Comprehensive inventory of major atmospheric CO₂ data sources from direct measurements to deep-time proxies
Purpose: This catalog documents major published sources of atmospheric CO₂ data across geological time, organized by measurement type and time range. Use this as a reference before building charts or analyzing CO₂ trends.
Coverage: From 540 million years ago (Cambrian period) to present day.
Quality tiers: Direct instrumental → Ice-core trapped air → Vetted proxy synthesis → Archive proxy data.
Highest quality: direct atmospheric sampling using infrared spectroscopy and related techniques.
| Source Name | Tier | Time Range | Resolution | Access URL | Notes |
|---|---|---|---|---|---|
| NOAA GML Mauna Loa | Direct | 1958 CE – present | Monthly | co2_mm_mlo.csv | The canonical Keeling Curve. Monthly averages from Mauna Loa Observatory, Hawaii. |
| NOAA GML Global Network | Direct | 1973 CE – present | Weekly/Monthly (flasks), Continuous (in-situ) | NOAA Data Viewer | Multiple stations worldwide. Barrow continuous measurements started 1973, Mauna Loa flasks 1974. Includes Cape Grim, South Pole, American Samoa, and many other sites. Both flask samples and continuous in-situ measurements. |
| Scripps CO₂ Program | Direct | 1958 CE – present | Monthly | Scripps CO2 data | Original Keeling program data, maintained independently from NOAA. Multiple stations worldwide. |
| Cape Grim Baseline Station | Direct | 1976 CE – present | Continuous + flask | CSIRO Cape Grim | Tasmania, Australia. Southern Hemisphere baseline station. CSIRO/BoM operation. Critical for SH CO₂ monitoring. |
| NOAA GML Data Portal (All Sites) | Direct | 1973 CE – present (site-dependent) | Site-dependent | NOAA GML Data Viewer | Comprehensive portal for all NOAA flask and in-situ measurements. Includes South Pole, American Samoa, Barrow/Utqiaġvik, and dozens more sites. |
| AGAGE Network | Direct | 1978 CE – present | Continuous (high frequency) | agage.mit.edu | Primarily a halocarbon and methane network, but several sites (Mace Head, Trinidad Head, Cape Grim, Ragged Point) also report CO₂. Independent of NOAA GML and Scripps. Useful cross-check for inter-network calibration and station contamination screening. |
These independent signals corroborate the instrumental CO₂ record and confirm the fossil-fuel origin of the modern rise. They are not concentration data sources but are essential context for defensible chart-building.
| Signal | Physical basis | Time range | Source | Significance |
|---|---|---|---|---|
| δ¹³C Suess Effect | Fossil fuel carbon is depleted in ¹³C relative to the atmospheric CO₂ background. As fossil CO₂ enters the atmosphere the ¹³C/¹²C ratio declines measurably. | 1950s CE – present (direct); ice cores, tree rings, corals extend the record further | Scripps ¹³C record; NOAA GML stable isotopes programme | Rules out volcanic or biospheric sources as primary drivers; neither produces the observed ¹³C depletion signature. Keeling 1979; Francey et al. 1999. |
| O₂/N₂ Ratio Decline | Combustion consumes atmospheric O₂. The O₂/N₂ ratio has declined consistently since direct measurements began in 1989, consistent with combustion stoichiometry (~1.4 O₂ consumed per CO₂ released, varying by fuel type). | 1989 CE – present | Scripps O₂ Program (Ralph Keeling, SIO) | Closes the mass balance and rules out oceanic degassing as the primary source — ocean outgassing would not consume atmospheric O₂. |
| Ocean Surface pCO₂ (SOCAT) | Surface ocean pCO₂ has risen in close proportion to atmospheric CO₂, consistent with Henry's Law equilibrium. Measured in a completely different medium from atmospheric monitoring. | 1950s CE – present | SOCAT; Bakker et al. 2016, Earth Syst. Sci. Data | >30 million shipboard and autonomous pCO₂ measurements across all ocean basins. Confirms the atmospheric CO₂ signal is a genuine global change, not a station artefact. |
High quality: ancient atmosphere preserved in air bubbles trapped in glacial ice. Age uncertainty increases with depth.
Antarctic cores are preferred for atmospheric CO₂ reconstruction because polar air masses are well mixed (so a single high-latitude site captures hemispheric/global CO₂ to within a few ppm), and Antarctic ice has substantially lower carbonate dust contamination than Greenland cores.
| Source Name | Tier | Time Range | Resolution | Access URL | Notes |
|---|---|---|---|---|---|
| Law Dome (DSS, DE08, DE08-2) | Ice Core | ~2000 BP – 2006 CE | ~20 years | lawdomeco2.txt | High-resolution Antarctic ice core. Bridges pre-industrial and modern periods. MacFarling Meure et al. 2006. |
| Law Dome Spline (2006) | Ice Core | 1006 CE – 1978 CE | Smoothed | law2006.txt | 20-year spline of Law Dome data. Often used for smooth pre-industrial baseline. |
| EPICA Dome C | Ice Core | ~800,000 BP – ~2000 BP | Variable (~100-1000 yr) | EPICA Dome C directory | Deepest Antarctic ice core. Covers 8 glacial cycles. Lüthi et al. 2008. |
| Vostok Ice Core | Ice Core | ~420,000 BP – ~2000 BP | Variable | Vostok directory | Classic Antarctic record. Petit et al. 1999; now superseded by EPICA for depth. |
| WAIS Divide | Ice Core | ~68,000 BP – ~0 BP | Variable | wais2014co2.txt | West Antarctic Ice Sheet core. High resolution for last glacial period. |
| Byrd Ice Core | Ice Core | ~90,000 BP – ~9,000 BP | Variable (detailed 50-15 kyr BP) | NOAA Byrd directory | West Antarctic core drilled 1968. Multiple studies with different time windows. Most detailed coverage for last glacial period (50-15 kyr BP). Neftel et al. 1988. |
| Siple Dome Ice Core | Ice Core | ~60,000 BP – ~9,000 BP | Variable | NOAA Siple Dome directory | West Antarctic core. Ahn et al. 2004. Often synchronized with Greenland records for inter-hemispheric comparison. |
| Taylor Dome Ice Core | Ice Core | ~60,000 BP – ~20,000 BP | Variable | NOAA Taylor Dome directory | East Antarctic core. Indermühle et al. 2000. Useful for last glacial period. |
| Talos Dome Ice Core | Ice Core | ~250,000 BP – recent | Variable | NOAA Talos Dome directory | East Antarctic core. Provides additional coverage for last 2-3 glacial cycles. |
| Dome Fuji Ice Core | Ice Core | ~720,000 BP – recent | Variable | NOAA Dome Fuji directory | Japanese drilling project. Deep East Antarctic core approaching EPICA depth. |
| EPICA Dronning Maud Land (EDML) | Ice Core | ~150,000 BP – recent | Variable | NOAA EDML directory | EPICA drilling in Dronning Maud Land. Complements Dome C record with different accumulation rate. |
| Antarctic Composite (NOAA 2015) | Ice Core | ~800,000 BP – 2001 CE | Variable | antarctica2015co2composite-noaa.txt | Merged composite from multiple Antarctic cores (EPICA, Vostok, WAIS, etc.). Standard reference for Quaternary context. |
| Lüthi et al. 2008 EPICA Synthesis | Ice Core | ~800,000 BP – recent | Composite synthesis | doi:10.1038/nature06949 | High-resolution 650-800 kyr BP data from EPICA Dome C. Landmark paper establishing 800 kyr record. See also Bereiter et al. 2015 for updated composite. |
| Bereiter et al. 2015 Antarctic Composite | Ice Core | ~800,000 BP – recent | Composite synthesis | doi:10.1002/2014GL061957 | Current standard 800-kyr composite for ongoing research. Corrects EPICA Dome C analytical artefacts for the 600–800 kyr interval. Most papers since 2015 use this as the reference baseline. Bereiter et al. 2015, GRL. (Also listed in Section 6.) |
Gas age vs ice age: In ice cores, the trapped gas bubbles are younger than the surrounding ice. Snow at the surface remains permeable to atmospheric gases until it compacts into ice at the base of the firn layer (typically 50–120 m deep), at which point bubbles seal off. Ice age and gas age are therefore modelled separately. The gas–ice age difference (Δage) depends on accumulation rate and temperature:
This matters when aligning CO₂ records with temperature proxies at glacial terminations — the gas age is always younger than the ice age at the same depth.
Note on Greenland cores: Greenland ice cores (NEEM, GISP2, GRIP) are excluded from this catalogue as primary CO₂ sources. Greenland ice has higher carbonate dust loading from Asian sources, and in situ acid–carbonate reactions within the ice can produce CO₂ artefacts that bias measurements high. Greenland cores remain essential for CH₄, δ¹⁸O, and dust records but are not used for atmospheric CO₂ reconstruction.
Moderate-to-high quality: harmonized, quality-controlled proxy compilation from multiple methods. This is the modern reference for deep-time CO₂.
| Source Name | Tier | Time Range | Proxy Methods Included | Access URL | Notes |
|---|---|---|---|---|---|
| CenCO2PIP Product (Nov 2023) | Vetted Proxy | 66 Ma – present | Boron isotopes Phytoplankton Stomata Leaf gas exchange Liverworts Land plant δ¹³C Paleosols Nahcolite | doi:10.5281/zenodo.10078932 | Primary modern reference. Hönisch et al. 2023 Science. Community synthesis with standardized uncertainty quantification. Use this for Cenozoic. |
| Rae et al. Marine Compilation (2021) | Vetted Proxy | 66 Ma – present | Boron isotopes Alkenones | PANGAEA 934290 | Marine-proxy focused. Rae et al. 2021. Cross-check source; CenCO2PIP supersedes for comprehensive use. |
| paleo-co2.org Live Compilation | Vetted Proxy | 66 Ma – present (extending) | Boron isotopes Alkenones Stomata Paleosols Others | paleo-co2.org | Live updated compilation maintained by Foster, Rae, and collaborators. Web interface allows filtering by proxy method, time range, and data quality. Complementary to the static CenCO2PIP product; useful for browsing recent additions before they appear in the next formal CenCO2PIP release. |
Variable quality: original published proxy values before full harmonization. Useful for extension to older periods, but treat as exploratory.
⚠️ Uncertainty estimates are not harmonised across archive entries. Values are reproduced as originally published. Individual papers use incompatible error-reporting conventions (1σ, 2σ, 95% CI, RMSE), and many older studies report no formal uncertainty at all. Do not treat individual study error bars as equivalent or directly comparable.
⚠️ Some archive entries are model-constrained estimates, not direct proxy records. GEOCARB-family outputs and similar mass-balance calculations included in archive compilations should be treated as model results (see Section 5) rather than independent observational evidence.
| Source Name | Tier | Time Range | Proxy Methods Included | Access URL | Notes |
|---|---|---|---|---|---|
| CO2PIP Archive (Jan 2025, latest) | Archive | Phanerozoic (target: ~540 Ma – present) | Boron isotopes Phytoplankton Stomata Leaf gas exchange Paleosols Liverworts Land plant δ¹³C Nahcolite | doi:10.5281/zenodo.5777278 (latest: v1.11, Jan 2025) | Phanerozoic archive of originally published paleo-CO₂ estimates. As-published values before CO2PIP re-processing and modernization. Project goal is full Phanerozoic coverage (~540 Ma). Use for pre-Cenozoic extension only. Heterogeneous methods and quality. Vetted product for pre-Cenozoic in progress (as of 2025-2026). |
| Foster, Royer & Lunt (2017) | Archive | 420 Ma – present | Boron isotopes Phytoplankton Stomata Paleosols Liverworts | doi:10.1038/ncomms14845 | Previous-generation Phanerozoic compilation (n=1,241 estimates from 112 studies). Applied uniform screening criteria. Now superseded by CO2PIP/CenCO2PIP for Cenozoic. Still useful for pre-Cenozoic comparison until CO2PIP vetted product is complete. |
Model-based CO₂ reconstructions that combine geochemical theory, carbon cycle dynamics, and geological constraints. These are NOT direct measurements but physics-based estimates. GEOCARB-family models are theory-heavy with relatively loose data constraints; cGENIE inversions are heavily data-constrained model–proxy hybrids.
| Model/Framework | Type | Time Range | Basis | Access/Reference | Notes |
|---|---|---|---|---|---|
| GEOCARB III/GEOCARBSULF | Geochemical box model | Phanerozoic (540 Ma – present) | Long-term carbon-climate modeling based on weathering, degassing, and carbon burial | Berner 1991, 1994, 2001 | Classic long-term CO₂ model. Widely cited but coarse temporal resolution (~10 Myr). Model output, not proxy data. Superseded by proxy compilations for detailed work but useful for context. GEOCARB III: Berner & Kothavala 2001 (American Journal of Science). GEOCARBSULF: Berner 2006 (Geochimica et Cosmochimica Acta). Earlier versions: Berner 1991, 1994. |
| COPSE Model | Biogeochemical Earth system model | Phanerozoic (540 Ma – present) | Couples carbon, oxygen, phosphorus, sulfur cycles with climate | Bergman et al. 2004; Lenton et al. 2018 | More complex than GEOCARB. Includes oxygen evolution. Model output. Useful for exploring biogeochemical feedbacks. |
| cGENIE Carbon-Cycle Inversions | Earth system model + data assimilation | Variable (often Cenozoic-focused) | Inverse modeling: fits model to proxy data to infer carbon cycle state | cGENIE model suite | Sophisticated inverse approach. Combines models with proxy constraints. Model-data hybrid. Not raw proxy data. |
| Historical Chemical CO₂ Measurements | Archived chemical analysis | ~1812–1961 CE | Historical wet-chemical CO₂ determinations | Beck compilation (controversial) | ⚠️ NOT RECOMMENDED. Ernst-Georg Beck's compilation of historical chemical measurements shows extreme variability (280-440 ppm swings). Widely rejected by mainstream science due to methodological issues, local contamination, and inconsistency with ice core data. Included here for completeness only. |
Specialized proxy datasets and compilations not covered in the main vetted/archive categories.
| Dataset/Compilation | Proxy Type | Time Range | Access/Reference | Notes |
|---|---|---|---|---|
| Köhler et al. Ice-Core Harmonizations | Ice core synthesis | ~800,000 BP – recent | Köhler et al. 2017, Clim. Past | Harmonized composite of multiple Antarctic ice cores with careful treatment of inter-core offsets. Recommended for Quaternary work. |
| Bereiter et al. 2015 Ice-Core Composite | Ice core synthesis | ~800,000 BP – recent | Bereiter et al. 2015, GRL | Updated EPICA Dome C corrections for analytical artifacts. Addresses 600-800 kyr issues. |
| Ginkgo Stomata Compilations | Fossil stomata | Mesozoic – Cenozoic | Multiple studies (e.g., Retallack 2001, Royer et al. 2001) | Ginkgo has long evolutionary history. Requires modern calibration. High uncertainty (±200-500+ ppm). |
| B/Ca Proxy Datasets | Boron/Calcium ratio | Cenozoic | Various studies (see CenCO2PIP for vetted data) | Alternative boron-based proxy. Related to carbonate chemistry. Less commonly used than δ¹¹B. |
| PAGES2k Integrations | Multi-proxy synthesis | Last 2,000 years | PAGES 2k Network | Primarily temperature reconstructions, but some include CO₂ context from ice cores. Not a CO₂-specific product. |
| IPCC AR6 Paleo CO₂ Assessments | Assessment synthesis | Full geological time | IPCC AR6 WG1 Chapter 2 | Policy-relevant assessment of paleo CO₂ knowledge as of ~2020-2021. Pre-dates CenCO2PIP publication (2023). Useful for context but superseded by newer compilations. |
Brief explanation of major proxy methods used to reconstruct ancient atmospheric CO₂.
| Proxy Method | Physical Basis | Typical Time Range | Typical Uncertainty | Key Limitations |
|---|---|---|---|---|
| Boron Isotopes (δ¹¹B) | Boron isotope fractionation in marine carbonates reflects seawater pH, which is linked to atmospheric CO₂ via ocean carbonate chemistry. | Cenozoic – Recent | ±40–100 ppm (typically ~±50 ppm) | Most robust marine proxy. Requires knowledge of seawater temperature, salinity, δ¹¹B of seawater, and careful sample selection. Uncertainty varies with calibration method and sample quality. |
| Phytoplankton (alkenones, ε_p) | Carbon isotope fractionation during photosynthesis in marine algae (haptophytes) is influenced by dissolved CO₂ concentration. | Cenozoic (past ~25 Ma most reliable) | ±200–600 ppm (analytical ±200-300 ppm alone) | Very large uncertainty. Extremely sensitive to algal growth rate, cell geometry, sea surface temperature (SST), local oceanographic conditions, and nutrient availability. Uncertainty increases markedly as εp approaches maximum values. Requires accurate SST reconstruction and site-specific calibration. Error range can reach ±600 ppm or more in deep time. Less reliable than boron isotopes. Recent recalibrations (mid-2020s, e.g. Stoll et al., Zhang et al.) have narrowed uncertainty for some Cenozoic intervals by improving treatment of cellular geometry and growth-rate effects; older alkenone estimates in archive datasets may not reflect these improvements. |
| Leaf Stomata (stomatal index/density) | Plants regulate stomatal density in response to atmospheric CO₂. Lower CO₂ → more stomata needed for gas exchange. | Mesozoic – Cenozoic | ±100–500+ ppm (often ±200–300 ppm) | Highly uncertain. Error can reach 200% of the estimate and is right-skewed (larger errors at higher CO₂). Requires modern calibration from same plant lineage. Limited to periods with fossil leaves. Strongly affected by water availability, light, and other environmental factors. |
| Leaf Gas Exchange (mechanistic model) | Physiological models of leaf carbon uptake and water loss, constrained by fossil leaf morphology and carbon isotopes. | Mesozoic – Cenozoic | ±150–500+ ppm | Model-dependent with large uncertainty. Requires assumptions about plant physiology (e.g., stomatal conductance, assimilation rate), environmental conditions, and ancient leaf anatomy. Different parameterizations can yield very different results. |
| Liverworts (carbon isotopes) | Liverworts lack stomata and take up CO₂ by diffusion across cell membranes. Their carbon isotope signature reflects atmospheric CO₂ concentration. | Paleozoic – Mesozoic | ±200–600+ ppm | Very sparse fossil record. Calibration highly uncertain for ancient species. Limited to rare localities with exceptional preservation. |
| Land Plant δ¹³C (various models) | Carbon isotope discrimination during photosynthesis is affected by atmospheric CO₂/O₂ ratio, plant physiology, and environmental stress. | Paleozoic – Cenozoic | ±300–1000+ ppm | Extremely model-dependent. Sensitive to atmospheric O₂ levels (which are also poorly constrained in deep time), plant functional type, soil moisture, light levels, and temperature. Different models can disagree by factor of 2–3. |
| Paleosols (pedogenic carbonates) | Soil carbonate nodules form in equilibrium with soil CO₂, which is related to atmospheric CO₂ via soil respiration models. | Paleozoic – Cenozoic | ±500–2000+ ppm | Very large uncertainty. Requires assumptions about soil respiration rate (highly variable), soil temperature, precipitation/evaporation, and carbonate formation depth. Strongly affected by post-depositional alteration (diagenesis). Values can be unreliable without careful geochemical screening. |
| Nahcolite (mineral stability) | Sodium bicarbonate mineral (nahcolite, NaHCO₃) forms only above a temperature-dependent atmospheric CO₂ threshold. Presence indicates minimum CO₂. | Paleozoic – Mesozoic (rare) | Provides minimum only (e.g., >1100 ppm) | Extremely rare. Only provides lower bound on CO₂, not absolute value. Temperature-dependent stability. Requires exceptional preservation in evaporite deposits. |
⚠️ Uncertainty increases dramatically going back in time:
⚠️ Common errors when plotting proxy data:
⚠️ For pre-Cenozoic data specifically:
⚠️ Do not splice incompatible temporal resolutions without acknowledgement:
Ice-core CO₂ records are inherently smoothed by firn diffusion and gradual bubble close-off. A measurement at Dome C represents an atmospheric average over several centuries to millennia; a Mauna Loa measurement represents hours to days. Directly overlaying these without noting the resolution difference can produce misleading "flat past, sudden modern spike" visualisations.
A century-long CO₂ excursion in the deep past would appear as a small bump (or be invisible) in a Dome C record, whilst the same excursion today would appear as a sharp spike in instrumental data. The rate of modern CO₂ rise is genuinely unprecedented in the Quaternary record (and supported by independent lines of evidence), but charts must acknowledge the resolution asymmetry to make this point honestly.
⚠️ Proxy methods are not always statistically independent:
Multiple proxy methods improve confidence when they converge, but independence should not be assumed. Common shared dependencies include:
Cross-validation is strongest when methods rely on genuinely different physical/chemical bases (e.g. boron isotopes vs leaf stomata vs nahcolite presence).
⚠️ Typical temporal resolution varies by many orders of magnitude:
| Dataset | Typical resolution |
|---|---|
| Instrumental (Mauna Loa, etc.) | Hours to months |
| Law Dome (DE08/DE08-2) | ~10–40 years |
| WAIS Divide | ~10–20 yr (Holocene), ~20–200 yr (glacial) |
| EPICA Dome C / Vostok | ~200–2,000 years |
| Cenozoic vetted proxies (CenCO2PIP) | ~10–100 kyr (interval-dependent) |
| Pre-Cenozoic archive proxies | ~Myr (highly variable) |
Resolution differences of three to six orders of magnitude exist between modern instrumental data and deep-time proxies. Visual comparisons across these scales require explicit acknowledgement.
⚠️ Clarification on "extreme outlier" thresholds:
The "<100 ppm" rule above is a simplification. For context: Quaternary glacial minima reach ~172–180 ppm, so Pleistocene values below ~170 ppm warrant scrutiny. For pre-Quaternary intervals, values below ~150 ppm are almost certainly artefacts. Very high values (>3000 ppm) may be plausible in early Paleozoic intervals but require verification against geological context.
⚠️ Holocene CO₂ stability and the early-anthropogenic hypothesis:
Pre-industrial Holocene CO₂ varied within a narrow ~260–285 ppm range. A small mid-Holocene minimum (~260 ppm, ~7 kyr BP) was followed by a gradual ~20 ppm rise into the late Holocene. Ruddiman and others have hypothesised this rise reflects early human land-use change (the "early-anthropogenic hypothesis"); the debate is unresolved.
This natural late-Holocene rise of ~20 ppm took roughly 5,000 years. The post-1850 industrial rise of ~135 ppm took roughly 175 years — approximately 200× faster in absolute rate. Charts showing a "modern spike" relative to "flat" Holocene CO₂ are quantitatively correct on rate, but should ideally make the rate comparison explicit rather than relying on visual impact alone.
⚠️ Age convention ambiguity — check before stitching datasets:
Different datasets use incompatible age reference systems:
| Convention | Meaning | Common in |
|---|---|---|
| CE / BCE | Calendar year, Gregorian | Instrumental, some ice-core |
| yr BP | Years before 1950 CE (by convention) | Radiocarbon, Holocene ice-core |
| kyr BP | Thousands of years before 1950 CE | Quaternary paleoclimate |
| Ma | Millions of years before present | Cenozoic / Phanerozoic proxy |
A value at "100 yr BP" is 1850 CE regardless of when the paper was published. When stitching ice-core (yr BP) data to instrumental (CE) data, ensure code accounts for the 1950 reference offset.
⚠️ Digitisation accuracy in archive datasets:
Some archive proxy estimates were digitised from published figures (e.g. using WebPlotDigitizer) rather than obtained from original data files. Digitisation error is typically ±1–5% of the plotted range and is rarely noted in compiled datasets. For pre-Cenozoic data where the full CO₂ range spans hundreds of ppm, this can introduce tens-of-ppm offsets that are invisible in reported uncertainty columns. Where original data files are available (Zenodo, PANGAEA, author repositories), prefer them over values taken from secondary compilations.
Recommended default data stack for composite CO₂ charts:
| Time range | Primary source |
|---|---|
| 1958 CE – present | NOAA GML Mauna Loa or Scripps CO₂ Program |
| 1850 CE – 1958 CE | Law Dome (DE08/DE08-2) bridging into instrumental |
| ~2 kyr BP – present | Law Dome composite |
| ~800 kyr BP – present | Bereiter et al. 2015 Antarctic composite |
| 66 Ma – 800 kyr BP | CenCO2PIP (Hönisch et al. 2023) |
| Pre-66 Ma | CO2PIP Archive — clearly labelled as exploratory |
Use a single tier per time window; do not overlay tiers within the same window unless explicitly visualising methodological agreement.
For modern analysis (1958–present): Use NOAA GML Mauna Loa data or Scripps CO₂ Program data (both are independent, high-quality records). This is the gold standard. Analytical uncertainty <0.5 ppm.
For pre-industrial baseline and recent centuries (last ~2000 years): Use Law Dome ice core data to bridge into the instrumental period. Analytical precision ~1.2 ppm, with potential inter-core offsets of a few ppm.
For Quaternary glacial cycles (last 800,000 years): Use Antarctic ice core composite (EPICA Dome C + WAIS Divide + others). Analytical precision ~1–2 ppm per measurement, but be aware of documented inter-core offsets of 2–5 ppm between different cores. Age uncertainty increases with depth. Excellent for showing natural variability over glacial-interglacial cycles.
For Cenozoic deep time (66 Ma – 800 kyr): Use CenCO2PIP vetted product. This is the modern reference synthesis with standardized methods and quality control. Uncertainty typically 40–200 ppm for best methods (boron isotopes), higher for others. This is the most reliable Cenozoic compilation available.
For pre-Cenozoic (Mesozoic/Paleozoic, 540–66 Ma): Use CO2PIP Archive data with extreme caution. Clearly label as "exploratory/provisional" and note that the vetted product is not yet available. Expect very large uncertainties (500–2000+ ppm) and many outliers. Cap display at reasonable values (e.g., 2000 ppm) unless specifically discussing extreme hothouse periods.
Default chart configuration for public audiences:
When in doubt: Show less data with clear quality tiers, not more data with hidden quality issues. Non-expert audiences cannot evaluate proxy quality themselves.