posited effects in a situation where the anthropogenic increases in CO2 concentrations happen to accompany those of temperatures.
In a kind of reductio ad absurdum, a similar situation would be encountered if the quantitative correlation observed
between the recent increases in atmospheric CO2 contents and the geographic displacement of the magnetic north pole (see Figure 3 at end of article)
were interpreted as a causality relationship – which could of course not be considered seriously in view of a complete physical implausibility!
1 Introduction
Perhaps the most important feature evidenced by the history of science
is how ideas that were unanimously accepted for very long periods of time have eventually been firmly rejected.
... As robust and convincing as they may appear, theories are rarely immune to various kinds of flaws that appear more or less rapidly
and serve as seeds for either major reformulation or for complete rejection,
as exemplified by geocentrism and the four-element theory.
With the reasonable premise that we are not any smarter than our predecessors,
an intriguing problem is to identify which of the currently accepted theories might fall into oblivion in the future
and make historians study why their demise did not take place earlier.
The goal, then, consists of spotting practical or theoretical weaknesses and assessing whether they are actually significant or not.
For this purpose, epistemological approaches are most valuable because they focus on fundamental principles without the need for delving into technical details.
A thesis for which such an approach can be followed is that anthropogenic emissions of CO2 and other greenhouse gases,
such as methane, have dire warming effects on the Earth's climate.
Because CO2 emissions are considered to be the single most important factor currently affecting climate,
unprecedented efforts are being formulated to achieve carbon-free societies within a few decades.
In view of the major social, environmental, political and economic issues raised by such a transition, two points deserve special attention.
The first concerns geochemical evidence available for the greenhouse effects of CO2 (and of CH4 as well) over periods of time long enough to encompass great climate cycles.
The second deals with the actual heuristic value of climate simulations, which appears to be generally acknowledged without having undergone real in-depth analyses.
Both points will, thus, be reviewed critically from an epistemological standpoint in the present study. ...
3 The temperature–CO2 relationship
3.1 The ice-core analyses
The ice cores drilled down to a depth of 3310 m at the Russian Vostok station have yielded the first comprehensive climate record spanning the last 423 kyr (Petit et al., 1999).
Including the current one, five great cycles of glaciation–deglaciation have been revealed.
The four most ancient cycles lasted from 87 to 123 kyr each,
during which Antarctic temperatures changed by about 10 ∘C
and atmospheric CO2 concentrations varied between 180 and 300 ppmv (parts per million by volume; ...
with the lowest values having slowed down but not impeded photosynthetic activity (Gerhart and Ward, 2010).
Another record extending back to 800 kyr
was subsequently obtained at the Dome C site of the European Project for Ice Coring in Antarctica (EPICA), 560 km south of Vostok (Lüthi et al., 2008).
The two series of analyses are very similar for their period of overlap.
Between 400 and 800 kyr, the Dome C record reveals four more glacial cycles over a 200 m depth.
Possibly because of perturbations and rearrangements of the accumulated ice, however,
the older material exhibits warming–cooling episodes in the form of broader features.
Although these additional cycles are valuable for studying transitions between glacial and interglacial conditions,
they will not be considered here because their lower resolution prevents further information from being drawn on the temperature–CO2 relationship.
... 3.2 Geochemical inferences
With the obvious exception of the ongoing cycle I, which began 18 kyr ago,
all others follow a common pattern
whereby a sharp glacial–interglacial transition is followed by a series of warming–cooling episodes of smaller magnitudes.
Even without ever having heard of Milankovitch cycles,
one would readily draw firm conclusions from the quasi-periodicity of these cycles and their common patterns.
In logical order, these inferences are as follows:
i. The major peaks were necessarily under astronomical control
because no natural phenomena on Earth exhibit anything approaching, even very distantly, such observed regularities with periods of tens of thousands of years.
ii. This astronomical control of glacial cycles was necessarily exerted through variations in the energy received by the Earth.
This energy could have been emitted only by the Sun.
At constant Sun power, its amount itself depends, in a complex manner, on a great many local and seasonal parameters, such as the extent of ice sheets.
iii. In the absence of photochemical production of CO2 in the atmosphere, increases in the amount of heat transferred by solar radiations
necessarily translate directly into either temperature increases or endothermic phase changes (e.g., ice melting) at the Earth's surface.
iv. Acting also first on temperature and ice volume, the opposite changes take place when the Earth's net radiation energy budget becomes negative.
v. The temperature variations themselves induce concentration changes of chemical species in the atmosphere;
for example, CO2, through variations of its overall solubility in seawater and the temperature dependence of the concentrations of carbonate species,
or CH4, through adjustment of biological activity.
vi. Barring any exceptional event, such as the impact of a giant meteorite or a mega-volcanic eruption, whose occurrences are not apparent in the Vostok record,
temperature changes were, thus, necessarily the triggering causes of all episodes.
In accordance with geological evidence (Lane et al., 2013),
the signature of even the most explosive volcanic event of the Quaternary, the Toba super-eruption 75 kyr ago, could not be found in the record.
vii. As indicated by the jagged nature of the record, temperature and CO2 fluctuations constantly took place with a variety of intensities and timescales,
with the shortest ones appearing as numerous spikes superimposed on the most recent temperature peaks.
viii. Since there is no reason why temperature spikes would have been less frequent or less intense in the most ancient than in the most recent part of the temperature record,
the contrast between the jagged temperature and smoother CO2 records was unlikely restricted to the most recent cycle.
Instead, it existed in all cycles before the temperature record was progressively smoothed out back in time as noted above.
4.2 The CO2 feedback
The simple temperature–CO2 relationship described in the previous section is not commonly considered, however, because it ignores greenhouse effects.
To conform to the Arrhenian paradigm, Petit et al. (1999) took over the idea that the initial rise in the atmospheric CO2 concentration
(induced by temperature increases at the onset of a Milankovitch cycle)
in turn amplifies the initial orbital forcing and is itself amplified by atmospheric feedbacks.
Epistemologically, such a four-step process must be examined in the light of the principle of parsimony, which was also first stated by Aristotle in his Topics.
“It is also a fault in deduction when a man proves something through a long chain of steps, when he might employ fewer steps.”
To be justified, therefore, such additional steps require unquestionable evidence.
Feedbacks are indeed possible, where a cause alternatively becomes an effect and an effect a cause.
By definition, however, such a mechanism implies the synchronicity of causes and effects to within the timescales of their mutual interactions.
In the Vostok case, the CO2 feedback would reinforce temperature increases during the warming periods
but also, reciprocally, temperature decreases during the cooling stages of the Milankovitch cycles.
From the dual way in which the feedback would work,
temperature decreases and increases should thus be similar for the same concentrations of greenhouse gases
regardless of the residence times of these gases in the atmosphere.
Now, the synchronicity of causes and effects is well respected only during the warming periods, where the time lag between the temperature and CO2 increases is very small.
As already emphasized, in contrast, synchronicity clearly breaks down during the cooling periods,
and this is most clearly so when temperatures sharply decrease while CO2 concentrations remain high.
This feature is most obvious in the glacial–interglacial transition of cycle III,
where the temperature peak is narrow and symmetrical whereas the CO2 peak exhibits the large shoulder representing the aforementioned 7 kyr time lag.
The feature is also clearly seen in cycle II, where the large jagged CO2 peak contrasts with the rapidly decreasing magnitude of the temperature peak.
Hence, the fact that temperature decreases do not depend in any noticeable way on CO2 concentrations in all cycles
clearly demonstrates that the synchronicity required by the feedback mechanism is lacking.
The CO2 feedback mechanism is, in addition, contradicted by the marked contrasts between the broad maxima in CO2 concentrations and doublets of sharp temperature peaks signaled by solid dots in Fig. 1. (below)
As indicated by the data for cycle IV, these contrasts are unlikely due to a resolution difference between the two parameters.
That they are not coincidental is revealed by their systematic observation only in those parts of Milankovitch cycles where insolation changes are the smallest (Fig. 1).
Also striking is the fact that, as shown by the dots included in the insolation plots at the top of Fig. 1, the temperature doublets are found, each time, at similar places within insolation cycles.
Put differently, the dotted peaks of Fig. 1 again demonstrate that temperature is sensitive to insolation changes
but not to CO2 concentration, a conclusion also consistent with the contrasting the jagged–smooth contrast of temperature and CO2 records.
Regarding CO2 feedback, the CH4 concentrations raise yet another difficulty that may be even more fundamental.
Like those of CO2, their variations could not be directly caused by changes in the solar energy transferred to the Earth's atmosphere.
They necessarily resulted from temperature changes.
If CO2 contents had exerted a noticeable feedback on temperatures, then the peak widths of the reported CO2 and CH4 concentrations should be highly correlated.
Such a causal correlation is actually nonexistent because, in marked contrast with the CO2 contents,
the CH4 concentrations show no time lags whatsoever with respect to temperatures.
Instead, these CH4 concentrations correlate remarkably well with temperatures,
as made clear by the fact that these two parameters have nearly the same peak widths (see Figure 2 at end of article).
More recent data have even revealed closer still correlations (e.g., Buizert et al., 2015).
Ironically, this clear synchronicity might make CH4, and not CO2, a potential match for a feedback mechanism.
But CH4 concentrations ranged from only 0.4 to 0.7 ppmv, which were about 500 times smaller than those of CO2 (Fig. 1)
and from 3 to 4 times lower than the current values.
If really significant in the past, a methane feedback would then cause today's temperatures to be considerably higher than observed.
Therefore, the ice-core data conversely also rule out any noticeable influence of methane.
5 Implications
5.1 The CO2 conundrum
As a rule, correlation does not necessarily imply causality.
In marked contrast, a lack of correlation resolutely rules out any causality.
Reconciling the driving role of CO2 assigned by climate models with the opposite conclusions drawn from the ice-core record thus seems fraught with considerable difficulties.
Hence, the ice-core results shift the burden of proof of any CO2 influence on temperature to the proponents of the feedback mechanism
and make, in addition, any climate sensitivity determinations problematic.
Current climate models are, in practice,
not open to falsifiability in Popper's (1959)
sense because they are
so complex,
involve so many physical parameters,
rely on so much data for their design and assessment,
lack proper error propagation estimations and
suffer from the fact that the observations they aim at reproducing cannot be changed at will to check their responses under widely different conditions.
Instead, models are claimed to be reliable thanks to their sound physical basis,
which is not supported by the present analysis,
whereas recourse is also made to the subjective notion of consensus to assert their validity.
Whether or not such a consensus prevails here does not need to be discussed at length because this notion is epistemologically irrelevant.
As already alluded to, the history of science is nothing more than a long stroll through the cemetery where ideas that were overwhelmingly accepted are now resting in peace.
For the present issue, the point has been remarkably well exemplified in the late 19th century and then again in mid-20th century
by the consensus successively reached for, then against and, finally, for the astronomical control of ice ages.
As stated for the latter period by Imbrie and Palmer Imbrie (1979), “during the 1930s and 1940s, most European geologists were won over by the Milankovitch theory”
and “the majority of scientists continued to favor the astronomical theory as late as 1950.
But the early 1950s saw a dramatic about-face since, by 1955, the astronomical theory was rejected by most geologists.”
The case against became particularly strong when the new technique of 14C dating “revealed a pattern of climatic change that was at variance at almost every point with the astronomical theory.”
Shortly before Hays, Imbrie and Shackelton published their landmark study (Hays et al., 1976), it followed that,
according to Imbrie and Palmer Imbrie, “by 1969, the majority of scientists were sufficiently impressed with the radiocarbon evidence against the Milankovitch theory
to eliminate the idea as a serious contender in the ice age sweepstakes.”
In contrast to climate simulations, the present analysis is open to falsifiability since its fallacy, if any, could be pointed out without ambiguity.
In this respect, one may stress that the approach followed here directly integrates, with the appropriate weights, all factors relevant to the problem
and that it is totally independent of any assumed physical mechanisms, interactions surmised between climate variables,
considerations on the CO2 cycle, statistical analyses of selected sets of data assumed to be representative of the problem and any other simulation features.
A cardinal rule in science is to reject a hypothesis that clearly contradicts the experimental findings it is supposed to account for,
especially if it also contradicts the most fundamental tenet of science,
the principle of non-contradiction, which is “the most certain of all” in Aristotle's words.
If the present analysis cannot be refuted, one should then reject the Arrhenian paradigm and conclude
(i) that changes in the concentration of atmospheric CO2 up to 300 ppm had minor effects at most on temperatures during the past 423 kyr,
(ii) that, as described in Sect. 4.1, the concentration of atmospheric CO2 simply adjusted during this period to the prevailing temperature conditions at the Earth's surface,
whose variations were mainly determined by insolation changes during Milankovitch cycles, and
(iii) that significant contributions of CO2 and CH4 to temperature changes at the Earth's surface remain unsubstantiated by direct, independent evidence.
Entertaining the possibility that temperature rises along Milankovitch cycles could have been triggered by increasing CO2 concentrations is in fact surprising as it would violate fundamental thermodynamics.
As known since Planck's work, radiation represents not only energy but also entropy.
Regardless of the particular ways in which radiation entropy is transferred to the Earth's surface and atmosphere, or lost from them,
the basic fact is that temperature and entropy are the intensive and extensive conjugate variables of thermal energy, respectively.
Under all circumstances, any temperature (or, more generally, enthalpy) changes of a system are thus necessarily driven by an entropy change (see Richet, 2001).
This is another way of stating that greenhouse gases can affect climate only via thermal changes.
As such, their effects would manifest themselves in any temperature record, which must be considered in this respect as thermograms in thermal analyses.
Obviously, one could alternatively claim that other factors than CO2–temperature interactions are involved in the very complex climate problem;
if so, however, an important aspect would be that changes in atmospheric CO2 contents of tens or even hundreds of parts per million certainly could not directly affect ice volume
or tipping points related to patterns of oceanic circulation,
to name a single important feature,
but could act only through thermal changes as described above.
The conclusion, thus, remains that it would not make sense to place so much emphasis on the effects of CO2 in either climate models or on emission reductions in environmental policies.
For ice cores, a first key factor that ensures reliable assessments of temperatures with respect to CO2 and CH4 concentrations is the determinations of the three parameters for the same ice fragments of known ages.
A second factor is the fact that the temperature variations of up to 12 ∘C during climate cycles observed in polar regions are much greater than the 2–3 ∘C that affected the entire Earth,
which accounts for the much higher resolution of the polar records.
And a third factor is the large timescales of these observations, which are more than 1000 times longer than those of climate simulations a
nd of available measurements of atmospheric temperatures and gas concentrations.
Of particular importance is also that the 423 kyr considered here are long enough to encompass four complete glacial cycles,
for which short-term fluctuations can be neglected, but short enough to not be affected by other factors,
such as changes in continent positions,
that play an important role over very long timescales.
In addition, the lack of correlations that support CO2 forcing is immune from the inevitable biases that arise
when unrelated sets of data are used for different parameters,
especially when some are derived indirectly from proxy studies or when investigations deal with short periods of time.
In the atmosphere, the maximum CO2 concentration of 300 ppm found in the Vostok record was reached again in the 1910s.
The main effect of such high concentrations was simply to increase considerably the subsequent CO2 time lag behind the temperature leads, without significant effects on past climate.
Hence, it is doubtful that any significant global warming could have been caused by human emissions during most of the 20th century
as a result of the additional 50 ppm CO2 increase observed until the 1980s.
Given the lack of evidence for feedback mechanisms particularly well demonstrated by the CH4 record,
it in fact remains to be determined from which level, if any, CO2 concentrations would begin to become relevant
and to ascertain the dire consequences of current CO2 levels.
The principle of parsimony thus points to any current warming as being just one of the recurrent fluctuations clearly recorded as spikes in the last two climate cycles,
which have not yet been averaged out in the Vostok record (Fig. 1)
and, surprisingly, seem to be overlooked in discussions of short-term temperature variations.
As often noted, it would in fact be an arbitrary assumption to posit that a system as chaotic and as highly heterogeneous as the Earth,
which must be described in terms of complex sets of coupled nonlinear equations,
would always evolve in a smooth manner over long periods of time.
Probably the most in-depth geochemical discussion of glacial cycles and climate effects of CO2 has been published by Broecker (2018),
who pointed out that some important features of past climate could not be accounted for in terms of CO2 variations.
Broecker nonetheless stated that “the geologic record makes a strong case that CO2 has been instrumental in driving past climate changes”,
adding that “as made clear by the record for the last 150 kyr”, CO2 “has not acted alone” because insolation cycles,
ocean circulation
or reorganization and
latitudinal temperature gradients
also contributed.
Although less important than insolation cycles, the other factors mentioned by Broecker certainly have to be taken into account in more detailed pictures of climate change.
Overall, however, the fundamental importance of the Antarctica records rests on the fact that glaciation–deglaciation cycles are the most conspicuous features of climate changes
and that the ensuing sea level variations necessarily affect the whole planet.
Of particular interest in this respect is the synchronicity of the episodes of warming and cooling found over long timescales between Greenland and Antarctica (Pedro et al., 2011).
In addition, the remarkable synchronicity of temperature and methane variations points to a lack of major latitudinal time lags,
since the methane budget appears largely controlled by tropical sources and sinks (Loulergue et al., 2008) and not by interactions with seawater (Reeburgh, 2007)."
Figure 1:
Figure 2:
Figure 3:
