Ice-Core Layer Counting

"It is relatively easy to count annual layers downward from the surface through considerable depths in the Greenland ice sheet." — Larry Vardiman, Institute for Creation Research, 1992¹

"Fundamentally, in counting any annual marker, we must ask whether it is absolutely unequivocal, or whether nonannual events could mimic or obscure a year." — Richard Alley and colleagues, GISP2 team, 1997²

"In the upper halves of the Greenland ice cores, creationist and uniformitarian age assignments agree to within a factor of two or three of one another." — Jake Hebert, Institute for Creation Research, 2021³

Summary

Snow that falls on Greenland in July is chemically different from snow that falls in January.  It carries different dust, different sea salt, different acidity, different oxygen isotopes.  Bury it, and if nothing smears the difference away, the ice preserves an alternating signal that can be counted like tree rings.

That is the whole method.  Everything else is a question about whether the signal survives, whether it is really annual, and how far down you can still see it.

Eight conclusions, stated first.

1. The counting stops at 60,000 years. The Greenland Ice Core Chronology 2005 (GICC05) is continuously counted layer by layer to 60,000 years before AD 2000, and below that the chronology comes from an ice-flow model, not from counting.⁴ The famous "800,000 years of ice" from Antarctica is not counted at all — East Antarctic accumulation is far too low for annual layers to survive, and those ages come entirely from flow models tuned to orbital templates.³ Anyone citing 800,000 counted layers has misread the literature, and the young-earth critique is right about that.
2. Inside the counted range, up to five physically independent signals have to agree. Sodium peaks in winter (marine air masses), calcium in spring (continental dust), ammonium in late spring (North American biomass burning), nitrate in summer, and the water isotopes trace a sine wave with summer maxima.⁵ These come from four different source regions and four different transport pathways. A single storm can fake one of them. Faking all five, in the right order, is a different proposition.
3. The counters publish their own error bars, and they are not small. Ambiguous layers in GICC05 are counted as half a year plus or minus half a year, and the ambiguity is summed down the core. The result — the maximum counting error — is "on average 1% in the Holocene and 5% in the Glacial," reaching 950 years over the 41,800–60,000-year section.⁴ GISP2's team give roughly 1–10% in the top 2,500 m, about 20% below that, and declare the core unusable below 2,800 m.⁶
4. The count has been checked against three different decay systems and lands within a few per cent of all of them. Argon-argon on the Laschamp geomagnetic excursion: 40,400 ± 1,000 years, against 41,250 ± 814 counted — 2.1%.⁴ Argon-argon on the North Atlantic Ash Zone II: 54,500 ± 1,000 against 55,380 ± 1,184 counted — 1.6%. Uranium-thorium on Chinese cave stalagmites, across the whole 60,000 years: agreement "within 800 years."⁴
5. The chronology was caught being wrong, by seven years, and fixed. The 774–775 CE burst of solar particles is fixed to a calendar year by tree rings on five continents. Its beryllium-10 signature in Greenland and Antarctic ice sat at 768 CE on the chronologies then in use.⁷ One named cause was a sulfate layer wrongly attributed to Vesuvius in AD 79 and used as a fixed tie point; another, in the authors' own words, was "use of manual-layer interpretation techniques that may favor interpretations consistent with a priori knowledge or existing chronologies (WDC)."⁷ That is a real admission of the exact bias the young-earth critique alleges, and it should be quoted. It should also be quoted with what followed: the corrected chronologies are good to better than ±5 years over 2,500 years, and 24 of 32 independently recorded historical dust-veil years now fall within ±3 years of an ice sulfate spike, where on the old chronology there was no statistically significant association at all.⁷
6. The young-earth case is that the deep "annual" cycles are storm cycles. Michael Oard's model: in a 700-year post-Flood ice age with enormous snowfall, one real year deep in the core "may be interpreted by uniformitarians as 100 or even 1,000 'annual cycles'."⁸ The mechanism is genuine — sub-annual oscillations do exist, and the GISP2 team say so themselves. What has never been published is a quantitative demonstration that they produce a factor of fourteen, consistently, in five signals at once.
7. The Lost Squadron argument should be retired. P-38 fighters abandoned on Greenland in 1942 were recovered in 1992 under about 268 feet of snow and firn. The figure is real; the inference is not, because the aircraft sit near the southeast coast, the wettest part of the ice sheet, while the deep cores are drilled at the dry interior divide. Vardiman, who introduced the argument, immediately warned that "life is never as simple as implied above."¹ Jake Hebert's 2018 treatment concedes the apples-to-oranges charge. Creation Ministries International's own article has carried an editorial correction for well over a decade.
8. The strongest unanswered problem for the young-earth reading is in the upper core, not the lower. Hebert concedes that the top halves of the Greenland cores agree with creationist estimates "to within a factor of two or three."³ But the GISP2 depth–age table puts the Holocene boundary at 1,678 m and 11,651 counted years. On a post-Flood model, essentially all of that ice fell in the roughly 4,000 years after the ice age ended, under something like modern climate, where the high-snowfall over-counting mechanisms do not apply. That is an over-count of about 2.9 in the part of the core where the layers are thickest and easiest to read — and about 5.0 in the ice below 500 m if the model is granted a perfectly counted top 500 m.⁹ No mechanism has been offered for either.

The thing that is actually surprising

Larry Vardiman's 1992 article is the founding young-earth treatment of ice cores, and it concedes more than most later ones do.  It states that Greenland layer counting is "relatively easy… through considerable depths," that historic volcanic eruptions cross-check the count "with a high degree of accuracy," and that Hammer and colleagues had already identified about a dozen historical eruptions in a central Greenland core.¹  His argument is not that the top of the core is uncountable.  It is that the bottom is, and that the deep chronology depends on models rather than counting — which, for Antarctica, is correct.

Part 1: What is actually being counted

The five signals

Snow arriving at the Greenland summit carries a chemical load that changes with the season, because its sources change with the season.  The GICC21 team list them:

• Sodium (Na⁺) peaks in winter, "because of increased advection of marine air masses." Winter storms bring sea spray inland.
• Calcium (Ca²⁺) peaks in spring, "because of enhanced transport from terrestrial reservoirs" — the continental dust season.
• Ammonium (NH₄⁺) peaks in late spring or summer, "because of enhanced biogenic activity in the North American continent," including boreal wildfire.
• Nitrate (NO₃⁻), also biogenic, peaks in summer.
• Water isotopes (δ¹⁸O, δD) "show a sinusoidal pattern with winter valleys and summer peaks, mainly representing temperature variations at the drill site."⁵

To these the GISP2 programme adds two physical signals:  visual stratigraphy, in which summer depth-hoar layers scatter light differently from winter wind slab, and laser-light scattering from dust, which resolves the dust season at millimetre scale.⁶

The important point is not that there are five signals but that they are driven by different physics in different places.  Sea salt comes from the North Atlantic.  Dust comes from central Asia.  Ammonium comes from North American fires.  Isotopes come from the local temperature.  A blizzard that puts a spurious peak into the sodium record has no reason to put a matching spurious peak into the ammonium record from a continent away.

How a year is identified

Manually:  a scientist looks at the parallel signals plotted against depth and marks where the sequence winter–spring–summer completes.  Increasingly also automatically:  the StratiCounter algorithm treats the multi-proxy record as a hidden semi-Markov model and "learns to recognize the specific annual pattern," returning both a layer count and a probability distribution over layer boundaries.⁵

Both are used, and GICC21's authors are explicit that neither alone suffices:  "A layer counting bias was found in all ice cores because of site-specific signal disturbances;  therefore the manual comparison of all ice cores was deemed necessary to increase timescale accuracy."⁵

That sentence is worth pausing on.  It says that automatic counting is biased at every site, and that the fix is human judgement across multiple cores.  A critic is entitled to note that human judgement is exactly what the Sigl 2015 correction blamed for a seven-year error.  The answer is that the residual error is measured, and the measurement is the next section.

How the uncertainty is computed

GICC05's convention is deliberately conservative.  Where the counters cannot decide whether a feature is a year boundary, they record it as half a year, plus or minus half a year, and sum the halves down the core.  The accumulated ambiguity is the maximum counting error (MCE), treated as a 2σ bound.⁴

The resulting numbers, from the paper's own text and Table 1:

Summarised in the paper's own words:  "The maximum counting error of the time scale is on average 1% in the Holocene and 5% in the Glacial."⁴

The authors also flag an error they cannot quantify, and it should be reported:  "There is potentially an additional uncertainty in the time scale arising from a systematic bias due to the assumed shape of annual layers…  Such a bias is, however, difficult to quantify and it is, therefore, not included in the uncertainty estimates given."⁴

For the last 3,800 years GICC21 does better:  its uncertainty rate is "1 year per century," and it revises GICC05 younger by about 13 years at 3,835 years ago.⁵

Part 2: Where the counting stops

This is the section most likely to surprise a reader who has met the subject through popular writing on either side.

Greenland, GICC05:  continuously counted to 60,000 years b2k.  Holocene from DYE-3, GRIP and NorthGRIP;  glacial from NorthGRIP alone.  Below 60,000 years, ages come from an ice-flow model.⁴

Greenland, GISP2:  counted to roughly 110,000 years at 2,800 m, with a stated error of about 20% between 2,500 and 2,800 m, and with the ice below 2,800 m deformed and excluded.⁶

Antarctica, WAIS Divide (WD2014):  annual-layer counted through 2,850 m of core, in a high-accumulation coastal-influenced setting.⁷

Antarctica, EPICA Dome C and Vostok:  not counted.  Accumulation at Dome C is a few centimetres of water equivalent a year;  annual layers are destroyed by diffusion long before they reach depth.  Ages come from flow models and orbital tuning.  Hebert states this correctly and it is worth quoting him:  "if accumulation rates are too low, as is the case for the deep cores in East Antarctica, direct counting of layers is not possible, so glaciologists use theoretical age-date models to assign ages to different depths within the ice."³

His further charge — that those models "implicitly assume 'millions of years' by taking the height of the ice sheet to be constant or nearly constant" — is a fair description of the standard assumption.  Whether it is circular depends on whether the model is being used to establish an old age or to interpolate within one established otherwise.  In practice the deep Antarctic chronologies are tuned to orbital and gas-record targets, which is a separate argument treated in Orbital Cycles in Rock.

None of the deep-ice ages are used as evidence in this series.  The counting evidence stops at 60,000 years, and that is sufficient for the question at hand.

Part 3: The cross-checks

A layer count could, in principle, be systematically wrong.  It could not be systematically wrong in a way that happens to track four unrelated clocks.

Against argon-argon

Two horizons in the Greenland glacial section have independent radiometric ages from lavas elsewhere:

• The Laschamp geomagnetic excursion, a near-collapse of the dipole field recorded worldwide, dated by ⁴⁰Ar/³⁹Ar on the Laschamp and Olby lava flows at 40,400 ± 1,000 years. Counted position in NorthGRIP: 41,250 ± 814. Difference 850 years, 2.1%, inside the combined uncertainty.⁴
• North Atlantic Ash Zone II, an Icelandic tephra, ⁴⁰Ar/³⁹Ar 54,500 ± 1,000; counted 55,380 ± 1,184. Difference 880 years, 1.6%.

Potassium-argon is an electron-capture decay.  Layer counting involves no decay at all.  These two have no physical reason to agree unless both are approximately right.

Against uranium-thorium

The Hulu Cave stalagmites in eastern China are dated by ²³⁰Th/²³⁴U, an alpha-decay chain with entirely different chemistry and different half-lives from potassium-argon.  Their oxygen-isotope record shows the same abrupt warmings (Dansgaard–Oeschger events) as Greenland ice.  Line the two up and:

"GICC05 compares well to the time scale of that record with absolute age differences of less than 800 years throughout the 60 ka period."⁴

and, on the interstadial onsets specifically:

"the two time scales agree within 500 yrs throughout the 60 ka period.  Such a good agreement would be unlikely if one of the records had a significant dating error."⁴

Two other caves agree — Kleegruben in Austria to within 300 years, Villars in France placing the onset of GI-12 at 46.6 ± 0.5 ka against the counted 46,860 ± 956.  One cave does not:  Moomi Cave on Socotra matches GI-12 to within 350 years but deviates by more than 1,000 years on adjacent interstadials, and the authors say so.⁴

Against cosmogenic isotopes and tree rings

This is the sharpest check, and it is also where the ice chronology was caught out.  It is described in Part 4.

Against history

Big eruptions leave sulfuric acid in the ice, and some of them are in the written record.  GICC21's Table 3 lines the two up:⁵

And in the other direction:  32 independent historical accounts of a dimmed or discoloured Sun — the oldest a Babylonian note from 255 BCE that "the disk of the sun looked like that of the moon" — were compared against volcanic years in the NEEM core.  24 of 32 fall within ±3 years, with a Monte Carlo p-value below 0.001.  On the old chronology, the same test returned no significant association at all.⁷

Part 4: The seven-year error, told properly

This episode is the single most useful thing in the ice-core literature for a reader trying to judge whether the field checks itself, and it should be told without spin in either direction.

The setup.  In 2012 Fusa Miyake and colleagues found that Japanese cedar rings recorded a sudden 12‰ jump in carbon-14 between 774 and 775 CE, "about 20 times larger than the change attributed to ordinary solar modulation."¹⁰  A spike that abrupt is atmospheric and therefore global.  It also produces beryllium-10, which falls out onto ice sheets.

The test.  Michael Sigl and colleagues measured beryllium-10 in Greenland and Antarctic cores and looked for the spike.  They found it — in both hemispheres, "exceeding the natural background concentration (>150%; 6σ) for one-to-two consecutive years."  On the existing chronologies it sat at 768 CE, seven years off.  A second anomaly compatible with the 994 CE event sat at 987 CE.⁷

The diagnosis.  Five named causes, verbatim:

"1) low effective resolution of some ice core measurements (NGRIP, GRIP);  2) use of only single (or few) parameters for annual-layer interpretation (GRIP, Dye-3);  3) intra-annual variations in various ice-core parameters falsely interpreted as layer boundaries (e.g., caused by summer melt in Dye-3);  4) use of tephra believed to originate from the 79 CE Vesuvian eruption as a fixed reference horizon to constrain the Greenland ice-core dating;  5) use of manual-layer interpretation techniques that may favor interpretations consistent with a priori knowledge or existing chronologies (WDC)."⁷

Causes 2, 3 and 5 are precisely the failure modes the young-earth literature has been alleging for thirty years.  They are here conceded, in Nature, by the people who built the chronologies.

A correction that has to be made.  Vesuvius AD 79 is often cited, including in some old-earth writing, as a confirmation of ice-core dating.  It is the opposite.  Sigl and colleagues rejected the attribution:  "in view of the new annual-layer dating results, we concluded that this layer dates to 87/88 CE," and note that the documentary evidence has Vesuvian ash blowing toward the eastern Mediterranean, not Greenland.⁷  Do not use Vesuvius as an ice-core tie point.

The repair, and what it bought.  The revised chronologies (NS1-2011 for Greenland, WD2014 for Antarctica) are constrained by the 775 CE beryllium spike plus three historically dated dust-veil years and the 1257 Samalas eruption.  They differ from the old ones "by up to 11 years (GICC05, Greenland) and 14 years (WDC06A-7, Antarctica)," and are "accurate to within less than ±5 years during the past 2,500 years."⁷

The validations not used in building them.  Three:  the Changbaishan tephra found in Greenland ice at 946–947 CE against a documented winter 946/47 eruption and a radiocarbon wiggle-match of 946 ± 3 CE on a tree killed by it;  the 994 CE beryllium rise appearing at 993 in the ice, matching Japanese tree-ring carbon-14 which places it between the 993 and 994 growing seasons;  and the 32 dust-veil accounts above.⁷

The scale of the error, in the terms of this debate.  Seven years in 1,240 is 0.6%.  If the deep layers were storms at the rate a young-earth model requires — roughly fourteen per year — the 775 CE spike would not have been seven layers deep at 1,240 counted years.  It would have been about seventeen thousand layers deep.

Part 5: The young-earth case at full strength

Vardiman 1992: the founding argument

Larry Vardiman's Impact article set the shape of the creationist position, and it is more careful than its reputation.  Its structure:

1. The Lost Squadron aircraft were buried under ~250 feet in ~50 years, so ice can accumulate fast (raised, then immediately hedged — see below).
2. At Camp Century, annual layer thickness runs from about fourteen inches near the surface to less than two inches near the bottom; take the mean and a 4,000-foot ice sheet forms in under 6,000 years.¹
3. Greenland counting is easy near the top and cross-checks well against historic eruptions, but confidence falls with depth; Hammer and colleagues expected ambiguity to limit counting to "less than about 8,500" layers.
4. Antarctic ages of 160,000 years come from oxygen-isotope-to-temperature-to-accumulation-rate modelling, not from counting, and depend on the assumption that accumulation has not varied — which a post-Flood model denies. "Perhaps as much as 95% of the ice near the poles could have accumulated in the first 500 years or so after the Flood."¹

Points 3 and 4 are substantially correct as descriptions of the method.  Point 2 is arithmetic on a real depth–age relation.  Point 1 is the weak one, and Vardiman knew it.

Oard 2001: the deep core is storms

Michael Oard's TJ paper is the technical core of the position.  Its argument:

• Creationists and glaciologists agree about the top of the core. "Annual layers in the very top section of the core are easily interpreted from the ratio of oxygen isotopes, and creationists agree with these interpretations."⁸
• Below that the models diverge. In a post-Flood ice age with roughly 6 m water-equivalent accumulation per year in central Greenland, an annual layer at 2,000 m depth would be about 2,000 mm thick rather than the conventional 20 mm — a factor of a hundred.⁸
• The number of measurements made down the core is chosen according to the expected layer thickness, so the resolution of the instruments is set by the model. "Significantly, their interpretation of annual layers from these variables has been determined by the thickness of the annual layers that they expect, based on their model."⁸
• Sub-annual oscillations demonstrably exist: a storm has warm and cold sectors, snow dunes redistribute accumulation, and multi-week warm and cold spells are routine at high latitudes.
• His authority for this is the GISP2 team itself:

"Fundamentally, in counting any annual marker, we must ask whether it is absolutely unequivocal, or whether nonannual events could mimic or obscure a year.  For the visible strata (and, we believe, for any other annual indicator at accumulation rates representative of central Greenland), it is almost certain that variability exists at the subseasonal or storm level, at the annual level, and for various longer periodicities (2-year, sunspot, etc.).  We certainly must entertain the possibility of misidentifying the deposit of a large storm or a snow dune as an entire year or missing a weak indication of a summer and thus picking a 2-year interval as 1 year."²

That quotation is genuine and accurately reproduced.  It is also incomplete in a way that matters:  the sentences that follow it in Alley et al. 1997 go on to quantify the resulting error, at roughly 1% in the Holocene, 5% in the Wisconsinan, and under 2% against an independent uranium-thorium marker at 30,500 years.²  The GISP2 team raise the storm problem in order to bound it, not to concede that it is unbounded.

Woodmorappe 2002: two cores 28 km apart that disagree

The sharpest specific objection in the young-earth ice-core literature is not about the deep core at all.  GRIP and GISP2 were drilled 28 km apart in central Greenland.  Over a few centuries around the thirteenth millennium before present, their counted layers do not match, and the reconciliation study says so in terms:

"then the problem is not missing core or other 'block' data loss.  Rather, the GRIP core lacks about half the annual layers throughout this interval, or the GISP2 ice contains many subannual structures which mimic annual bands, or the layers are in fact annual but one of the counts is erroneous."¹¹

Woodmorappe's conclusion:  the correlation between cores is achieved by "ad hoc contractions and expansions (i.e. 'accordioned')," and "The lack of close correspondence between the two cores, especially in large segments, in spite of their geographic proximity, favours a catastrophic storm-dominated accumulation."¹¹

The observation is correct and it is not disputed.  What has happened since is that the field agreed with it and rebuilt the chronology.  GICC05 does not rest on GISP2 or on single-parameter counting;  it is built from DYE-3, GRIP and NorthGRIP together, with up to five parallel chemical signals, and its ambiguous layers are carried explicitly as the maximum counting error rather than resolved by fiat.  GICC21's abstract states the problem in Woodmorappe's own terms — "A layer counting bias was found in all ice cores because of site-specific signal disturbances;  therefore the manual comparison of all ice cores was deemed necessary" — and then quantifies what is left at one year per century.⁵

So the right reading of this episode is neither that the cores are unreliable nor that the discrepancy was imaginary.  It is that a real inter-core discrepancy of a few hundred years, in a 14,500-year record, drove a methodological rebuild whose residual error is now measured against external markers at 0.6%.

Hebert 2014–2021: resolution and circularity

Jake Hebert's contribution is the most recent and the most methodological.  His strongest single point:

"uniformitarian glaciologists increased their assigned age for the bottom of the GISP2 core from 85,000 years to 111,000 years simply by increasing the resolution of their dust measurements in the bottom 500 m of the core!"³

The fact is correct.  The inference — that the increase was driven by a desire to match the orbitally tuned marine record — is not obviously wrong and deserves a straight answer.  Here is one:  annual layers at that depth are about 4 mm thick, and the earlier dust measurements were made at 8 mm resolution.  An instrument that samples every 8 mm cannot resolve a 4 mm layer;  it must under-count, and by roughly a factor of two, which is roughly the size of the revision.  Increasing resolution to 1 mm was not a fudge, it was the necessary fix.  But the concern was legitimate and deserved that answer rather than a wave-off.

Hebert also states plainly what the young-earth programme still lacks:

"I think creationists have refuted claims that ice cores demand an old Earth.  However, we have done very little actual modelling of the ice sheets."³

There is also a mechanistic proposal that deserves noting, because it is one of the few in this literature that tries to explain an ice-core observation rather than to doubt it.  Larry Vardiman's 1997 paper argues that the long decline in the oxygen-isotope ratio through the "ice age" section of the cores need not record a hundred thousand years of ocean temperature change:  it can be produced in thousands of years by isotopic fractionation and trajectory dispersion as an ice shelf grows outward, since snow falling far from the shelf edge is isotopically lighter.¹²  Whether the mechanism can carry the observed amplitude has not been tested by anyone;  it is a hypothesis with a physical basis and no follow-up.

The existing model is Vardiman's 1994 one-dimensional analytic treatment, which assumes high ice-age accumulation decaying exponentially to modern rates and treats thinning as proportional to current thickness rather than solving for stress.³  Hebert has since computed annual-layer thicknesses within it and has published two further steps toward a physics-based treatment — a shallow isothermal ice dome with a frozen base in 2022, and an ice ridge in 2023.¹³  A 2025 paper by Kady Kunisaki takes up the sediment and ice-core records together within the same framework.¹⁴  That is honest, ongoing work, and it should be watched.  One result from it is worth reporting because it is counter-intuitive and comes from their own model:  "in creation-based ice models the thinnest annual layers are not necessarily found at the very bottom of the ice, but can occur at intermediate depths."³

The Lost Squadron

In July 1942, six P-38 fighters and two B-17 bombers made forced landings on the Greenland ice cap.  In 1992 one P-38 was recovered from beneath roughly 268 feet of snow and ice.  The argument writes itself:  five feet a year, so a 10,000-foot ice sheet takes 2,000 years.

Why it does not work.  The aircraft came down near the southeast coast, in the wettest sector of the Greenland ice sheet, where accumulation exceeds 1,500 mm of water equivalent per year — a figure mapped in the Journal of Glaciology the year before the recovery.¹⁵  The deep cores are drilled at the interior divide, which receives roughly a sixth of that.  The 268 feet is mostly firn, not ice:  about 53 m water equivalent, which is an ordinary fifty years for that location.  At the SE-Dome site 220 km away, the 1963 nuclear-test tritium horizon lies at 81 m depth — the same burial in the same fifty-odd years, dated three independent ways.¹⁶

Who declines the argument is the tell.  Vardiman raised it in 1992 and warned in the next paragraph that "life is never as simple as implied above," listing compaction, horizontal thinning, and site-to-site precipitation differences.¹  Oard's technical ice-core papers do not use it.  Hebert's 2018 treatment concedes the apples-to-oranges charge and demotes the aircraft to an illustration of post-Flood snowfall rather than a refutation of ice-core dating.  Creation Ministries International's own article has carried an editorial note for well over a decade conceding that published accumulation rates make the burial depth unsurprising.¹⁷

It survives only in popular presentations.  It should be dropped.

Part 6: The problem in the upper core

The debate usually concentrates on the deep, thin, ambiguous layers.  The harder problem for a young-earth reading is in the thick, easy, unambiguous ones.

Take the GISP2 depth–age table, which is public.⁹  The Holocene–Pleistocene boundary sits at 1,678 m, with 11,651 counted years above it.  In an Oard-style chronology the ice age ended about 4,000 years ago;  everything above that boundary therefore fell in the last ~4,000 years, under climate not very different from today's.

That is the regime where all sides agree the counting works.  The layers are thick.  The isotope cycle is clean.  Vardiman calls counting there "relatively easy."  Hebert allows agreement "to within a factor of two or three."  But a factor of two or three is not enough:  11,651 counted years in ~4,000 real years is a factor of about 2.9, and it has to be achieved in ice that fell after the high-snowfall, high-storm conditions that the model invokes to explain over-counting had ended.

And the requirement gets worse if the model is given the part it most obviously deserves.  The same table puts 500 m depth at 2,097 counted years.  Grant that the top 500 m is counted perfectly — recent ice, thick layers, historic volcanic markers all through it — and the remaining 9,554 counted years have to fit into 4,000 − 2,097 = 1,903 real years.  That is a required over-count of 5.0 in the ice between 500 m and 1,678 m.⁹

More snowfall does not help.  Thicker layers are easier to resolve, not harder.  The mechanism the model needs — many storm cycles miscounted as years — works best where layers are thin and ambiguous, which is exactly where this ice is not.

This is, in the assessment of this article, the strongest single unanswered objection to the young-earth reading of the ice cores, and it is computable from a table anyone can download.

Objections and responses

"Antarctic ice is 800,000 years old, so the earth is old." Not by counting it is not.  East Antarctic accumulation is too low for annual layers to survive to depth;  those ages come from flow models and orbital tuning.  The young-earth critique is correct on this point and it should be conceded plainly.  The counting evidence is Greenland, and it stops at 60,000 years.

"Layers can be counted wrong, and the team admits it." They do admit it, they quantify it, and the quantification is the point.  Ambiguous layers are counted at half weight;  the accumulated ambiguity is 1% in the Holocene and 5% in the glacial;  and the residual error against external clocks has been measured at 0.6% over the last 1,240 years and about 2% against argon-argon at 41,000.  A method whose error is measured is not thereby refuted.

"A single storm can produce a false layer." Yes, in one signal.  The counting requires the signals to agree, and they come from different continents and different physical processes.  A North Atlantic storm has no mechanism for putting a spurious ammonium peak into snow whose ammonium comes from North American wildfire.

"Snowfall was far higher after the Flood, so the deep layers are storms." This is the real argument, and it needs a number.  Sixty thousand counted layers in about 4,300 years requires roughly fourteen countable oscillations per year, sustained, in five parallel signals, for the entire depth of the core.  No published creationist model produces that figure, and Hebert states that ice-sheet modelling on the creationist side has barely begun.

"The count was raised from 85,000 to 111,000 years to match the marine record.  That is circular." The revision followed an eightfold increase in dust-measurement resolution, from 8 mm to 1 mm, in a section where layers are about 4 mm thick.  The original measurement physically could not resolve those layers.  The suspicion was reasonable;  the explanation is mechanical.

"The Vesuvius layer proves the chronology works." It does not, and this claim should be dropped by whoever is making it.  The Vesuvius attribution was withdrawn in 2015;  the layer dates to 87/88 CE, and using it as a tie point was one of the named causes of the seven-year error.

"They tuned the ice chronology to the tree rings, so of course they agree." The direction of the correction refutes this.  The chronology was not tuned to the tree rings before 2015 — that is why it was seven years wrong.  The tree-ring marker was applied from outside, found a discrepancy, and forced a revision.  A tuned system does not do that.

"The Lost Squadron planes were under 268 feet after fifty years." They were, at a site receiving more than 1,500 mm water equivalent a year.  The cores are drilled where accumulation is about a sixth of that.  The argument has been withdrawn or hedged by every creationist author who has examined it closely.

Issues and unanswered questions

• The "shape of annual layers" bias is unquantified, by the counters' own statement, and is not included in the stated uncertainties.⁴
• Around 30,000 years the Greenland scale carries about 1,000 uncertain years, and around 40,000 there is an unresolved conflict of about 1,000 years between a floating New Zealand kauri sequence placed by beryllium-10 against the ice and the same sequence placed by carbon-14 against Hulu Cave. The specialists conclude the ice placement is the wrong one.¹⁸
• GICC05 systematically over-counts the Holocene by about 65 years — 0.55%, in the direction of too many layers, not too few.¹⁸ Two independently counted chronologies in opposite hemispheres differ by 0.63% in accumulated years, and IntCal20 applies a blunt ×1.0063 correction to reconcile them.
• Moomi Cave, Socotra, does not fit. It agrees on one interstadial and deviates by more than a thousand years on the adjacent ones. No explanation has been published.
• Alley et al. 1997 has not been read in full for this article. The quotation Oard uses and the following sentences that quantify the error are verified; the rest of the paper has not been obtained directly.

Verdict

Ice-core layer counting is a good clock over a limited range and a poor one outside it, and both halves of that sentence get lost in popular argument.

The young-earth critique gets three things right.  Antarctic deep ice is not counted.  Greenland counting is genuinely difficult at depth and its error was genuinely underestimated.  And the field has, on the record, tuned chronologies to markers that turned out to be misidentified.

It gets the conclusion wrong for a reason that can be stated in one line:  the residual error, after all of that, has been measured against three unrelated decay systems and one absolutely dated cosmic-ray event, and it comes out between 0.6% and 5%.  A young-earth chronology needs it to be about 1,300%.

And the sharpest problem is not in the disputed deep ice at all.  It is in the top 1,678 metres, where everyone agrees the counting works, and where a post-Flood model still has to explain 11,651 layers in about 4,000 years.

Further reading

• Non-Radiometric Clocks — the overview of this series
• Tree-Ring Dating — the 774 CE marker that corrected the ice chronologies
• The Ice Age — the post-Flood ice-age model this argument sits inside
• Radiocarbon Dating

Sources