The standard and modified ECC En-Sci ozonesondes were used for the O3 and SO2 sonde measurements in this study. The basic functioning of the ECC ozonesonde is described in Komhyr (1969) and Morris et al. (2010). The ECC sensor is composed of platinum cathode and anode electrodes, each in its own cell, immersed in a diluted and saturated solution of potassium iodide (KI), respectively. The cells are connected by an ion bridge allowing for the transfer of electrical charges while maintaining the separation of the solutions (Eqs. 1 and 2). When the cells are charged with the solution, a transient potential difference is generated that is dissipated through the redistribution of charge across the ion bridge. The following equilibria are established from these reactions: 13I-⇌I3-+2e-(anode),2I2⇌2e-→2I-(cathode). Sampled air is pumped into the cathode cell, and the presence of O3 initiates a reaction (Eq. 3) that causes an imbalance in favor of [I2] in the cathode solution. 32KI+O3+H2O→2KOH+I2+O2 To rebalance the electrochemical potential of the cell, the iodine and iodide redox reactions in Eqs. (4) and (5) result in a flow of electrons from the anode to the cathode via the ion bridge. This cell current, measured by an external ammeter, is proportional to the O3 concentration. 43I-→I3-+2e-(anode)5I2+2e-→2I-(cathode) When SO2 is present in the sample air, an additional reaction (Eq. 6) occurs in the cathode cell of the ECC, supplying the two electrons needed to rebalance the cathode cell after the O3 reaction (Eq. 3) (Komhyr, 1969; Morris et al., 2010). 6SO2+2H2O→SO42-+4H++2e- Thus, each SO2 molecule in the sampled air has the effect of canceling the measurement of one O3 molecule. In effect, the standard ECC ozonesonde reports [O3]-[SO2] for its measurement. In most places and at most times, [SO2]≪[O3], so there is not a significant impact on the O3 measurements, but in places downwind of SO2 sources (e.g., coal-burning power plants or volcanoes), the O3 measurement will be negatively impacted.

Several SO2 and O3 instruments were used for validation of the SO2 sonde during laboratory and field testing. A calibration system was used to produce controlled concentrations of SO2 and O3. The calibration system relied on the operation of flow controllers or restrictors, an SO2 ultra-high-purity (UHP) gas cylinder (4.87 ppm; Scott-Marrin, Inc., Riverside, CA) and/or a UV photometric O3 calibrator (49C PS; Thermo Fisher Scientific, Franklin, MA), and zero air to produce desired pre-set concentrations of SO2 and/or O3. The zero-air setup used for the field and laboratory testing was achieved using a dry zero-air UHP gas cylinder or else generated by scrubbing ambient air through activated charcoal and Purafil SP (Purafil, Inc., Doraville, GA) canisters. The Thermo 43i-TL SO2 analyzer (LLOD: 60–90 pptv at 5 min averaging) and the 49i O3 analyzer (LLOD: 1.5 ppbv at 5 min averaging) were also used during laboratory testing, while a Thermo 43c-TL SO2 analyzer was used during field testing in Hawai'i. These instruments were set to report 10 s average measurements.

The single-sonde SO2 system included three major modifications to the En-Sci ECC ozonesonde: (1) the application of a positively biased current to the cathode cell, (2) the addition of an O3 removal filter, and (3) a sample dryer (Fig. S1). The first version of the SO2 system (SO2 sonde v1.0) included the first two modifications: the bias current and an O3 removal filter. The bias current sets the upper limit of detection (ULOD) for the SO2 sonde and is set prior to measurement. The O3 removal filter is placed in line with the inlet, allowing O3-free air to be sampled in the SO2 sonde. In the ECC, O3 produces a positive response signal while SO2 produces a negative signal when sufficient O3 is present (i.e., positive signal). With these two modifications, SO2 can be measured directly as the reduction of the cell current from the pre-set biased current (Flynn and Morris, 2021). Unlike the dual-sonde system, this approach allows for direct SO2 measurements rather than an inference by subtraction of signals from two separate instruments. A sample dryer was added to the SO2 sonde in the second version (v1.1) to combat humidity issues discovered after initial field tests. The addition of the dryer corrected the highly varying instrument sensitivity observed in the field. All components of the SO2 sonde fit within a standard ozonesonde foam box (approximately 8 in. ×8 in. ×10 in.) except for the inlet filter. The free-release balloon payload's total mass is approximately 1 kg. The patent publication and Fig. S1 provide a detailed description and schematic of the SO2 sonde (Flynn and Morris, 2021).

The bias current is supplied by inserting into the cathode cell an additional platinum electrode powered by a 9 V battery (Fig. S1) (Flynn and Morris, 2021). To maintain consistent power, the circuit uses a 5 V regulator. Varying the resistance allows for a range of bias currents to be introduced. The current version of the SO2 sonde uses a fixed resistor which requires a priori knowledge of the desired SO2 concentration range. The desired resistor is installed in series with the battery and the electrode. An earlier laboratory test compared the SO2 sonde measurements (initially configured without an O3 removal filter) to those made by a 43i-TL SO2 analyzer (Fig. 1, Table 1). O3 and SO2 gases were introduced using the laboratory calibration setup and a manifold to allow the sonde and the Thermo trace gas instruments to sample the same air. Results in Fig. 1 show 60 s averaged data. The test included (A) input of O3 without an added bias current; (B) the same input of O3 with the addition of a bias current (equivalent to approximately 90 ppbv of O3); and the addition of SO2 to the O3 with the enhanced bias signal where the SO2 concentration was either (C) smaller or (D and E) larger than the O3 concentration. During (A), measurements made by O3 and SO2 sondes compare well to measurements made by the Thermo instruments (Fig. 1, Table 1). The test included (E) the response of the SO2 sonde to a reduction of the O3 concentration, resulting in an equivalent decrease in signal, followed by (G–I) a reduction in the SO2 concentration, resulting in an equivalent increase in signal. At (F), the SO2 concentration exceeded the bias current (90 ppbv), producing a signal equivalent to 2.9±0.1 ppbv. The sonde successfully measured SO2 both with and without O3 with approximately 97 % efficiency.

Examination of the SO2 sonde data showed that noise was proportional to the measured signal, with 1-σ noise at approximately 0.2 %–0.3 % of the measured signal. Because increases in the SO2 concentrations result in decreases in the signal (i.e., lower cell currents), the magnitude of the applied bias current determines the saturation point (i.e., ULOD) of the SO2 sonde; saturation occurs when the measured cell current drops to zero. Applying a higher bias current increases the ULOD but also increases noise and the LLOD. The reported LLODs of bias currents are calculated as 3σ relative to the baseline signal when sampling zero air. During laboratory testing, the LLOD (3σ) was calculated for a range of applied bias currents (0.25 to 10.0 µA). The LLOD for the varying bias current of 0.25 to 10.0 µA ranged from approximately 0.002 to 0.084 µA, respectively. Results of calculated LLOD of a 0.25 µA bias current at varying replicated altitudes are included in Table S1 in the Supplement. At the surface, the LLOD of 20 s averaged measurements is 0.17 ppbv. The final version of the SO2 sonde (v1.1) requires the bias current to be selected prior to measurement. If the bias current is set too low, a measurement of larger-than-expected SO2 concentrations can saturate the sensor while a bias current that is set too high will have higher LLOD due to the increase in noise. The applied magnitude of the bias current can be best determined based on known SO2 sources including volcanic emissions and urban and/or industrial emissions.

Since the ECC responds to both O3 and SO2, an O3 removal filter was developed to remove interference from O3 in the sample. This proprietary O3 removal filter is placed upstream of the sonde inlet (Flynn and Morris, 2021). During laboratory testing, the O3 removal filter was exposed to a continual concentration of 487±3 ppbv of O3 and a varying concentration of SO2 ranging from 0 to 111±1 ppbv (Fig. 2). The O3 was effectively and consistently removed from the sampled air by the O3 removal filter as SO2 was diluted. The testing included measurements with (gray background) and without (white background) the O3 removal filter. The SO2 and O3 concentrations measured by the Thermo 43i-TL and 49i instruments, respectively, and changes in SO2 dilution levels are also indicated in Fig. 2. The O3 removal filter destroyed the O3 at all SO2 dilution levels to below the detection limit of the O3 instrument. By comparing the Thermo 43i-TL SO2 analyzer measurements with and without the O3 removal filter, SO2 passed through the filter with 88 % efficiency (Fig. 3a). The transmission efficiency was calculated by taking the ratio of SO2 measured by the sonde to that measured by the analyzer. The SO2 transmission efficiency increased to 97 % when testing the O3 removal filter with the dry zero-air UHP gas cylinder (Fig. 3b) instead of the zero-air generator that processes ambient laboratory air (Fig. 3a). Additional testing of the O3 removal filter demonstrated that the filter removed approximately 1 ppm of O3 at sea level with >99.9 % in O3 removal efficiency, concentrations below the detection limit of the Thermo 49i O3 monitor.

The SO2 sonde v1.0 had highly varying sensitivities during the initial field tests. The instrument sensitivity was determined by regression analysis of the sonde's cell current to the SO2 concentration measured by an SO2 analyzer. The variability in the sensitivities was hypothesized to be due to differing levels of humidity during each SO2 sonde launch. SO2 is soluble in water and through multiphase reactions can be oxidized to sulfuric acid in the atmosphere in the presence of water vapor (e.g., precipitation, clouds, fog) (Carmichael and Peters, 1979; Zhang et al., 2013; Terraglio and Manganelli, 1967). Factors including liquid water content, aerosol composition, aerosol loading, and pH of the water are important in determining the adsorption and oxidation rates of SO2 (Liu et al., 2021). When air with elevated humidity is flowing through a filter, SO2 gas is likely adsorbing on the filter, causing lower SO2 transmission efficiency due to the potential uptake of SO2 in water on the filter. Several laboratory tests confirmed the need to remove water from the sample upstream of the O3 removal filter to improve the measurement of SO2. A desiccant membrane dryer (Perma Pure LLC, Lakewood, NJ) composed of a Nafion™ tube in silica gel desiccant was placed in-line upstream of the O3 removal filter. This sample dryer is lightweight, relatively inexpensive, and does not require power.

Laboratory tests included exposing the SO2 sonde, with and without a sample dryer, to controlled levels of humidity and SO2. Without removing water vapor, the SO2 transmission efficiency decreases as humidity increases, particularly above 50 % RH (Fig. 6). As the O3 removal filter is humidified, the SO2 transmission efficiency decreases. With the sample dryer in place, each of the laboratory SO2 transmission efficiency (17–18 and 21 May 2018) tests varied by an average of <1 % across a range of 0 %–85 % RH (Fig. 6).

The dryer's useful lifetime was determined by continuously exposing it to high-humidity (>95 % RH at approximately 23 ∘C) sample stream. The downstream RH climbed from 5 % to 16 % after 2.3 h and to 25 % after 6.3 h. At these downstream RH levels, the SO2 transmission efficiency remained above 95 %. A typical SO2 sonde's measurement time per flight, including pre-flight calibration, is approximately 3 h. The dryer's useful lifetime is likely much longer than required for a balloon flight since exposure to 95 % RH conditions for several hours is highly unusual outside of hurricanes and tropical systems. SO2 sonde and Thermo 43c-TL measurements were strongly correlated (r2=0.99) during a multipoint calibration conducted using the O3 removal filter and the dryer under relatively high humidity levels. During that calibration, the SO2 sonde's sensitivity was 45.43±0.17 ppbvµA-1. By comparison, the average sensitivity during the initial Hawaii deployment was 84.6±31.7 ppbvµA-1 across 10 sondes. The sample dryer, therefore, improved both the sensitivity and stability of the measurements observed. The addition of the sample dryer is necessary for providing accurate ambient SO2 measurements.

The first deployment of the SO2 sonde v1.0 was during NASA's HyspIRI HyTES Hawaii Campaign (H3C) from 3–10 February 2018, near Kīlauea volcano. The instrument was tested in flights on free-release balloons and a tethered balloon system (TBS) and at ground level with measurements in Hawai'i Volcanoes National Park (HVNP) downwind of Kīlauea's summit crater, Halema'uma'u. During the ground-level testing, an SO2 sonde and a Thermo 43c-TL SO2 analyzer's sample inlet were mounted on the top of a van for co-located sampling.

Figure 4a depicts the measurements taken during the first encounter with an SO2 plume while driving through the HVNP on 3 February 2018. The strongly correlated SO2 sonde and Thermo 43c-TL measurements (r2=0.99) reached upward of ∼940 ppbv. The SO2 sonde had a sensitivity of 118.4±0.4 ppbvµA-1, determined by regression analysis of the sonde's cell current with the Thermo 43c-TL concentrations (Fig. 4a). The SO2 sonde sensitivity varied significantly during the field deployment. During surface measurements on 10 February 2018, earlier zero-air calibrations measured a sensitivity of 86.5±1.5 ppbvµA-1, while measurements during an SO2 plume event, with peak concentrations of up to 400 ppbv, found the SO2 sonde's sensitivity was 73.9±0.6 ppbvµA-1 (Fig. 4b). Although the SO2 sonde sensitivity varied significantly in 10 subsequent calibrations (84.6±31.7 ppbvµA-1), the measurements remained strongly correlated (range: r2= 0.94–0.99). The variability in the sensitivity in the field was due to changes in the ambient RH impacting the SO2 transmission efficiency of the O3 removal filter. This hypothesis was confirmed by laboratory RH testing and discussed in Sect. 3.3 and 3.4.

On 23 March 2018, a traditional SO2 dual-sonde payload (Morris et al., 2010) and the SO2 sonde v1.0 were launched using a free-release balloon flight from the Universidad de Costa Rica's campus in San Jose (approximately 31 km downwind of Turrialba volcano). This flight provided the first direct in situ comparison of the two SO2 sonde methods. Figure 5 shows the response of the SO2 sonde v1.0 and the calculated SO2 dual-sonde profile. The dual-sonde SO2 method can only report concentrations of SO2 up to a maximum of the concentration of O3 present. Furthermore, because the SO2 concentration is determined by subtracting the signals from two instruments, its uncertainty is higher than the uncertainty of a measurement from a single instrument. When [SO2]>[O3], the dual sonde's unfiltered ozonesonde signal goes to zero, as happened for the Turrialba sonde launch between 3 and 5 km (Fig. 5). The SO2 saturates the cathode solution in the unfiltered sonde, not recovering until enough ambient O3 has been processed to rebalance the cell, resulting in a distorted profile (Fig. 5). For this flight, the SO2 sonde was configured to its maximum range (ULOD of approximately 450 ppbv at standard pressure) and was able to capture both the small plume below 2 km above mean sea level (a.m.s.l.) (approximately 18 ppbv) and the primary plume between 3 and 4 km a.m.s.l. (approximately 230 ppbv). The SO2 sonde v1.0 was able to capture the full shape of the profile, including the peak values and structure of the plume. The SO2 sonde v1.0 reports the top of the plume around 4 km a.m.s.l., whereas the dual sonde remains saturated until closer to 5 km a.m.s.l. Thus, the dual-sonde SO2 profiles, when saturated by high concentrations of SO2, erroneously appear to have a greater vertical extent. Further, the SO2 sonde v1.0 showed no interference from O3 at altitudes from the surface to 24.4 km a.m.s.l., with O3 concentrations in the stratospheric O3 layer reaching >4 ppmv (not shown), demonstrating the effectiveness of the O3 filter. The SO2 VCD was 8.3 DU (Dobson units, 1 DU =2.69×1016 molec. cm-2) for the SO2 sonde but was only 3.4 DU for the dual-sonde measurement. Thus, once saturated, the dual-sonde method is likely to underestimate the SO2 VCD.

The SO2 sonde v1.1 was tested in Ft. Mackay (57.1206∘ N, 111.4241∘ W), Alberta, in the Athabasca oil sands from 10–16 June 2018 (Fig. S2c). This field project, conducted in conjunction with Environment Canada and York University, evaluated SO2 emissions from industrial activities in and near the oil sands region using a combination of TBS and ground-based measurements. The SO2 sonde v1.1 was flown on the York TBS payload, recording measurements from the ground to 300 m above ground level (a.g.l.; 650 m a.m.s.l.). This deployment provided a dilute anthropogenic plume to test the SO2 sonde in a high-sensitivity, low-range configuration. The average sensitivity of the SO2 sonde v1.1 during the project was 51±1.2 ppbvµA-1. The SO2 sonde was configured to sample in a range from ∼ 0.5–25 ppbv of SO2. The TBS SO2 sonde's vertical profiles were averaged into 10 m altitude bins that measured SO2 concentration ranges that are more representative of anthropogenically impacted SO2 rather than large volcanic plumes (Fig. 7). This field deployment also demonstrated the performance of the sonde at sub-ppbv levels of ambient SO2.

In response to the larger eruption that started in May 2018, the SO2 sonde v1.1 was deployed to Hawai'i for the NASA-funded Big Island SO2 Survey (BISOS). The SO2 sonde launches occurred from Kahuku Ranch (19.0549∘ N, 155.6934∘ W) and Na'alehu Elementary School (19.0610∘ N, 155.5788∘ W) approximately 90 km downwind of Kīlauea's LERZ (Fig. S2d). The site's distance from the source allowed the plume to disperse and dilute compared with measurements at the vent. An SO2 plume was detected during seven of the nine free-release balloon launches during the June 2018 BISOS campaign. The 10 SO2 sonde v1.1 calibrations performed during BISOS had an SO2 sensitivity of 47.0±5.8 ppbvµA-1 and were similar to the laboratory results (45.43±0.17 ppbvµA-1).

With the anticipated levels of SO2, the sondes were configured to sample in the range of 10–450 ppbv of SO2. Figure 8 shows four distinctive SO2 profiles, and Table 2 includes the VCDs for each flight. No plumes above 5 km a.m.s.l. were detected. All but one of the observed SO2 plumes were below the capping inversion of the planetary boundary layer (PBL). On 22 June (Fig. 8a), the ascent profile shows SO2 below 3 km a.m.s.l. peaking at nearly 100 ppbv and additional features between 3 and 4 km a.m.s.l. peaking at 20–35 ppbv (Tang et al., 2020). The latter peaks were correlated with higher RH, perhaps the result of steam from a vent or the ocean entry points having broken through the inversion. The early afternoon 28 June profile (Fig. 8b) shows the highest concentration (325 ppbv) for a resolved SO2 plume during the BISOS campaign. Typical for the trade winds, NOAA HYSPLIT trajectories (Stein et al., 2015) showed the winds were out of the NE, consistent with the plume's transport from vents in the LERZ or the lava ocean entry points. Although the descent profile from a 29 June early afternoon launch lost the signal at 0.58 km a.m.s.l., Fig. 8c shows an SO2 plume over the ocean with a peak concentration of 188 ppbv at 0.74 km a.m.s.l. HYSPLIT trajectories again showed the winds were out of the NE. Lastly, the SO2 plume detected during the ascent of the 30 June launch (Fig. 8d) exceeded the ULOD between 1 and 3 km a.m.s.l. for the SO2 sonde configuration used. The distorted SO2 enhancement extending above the PBL as determined by the temperature inversion is most likely an artifact of the saturated sonde, similar to what was seen in the dual-sonde profile from Costa Rica (Fig. 5). As the RH remains low above the PBL, it is most likely that the SO2 is contained entirely within the PBL.