Spatial and Temporal Variations in SO2 and PM2.5 Levels Around Kīlauea Volcano, Hawai'i During 2007–2018

Among the hazards posed by volcanoes are the emissions of gases and particles that can affect air quality and damage agriculture and infrastructure. A recent intense episode of volcanic degassing associated with severe impacts on air quality accompanied the 2018 lower East Rift Zone (LERZ) eruption of Kīlauea volcano, Hawai'i. This resulted in a major increase in gas emission rates with respect to usual emission values for this volcano, along with a shift in the source of the dominant plume to a populated area on the lower flank of the volcano. This led to reduced air quality in downwind communities. We analyse open-access data from the permanent air quality monitoring networks operated by the Hawai'i Department of Health (HDOH) and National Park Service (NPS), and report on measurements of atmospheric sulfur dioxide (SO2) between 2007 and 2018 and PM2.5 (aerosol particulate matter with diameter <2.5 μm) between 2010 and 2018. Additional air quality data were collected through a community-operated network of low-cost PM2.5 sensors during the 2018 LERZ eruption. From 2007 to 2018 the two most significant escalations in Kīlauea's volcanic emissions were: the summit eruption that began in 2008 (Kīlauea emissions averaged 5–6 kt/day SO2 from 2008 until summit activity decreased in May 2018) and the LERZ eruption in 2018 when SO2 emission rates reached a monthly average of 200 kt/day during June. In this paper we focus on characterizing the airborne pollutants arising from the 2018 LERZ eruption and the spatial distribution and severity of volcanic air pollution events across the Island of Hawai'i. The LERZ eruption caused the most frequent and severe exceedances of the Environmental Protection Agency (EPA) PM2.5 air quality threshold (35 μg/m3 as a daily average) in Hawai'i in the period 2010–2018. In Kona, for example, the maximum 24-h-mean mass concentration of PM2.5 was recorded as 59 μg/m3 on the twenty-ninth of May 2018, which was one of eight recorded exceedances of the EPA air quality threshold during the 2018 LERZ eruption, where there had been no exceedances in the previous 8 years as measured by the HDOH and NPS networks. SO2 air pollution during the LERZ eruption was most severe in communities in the south and west of the island, as measured by selected HDOH and NPS stations in this study, with a maximum 24-h-mean mass concentration of 728 μg/m3 recorded in Ocean View (100 km west of the LERZ emission source) in May 2018. Data from the low-cost sensor network correlated well with data from the HDOH PM2.5 instruments, confirming that these low-cost sensors provide a robust means to augment reference-grade instrument networks.

Among the hazards posed by volcanoes are the emissions of gases and particles that can affect air quality and damage agriculture and infrastructure. A recent intense episode of volcanic degassing associated with severe impacts on air quality accompanied the 2018 lower East Rift Zone (LERZ) eruption of Kīlauea volcano, Hawai'i. This resulted in a major increase in gas emission rates with respect to usual emission values for this volcano, along with a shift in the source of the dominant plume to a populated area on the lower flank of the volcano. This led to reduced air quality in downwind communities. We analyse open-access data from the permanent air quality monitoring networks operated by the Hawai'i Department of Health (HDOH) and National Park Service (NPS), and report on measurements of atmospheric sulfur dioxide (SO 2 ) between 2007 and 2018 and PM 2.5 (aerosol particulate matter with diameter <2.5 µm) between 2010 and 2018. Additional air quality data were collected through a community-operated network of low-cost PM 2.5 sensors during the 2018 LERZ eruption. From 2007 to 2018 the two most significant escalations in Kīlauea's volcanic emissions were: the summit eruption that began in 2008 (Kīlauea emissions averaged 5-6 kt/day SO 2 from 2008 until summit activity decreased in May 2018) and the LERZ eruption in 2018 when SO 2 emission rates reached a monthly average of 200 kt/day during June. In this paper we focus on characterizing the airborne pollutants arising from the 2018 LERZ eruption and the spatial distribution and severity of volcanic air pollution events across the Island of Hawai'i. The LERZ eruption caused the most frequent and severe exceedances of the Environmental Protection Agency (EPA) PM 2.5 air quality threshold (35 µg/m 3 as a daily average) in Hawai'i in the period 2010-2018. In Kona, for example, the maximum 24-h-mean mass concentration of PM 2.5 was recorded as 59 µg/m 3 on the twenty-ninth of May 2018, which was one of eight recorded exceedances of the EPA air quality threshold during the 2018 LERZ eruption, where there had been no exceedances in the previous 8 years as measured by the HDOH and NPS networks. SO 2 air pollution during the LERZ eruption was most severe in communities in the south and west of the island, as measured by selected HDOH and NPS stations in this study, with a maximum 24-h-mean mass concentration of 728 µg/m 3 recorded in Ocean View (100 km west of the LERZ emission source) in May 2018. Data from the low-cost sensor network correlated well with data from the HDOH PM 2.5 instruments, confirming that these low-cost sensors provide a robust means to augment reference-grade instrument networks. Keywords: Kīlauea,volcano,PM 2.5 , SO 2 , emissions, air quality, Hawai'i

INTRODUCTION
Volcanic clouds are complex, evolving mixtures of volcanic and atmospheric gases, primary and secondary aerosol particles, ash and dust (Oppenheimer and McGonigle, 2004;Pfeffer et al., 2006b;von Glasow et al., 2009;Oppenheimer et al., 2010;Langmann, 2014). As well as the potential for global climatic consequences from explosive or large-scale volcanic emissions, low altitude volcanic clouds can have important impacts on air quality, human and animal health, and the environment on the local to regional scale (Hansell and Oppenheimer, 2004;Barsotti et al., 2010;Mather, 2015;Schmidt et al., 2015;Andronico and Del Carlo, 2016;Tam et al., 2016;Ilyinskaya et al., 2017).
Gaseous sulfur dioxide (SO 2 ) is usually highly concentrated in volcanic emissions compared to the background atmosphere and is often the focus of gas emission monitoring due to the relative ease of its measurement and its important environmental and air quality consequences (Cadle et al., 1971;Lambert et al., 1988;Loughlin et al., 2012;Schmidt et al., 2015). Population sub-groups including children, asthmatics and cardiac-or respiratory-compromised individuals are particularly vulnerable to exposure to SO 2 (ATSDR, 1998;CRI, 2004). For example, exposure to mass concentrations of 1,310 µg/m 3 SO 2 for 3 min can induce respiratory attacks in asthmatic individuals (Balmes et al., 1987;ATSDR, 1998). In 2010 the U.S. Environmental Protection Agency (EPA) set the National Ambient Air Quality Standard (NAAQS) for SO 2 mass concentration exposure limits at 195 µg/m 3 as an hourly average (EPA, 2010). Persistent volcanic SO 2 emissions on the Island of Hawai'i led the state of Hawai'i being designated as unclassifiable for the EPA 2010 NAAQS, and as such Hawai'i uses the pre-2010 EPA SO 2 exposure limit of 366 µg/m 3 as 24-h average (EPA, 2013). The European Commission (EC) air quality standards recommend a SO 2 mass concentration threshold of 350 µg/m 3 for a 3-h average, and 125 µg/m 3 as a daily average (EC, 2018).
Particulate matter (PM) in volcanic clouds is also significant in the context of environment and health. The chemical composition of volcanogenic PM 2.5 (PM with diameter <2.5 µm) is highly heterogeneous. Typical chemical species include sulfates (primary emissions or formed via oxidation of sulfur gases) (Cadle et al., 1971;Stockwell and Calvert, 1983;Allen et al., 2002;Mather et al., 2003;Langmann, 2014) and halides, with an array of metals and metalloids including environmentally-harmful species such as lead and cadmium (Longo, 2013;Langmann, 2014). PM 2.5 is a well-established indicator for air quality, since it commonly includes particulates derived from transport and industrial sources, fine wind-blown mineral dust, ambient matter, and volcanic material (Lim et al., 2012;Tam et al., 2016;Holgate, 2017;Butwin et al., 2019). It has been estimated that the health burden due to exposure to ambient PM 2.5 globally amounts to more than three million premature deaths each year (Lim et al., 2012), and is especially linked to increases in death from cardiovascular and respiratory diseases in vulnerable individuals (Holgate, 2017). EPA NAAQS thresholds recommend a PM 2.5 mass concentration exposure limit of 35 µg/m 3 as a daily average (EPA, 2013). This is higher than the 24-h mean exposure guideline of 25 µg/m 3 established by the World Health Organization (WHO, 2005).

KĪLAUEA ERUPTIVE ACTIVITY, 2007-2018
Kīlauea volcano on the Island of Hawai'i consists of a summit caldera at 1,200 m a.s.l. and rift zones to the south-west and east. From 1983 until 2008, activity at Kīlauea was concentrated on the middle East Rift Zone (ERZ), primarily near the Pu'u 'Ō'ō vent (Elias and Sutton, 2007;Poland et al., 2008) ( Figure 1A). Between 2002 and 2006, average SO 2 flux from the ERZ was 1.7 ± 0.7 kt/day, while emissions from the summit were low at 0.1 kt/day (Elias and Sutton, 2007). From November 2007 to March 2008, SO 2 emissions at the summit increased to levels 10 times the long-term background (Wooten et al., 2009) (Figure 2A). On the twelfth of March 2008, a new vent opened within the Halema'uma'u summit crater, leading to sporadic explosive eruptions and increased degassing of SO 2 .
Kīlauea's SO 2 emissions peaked in the summer of 2008, when a total emission rate of up to 20 kt/day was measured by satellite sensors (Beirle et al., 2014). At this time, emissions from both the ERZ and the summit were significant, with the two sources contributing variable amounts to the total degassing rate (Elias and Sutton, 2012). For the period 2009-2017, the dynamic activity at Kīlauea was reflected in variable emissions, with a longterm average of 5-6 kt/day based on satellite and ground-based measurements (Eguchi et al., 2011;Elias and Sutton, 2012;Carn et al., 2016;Elias et al., 2018) (Figure 2A). Lava was first observed in the Halema'uma'u summit crater in September 2008, with a permanent lava lake visible from February 2010 until May 2018 (Patrick et al., 2013;Neal et al., 2019). The 2018 Kīlauea eruption in the lower East Rift Zone (LERZ) began following the collapse of the Pu'u 'Ō'ō vent on the thirtieth of April (Neal et al. 2019;HVO, 2018). Twenty-four fissures opened over a distance of 6.8 km in the vicinity of Leilani Estates ( Figure 1A). During the first week of the LERZ eruption, spattering activity at individual fissures was typically short-lived (minutes to hours in duration) and lava was viscous with spatter deposition within tens of meters of individual fissures. On the eighteenth of May, the eruptive style evolved to less viscous lava and resulted in fast-moving lava flows which reached the ocean 2 days later (HVO, 2018). By the end of May 2018, activity had become focused at Fissure 8, and this remained the dominant fissure for the remainder of the LERZ eruption (Neal et al., 2019). Lava fountains from Fissure 8 reached heights of 80 m, and lava effusion rates ranged from 50 to 200 m 3 /s (Neal et al., 2019). Lava from Fissure 8 flowed in a semi-stable channel to the ocean and eventually covered an area of land 35.5 km 2 in size (Neal et al. 2019;HVO, 2018). This eruption was the largest along Kīlauea's LERZ in the last two centuries and had farreaching impacts around the Island of Hawai'i. With the collapse of the Pu'u 'Ō'ō vent and draining of Kīlauea's summit magma reservoir, the dominant source of volcanic SO 2 became the LERZ eruptive vents. SO 2 emissions reached an average of 200 kt/day in June 2018 , severely impacting island-wide air quality. The eruption declined rapidly at the end of July and lava effusion ceased on the fourth of August 2018 (Neal et al., 2019).

Downwind Processes and Impacts
Since initiation of intermittent fountaining activity at Kīlauea in 1983, SO 2 emissions have been a health concern among downwind communities on the Island of Hawai'i. As emissions from Kīlauea are dispersed downwind, communities are exposed to volcanic smog, locally known as vog, predominantly composed of SO 2 and fine particles of sulfuric acid aerosol (Longo, 2009;Longo et al., 2010;Halliday et al., 2015;Tam et al., 2016;Elias and Sutton, 2017). Prevailing trade winds from the north-east, particularly during the period from April to October, carry Kīlauea's emissions over the communities to the south and west (Longo et al., 2005(Longo et al., , 2008Michaud et al., 2007;Tam et al., 2016;Elias and Sutton, 2017) (Figure 1A). The Island of Hawai'i has relatively low population density in the south ( Figure 1B), with 4,400 residents in Ocean View and~1,300 residents in Pāhala. Trade winds from the north-east are influenced by the high topography of Mauna Loa and Mauna Kea, generating more localized air movement in the lee of the island on the west coast (Michaud et al., 2007) (Figure 1A). This wind shadow allows a potentially longer residence time for air pollutants (volcanogenic or otherwise) along the densely-populated western coastline ( Figure 1B). During the winter months (November to March), the trade winds weaken and southerly and westerly winds may distribute vog toward the densely-populated eastern coastline of the island (Wyrtki and Meyers, 1976;Mannino et al., 1996;Michaud et al., 2004) (Figure 1B).
Numerous studies have investigated the impact of Kīlauea's SO 2 emissions on the health of island residents, even at the relatively low levels of degassing prior to the emergence of the lava lake in 2008. Mannino et al. (1996) reviewed the frequency of visits to emergency departments and hospitalizations for respiratory issues during periods of continuous and discontinuous SO 2 emissions throughout the 1980s. Communities on the western side of the island frequently exposed to vog were found to have higher rates of hospitalizations for chronic obstructive pulmonary disease than the east-coast city of Hilo, which is rarely exposed to vog. Periods of weakened north-easterly trade winds coincided with a 15 % increase in emergency department visits for asthma in Hilo (Mannino et al., 1996). In 2004, the health of Hawai'i residents in vog-exposed and -unexposed communities was surveyed (Longo et al., 2008;Longo, 2009). Those in exposed communities were found to have a significantly increased prevalence of cough, phlegm, sinus congestion, rhinorrhoea, wheezing, eye irritation, and bronchitis than those in unexposed communities. Following the increase in SO 2 flux from Kīlauea's summit in 2008, Longo (2013) reassessed the vog-related health impacts on the residents of the Island of Hawai'i. The magnitude of cardio-respiratory issues in vog-exposed communities was found to have increased as compared to 2004 (Longo et al., 2008), with the risk factor of acute cardiac events in persons aged >50 years increased by 12 % (Longo, 2013). A study by Tam et al. (2016) investigated the effects of vog on the respiratory health of school children across Hawai'i, finding that chronic exposure to vog was associated with increased prevalence of cough and potential decrease in lung function, but not with the prevalence of asthma or bronchitis. The unprecedented emission rates of the 2018 LERZ eruption has presented a continued motivation to further characterize the severity and distribution of volcanic air pollution during elevated volcanic activity.
The 2018 Kīlauea LERZ eruption provided a unique opportunity to study the impacts arising from a large low-altitude cloud rich in SO 2 in a populated and well-instrumented part of the world. Here we used open-access data from a network of reference-grade instruments in populated areas around the Island of Hawai'i to determine the severity of SO 2 and PM 2.5 impact on air quality from the LERZ eruption. We compare air quality during the LERZ eruption to that from a lower emission period, defined to be January 2007 to December 2017 for SO 2 and January 2010 to December 2017 for PM 2.5 . We examine a network of community-operated PM 2.5 instruments and compare their measurements to those from reference-grade instruments. From these data, we demonstrate that SO 2 and PM 2.5 mass concentrations during the 2018 LERZ eruption in selected communities around the island were of a higher magnitude than during volcanic activity from Kīlauea during 2007 to 2017.

Continuous SO 2 and PM 2.5 Air Quality Monitoring
Hawai'i Department of Health (HDOH) ambient air quality stations continuously monitor SO 2 and PM 2.5 mass concentrations around the Island of Hawai'i ( Figure 1A). SO 2 is measured by a pulsed fluorescence spectroscopy analyzer (model 43i manufactured by Thermo Scientific) that is designated by the EPA for measurements in the range of 0-1,000 ppb, with a lower detectable SO 2 limit of 0.5 ppb and a precision of 1 ppb (Thermo Scientific 2010; EPA, 2016). FEM-designated instruments (Forum for Environmental Measurements), such as this SO 2 analyzer, promote consistency in measurements and laboratory conditions ensuring that the instruments are of reference-grade quality (EPA, 2016). Following EPA regulations, the analyzers undergo in-situ calibration checks weekly, with a multi-point calibration run every 6 months. PM 2.5 mass concentrations are measured by a Beta Attenuation Monitor (BAM) with a 60-min sampling rate. The BAM instrument (model BAM-1020, manufactured by Met One Instruments) is FEM-designated for measurements of particles in the size range of 0-1,000 µm (with PM 2.5 being a small subset of the measured particle size range), with a resolution of 1 µm particle diameter and a lower detection limit of 4 µg/m 3 (MetOne 2008; EPA, 2016). The BAM instruments undergo calibration and auditing every 6 months. Permanent HDOH ambient air quality stations are kept in air-conditioned enclosures to maintain long-term stability. Data from the air quality stations are streamed in near-real time to the HDOH website, which is open-access and publicly-available (HDOH, 2019).
Data from the HDOH and NPS station networks used in this study have been categorized into regions for the purpose of data analysis. The western region includes Kona HDOH station on the west coast of the Island of Hawai'i. The southern region includes Volcano Observatory NPS station and Pāhala and Ocean View HDOH stations. The eastern region includes the Hilo and Mountain View HDOH stations. We compare the HDOH and NPS SO 2 timeseries data against the European Commission 24-h air quality threshold (125 µg/m 3 ) and the Hawai'i 24-h ambient air quality standard (366 µg/m 3 ). The HDOH PM 2.5 timeseries data is compared to the World Health Organization 24-h exposure limit (25 µg/m 3 ) and the Environmental Protection Agency NAAQS 24-h limit (35 µg/m 3 ).

Community-Operated PM 2.5 Instruments
PurpleAir (Utah, USA) instruments are low-cost (approximately $250 per unit) particulate sensors that are purchased and operated by individuals and provide open access data online (PurpleAir, 2019). PurpleAir instruments contain Plantower PMS5003 nephelometer sensors, which use a small fan to draw air through a laser-induced light, and a photo-diode detector converts 90°-scattered light into a voltage pulse (Kelly et al., 2017). PMS5003 sensors have a 10 s response time and detect particles between 0.3 and 10 µm in diameter (Kelly et al., 2017;Sayahi et al., 2019). The maximum consistency error of the sensors is stated by the manufacturer to be ± 10 µg/m 3 between 0 and 100 µg/m 3 (Plantower, 2016). The instruments are factory calibrated prior to sale (PurpleAir, 2019). PM mass concentration measurements are calculated using an atmospheric calibration factor, details of which are not provided by the manufacturer (Kelly et al., 2017;Zheng et al., 2018;Sayahi et al., 2019). PurpleAir instruments contain two Plantower PMS5003 sensors mounted in one housing, allowing self-consistency checks to alert when significant differences are reported between the internal sensors. An ESP8266 wireless chip is included in PurpleAir instruments to upload data via WiFi to an online cloud database, which is open-access (Sayahi et al., 2019).
Prior to the 2018 LERZ eruption, six community-operated PurpleAir instruments were located on the Island of Hawai'i, three of which were in the Kona area in the western region. Installation dates ranged from August 2017 to February 2018. Following the onset of the LERZ eruption and island-wide increase in atmospheric pollutants, the number of PurpleAir instruments increased, with a further twenty instruments installed across the western region of the island over the course of May to July 2018.
We carried out fieldwork during the 2018 LERZ eruption and co-located PurpleAir instruments with HDOH ambient air quality stations at Kona and Ocean View ( Figure 1A). Two PurpleAir instruments were installed at the Kona station, one from the fifth of June to September 2018 and a second from the nineteenth of July to the third of August, and one PurpleAir instrument was installed at the Ocean View station from the nineteenth of July to the third of August 2018. The PurpleAir instruments were installed close to the inlet for the BAM instruments, on the roofs of the air quality shelters and away from obstructions. Other than the coordinates of the instrument, no environmental information is provided in association with the PurpleAir instruments. For this reason, measurement uncertainty arising from factors such as installation in proximity to potential contamination sources in community-operated PurpleAir instruments should be considered significantly higher than those placed in optimal conditions near HDOH stations. Community-operated PurpleAir instruments were selected for analysis along the region highlighted in blue in Figure 1A. Kona experienced elevated SO 2 mass concentrations during the 2018 LERZ eruption, with a peak 24-h-mean mass concentration of 136 µg/m 3 measured at the Kona station. During the 2018 LERZ, SO 2 mass concentrations in Kona did not exceed the Hawai'i SO 2 threshold of 366 µg/m 3 ( Figure 2B), but did exceed the 125 µg/m 3 EC threshold on one occasion ( Table 1).

Southern Region: Volcano Observatory, Pāhala and Ocean View
Following initiation of Kīlauea's summit activity in 2008, the HDOH-operated Pāhala station and the NPS-operated Volcano Observatory station routinely recorded high concentrations of SO 2 (Figure 2C), with SO 2 mass concentrations exceeding the Hawai'i 366 µg/m 3 24-h-mean threshold 0.7 % of the time at both Volcano Observatory (twenty-eight exceedance events) and Pāhala (thirty exceedance events) ( Table 1). The maximum 24h-mean mass concentration recorded by the NPS station at Volcano Observatory during 2007-2017 was 1,068 µg/m 3 , and by the HDOH station in Pāhala was 776 µg/m 3 . The Ocean View HDOH station is located farther to the south-west than Pāhala and Volcano Observatory, at a greater distance from Kīlauea's summit and the ERZ (Figure 1A). During the period 2010-2017, SO 2 mass concentrations recorded at Ocean View exceeded the Hawai'i 24-h-mean threshold 0.1 % of the time (three exceedance events). The maximum 24-h-mean mass concentration recorded at Ocean View was 403 µg/m 3 , significantly lower than measured at Volcano Observatory and Pāhala (Table 1).
In comparison, during the 3-months of the 2018 LERZ eruption, SO 2 mass concentrations exceeded the Hawai'i 366 µg/m 3 threshold 2.1 % of the time at Volcano Observatory (two exceedance events), 5.3 % of the time at Pāhala (five exceedance events) and 4.2 % at Ocean View (four exceedance events). Maximum 24-h-mean mass concentrations at Volcano Observatory and Pāhala were lower than those measured during 2008-2017 (450 and 555 µg/m 3 , respectively), but the relative frequency of exceedance events increased (Table 1). During the 2018 LERZ eruption, the Ocean View station recorded a peak 24-h-mean mass concentration of 728 µg/m 3 , almost double the previous peak measurement of 403 µg/m 3 recorded at that station in January 2016.

Eastern Region: Hilo
During the period 2007-2017, SO 2 mass concentrations in Hilo followed a distinct seasonality (Figure 2D). Peak SO 2 mass concentrations were commonly observed in Hilo in November to March (average monthly concentration of 11 µg/m 3 ) with low mass concentrations in the intervening months of April to October (average monthly mass concentration of 3.6 µg/m 3 , as calculated from 2007 to 2017) ( Figure 2D). Exceedances of the EC 24-h-mean threshold (125 µg/m 3 ) rarely occurred outside this peak season. In the period 2007-2017 there were twenty-one exceedance events during November to March, compared with just three between April to October. The seasonal variations in SO 2 mass concentrations observed in Hilo can be explained by the strong prevalence of northeasterly trade winds during April to October (Wyrtki and Meyers, 1976) (Figure 1A). During these months, emissions from Kīlauea's summit and the ERZ were dispersed predominantly to the south-west of the Island. The trade winds weaken between November to March, allowing SO 2 to be dispersed to the east of the island (Mannino et al., 1996;Michaud et al., 2004;Elias and Sutton, 2017).
During the third of May to fourth of August LERZ eruption, the HDOH station in Hilo recorded a maximum 24-h-mean mass concentration of 144 µg/m 3 , which was recorded on the twenty-first of June (Table 1), and was the only exceedance of the EC 24-h-mean threshold during the 3-month eruption. The SO 2 mass concentrations measured during the LERZ eruption were lower than the average measurements during the 2007-2017 period. SO 2 mass concentrations in Hilo are usually low during the months when the LERZ eruption occurred. Nevertheless, during the 2018 LERZ eruption, SO 2 mass concentrations in Hilo rose significantly above the average for the season (average 24-h-mean SO 2 mass concentration during 2018 LERZ eruption was 6.9 µg/m 3 , in comparison to the usual seasonal average of 3.6 µg/m 3 ).

Western Region: Kona
In the period 2010-2017, PM 2.5 recorded by HDOH Kona station never exceeded the EPA 24-h-mean threshold of 35 µg/m 3 ( Figure 3A). The WHO 24-h-mean guideline of 25 µg/m 3 was exceeded 1.2 % of the time at the Kona site (33 exceedance events). The maximum PM 2.5 24-h-mean mass concentrations was 33 µg/m 3 , recorded in April 2016 (Table 1).
During the 2018 LERZ eruption, PM 2.5 exceeded the 24h-mean 35 µg/m 3 EPA threshold 8.4 % of the time at the Kona station (eight exceedance events). The lower guideline of 25 µg/m 3 established by the WHO was exceeded 34.7 % of the time at the Kona site (33 exceedance events). The maximum All units in µg/m 3 . Exceedances of air quality standards are indicated for 24 h means, and calculated as percentage of total measurement duration. Distances between emission points and measurements sites are straight line distances; the emissions will not always follow the most direct route from near-to far-field. Note: no data available for PM 2. recorded in Ocean View during the 2018 LERZ eruption was 56 µg/m 3 , somewhat higher than the peak 24-h-mean mass concentration recorded at Ocean View in 2016 (42 µg/m 3 ). During the 2018 LERZ eruption, PM 2.5 mass concentrations in Mountain View were higher than the average mass concentrations for 2010-2017 (7 and 4 µg/m 3 , respectively) ( Table 1). However, during the 2018 LERZ eruption, PM 2.5 mass concentrations did not exceed either the EPA threshold of 35 µg/m 3 or the WHO guideline of 25 µg/m 3 . Kīlauea volcano can be identified (Figure 2A), which were registered by the air quality monitoring instruments around the Island of Hawai'i ( Table 1, section 4.1 and 4.2).

Dispersal of Volcanic Emissions
High mass concentrations of SO 2 and PM 2.5 generally occurred in the southern and western parts of the Island of Hawai'i. Prevailing trade winds from the north-east dispersed SO 2 emissions from Kīlauea volcano toward communities in the south and west of the island, as reported in previous studies (Longo et al., 2005(Longo et al., , 2008Michaud et al., 2007;Tam et al., 2016). During the 2018 LERZ eruption, the EPA 24-h-mean threshold for PM 2.5 (35 µg/m 3 ) and the Hawai'i threshold for SO 2 (366 µg/m 3 ) were exceeded in the south and west of the island (Figure 4). HDOH stations in the south of the island, 35-100 km away from Fissure 8, recorded 24-h events where SO 2 exceeded Hawai'i thresholds (note that there were no HDOH or NPS permanent monitoring stations for SO 2 in proximal location to the 2018 LERZ eruption site), but PM 2.5 EPA exceedance events only occurred at HDOH stations 100 km or further away from Fissure 8 (Figure 4). This spatial variance between distribution of PM 2.5 and SO 2 is well-documented and thought to reflect the timescale of oxidation of sulfur dioxide gas into sulfate aerosol during dispersion (Cadle et al., 1971;Stockwell and Calvert, 1983;Porter et al., 2002;Ilyinskaya et al., 2017).
Volcanic emissions at source commonly consist of a mixture of silicate ash particles, various gases and non-silicate aerosol (Oppenheimer and McGonigle, 2004;von Glasow et al., 2009;Langmann, 2014). The lifetime of SO 2 in the lower troposphere is generally considered to be on the order of 1-3 days to a week (Allen et al., 2002;Rotstayn and Lohmann, 2002;Pfeffer et al., 2006a;Pattantyus et al., 2018), the rate of conversion depending on relative humidity and temperature, the availability of oxidants, and interaction with cloud or fog (Saxena and Seigneur, 1987;Oppenheimer et al., 1998). However, the SO 2 oxidation pathways in a volcanic cloud are not necessarily the same as under background conditions (Galeazzo et al., 2018). Through a variety of reaction pathways (including oxidation with the hydroxyl radical, OH, and with hydrogen peroxide, H 2 O 2 , and O 3 ), SO 2 in volcanic clouds is gradually converted to sulfate aerosol (Stockwell and Calvert, 1983;Allen et al., 2002), which is a dominant component of volcanic PM 2.5 (Tam et al., 2016;Pattantyus et al., 2018). The conversion rate of SO 2 to sulfate aerosol is important for estimating the potential hazard of volcanic PM 2.5 to human health and the downwind environments (Kroll et al., 2015).
HDOH stations measure the ambient air, which contains SO 2 and PM 2.5 derived from anthropogenic sources as well as natural non-volcanic and volcanic sources. In order to determine the influence of the volcanic eruption on the measured SO 2 and aerosol abundances, it is first necessary to calculate the volcanic component of the HDOH measurements. During the 2018 LERZ eruption, Fissure 8 was the dominant source of volcanic SO 2 emissions on the island, and following the decline of the eruption, the SO 2 and PM 2.5 mass concentrations decreased to below pre-LERZ eruption levels at all HDOH stations analyzed in this study (Figures 2, 3). The mass of pollutants recorded at HDOH stations during this post-LERZ eruption period (mid-August 2018 to the first of February 2019) are therefore used to define the background abundances arising from all other nonvolcanic sources. During this time there was some SO 2 emitted from Kīlauea's summit but at the lowest rate measured in decades at 0.1 kt/year . The volcanic component of the HDOH measurements was calculated by subtracting the average PM 2.5 and SO 2 mass concentration following the end of the 2018 LERZ eruption for each station from the mass concentrations measured during the LERZ eruption, to estimate the volcanogenic component. The sulfate aerosol component within the volcanic PM 2.5 mass concentration was then estimated to be in the range 77-92 %, following the methods of Mather et al. (2012) and Kroll et al. (2015), of PM 2.5 composition from Kīlauea.
Estimating the conversion rate of SO 2 to sulfate from SO 2 and SO 4 2datasets is not straightforward because several processes can occur simultaneously, including SO 2 oxidation to sulfate, dispersion-dilution and deposition of SO 2 and/or SO 4 2to the surface.
Here, a first-order decay constant for SO 2 is estimated by the relationship between volcanic components of SO 2 and SO 4 2-, as follows; where S gas is the sulfur component of the volcanic SO 2 mass concentration (µg/m 3 ), S total is the sum of sulfur components of the volcanic SO 4 2and SO 2 mass concentrations (µg/m 3 ), t is the age of the volcanic cloud (seconds) and −k is a first-order decay constant.
FIGURE 5 | Relationship between the age of the volcanic cloud and the volcanic sulfur components measured at Pāhala, Ocean View and Kona HDOH ambient air quality stations. S gas is the sulfur component of the volcanic SO 2 mass concentration, S total is the sum of sulfur components of the volcanic SO 4 2and SO 2 mass concentrations. Age is derived from HYSPLIT back-trajectory simulations (Ilyinskaya et al., in preparation). Y axis error bars indicate range of SO 4 2component in volcanic PM 2.5 composition from Kīlauea. X axis error bars indicate one standard deviation of HYSPLIT back-trajectory results specific to Kona, Ocean View and Pāhala. Correlation trendline based on the data points, with a first-order decay constant of 3.8 x 10 -6 s -1 with a 95 % confidence interval of ± 1.26 x 10 -6 s -1 .
The age of the volcanic cloud is here considered to be the time between emission of the cloud at the LERZ eruption source point and subsequent measurement at the HDOH station. From back-trajectory HYSPLIT simulations run between the HDOH ambient air quality stations and Fissure 8, the average age of the emissions and dispersal distance was calculated for dates between the eighteenth of July to the second of August 2018 (Ilyinskaya et al., in preparation). An estimate of the average first-order decay constant from our data-set is indicated in Figure 5. Although there is considerable scatter in the data, a broad trend of decreasing S fraction in the gas phase is apparent and a linear fit allows us to estimate a first-order rate constant of 3.8 × 10 -6 s -1 with a 95 % confidence interval of ± 1.26 × 10 -6 s -1 . This first-order decay constant for SO 2 relative to total sulfur can represent an estimate of the average rate of SO 2 oxidation to sulfate only if negligible sulfur deposition has occurred. Nevertheless, our value is similar to the SO 2 oxidation rate calculated by Kroll et al. (2015) from direct measurements of sulfur in gas and particle phase in Kīlauea's emission cloud from the summit to Pāhala. Kroll et al. (2015) identified a diurnal cycle in measured sulfate as a fraction of total sulfur, from which they calculated a noontime instantaneous SO 2 oxidation rate of 2.4 × 10 -6 s -1 .

Reliability Assessment of Community-Operated PM 2.5 Instruments
A subset of the community-operated PurpleAir instruments on the Island of Hawai'i were selected for intercomparison with the established institutional data-sets across the western region of the island. The low-cost and portable nature of the PurpleAir instruments ( Figure 6F) facilitated installation of a monitoring network across the western region (Figure 6A), with a high spatial resolution of measurements in comparison to the locations of HDOH ambient air quality sites ( Figure 6A). This can be advantageous to capture the effects of local topographic and meteorological factors, which may influence dispersion of and deposition from volcanic plumes. Mass concentrations from PurpleAir instruments and HDOH PM 2.5 instruments during the course of the 2018 LERZ eruption are presented in Figures 6B-E.
Small differences were found between individual PurpleAir instruments in the same location. Two PurpleAir instruments ( Figure 6A, PurpleAir references 10 and 11) were co-located at the Kona HDOH station, and the PurpleAir instruments ran together for 16 days. During this time, the maximum absolute differences in 24-h average measurements between the PurpleAir instruments was 2.3 µg/m 3 ; which was 6 % of the total measured concentration. Correlation between the two PurpleAir instruments was very strong (Pearson's r = 0.99). Similar results were found by Malings et al. (2019) with co-location of nine PurpleAir instruments at a site in Pennsylvania for a period of 66 days (Pearson's r > 0.9). The high correlation between co-located individual PurpleAir instruments indicates high standardization. With this being the case, relative PM 2.5 mass concentrations measured by PurpleAir instruments over a wider geographical area should be comparably reliable.
Three PurpleAir instruments were co-located with HDOH PM 2.5 instruments (BAM) to determine the accuracy of PurpleAir instrument PM 2.5 measurements in relation to reference-grade instruments ( Figure 6A, PurpleAir references 10, 11, and 20). The co-located PurpleAir measurements correlated well with the BAM measurements, with Pearson's r values of 0.99, 0.97, and 0.91 (Figures 7A,B). However, PurpleAir instruments did record higher mass concentrations of PM 2.5 in comparison to the BAM analyzers (Figures 7A,B). PurpleAir reference 10 (co-located with Kona HDOH station) (Figure 6A) recorded the greatest measurement off-set, with up to 40 % higher mass concentrations of PM 2.5 relative to the Kona BAM analyzer measurements (Figure 7A). This trend was also found when the community-operated PurpleAir instruments across the western region of the island were compared to the BAM at Kona HDOH station (Figure 7C), with strong correlation between PurpleAir instruments and BAM (Pearson's r = 0.92) but an average 30 % higher mass concentrations measured by PurpleAir instruments within 10 km of the Kona BAM instrument.

Previous testing of PurpleAir instruments in Pennsylvania and
California has yielded similar findings (AQ-SPEC, 2017;Malings et al., 2019), with over-estimation of PM 2.5 mass concentrations measured by PurpleAir instruments relative to BAM referencegrade analyzers.
Some discrepancy between PurpleAir and BAM measurements may be expected, as it is well-known that low-cost instruments measuring PM 2.5 with light scattering methods have not historically agreed with measurements obtained from reference grade instruments with different operating principles (Watson et al., 1998;Wilson et al., 2002;Chow et al., 2008;Burkart et al., 2010). In comparison with sensors of other operating principles, light-scattering optical particle sensors have been shown to suffer effects of relative humidity Crilley et al., 2018), since the operating principle relies on indirect measurement of particle size and shape based on scattered light, and an assumed particle shape and refractive index. Conversely, BAM instruments measure direct changes in aerosol mass concentrations based on the loss of electrons on a filter which the aerosol has been deposited on (Watson et al., 1998;Manikonda et al., 2016), and so are not influenced by the hygroscopic growth of individual particles. Zheng et al. (2018) analyzed the performance of Plantower PMS3003 (an earlier version of the Plantower PMS5003 housed in PurpleAir instruments used in this study) against a reference grade scattered light spectrometer (with good correlation of R 2 = 0.8) and a BAM instrument (with lower correlation of R 2 = 0.5). They concluded that a likely explanation contributing to the discrepancy is the potential for hygroscopic growth of aerosol particles due to ambient humidity, which alters the light-scattering properties of the aerosol and therefore the measurement made by the optical sensor (Watson et al., 1998;Cabada et al., 2004;Spinetti and Buongiorno, 2007;Jayaratne et al., 2018). In this instance, Zheng et al. (2018) found that the low-cost light-scattering sensor correlated best with the reference-grade instrument of the same operating principle, finding a lower correlation against the reference-grade instrument operating on principles other than light-scattering. The PurpleAir instrument over-estimation of PM 2.5 relative to the BAM reference-grade instrument may therefore be due to influences of humidity acting on the light scattering (Zheng et al., 2018).
An additional consideration in explaining the overestimation of PM 2.5 by the PurpleAir instruments relative to the referencegrade BAM is the particle density of the measured particulates. The Plantower PMS5003 sensors contained within the PurpleAir instruments provide mass concentration measurements which are calculated by an unreported atmospheric calibration factor (Zheng et al., 2018). We assume that this calibration factor uses an average particle density, likely similar to that for an urban environment, such as 1.65 g cm -3 (Pitz et al., 2003;Liu et al., 2015;Crilley et al., 2018). If the Plantower PMS5003 sensor measures ambient air with a particle density dissimilar to the average particle density used in the atmospheric calibration factor, the resulting sensor output would be biased. For example, introduction of volcanic aerosol, primarily sulfate with an average particle density of 1.77 g cm -3 (Sarangi et al., 2016), into the air measured by the sensor could result in sensor output bias as a result of dissimilarity between the real and assumed particle density.
Despite the PM 2.5 over-estimation, the strong correlation between the PurpleAir and BAM instruments indicates that the PurpleAir instruments provided qualitatively valuable measurements of the atmospheric conditions. The high degree of intra-instrument performance, similar to findings by Malings et al. (2019), indicates that the PurpleAir instruments are reliable for determining relative variations in PM 2.5 . The dense network of instruments with a high spatial resolution along the western region of Hawai'i during the 2018 LERZ eruption gives a good indication of the relative amounts of PM 2.5 across the region, at a finer spatial resolution than available from the sparselylocated BAM instruments ( Figure 6A). Low-cost communityoperated networks, such as the PurpleAir instruments across the western region of Hawai'i during the 2018 LERZ eruption, can therefore be invaluable in providing insights into smallerscale heterogeneities in air quality across a regional area at a scale inaccessible by the usually more disperse reference-grade instruments (Figures 6B-E). Additionally, as far as the authors are aware, this is the first validation of PurpleAir instruments in a volcanic environment. Their strong correlation to the BAM instruments along the west coast of the Island of Hawai'i during the 2018 LERZ eruption indicates that they are suitable for augmenting reference-grade instrument networks in periods of volcanic unrest. Considering that these instruments also provide a source of open-access data to the public, they present an opportunity to improve community awareness and inclusion of the general public in hazard assessment of downwind volcanic PM 2.5 air pollution.

CONCLUSIONS
Kīlauea's 2018 eruption was the largest LERZ eruption in the last two centuries. SO 2 emissions reached a monthly average of 200 kt/day during June , significantly exceeding emissions from Kīlauea during 2008-2017, which averaged 5-6 kt/day (Eguchi et al., 2011;Beirle et al., 2014;Carn et al., 2016;Elias et al., 2018). During the 2018 LERZ eruption, SO 2 mass concentrations exceeding the Hawai'i 24-h-mean threshold (366 µg/m 3 ) primarily occurred in the south of the island, at Volcano Observatory and in Pāhala and Ocean View (2.1, 5.3, and 4.2 % of the time during the 3-month long eruption, respectively). SO 2 mass concentrations were elevated at the HDOH Kona station (average 24-h-mean mass concentration of 39 µg/m 3 during the LERZ eruption, relative to 7 µg/m 3 during 2007-2017), but mass concentrations were highest in Ocean View, Pāhala and at Volcano Observatory (average 24-h-mean mass concentrations of 117 , 129 , and 56 µg/m 3 , respectively). The Hawai'i 24-h-mean threshold was exceeded five times in Pāhala and twice at Volcano Observatory, but peak mass concentrations did not exceed those from the period 2007-2017. In Ocean View, the Hawai'i 24-h-mean threshold was exceeded four times and 24-h-mean mass concentrations peaked at 728 µg/m 3 , almost double the previous peak measurement of 403 µg/m 3 recorded at that station in January 2016. PM 2.5 mass concentrations recorded at HDOH stations around the island from 2010 to 2017 rarely exceeded the EPA 24h-mean threshold of 35 µg/m 3 (0.1 % of the time at both Ocean View and Pāhala). The lower 24-h-mean PM 2.5 limit (25 µg/m 3 ) set by the World Health Organization was exceeded with greater frequency, particularly in Kona and Ocean View (1.2 and 0.9 % of the time, respectively). During the 2018 LERZ eruption, PM 2.5 air pollution was significantly higher than 2010-2017 levels in Kona and Ocean View, exceeding WHO guidelines 34.7 and 46.3 % of the time, respectively. Peak 24-h-mean mass concentrations in Ocean View were recorded at 56 µg/m 3 , and 3 consecutive days in June were recorded with mean 24-h mass concentrations exceeding 35 µg/m 3 . The Kona HDOH stations recorded eight 24-h periods which exceeded EPA thresholds, unprecedented in the 2010-2017 period.
Following the decline of the 2018 LERZ eruption, mass concentrations of both SO 2 and PM 2.5 measured at the HDOH stations decreased to below pre-LERZ eruption levels, indicating that a large proportion of the air quality anomalies measured during the eruption were volcanogenic. The post-LERZ HDOH measurements are here assumed to be representative of the background (non-volcanically-perturbed) atmosphere, and subtracting these abundances from those recorded during the LERZ eruption provides an estimate of the purely volcanogenic PM 2.5 and SO 2 . The sulfate aerosol component within the volcanic PM 2.5 is calculated as between 77 and 92 %, following Mather et al. (2012). HYSPLIT back-trajectory simulations provide an estimate of emission age following dispersion from source to measurement point (Ilyinskaya et al., in preparation), and a first-order SO 2 decay constant is estimated at 3.8 × 10 -6 s -1 .
Community-operated PurpleAir instruments provided a high spatial resolution network across the western region of the island, informing the public regarding PM 2.5 mass concentrations in their locality. Low measurement variability (Pearson's r = 0.99) was found between co-located individual PurpleAir instruments, indicating a high level of intrainstrument performance. Observations recorded by co-located PurpleAir and BAM instruments correlated well (Pearson's r = 0.99, 0.97, and 0.94), but PurpleAir instruments were found to overestimate the PM 2.5 mass concentration by up to 40 %, relative to the BAM instrument. This likely reflects inherent differences in instrument operating principles and may be associated with changes in optical-properties of aerosol arising from hygroscopic growth in ambient humidity. Nevertheless, the PurpleAir instruments are suitable for providing a lowcost network to augment reference-grade instruments, and contribute an open-access source of readily-available information to the public leading to development of community awareness toward air quality.
This study has assessed the impacts to air quality in downwind communities around the Island of Hawai'i from 2007 to 2018. Spatial variability of air quality around the island during the 2018 LERZ eruption was comparable to patterns identified over the previous decade, but PM 2.5 and SO 2 pollution levels resulting from the 2018 LERZ eruption were significantly higher in western and southern regions of the island. A study of the potential health burden of these significant impacts on air quality might further illuminate aspects of the dose-response to volcanogenic emissions in addition to the impacts of public health protection measures put in place during the eruption.

DATA AVAILABILITY STATEMENT
The datasets analyzed for this study can be found in the Hawai'i Department of Health Ambient Air Quality Data repository [http://health.hawaii.gov/cab/hawaii-ambientair-quality-data/], and in the PurpleAir sensor list repository [https://www.purpleair.com/sensorlist].

AUTHOR CONTRIBUTIONS
RW performed the data analysis and wrote the original draft. EI, AS, TR, MP, and TM contributed to data interpretation. All co-authors contributed to draft review and editing.

FUNDING
RW is funded by the Leeds-York Natural Environment Research Council (NERC) Doctoral Training Partnership (DTP) NE/L002574/1, in CASE partnership with the Icelandic Meteorological Office. RW, EI, EM, PW, and EL were funded for campaign-fieldwork by the Centre for the Observation and Modeling of Earthquakes, Volcanoes and Tectonics (COMET) grant COME3001. The authors also acknowledge support to EI from the NERC urgency grant NE/M021130/1 Source and longevity of sulfur in Icelandic flood basalt eruption plumes and GCRF NE/P015271/1 and NE/R009465/1 Unseen but not unfelt: resilience to persistent volcanic emissions (UNRESP). We acknowledge support from the NERC ESS DTP to PW, EPSRC CASE funding to EM, Alfred P. Sloan Foundation through support of the Deep Carbon Observatory to ME and EL and Leverhulme Early Career Fellowship to EL, and ANR Projet de Recherche Collaborative VOLC-HAL-CLIM ANR-18-CE01-0018 and Orléans Labex Voltaire ANR-10-LABX-100-0 to TR. CO acknowledges support from NERC grant NE/N009312/1, and EI, TM, ME, and CO further acknowledge support from NERC grant NE/S00436X/1: Volcanic plume understanding and forecasting: Integrating remote-sensing, in-situ observations and models (V-PLUS).