Volume 107, Issue B9 p. ECV 3-1-ECV 3-15
Chemistry and Physics of Minerals and Rocks/Volcanology
Free Access

Carbon dioxide emission rate of Kīlauea Volcano: Implications for primary magma and the summit reservoir

T. M. Gerlach

T. M. Gerlach

U.S. Geological Survey, Cascades Volcano Observatory, Vancouver, Washington, USA

Search for more papers by this author
K. A. McGee

K. A. McGee

U.S. Geological Survey, Cascades Volcano Observatory, Vancouver, Washington, USA

Search for more papers by this author
T. Elias

T. Elias

U.S. Geological Survey, Hawaiian Volcano Observatory, Hawai‘i National Park, Hawaii, USA

Search for more papers by this author
A. J. Sutton

A. J. Sutton

U.S. Geological Survey, Hawaiian Volcano Observatory, Hawai‘i National Park, Hawaii, USA

Search for more papers by this author
M. P. Doukas

M. P. Doukas

U.S. Geological Survey, Cascades Volcano Observatory, Vancouver, Washington, USA

Search for more papers by this author
First published: 18 September 2002
Citations: 121

Abstract

[1] We report a CO2 emission rate of 8500 metric tons per day (t d−1) for the summit of Kīlauea Volcano, several times larger than previous estimates. It is based on three sets of measurements over 4 years of synchronous SO2 emission rates and volcanic CO2/SO2 concentration ratios for the summit correlation spectrometer (COSPEC) traverse. Volcanic CO2/SO2 for the traverse is representative of the global ratio for summit emissions. The summit CO2 emission rate is nearly constant, despite large temporal variations in summit CO2/SO2 and SO2 emission rates. Summit CO2 emissions comprise most of Kīlauea's total CO2 output (∼9000 t d−1). The bulk CO2 content of primary magma determined from CO2 emission and magma supply rate data is ∼0.70 wt %. Most of the CO2 is present as exsolved vapor at summit reservoir depths, making the primary magma strongly buoyant. Turbulent mixing with resident reservoir magma, however, prevents frequent eruptions of buoyant primary magma in the summit region. CO2 emissions confirm that the magma supply enters the edifice through the summit reservoir. A persistent several hundred parts per million CO2 anomaly arises from the entry of magma into the summit reservoir beneath a square kilometer area east of Halemaumau pit crater. Since most of the CO2 in primary magma is degassed in the summit, the summit CO2 emission rate is an effective proxy for the magma supply rate. Both scrubbing of SO2 and solubility controls on CO2 and S in basaltic melt cause high CO2/SO2 in summit emissions and spatially uncorrelated distributions of CO2 and SO2 in the summit plume.

1. Introduction

[2] Kīlauea Volcano is among the most active and best studied volcanoes, long providing a natural laboratory for learning how hot spot volcanoes work. A broad understanding of Kīlauea has emerged over many years of study [Tilling and Dvorak, 1993]. Located on the southeastern flank of Mauna Loa Volcano, Kīlauea's southern flank is detached from the stable shield to the north by the east and southwest rift zones and the intervening Koa‘e fault system (Figure 1). The most active eruption sites are within the summit caldera region and along the rift zones. However, all eruptions and intrusions at Kīlauea, whether within the caldera or along the rift zones, appear to involve magma that first passes through a summit magma reservoir complex between about 1–2 km and 7 km beneath the summit. The primary magma supplied to the summit reservoir arrives at a relatively steady rate from mantle sources, but reservoir storage is transitory, as magma is discharged to the surface or into the rift zones. Nearly continuous eruptions in the summit caldera typified the activity for over a hundred years before 1924. Since then, most of the magma is intruded into the rift zones, following a period of summit reservoir storage, and most of the eruptions take place at sites on the east rift zone (ERZ).

Details are in the caption following the image
Location map for Kīlauea Volcano.

[3] Quiescent magma degassing during summit reservoir residence is thought to be of primary importance in removing CO2 from the mantle-supplied magma [Gerlach and Graeber, 1985; Greenland et al., 1985]. Since the Hawaiian hot spot is one of the Earth's most vigorous and best understood hot spots, the CO2 emission rate of Kīlauea is a benchmark of interest for volcanic CO2 discharge. Only two measurements of Kīlauea's summit CO2 emission rate are available from past work [Greenland et al., 1985], however, and although made within just two months of each other, they differ by over a factor of 2.

[4] In the summit region of Kīlauea, prevailing northeasterly trade winds carry gas emissions from the summit caldera to the southwest across the southern portion of Crater Rim Drive (Figure 2). No fumaroles or visible forms of gas emission exist downwind of this stretch of Crater Rim Drive. It has therefore been used regularly since 1979 for vehicle-based correlation spectrometer (COSPEC) measurements to constrain the summit SO2 emission rate [Casadevall et al., 1987; Elias et al., 1998; Sutton et al., 2001]. Exploratory measurements in 1995 with a nondispersive infrared (NDIR) CO2 analyzer and a closed-path Fourier transform infrared (CP-FTIR) spectrometer revealed high concentrations of CO2 and SO2 in the air along the summit COSPEC traverse (hereinafter simply “the traverse”). The high concentrations of CO2 and SO2 are easily distinguished from ambient levels, making it a straightforward task to determine volcanic CO2 and SO2 concentrations within the summit plume along the traverse.

Details are in the caption following the image
Summit region of Kīlauea Volcano. Heavy black lines outline innermost caldera boundaries and fissure and fault systems. Hachured enclosures are summit area craters. SWRZ and ERZ indicate the southwest rift zone and east rift zone (Figure 1) in the summit region. Puhimau thermal area is a site of warm bare ground in the upper ERZ. The borehole is the Keller research hole. The highlighted portion of Crater Rim Drive between stations 12 and 32 indicates the summit COSPEC traverse. Under normal trade wind conditions (vector), the summit plume is blown across the traverse. Points a, b, c, and d identify traverse segments with high plume CO2 concentrations (see text). Dashed lines project road distances for the 10 legs of the traverse to wind-normal distances.

[5] In this study, we use the volcanic CO2-to-SO2 ratio of the plume along the traverse and the synchronous SO2 emission rate to constrain Kīlauea's summit CO2 emission rate. We assess the spatial and temporal variability of summit CO2 and SO2 emissions and use the summit CO2 emission pattern to infer where primary magma enters the summit reservoir. We compare the new summit CO2 emission rate with previous estimates and use it (together with rift zone CO2 emission data) to set limits on the total CO2 output of the volcano and to constrain the bulk CO2 content of the primary magma. Then we determine the concentrations of dissolved and exsolved CO2 in primary magma and evaluate the importance of excess gas in making primary magma buoyant at summit reservoir depths. Next, we investigate the likelihood that primary magma will mix with resident reservoir magma. Finally, we examine the impact of SO2 scrubbing and of CO2 and S solubility in basaltic magma on the summit emissions, and we consider the summit CO2 emission rate as a proxy for the magma supply rate.

2. Procedures and Methods

[6] Twelve summit CO2 emission rate experiments were carried out on the summit traverse on three occasions over 4 years: the afternoon of 20 September 1995, the morning and afternoon of 20 October 1998, and the morning of 6 May 1999. Dry weather, clear skies, and northeasterly trade winds prevailed in each experiment.

2.1. Approach

[7] An indirect approach to determining the summit CO2 emission rate is to constrain it from the summit SO2 emission rate and the average CO2-to-SO2 ratio of summit emissions. In this approach, the mass emission rate of CO2 (ECO2) from the summit is calculated from
urn:x-wiley:01480227:media:jgrb12998:jgrb12998-math-0001
where CO2/SO2 is the average molar concentration ratio of the global summit emission, ESO2 is the mass emission rate of SO2 from the summit, and 0.69 is the ratio of the molecular weight of CO2 to SO2. Fractionation of the gases, whether before or after emission, does not cause error in ECO2; for example, scrubbing of SO2 by water (in the ground or the air) causes a higher CO2/SO2 and proportionately lower ESO2, so that the correct ECO2 is retrieved.

2.2. SO2 Emission Rate Measurements

[8] Summit SO2 emission rates were determined by vehicle-based traverse measurements with a COSPEC IV [Casadevall et al., 1987; Elias et al., 1998; Sutton et al., 2001]. Wind data for 20 September 1995 were obtained 1 m above the ground, before and after measurements, and corrected for ground effects [Elias et al., 1998; Sutton et al., 2001]; wind data on 20 October 1998 and 6 May 1999 were obtained 3 m above ground level with a continuous, summit wind-monitoring system [Elias et al., 1998]. Uncertainties in the wind speeds are about ±10%.

2.3. CO2 and SO2 Concentration Measurements

[9] Measurements of molar CO2 and SO2 concentrations in the air along the traverse were made with a LI-COR NDIR CO2 analyzer and a MIDAC CP-FTIR spectrometer, simultaneously with COSPEC measurements. The instruments and their use on volcanic emissions are described elsewhere [McGee, 1996; Gerlach et al., 1997, 1998, 1999; McGee and Gerlach, 1998].

2.4. Traverse Procedures and Instrument Configurations

[10] The CP-FTIR and CO2 analyzer systems were mounted on a vehicle, together with a COSPEC IV arranged as usual (telescope and upward directed right angle mirror positioned outside an open window about 1.5 m above road level), and driven at a nearly constant, low speed across the plume along the traverse. Sample intakes for the CP-FTIR and CO2 analyzer were through forward directed elbows at the ends of rigid PVC tubes placed outside an open window about 2.5 m above road level (well upstream of the vehicle's exhaust) and connected by flexible tubing to instrument inlet ports. Flow rates through the CO2 analyzer and CP-FTIR were 0.0167 L s−1 and 3.0 L s−1 respectively, resulting in air turnover times for the respective sample cells of about 0.7 s and 3.7 s. The CO2 analyzer produced one analysis per second. The CP-FTIR generated one SO2 analysis every 5–8 s, depending on instrument settings. Data sets were collected in both directions in 7–8 min at average speeds of 11.4 to 12.2 m s−1. At typical vehicle speeds, the CO2 analyzer averaged CO2 in air over traverse distances of ∼12 m; the CP-FTIR averaged SO2 over ∼60–95 m, depending on the pre-set sample time interval (5–8 s). The traverses yielded 440–473 CO2 measurements and 60–90 SO2 measurements, depending on vehicle speed and, in the case of SO2, the sample time interval.

2.5. Data Analysis and Estimation of Summit CO2/SO2

[11] Determination of the CO2 emission rate by equation (1) depends critically on CO2/SO2. Figure 3 illustrates the procedure used to estimate CO2/SO2. Figures 3a and 3b show CO2 and SO2 concentrations (molar ppm) measured along the traverse, together with ambient baseline fits. The high concentrations of CO2 and SO2 along the traverse eliminated ambiguity in distinguishing volcanic and ambient concentration levels. A baseline-processing algorithm available in peak analysis software (PeakFit, version 4) was used to select and fit ambient data points. Figures 3c and 3d show concentrations of volcanic CO2 and SO2 obtained after subtracting the baseline fits from the data. Occasional small negative concentrations generated by baseline subtraction in the ambient regions were zeroed.

Details are in the caption following the image
Derivation of volcanic CO2 and SO2 concentrations (ppm molar) in the plume along the summit COSPEC traverse for experiment H on 20 October 1998. (a, b) Measured CO2 and SO2 concentrations (solid) and ambient CO2 and SO2 baselines (dash); distances are road distances from station 12 to station 32 (Figure 2). (c, d) Volcanic CO2 and SO2 concentrations after subtracting ambient baseline concentrations; sporadic small negative values at ends of traverse have been zeroed; distances are road distances as in Figures 3a and 3b. (e, f) Volcanic CO2 and SO2 concentration profiles after converting road distances to wind-normal distances (see text and Figure 2).

[12] In Figures 3e and 3f, the road distances between data points are corrected to wind-normal distances. The traverse is crooked but resolvable to 10 legs (Figure 2), generally not normal to the wind direction. This problem, common in road-based COSPEC traverses [Milan et al., 1976], is handled by simple trigonometric calculation of wind-normal distances along traverse legs. The correction is illustrated by wind-normal projections of traverse legs in Figure 2. Wind-normal distances, being generally shorter than corresponding road distances, contribute less area under the concentration profiles (Figures 3e and 3f) than do road distances (Figures 3c and 3d). In effect, the data are weighted, with those points along traverse legs more nearly normal to the wind contributing proportionately more area under concentration profiles than those at angles less nearly normal to the wind. The areas (in units of ppm m) under the corrected concentration profiles were calculated precisely by integration in the peak analysis software. Since the wind-normal traverse length is the same for each species, the ratio of the areas is equal to the average wind-normal CO2/SO2 concentration ratio of the traverse. We used this value in equation (1) to estimate CO2/SO2 the average CO2-to-SO2 concentration ratio of the global summit emission.

3. Results

[13] Figures 4 and 5 and Tables 1 and 2 present the results of the 12 summit CO2 emission rate experiments carried out on 3 days over the 1995–1999 period. Figures 4 and 5 display concentrations of volcanic CO2 and SO2, i.e., concentrations above ambient, as a function of road distance and wind-normal distance along the traverse. Table 1 contains all the data used to calculate the CO2 emission rates from equation (1). Emission rates for CO2 and SO2 are in metric tons per day (t d−1)—i.e., 1 t d−1 is 103 kg d−1. Table 2 gives statistics for volcanic CO2 and SO2.

Details are in the caption following the image
Volcanic CO2 concentrations in ppm (molar) in the plume along the summit COSPEC traverse (Figure 2) after subtraction of ambient CO2. Text on graphs designates experiments A, B, … S (Table 1). Black line plots show concentrations as a function of road distance along the traverse from station 12 (Figure 2); gray line plots show concentrations as a function of wind-normal distances, which vary with wind direction (Table 1 and Figure 2).
Details are in the caption following the image
Same as Figure 4 but for SO2.
Table 1. Results of Summit CO2 Emission Rate Experiments
Experiment Hawaiian Local Time Wind Direction (East of North), deg Wind Speed, m s−1 Wind-Normal Traverse Length, m Volcanic CO2 Area,a ppm m Volcanic SO2 Area,a ppm m Summit Emission CO2/SO2,b molar SO2 Emission Rate,c t d−1 CO2 Emission Rate,d t d−1
20 September 1995
A 1530 40 4.6 3435 114,927 2076 55.4 239.5 9100
B 1554 40 4.6 3435 110,002 2040 53.9 221.7 8200
C 1618 40 4.6 3435 133,330 2493 53.5 221.8 8200
Mean 8500
SD 500
20 October 1998
H 1149 25 4.7 3911 212,563 2936 72.4 152.4 7600
I 1203 25 4.7 3911 197,588 2737 72.2 175.5 8700
J 1210 25 4.7 3911 220,190 2522 87.3 140.8 8500
K 1634 14 3.5 4088 287,858 2932 98.2 127.5 8600
Mean 8400
SD 500
6 May 1999
O 1016 35 4.0 3621 99,321 753.8 131.8 102.7 9300
P 1031 35 4.0 3621 105,216 574.5 183.1 61.5 7800
Q 1044 35 4.0 3621 73,806 623.4 118.4 102.1 8300
R 1057 35 4.0 3621 67,885 467.1 145.3 85.8 8600
S 1109 35 4.0 3621 74,178 534.4 138.8 94.0 9000
Mean 8600
SD 600
  • a Based on integration of wind-normalized CO2 and SO2 concentration profiles (Figures 4 and 5), as described in text.
  • b Traverse-based estimate of summit CO2/SO2 in equation (1) calculated from volcanic CO2 and SO2 areas, as described in text.
  • c Significant figures inflated to reduce round-off errors in calculations with equation (1).
  • d Calculated from equation (1) for summit CO2/SO2 and SO2 emission rate shown and rounded to nearest 100 t d−1.
Table 2. Volcanic CO2 and SO2 Concentrations Along the Summit COSPEC Traversea
Experiment CO2 SO2
N Mean SD Range N Mean SD Range
A 440 27.7 46.3 240.7 90 0.67 0.65 2.77
B 440 25.2 44.2 220.0 89 0.56 0.42 1.62
C 440 32.5 47.3 292.3 89 0.77 0.91 7.15
H 461 45.6 90.2 754.1 62 0.69 0.73 3.79
I 462 42.7 78.3 685.3 62 0.66 0.50 1.68
J 461 48.3 111.7 964.8 61 0.60 0.47 1.75
K 465 61.2 98.9 810.7 60 0.68 0.57 2.73
O 473 23.4 55.8 620.6 87 0.20 0.22 1.06
P 461 22.9 64.3 714.5 85 0.15 0.23 0.88
Q 462 17.1 37.3 278.9 83 0.17 0.20 0.91
R 463 17.1 33.2 325.8 74 0.14 0.20 0.95
S 469 16.8 41.7 406.8 74 0.16 0.35 2.53
  • a N is number of samples in traverse experiment. Mean, standard deviation, and range are in units of ppm (molar). Range values are also the maximums, since minimums are zero.

3.1. Summit CO2 Emission Rates

[14] The means of the summit CO2 emission rates on the three days of experiments fall in a narrow range of 8400–8600 t d−1 (Table 1). Standard deviations are within 6–7% of the means. Since all CO2 emission rates are in close agreement, we adopt the 8500 t d−1 mean of the 12 experiments as the best estimate of the summit CO2 emission rate. The 95% confidence interval of this estimate is ±300 t d−1.

[15] The narrow range of the summit CO2 emission rates is noteworthy given the temporal variation of the SO2 emission rates and traverse-based estimates of the global summit CO2/SO2 (Table 1). The CO2/SO2 estimates vary inversely and linearly over a wide range SO2 emission rates (Figure 6), keeping the summit CO2 emission rates nearly constant; a negative correlation exists in the 12 data sets over the long-term (r = −0.95) and in the short-term variations on a particular date. Thus, although the concentration profiles of Figures 4 and 5 are by no means “carbon copies” of each other, and despite large and more or less independent variations of volcanic CO2 and SO2 along the traverse, a strong underlying constraint affects CO2/SO2.

Details are in the caption following the image
Traverse-based estimates of global summit CO2/SO2 versus summit SO2 emission rates for all experiments (Table 1). Results for experiments B and C overlap. Solid line is calculated least squares fit (r2 = 0.90).

3.2 Segregation of CO2 and SO2

[16] Concentrations of volcanic CO2 are higher along the central to eastern part of the traverse at road distances >2.0–2.5 km (Figure 4). Most of the CO2 occurs at road distances between ∼2.1 km and ∼3.7 km (Figure 4), corresponding to locations between a and d along the traverse (Figure 2). Indeed, the traverse section between a and d contributes 70–90% of the summit CO2 emission rate (6000–7700 t d−1). The biggest CO2 peaks commonly occur between ∼2.6 km and ∼3.1 km (b to c, Figure 2), extending in some cases to ∼3.5 km, about midway between c and d. Secondary CO2 peaks often form at ∼4.0 km and ∼4.5 km between d and Keanakāko‘i Crater.

[17] In contrast, SO2 is more ubiquitous but tends to be higher along the western half of the traverse at road distances <3.0 km (Figure 5). Only 25–60% of the area under its wind-normalized profiles (Figure 5) is contributed by the a–d section of traverse; it tends to form broader, more spread out highs, especially to the west downwind of Halemaumau (Figure 2). SO2 profiles tend to lack the extremely anomalous peaks observed for CO2 along the a–d section of traverse. In part, this reflects smoothing of SO2 data by the larger averaging distance of the CP-FTIR, but this should not remove SO2 peaks altogether. Furthermore, it does not explain why SO2 peaks tend to occur to the west of CO2 peaks, or why higher concentrations in general extend farther west for SO2 than for CO2.

[18] Thus volcanic CO2 and SO2 concentrations generally lack spatial correlation along the traverse. Figure 7 highlights the segregation of volcanic CO2 and SO2 along the traverse for all measured concentrations greater than one standard deviation above experiment means (Table 2). Nearly all the high SO2 concentrations occur along the western half of the traverse. High CO2 concentrations begin near 2.1 km (a, Figure 2) and rise sharply at ∼2.6 km (b, Figure 2). They are confined mainly to the eastern half of the traverse with a large cluster between 2.6 km and 3.5 km and two small clusters at 4.0 km and 4.5 km.

Details are in the caption following the image
Volcanic CO2 (solid circles) and SO2 (open circles) concentrations greater than one standard deviation above their respective means (Table 2) in experiments along the summit COSPEC traverse. The x axis is traverse road distance from station 12 (Figure 2).

3.3. High CO2/SO2

[19] Although the concentrations of volcanic CO2 and SO2 vary greatly and independently, CO2 generally exceeds SO2 (commonly by a great deal) all along the traverse. (This may not be readily apparent from Figures 4 and 5 because of huge differences in y axis scaling.) The dominance of CO2 is reflected in the global summit CO2/SO2 values, which range from 54 to 183 (Table 1). Estimates of CO2/SO2 based on means from direct averaging of traverse CO2 and SO2 concentrations (Table 2) are also high (41 to 156) without weighting for wind normalization.

4. Discussion

4.1. Possible Sources of Error

4.1.1. Misrepresentative CO2/SO2 Values

[20] Obtaining a CO2/SO2 representative of the global summit emission is essential when using equation (1) to constrain the summit CO2 emission rate. The negative correlation of traverse-estimated CO2/SO2 with SO2 emission rate in both the long and short term (Figure 6) and the nearly constant summit emission rate obtained for CO2, despite the variable SO2 emission rates and the independent variations of CO2 and SO2 within the plume, are most improbable unless the CO2/SO2 estimates are representative. In our opinion, this is attributable to (1) the high density of data along the traverse and (2) the sampling of the plume where it is close to its sources, concentrated, and compact. Fixed-wing, helicopter, and remotely piloted aircraft surveys of CO2 and SO2 concentrations at several elevations confirm that the core of the plume extends from ground level to a few hundred meters above the ground along the traverse (A. J. Sutton, U.S. Geological Survey, unpublished data, 2001).

4.1.2. Contamination

[21] The need to make COSPEC measurements during daylight in a national park made occasional vehicle encounters inevitable. CO2 from vehicle exhaust is thus a potential pollutant along the traverse. Positioning the instrument intake 2.5 m above road level reduced the impact of vehicle exhaust. The CO2 record showed that exhaust effects ranged from insignificant to small spikes (<10 ppm for 2–4 s) that were easily removed. Finally, to verify that the summit CO2 anomaly was volcanic and not due to aggregate park vehicle exhaust, we ran two CO2-measuring experiments at night when no traffic was present. The concentration profiles had the same pattern as in the daytime experiments, with a similar prominent peak along the traverse between a and d (Figure 2).

4.1.3. Overlooked Emission Sources

[22] We occasionally extended the measurements beyond the traverse in search of additional emission sources in the summit and upper east rift region. Measurements along Crater Rim Drive around the entire caldera (Figure 2) revealed no additional sources. We made no additional measurements south (downwind) of the traverse; although there are no fumaroles or obvious signs of gas release (steam, bluish SO2 fume, SO2 odor) in this region, we cannot reject the possibility of some diffusive CO2 release. Measurements downwind of the upper ERZ along Chain of Craters Road from its intersection with Crater Rim Drive (Figure 2) to Mauna Ulu (∼2 km southeast of Figure 2) showed no additional sources. Finally, we investigated a site of steaming ground known as the Puhimau thermal area (Figure 2). This area does release CO2, but soil efflux surveys give emissions of <25 t d−1 (K. A. McGee, U.S. Geological Survey, unpublished data, 2001).

4.2. Comparisons With Previous Summit CO2 Emission Rate Estimates

[23] Our results for Kīlauea's summit CO2 emission rate are significantly larger and more consistent than previous results [Greenland et al., 1985] of 3600 t d−1 and 1600 t d−1 obtained respectively on 9 December 1983 and 13 February 1984. While it is possible that the summit CO2 emission rate was much lower and more variable in late 1983 to early 1984, it is more likely that differences in technique account for the dissimilar results.

[24] The previous results are based on fixed-wing aircraft profiling [Harris et al., 1981] downwind of the summit with a MIRAN CO2 analyzer several hundred meters above ground, thus possibly above the core of the plume. The earlier studies report no high CO2 levels rising several hundred ppm above ambient, as we observed. Because the gas emissions of the summit caldera come from thousands of distributed small sources, our CO2 measurements, made much closer to these sources than the airborne measurements, should be less affected by dispersion, and more fully capture the total summit CO2 output.

[25] Finally, there are important differences in the CO2 analyzers. The MIRAN lacks the sensitivity of the LI-COR CO2 analyzer. For example, the LI-COR's sample cell is small (12 cm3) compared to the MIRAN's (5600 cm3), and at the flow rates used in this study, its cell volume is replaced every 0.7 s, compared to 14 s for the MIRAN. At traverse vehicle speeds, the LI-COR cell samples CO2 over distances of ∼8 m, compared to ∼1 km for the MIRAN sample cell at typical aircraft speeds (70–80 m s−1). Gerlach et al. [1997] discuss additional advantages of the LI-COR over the MIRAN in volcanic plume applications.

4.3. Sources of Summit CO2 and SO2 Emissions

4.3.1. Surface Sources

[26] The CO2 and SO2 profiles (Figures 4 and 5) and emission rates (Table 1) reflect the output of countless surface sources within the summit caldera region. The summit gas emissions are not simply the product of a few powerful fumarole vents. Instead, innumerable weakly degassing cracks, fractures, low-temperature fumaroles, and patches of steaming ground distributed throughout the caldera region comprise the surface source of summit CO2 and SO2 emissions. Most important are the thousands of small fumaroles of the summit caldera region immediately upwind of the traverse, particularly near and within Halemaumau pit crater and to the east and southeast of Halemaumau, including especially those associated with the 1971 and 1974 fissures (Figure 2). Individual fumaroles are, with few exceptions, characterized by low temperatures (boiling point (∼95°C) or less) and weak output of mostly steam, air, and moderate amounts of CO2 and SO2 in proportions generally greater than ten to one (A. J. Sutton and T. Elias, U.S. Geological Survey, unpublished data, 2001). The highest temperature (typically 85–95°C) and most SO2-enriched fumaroles (some with CO2/SO2 < 10) occur around the rim of and inside Halemaumau [Shinohara, 1999; T. M. Gerlach, U.S. Geological Survey, unpublished data, 2001]. Fumaroles of the 1971 and 1974 fissures and other areas generally east and southeast of Halemaumau have lower temperatures (70–90°C), more air, and higher CO2/SO2. The remaining emissions (mostly on the caldera floor north of the above sites) arise from steaming areas of mainly water vapor and air at ∼50–65°C.

[27] The location of the main CO2 anomaly (Figures 4 and 7) indicates the greatest CO2 contributions arise from sources upwind of the a–d section of the traverse (Figure 2). Preliminary measurements by backpacking the CO2 analyzer define a roughly km2-region northeast of a–d, bounded on the west by Halemaumau (Figure 2), where ground level CO2 concentrations of 500–1500 ppm are common, rising to ∼3000 ppm in the vicinity of 1971 fissure and a nearby upwind steaming area. The more westerly locations of the higher SO2 concentrations (Figures 5 and 7) indicate greater SO2 contributions from in and around Halemaumau, generally consistent with COSPEC measurements [Elias et al., 1998; Sutton et al., 2001].

4.3.2. Summit Magma Reservoir Source

[28] Several investigators have interpreted the isotopic composition of carbon in CO2 from summit fumaroles (δ13CCO2 = −4 to −3‰) to result from degassing of mantle basalt [Gerlach and Thomas, 1986; Friedman et al., 1987; Gerlach and Taylor, 1990; Hilton et al., 1997]. A similar origin is indicated for SO2, the principal sulfur gas species of summit fumarole. Fumarole data generally give SO2/H2S values ≥10 [Shinohara, 1999; T. M. Gerlach, U.S. Geological Survey, unpublished data, 2001]. These observations favor a shallow magma-degassing origin for the SO2, since basaltic magmas degas SO2 as the principal sulfur gas at high temperatures and low pressures [Gerlach and Nordlie, 1975].

[29] Kauahikaua [1993] summarizes evidence for a high-elevation groundwater aquifer hosting a hydrothermal system to depths of ∼1–2 km in the summit region. A research borehole in the southern margin of the caldera (Figure 2) reaches the hydrothermal system at ∼0.5 km depth [Tilling and Jones, 1996]. Fournier [1987] reasons that a vapor-only zone above shallow magma penetrates through the hydrothermal system in places under the summit caldera where temperatures are too high and pore fluid pressures too low for liquid of any composition to exist. A dry, vapor transport zone is consistent with a basalt-degassing origin for the SO2 and the predominance of SO2 over H2S in fumarole emissions [Symonds et al., 2001].

[30] Thus summit CO2 and SO2 emissions point to a source of shallow degassing magma. Shallow magma is also the inferred source of a summit aseismic zone at 2–7 km depth [Koyanagi et al., 1976]. Additionally, several geophysical studies indicate shallow magma in a summit reservoir complex beneath the southern part of the caldera, connected at its base to a pipe-like conduit supplying magma from the mantle and at its margins to conduits that inject intrusions into the rift zones [Ryan et al., 1981; Klein et al., 1987; Thurber, 1987; Yang et al., 1992; Tilling and Dvorak, 1993; Dawson et al., 1999].

[31] A recent high-resolution (0.5-km) P and S wave velocity model for the summit caldera region defines a volume of high VP/VS interpreted as a magma reservoir 1–4 km beneath the southern part of the caldera, about 1–2 km south of Halemaumau (Figure 8) [Dawson et al., 1999]. A second high VP/VS anomaly, at similar depths under the Puhimau thermal area (Figure 2), is interpreted as an upper east rift magma reservoir. The model also shows an inverted V-shaped pattern of high VP/VS values passing through the central caldera region, inferred to be a connecting pathway transporting magma from the southern reservoir (where it is assumed magma arrives) to the east rift reservoir. However, the large emission of CO2 just upwind of the a–d section of traverse (Figure 2), suggests to us that primary magma enters the summit reservoir near the volume of high VP/VS immediately east-southeast of Halemaumau (Figure 8) and degasses its exsolved load of CO2 under the square kilometer region with high ground level CO2 noted above. We suggest that magma largely depleted of excess CO2 moves subsequently along the inverted V-shaped pathway toward the storage reservoirs to the south or in the upper ERZ.

Details are in the caption following the image
Map view of VP/VS model for 1.0 to 1.5 km depths based on high-resolution velocity models for the caldera region obtained by tomographic inversion of P and S wave arrival times (adapted from Dawson et al. [1999]). The high VP/VS volumes are attributed to the presence of magma in reservoirs beneath the southern part of the caldera and the upper ERZ, and along an interconnecting conduit [Dawson et al., 1999].

4.4. Total CO2 Emission Rate

[32] The total CO2 emission rate of Kīlauea is at least 8800 t d−1; this includes the 8500-t d−1 summit emission rate, the 240–300 t d−1 emission from Pu'u equation image cone [Gerlach et al., 1998], the chief degassing site of the current ERZ eruption (Figure 1), and the <25 t d−1 from the Puhimau thermal area of the upper ERZ (Figure 2). Additional sources are expected to be minor and unlikely to raise the total above 9000 t d−1. Thus summit degassing alone comprises between 94 and 97% of the total CO2 output.

4.5. CO2 Content of Primary Magma

4.5.1. Bulk CO2 Content

[33] Clague et al. [1995] derive an average primary magma for Kīlauea from a series of glasses with near-primary compositions. It consists of MgO-rich (16.5 wt %) crystal-free liquid at 1346°C with a density of 2.67 g cm−3, a viscosity of 0.4 Pa s, and dissolved volatile contents of 0.37 wt % H2O, 0.10 wt % S, and 0.01 wt % Cl. Its CO2 content cannot be derived from the glasses due to exsolution effects but can be constrained from magma supply and CO2 emission rate data.

[34] Cayol et al. [2000] determine a magma supply rate to the volcano of 0.18 km3 yr−1 for the period 1961–1991. This estimate almost doubles previous estimates of ∼0.1 km3 yr−1 [Swanson, 1972; Dvorak and Dzurisin, 1993; Denlinger, 1997] by including the volume of magma discharged to the rift zones without associated summit deformation. The 0.18 km3 yr−1 magma supply rate and the 8500 ± 300 t d−1 summit CO2 emission rate give a minimum CO2 content for primary magma of 0.65 ± 0.02 wt %. Including currently known rift zone CO2 emissions increases the amount to 0.67 wt %, but this estimate is undoubtedly also low. The upper limit for the current total CO2 emission (9000 t d−1) raises the result to 0.69 wt %; however, this estimate could still overlook CO2 in magma that is injected into the rift zones but never erupted. Studies of glasses from the submarine portion of the ERZ suggest that magma stored in rift zones contains ∼0.05 wt % CO2 [Dixon et al., 1991]. We take the sum of this amount and the 0.65 wt % implied by summit CO2 degassing alone as the best estimate of the CO2 content of primary magma, i.e., 0.70 wt % (estimated error ∼ ± 0.05 wt %). For comparison, the 0.1-km3 y−1 supply rate would give a primary magma CO2 content of 1.21 ± 0.10 wt %.

[35] The 0.70 wt % estimate for the CO2 content of Kīlauea primary magma is significantly higher than previous estimates of 0.32 wt %, derived from the earlier CO2 emission and magma supply rate estimates [Greenland et al., 1985], and 0.2 wt %, based on dissolved and vapor bubble CO2 in melt inclusions of the 1959 Kīlauea Iki eruption [Anderson and Brown, 1993]. It is not significantly different from the earlier estimate of 0.65 ± 0.06 wt % based on CO2-rich type I volcanic gases of summit lava lake eruptions [Gerlach and Graeber, 1985]. We note that the primary concentrations derived by Gerlach and Graeber for other volatiles (0.30 wt % H2O, 0.13 wt % S, and 0.01 wt % Cl) agree reasonably well with those derived by Clague et al. [1995] from the near primary-composition glasses (summarized above).

[36] Finally, we caution that because ascending magma may stall and solidify in the sub-volcano lithosphere, while its CO2-rich fluid phase continues to ascend, the 0.70 wt % CO2 of the primary magma supplied to Kīlauea may exceed the CO2 content of the mantle source magma.

4.5.2. Dissolved and Exsolved CO2

[37] We estimate a saturation pressure of ∼0.95 GPa for the 0.70 wt % CO2 in primary magma, assuming that CO2 solubility in picritic melts is similar to that of tholeiitic basalt melts [Pan et al., 1991]. The corresponding saturation depth is ∼30 km at inferred crust and upper mantle densities under the volcano [Ryan, 1987].

[38] The dissolved CO2 content of primary magma arriving at the 7-km-deep base of the summit reservoir complex (∼180 MPa [Ryan, 1987]) would be ∼0.09 wt %, based on solubility data for CO2–H2O in tholeiitic basalt liquid [Dixon et al., 1995]. Thus, of the bulk 0.70 wt % primary CO2, approximately 0.61 wt % would enter the reservoir as exsolved gas. Further exsolution of dissolved CO2 would occur at shallower depths. In the shallowest portion of the reservoir at 1–2 km (∼23–45 MPa [Ryan, 1987]), evolved basaltic melt with ∼0.7 wt % H2O [Wallace and Anderson, 1998] would dissolve only ∼0.01–0.02 wt % CO2 [Dixon et al., 1995]. Thus some 87–99% of the CO2 supplied in primary magma is present as exsolved gas at summit reservoir conditions and potentially degassed from the reservoir. Solubility data are therefore consistent with the high proportion (94–97%) of summit CO2 emissions in the total CO2 output of Kīlauea, the low CO2 emission rates of ERZ sources, and the average 0.05 wt % CO2 inferred for rift zone magma after summit degassing.

4.6 Buoyancy Effects of CO2 on Primary Magma

[39] The MRK equation of state [Holloway, 1977] gives a density of 0.44 g cm−3 for CO2 vapor in primary magma at 7 km depth (180 MPa) and 1346°C. Primary magma entering the base of the reservoir containing 0.70 wt % CO2 (0.61 wt % exsolved, 0.09 wt % dissolved) would therefore contain 3.6 vol % CO2 vapor and have a bulk density of 2.59 g cm−3. Minor exsolved H2O would further reduce its density, probably to <2.57 g cm−3. Liquid densities in the summit reservoir decrease from 2.67 g cm−3 for primary liquid containing 16.5 wt % MgO to 2.61 g cm−3 for evolved liquid with 7.1 wt % MgO [Clague et al., 1995]. However, owing to crystal content, calculations based on data for ERZ submarine lavas indicate significantly higher bulk densities (>2.67 g cm−3) for reservoir magma containing evolved liquids [Clague et al., 1995]. Thus primary magma entering the base of the reservoir is clearly buoyant.

4.7. Mixing of Primary and Reservoir Magmas

[40] Primary magma as buoyant as indicated above would tend to ascend through the reservoir and erupt [Anderson, 1995]. However, eruptions of high-MgO liquids are rare in the summit area. This could mean primary magma contains appreciably less CO2 than derived above. Alternatively, buoyant primary magma might mix with reservoir magma, thus preventing ascent and eruption. Finally, bubble escape may cause primary magma to lose its buoyancy but only in the presence of very crystal-poor reservoir magma containing mostly low-density evolved liquid.

[41] To address these issues, we have used Huppert et al.'s [1986] investigation of magma chamber replenishment by “light inputs”. Light primary magma injected into the summit reservoir from an orifice with a diameter much smaller than the reservoir height can be considered to issue from a point source just below the vent opening. The Reynolds numbers for the “internal” fluid (i.e., primary magma), Rep, and the “external” fluid (i.e., reservoir magma), Rer, may then be expressed as follows:
urn:x-wiley:01480227:media:jgrb12998:jgrb12998-math-0003
urn:x-wiley:01480227:media:jgrb12998:jgrb12998-math-0004
Rep and Rer are defined by four parameters: the primary magma supply rate, Q; the kinematic viscosities of the primary and reservoir magmas, vp and vr, respectively; and the reduced gravity, equation image = g(drdp)dr−1, where g is the acceleration of gravity and dp and dr are the respective densities of primary and reservoir magmas (dp < dr). Huppert et al. [1986] investigated laboratory replenishment models covering a wide range of Rep and Rer values and mapped the fields of fluid motion on a Reynolds number regime diagram.

[42] We calculated Rep and Rer values for primary magma replenishing a summit reservoir containing a range of plausible liquids and magmas (liquid + crystals). The calculations assume the supply rate (Q = 0.18 km3 yr−1), density (dp = 2.59 g cm−3), viscosity (0.4 Pa s), and kinematic viscosity (vp = 1.54 cm2 s−1) of primary magma entering the reservoir with a bulk CO2 content of 0.70 wt %. To represent reservoir values for dr and vr, we used the liquid and liquid + crystal densities and viscosities calculated by Clague et al. [1995, Table 1] for glass rinds on 24 lavas from the submarine portion of the ERZ. The results (Figure 9a) plot deeply within the turbulent field of Huppert et al.'s [1986] Reynolds number regime diagram. We highlight the results for the six magmas with the lowest crystal content (6.1–9.4%, vesicle-free basis from point count data) and the five liquids with the highest MgO content (6.6–7.0 wt %), as these are considered less affected by olivine entrainment and crystallization during rift zone transport [Clague et al., 1995].

Details are in the caption following the image
(a) Reynolds numbers for primary magma (Rep) and reservoir magma (Rer) during reservoir replenishment by primary magma entering the base of the summit reservoir at 7 km. Squares and circles represent 24 plausible reservoir liquids and magmas (liquid + crystals), respectively (some points overlap), based on glass rinds of Puna Ridge lavas [Clague et al., 1995], as discussed in the text. Five liquids containing >6.5 wt % MgO (solid squares) and six magmas containing <10% crystals (solid circles) are highlighted, as discussed in the text. Heavy line is boundary of turbulent flow field [from Huppert et al., 1986, Figure 7]. Lines emanating from a reservoir magma with 6.9% crystals show changes for an order of magnitude increase in the viscosity of primary magma (dashed), an order of magnitude decrease in primary magma supply rate (dotted), an order of magnitude increase in the viscosity of reservoir magma (dash-dotted), and an order of magnitude decrease in the density difference between reservoir magma and primary magma (solid). (b) Same magmas as in Figure 9a but assuming total loss of CO2-rich bubbles from the primary magma, as discussed in the text (some points overlap).

[43] The sensitivity of Rep and Rer values to turbulence-reducing order of magnitude decreases in momentum terms Q and drdp, and order of magnitude increases in viscosity terms vp and vr, is shown in Figure 9a for a model reservoir magma with 6.9% crystals. In all cases, results remain within the turbulent field. Even if bubble escape removes all vapor from primary magma, vapor-free primary magma remains buoyant (due to the crystal content of the model reservoir magmas), and replenishment continues to be turbulent (Figure 9b). CO2 vapor becomes indispensable to the buoyancy of primary magma and to turbulent replenishment only if the crystal content of reservoir magma falls below 5%. Such low crystal contents may not be feasible in the lower part of the reservoir because of olivine cumulate [Clague and Denlinger, 1994].

[44] Thus the light input replenishment model robustly predicts turbulent mixing above the reservoir entry. Primary magma will rise from the entry as a turbulent plume, entrain reservoir magma, and mix with it. CO2-rich bubbles in primary magma will be advected along with the turbulent plume and also mixed with reservoir magma [Phillips and Woods, 2001], although turbulence-induced bubble coalescence may accelerate their escape. The vertical extent of turbulent mixing, however, depends on reservoir geometry and other factors presently unknown.

[45] We conclude that turbulent mixing with reservoir magma generally prevents summit eruptions of buoyant primary magma. Nevertheless, rare events may permit summit eruptions of near primary magma, e.g., if the magma supply bypasses most of the reservoir [Helz, 1987], or if greatly elevated supply rates override reservoir capacity [Eaton et al., 1987].

4.8. High CO2/SO2 of Summit Emissions

[46] Turbulent mixing promotes the extraction of S and H2O from reservoir magma through repeated exposure to the purging effects of CO2-rich vapor. However, only a portion of the H2O and S in reservoir melt partitions into the CO2-rich bubbles because of solubility controls [Wallace and Carmichael, 1992; Dixon et al., 1995; Wallace and Anderson, 1998, 2000]. The partial degassing of S (as SO2) and almost complete degassing of CO2 from summit reservoir magma contributes to the high CO2/SO2 of summit emissions (Table 1). The low CO2/SO2 values of gas emissions in the current ERZ eruption, involving magma previously degassed in the summit reservoir, reflect the fractionation occurring during summit degassing; for example, the CO2/SO2 for 204 gas samples since the start of the eruption in 1983 average only ∼0.2 [Gerlach et al., 1998]. The high CO2 and low SO2 emission rates of the summit (Table 1) are similarly complemented by the low CO2 (≤300 t d−1) and high SO2 (∼2,000 t d−1) emission rates of the ERZ eruption [Elias et al., 1998; Gerlach et al., 1998; Sutton et al., 2001].

[47] Scrubbing of degassed SO2 by contact with liquid water [Symonds et al., 2001] is undoubtedly a further factor, in addition to melt solubility controls, contributing to the high summit CO2/SO2. High sulfate waters (2000–4300 ppm) in a research borehole in the southern part of the caldera (Figure 2) indicate hydrothermal scrubbing of SO2 (S. Hurwitz, U.S. Geological Survey, unpublished data, 2001). In addition, the scrubbing of SO2 observed in condensed steam of Halemaumau fumaroles [Gerlach et al., 1991] is expected to be a widespread phenomenon, since almost all summit fumaroles are at or below the local boiling point (∼95°C). Water droplets in the air could also scrub SO2, but we doubt this process significantly affected our data, which were obtained in dry weather a short distance downwind of emission sources.

4.9. Temporal Trends in Summit SO2 Emission Rates and CO2/SO2

[48] The decrease in summit SO2 emission rates from the 1995 to 1999 experiments (Table 1) is part of a long-term trend since 1987 (Figure 10). The decline coincides with a steady deflation of the summit region and a drop in the frequency of shallow, short-period earthquakes under the caldera, both consistent with decreasing magma pressure at shallow depths in the summit reservoir [Sutton et al., 2001]. Thus the decrease in summit SO2 emissions appears to correlate with a decrease in the amount of shallow magma.

Details are in the caption following the image
Summit SO2 emission rates from 1987 to 2000 based on data from Elias et al. [1998]. Heavy line is linear regression (r2 = 0.74). Dashed vertical lines indicate dates of CO2 emission-rate experiments of the present study (Table 1).

[49] A shrinking quantity of shallow magma could be an important factor affecting the availability of S for degassing and the extent of SO2 scrubbing. Dissolved S decreases in silicate melts with cooling, although this effect is generally superseded in crystallizing basalt by Fe-enrichment, which increases S solubility [Wallace and Carmichael, 1992; Wallace and Anderson, 2000]. However, Kilauean basalt does not show Fe enrichment, because crystallizing olivine contains approximately the same weight percent FeO as the associated melt [Wallace and Anderson, 2000]. The solubility of S therefore decreases in cooling reservoir magma. Data for S in ERZ submarine glasses are consistent with lowering of S by degassing in the summit reservoir [Wallace and Carmichael, 1992], suggesting that degassing removes S as its solubility falls in cooling reservoir magma; the glasses do not show evidence of S removal by saturation with sulfide phases [Johnson et al., 1994]. A declining amount of shallower (i.e., cooler) reservoir magma would therefore diminish the amount of S available for degassing and cause a falling summit SO2 emission rate and rising summit CO2/SO2, consistent with observed trends (Table 1 and Figure 6). A declining amount of shallow magma would also permit deeper incursion of water and enhanced SO2 scrubbing, thus further reducing SO2 emissions and increasing CO2/SO2.

4.10. Spatial Variations in Summit CO2 and SO2

[50] Melt solubility controls and scrubbing can also account for segregation effects and independent variations of CO2 and SO2 observed in the summit emissions (Figures 4, 5, and 7). For example, the absence of the highest SO2 concentrations along the a–d section of the traverse showing the highest CO2 may reflect solubility controls resulting in the degassing of a high proportion of the CO2 and a low proportion of the S in deeper (hotter) magma near the entry to the summit reservoir east of Halemaumau. On the other hand, the observations are also consistent with a hypothesis of extensive SO2 scrubbing of magmatic gas released beneath this area. The tendency for higher SO2 and lower CO2 along the western portion of the traverse downwind of Halemaumau may signify the degassing of shallower (cooler) magma in route to the southern reservoir (Figure 8) after prior (deeper) degassing of CO2 at the entry east of Halemaumau; however, it could also mean that SO2 scrubbing is less extensive to the west. The generally lower SO2 to the east may indicate deeper transport of magma to the east rift reservoir (Figure 8), or more extensive SO2 scrubbing to the east.

[51] The end result is that summit CO2 and SO2 emissions are poorly correlated in space. Figuratively, it is as if there is a summit plume for CO2 and another for SO2, each with its own characteristic configuration and distribution and accentuating different processes, a matter of importance in monitoring summit gases. The “CO2 plume” is clearly the target of choice for following the degassing of deeper magma, especially primary magma supplied to the summit reservoir. The “SO2 plume” is preferred for assessing the degassing of shallow magma (including erupting summit lava) and interactions of magmatic gas with water.

4.11. Comments on Estimating Summit CO2/SO2

[52] It is apparent from this study that a summit CO2/SO2 value based on a few summit fumarole samples, or a few spot samples along the traverse through the plume, could give misleading results for the emission rate of CO2 obtained by equation (1). Fumarole sampling is especially apt to give low results, since the usual practice is to sample fumaroles with the highest temperatures. At Kīlauea, the gases of such fumaroles are likely to be less affected by scrubbing and to come from relatively shallow (more degassed) magma; as a result, they tend to have rather low CO2/SO2 values. For example, the hotter fumaroles in the summit area occur inside and around Halemaumau pit crater and tend to have CO2/SO2 values ≤10 [Shinohara, 1999]; indeed, the hottest fumarole (280–300°C) is inside the crater and has had CO2/SO2 values of 3–4 over several years since its discovery in 1990 (T. M. Gerlach, U.S. Geological Survey, unpublished data, 2001). These fumaroles are not representative of global summit emissions, and a CO2/SO2 based on them would give a vastly underestimated CO2 emission rate. On the other hand, a CO2/SO2 based on fumarole sampling of degassing sources upwind of the ad portion of the traverse, producing the observed higher CO2 and lower SO2 concentrations, would probably be too high and overestimate CO2 emissions. The challenge facing a fumarole sampling approach to constraining the summit CO2 emission rate by equation (1) is to devise a strategy that gives reliable CO2/SO2 estimates from a practical number of samples.

4.12. Summit CO2 Emission as a Proxy for Primary Magma Supply

[53] The rate of supply of mantle magma is a fundamental force driving much of Kīlauea's activity [Dvorak and Dzurisin, 1993]. A practical method for frequent monitoring of the supply rate is therefore desirable. The observation that most of the CO2 in primary magma degases in the summit region, is material evidence of magma supply and confirms the long-held hypothesis that the magma supplied to Kīlauea enters the edifice through the summit reservoir [Tilling and Dvorak, 1993]. Since the summit CO2 emission rate is linked to magma supply, it should be monitored as a proxy for the magma supply rate. Furthermore, there is reason to expect the summit CO2 emission rate responds rapidly to changes in the magma supply rate. On 1 February 1996, for example, a fivefold increase in the CO2/SO2 of a monitored summit fumarole occurred within fifteen minutes of the onset of a seismic crisis event attributed to a sharp increase in magma supply [Thornber et al., 1996].

5. Conclusions

  1. The summit CO2 emission rate is several times larger than earlier estimates and nearly constant at 8500 ± 300 t d−1.
  2. Summit CO2 emissions comprise almost all of Kīlauea's total CO2 output, which is probably not in excess of 9,000 t d−1.
  3. Primary magma supplied to Kīlauea enters the edifice through the summit reservoir via an entry beneath a square kilometer area immediately east of Halemaumau.
  4. The bulk CO2 content of the primary magma supplied to Kīlauea is 0.70 ± 0.05 wt %; this is probably an upper limit for the CO2 content of magma generated in the mantle source.
  5. Most of the CO2 in primary magma is degassed in the summit region, making the summit CO2 emission rate a convenient proxy for Kīlauea's magma supply rate.
  6. The CO2 of primary magma is present mainly as vapor at reservoir depths and makes the primary magma strongly buoyant.
  7. Turbulent mixing with resident reservoir magma prevents frequent eruption of the buoyant primary magma in the summit region.
  8. Volcanic CO2/SO2 of the plume along the summit COSPEC traverse is representative of the global CO2/SO2 of the summit emissions.
  9. Solubility controls on CO2 and S in basaltic melt and scrubbing of degassed SO2 cause the high CO2/SO2 of summit emissions and the poorly correlated spatial distributions of summit CO2 and SO2.
  10. Summit CO2 emissions are generally more informative about deeper magma degassing, while summit SO2 emissions are more revealing of shallower magma degassing and interactions of magmatic gases with water.

Acknowledgments

[54] We have benefited from and are grateful for discussions, insights and unpublished results graciously shared with us by Bernard Chouet, Dave Clague, Phil Dawson, Roger Denlinger, James Dieterich, Dan Dzurisin, Herbert Huppert, Shaul Hurwitz, Dick Iverson, Jim Kauahikaua, Larry Mastin, Arnold Okamura, Paul Okubo, Don Swanson, Tom Sisson, Carl Thornber, Joe Walder, and Pete Zemek. We thank John Eichelberger, Toby Fischer, Fraser Goff, Larry Mastin, and Don Swanson for reviews of earlier versions of the paper, and Bobbie Myers for assistance with the figures. Funding from the U.S. Geological Survey Volcano Hazards and Earth Surface Dynamics Programs is acknowledged.