1. Introduction
a. Kīlauea and vog hazard
Kīlauea volcano, located along the southern shore of the island of Hawai‘i, is an active shield volcano that has been erupting nearly continuously since 1983. Accompanying the sustained eruption, volcanic gases and ash particles have also been released, predominantly from two distinct degassing sources: (i) the Pu‘u‘ō‘ō vent along the east rift zone (ERZ) and (ii) the Halema‘uma‘u vent within the summit caldera (Fig. 1a).
(a) Map showing the island of Hawai‘i with the locations of two volcanic vents (filled circles), the locations of Hilo, Pahala, Ocean View, and Kona (open circles), and the mean annual rainfall pattern (color; mm). (b) The 600 km × 600 km domain used for the WRF simulations with smoothed terrain and 500-m elevation contours. The two filled circles represent the two volcanic vent locations at the summit (Halema‘uma‘u crater, left filled circle) and the east rift zone (ERZ; Pu‘u‘ō‘ō vent, right filled circle) where volcanic aerosols originate.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0260.1
(a) Map showing the island of Hawai‘i with the locations of two volcanic vents (filled circles), the locations of Hilo, Pahala, Ocean View, and Kona (open circles), and the mean annual rainfall pattern (color; mm). (b) The 600 km × 600 km domain used for the WRF simulations with smoothed terrain and 500-m elevation contours. The two filled circles represent the two volcanic vent locations at the summit (Halema‘uma‘u crater, left filled circle) and the east rift zone (ERZ; Pu‘u‘ō‘ō vent, right filled circle) where volcanic aerosols originate.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0260.1
(a) Map showing the island of Hawai‘i with the locations of two volcanic vents (filled circles), the locations of Hilo, Pahala, Ocean View, and Kona (open circles), and the mean annual rainfall pattern (color; mm). (b) The 600 km × 600 km domain used for the WRF simulations with smoothed terrain and 500-m elevation contours. The two filled circles represent the two volcanic vent locations at the summit (Halema‘uma‘u crater, left filled circle) and the east rift zone (ERZ; Pu‘u‘ō‘ō vent, right filled circle) where volcanic aerosols originate.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0260.1
In 2018, Kīlauea experienced a historic rift eruption and summit collapse leading to the release of more than 50 000 tons of sulfur dioxide (SO2) per day from the newly opened fissures in the lower ERZ (Neal et al. 2019). These events severely affected air quality across the island of Hawai‘i and impacted other Hawaiian Islands and ocean areas over 6000 km downstream. Kīlauea activity decreased in 2019 but increased at the end of 2020 with the summit caldera once again filling with lava and increased SO2 emissions. Prior to the 2018 rift eruption event and the 2020 summit event, it is estimated that an average of 5100 tons of SO2 were released daily from Kīlauea during the period of 2014–17 (Elias et al. 2018). This high SO2 source as well as the subsequent formation of sulfate aerosols (locally known as “vog,” short for volcanic smog) has continuously threatened the local communities over the past decade. The sulfate particulate matter has a mean diameter smaller than 2.5 μm (PM2.5), which can stay suspended in the atmosphere for long durations and is easily carried long distances, increasing its chances of being inhaled. Health studies by Longo et al. (2010) and Tam et al. (2016) reported a notable increase of upper respiratory symptom prevalence in areas of persistent vog pollution. In addition to the health effects from vog, locals believe and studies have hypothesized that vog is impacting rainfall on the island of Hawai‘i (Giambelluca et al. 2013). However, no peer-reviewed studies have been carried out on the topic.
Given the variety of vog impacts, including effects on agriculture, infrastructure, human health, and potentially rainfall, many efforts are underway to monitor the SO2 emission rate (Elias and Sutton 2012; Elias et al. 2018), study vog chemistry (Pattantyus et al. 2018), and forecast vog dispersion (Businger et al. 2015). The SO2 emission rates from the Kīlauea volcano have been regularly monitored by the U.S. Geological Survey (USGS) since 1979. The vehicle-based measurements are obtained downwind of the Kīlauea summit and ERZ, respectively (Elias et al. 2018). Because the concentration of sulfate aerosol is not measured operationally, previous studies have aimed to quantify the conversion rate of SO2 to sulfate aerosol downwind of the Kīlauea volcano (Porter et al. 2002; Beirle et al. 2014; Kroll et al. 2015), with reported SO2 loss rates ranging from 5.3 × 10−7 to 5.5 × 10−5 s−1 suggesting an SO2 half-life as short as 3.5 h with most studies giving values around 6 h. The details of the sulfur chemistry and major processes responsible for sulfate aerosol formation can be found in a review by Pattantyus et al. (2018). To mitigate the effects of the ongoing eruption to public health, in 2010, researchers at the University of Hawai‘i at Mānoa began the Vog Measurement and Prediction (VMAP) program to operationally forecast the trajectory and dispersion of the two major pollutant species (SO2 and sulfate,
The dispersion of vog is primarily dependent on the atmospheric motions in the local environment. Over the island of Hawai‘i, the local circulation can be complicated due to the interactions among trade winds, orographic effects, diurnal thermal effects, and synoptic weather systems (Chen and Nash 1994; Yang and Chen 2003). Northeasterly trade winds dominate throughout the year, especially during summer. Under trade wind conditions, the areas southwest of Kīlauea are most frequently affected by “fresh” vog composed primarily of SO2, while the onshore and offshore regions farther downwind and along the Kona coast are affected by “aged” vog, composed primarily of
b. Rainfall on Hawai‘i Island
The Hawaiian Islands have one of the most varied spatial distributions of rainfall on Earth (Giambelluca et al. 2013). Based on previous work incorporated into the Rainfall Atlas of Hawai‘i (Giambelluca et al. 2013; Frazier et al. 2016), the mean annual rainfall for a 93-yr period (1920–2012) is shown in Fig. 1a. High average rainfall amounts are found on the windward (northeast facing) mountain slopes due to persistent orographic uplift of the moist trade wind flow, with a maximum of 7408 mm yr−1 occurring on the windward lowlands to the northwest of Hilo. The driest areas are found mainly on the upper slopes of the tallest mountains, Mauna Kea and Mauna Loa, where the elevations are well above the trade wind inversion (TWI), with an average height of ~2200 m (Chen and Feng 1995), above which cloud growth is inhibited. The leeside (west facing) lowland is also dry as a result of the rain shadow created by Mauna Kea and Mauna Loa. However, when the incoming trade winds are weak, thermally driven local circulations (land-breeze/sea-breeze circulations or mountain–valley winds) may enhance or alter the local wind patterns (Chen and Nash 1994). Locally driven wind patterns are especially notable in the leeside Kona region, where a rainfall maximum occurs in the summer (dry season for the rest of the island) when solar insolation and therefore thermal circulations driving the onshore flow are both strongest (Huang and Chen 2019).
Recent statistical studies also suggest that the rainfall in Hawai‘i exhibits both a signal of interannual and interdecadal variations that are strongly modulated by the leading modes of tropical atmosphere-ocean interaction, in particular, El Niño–Southern Oscillation (ENSO), Pacific decadal oscillation (PDO), and Pacific–North American (PNA) teleconnection pattern (Chu and Chen 2005; Diaz and Giambelluca 2012; Frazier et al. 2018). In the past several decades, a clear drying trend in rainfall has been noted in over 90% of the state of Hawai‘i since 1920 (Frazier and Giambelluca 2017). It is not clear whether this drying trend is a result of natural variations in rainfall, or whether it represents a phase shift toward a drier mean state, potentially caused by anthropogenic effects (Frazier et al. 2018). Specifically for the island of Hawai‘i, Frazier and Giambelluca (2017) performed a running trend analysis and found that long-term rainfall trends are not consistent between the windward and leeward sides of the island. On the windward side, the long-term trend exhibits clear decadal variations that oscillate between positive and negative rainfall amounts in approximate 10-yr periods. However, for the leeward side, especially for the Kona region, there has been a persistent drying trend since 1920.
This spatial difference of the windward/leeward rainfall trend cannot be accounted for by large-scale natural variability or the background warming climate signal. It is likely that local factors contribute to the long-term rainfall variation in the leeside region of the island of Hawai‘i. One possible explanation for a decline in leeside rainfall suggested by Frazier and Giambelluca (2017) is related to a minor long-term change in trade wind direction, frequency, and/or intensity. Furthermore, Giambelluca et al. (2013) hypothesized that aerosol particles from the Kīlauea eruption could be impacting leeside rainfall because the region with the largest persistent drying trend is located downwind of the volcanic vents. This theory of a volcanic connection to reduced rainfall is widely accepted throughout the nonscientific community in Hawai‘i.
c. Aerosol–cloud interactions
It is well known that aerosols can serve as cloud condensation nuclei (CCN) and thus have a substantial effect on cloud microphysical properties and precipitation formation. An increase of aerosol loading with a fixed liquid water path leads to more numerous, but smaller, cloud droplets (Warner and Twomey 1967). A larger cloud droplet aggregated cross-sectional area results in more incoming shortwave radiation being reflected, referred to as the first indirect effect (Twomey 1974). Furthermore, the increased droplet number concentration decreases the precipitation efficiency, thereby increasing the cloud amount and decreasing the precipitation amount, referred to as the second indirect effect or cloud lifetime effect (Albrecht 1989). This tends to lead toward a downwind shift in precipitation location on mountains (Muhlbauer and Lohmann 2008). The sensitivity of clouds and precipitation to a change of aerosol can be cloud regime dependent (Stevens and Feingold 2009).
Here we focus on trade wind cumulus clouds, a type of small, shallow, convective cloud that forms from surface-based boundary layer mixing, or from the breakup of a stratocumulus layer. While both the first and second indirect effect have been well documented in trade cumulus clouds, detecting these effects in observations and capturing them in models is more difficult, sometimes elusive, inconclusive, or contradictory. Some previous studies have reported an invigoration effect of aerosols on trade cumulus (Yuan et al. 2011; Koren et al. 2014), while others (Jiang et al. 2010; Xue et al. 2008) suggest that increasing aerosol loading will cause suppression of the shallow convective trade cumulus cloud field. Dagan et al. (2017) and Liu et al. (2019) suggest that aerosol–cloud interactions depend on environmental conditions and that there can be nonmonotonic aerosol effects within convective clouds that depend on aerosol loading.
The continuous long-term degassing of Kīlauea provides a unique natural laboratory for studying the volcanic aerosol impacts on the downstream cloud properties. The background aerosol loading is low due to the remote marine location as is the anthropogenic aerosol loading. For satellite studies, the ocean provides a homogeneous background (Yuan et al. 2011). For these reasons, many previous studies have chosen the environment downwind of the island of Hawai‘i to study aerosol–cloud interactions (Yuan et al. 2011; Eguchi et al. 2011; Ebmeier et al. 2014; Mace and Abernathy 2016; Malavelle et al. 2017). While the environment is relatively simple, not all studies agree. For example, Yuan et al. (2011) found a decreased cloud droplet size, decreased precipitation efficiency, increased cloud amount and higher cloud tops in the trade cumulus cloud field during a high aerosol loading event, providing evidence for aerosol indirect effects. Malavelle et al. (2017) also studied the effects of volcanic aerosols on clouds and found that increasing aerosol may result in brighter clouds, but the second indirect effects appear small in their study. Some cloud properties, like cloud amount and cloud liquid water path also do not vary with the change of aerosol amount and are well buffered against the aerosols changes (Malavelle et al. 2017; Stevens and Feingold 2009).
This study is driven by the following facts and questions: the Kīlauea volcano has been continuously emitting gases that negatively impact public health and become sulfate aerosol, an efficient CCN. Studies documenting the effects of these volcanic aerosols are not always consistent in their details and prior work has shown that aerosol–cloud interactions are cloud regime dependent, nonmonotonic, and localized. Driven by the importance of aerosol–cloud interactions, especially to local rainfall, we investigate the impacts of vog on cloud and precipitation formed over the island of Hawai‘i itself. Freshwater from rainfall is a vital natural resource for the islands, so the leeside decline in rainfall over the past century and islandwide rainfall declines are causes for concern. Is it possible that rainfall trends on the island of Hawai‘i are driven or modulated by aerosols from volcanic emissions? The possible impacts of vog on rainfall are investigated using observed rainfall, observed volcanic emissions, air quality records, and an aerosol-aware numerical model.
The article is organized as follows: Section 2a describes observational datasets, followed by details of the numerical model setup in section 2b. Results from the statistical analysis of rainfall and emission observations are discussed in section 3. Section 4 presents results of a suite of numerical model simulations, including a comparison to air quality records and sensitivity experiments and results are discussed and summarized in sections 5 and 6, respectively.
2. Methodology
a. Observations
1) Daily rainfall
Significant efforts have been undertaken to combine and quality control a historic dataset containing high-spatial-resolution and high-temporal-resolution rainfall datasets across the Hawaiian Islands (Giambelluca et al. 2013; Frazier et al. 2016; Longman et al. 2018, 2019). A study by Longman et al. (2018) compiled daily rain gauge observations from heterogeneous networks for a 25-yr period from 1990 to 2014. Using the compiled point measurements, Longman et al. (2019) further generated a series of high-resolution (250-m) gridded daily rainfall maps for the same 25-yr period using a climatologically aided interpolation (CAI) approach (Willmott and Robeson 1995). In the CAI approach, rain gauge station anomalies are interpolated using an optimized inverse distance weighting (Li and Heap 2008) and then combined with mean maps (Giambelluca et al. 2013) to produce daily rainfall maps. In this study, the gridded daily rainfall maps for the island of Hawai‘i are utilized for a 1-yr period from January to December 2014.
In addition to the daily rainfall gridded maps, 201 individual rain gauge records from 2014 to 2017 are utilized (Longman et al. 2020), which have been serially completed using the normal ratio method (same as Longman et al. 2018) to partially gap fill the stations first with strong statistical relationships and second with the inverse distance-weighting algorithm. This set of 201 individual rain gauge records provides a longer rainfall record with lower spatial completeness than the gridded maps.
2) Kīlauea SO2 emissions rates
SO2 emission rates from Kīlauea have been regularly measured by the U.S. Geological Survey’s Hawaiian Volcano Observatory (HVO) since 1979, facilitated by convenient access by a road located downwind of the summit and ERZ vents which both act as volcanic degassing sites (Elias and Sutton 2012; Elias et al. 2018). The records from both the summit and ERZ vents, the two predominant emission sources from the Kīlauea volcano, are publicly available from 1992 to 2010 and from 2014 to 2017 and have been compiled in Fig. 2. With the onset of the summit eruption in 2008, a shift in the relative contributions of SO2 by the summit and ERZ vents is clearly visible. Prior to the 2008 event, the average SO2 emissions by the summit vent was only 250 tons day−1, accounting for less than 10% of the total emissions, which were mostly coming from the ERZ vent. However, during the most recent emission period from 2014 to 2017, the summit vent emissions became an order of magnitude greater than the ERZ vent emissions.
SO2 emission rates (tons day−1) from Kīlauea’s ERZ (black) and summit (red) from 1992 to 2017. Notice the increase in emissions in 2008, when a new vent opened, and missing data from 2011 through 2013.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0260.1
SO2 emission rates (tons day−1) from Kīlauea’s ERZ (black) and summit (red) from 1992 to 2017. Notice the increase in emissions in 2008, when a new vent opened, and missing data from 2011 through 2013.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0260.1
SO2 emission rates (tons day−1) from Kīlauea’s ERZ (black) and summit (red) from 1992 to 2017. Notice the increase in emissions in 2008, when a new vent opened, and missing data from 2011 through 2013.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0260.1
The SO2 emission rate is obtained by the SO2 load determined in a cross section of the plume and multiplied by the plume speed (Elias et al. 2018). However, the extremely dense volcanic plume and high SO2 amounts produced by the summit lava lake posed a challenge to conventional measurement techniques (Elias et al. 2018), resulting in a gross underestimation of the measured summit SO2 emission rates by anywhere from 20% to 90% for the 2008–10 period (Elias and Sutton 2012; Businger et al. 2015). In 2012, a new array of 10 upward looking UV spectrometers (FLYSPECs) was installed ~3 km downwind of the summit (Horton et al. 2006) and a new dual fit window (DFW) retrieval method has been utilized since 2014. The DFW method adds a second spectral window at longer wavelengths (319–330 nm) where SO2 is less absorbing to the conventional spectral window at the 305–315 nm range (Businger et al. 2015). According to Elias et al. (2018), the uncertainty of SO2 column densities calculated using this DFW method are within −6% to +22% of results obtained with other retrieval methods and do not appear to be sensitive to water vapor. The new instrumentation, more robust retrieval algorithms, and higher temporal resolution, mean that more accurate SO2 retrievals have been achieved from 2014 to 2017. This improved dataset of SO2 emissions makes it possible for us to examine the relationship between SO2 and daily rainfall. Based on the mutual availability of both high quality SO2 emissions and rainfall records, the time period from January 2014 through December 2017 is analyzed, with a total of 665 days of summit measurements and 151 days of ERZ measurements.
3) Hawai‘i ambient air quality data
In addition to emissions records, air quality measurements by the Clean Air Branch of the Hawai‘i State Department of Health (DOH) at three locations are also utilized: Kona (19.51°N, 155.91°W), Ocean View (19.12°N, 155.78°W), and Pahala (19.20°N, 155.48°W) (Fig. 1a; State of Hawai‘i Department of Health 2020). The DOH stations observe hourly SO2, wind speed and direction, which are available provisionally at the DOH website (http://air.doh.hawaii.gov/Report). Hourly observations from 2014 to 2017 are compiled to produce a 4-yr average diurnal cycle of SO2 and wind at each location. Due to the preliminary nature of the measurements, they are used only for qualitative comparison to numerical model results.
b. Numerical model setup
Idealized numerical model simulations were performed using the Weather Research and Forecasting (WRF) Model (Skamarock et al. 2008), version 3.9.1. The model domain has 400 × 400 grid spaces with 1.5-km horizontal resolution (Fig. 1b) and 99 vertical levels spaced from 70 m near the surface to 250 m near the midtroposphere. The model was integrated for 24 h, one diurnal cycle beginning from 0000 HST (midnight local time), with a 12-s time step and hourly output. The model topography is derived from Global Multiresolution Terrain Elevation Data (Danielson and Gesh 2011) and land-use categories are derived from 21-category MODIS IGBP data. The static fields were interpolated onto the 1.5-km grid. For the purpose of representing the diurnal forcing and cloud-radiative effects realistically, both the RRTMG cloud-radiation scheme (Iacono et al. 2008) and the five-layer thermal diffusion scheme (Dudhia 1996) were used. Warm rain processes were explicitly resolved using the Thompson–Eidhammer aerosol-aware cloud microphysics scheme (Thompson and Eidhammer 2014). Use of a planetary boundary layer (PBL) scheme led to the production of unrealistic rainfall across the entire domain. As such, no PBL scheme was used and, instead, mixing was handled by the Smagorinsky 3D first-order closure with mixing terms evaluated in physical space (diff_opt = 2, km_opt = 3) (Skamarock et al. 2008).
The simulation was initialized with a widely tested summertime trade wind sounding with slight modifications (Fig. 3a) which is a composite from aircraft soundings and Hilo radiosonde launches during the Hawaiian Rainband Project (HaRP) field campaign (Feng and Chen 2001). In our study, the mixing ratio of this classic HaRP sounding was reduced by 10% throughout the entire atmosphere because early simulation tests produced unrealistically large rainfall accumulations. The sounding is characterized by a trade wind inversion (TWI) layer at about 2.1 km (1.5°C increase over 250 m) and low level easterly flow of ~7–8 m s−1. The lifted condensation level (LCL) is located at 921 hPa (approximately 840 m) and the sounding contains no convective available potential energy (CAPE). The sea surface temperature is set to a constant 298 K throughout the simulation duration.
(a) The atmospheric sounding and (b) profile of the background aerosol number concentration (cm−3) used in the model. The temperature profile (red, °C) and dewpoint temperature profile (blue, °C) are accompanied by the other standard lines including pressure, skewed temperature, and dry and moist adiabats on this typical skew T–lnp chart. Wind barbs (m s−1) are included on the right side of (a).
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0260.1
(a) The atmospheric sounding and (b) profile of the background aerosol number concentration (cm−3) used in the model. The temperature profile (red, °C) and dewpoint temperature profile (blue, °C) are accompanied by the other standard lines including pressure, skewed temperature, and dry and moist adiabats on this typical skew T–lnp chart. Wind barbs (m s−1) are included on the right side of (a).
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0260.1
(a) The atmospheric sounding and (b) profile of the background aerosol number concentration (cm−3) used in the model. The temperature profile (red, °C) and dewpoint temperature profile (blue, °C) are accompanied by the other standard lines including pressure, skewed temperature, and dry and moist adiabats on this typical skew T–lnp chart. Wind barbs (m s−1) are included on the right side of (a).
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0260.1
The initial background aerosol profile is shown in Fig. 3b, with a near-surface aerosol number concentration of ~350 cm−3 that exponentially decreases with height to roughly 55 cm−3 beginning around 400 hPa. This background aerosol profile represents a clean maritime environment (similar to Nugent et al. 2016). To study the effects of additional volcanic aerosols on clouds and precipitation, we also impose two aerosol point sources at the locations of the volcanic vents (marked in Figs. 1a,b). Aerosols are coupled and interactive with hydrometeors in the Thompson–Eidhammer microphysics scheme including source terms from aerosol regeneration upon cloud droplet and rain drop evaporation and surface emissions, and sink terms from cloud droplet activation and scavenging by precipitation (Thompson and Eidhammer 2014). The number of aerosols activated to become cloud droplets in a grid box is based on the number of aerosols present, vertical velocity, temperature, aerosol size, and hygroscopicity. In this study, the sulfate-like aerosols we simulate all have an assumed particle distribution with a mean size of 0.04 μm and a hygroscopicity of 0.4.
Aerosols in the model are emitted directly from the two point sources on Hawai‘i Island. Because the number concentration of sulfate aerosol is not observed nor measured, and the only available observations are of SO2 concentration, we compare the results of a control case with no volcanic emissions to three sensitivity studies with varying aerosol emission rates. Sensitivity studies for volcanic emissions of 100, 1000, and 5000 aerosol particles cm−3 s−1 from each volcanic vent (summit and ERZ) are investigated. We will refer to these experiments as Na100, Na1000, and Na5000, respectively. The source quantities are given in units of aerosol particles cm−3 s−1, equivalent to a surface flux spread over the lowest model level of depth 70 m. A source of 100 cm−3 s−1, for example, is equivalent to a surface flux of 7 × 109 aerosol particles m−2 s−1.
It is important to note that the treatment of volcanic aerosol emissions is simplified in these idealized simulations. There are no chemical processes, such as the conversion from SO2 to sulfate, just physical processes that control the interaction of aerosols with hydrometeors. Explicitly representing the chemical processes is significantly more complex and not necessary for the study objectives.
3. Statistical analysis of observational datasets
To assess whether vog is affecting local rainfall, the rainfall differences between high and low-emission days are compared. This is done separately for the two rainfall datasets, first with the daily rainfall maps for 2014 and second with individual daily rain gauge observations from 2014 to 2017. High- and low-emission days are determined based on Kīlauea SO2 emissions rates (Fig. 4a). The primary degassing source of Kīlauea during 2014–17 was from the summit lava lake, with emissions ranging from 700 to 2900 tons day−1. Summit emissions were an order of magnitude greater than ERZ emissions so both analyses only include days when summit measurements are available. High-emission (low-emission) days are defined as having a daily summit SO2 emission rate greater (less) than 2000 tons day−1. In 2014, a total of 102 high-emission days and 68 low-emission days were identified based on these bounds, in the period 2014–17, 438 high-emission days and 227 low-emission days were identified.
(a) SO2 emission rates from Kīlauea’s ERZ (black) and summit (red) from 2014 to 2017. The dashed lines indicate the threshold of 2000 tons day−1. Daily rainfall maps are composited for the days of 2014 when (b) summit SO2 > 2000 tons day−1 and (c) summit SO2 < 2000 tons day−1, and individual rain gauge stations for the days of 2014–17 when (e) summit SO2 > 2000 tons day−1 and (f) summit SO2 < 2000 tons day−1. Maps of the Pearson correlation coefficients between observed summit SO2 emission rates and (d) daily rainfall maps from 2014 and (g) daily rain gauge rainfall from 2014 to 2017. Notice the change in scale from (d) to (g). Stippling indicates areas where the correlation coefficient is statistically significant at the 95% confidence level.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0260.1
(a) SO2 emission rates from Kīlauea’s ERZ (black) and summit (red) from 2014 to 2017. The dashed lines indicate the threshold of 2000 tons day−1. Daily rainfall maps are composited for the days of 2014 when (b) summit SO2 > 2000 tons day−1 and (c) summit SO2 < 2000 tons day−1, and individual rain gauge stations for the days of 2014–17 when (e) summit SO2 > 2000 tons day−1 and (f) summit SO2 < 2000 tons day−1. Maps of the Pearson correlation coefficients between observed summit SO2 emission rates and (d) daily rainfall maps from 2014 and (g) daily rain gauge rainfall from 2014 to 2017. Notice the change in scale from (d) to (g). Stippling indicates areas where the correlation coefficient is statistically significant at the 95% confidence level.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0260.1
(a) SO2 emission rates from Kīlauea’s ERZ (black) and summit (red) from 2014 to 2017. The dashed lines indicate the threshold of 2000 tons day−1. Daily rainfall maps are composited for the days of 2014 when (b) summit SO2 > 2000 tons day−1 and (c) summit SO2 < 2000 tons day−1, and individual rain gauge stations for the days of 2014–17 when (e) summit SO2 > 2000 tons day−1 and (f) summit SO2 < 2000 tons day−1. Maps of the Pearson correlation coefficients between observed summit SO2 emission rates and (d) daily rainfall maps from 2014 and (g) daily rain gauge rainfall from 2014 to 2017. Notice the change in scale from (d) to (g). Stippling indicates areas where the correlation coefficient is statistically significant at the 95% confidence level.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0260.1
The corresponding composite rainfall maps were generated for high- and low-emission days for the 2014 rainfall maps (Figs. 4b,c) and for the 2014–17 rain gauge record (Figs. 4e,f). In both analyses, the primary rainfall differences are found downwind of Kīlauea, where it is observed to be drier under high-emission days; in the rainfall maps, the difference between the two emission categories is 8 mm on average. For the rest of the island of Hawai‘i, the overall rainfall patterns for the high- and low-emission categories are similar, meaning that the large-scale background environmental conditions are not significantly different and the localized reduction of rainfall downwind is a result of the high volcanic aerosol emissions.
To quantify the relationship between rainfall and SO2 emissions rates, the Pearson correlation coefficient was calculated using all 170 days of summit SO2 sampling records and the corresponding rainfall maps in the calendar year 2014 (Fig. 4d), and 665 days of summit SO2 sampling records and corresponding daily rainfall from the rain gauge records from 2014 to 2017 (Fig. 4g). An anticorrelation is identified over the southern flank of the island, significant at the 95% confidence level in the 2014 rainfall maps, indicating that increasing SO2 is associated with lower rainfall amounts downwind of Kīlauea. The region with a significant anticorrelation is the same region identified by the composite rainfall maps for high- and low-emission days. Weak positive correlations have also been found over the leeside Kona region, Kohala Mountain, and the Waimea Saddle as well as weak negative correlations over the windward region, but these results are not statistically significant. Results from the longer rain gauge record are consistent, but somewhat weaker, although values larger than ±0.078 are significant at the 95% confidence level.
Based on statistical analysis of SO2 emission records and rainfall observations, it is evident that rainfall downwind of Kīlauea volcano is modulated by the quantity of SO2 emissions. We hypothesize that the formed aerosol particles may be advected and trapped under the TWI in the usual prevailing trade wind conditions, thereby affecting the evolution of cloud and rain processes in the downwind region through the second indirect effect. However, with a lack of sufficient density and quality of aerosol and cloud microphysical observations, quantitative analysis of these effects from observations alone is impossible. Due to this limitation, we take a model-driven approach to test our hypotheses and gain further insight into the detailed microphysical and dynamical processes at play.
4. Numerical model results
a. Control simulation: Trade wind flow
The control simulation has no volcanic aerosol sources, only the background aerosol distribution (Fig. 3b). This allows for a diagnosis of the diurnal variation of airflow, thermodynamic fields, cloud, and spatial distributions of rainfall (Fig. 5) over the island of Hawai‘i and adjacent ocean before considering the potential aerosol impacts of volcanic emissions. In the control simulation, clouds begin to form at 0900 HST (Fig. 5b) when the localized upslope-sea-breeze circulation is pronounced beneath the TWI (~2 km) around the island due to morning surface heating (Fig. 5a). The simulated thermal contrast between the slope surface over the windward slope and the adjacent areas at the same elevation at 0900 HST is near 1°–4°C. This thermal contrast is directly related to the strength of the surface heat flux driving the sea-breeze circulation. With continued solar heating, the onset of rain formation occurs at 1200 HST, preceded by a stronger onshore wind component and a higher cloud amount over the mountain slopes. Peak convection occurs at 1500 HST, with maximum vertically integrated cloud water content of 1.5 kg m−2 and rainwater content of 0.9 kg m−2 located at the southern flank of the island. During the evening transition (1800 HST), without strong surface heating during sunset, the air–ground temperature difference diminishes. The surface airflow on the windward side begins to split and develop a large component parallel to topographic contours due to orographic blocking as the land cools (Fig. 5j). Both cloud and rain content decrease over the island and increase in the leeside wake circulations, which may be related to the advection of island convection offshore and increased instability offshore in the evening as cloud tops cool. The downstream clouds have a higher elevation than the over-island clouds suggesting their presence is related to cloud-top detrainment.
(left) Simulated surface wind (vectors, m s−1) and the differences between the 2-m surface temperature and upstream open-ocean temperature at the same elevation (color, °C). (center) Vertically integrated cloud water content (kg m−2). (right) Vertically integrated rainwater content (kg m−2). Rows show different simulation times of the control run, beginning with (top) 0900 HST, every 3 h, to (bottom) 1800 HST. Black contour lines show terrain height in 500-m increments.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0260.1
(left) Simulated surface wind (vectors, m s−1) and the differences between the 2-m surface temperature and upstream open-ocean temperature at the same elevation (color, °C). (center) Vertically integrated cloud water content (kg m−2). (right) Vertically integrated rainwater content (kg m−2). Rows show different simulation times of the control run, beginning with (top) 0900 HST, every 3 h, to (bottom) 1800 HST. Black contour lines show terrain height in 500-m increments.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0260.1
(left) Simulated surface wind (vectors, m s−1) and the differences between the 2-m surface temperature and upstream open-ocean temperature at the same elevation (color, °C). (center) Vertically integrated cloud water content (kg m−2). (right) Vertically integrated rainwater content (kg m−2). Rows show different simulation times of the control run, beginning with (top) 0900 HST, every 3 h, to (bottom) 1800 HST. Black contour lines show terrain height in 500-m increments.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0260.1
Idealized model simulations are initialized with the HaRP composite sounding, representative of upstream early morning conditions. The results are therefore not designed to reproduce the weather on any particular day. The goal of these idealized simulations is for this typical HaRP sounding composite to produce a similar mean rainfall pattern to one experienced on a majority of days throughout the year (Fig. 1a). In addition, we focus on the rainfall in the region downwind of the Kīlauea volcano and are much less concerned with windward rainfall. Previous HaRP observations suggest that a large portion of rainfall on the windward side of Hawai‘i Island originates far upstream as trade wind showers (Chen and Feng 2001; Frye and Chen 2001). This source of windward rainfall is not included in these relatively simple idealized simulations due to the lack of upstream convection and therefore rainfall amounts on the northeast facing windward slopes are severely underestimated. Despite this windward rainfall difference due to experiment design, when our simulations are compared with HaRP observations (Chen and Wang 1994; Chen and Feng 1995), we find that our idealized simulation results can still capture the key dynamic- and thermal-induced features.
b. Diurnal variations of simulated aerosol
With successful simulations of the land–sea thermal contrast and orographic airflow regimes over one diurnal heating cycle, we next evaluate whether the added aerosols are successfully simulated using hourly DOH air quality records. Observed SO2 averaged over 4 years at three operational stations downstream of Kīlauea (labeled in Fig. 1a) are compared to the aerosols in the simulation with the highest volcanic aerosol source (Na5000). The comparison in Fig. 6 shows important changes in simulated aerosol with height and observed SO2 at the surface throughout one diurnal day that are consistent with one another. Both observed SO2 and simulated aerosol concentrations in Pahala increase in concentration and altitude and peak at 0900 HST before rapidly dropping off. Ocean View experiences a steady stream of aerosols throughout much of the day, with the highest concentrations between 0900 and 1200 HST. Finally, Kona experiences only a low concentration of aerosols that increases slightly in the late afternoon around 1500 HST.
Diurnal variations of (left) simulated aerosol number concentrations with altitude (color; cm−3) at (a) Pahala, (c) Ocean View, and (e) Kona and (right) observed SO2 (line; ppm) and wind vectors at (b) Pahala, (d) Ocean View, and (f) Kona from DOH air quality monitoring stations averaged hourly over 4 years. Note the changes in scale in the right column.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0260.1
Diurnal variations of (left) simulated aerosol number concentrations with altitude (color; cm−3) at (a) Pahala, (c) Ocean View, and (e) Kona and (right) observed SO2 (line; ppm) and wind vectors at (b) Pahala, (d) Ocean View, and (f) Kona from DOH air quality monitoring stations averaged hourly over 4 years. Note the changes in scale in the right column.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0260.1
Diurnal variations of (left) simulated aerosol number concentrations with altitude (color; cm−3) at (a) Pahala, (c) Ocean View, and (e) Kona and (right) observed SO2 (line; ppm) and wind vectors at (b) Pahala, (d) Ocean View, and (f) Kona from DOH air quality monitoring stations averaged hourly over 4 years. Note the changes in scale in the right column.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0260.1
To account for the variation in SO2 and aerosol, the diurnal variation of the observed surface wind vectors have also been plotted in Figs. 6b, 6d, and 6f. Consistent with Fig. 5, all three stations show that aerosol transport is strongly dominated by the thermally driven surface wind, with daytime upslope/sea-breeze flow and nighttime downslope/land-breeze flow. For Pahala station (Fig. 6b), the strong northeasterly upslope flow beginning at 1000 HST rapidly transports the existing aerosols to higher elevation until sundown when downslope motion starts to rebuild the high concentration. The evolution of the mixing-layer depth over the island as it warms likely also plays a role. The concentration at Ocean View is still high, but more dispersed than at Pahala, and slightly lags the Pahala pattern with less of a distinct diurnal cycle. For Kona (Fig. 6f), the daytime sea-breeze flow brings substantial moisture onshore in the afternoon to generate rainfall, but also SO2 and aerosols that have been swept downwind and wrapped around the island topography. Both observed SO2 and simulated aerosol concentrations show that vog influence in Kona is nearly a tenth of that in Pahala and much lower than Ocean View as well with a larger temporal lag.
c. Aerosol impacts on cloud microphysics
Three model sensitivity tests with varying aerosol sources from the two volcanic vent locations (summit and ERZ) are examined to determine the impact of volcanic aerosols on cloud properties and precipitation. Figure 7 presents all simulations at 1400 HST when convection has fully developed over the island and the added aerosol has interacted with the surrounding environment. Under the prevailing trade wind conditions, the emitted aerosol is advected from the vent sources toward the west-southwest (Figs. 7d,g,j), with a high chance of interacting with the orographic convection forming over the southern flank of the island. For the high-emission cases (Na1000 and Na5000, Figs. 7g,j), the long aerosol plume extends farther downstream, embedded in the southern rotating wake vortex, which was well described by Smith and Grubišić (1993) during the HaRP field experiment. A slice through the number concentration and mean volume diameter of cloud droplets fields at the elevation of the trade wind inversion (~2 km) finds that along with the increasing aerosol concentrations, there is both an increase of cloud droplet number concentration and a decrease of mean volume diameter in regions downwind of Kīlauea (Figs. 7h,i,k,l). The maximum cloud droplet size decreases from 55 μm in the control case to approximately 20 μm in the Na5000 case. The small cloud droplet sizes also affect the radiation budget by reflecting higher amounts of shortwave radiation.
WRF simulation results at 1400 HST when the clouds are fully developed in (top to bottom) the control run, Na100, Na1000, and Na5000 simulations, respectively. Columns show (left) the vertically integrated aerosol number concentration (cm−2), (center) the cloud droplet number concentration (cm−3) at 2-km height, and (right) the mean volume diameter (μm) of cloud droplets at 2-km height. The areas near the center of the island have no data because they intersect the high terrain. Terrain contours are at increments of (left) 1 km and (center),(right) 500 m.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0260.1
WRF simulation results at 1400 HST when the clouds are fully developed in (top to bottom) the control run, Na100, Na1000, and Na5000 simulations, respectively. Columns show (left) the vertically integrated aerosol number concentration (cm−2), (center) the cloud droplet number concentration (cm−3) at 2-km height, and (right) the mean volume diameter (μm) of cloud droplets at 2-km height. The areas near the center of the island have no data because they intersect the high terrain. Terrain contours are at increments of (left) 1 km and (center),(right) 500 m.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0260.1
WRF simulation results at 1400 HST when the clouds are fully developed in (top to bottom) the control run, Na100, Na1000, and Na5000 simulations, respectively. Columns show (left) the vertically integrated aerosol number concentration (cm−2), (center) the cloud droplet number concentration (cm−3) at 2-km height, and (right) the mean volume diameter (μm) of cloud droplets at 2-km height. The areas near the center of the island have no data because they intersect the high terrain. Terrain contours are at increments of (left) 1 km and (center),(right) 500 m.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0260.1
d. Aerosol impacts on precipitation
Previous studies have documented that increasing numbers of small cloud droplets in polluted scenarios will affect the precipitation efficiency (Albrecht 1989; Stevens and Seifert 2008). With this in mind, the simulated 24-h rainfall amounts for the four model experiments are compared in Fig. 8. The model successfully simulates important features in the mean rainfall distribution downwind of Kīlauea (Fig. 1) but poorly simulates windward island rainfall. In the model simulations, there is little to no rainfall on the windward slopes but because vog rarely pollutes the windward slopes it is not of direct concern for the aims of our study. The key point of comparison is the realistic rainfall northwest of Pahala with reasonable rainfall rates, amount, location, and coverage. Consistent with observations (Fig. 4), the model sensitivity studies show a monotonic reduction of rainfall accumulation over the southern flank of the island when aerosol impacts are considered. The rainfall difference between sensitivity tests and the control run are as large as 21.1 mm day−1 (55% less) and 28.5mm day−1 (73% less) in the Na1000 and Na5000 simulations, respectively. However, the pattern of the rainfall change is especially notable. While most regions experience less rainfall in the higher aerosol emission simulations, some regions (e.g., near the peak of southern flank) have either no change (Fig. 8e) or even a slight increase in rainfall as compared to the control run (Figs. 8f,g), suggesting that the orographic precipitation distribution has shifted.
The simulated 24-h rainfall accumulation (mm day−1) from the (a) control run, (b) Na100, (c) Na1000, and (d) Na5000 simulations. Also shown are rainfall differences (mm) between the control run and three sensitivity tests: (e) Na100 − control, (f) Na1000 − control, and (g) Na5000 − control. Note the asymmetric color scale in (e)–(g).
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0260.1
The simulated 24-h rainfall accumulation (mm day−1) from the (a) control run, (b) Na100, (c) Na1000, and (d) Na5000 simulations. Also shown are rainfall differences (mm) between the control run and three sensitivity tests: (e) Na100 − control, (f) Na1000 − control, and (g) Na5000 − control. Note the asymmetric color scale in (e)–(g).
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0260.1
The simulated 24-h rainfall accumulation (mm day−1) from the (a) control run, (b) Na100, (c) Na1000, and (d) Na5000 simulations. Also shown are rainfall differences (mm) between the control run and three sensitivity tests: (e) Na100 − control, (f) Na1000 − control, and (g) Na5000 − control. Note the asymmetric color scale in (e)–(g).
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0260.1
To quantify the overall effects and describe the cloud and rain response to aerosol loading changes, we first present in Fig. 9 the temporal evolution of bulk properties of clouds and key microphysical processes for cloud evolution (e.g., condensation, evaporation, autoconversion, and accretion). We focus on the southern flank region that is heavily exposed to upstream aerosol sources and experiences significant rainfall accumulation reduction (black box labeled in Fig. 8g). As expected, the lower precipitation efficiency in higher-emission cases shows that more liquid water mass tends to be stored as cloud water instead of being transformed into rainwater (Figs. 9a,b). Further analysis of the cloud water budget finds that simulations with increased aerosol emissions do not result in a significant change in cloud condensation efficiency (Fig. 9c), but instead result in a higher cloud evaporation rate (Fig. 9d). For cases where the environment has already had sufficient aerosol loading, however, further increases in aerosol amount do not increase the cloud evaporation rate (red and blue lines in Fig. 9d). When analyzing the rainwater budget, it is found that the suppression of surface rain rate (Fig. 9h) in high-emission cases is mainly caused by the delay and reduction of the autoconversion rate (Fig. 9e) which is further amplified in changes to the accretion rate (Fig. 9f).
The temporal evolution of key cloud and rain processes in the four sensitivity simulations: control (black), Na100 (yellow), Na1000 (blue), and Na5000 (red). Total integrated (a) cloud water (kg), (b) rainwater (kg), (c) cloud condensation rate (kg s−1), (d) cloud evaporation rate (kg s−1), (e) autoconversion rate (kg s−1), (f) accretion rate (kg s−1), (g) rain evaporation rate (kg s−1), and (h) surface rain rate (mm h−1) through time within the box in Fig. 8g. Note the changes in the y-axis scale.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0260.1
The temporal evolution of key cloud and rain processes in the four sensitivity simulations: control (black), Na100 (yellow), Na1000 (blue), and Na5000 (red). Total integrated (a) cloud water (kg), (b) rainwater (kg), (c) cloud condensation rate (kg s−1), (d) cloud evaporation rate (kg s−1), (e) autoconversion rate (kg s−1), (f) accretion rate (kg s−1), (g) rain evaporation rate (kg s−1), and (h) surface rain rate (mm h−1) through time within the box in Fig. 8g. Note the changes in the y-axis scale.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0260.1
The temporal evolution of key cloud and rain processes in the four sensitivity simulations: control (black), Na100 (yellow), Na1000 (blue), and Na5000 (red). Total integrated (a) cloud water (kg), (b) rainwater (kg), (c) cloud condensation rate (kg s−1), (d) cloud evaporation rate (kg s−1), (e) autoconversion rate (kg s−1), (f) accretion rate (kg s−1), (g) rain evaporation rate (kg s−1), and (h) surface rain rate (mm h−1) through time within the box in Fig. 8g. Note the changes in the y-axis scale.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0260.1
Changes in precipitation from aerosol loading can also come from microphysical feedbacks that lead to dynamical feedbacks on clouds. For example, based on the cloud lifetime effect, one would expect high aerosol loading to lead to a larger cloud fraction. This effect is seen in the simulations, whereby the Na1000 and Na5000 simulations have a slightly larger cloud fraction, but the difference between all sensitivity simulations is small at <5%. Cloud-top height and cloud depth differences between all sensitivity simulations are also small at <200 m and vary over time. Taller clouds are typically expected to produce more rain, but the taller clouds are found in the Na1000 and Na5000 cases which produce less rain, showing that microphysical changes are dominant. Despite the existence of minor cloud structure changes, the significance of these changes is difficult to diagnose and the largest difference between the simulations remain the cloud microphysical changes associated with precipitation formation.
For the purpose of investigating the effects of aerosol on the precipitation distribution, Figs. 10a–d show Hovmöller diagrams of surface rain rate for the four cases along the dashed line cross section shown in Fig. 8g, which is perpendicular to the terrain contours and lies across the center of convection within the black box. In all cases, surface rainfall starts around noon (1200 HST) and dissipates around 1800 HST in response to daytime heating. However, the timing and location of the peak rainfall intensity and accumulated 24-h rainfall amounts (Fig. 10e) vary due to the microphysical feedbacks from differing aerosol loading. In the control run, the location that receives the most rainfall (19 mm day−1) is found near the middle of the upslope region. Adding a small aerosol source in Na100 results in a small decrease in the rainfall peak to 16 mm day−1 due to the suppression of both autoconversion and accretion (Figs. 9e,f), but with no significant change in the peak location. Further increases in the emission source in Na1000 and Na5000 result in decreasing rainfall amounts, and the location of the rainfall maximum shifts downstream toward the terrain peak.
Hovmöller diagrams of surface rain rate (mm h−1) for (a) control, (b) Na100, (c) Na1000, and (d) Na5000, along with (e) daily accumulated surface rain rate (mm day−1) and (f) terrain height (m) along the dashed-line cross section in Fig. 8g.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0260.1
Hovmöller diagrams of surface rain rate (mm h−1) for (a) control, (b) Na100, (c) Na1000, and (d) Na5000, along with (e) daily accumulated surface rain rate (mm day−1) and (f) terrain height (m) along the dashed-line cross section in Fig. 8g.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0260.1
Hovmöller diagrams of surface rain rate (mm h−1) for (a) control, (b) Na100, (c) Na1000, and (d) Na5000, along with (e) daily accumulated surface rain rate (mm day−1) and (f) terrain height (m) along the dashed-line cross section in Fig. 8g.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0260.1
The results are supported by a previous study by Muhlbauer and Lohmann (2008), who approached the same question of aerosols in warm clouds under a pure dynamic flow regime (no heating) within an idealized 2D model setup. In this study, the additional thermal effects and the application of realistic 3D complex terrain make the results more complicated. This is expected because studies like Seo et al. (2020) found the upslope geometry will control the aerosol effects on orographic precipitation. They show that aerosol effects on orographic precipitation are more clearly seen in cases with a narrow windward width compared to cases with a symmetric mountain and wide windward width (Seo et al. 2020).
5. Discussion
Long-term trends of Hawaiian rainfall show a monotonic drying trend on the leeside of Hawai‘i Island since 1920 (Frazier and Giambelluca 2017). The results presented here cannot alone account for this long-term drying trend, as only one case study is analyzed, volcanic emission records prior to 1979 are not available, and early SO2 observation are infrequent (Elias et al. 1998). The drying evident in the observations and simulations due to volcanic emissions is on the downstream mountainside, within the area found to have a statistically significant long-term drying signal especially in the summer dry season when the trade wind pattern is dominant (Frazier and Giambelluca 2017). However, while the downstream mountainside shows a pattern of drying due to volcanic emissions in observations and simulations (Fig. 8), the leeside Kona region does not. The results presented herein also cannot account for the variations in rainfall that occur on the windward side of Hawai‘i Island, nor rainfall variations that occur on other Hawaiian Islands.
The short 24-h simulation time does not allow for the buildup of aerosol over longer time scales, for example, the potential build up of recirculated aerosol in the southern eddy downwind of Hawai‘i Island (e.g., Fig. 18 in Smith and Grubišić 1993). It is possible that our results do not find significant rainfall changes in the Kona region but longer duration simulations with older recirculating and “aged” aerosol may find a more significant impact there.
Low aerosol values are both simulated and observed in Kona. In the simulations, the low aerosol concentrations are primarily due to dispersion downstream of Kīlauea, and the cold start. In observations, the low concentrations of SO2 in air quality records can be due to both dispersion and the conversion of SO2 to sulfate. Still, the model and observations match well, leading to the unsurprising result that simulated cloud and precipitation are more impacted by microphysical processes near Pahala than Kona.
Here we only consider one particular environmental condition, typical trade wind flow, which occurs approximately 65%–90% of the time, depending on the season (Garza et al. 2012). In reality, the aerosol impacts on cloud microphysics, structure, and dynamics may vary depending on the specific environmental conditions like wind speed, moisture content, or atmospheric stability (Stevens and Feingold 2009; Dagan et al. 2015; Nugent et al. 2016). Accumulated effects over long-term time scales need to integrate over all weather conditions and emission variability. Even the relatively small changes found on the single simulated typical day are likely to have significant climate effects when accumulated over long time scales.
The impact of volcanic emissions on rainfall has been long argued by the islands’ populace, but no peer-reviewed scientific study ever looked into this. Primary results from our study act as a confirmation to this anecdotal evidence. Furthermore, while aerosol–cloud impacts on precipitation have been quantified downstream of volcanoes from prior satellite studies over the open ocean (e.g., Malavelle et al. 2017; Yuan et al. 2011), this is the first study we know of that quantifies the impacts of volcanic emissions on rainfall on the same volcanic island. This study provides precedent to the idea that local aerosol sources can and do impact nearby rainfall. While this study examined volcanic aerosols, it is also likely that other large local aerosol sources can impact their local environment as well.
This study has important applications for local residents of Hawai‘i Island. The change in surface aerosol concentration and air quality over the course of a day, seen clearly and consistently both in observations and simulations (Fig. 6) verifies the model performance both in dynamics and in aerosol advection and treatment. Additional application of real case studies of this type can be valuable to local communities for understanding the diurnal fluctuations in air quality that are sensitive to location downwind of Kīlauea, time of day, and local winds that are variable but predictable. The diurnal air quality observations show safe and unsafe times of day to be outside and the model simulations provide insight into the underlying mechanisms. The World Health Organization (WHO) has set a safe-level guideline of <125 μg m−3 of SO2 in a 24-h average, equivalent to ~0.045 ppm (World Health Organization 2006). Figure 6b shows this value being exceeded at Pahala in the 4-yr averaged DOH air quality observations between 0400 and 1000 HST. Despite the differences inherent in a 24-h exposure limit and a 4-yr hourly average, research shows negative health impacts from SO2 occurs from high levels of SO2 and/or long-duration exposure (Tam et al. 2016), suggesting it is especially unsafe in Pahala at specific times of day.
6. Summary
In this study, the impacts of volcanic aerosol on cloud and precipitation over the island of Hawai‘i have been studied using both observational datasets and numerical model simulations. Statistical analysis shows that rainfall downwind of the Kīlauea volcano is strongly modulated by the amount of SO2 emissions. Comparing the composited rainfall maps between high and low SO2 emissions days, it is evident that the downstream mountainside is likely to have less rainfall (8 mm day−1) on days with high SO2 emissions. The rainfall amount within downwind regions exhibits a significant anticorrelation (down to −0.3) with the SO2 emission amount, while the rest of the island has a relatively weak and nonsignificant correlation (<|±0.15|), emphasizing that the rainfall variations are attributable to localized vog, not large-scale climate variability or weather conditions that vary regularly throughout the year, independent of emission strength.
To account for the observed features and gain insight into the detailed microphysical and dynamical processes, one full diurnal day is examined with sensitivity simulations that have varied volcanic aerosol emission rates from the Kīlauea volcano. These are tested with the aerosol-aware Thompson–Eidhammer microphysics scheme in the WRF Model to focus on the small-scale microphysical pattern development. The simulation of one diurnal day allows for a unique comparison of cloud and rainfall development downwind of the Kīlauea volcano. Consistent with observational analysis, a monotonic decrease in accumulated daily precipitation amount with increasing aerosol emissions is found within the regions downwind of Kīlauea (Fig. 8). Under higher-emission conditions (Na1000 and Na5000), both the autoconversion and accretion warm rain processes are delayed (60–90 min) and inhibited (up to 90% less), and clouds tend to readily evaporate their smaller cloud droplets (up to 40% more) when compared to the control simulation (Fig. 9). In addition to having a lower rainfall amount (Fig. 8), the high emission simulations also show a change in the spatial and temporal distribution of rainfall (Fig. 10). The rainfall maximum shifts downstream toward the terrain peak and rain falls at later times, due to microphysical feedbacks resulting from the increased atmospheric aerosol number concentration, which retard rainfall formation processes.
The results of this study confirm an age-old idea that volcanic aerosol emissions from the Kīlauea volcano influence rainfall on Hawai‘i Island. Additional research can make significant contributions in future. For example, decreasing the grid spacing will allow for improved resolution of aerosol–cloud interactions, mountain meteorology, and local air quality. Adding variations in temperature or humidity to the upwind trade wind region, or additional time and space to generate small-scale turbulence would allow for more realistic windward rainfall (Kirshbaum and Smith 2009). Longer-term simulations driven by reanalysis of real environmental conditions will allow for a full variety of cloud and rainfall patterns to test the sensitivity of vog on rainfall climatologically in addition to allowing for the buildup of aerosol over time, which may lead to stronger impacts on the Kona region. Finally, vog chemistry is not explicitly represented in this study, so future work could include a coupled chemistry model like WRF-Chem.
Acknowledgments
The research presented in this paper was performed on Cheyenne (doi:10.5065/D6RX99HX), a high-performance computing resource provided by NCAR’s Computational and Information Systems Laboratory, sponsored by the National Science Foundation. Both Tianqi Zuo and Alison D. Nugent were supported by NSF AGS Grant 1854443. We thanks Elias Tamar and Lacey Holland for helpful discussions about vog emission rates, Andre Pattantyus for discussions about SO2 conversion rates to sulfate aerosol, and Abby Frazier and Thomas Giambelluca for discussions on rainfall patterns and vog hypotheses. We also thank Yi-Leng Chen for providing the HaRP sounding used in the simulations, and Katherine Ackerman for inspiring research involving SO2 DOH observations. This is SOEST Contribution Number 11363.
Data availability statement
Gridded daily rainfall datasets and daily rain gauge datasets used in this research were obtained from Longman et al. (2019, 2020). Kīlauea volcanic aerosol SO2 emissions were obtained from Elias and Sutton (2012) and Elias et al. (2018). Air quality and wind records were obtained from publicly available resources through the State of Hawai‘i Department of Health (2020).
REFERENCES
Albrecht, B. A., 1989: Aerosols, cloud microphysics, and fractional cloudiness. Science, 245, 1227–1230, https://doi.org/10.1126/science.245.4923.1227.
Beirle, S., C. Hörmann, M. Penning De Vries, S. Dörner, C. Kern, and T. Wagner, 2014: Estimating the volcanic emission rate and atmospheric lifetime of SO2. Atmos. Chem. Phys., 14, 8309–8322, https://doi.org/10.5194/acp-14-8309-2014.
Businger, S., R. Huff, A. Pattantyus, K. Horton, A. J. Sutton, T. Elias, and T. Cherubini, 2015: Observing and forecasting vog dispersion from Kīlauea volcano, Hawaii. Bull. Amer. Meteor. Soc., 96, 1667–1686, https://doi.org/10.1175/BAMS-D-14-00150.1.
Chen, Y.-L., and A. J. Nash, 1994: Diurnal variation of surface airflow and rainfall frequencies on the island of Hawaii. Mon. Wea. Rev., 122, 34–56, https://doi.org/10.1175/1520-0493(1994)122<0034:DVOSAA>2.0.CO;2.
Chen, Y.-L., and J.-J. Wang, 1994: Diurnal variation of surface thermodynamic fields on the island of Hawaii. Mon. Wea. Rev., 122, 2125–2138, https://doi.org/10.1175/1520-0493(1994)122<2125:DVOSTF>2.0.CO;2.
Chen, Y.-L., and J. Feng, 1995: The influences of inversion height on precipitation and airflow over the island of Hawaii. Mon. Wea. Rev., 123, 1660–1676, https://doi.org/10.1175/1520-0493(1995)123<1660:TIOIHO>2.0.CO;2.
Chen, Y.-L., and J. Feng, 2001: Numerical simulations of airflow and cloud distributions over the windward side of the island of Hawaii. Part I: The effects of trade wind inversion. Mon. Wea. Rev., 129, 1117–1134, https://doi.org/10.1175/1520-0493(2001)129<1117:NSOAAC>2.0.CO;2.
Chu, P. S., and H. Chen, 2005: Interannual and interdecadal rainfall variations in the Hawaiian Islands. J. Climate, 18, 4796–4813, https://doi.org/10.1175/JCLI3578.1.
Dagan, G., I. Koren, and O. Altaratz, 2015: Competition between core and periphery-based processes in warm convective clouds—From invigoration to suppression. Atmos. Chem. Phys., 15, 2749–2760, https://doi.org/10.5194/acp-15-2749-2015.
Dagan, G., I. Koren, O. Altaratz, and R. H. Heiblum, 2017: Time-dependent, non-monotonic response of warm convective cloud fields to changes in aerosol loading. Atmos. Chem. Phys., 17, 7435–7444, https://doi.org/10.5194/acp-17-7435-2017.
Danielson, J., and D. Gesh, 2011: Global Multi-Resolution Terrain Elevation Data 2010 (GMTED2010). USGS Open-File Rep. 2011-1073, 23 pp., https://doi.org/10.3133/ofr20111073.
Diaz, H. F., and T. W. Giambelluca, 2012: Changes in atmospheric circulation patterns associated with high and low rainfall regimes in the Hawaiian Islands region on multiple time scales. Global Planet. Change, 98–99, 97–108, https://doi.org/10.1016/j.gloplacha.2012.08.011.
Dudhia, J., 1996: A multi-layer soil temperature model for MM5. Preprints, Sixth PSU/NCAR Mesoscale Model Users’ Workshop, Boulder, CO, NCAR, 49–50.
Ebmeier, S. K., A. M. Sayer, R. G. Grainger, T. A. Mather, and E. Carboni, 2014: Systematic satellite observations of the impact of aerosols from passive volcanic degassing on local cloud properties. Atmos. Chem. Phys., 14, 10 601–10 618, https://doi.org/10.5194/acp-14-10601-2014.
Eguchi, K., I. Uno, K. Yumimoto, T. Takemura, T. Y. Nakajima, M. Uematsu, and Z. Liu, 2011: Modulation of cloud droplets and radiation over the North Pacific by sulfate aerosol erupted from Mount Kīlauea. SOLA, 7, 77–80, https://doi.org/10.2151/sola.2011-020.
Elias, T., and A. J. Sutton, 2012: Sulfur dioxide emission rates from Kīlauea volcano, Hawai‘i, 2007–2010. USGS Open-File Rep. 2012-1107, 25 pp., https://doi.org/10.3133/ofr20121107.
Elias, T., A. J. Sutton, J. B. Stokes, and T. J. Casadevall, 1998: Sulfur dioxide emission rates of Kīlauea volcano, Hawaii, 1979-1997. USGS Open-File Rep. 98-462, 41 pp., https://pubs.er.usgs.gov/publication/ofr98462.
Elias, T., C. Kern, K. A. Horton, A. J. Sutton, and H. Garbeil, 2018: Measuring SO2 emission rates at Kīlauea volcano, Hawaii, using an array of upward-looking UV spectrometers, 2014–2017. Front. Earth Sci., 6, 214, https://doi.org/10.3389/feart.2018.00214.
Feng, J., and Y. L. Chen, 2001: Numerical simulations of airflow and cloud distributions over the windward side of the island of Hawaii. Part II: Nocturnal flow regime. Mon. Wea. Rev., 129, 1135–1147, https://doi.org/10.1175/1520-0493(2001)129<1135:NSOAAC>2.0.CO;2.
Frazier, A. G., and T. W. Giambelluca, 2017: Spatial trend analysis of Hawaiian rainfall from 1920 to 2012. Int. J. Climatol., 37, 2522–2531, https://doi.org/10.1002/joc.4862.
Frazier, A. G., T. W. Giambelluca, H. F. Diaz, and H. L. Needham, 2016: Comparison of geostatistical approaches to spatially interpolate month-year rainfall for the Hawaiian Islands. Int. J. Climatol., 36, 1459–1470, https://doi.org/10.1002/joc.4437.
Frazier, A. G., O. Elison Timm, T. W. Giambelluca, and H. F. Diaz, 2018: The influence of ENSO, PDO and PNA on secular rainfall variations in Hawai‘i. Climate Dyn., 51, 2127–2140, https://doi.org/10.1007/s00382-017-4003-4.
Frye, J. L., and Y. L. Chen, 2001: Evolution of downslope flow under strong opposing trade winds and frequent trade-wind rainshowers over the island of Hawaii. Mon. Wea. Rev., 129, 956–977, https://doi.org/10.1175/1520-0493(2001)129<0956:EODFUS>2.0.CO;2.
Garza, J. A., P. S. Chu, C. W. Norton, and T. A. Schroeder, 2012: Changes of the prevailing trade winds over the islands of Hawaii and the North Pacific. J. Geophys. Res., 117, D11109, https://doi.org/10.1029/2011JD016888.
Giambelluca, T. W., Q. Chen, A. G. Frazier, J. P. Price, Y. L. Chen, P. S. Chu, J. K. Eischeid, and D. M. Delparte, 2013: Online Rainfall Atlas of Hawai‘i. Bull. Amer. Meteor. Soc., 94, 313–316, https://doi.org/10.1175/BAMS-D-11-00228.1.
Horton, K. A., G. Williams-Jones, H. Garbeil, T. Elias, A. J. Sutton, P. Mouginis-Mark, J. N. Porter, and S. Clegg, 2006: Real-time measurement of volcanic SO2 emissions: Validation of a new UV correlation spectrometer (FLYSPEC). Bull. Volcanol., 68, 323–327, https://doi.org/10.1007/s00445-005-0014-9.
Huang, Y. F., and Y. L. Chen, 2019: Numerical simulations of seasonal variations of rainfall over the island of Hawaii. J. Appl. Meteor. Climatol., 58, 1219–1232, https://doi.org/10.1175/JAMC-D-18-0078.1.
Iacono, M. J., J. S. Delamere, E. J. Mlawer, M. W. Shephard, S. A. Clough, and W. D. Collins, 2008: Radiative forcing by long-lived greenhouse gases: Calculations with the AER radiative transfer models. J. Geophys. Res., 113, D13103, https://doi.org/10.1029/2008JD009944.
Jiang, H., G. Feingold, and A. Sorooshian, 2010: Effect of aerosol on the susceptibility and efficiency of precipitation in warm trade cumulus clouds. J. Atmos. Sci., 67, 3525–3540, https://doi.org/10.1175/2010JAS3484.1.
Kirshbaum, D. J., and R. B. Smith, 2009: Orographic precipitation in the tropics: Large-eddy simulations and theory. J. Atmos. Sci., 66, 2559–2578, https://doi.org/10.1175/2009JAS2990.1.
Koren, I., G. Dagan, and O. Altaratz, 2014: From aerosol-limited to invigoration of warm convective clouds. Science, 344, 1143–1146, https://doi.org/10.1126/science.1252595.
Kroll, J. H., and Coauthors, 2015: Atmospheric evolution of sulfur emissions from Kīlauea: Real-time measurements of oxidation, dilution, and neutralization within a volcanic plume. Environ. Sci. Technol., 49, 4129–4137, https://doi.org/10.1021/es506119x.
Li, J., and A. D. Heap, 2008: A review of spatial interpolation methods for environmental scientists. Geoscience Australia Doc. 2008/23, 137 pp.
Liu, H., and Coauthors, 2019: Non-monotonic aerosol effect on precipitation in convective clouds over tropical oceans. Sci. Rep., 9, 7809, https://doi.org/10.1038/s41598-019-44284-2.
Longman, R. J., and Coauthors, 2018: Data descriptor: Compilation of climate data from heterogeneous networks across the Hawaiian Islands. Sci. Data, 5, 180012, https://doi.org/10.1038/sdata.2018.12.
Longman, R. J., and Coauthors, 2019: High-resolution gridded daily rainfall and temperature for the Hawaiian Islands (1990–2014). J. Hydrometeor., 20, 489–508, https://doi.org/10.1175/JHM-D-18-0112.1.
Longman, R. J., A. J. Newman, T. W. Giambelluca, and M. Lucas, 2020: Characterizing the uncertainty and assessing the value of gap-filled daily rainfall data in Hawaii. J. Appl. Meteor. Climatol., 59, 1261–1276, https://doi.org/10.1175/JAMC-D-20-0007.1.
Longo, B. M., W. Yang, J. B. Green, F. L. Crosby, and V. L. Crosby, 2010: Acute health effects associated with exposure to volcanic air pollution (vog) from increased activity at Kīlauea volcano in 2008. J. Toxicol. Environ. Health, 73A, 1370–1381, https://doi.org/10.1080/15287394.2010.497440.
Mace, G. G., and A. C. Abernathy, 2016: Observational evidence for aerosol invigoration in shallow cumulus downstream of Mount Kīlauea. Geophys. Res. Lett., 43, 2981–2988, https://doi.org/10.1002/2016GL067830.
Malavelle, F. F., and Coauthors, 2017: Strong constraints on aerosol-cloud interactions from volcanic eruptions. Nature, 546, 485–491, https://doi.org/10.1038/nature22974.
Muhlbauer, A., and U. Lohmann, 2008: Sensitivity studies of the role of aerosols in warm-phase orographic precipitation in different dynamical flow regimes. J. Atmos. Sci., 65, 2522–2542, https://doi.org/10.1175/2007JAS2492.1.
Neal, C. A., S. R. Brantley, L. Antolik, and J. L. Babb, 2019: The 2018 rift eruption and summit collapse of Kīlauea volcano. Science, 363, 367–374, https://doi.org/10.1126/science.aav7046.
Nugent, A. D., C. D. Watson, G. Thompson, and R. B. Smith, 2016: Aerosol impacts on thermally driven orographic convection. J. Atmos. Sci., 73, 3115–3132, https://doi.org/10.1175/JAS-D-15-0320.1.
Pattantyus, A. K., S. Businger, and S. G. Howell, 2018: Review of sulfur dioxide to sulfate aerosol chemistry at Kīlauea volcano, Hawai‘i. Atmos. Environ., 185, 262–271, https://doi.org/10.1016/j.atmosenv.2018.04.055.
Porter, J. N., and Coauthors, 2002: Sun photometer and lidar measurements of the plume from the Hawaii Kīlauea volcano Pu‘u O‘o vent: Aerosol flux and SO2 lifetime. Geophys. Res. Lett., 29, 1783, https://doi.org/10.1029/2002GL014744.
Seo, J. M., H. Lee, S. Moon, and J. J. Baik, 2020: How mountain geometry affects aerosol-cloud-precipitation interactions: Part I. shallow convective clouds. J. Meteor. Soc. Japan, 98, 43–60, https://doi.org/10.2151/jmsj.2020-003.
Skamarock, W. C., and Coauthors, 2008: A description of the Advanced Research WRF version 3. NCAR Tech. Note NCAR/TN-475+STR, 113 pp., http://doi.org/10.5065/D68S4MVH.
Smith, R. B., and V. Grubišić, 1993: Aerial observations of Hawaii’s wake. J. Atmos. Sci., 50, 3728–3750, https://doi.org/10.1175/1520-0469(1993)050<3728:AOOHW>2.0.CO;2.
State of Hawai‘i Department of Health, 2020: Annual summary 2018. State of Hawai‘i Department of Health Rep., 42 pp., https://health.hawaii.gov/cab/files/2020/05/aqbook_2018.pdf.
Stevens, B., and A. Seifert, 2008: Understanding macrophysical outcomes of microphysical choices in simulations of shallow cumulus convection. J. Meteor. Soc. Japan, 86A, 143–162, https://doi.org/10.2151/jmsj.86A.143.
Stevens, B., and G. Feingold, 2009: Untangling aerosol effects on clouds and precipitation in a buffered system. Nature, 461, 607–613, https://doi.org/10.1038/nature08281.
Tam, E., and Coauthors, 2016: Volcanic air pollution over the island of Hawai‘i: Emissions, dispersal, and composition. Association with respiratory symptoms and lung function in Hawai‘i Island school children. Environ. Int., 92–93, 543–552, https://doi.org/10.1016/j.envint.2016.03.025.
Thompson, G., and T. Eidhammer, 2014: A study of aerosol impacts on clouds and precipitation development in a large winter cyclone. J. Atmos. Sci., 71, 3636–3658, https://doi.org/10.1175/JAS-D-13-0305.1.
Twomey, S., 1974: Pollution and the planetary albedo. Atmos. Environ., 8, 1251–1256, https://doi.org/10.1016/0004-6981(74)90004-3.
Warner, J., and S. Twomey, 1967: The production of cloud nuclei by cane fires and the effect on cloud droplet concentration. J. Atmos. Sci., 24, 704–706, https://doi.org/10.1175/1520-0469(1967)024<0704:TPOCNB>2.0.CO;2.
Willmott, C. J., and S. M. Robeson, 1995: Climatologically aided interpolation (CAI) of terrestrial air temperature. Int. J. Climatol., 15, 221–229, https://doi.org/10.1002/joc.3370150207.
World Health Organization, 2006: Air quality guidelines: Global update—2005. WHO Rep., 496 pp.
Xue, H., G. Feingold, and B. Stevens, 2008: Aerosol effects on clouds, precipitation, and the organization of shallow cumulus convection. J. Atmos. Sci., 65, 392–406, https://doi.org/10.1175/2007JAS2428.1.
Yang, Y., and Y. L. Chen, 2003: Circulations and rainfall on the lee side of the island of Hawaii during HaRP. Mon. Wea. Rev., 131, 2525–2542, https://doi.org/10.1175/1520-0493(2003)131<2525:CAROTL>2.0.CO;2.
Yuan, T., L. A. Remer, and H. Yu, 2011: Microphysical, macrophysical and radiative signatures of volcanic aerosols in trade wind cumulus observed by the A-Train. Atmos. Chem. Phys., 11, 7119–7132, https://doi.org/10.5194/acp-11-7119-2011.