Radiocarbon dates have been obtained for 30 charcoal samples corresponding to 27 surface lava flows from the Mauna Loa and Kilauea volcanoes on the Island of Hawaii. The submitted charcoal was a mixture of fresh and archived material. Preparation and analysis was undertaken at the NERC Radiocarbon Laboratory in Glasgow, Scotland, and the associated SUERC Accelerator Mass Spectrometry facility. The resulting dates range from 390 years B.P. to 12,910 years B.P. with corresponding error bars an order of magnitude smaller than previously obtained using the gas-counting method. The new and revised 14C data set can aid hazard and risk assessment on the island. The data presented here also have implications for geomagnetic modelling, which at present is limited by large dating errors.
 The Island of Hawaii is the youngest island in the Hawaiian archipelago, located in the central-northern Pacific Ocean. The island is made up of five volcanoes (in order of decreasing age): Kohala, Mauna Kea, Hualalai, Mauna Loa and Kilauea; the three youngest of which are still active. 80% of the surface flows on Hualalai are less than 5000 years old and in general the volcano erupts every few hundred years. Mauna Loa has had 33 historical eruptions that have contributed to lava flows covering 50.5% of the island. The youngest of the volcanoes, Kilauea, is also the most active, having been erupting almost continuously from the Pu'u'O'o vent since 1983.
 The Hawaiian radiocarbon programme was initiated by the USGS in the late 1940s with the aim of understanding the history and timing of pre-historic volcanism in Hawaii with regard to hazard and risk assessment. Rubin et al.  documents the largest compilation of 14C dates with 185 ages from Mauna Loa lava flows, 90 from Kilauea flows and 30 from Hualalai flows. Additional radiocarbon ages can be found in work by Neal and Lockwood , which focuses on the summit region of the Kilauea volcano.
 A major limitation associated with the Hawaiian radiocarbon database stems from the small amount of charcoal material submitted for analysis. The ages given by Rubin et al.  were all processed at the USGS Radiocarbon Laboratory, Reston, Virginia (laboratory code W or WW) using the gas-counting method. This early dating technique requires 5–10 g of organic material, but many sites yield insufficient amounts without large-scale quarrying. To remove any contamination, the samples need to be pretreated with acids that remove carbonates from percolating ground water, and alkalis that remove humics, i.e. the mobile decay products of biological material. This process is very destructive and, in some cases where samples were too small to leave adequate material after pretreatment, a 14C age was obtained without pretreatment (J. McGeehin, U.S. Geological Survey, personal communication, 2006). In addition, the gas-counting equipment used was not sophisticated enough to account and correct for isotopic fractionation. Unsurprisingly, recent ground-mapping of the lava flows show some radiocarbon ages to be inconsistent with constraints set by stratigraphy [Trusdell and Lockwood, 2007].
 In this paper, a number of lava flows were targeted for dating using the more modern and more accurate accelerator mass spectrometry (AMS) equipment. These included flows that had already been dated, but had large errors on the age (1σ > 100 years), or dated flows with radiocarbon laboratory numbers in the W3000's, implying analysis was carried out while the gas-counting method was still in its infancy. A significant advantage of the AMS method over the gas-counting method is sample size. AMS typically requires 10–100 mg of charcoal — 1000 times less than that required for gas-counting. However, because of the reduced sample size, pretreatment needs to be extremely rigorous since a small amount of contamination can lead to substantial errors. A total of 30 samples were submitted to the Natural Environment Research Council (NERC) Radiocarbon Laboratory and analysed at the Scottish Universities Environmental Research Centre (SUERC) AMS facility. The charcoal was a mixture of fresh material from new locations, fresh material from previously sampled locations and archived material that had either not been submitted before or was surplus after earlier dating.
2. Sample Collection
 For the formation of charcoal, there first needs to be woody material present. In Hawaii, the characteristic low viscosity lava results in thin flows that are conducive to rapid forest growth between volcanic eruptions. In particular, the lower viscosity pahoehoe-type lava will smother any underlying vegetation, completely excluding oxygen and thus creating high quality, vitreous charcoal. Practical guidelines for the recovery of such charcoal beneath young basaltic lava flows are outlined by Lockwood and Lipman . In this study, four new charcoal localities were found and eight known charcoal sites were revisited to collect fresh material. Surplus charcoal material from 13 previous excavations, archived in a repository within the U.S. Geological Survey's Hawaiian Volcano Observatory, was also submitted, along with a further five samples that had been collected by the USGS in the 1990s, but had not been previously analysed. Figure 1 shows the location of the 30 charcoal sampling sites.
 The raw 14C results can be found in the auxiliary material, Data Set S1. The radiocarbon ages have been corrected for isotopic fractionation and are reported as conventional radiocarbon years before present (B.P.) relative to 1950 AD, with a 1σ error. Where there is more than one age representing a flow, the weighted mean 14C age has been calculated and rounded according to Stuiver and Polach . The t-test was used to determine whether the duplicate ages were statistically estimating the same date at the 95% confidence level. Conversion from mean 14C age to mean calibrated age in calendar years AD/BC was done using Version 5.0.1 of the CALIB Radiocarbon Calibration Program [Stuiver and Reimer, 1993] in conjunction with the IntCal04 Atmospheric 14C Curve [Reimer et al., 2004]. Table 1 lists the weighted mean 14C ages (years B.P.) and the corresponding mean calibrated ages (calendar years AD/BC).
Table 1. Weighted Mean 14C Ages, Calibrated Calendar Ages and Associated Paleomagnetic Informationa
Mean 14C Age (yrs B.P. ±1σ)
Calibrated Calendar Age (cal AD/BC)
F ± 1σ
Map number corresponds to the sample location number in Figure 1. Mean 14C age and 1σ error is the weighted mean radiocarbon age for a single flow in years B.P., rounded according to Stuiver and Polach . Most probable calibrated calendar age (years AD or BC) calculated using Version 5.0.1 of the CALIB Radiocarbon Calibration Program [Stuiver and Reimer, 1993] in conjunction with the IntCal04 Atmospheric 14C Curve [Reimer et al., 2004]. Any paleomagnetic data associated with dated flows are also listed: unique paleomagnetic code; paleointensity (μT); declination (deg); inclination (deg), α95 cone of confidence; reference source.
D. Champion, U.S. Geological Survey, unpublished data, 2001.
 Of the 30 samples analysed in this suite of radiocarbon experiments, 22 can be associated with previously dated flows. The revised 14C ages all have smaller errors (Figure 2). In particular, the 1σ errors have been reduced by an order of magnitude for charcoal locations #1 + 2 and location #12: the original errors of 100 and 150 years have been reduced to 20 and 30 years respectively. Charcoal from location #18 was re-dated because the gas-counting analysis could only produce a minimum age of 3000 years B.P. This would have been due to the sample count-rate and background count-rate being indistinguishable at the two standard deviation level. The more sensitive AMS equipment produced a consistent, but slightly older age of 3094 ± 28 years B.P.
 Reliable ages were obtained for 60% of the charcoal samples submitted. In 14 cases, the new ages agreed with the previously published ages to within one standard deviation. A further four cases (corresponding to charcoal locations #9, 11, 14 and 25) had not been formerly dated. The ages produced here are compatible with stratigraphic constraints in the area.
 Seven lava flows failed the t-test, implying that the duplicate 14C ages are not statistically estimating the same date. Two flows selected for re-dating because of large errors on the previously published age were from Kilauea Crater (charcoal location #6) and Volcano Village (charcoal location #13). The AMS ages were significantly younger in both cases. However, neither of the flows are overlain by younger flows. Therefore, with no younger age limit, the results must be regarded as indeterminate. A similar situation is seen regarding a flow with a radiocarbon age generated when the gas-counting method was still developing (#16, laboratory number W3876). The AMS analysis was performed on surplus charcoal from the original excavation and the resulting age complies with constraints from surrounding flows. However, it is 146 years older than the previously published age, causing the two ages to fail the t-test. But, since the AMS technique and equipment can be considered to be more superior to the gas-counting method, more confidence is placed in the AMS age. Likewise, the two AMS ages relating to charcoal from locations #21 + 22 are older than gas-counting age for the flow and the three samples together fail the statistical analysis. When taking into account only the AMS ages, a mean 14C age of 3800 years B.P. is obtained that is statistically robust. An alternative explanation regarding the slightly older ages for #16, 21 and 22 with respect to the new AMS dates is that the older method did not adequately clean the samples of modern charcoal/carbon. Modern carbon (rootlets, detritus) too small to be seen with the naked eye may still have been present when the radiocarbon was analysed. The outside of twigs can look clean, yet when they are broken open tiny modern rootlets can often be found, making contamination a real concern. The modern AMS technique will remove most, if not all modern carbon.
 A poor 14C result was obtained for charcoal from location #4. The gas-counting age is 28,120 years B.P., the AMS age dramatically younger at 739 years B.P. The younger age came from a freshly collected sample. The sampling site was in a flowing river bed and the submitted charcoal was of poor quality, comprising mainly black/brown soily material. Despite rigorous and repeated chemical pretreatment, a significant amount of contamination was still present (C. Bryant, NERC Radiocarbon Laboratory, personal communication, 2006). This modern contamination is the likely cause of the significant younging and makes the age of 739 years B.P. unacceptable.
 The lava flow corresponding to charcoal locations #29 + 30 had not been previously dated; the assigned age of 16,390 years results from its stratigraphic relation to surrounding flows that have been dated. Two samples representing this flow were submitted for AMS analysis, one fresh and one collected in 1993 by the USGS that had not been previously analysed. The latter sample gave an age of 16,606 ± 171 years B.P., in agreement with the age from stratigraphy. The fresher sample gave an age of 13,945 ± 177 years B.P. This is too young with respect to the stratigraphy of the area. The sample collected in 1993 was of better quality (rootlets as opposed to black sooty material). Therefore, the younger age is rejected.
5. Paleomagnetic and Geomagnetic Implications
 Geomagnetic secular variation (SV) occurs on timescales of tens to thousands of years and many characteristic geomagnetic changes occur with periods longer than those covered by historical records. Therefore, paleomagnetism on archaeological and geological materials is used to provide information on paleosecular variation (PSV). However, in contrast to historical observations that can be dated with an accuracy of typically less than a day, paleomagnetic measurements have large dating uncertainties, which are often longer than the dominant periods of SV.
 Accurately dated lava flows that have reliable associated paleomagnetic measurements are ideal to use as tie-point for correlation with other data. For example, long boreholes such as the Hawaii Scientific Drilling Project (HSDP) [Laj et al., 1999] and Scientific Observation Holes 1 and 4 (SOH-1 and 4) [Teanby et al., 2002; Gratton et al., 2005; Laj et al., 2002]), are very difficult to date. No organic material was recovered in any of the cores so K-Ar and 40Ar/39Ar dating techniques were used instead. However, Hawaiian lavas have a tholeiitic composition with low potassium content. Their age is very young in comparison to the potassium half-life and consequently very little radiogenic argon is produced. The K-Ar dating of SOH-4 was the most reliable and the depth-age profile for that borehole was used as a proxy for SOH-1 chronology, which was then stretched to correlate with features seen in the paleomagnetic records from independently dated surface lava flows.
 The paleomagnetic intensity data set from dated surface lava flows is comparatively limited in size with only 81 measurements. The new and revised radiocarbon dates generated in this study apply to a quarter of these paleomagnetic data (Table 1). In particular, three of the flows that have been re-dated have very robust paleointensity estimates: flow 9B289 (location #7), flow 3A001 (location #20) and flow 6B085 (location #24). With regard to a secular variation curve, these three flows are well constrained in both time and intensity and thus ideal to use as tie-points for correlation with other data. The paleointensity, inclination and declination from all published 14C-dated Hawaiian surface lava flows is shown in Figure 3, modified according to the new and revised ages generated in this study. Furthermore, global paleomagnetic databases must incorporate these revised ages. Geomagnetic models, such as CALS7K.2, are based on these large data compilations but their resolution is currently limited by large dating errors. The work in this paper begins to address this problem.
 30 charcoal samples from 27 Hawaiian surface lava flows were submitted to the NERC Radiocarbon Laboratory for processing at the SUERC AMS facility. The samples were a mixture of fresh material from new locations, fresh material from previously sampled locations and archived material that had either not been submitted before or was surplus after earlier dating. The quantity of material submitted was sufficient to permit rigorous chemical pretreatment to remove any contamination.
 Where individual flows have multiple 14C ages, they were combined to give a weighted mean 14C age. The t-test was then employed to test whether the duplicate ages were estimating the same date. All ages were calibrated to give a calendar age for each flow.
 For 18 flows, revised 14C ages have been obtained with dramatically lower 1σ error bars. Of the remaining nine flows, three AMS ages are geologically plausible, despite not agreeing with the previous gas-counting ages, and two AMS ages are deemed indeterminate because no younger age bounds from overlying flows can be applied. The results associated with the other four flows were statistically and geologically unreliable.
 The new 14C data presented here has implications for many contemporary paleomagnetic studies on Hawaii. The results could also affect global geomagnetic models that are currently regulated by inadequately large errors in time.
 We would like to thank John “Jack” Lockwood, formerly of the HVO, for all his guidance in the field. We are also grateful to Charlotte Bryant, Callum Murray and the staff at the NERC Radiocarbon Laboratory for all their help and support. The results in this paper form Radiocarbon Dating Allocation 1069.0404, awarded to D. Gubbins FRS. This work was partially supported by NERC grant A/S/2001/01088. N. Pressling was supported by a tied NERC studentship.