New radiocarbon calibration curves, IntCal04 and Marine04, have been constructed and internationally ratified to replace the terrestrial and marine components of IntCal98. The new calibration data sets extend an additional 2000 yr, from 0–26 cal kyr BP (Before Present, 0 cal BP = AD 1950), and provide much higher resolution, greater precision, and more detailed structure than IntCal98. For the Marine04 curve, dendrochronologically-dated tree-ring samples, converted with a box diffusion model to marine mixed-layer ages, cover the period from 0–10.5 cal kyr BP. Beyond 10.5 cal kyr BP, high-resolution marine data become available from foraminifera in varved sediments and U/Th-dated corals. The marine records are corrected with site-specific 14C reservoir age information to provide a single global marine mixed-layer calibration from 10.5–26.0 cal kyr BP. A substantial enhancement relative to IntCal98 is the introduction of a random walk model, which takes into account the uncertainty in both the calendar age and the 14C age to calculate the underlying calibration curve (Buck and Blackwell, this issue). The marine data sets and calibration curve for marine samples from the surface mixed layer (Marine04) are discussed here. The tree-ring data sets, sources of uncertainty, and regional offsets are presented in detail in a companion paper by Reimer et al. (this issue).

The first meeting of the IntCal04 working group took place at Queen's University Belfast from April 15 to 17, 2002. The participants are listed as co-authors of this report. The meeting considered criteria for the acceptance of data into the next official calibration dataset, the importance of including reliable estimates of uncertainty in both the radiocarbon ages and the cal ages, and potential methods for combining datasets. This preliminary report summarizes the criteria that were discussed, but does not yet give specific recommendations for inclusion or exclusion of individual datasets.

This paper includes a brief history of the calibration of the radiocarbon time scale from the first recognition of the necessity of calibration in 1962 to INTCAL98. Thirty-six years of effort by dendrochronologists and the 14C community have pushed the tree-ring calibration back to 11,854 yr BP. All of this part of the calibration has been done by high-precision beta counting. Uranium-thorium (U-Th) dating of coral samples coupled with accelerator mass spectrometry (AMS) measurement of 14C has extended a fairly detailed calibration back beyond the Bølling warm episode to 15,000 BP. Earlier than 15,000 BP, piecewise linear approximation extends INTCAL98 calibration to 24,200 BP.

Blending 1-, 2-, 3-, 10-, and 20-yr tree-ring samples containing regional and data offsets into a decadal time scale does not make an ideal error and bias free δ14C record. Nevertheless, spectral analysis reveals some statistically significant fundamental frequencies as well as interesting “beat” frequencies and the second harmonic of the around 208-yr cycle that is considered to be solar in origin. Although, some very prominent peaks such as the 88-yr (Gleissberg) are clearly solar in origin, some of the lower frequencies such as of the 512-yr period may have an origin in thermohaline circulation. Thus, INTCAL98 provides useful data for geophysical and solar physics research. Lastly, single year δ14C analysis would be useful for revealing invaluable information for solar physics, astrophysics and geophysics not accessible by decadal data. We provide several examples.

Measurements on same-age tree-ring samples from proximal Ural Mountain trees by the Ioffe Institute research group and at the University of Arizona demonstrate a variance corresponding to a standard deviation of ±5.1% for Ioffe compared to ±2.1% for Tucson. There is also a calibration difference of 4.3 ±1.2 ‰. Comparison of the same years measured in Seattle on wood from the Pacific Northwest shows an offset of 2.2 ± 0.5 ‰. This is not a calibration error, but rather is expected from the well-documented evidence for divergence and upwelling of 14C-depleted CO2 along the west coast of North America.

Recently, in an intercomparison of the Hohenheim German oak chronology (Becker 1993) and the Göttingen chronology (Leuschner and Delorme 1988), an error was discovered in the former (Leuschner, in preparation). Due to an error in adding sections at 5241 bc, 41 yr are missing in the published Hohenheim chronology. After correction of the error, the two chronologies synchronize over their entire common length, back to 7200 bc.

We selected SN1006, the brightest and closest to Earth of all supernovas historically observed, for a study of 14C production by e−,e+-bremsstrahlung cascades initiated by hard γ rays (>10 MeV) from that event. During the cascade, bremsstrahlung energies eventually fall within a giant (n,γ), (n,2γ) cross-section, peaking at 23 MeV and approaching effectively zero below 10 MeV and above 40 MeV. The neutrons are absorbed primarily in the reaction 14N(n,p)14C. Cellulose from single-year tree rings from ad 1003 to ad 1020 was measured to determine ∆14C. Three years after the first visual observation of SN1006, ∆14C rose and remained above pre-ad 1009 values until ad 1018. Comparison of the 7 years before ad 1009 with the 9 years following show an average increase of 6.1 ± 1.6 (s.d.)‰ (significant at the 99.6% confidence level). Such a pulse of 14C requires a total production of neutrons of 17.1 × 107n cm−2e, implying an input of 11.3 × 104 ergs cm−2e γ-ray energy. This requires the total supernova γ-ray energy (>10 MeV) to have been 1 × 1050 ergs.

Although a coauthor of the paper of Kahn et al. (1995), I was not available in time to review the final copy. Consequently, I take this opportunity to make further comments and, hopefully, clarify some of the content and conclusions.

Using present global warming and paleoclimatic records from climatically sensitive regions as a frame of reference, we infer that global temperature changes did not exceed ±0.5°C during the current interglacial or ±2.0°C during the last glacial period. In order to completely explain such fluctuations by solar irradiance changes, all solar and terrestrial factors must be optimized. Since this is unlikely, we conclude that solar forcing of pre-anthropogenic climate change is a significant and perhaps dominant factor but other processes must also be significant. Solar irradiance changes alone cannot result in the temperature decrease required to change from interglacial to glacial as suggested by Opik (1965).

The coarse structure of the 14C spectrum consists of a secular trend curve that may be closely fit by a sinusoidal curve with period ca 11,000 yr and half amplitude ±51. This long-term trend is the result of changes in the earth's geomagnetic dipole moment. Consequently, it modulates solar components of the 14C spectrum but does not appear to modulate a component of the spectrum of ca 2300-yr period. The ca 2300-yr period is of uncertain origin but may be due to changes in climate because it also appears in the δ18O spectrum of ice cores. This component strongly modulates the well-known ca 200-yr period of the spectrum's fine structure. The hyperfine structure consists of two components that fluctuate with the 11-yr solar cycle. One component results from solar-wind modulation of the galactic cosmic rays and has a half-amplitude of ca ±1.5. The other component is the result of 14C production by solar cosmic rays that arrive more randomly but rise and fall with the 11-yr cycle and appear to dominate the fluctuation of the galactic cosmic-ray-produced component by a factor of two.

The production rate of 14C during the Holocene averaged 2.4 ± 0.2 atoms 14C/cme 2 sec. Neutrons produced by galactic cosmic rays account for 90% of the 14C production with the remaining 10% resulting from neutrons produced by protons from solar flares. Production and decay of 14C can be reconciled by including 14C permanently or temporarily stored in sediments. Sedimentary reservoirs contain ca 30% of all terrestrial 14C. The lagoons, bays, marshes and deltas of the coastal wetlands alone account for 12% of the 14C inventory. The capacity of the coastal wetlands to store carbon has become the subject of renewed interest.

New results of radiocarbon dating of ca 100 decadal bristlecone pine samples from 6554 BC to 6084 BC and from 5820 to 5350 BC are presented. Using 3 new 2.5L counters filled to ca 3atm with carbon dioxide, high-precision dating has been performed by this laboratory for more than two years. Demonstration of the precision and accuracy of these counters is presented using ±2‰ measurements from the Spörer minimum period. For the older samples, ±3‰ measurements were made using ca 12-day counting times. Results are presented both as C age bp vs dendro-year BC, particularly for calibration purposes, and as Δ14C vs time.

New Arizona high precision Δ14C data back to 6500 BC plot close to an 11,300-yr period sinusoid extrapolated from the post 5300 BC data (offset = +32‰, half amplitude = 51‰ and phase lag = 2.29 radians). The trend curve is modulated by high latitude components of the non-dipole field with a fundamental period of 2400 yr. Based upon a model of Lund and Banerjee (1985), the non-dipole field rotates and every 1200 yr the high latitude maxima pass over the north magnetic pole and near the south magnetic pole in reversed polarity. This modulates cosmic ray production producing extended maxima ca AD 1700, 700 bc, 3100 bc, and 5500 bc. The 2400 period appears to be stationary. The magnetic field also modulates the amplitude of the solar activity induced cycles of periods 200, 80, and 11 yr as can be seen in the Zürich-Bern Camp Century ice core data as well as in the Δ14 C fluctuation data. Reinterpretation of the Camp Century 10Be data indicates that it is in agreement with magnetic field as well as solar activity modulation of terrestrial radioisotope production.

A constraint on radiocarbon reservoir models is that the DC gain of a system (system transfer function at zero frequency) should equal the equilibrium ratio of the atmospheric radiocarbon mass to the production rate. The simple one-box model is essentially a “black box” but the value of the single residence time is theoretically equal to the DC gain. Using a sunspot-production rate algorithm as the forcing function, predictions of the one-box model match the 14C data from ad 1700 to 1900 better than the 3-box, 5-box and box-diffusion models. The more complex models tend to pile up 14C in the atmosphere because their DC gains are too high, and they over-attenuate the de Vries “wiggles”. The DC gains can be reduced to more acceptable levels by adjusting model parameters, particularly the sizes of the ocean reservoirs. Better fits to the “wiggles” are also obtained by parameter adjustment. Water content of deep-sea sediments constitutes an extra reservoir for dead carbon, and should help reduce system DC gain.

Radiocarbon dates calculated from the ratio of modern carbon-14 activity and sample activity and the half-life of carbon-14 need to be calibrated to compensate for temporal variations in the concentration of carbon-14 in the atmosphere. Development of a suitable calibration scheme has been an ongoing process of the last twenty years, ever since the discovery of variations in historical times of the atmospheric radiocarbon content which parallel climatic and solar phenomena (de Vries, 1958; 1959) and the recent depletion due to industrial effects (Suess, 1955).

The C14 measurements reported here were madein this laboratory between November 15, 1965 andJune 15, 1966. Sample descriptions are classified asfollows:

1. I. GeochemicalSamples

2. II. Geologic-PaleoclimaticSamples

3. III. EarlyMan-Alluvial StratigraphySamples

4. IV. ArchaeologicSamples

The C14 measurements reported here were made in this laboratory between November 1, 1963 and November 15, 1965. Sample descriptions are classified as follows:

1. I. Geochemical Samples

2. II. Geologic-Paleoclimatic Samples

3. III. Early Man-Alluvial Stratigraphy Samples

4. IV. Archaeologic Samples

With few exceptions the C14 measurements reported here were made in this laboratory between October 1, 1962 and November 1, 1963.