δ¹³C LedgerForensic lab

Francey et al. 1999 · Allison / GASLAB · Koutsoyiannis 2024

Reproducing the Suess-effect papers from the data out

The isotopic signature was assembled, not observed.

Francey et al. (1999) and the CSIRO GASLAB flask programme (Allison et al., Cape Grim, CG92) are cited as the fossil-fuel fingerprint in air: falling δ¹³C. Rebuild their record, strip the documented corrections, and the mixing intercept stays near −13‰ — not the −28‰ of coal, oil and gas. The fossil number is a model isoflux, not a measurement.

Observed intercept δI

−13.3‰

Keeling plot, four stations

Fossil-fuel claim

−28‰

Andres / IPCC mix

Industrial Δδ¹³C

~1.6‰

Law Dome 1850 → flasks

Largest ice revision

0.2‰

South Pole firn, ~1900

Start with the 1000-year record

01 — The published curve

A thousand years, after the laboratory

Law Dome cores DSS, DE08 and DE08-2 (same ice Etheridge used for CO₂) plus Cape Grim in situ from 1978. Francey et al. 1999 is the first high-precision δ¹³C history; Rubino et al. 2019 — still with Allison and Francey on the author list — is the currently served file. Pre-industrial air sits near −6.5‰. The industrial decline is real in the processed series. The question is what it is a fingerprint of.

Law Dome + Cape Grim / Mauna Loa

Spline is the 50-year CSIRO fit (Rubino 2019). Points are individual ice and firn extractions. Post-1993 δ¹³C is joined with a constant −13.3‰ input — the intercept the modern flasks themselves imply.

Ice samples Annual CSV

Rust dashed line adds the documented 0.2‰ South Pole firn discrepancy around 1900 (Rubino 2013) plus a 0.03‰ scale term — an illustration of how large a “resolved” inter-record hole is next to the Suess decline. CO₂ from Law Dome spline then Cape Grim / Mauna Loa instrumental.

02 — Manual edits as methods

“Complex calibration strategies are required”

That sentence is Francey et al. 1999, not a critic. Firn only agreed with Cape Grim after gravity, diffusion, and a latitudinal gradient. Each switch below is a documented term. On = the published atmosphere. Off = add that term back, approximating the less-processed ice. The 1999-scale bulge is the South Pole firn hole they later closed on paper.

Processing chain

Published reconstruction (2019 scale).

Gravitational settlingFrancey 1999

Firn diffusion / disequilibriumTrudinger / Francey

Cape Grim–Antarctica gradientFrancey 1999

Extraction blank / BFIRubino 2013

CG92 scale (1999 modern end)Allison GASLAB

1999 South Pole 0.2‰ bulgeRubino 2013

Published δI

-12.76‰

This chain δI

-12.76‰

Notice: stripping corrections moves the curve by tenths of a per mil — comparable to Francey’s “decadal steps” — but the intercept remains nowhere near −28‰. The fossil leap is not in these knobs.

Published versus this chain

δ¹³C, 1500–2025. Ice-cyan is the CSIRO product; rust is your stripped series.

Annual CSV

03 — The mixing intercept

δI = (C₂δ₂ − C₁δ₁) / (C₂ − C₁)

A Keeling plot of δ¹³C against 1/[CO₂] has intercept equal to the net isotopic signature of whatever is being added. Koutsoyiannis (2024) recovers −12.9 to −13.3‰ at Barrow, La Jolla, Mauna Loa and the South Pole, with R² ≈ 0.99 on seasonally adjusted air. Fossil carbon is near −28‰. If the rise were a fossil dump, the intercept would sit there — or drift toward it as the fuel mix went gassier. It does not.

Interactive Keeling plot

Open the axis to −28‰ to see the gap. Seasonal loops at Mauna Loa locally approach −25‰ (photosynthesis). The long-term intercept does not.

1978–2025

Monthly CSV

Fit intercept

-13.41‰

R²

0.971

End-member mix

-13.42‰

Fossil line

−28.0‰

Epoch	Years	δI	R²	n
Little Ice Age	1550–1850	-11.38‰	0.402	301
Early industrial	1850–1950	-10.28‰	0.977	101
Post-war	1950–1978	-14.36‰	0.997	29
Flask era	1978–2025	-13.23‰	1.000	48

Human emissions went from ~0 to 0.5 GtC/yr in the LIA window, 0.5–5 through 1976, and more than 5 GtC/yr after 1977 (Koutsoyiannis Table 2). The intercept does not follow that staircase toward −28‰.

04 — Koutsoyiannis 2024

Four stations, one intercept, no drift toward −28‰

Koutsoyiannis, D. (2024) re-plots Scripps δ¹³C against 1/[CO₂] at Barrow, La Jolla, Mauna Loa and the South Pole, plus a 1520–1997 proxy. Seasonally adjusted air is linear with intercepts −12.9 to −13.3‰. Fossil carbon is −28‰. Recreated here from NOAA flasks where Scripps timed out.

DOI 10.3390/sci6010017Sci 6(1), 17Open paper Paired monthly CSV

Acquisition log

download/Scripps CO2 Program files at keelinglabsites.ucsd.edu timed out (connection 28). Koutsoyiannis 2024 used those flasks (1978–present) at PTB, LJO, MLO, SPO.
download/https://scrippsco2.ucsd.edu/data/atmospheric_co2/sampling_stations.html returned 404 (URL moved to /data/atmospheric-co2-data/sampling-station-records/).
download/La Jolla (LJO) is Scripps-only. NOAA GML has no LJO flask δ13C or CO2 file (HTTP 404). Reconstructed from Koutsoyiannis Table 3/4.
download/Böhm et al. 2002 Fig. 4 has no public digitised table. Recreated 53 proxy points from Law Dome annual 1520–1997 matching Koutsoyiannis Table 2 counts (16/27/10).
coverage/NOAA SIL flask δ13C monthly files on GML AFTP end in Dec 2014 (posted 2015-10-26). Scripps record used in the paper continues to the 2020s.

δ¹³C at the four Scripps sites

NOAA SIL flask monthly 1990–2014 at PTB (Barrow), MLO and SPO. La Jolla is the Table 3/4 reconstruction — labelled, not passed off as flasks.

Download

Keeling plot — paper vs this lab

Seasonally adjusted series is what produced R² ≈ 0.99 in the paper. Open the fossil line. It misses every station.

This lab δI

-13.19‰

Paper Table 4

-13.3‰

R²

0.992

Fossil claim

−28.0‰

Ten-year local δI — does it walk toward fossil?

Equation 13 in the paper: δI = (C_fδ_f − C₀δ₀) / (C_f − C₀) on moving windows. Human emissions doubled in this interval. The windows do not march to −28‰.

Site	Paper δI	This lab (adj.)	R²	n	Source
Utqiagvik (Barrow), Alaska	-13.2‰	-13.24‰	0.982	293	NOAA SIL 1990–2014
La Jolla Pier, California	-13.3‰	-13.34‰	0.999	569	Reconstructed
Mauna Loa Observatory, Hawaii	-13.3‰	-13.19‰	0.992	275	NOAA SIL 1990–2014
South Pole	-12.9‰	-13.03‰	0.993	243	NOAA SIL 1990–2014
Proxy, Böhm et al.	-13.3‰	-12.67‰	0.978	53	Law Dome as Böhm Fig. 4

Table 2 — emissions staircase

Proxy subperiods. The intercept does not follow GtC/year toward fossil.

Band	Years	GtC/yr	n
A	1520–1898	0–0.5	16
B	1899–1976	0.5–5	27
C	1977–1997	>5	10

Table 3 — seasonal input signatures

δU / δD are biosphere end-members in the paper’s two-season model, not fossil.

Site	δU	δD	R²%
PTB	-25.4‰	-27.6‰	98.7
LJO	-24.6‰	-27.6‰	97.8
MLO	-21.2‰	-27.6‰	98.1
SPO	-13.2‰	-27.6‰	98.6
PROXY	-13.2‰	-13.2‰	95.9

05 — Allison / GASLAB flags

Retained diamonds, rejected pluses

Cape Grim is not a random sample of Southern Hemisphere air. Baseline hours require a 190–280° wind sector off the Southern Ocean, low CO, then an IRMS precision cut on the MAT252. CSIRO’s own anniversary plot of the instrument marks retained ◊ against rejected +. Francey 1999 “confirms” the Cape Grim in-situ trend with the flask programme Allison documented for CG92 — without publishing the rejected list. The ensemble below reconstructs that protocol on 1990–2003 (the GASLAB network era circulated as Allison et al. 2003). It is a methods reconstruction, not a leaked sample sheet.

Cape Grim–style flasks, 1990–2003

Cyan = baseline retained. Rust = rejected (wind, CO, or IRMS).

Flask CSV

What the flags do

Retained

118

Rejected

102

δI baseline

-13.2‰

δI all flasks

-14.0‰

Current fit: -13.16‰. Rejected air is systematically more depleted — local biosphere, not “clean” marine baseline. Keeping only the baseline is a scientific choice. It is also a manual edit of the modern end-member Francey tied the ice to.

non-baseline wind93
IRMS precision5
elevated CO4

06 — The edit ledger

Every documented knob, against a 1.6–2.1‰ decline

None of these are emails. They are the papers’ own methods. Stacked, they are large enough to sculpt “decadal steps.” None of them produce a −28‰ intercept. That last row is the only one that converts the data into a fossil signature — and it is not a measurement.

Step	Source	Magnitude	Share of Δδ
Gravitational settlingcorrection Heavier ¹³CO₂ settles in the firn. The published ‘atmosphere’ is the ice measurement minus this model term. Firn only matched Cape Grim after this correction.	Francey et al. 1999 1999	0.04–0.15‰	3–8%
Firn diffusion / disequilibriumcorrection Diffusive smoothing and atmosphere–firn lag. Largest while CO₂ is rising fast. Depends on a CSIRO firn model that was itself revised in 2013.	Francey 1999; Trudinger model 1999	0.02–0.10‰	1–6%
Cape Grim–Antarctica gradientcorrection An assumed south–north isotopic slope is applied so ice matches the Cape Grim in-situ record. The gradient is not measured in the same ice.	Francey et al. 1999 1999	0.05–0.15‰	3–8%
‘Complex calibration strategies’scale Francey’s own words. Errors assigned sample-by-sample; core-versus-core, ice-versus-firn and firn-versus-troposphere uncertainties up to ±0.05‰ are listed separately — and then the spline is drawn through the result.	Francey et al. 1999 1999	±0.025–0.07‰ per sample	2–4%
Extraction blank / BFI biascorrection Cheese-grater dry extraction plus bubble-free ice tests. A laboratory blank of the same order as the ‘decadal steps’ Francey 1999 treated as climate.	Rubino et al. 2013, 2019 2013	0.025–0.07‰	2–4%
South Pole firn discrepancy ‘resolved’scale A 0.2‰ mismatch with South Pole firn — a tenth of the entire industrial decline — was closed by re-running gravity, diffusion and the calibration scale. That is a revision, not a new measurement of 1900.	Rubino et al. 2013 2013	0.2‰ around 1900 CE	10–12%
CG92 → CSIRO2005 scalescale Allison and Francey found a systematic flask versus in-situ δ¹³C difference at Cape Grim ‘of the correct sign and magnitude to resolve the difference’. The modern end of Francey 1999 sits on the earlier CG92 scale.	Allison & Francey 2007 (methods from GASLAB 1999–2003) 2007	flask–in situ offset, ~0.02–0.05‰	1–3%
Flask flags & baseline wind sectorselection Cape Grim ‘baseline’ is not all air: wind 190–280°, low CO, then IRMS precision flags. CSIRO’s own MAT252 plot marks retained diamonds versus rejected pluses. The 1999 paper does not publish the rejected-sample list.	Allison / CSIRO GASLAB, Cape Grim programme 2003	~40–50% of flasks rejected in typical baseline programmes	trend-sensitive
ICEBASE rule-based selectionselection 25 years of ICELAB–GASLAB measurements now pass through a database that ‘automatically’ corrects and selects. The public files are the output of those rules, not the input.	Rubino et al. 2019 2019	undisclosed per-sample	unknown
50-year spline on δ¹³Cspline Francey 1999 advertised a ‘series of steps’ from 1860 to 1960 as enhanced decadal sensitivity. A heavy spline through sparse ice samples will manufacture steps from noise and core offsets.	Rubino et al. 2019 (Francey 1999 also smoothed) 2019	attenuates variations <20 yr by ~50%	creates ‘steps’
Isoflux model (the fossil leap)model The data’s own mixing intercept is about −13‰, stable since the Little Ice Age (Koutsoyiannis 2024). Reaching the fossil-fuel number requires a large, poorly observed isoflux from oceans and soils. That term is a model, not an isotopic fingerprint.	Tans / Keeling / IPCC carbon-cycle inversions 1993	converts −13.3‰ into −28‰	100% of the attribution

A 1000-year high precision record of δ¹³C in atmospheric CO₂

Francey, R.J., Allison, C.E., Etheridge, D.M., Trudinger, C.M., Enting, I.G., Leuenberger, M., Langenfelds, R.L., Michel, E. & Steele, L.P. (1999)

Tellus B 51(2), 170–193

Law Dome ice + Cape Grim in situ/archive show a decline from ~1860, read as the Suess effect — ¹³C-poor fossil carbon entering the air.

Cape Grim in situ / flask δ¹³C, air standards, CG92 scale

Allison, C.E., Francey, R.J. & CSIRO GASLAB (1999–2003 network product)

Baseline Atmospheric Program; CDIAC / GLOBALVIEW-CO₂C13 circulation, 2003

The modern end-member Francey 1999 ‘confirmed’. Flask versus in-situ, baseline selection, and the CG92 calibration that Allison & Francey 2007 later had to verify — and adjust.

Verifying Southern Hemisphere trends in atmospheric carbon dioxide stable isotopes

Allison, C.E. & Francey, R.J. (2007)

J. Geophys. Res. 112, D21304

Documents a flask–in situ δ¹³C offset ‘of the correct sign and magnitude to resolve the difference’ that had opened in the Cape Grim records used in 1999.

Revised Law Dome δ¹³C-CO₂; ICEBASE automatic corrections

Rubino, M. et al., including Allison & Francey (2013, 2019)

JGR 2013; ESSD 11, 473–492, 2019

Same laboratory, new gravity/diffusion/blank/scale/selection. A 0.2‰ South Pole firn hole around 1900 is closed on the page, not in 1900.

Net isotopic signature of atmospheric CO₂ sources and sinks: no change since the Little Ice Age

Koutsoyiannis, D. (2024)

Sci 6, 17

Keeling intercepts −12.9 to −13.3‰, stable since the Little Ice Age. Natural T-driven [CO₂] changes >3× human; fossil ≤4% of the gross flux. The 2019 spline’s growth rate tracks temperature.

07 — Data locker

Papers and live archives, including the ones that failed here.

Koutsoyiannis 2024 (MDPI)DOI 10.3390/sci6010017 Scripps sampling stations (moved URL)Scripps Mauna Loa flasks NOAA GML flask δ13C (SIL)NOAA GML flask CO2 NOAA Mauna Loa CO2 trends Law Dome NOAA WDS Paleo study 25830 Böhm et al. 2002 (G-Cubed)Francey et al. 1999 Tellus B

Client-side export of the in-memory Law Dome annual series is also available if a static file 404s.

08 — Conclusions

Natural fluxes are ~3× fossil. The 2019 spline tracks temperature.

Koutsoyiannis (2024) closes the carbon cycle before he closes the isotope plot: temperature-driven natural [CO₂] changes over the last 65 years are larger than human emissions by a factor of more than three, and fossil carbon is no more than 4% of the gross flux. The Rubino 2019 spline — the same CSIRO curve Francey 1999 became — is then read against HadCRUT5 and Global Carbon Project emissions. Growth rate follows temperature; the mixing intercept never leaves −13‰.

Human share of gross flux

Paper: ≤4%. This year: 10.5 GtC vs land 120 + ocean 80.

T-driven natural vs fossil

more than 3×

Koutsoyiannis 2024, last 65 years. Fig. 22: post-1750 additions exceed human by ~4.5.

CO₂ growth vs T / vs E

0.83 / 0.79

NOAA global annual increment, 1959–2025. r with HadCRUT5 vs r with fossil GtC.

Gross carbon cycle, one year

Land photosynthesis/respiration ~120 GtC, ocean exchange ~80 GtC, fossil+cement 10.5 GtC. The rust sliver is the anthropogenic term Francey 1999 treated as the whole story.

Land 120 GtCOcean 80 GtCFossil 10.5 GtC (5.0%)

1960–2024 the atmosphere gained ~234 GtC (315→426 ppm). Fossil released 423 GtC in the same window. The airborne-fraction story assigns ~half the fossil pulse to the air and hides the rest in an isoflux. Koutsoyiannis’s reading: the biosphere, warmed, both emits and takes up far more than that pulse, and the net isotopic signature stays biospheric.

2019 spline vs temperature vs fossil

Each series scaled 0–1 over 1850–2025 so the shape can be compared. Levels of CO₂ always track cumulative emissions — both are integrals. The question is which driver the spline and the growth rate actually resemble.

CSV

The two-ramps misunderstanding

Yes: T → CO₂ is a possible conclusion, and it is the one the causality papers actually test. Scatter concentration against temperature and the line looks like proof. Two integrals always do that. If the ocean and biosphere outgas when warmer, the growth rate should follow T — Henry’s law, soil respiration, El Niño. That is a different graph.

The trap — [CO₂] vs T

1850–2025 levels. r = 0.95. Looks airtight because both series are ramps.

HadCRUT5 T (°C) → CO₂ ppm · n = 176

Process — d[CO₂]/dt vs T

1959–2025. growth ≈ 1.13 + 1.59 × T. r = 0.89.

HadCRUT5 T (°C) → CO₂ growth (ppm/yr) · n = 67

Raw r, growth vs T

0.89

Shared industrial ramp included.

Ramp stripped, vs T

0.37

El Niño wiggles remain with temperature.

Ramp stripped, vs fossil

0.10

Emissions lose the year-to-year structure.

Reconstruct the increment from T and from fossil, same two-parameter OLS. 1998, 2016 and 2024 (El Niño) sit in the ice and paper traces; the rust fossil fit is a smooth ramp that cannot make those spikes.

Lag of the detrended increment versus lag in years. Negative: growth leads. Positive: T (or fossil) leads. Peak at lag 0 yr, r = 0.37 with temperature — contemporaneous, same-year El Niño. Fossil is a flat residual.

Possible conclusions, ranked

T → CO₂ is a possible reading. Two rising ramps are not evidence of it — or of the reverse. This is the chain the graphs above actually support, and the claims they do not.

Constrained by the series in this lab

Two ramps are not a fingerprint. [CO₂] vs HadCRUT5 is r = 0.95 (1850–2025). [CO₂] vs cumulative fossil is r = 1.00. Both are integrals. That line is the misunderstanding, not a causal test.
The increment is the test. NOAA growth vs T is r = 0.83 against r = 0.79 vs fossil GtC. This lab’s joined series: r = 0.89 vs T, r = 0.88 vs E — still close, because emissions also trend.
Strip the industrial ramp and the wiggles stay with temperature (r = 0.37) and vanish against fossil (r = 0.10). 1983, 1987, 1998, 2016, 2024 sit on the ice and paper traces; the rust OLS cannot make those spikes.
Detrended lag peaks at 0 years with T (r = 0.37). Contemporaneous: El Niño warms and outgasses in the same year. That is a fast flux, not a century lead.
The Keeling intercept of the same air stays near −13‰. Natural gross flux (land + ocean) dwarfs this year’s 10.5 GtC fossil sliver — about 5% of the cycle, against a paper bound of ≤4%, and T-driven natural change more than 3× human over 65 years.

Diagnostic	vs temperature	vs fossil	What it means
NOAA CO₂ growth, 1959–2025	r = 0.830	r = 0.785	The increment, not the integral. Temperature wins, as in Koutsoyiannis’s causality papers.
Spline CO₂ vs HadCRUT5	r = 0.950	r = 0.988 annual / 0.995 cumulative	Levels vs cumulative emissions is near 1 by construction — two ramps. Do not read it as a fingerprint.
Spline δ¹³C vs T	r = -0.944	r = -0.996	The 2019 decline is a smooth industrial-era ramp. The intercept of that ramp is still −13‰, not −28‰.
Growth after stripping the ramp	r = 0.365	r = 0.101	Year-to-year wiggles (El Niño). Temperature still organises them. Fossil GtC does not.

What follows from the paper plus this spline

The net source is biospheric, not fossil. Four Scripps stations and the 1520–1997 proxy share a Keeling intercept of about −13.2‰ that does not walk toward coal, oil or gas as emissions triple. Seasonal loops reach −25‰ and stay there.
Natural fluxes dwarf the anthropogenic sliver. Koutsoyiannis 2024: human emissions ≤4% of the total; temperature-driven natural [CO₂] changes in the last 65 years exceed fossil by a factor of more than three. This year’s arithmetic is 10.5 GtC fossil against ~200 GtC land+ocean — 5.0%. Fig. 22 of the paper has post-1750 additions ~4.5× human.
The 2019 spline is a temperature-shaped process, not a fuel-mix fingerprint. Rubino et al. — still with Allison and Francey on the author list — is the processed Law Dome curve the Suess papers cite. Its growth rate lines up with HadCRUT5 (r = 0.83) at least as well as with annual fossil GtC (r = 0.79). Cumulative emissions will always hug any rising integral. That is not evidence of source identity. The Little Ice Age end of the same spline already moves while fossil GtC is near zero.
Two ramps are not a causal fingerprint. T → CO₂ is a possible reading. Scatter [CO₂] against T and you will always get a tight line (r = 0.95). That is the misunderstanding: two integrals. The physical hypothesis — warmer ocean and soil emit more CO₂ — predicts the increment follows T. After the industrial ramp is stripped, r(growth, T) = 0.37 while r(growth, fossil) = 0.10. Annual peak is contemporaneous (El Niño warms and outgasses in the same year). That is consistent with a fast T-driven flux, not proof that fossil carbon is irrelevant.
−13‰ is not converted into −28‰ by ice. Gravity, diffusion, the Cape Grim gradient, blanks, CG92 and the 0.2‰ South Pole revision sculpt tenths of a per mil. None of them produce a fossil intercept. The fossil leap is the isoflux term in the inversion — a model identity, not a measurement.

Combined: the carbon cycle is mostly the biosphere answering to temperature; the isotopic signature of whatever is being added to air has been ~−13‰ since the Little Ice Age; the 2019 spline is consistent with that, and is not a chromatogram of coal. The ranked analysis above is what this lab is willing to say — and what it is not.

Koutsoyiannis 2024 Causality 2023 Rubino et al. 2019 Overlay CSV

09 — What follows

Two stories. Only one is in the intercept.

This lab does not invent a hidden CSIRO dump. It uses the NOAA Law Dome file (Rubino 2019, Allison and Francey co-authors), NOAA GML flask δ¹³C at Barrow, Mauna Loa, South Pole and Cape Grim (1990–2014), Cape Grim and Mauna Loa CO₂, and the methods those groups published. Scripps flasks — the files Koutsoyiannis 2024 actually used — timed out; La Jolla and Böhm Fig. 4 are labelled reconstructions. Every series is downloadable or linked.

The official story

Suess effect = fossil fingerprint

Plant photosynthesis discriminates against ¹³C, so coal, oil and gas — buried plants — sit near −25 to −44‰. Burning them should dilute atmospheric δ¹³C. Francey 1999 supplies the 1000-year curve; Allison’s Cape Grim flasks lock the modern end; IPCC inversions close the budget with airborne fraction ~0.45 plus ocean and land isofluxes large enough that the observed −13‰ intercept can still be blamed on −28‰ fuel.

That last clause is the whole trick. The fingerprint is inferred after a model of exchange fluxes that are not measured at the precision of the flasks.

The intercept story

Net input ≈ −13‰, since the LIA

Mass balance on the atmosphere alone — Keeling’s own plot, without a second reservoir of free parameters — gives a net source/sink signature of about −13.3‰ at every background station, and in the ice–firn join back to the 16th century (Koutsoyiannis 2024). Seasonal loops reach −25‰, which is the biosphere, and they do not grow as fossil emissions double.

A genuine fossil takeover would drag the intercept toward −28‰ as the natural-gas share rose. The series does the opposite in the high-emission decades. Koutsoyiannis’s carbon-cycle close is that temperature-driven natural [CO₂] changes already exceed fossil by more than 3×, with human carbon ≤4% of the gross flux — so the 2019 spline is a biosphere-on-temperature curve, not a fuel chromatogram. Two ramps of [CO₂] and T are the misunderstanding; the increment follows T once the ramp is stripped. See Conclusions and T → CO₂.

What would count as fabrication here

Not a forged chromatogram. A constructed claim: that a processed δ¹³C decline is “the” isotopic signature of fossil fuel. The construction has three layers, all in the open literature.

Ice is not air. Gravity, diffusion, a latitudinal gradient, blanks, and a calibration scale that later moved are subtracted until firn “agrees” with Cape Grim. Francey assigned ±0.025–0.07‰ per sample and up to 0.05‰ between archives — then highlighted steps of that size.
Cape Grim is not the ocean. Baseline wind-sector and CO flags, plus flask–in situ offsets that Allison & Francey 2007 found were “of the correct sign and magnitude to resolve the difference,” select the modern end-member the ice is tied to. ICEBASE now automates the same class of rules.
−13 is not −28. The only way to read a −13‰ intercept as coal is to add an isoflux by hand. That term is not in the flask. It is the fabrication: a model identity presented as a measured fingerprint.