Background—Application of Carbon-14 to Petroleum Spills

Carbon-14 is a general tracer that can be used to identify petroleum-derived carbon (C) in groundwater and in soil CO₂. Carbon-14 is a radioactive isotope of C that is formed in the atmosphere and has a half-life of about 5,700 years (Woods Hole Oceanographic Institution, National Ocean Sciences Accelerator Mass Spectrometry, 2023b). Radiocarbon is continually produced in the atmosphere by interactions between energetic neutrons created by cosmic radiation and atmospheric nitrogen (Suess, 1965). In the atmosphere, ¹⁴C combines with oxygen to form ¹⁴CO₂, which is then fixed by plants and algae (Tans and others, 1979). When C exchange with the atmosphere ceases because of death or senescence of photosynthetic tissues, the ¹⁴C concentration of dead organic matter declines from radioactive decay as time elapses (Larsen and others, 2018). Given enough time (about 60,000 years), plant-derived C such as that detected in fossil fuels (petroleum and coal) can become “radiocarbon dead,” being entirely depleted in ¹⁴C (Graven and others, 2017).

In groundwater, plumes migrating away from petroleum fuel spills contain dissolved organic carbon (DOC) in the form of dissolved fuel compounds and partial transformation products of the fuel compounds originating in the fuel in addition to background DOC (Bekins and others, 2016). In the unsaturated zone or soil gas, plumes of CO₂ from the biodegradation of petroleum fuel can exist near source areas along with CO₂ from background soil respiration (Sihota and others, 2011; Wozney and others, 2021). In groundwater DOC and soil CO₂, the ¹⁴C content of samples decreases as the proportion of petroleum-derived C increases (Sihota and others, 2011; Wozney and others, 2021). Samples with no petroleum-derived C will typically reflect background ¹⁴C content.

The primary motivations for this study were to use measurements of (1) ¹⁴C in groundwater DOC, along with other biodegradation indicators, to track migration of petroleum-derived DOC away from the facility and (2) ¹⁴C in soil CO₂ to identify locations at the facility where petroleum may be biodegrading in the subsurface (fig. 2).

Purpose and Scope

The purpose of this report is to present results and interpretation of groundwater and soil gas samples collected in and around the facility in O‘ahu, Hawai‘i. Samples were collected to understand the migration of petroleum-derived DOC in groundwater away from the facility and to identify locations at the facility where petroleum might be biodegrading in the subsurface.

Site Description

The study site is on the island of O‘ahu, Hawai‘i. The following sections describe the location and important features of the facility and the history of known spills. Short descriptions of the geology and hydrology are provided next and are followed by a description of existing groundwater concentration data used for comparison with the results from this study.

Geology

Most of the facility is underlain by the Ko‘olau Basalt, which ranges in age from 2 to 3 million years (Ozawa and others, 2005). The far western edge of the facility lies on tuff cone deposits of the rejuvenated Honolulu Volcanics. Highly weathered basalt known as saprolite is present in the top 25–50 ft below Red Hill (fig. 3). The basalt rocks consist of individual lava flow layers containing ‘a‘ā flows with clinker zones and commonly massive flow cores and pāhoehoe flows. Interflow voids, joints, and lava tubes associated with pāhoehoe flows contribute to the overall permeability of the volcanic aquifer (fig. 3, fig. 4). Based on observations from outcrops near the facility, the average dip azimuth and dip angle of the lava flows were 246 degrees (Department of the Navy, 2023a) and 2.9 degrees, respectively (Department of the Navy, 2019). The adjacent South Hālawa and Moanalua Stream valleys are underlain by alluvial valley-fill deposits and variably weathered basalt, including saprolite. Measurements of outcrops in South Hālawa, Moanalua, and Waimalu Stream valleys quantified the proportion of clinker to total ‘a‘ā as 15–45 percent (Wentworth and Macdonald, 1953). Rocky outcrops on Red Hill contain numerous vertical fractures that may provide vertical flow connections across otherwise low permeability ‘a‘ā flow cores (fig. 4; Beckett and others, 2022). Extensive vertical connectivity may also be inferred from the rapid transport that occurred from the tunnel floor of adit 3 to the water table about 80 ft below after the November 2021 fuel spill.

Groundwater in the uncontaminated basalt aquifers of O‘ahu is relatively low in DOC with concentrations of 0.19–0.27 milligram per liter (mg/L; Chapelle and others, 2013). For reference, a literature review of uncontaminated aquifer DOC values indicated a range for other sites of 0.1–4 mg/L (Regan and others, 2017). The DOC in uncontaminated groundwater samples from O‘ahu seems to have low to no bioavailability based on the high dissolved oxygen concentrations and lack of correlation between DOC and dissolved oxygen (Chapelle and others, 2013). The Ko‘olau Basalt contains organic C in paleosol layers between lava flows (Wentworth and Macdonald, 1953) that is ¹⁴C dead (Doell and Dalrymple, 1973). Addition of DOC from the paleosol layers will create a DOC pool in groundwater that is older than modern age. Paleosols may also contain pyrogenic C because lava flows engulf and burn plant matter under low oxygen conditions as they advance through forested areas (Wentworth and Macdonald, 1953).

Groundwater

This section describes the methods used in the groundwater part of this study. The wells sampled and the sampling methods are described first and followed by a description of sample preservation methods and analytical procedures used in the field and laboratory to determine concentrations of constituents in the water.

Soil Gas

This section covers the strategy and methods used for soil gas sampling and analyses. It begins with an explanation of how sample locations were chosen and continues with a description of the sampling and analysis methods.

Soil Gas Sampling Design

At the facility, fluxes of CO₂ depleted in ¹⁴C (ancient CO₂) have been used to estimate biodegradation rates of spilled fuel below the facility (Department of the Navy, 2019; McHugh and others, 2020). In 2017, ancient CO₂ was detected above tank 5 and on the southeast-facing slope of Red Hill downhill from tanks 11 and 12 (sites T2, T3, and T8 on fig. 5; Department of the Navy, 2019; McHugh and others, 2020). The presence of ancient CO₂ south of the tank farm at the facility at site T8 (fig. 5) was concerning because it might indicate that spilled fuel migrated south away from the tanks in the direction of the dip of the lava flows. To investigate whether other locations had signs of underlying biodegrading fuel, a series of soil gas measurements was made over the crest of Red Hill, along the southeast-facing slope of the ridge and above the November 2021 spill areas near adit 3 (fig. 5, table 4, adit 3 shown on fig. 1). The purpose of these measurements was to replicate previous results (Department of the Navy, 2019; McHugh and others, 2020) and determine if evidence for subsurface biodegradation of fuel was present at other locations on the surface of the facility, including the area overlying adit 3, the site of the November 2021 spill (Department of the Navy, 2022). Three methods were used to evaluate the presence of ancient C in soil CO₂ at land surface: (1) soil gas sample collection with soil probes for measuring ¹⁴C content of soil CO₂, (2) soil C traps (E-Flux, Fort Collins, Colo.) for measuring soil CO₂ efflux and ¹⁴C content of CO₂ leaving the soil, and (3) soil CO₂ efflux measurements made with a dynamic closed chamber (fig. 5, table 4).

Table 4.    

Site information for soil gas sampling sites at the Red Hill Bulk Fuel Storage Facility, O‘ahu, Hawai‘i.

[USGS, U.S. Geological Survey; HI, Hawai‘i; --, no samples collected; site information from U.S. Geological Survey (2024)]

Table 4.    Site information for soil gas sampling sites at the Red Hill Bulk Fuel Storage Facility, O‘ahu, Hawai‘i.

Site short name	USGS site identifier	USGS station name	Approximate land surface altitude, in feet above local mean sea level	Number of regular samples
				Dynamic closed chamber	Soil gas probe		Soil carbon trap
A3–01	212214157542302	A3–01 Red Hill, Oahu, HI	160	--		1		--
A3–02	212215157542302	A3–02 Red Hill, Oahu, HI	150	--		1		--
A3–03	212215157542301	A3–03 Red Hill, Oahu, HI	135	--		1		--
A3–04	212214157542101	A3–04 Red Hill, Oahu, HI	170	--		1		--
A3–05	212213157541603	A3–05 Red Hill, Oahu, HI	190	--		1		--
A3-06	212213157541602	A3–06 Red Hill, Oahu, HI	200	--		1		--
A3–07	212213157541701	A3–07 Red Hill, Oahu, HI	200	--		1		--
A3–08	212213157541601	A3–08 Red Hill, Oahu, HI	215	--		1		--
T2	212217157534801	T2 Red Hill, Oahu, HI	495	6		2		2
T22–01	212215157535201	T22–01 Red Hill, Oahu, HI	460	6		1		--
T22–02	212216157535001	T22–02 Red Hill, Oahu, HI	475	6		1		--
T22–03	212217157534601	T22–03 Red Hill, Oahu, HI	515	6		1		--
T22–04	212219157534401	T22–04 Red Hill, Oahu, HI	520	6		1		1
T22–05	212220157534001	T22–05 Red Hill, Oahu, HI	535	7		--		--
T22–06	212220157534002	T22–06 Red Hill, Oahu, HI	545	6		1		--
T22–07	212221157533601	T22–07 Red Hill, Oahu, HI	555	6		1		--
T22–08	212222157533401	T22–08 Red Hill, Oahu, HI	555	6		1		--
T22–09	212218157534201	T22–09 Red Hill, Oahu, HI	515	6		1		--
T22–10	212224157534802	T22–10 Red Hill, Oahu, HI	235	6		--		--
T23–01	212211157535201	T23–01 Red Hill, Oahu, HI	420	--		1		--
T23–02	212208157534801	T23–02 Red Hill, Oahu, HI	430	--		1		--
T23–05	212211157534101	T23–05 Red Hill, Oahu, HI	405	--		1		1
T23–06	212213157533801	T23–06 Red Hill, Oahu, HI	440	--		1		--
T23–07	212212157534602	T23–07 Red Hill, Oahu, HI	440	--		1		--
T23–08	212208157534501	T23–08 Red Hill, Oahu, HI	370	--		1		--
T23–09	212215157534301	T23–09 Red Hill, Oahu, HI	490	--		1		--
T23–10	212218157533701	T23–10 Red Hill, Oahu, HI	530	--		1		--
T3	212212157534701	T3 Red Hill, Oahu, HI	495	--		1		1
T5	212224157534801	T5 Red Hill, Oahu, HI	235	6		2		2
T8	212212157533801	T8 Red Hill, Oahu, HI	415	--		1		1

Comparison of Nonvolatile Dissolved Organic Carbon to Routinely Collected Analytes

A comparison of concentrations of NVDOC and DRO is shown in figure 6A. For this direct comparison of NVDOC and DRO, DRO concentrations reported in Jaeschke and others (2023) were divided by 1,000 and multiplied by 0.87 to convert from micrograms per liter into milligrams of C per liter (Zito and others, 2019). Nondetects for DRO were assigned a value of 20 μg/L (the laboratory’s limit of detection) before converting to milligrams of C per liter. The designated background well (Moanalua W–2) had an NVDOC concentration of 0.21 mg/L and DRO less than the detection limit. The NVDOC value is flagged as “estimated” by the analytical laboratory because it is less than the reporting limit but greater than the detection limit. Only well DH43 and one sample at RHMW12A had NVDOC concentrations within the range (0.19−0.27 mg/L) typical of DOC in the basalt aquifers of O‘ahu (Chapelle and others, 2013). Samples from the remaining wells all had elevated values of NVDOC compared to the background value. A linear regression through the data has a slope of 2.79 (coefficient of determination of 0.82), indicating that the NVDOC concentrations are about three times the DRO concentrations in the same wells. The NVDOC analysis method quantifies hydrocarbon oxidation products that are more polar than hydrocarbons and therefore missed in the DRO analysis (Bekins and others, 2016). Assuming no other substantial NVDOC inputs to groundwater near the facility, this result indicates that elevated NVDOC provides a measure of groundwater containing hydrocarbon oxidation products that might help to trace migration pathways of solutes away from the facility.

NVDOC is operationally defined as the organic fraction passing through a 0.45-micrometer filter and typically accounts for more than 90 percent of TOC in natural groundwaters (Regan and others, 2017). In this study, samples were filtered with a 0.22-micrometer filter (to filter out particulates and microorganisms) and compared with unfiltered TOC values from the same wells (fig. 6B, table 5). For this plot, samples with nondetects for TOC were assigned a value of 0.8 mg/L, which is the laboratory limit of detection for TOC. This comparison is provided because TOC rather than NVDOC has been routinely measured as part of the facility’s monitoring. The regression line shows a strong positive linear relation (coefficient of determination of 0.82) between these variables. In general, NVDOC is expected to be lower than TOC because NVDOC samples are filtered before analysis and TOC samples are unfiltered. The slope of the relation is about 0.8, indicating, as expected, that NVDOC is about 80 percent of the TOC; however, NVDOC concentrations were greater than TOC in four of the nine samples with detections of TOC, and at low TOC values, the regression gives NVDOC values more than 100 percent of TOC. The reason for this result is unknown. NVDOC is a more sensitive analytical approach. TOC was not detected in well RHMW09 to the south of the facility, or in wells RHMW11–05 and RHMW12A, which are screened below the water table; however, NVDOC was detected in these wells at concentrations of 0.41, 0.35, and 0.28 mg/L, respectively. These results indicate that NVDOC may be a better measure than TOC of the presence of hydrocarbon oxidation products.

Bailed and Pumped Sample Comparison

A component of this study included a comparison of bailed and pumped sample concentrations at three wells. The purpose was to determine if one sampling method resulted in consistently higher analyte concentrations. The spilled fuel at the facility is a light nonaqueous phase liquid, which is less dense than water and therefore will tend to form plumes in the uppermost part of the aquifer near the water table and near the source. Bailed samples from wells near the source were hypothesized to have greater concentrations of organic constituents (NVDOC and DRO) and biodegradation indicators (Fe, alkalinity) compared to pumped samples because bailed samples were collected near the water table. At locations farther from the source, the opposite might be true because plumes often migrate deeper with distance because of the addition of recharge at the water table or the presence of downward pressure gradients. In this study, bailer samples were collected from the upper 3−5 ft below the water table at wells RHMW2254–01, RHMW02, and RHMW03, whereas the pump intakes were set deeper at about 6, 9, and 8 ft below the water table, respectively (AECOM, written commun., August 3, 2023).

The shaft (RHMW2254–01) had higher DRO, NVDOC, and Fe in the bailed samples compared to the pumped samples (fig. 7, table 6). In contrast, well RHMW02 had lower DRO and NVDOC in the bailed samples compared to the pumped samples. At well RHMW03, the NVDOC and DRO results were mixed. Major ions (Al, alkalinity, B, Ba, Br, Ca, Cl, Cu, F, Fe, K, Li, Mg, Mn, Na, NO₃, PO₄, Si, SO₄, Sr, Zn) typically had higher concentrations in bailed samples compared to pumped samples; about 70 percent of the bailed samples had higher concentrations of major ions compared to the pumped samples (table 6). At the shaft (RHMW2254–01), all detected major ion concentrations were higher in the bailed samples compared to the pumped samples. Well RHMW03 had more analytes with lower concentrations in bailed samples compared to pumped samples relative to wells RHMW2254–01 and RHMW02.

In summary, the elevated NVDOC, DRO, Fe, and alkalinity concentrations in the bailed compared to the pumped samples at the shaft (RHMW2254–01) are consistent with a nearby source of light nonaqueous phase liquid. Differences in concentrations between the bailed and pumped samples were more variable at the other two wells, indicating that, in those locations, the source of light nonaqueous phase liquid is not as localized near the water table. The interpretation of RHMW03 data is discussed further with the alkalinity data in the following section.

Distribution of Petroleum Indicators

A plot of Kendall’s tau (τ) correlation coefficients between analytes measured in groundwater samples in this study is shown in figure 8. These analytes have been determined in previous studies to indicate petroleum degradation (National Research Council, 2000). Data from regular field samples were included for this analysis; field blanks and duplicates were excluded. Pumped samples from pumped-bailed sample pairs were also excluded because ¹⁴C content was not measured in the pumped samples. For all analytes except DRO, nondetects were assigned the value of the reported limit in the data release (for example, a value of LT 0.5 is assigned a value of 0.5). For DRO, nondetects were assigned a value of 20 µg/L, which is the laboratory’s limit of detection. In the case of ties, Kendall’s τ is computed according to Kendall (1945).

The results indicate that NVDOC correlates with percentage ancient C in NVDOC (τ=0.36) and with DRO (τ=0.52). NVDOC also correlates with Fe and Mn (τ=0.43, 0.52 respectively), consistent with biodegradation under Fe- and Mn-reducing conditions. NVDOC does not correlate as strongly with either NO₃, SO₄, specific conductance, pH, or alkalinity (τ=−0.14, −0.15, −0.16, −0.24, 0.18 respectively), indicating that these traditional measures of biodegradation might not be useful to delineate the contaminant plume at this site. Instead, specific conductance correlates well with SO₄, indicating that mixing of groundwater with a seawater source has a larger effect on the SO₄ concentration than biodegradation. The inverse correlation of pH with alkalinity is expected. In summary, the indicators that are useful for tracking the plume migration are NVDOC, percentage ancient C, Fe, and Mn.

NVDOC concentrations in the samples from the wells are compared to the ancient C contents of the NVDOC in figure 9. This plot includes all regular field samples that had measurements of NVDOC and ¹⁴C; field duplicates are excluded. The NVDOC and ¹⁴C data together indicate the presence of hydrocarbon oxidation products in groundwater below the facility. At background well Moanalua W–2, the ancient C is 38.20 percent. Higher ancient C values than this are indicative of the addition of hydrocarbon oxidation products to the NVDOC pool. The highest percentage ancient C (92.99 percent) and the highest NVDOC (6.45 mg/L) were detected at well RHMW02, indicating that groundwater near this well is the most affected by a fuel source. Overall, a trend of decreasing percentage ancient C with decreasing NVDOC concentrations indicates that the hydrocarbon oxidation products are either biodegrading or mixing with uncontaminated groundwater or both.

Two outliers are noteworthy. Groundwater from well RHMW05 has a relatively high NVDOC concentration of 2.22 mg/L but the ancient C is 34.95 percent, which is lower than the background value of 38.20 percent. Groundwater from this well had a DRO concentration of 380 µg/L, and DRO was detected at RHMW05 in 13 percent of samples collected in 2023 (table 1). Conversely, groundwater from well RHMW11–05 had a high ancient C value of 64.29 percent but a barely elevated NVDOC concentration of 0.35 mg/L. DRO values in this well were estimated concentrations of 66 and 84 µg/L. Based on the ancient C percentage, groundwater from this well could be affected by hydrocarbon oxidation products; however, the sampled interval of this well is −65 to −79 ft relative to local mean sea level or about 85−100 ft below the water table. One possible explanation for the very low NVDOC concentration is dilution and dispersion of the plume during vertical migration from the source to the sampled interval. Examining other aspects of the NVDOC using techniques in Podgorski and others (2021) such as optical properties or ultrahigh resolution mass spectrometry could be used to determine if NVDOC matches the characteristics of the natural NVDOC or that of the hydrocarbon oxidation products.

The background ancient C value of 38.20 percent is surprising because there is 0 percent ancient C at sites with shallow water tables such as the Bemidji site. The ancient C percentage of 38.20 percent corresponds to an age of 3,870 years for the background NVDOC. Such a large age is unreasonable for the transport time of NVDOC from the surface by the groundwater based on recharge dates of the 1940s to the 1980s for groundwater in this part of O‘ahu (Hunt, 2004). Other investigators have determined that the use of DOC for estimating groundwater age is questionable for thick unsaturated zones because of the potential for old, buried soil C to be transported in recharge water (Thomas and others, 2021). A possible source of the ancient NVDOC is burned vegetation between lava flows. This hypothesis could be tested by examining the optical properties of the background NVDOC or characterizing it with ultrahigh resolution mass spectrometry (Podgorski and others, 2021).

Maps showing contours of NVDOC, ¹⁴C content (percentage ancient C), DRO, and alkalinity are shown in figures 10–13. The data are combined from sample days in September 2022, January 2023–February 2023, and April 2023. Only regular field samples were used for these plots; field duplicates were excluded. Most wells were only sampled once, but wells RHMW02 and RHMW03 were sampled in September 2022 and January 2023. For these sites, the average of the September 2022 and January 2023 samples was used for plotting each analyte except DRO in well RHMW03 (fig.  2). Only the September 2022 concentration was used for RHMW03 because the January 2023 concentration was less than detection and nondetects occur infrequently at this well (table 1). Values plotted are from samples acquired with a bailer, except for samples from RHMW11–05, RHMW14–03, DH43, RHP05 and RHMW12A. Well locations (U.S. Geological Survey, 2024) and chemical concentration data were used to interpolate a concentration surface for each analyte using the Natural Neighbor method provided in the Spatial Analyst Toolbox of ArcGIS Pro version 3.1.3 (Esri, 2023). From these surfaces, concentration contour lines were created using the Contour tool in the same toolbox (Esri, 2023).

A contour map of NVDOC concentrations in groundwater is shown in figure 10. The lowest value of 0.21 mg/L, measured at the Moanalua well W–2, is assumed to represent background groundwater. It falls in the range of previous values measured by Chapelle and others (2013) on O‘ahu (0.19−0.27 mg/L) and is low for natural groundwater (Regan and others, 2017). Values measured in the three blank samples were 0.20, 0.26, and 0.30 mg/L, which are in the same range as the background sample; thus, values greater than 0.3 mg/L are considered elevated. Values less than 0.3 mg/L were detected at wells DH43 (0.21, 0.23 mg/L) and RHMW12A (0.27, 0.28 mg/L). The sampled interval of RHMW12A is −174 to −194 ft relative to local mean sea level (194–215 ft below the water table); thus, the low values at RHMW12A likely reflect the natural background NVDOC. Concentrations in groundwater at all other wells exceed 0.3 mg/L with the highest at well RHMW02 (3.33, 6.45 mg/L). The second highest concentration was from well RHMW03 (2.74, 2.92). Away from the ridge, concentrations drop rapidly by a factor of 5.2 in the approximately 700 ft between RHMW03 and RHMW10. Values at the two sampled periphery wells, RHP04A and RHP05, were 0.63 and 0.38 mg/L, respectively. Examining the overall shape and extent of the contours indicates that wells on the north and south side of the ridge are elevated; however, elevated values do not extend to well DH43 on the south side or well RHMW12A on the north side. The concentrations of NVDOC in groundwater decrease to about 10 percent of the highest concentration measured under the tanks within a distance of 700 feet to the south and 1,000 feet to the north of the ridge. The NVDOC data alone indicate no clear direction of contaminant migration away from the sources associated with the facility spills.

A contour map of the percentage of ancient C in NVDOC is shown in figure 11. As in figure 10, groundwater from well RHMW02 is most affected by hydrocarbon oxidation products with more than 90 percent ancient C in NVDOC. The contours in figure 11 indicate a broad south-southeast to north-northwest pattern of migration of ancient C from RHMW02 toward wells RHMW14–03 and RHMW11–05. However, the sampled intervals of wells RHMW14–03 and RHMW11–05 are quite deep at −140 to −158 ft relative to local mean sea level (160–179 ft below the water table) and −66 to −79 ft relative to local mean sea level (86–100 ft below the water table), respectively, calling into question whether or not the elevated ancient C values measured in these wells reflect compounds from a fuel source. The samples could be analyzed by optical methods to determine if the dissolved organic C is natural or from the facility (Podgorski and others, 2021). Groundwater from wells RHMW04, RHMW10, RHP05, RHP04A, and Red Hill Shaft RHMW2254–01 have slightly elevated ancient C values of 40–50 percent. Other techniques are needed to analyze the NVDOC from these wells to determine if the ancient C comes from biodegradation of fuel hydrocarbons.

A contour map of DRO concentrations in groundwater is shown in figure 12. The lowest contour line is set to 200 µg/L. Groundwater in the four tunnel wells exceeded the reporting limit of 190–220 µg/L, with the highest values measured at well RHMW02 (2,500 µg/L in September 2022 and 1,100 µg/L in January 2023). The values drop to the south and north of the facility. As with the ancient C map (fig. 11), the contours in this figure indicate a broad south-southeast to north-northwest direction of migration of DRO from well RHMW02 toward wells RHMW14–03 and RHMW11–05, but DRO was less than detection at RHMW14–03. The contours are dominated by the high value at well RHMW02 and provide little new information. Comparing this DRO map with the NVDOC plot in figure 10 and the ancient C map in figure 11 illustrates how the lower detection level of NVDOC combined with percentage ancient C can provide greater insight into the migration directions revealing the presence of elevated NVDOC and ancient C on the south side of the ridge. The DRO results in this study compare well with the monitoring data from 2016 to 2023 summarized in table 1. The frequencies of detection of DRO in all samples from the three tunnels wells, RHMW01R, RHMW02, and RHMW03, were 0.99, 1.00, and 0.93, respectively. For RHMW05, the overall frequency of DRO detection was 0.36 with higher frequencies in 2021 and 2022 compared to other years.

A contour map of carbonate alkalinity is shown in figure 13. All biodegradation reactions except fermentation produce CO₂, which raises the carbonate alkalinity. The background value at Moanalua well W–2 (fig. 2) is 84.1 mg/L as HCO₃. Higher values than this are considered an indicator of biodegradation, whereas lower values indicate less evidence for presence of a contaminant plume. Groundwater samples with alkalinity values less than the background alkalinity value were from wells RHMW12A, RHMW14–03, RHMW2254–01, RHMW09, RHMW10, RHMW05, and RHMW19. The elevated values on the map indicate a different pattern than the NVDOC (fig. 10) and the percentage ancient C (fig. 11). In this case, the highest values are at well RHMW03 (335 and 343 mg/L as HCO₃ for bailed samples) and the second highest values are at RHMW02 (214 and 240 mg/L as HCO₃ for bailed samples). These data indicate that biodegradation of fuel in the unsaturated zone above well RHMW03 results in transport to the water table of CO₂. Also at this well, NVDOC was higher in the bailed sample compared to the pumped sample (fig. 7) consistent with a source of partial transformation products in the unsaturated zone. Previous studies have established that aerobic biodegradation of petroleum hydrocarbons and methane lead to increased temperatures in the vadose zone and underlying groundwater (for example, Sweeney and Ririe, 2014; Warren and Bekins, 2015; McHugh and others, 2020). Mitchell and Oki (2018) determined that the groundwater temperature in well RWMW03 was 2.4–2.5 °C warmer than RHMW02. Likewise, McHugh and others (2020) measured temperatures about 2 °C higher at RHMW03 than RHMW02. These observations of increased temperatures support the conclusion from the alkalinity data that there is aerobic biodegradation of fuel in the unsaturated zone above well RHMW03. Samples from both periphery wells, RHP04A and RHP05, provide evidence of increased alkalinity consistent with biodegradation. Alkalinity at Red Hill Shaft RHMW2254–01 is not elevated. Groundwater from well DH43 had elevated alkalinity but no other clear signs of the presence of hydrocarbon oxidation products based on the NVDOC concentrations and ancient carbon percentage. Petroleum hydrocarbons in the diesel and oil range were detected at well DH43 in 2022 on February 23 (2.53 mg/L), March 6 (1.83 mg/L), and May 5 (0.46–0.06 mg/L) (Honolulu Board of Water Supply, 2024). These concentrations decreased with time from greater than 2 mg/L in February to less than 0.5 mg/L in May. The decreasing trend is consistent with the lack of evidence for fuel contamination by the time of sampling on April 11, 2023, in this study.

Carbon-14 in Soil Probe Samples

All but two soil gas samples at the facility had ¹⁴C age-corrected percentage modern values higher than annual average atmospheric ¹⁴C during the study period (Jaeschke and others, 2023). The annual average atmospheric ¹⁴C was estimated to be about 99 and 98 percent modern in 2022 and 2023, respectively (table 7). These values were estimated with equation 1.6 in appendix 1. Equation 1.6 is a linear regression equation fit to annual average atmospheric ¹⁴C values for 2005–15 from Graven and others (2017). Samples with ¹⁴C contents greater than atmospheric ¹⁴C indicate that ancient fossil fuel-derived CO₂ is not present in detectable quantities in soil gas at land surface.

Two samples at site T3 (fig. 5), a regular sample and a field duplicate sample, had ¹⁴C contents (97.85 and 97.35 percent modern, age corrected) slightly less than the estimated annual average atmospheric ¹⁴C of 98 percent modern in 2023. This difference is too small to confidently classify the T3 results as a detection of ancient C because of variability in atmospheric ¹⁴C and measurement error. During 2020, 12 samples of atmospheric CO₂ were collected at Mauna Loa Observatory and analyzed for ¹⁴C content (about 200 miles from the facility). These samples varied by about 1 percent modern (10 per mil); similar annual ranges in atmospheric ¹⁴C were observed at stations in La Jolla, Calif.; Samoa; and the South Pole (Graven and others, 2022). This atmospheric variation in ¹⁴C, coupled with the laboratory measurement error (0.2 percent modern), and the standard deviation of the two samples at site T3 (0.35 percent modern) indicates that the ¹⁴C contents of CO₂ at site T3 are not distinguishable from atmospheric ¹⁴C contents and are therefore not classified as detections of ancient C.

The average ¹⁴C content of CO₂ in soil gas samples (including regular and duplicate samples) at the facility was 100.43 percent modern (age corrected), which is about 2 percent higher than average atmospheric ¹⁴C contents in 2022 and 2023. Because the ¹⁴C of soil CO₂ is generally higher than atmospheric ¹⁴C at the facility, it indicates that a substantial part of the soil CO₂ is produced from biodegradation of soil organic matter containing “bomb” ¹⁴C deposited in the soil since the 1950s and early 1960s. After peaking in the late 1960s, atmospheric ¹⁴C contents have been declining (Graven and others, 2017). Between June 2017 and June 2023, atmospheric ¹⁴C contents decreased from about 101 to 98 percent modern (age corrected) according to equation 1.6. Positive ¹⁴C values (values of ∆¹⁴C greater than 0 or percentage modern values greater than 100) of soil CO₂ indicate “bomb” ¹⁴C whereas negative ¹⁴C values indicate CO₂ that is derived from a soil C pool with an average turnover time long enough to reflect radioactive decay of cosmogenic ¹⁴C (Wang and others, 2000).

The ¹⁴C contents of CO₂ in soil probe samples collected at the facility in 2022 and 2023 varied by almost 6 percent, from 97.35 to 103.06 percent modern (age corrected by applying equations 1.1 and 1.2 in appendix 1 to ¹⁴C data in Jaeschke and others [2023]). The wide range of soil ¹⁴C values is likely the result of variability across sample sites in soil moisture, temperature, and vegetation types. These are characteristics that are known to affect root respiration rates (Raich and Tufekcioglu, 2000). The sites sampled at the facility had different vegetation types, ranging from grass covered to scrub forest, to forested. Surface soil temperatures were variable during sample collection, ranging from about 23 to 36 °C for samples collected in September 2022 (Jaeschke and others, 2023). Soil moisture in the upper 10 cm of soil also varied substantially, from about 5 to 31 percent, during the September 2022 sample collection (Jaeschke and others, 2023). Soil temperature and moisture were not measured during the January 2023 sample collection, but samples were collected at different times of day and some sites were rarely exposed to direct sunlight because of their position on slopes.

Soil Carbon Traps

In 2022 and 2023, 8 traps were deployed at 6 sites at the facility (fig. 5). No detections of ancient fossil fuel-derived CO₂ were reported by the laboratory for any of these trap samples (table 7; Jaeschke and others, 2023). These results are consistent with the soil probe samples and support the conclusion that no detectable ancient fossil fuel-derived CO₂ was present in soil at the locations sampled for this study; however, the 2022–23 trap results differ from 2017 reported detections at sites T1, T2, T3, and T8 (Department of the Navy, 2019; McHugh and others, 2020). Detections of ancient fossil fuel-derived C were also reported for tunnel air intake samples and upper and lower tunnel air exhaust samples collected in 2017 (table 7; Department of the Navy, 2019; McHugh and others, 2020). The mass of ancient fossil fuel CO₂ trapped on the sorbent material (CO₂ ff) for each of these samples was calculated by equation 1 using analytical data measured in the laboratory. The values used for these calculations are provided in table 1.1 in appendix 1 and Jaeschke and others (2023). The reported masses of ancient fossil fuel-derived CO₂ (CO_{2 ff}) for 2017 trap samples were near zero (0.002–0.085 gram, table 7), which is the operational “detection limit” of equation 1.

Positive CO_{2 ff} values from equation 1 are classified as detections, but CO2 ff values that are less than or equal to zero are classified as nondetects. Nondetects are represented as “ND” in table 7, table 8, and table 1.1 in appendix 1. The calculated CO_{2 ff} values from equation 1 that are less than 0 are shown below the zero line in figure 14A and B. Positive CO_{2 ff} values are shown as numeric values in table 7, table 8, and table 1.1 in appendix 1 and shown above the zero line in figure 14A and B. The laboratory provided ancient C efflux rates in micromoles per square meter per second (µmol/m²/s) for all samples classified as detections. In the following discussion, the sensitivity of classifying results as a detection or nondetect from equation 1 is examined.

Equation 1 corrects sample results for the CO₂ content and ¹⁴C composition of the trap sorbent from the travel blank (table 1.1 in appendix 1). Equation 1 relies on the variable ff_sample, which is the fossil fuel fraction of a sample calculated according to equation 1.5 in appendix 1. Equation 1.5 relies on an assumed value of age-corrected background C content (pMC_bg). As pMC_bg for a given sample and travel blank combination increases, the calculated ancient fossil fuel CO₂ mass also increases (table 8, fig. 14). Therefore, for samples near the reporting limit, classification of results as detections by equation 1 is sensitive to the assumed age-corrected background ¹⁴C content value (pMC_bg).

E-Flux assumed that age-corrected background CO₂ from natural soil respiration processes (pMC_bg) is equal to the atmospheric ¹⁴C content (Department of the Navy, 2019; Jaeschke and others, 2024). Under this assumption, E-Flux used a fixed value of 102 for pMC_bg for the 2022–23 samples (Jaeschke and others, 2023) based on literature values (Cerling and others, 2016; Larsen and others, 2018). E-Flux assumed a fixed value of 105 for pMC_bg for the 2017 samples (Department of the Navy, 2019) also based on a literature value available at the time of sample analysis (Hua and others, 2013). E-Flux’s approach is typical in that literature-based annual average ¹⁴C values are considered to be representative of conditions over very large areas, such as the entire northern or southern hemisphere (Graven and others, 2017, 2020). However, this study’s soil probe field data and other comprehensive published compilations of atmospheric ¹⁴C data (Graven and others, 2017, 2020) indicate the age-corrected background ¹⁴C values used by E-Flux were likely high and indicate the trap results could be reevaluated using more reasonable values for background ¹⁴C content.

Evaluating possible ancient fossil fuel CO₂ detections across a range of reasonable pMC_bg values is a careful approach that provides more context for classifying sample results as detections compared to evaluating possible detections at a single literature value. For the following evaluations, samples are considered to have likely detections when equation 1 produces positive CO_{2 ff} values across the entire site-specific reasonable range of pMC_bg values listed in table 7 and table 8 and shown on figure 14. Each sample was evaluated by plugging in the age-corrected background ¹⁴C contents (pMC_bg) listed in table 8 into equation 1.5 and then using the resulting ff_sample values in equation 1 along with the reported sample values in table 1.1 in appendix 1.

Reasonable ranges for age-corrected background ¹⁴C contents in CO₂ (pMC_bg) are from published atmospheric ¹⁴C data and from soil probe samples collected in this study (table 7). The ¹⁴C content of atmospheric CO₂ is trending down through time (Graven and others, 2017, 202225). Data from this study demonstrate that ¹⁴C contents of soil CO₂ are variable across the facility. In 2022 and 2023, the average age-corrected atmospheric ¹⁴C was 98 to 99 percent modern (eq. 1.6 in appendix 1, based on data from Graven and others [2017]) and the average ¹⁴C of soil CO₂ from soil probe samples was 100.43 percent modern (age-corrected), or about 2 percent higher than atmospheric ¹⁴C. Therefore, 2022–23 samples were evaluated over a reasonable pMC_bg range of 98–101. In some cases, a soil probe sample had a ¹⁴C content more than 2 percent greater than the average atmospheric value, so the soil probe sample concentration was set as the upper limit to the reasonable range rather than 101 (table 7, table 8). Assuming the ¹⁴C of soil CO₂ in 2017 was also, on average, about 2 percent higher than annual average atmospheric ¹⁴C, a reasonable pMC_bg range of 101–103 is appropriate for evaluating the 2017 samples. Evaluation of the 2022–23 samples across the pMC_bg range of 98–101 still indicates no detectable ancient fossil-fuel derived CO₂ for all samples (table 8).

The reevaluation of reported ancient fossil fuel ¹⁴C results from the 2017 samples collected at the facility (Department of the Navy, 2019; McHugh and others, 2020) is listed in table 8 and figure 14A and B. Each line in figure 14A and B represents one field sample being evaluated over a range of age-corrected pMC_bg values. The gray shading on figure 14A and B shows the reasonable range of age-corrected background ¹⁴C content (pMC_bg) values (x-axis) over which the 2017 samples were evaluated. The reasonable ranges do not overlap the background ¹⁴C content of 105 percent modern, which is the value used by E-Flux in sample reports (Department of the Navy, 2019). When the pMC_bg value is 105, all four soil trap samples (T1, T2, T3, T8) are classified as detections (above the zero line), as was reported previously (fig. 14A); however, within the reasonable range of pMC_bg values, none of the samples’ calculated ancient CO₂ masses are entirely greater than zero. Therefore, according to the metrics proposed in this report, none of the 2017 soil trap samples are classified as detections because none of the samples are always positive across the reasonable range of pMC_bg values (fig. 14A and B). Furthermore, when pMC_bg is set to 105 for the 2022–23 samples, the range of calculated ancient fossil fuel CO₂ masses for the 2022–23 samples overlaps with the calculated masses for the 2017 samples (table 8). This is further evidence that the analytical results for the soil trap samples are similar from the two studies and the detections of ancient C reported from the 2017 samples were affected by the selection of a high age-corrected background ¹⁴C value.

In 2017, cartridges were used to collect CO₂ from the tunnel air intake and exhaust systems (Department of the Navy, 2019; McHugh and others, 2020). These cartridges were not placed on soil so evaluating the trap results over a range of possible atmospheric ¹⁴C contents was more appropriate than evaluating trap results over a range of ¹⁴C from soil CO₂. The average atmospheric value in 2017 was about 101 percent modern (age-corrected). Previous studies have determined that atmospheric ¹⁴C varies by about 1 percent modern (10 per mil) within 1 year at the Mauna Loa Observatory and other stations (Graven and others, 2022); therefore, the exhaust samples were evaluated over a pMC_bg range of 100–102 percent modern (table 8, fig. 14B). Both lower tunnel exhaust cartridge samples were classified as detections because the calculated ancient fossil fuel CO₂ mass was always greater than zero over the reasonable range of age-corrected pMC_bg values (fig. 14B, table 8). No other tunnel intake or upper tunnel exhaust samples were classified as detections because they were not always positive over the reasonable range of age-corrected background ¹⁴C values (table 8, fig. 14B).

In conclusion, only the lower tunnel exhaust samples from 2017 are classified as detections using a method that accounts for reasonable ranges of ¹⁴C contents of background CO₂. All other intake, exhaust, and soil trap samples collected in 2017 and 2022–23 are not classified as detections.

Dynamic Closed Chamber Soil Carbon Dioxide Efflux

Soil CO₂ effluxes measured with a dynamic closed chamber were highly variable at the facility, ranging from site average values of 1.34 to 7.72 μmol/m²/s (fig. 15). As previously discussed, none of the sites sampled in 2022–23 had detectable ancient C in soil CO₂; therefore, the variability in measured soil CO₂ effluxes only represents variability in natural soil respiration rather than variability caused by mixtures of natural soil respiration and petroleum biodegradation.

The variability in natural soil respiration at the facility is large compared to variability in measured soil CO₂ effluxes at the well-documented Bemidji site. At the Bemidji site, the magnitude of soil CO₂ effluxes has been used to distinguish sampling locations unaffected by petroleum biodegradation from “affected” locations where petroleum biodegradation is occurring (Sihota and others, 2011). At the Bemidji site, 30 years after a crude oil spill occurred, the average CO₂ efflux at unaffected locations was 3.29 µmol/m²/s compared to an average of 5.89 µmol/m²/s at affected locations where oil was degrading (Sihota and others, 2011). The individual efflux measurements at affected and unaffected locations were also distinct. The effluxes ranged from 2.63 to 4.45 µmol/m²/s at the unaffected locations and from 5.21 to 7.12 µmol/m²/s at the affected locations (Sihota and others, 2011). The range of natural soil respiration effluxes observed at the facility (6.38 µmol/m²/s) exceeds the range of mixed natural respiration and petroleum degradation effluxes observed at the Bemidji site (4.49 µmol/m²/s). Because of the inherent variability in natural soil respiration at the facility, a substantial increase in CO₂ efflux from petroleum biodegradation would be needed to distinguish natural and contaminant effluxes based solely on the magnitude of fluxes. The magnitude of soil CO₂ effluxes alone could not be used to estimate natural source zone depletion rates at the facility.