The perfect storm: temporal analysis of air during the world’s most deadly epidemic thunderstorm asthma (ETSA) event in Melbourne

Abstract

Background:

There have been 26 epidemic thunderstorm asthma (ETSA) events worldwide, with Melbourne at the epicentre of ETSA with 7 recorded events, and in 2016 experienced the deadliest ETSA event ever recorded. Health services and emergency departments were overwhelmed with thousands requiring medical care for acute asthma and 10 people died.

Objectives:

This multidisciplinary study was conducted across various health and science departments with the aim of improving our collective understanding of the mechanism behind ETSA.

Design:

This study involved time-resolved analysis of atmospheric sampling of the air for pollen and fungal spores, and intact and ruptured pollen compared with different weather parameters, pollution levels and clinical asthma presentations.

Methods:

Time-resolved pollen and fungal spore data collected by Deakin AirWATCH Burwood, underwent 3-h analysis, to better reflect the ‘before’, ‘during’ and ‘after’ ETSA time points, on the days leading up to and following the Melbourne 2016 event. Linear correlations were conducted with atmospheric pollution data provided by the Environment Protection Authority (EPA) of Victoria, weather data sourced from Bureau of Meteorology (BOM) and clinical asthma presentation data from the Victorian Agency for Health Information (VAHI) of Department of Health.

Results:

Counts of ruptured grass pollen grains increased 250% when the thunderstorm outflow reached Burwood. Increased PM10, high relative humidity, decreased temperature and low ozone concentrations observed in the storm outflow were correlated with increased levels of ruptured grass pollen. In particular, high ozone levels observed 6 h prior to this ETSA event may be a critical early indicator of impending ETSA event, since high ozone levels have been linked to increasing pollen allergen content and reducing pollen integrity, which may in turn contribute to enhanced pollen rupture.

Conclusion:

The findings presented in this article highlight the importance of including ruptured pollen and time-resolved analysis to forecast ETSA events and thus save lives.

Introduction

The largest and most devastating epidemic thunderstorm asthma (ETSA) event was recorded on 21 November 2016 in Melbourne, Australia. Thousands of acute respiratory cases and overwhelmed emergency departments following a thunderstorm led to 10 untimely deaths. ¹ The severity of this event served as a wake-up call to address the potential deadly risk of pollen as a trigger for allergic asthma en masse. Despite a number of previous thunderstorm asthma events occurring in Melbourne,^1–3 there has been significant growth in research and health investment since the 2016 ETSA event. We have recently reported the detailed clinical aspects of the 2016 Melbourne ETSA event ¹ and here, we report the detailed time-resolved environmental aspects of the same event with a view to understanding the mechanism(s) that underline ETSA.

Up to 87% of those presented to hospital emergency departments at the eight major Melbourne health services during the 2016 ETSA suffered from allergic rhinitis (hay fever). ⁴ Grass pollen is widely accepted as a major cause of allergic rhinitis, and additionally it is a known trigger of allergic asthma. ² However, pollen from Parietaria species, which is an urticacea weed that is widespread in the Naples and Mediterranean area, has also been implicated in the cause of thunderstorm-related asthma. ⁵ A recent report from the Australasian Society of Clinical Immunology and Allergy (ASCIA) confirms that 50–80% of patients who suffer from asthma also experience allergic rhinitis, while 20–30% of patients who present with allergic rhinitis also suffer from asthma. ⁵ However, some asthma prevalence may not be reported, with the proportion of undiagnosed cases among asthma sufferers estimated to be 20–70%. ⁶

Grass pollen is well accepted as a major trigger of ETSA and has grains that can vary in size from ~35 to 60 μm, making them too large to access the lower airways. However, sub-pollen particles of <5 μm are capable of this where they can trigger an asthmatic response. ⁷ Our pioneering The Lancet paper 3 decades ago in 1992 ⁸ was shortly followed by our observations with the possible aetiology of two consecutive epidemics of asthma, as being linked to the presence of thunderstorms. ⁹ By drawing on experience from previous ETSA events, and in combination with our research on the molecular cloning of grass pollen allergens, led us to successfully demonstrate how grass pollen can rupture to trigger IgE-mediated (allergic immune) responses.^2,8

To date, 26 known cases of ETSA have been documented worldwide since 1983, with 11 occurring in Australia, and 7 in Melbourne alone. ³ Collectively, these events have resulted in over 15,000 individuals suffering from severe asthmatic symptoms with 15 deaths attributed to ETSA events across the world. More than two thirds of reported asthma hospitalisations and deaths occurred in Australia. Research suggests environmental factors such as high concentrations of aeroallergens, primarily ryegrass (Lolium perenne) pollen in southern states of Australia, along with specific weather conditions lead to the rupture of ryegrass pollen grains and subsequent release of fine respirable allergen-bearing starch granules (<2.5 mm), containing the major ryegrass allergen Lol p 5.^2,8,10 Indeed, a recent publication has identified that patients with ETSA are more likely to demonstrate Lol p 5–specific serum IgE. ¹¹ From here, it is thought that thunderstorm outflows then carry these respirable particles down to the ground ready for distribution, and we further hypothesise that a possible vortex mechanism amplifies the allergenic content of the air as it travels across the ground towards highly populated metropolis (Figure 1). This exposes previously sensitised individuals to a sudden increase in the concentration of allergenic sub-pollen particles, consequently triggering allergic asthma^10,12. Indeed, Marks and colleagues in 2001 demonstrated an association between asthma outbreaks and thunderstorm outflows from the Wagga Wagga epidemic of 1997. ¹³

Figure 1.

Thunderstorm asthma. The proposed mechanism of pollen transport and rupturing during a thunderstorm asthma event. It is proposed that grass pollen grains are picked up and carried into the clouds where they rupture due to low temperature and high humidity, following high ozone levels (findings of this study), releasing highly allergenic micronic particles (1). Outflows dump these highly allergenic micronic particles at the ground level, which are concentrated further by our hypothesised vortex action as it travels towards the unsuspecting susceptible population (2).

In response to the 2016 event, the Victorian Government funded the establishment of further air monitoring sites across Victoria. Known as the Victorian Thunderstorm Asthma Pollen Surveillance (VicTAPS) network, this system measures atmospheric pollen and spore concentrations, providing data previously lacking for much of regional Victoria. The Deakin AirWATCH Burwood spore trap that captured the time-resolved and analysed 2016 ETSA data presented here, is located within the local government municipality of Whitehorse, which is characterised by 18.4% tree canopy coverage. Grass is grown extensively across verges, parklands, urban gardens and along waterways and amounts to approximately 27.9 of vegetation cover in the eastern regions of Melbourne. ¹⁴ The Deakin AirWATCH network, along with other monitoring sites, not only benefits the public with their allergen avoidance programmes but can also directly contributes to clinical studies and inform significant research.

Our understanding of thunderstorm asthma has improved dramatically since the 2016 ETSA event, specifically relating to susceptibility, population risk, alleviation strategies and its effects on the health services.^1,3,15 However, we still have major gaps in understanding of the meteorological and environmental trigger factors that are critical in wholly understanding the mechanism of ETSA.

Thus, in this article, we aim to link the much-needed data between grass pollen, environmental and meteorological conditions and asthma emergency presentations during the 2016 thunderstorm asthma epidemic that devastated Melbourne. Specifically, this article, for the first time, couples 3-h time-resolved pollen intelligence obtained from Deakin AirWATCH pollen counting and forecasting facility ¹⁶ with meteorological data obtained from the Bureau of Meteorology (BOM), environmental data obtained from the Environment Protection Authority (EPA) of Victoria and asthma hospital emergency department presentations documented by the Victorian Government’s Department of Health, and so improves our understanding of thunderstorm asthma. This is high-impact multidisciplinary research since the occurrence of thunderstorm-associated allergic asthma is estimated to occur twice every 3 years. ¹⁷ This, in combination with the effects of climate change increasing pollen production, ¹⁸ more common occurrence of thunderstorms ¹⁹ and the high proportion of at-risk individuals, ²⁰ is highly likely to make ETSA an emerging and continued health threat.

Materials and methods

Atmospheric sampling

Air sampling was conducted continuously throughout the allergy season at the Deakin AirWATCH site at Burwood using a Burkard pollen and spore trap (Burkard Manufacturing Co. Ltd., UK). The sampler is designed to imitate the volume of air inhaled by the human lung, drawing in atmospheric air at a rate of 10 to 11 l/min. A glycerol adhesive 7-day tape was placed on a rotating drum to trap and hold the pollen, allowing for the 3-h analysis over the more conventional 24-h sampling methods. ¹⁶ The pollen was analysed and presented in 3-h segments to coincide with 3 critical time points of ‘before’ the thunderstorm event (i.e. 3–6 p.m.), ‘during’ the thunderstorm event (i.e. 6–9 p.m.) and ‘after’ the thunderstorm event (i.e. 9–12 a.m.). Samples were taken weekly (i.e. 7 days) over the pollen season; however, this article will only focus on the dates leading up to and following the Melbourne ETSA event (i.e. 19–27 November 2016). The glycerol tapes enclosing the trapped pollen and spore samples were mounted onto glass slides, covered with a glass coverslip and analysed using an Olympus BX50 light microscope at a magnification of 400× (40× objective by 10× ocular lens), according to our established methods. ¹⁹ The above protocols were conducted as per the Australian Airborne Pollen and Spore Monitoring Network Interim Standard and Protocols. ²¹ A pollen atlas compiled by Deakin AirWATCH over the extensive past history of pollen monitoring was used as a reference to identify and tally grass (both ruptured and intact), tree and weed pollen as well as yeast and fungal spores. Since the critically important starch granules cannot be identified by the pollen counter, the ruptured pollen served as the surrogate. The total concentration of pollen grains and fungal spores was then expressed per m³ of air over the 3-h time points.

Asthma presentations

Asthma presentation data were sourced directly from the Victorian Agency for Health Information (VAHI) via collaboration with Department of Health to analyse the association between airborne grass pollen and respiratory health during thunderstorm events. Human ethics exemption was obtained from Deakin University (2020-392: That Melbourne day: time resolved analysis of air during the world’s deadliest epidemic thunderstorm asthma) and approved due to the pre-existing and non-identifiable nature of the data set. Recorded hospital presentations obtained were time-stamped data from five emergency departments across metropolitan Melbourne. Namely, this included Austin Hospital (about 10 km north-east from the city of Melbourne), Box Hill Hospital (about 15 km east from the city of Melbourne), Maroondah Hospital (about 30 km east from the city of Melbourne), Monash Medical Centre (about 20 km south-east from the city of Melbourne) and Royal Melbourne Hospital (about 2 km north from the city of Melbourne). Patient presentations were categorised under two codes: J450 (childhood asthma) and J459 (general asthma). As hospitals may use the abovementioned codes interchangeably or for additional classifications, these categories were not analysed separately from one another and instead combined to properly correlate with environmental parameters.

Meteorological data

Time-resolved 3-h weather data were sourced via collaboration from the BOM to observe the potential impact on atmospheric concentration of airborne allergens. Measurements from the Viewbank station were obtained in 30-min increments recorded over a 24-h period. Parameters analysed include precipitation (mm), air temperature (°C), dew point temperature (°C), relative humidity (%), wind speed (km/h), wind direction (o), maximum wind gusts (km/h) and barometric air pressure (hPa).

The Bureau of Meteorology Storm Confirmation Tool was used to confirm the presence or absence of thunderstorms on comparative dates in November that incurred the poorest air quality (i.e. 14 November 2017, 2 November 2018 and 21 November 2019). Thunderstorms were categorised as the presence of thunderstorms (lightning) and/or strong winds (averaging from 26 knots and up to 33 knots) observed within 50 km of the Melbourne suburb of Burwood.

Environmental data

Atmospheric pollution data in 1-h increments were sourced from the Victorian Environment Protection Authority (EPA) to determine the possible association on aeroallergen levels and conditions. Time-resolved 3-h parameters analysed from the Alphington station include particulate matter (PM10), nitrogen dioxide (NO₂), ozone (O₃) and the Air Pollution Index (API).

Statistical analysis

Bivariate Pearson’s correlation was performed using the statistical software package SPSS 27. Linear correlations were performed between total asthma presentations (including a time delay of 3 h), individual hospital presentations (including a time delay of 3 h), individual weather/environmental parameters and concentrations of total pollen, intact grass pollen and ruptured grass pollen. Correlations were also investigated between weather parameters and environmental factors and the grass pollen rupturing ratio, which represented the proportion of ruptured grass as a proportion of total pollen.

Results

The perfect storm

The 21 November 2016 was the hottest day of the new summer season that year in Victoria with temperatures peaking at 40–44°C around greater Melbourne. The wind reached a peak speed of 40 km/h prior to 3 p.m. Australian Eastern Daylight Time (AEDT). A storm front developed across a north–south line west of Melbourne and Geelong. The arrival of the thunderstorm outflow at approximately 6:18 p.m. AEDT in Burwood, found wind speed in excess of 60 km/h, and coincided with a peak in airborne particles, giving rise to higher pollen and spore concentrations compared with the time points ‘before’ and ‘after’ the event (Figures 2 and 3). These winds held steady at 60 km/h for approximately 12 min, indicating that the storm outflow was close to 12 kilometres wide. The storm front continued eastward through greater Melbourne between 6 p.m. AEDT and 6:30 p.m. AEDT. The storm outflow coincided with an elevated dust event as the concentration of particle matter (PM10) exceeded the associated air quality guidelines, and this was visible to the naked eye on the adhesive tape as indicated in Figure 3(a). After the passage of maximum wind gusts, wind speed dropped by roughly 50% while varying in direction, temperature rapidly decreased by 10°C, and relative humidity increased by 70–80%. Within minutes after the passing of the thunderstorm, meteorological data showed the beginning of heavy rainfall (1–4 mm), along with greatly diminished wind speeds and a reduction in grass pollen concentrations and PM10 levels.

Figure 2.

Daily total grass and fungal spore concentrations for the days prior to and after the Melbourne 21 November 2016 ETSA event. Daily atmospheric pollen and fungal spores collected on an adhesive tape via a Burkard pollen and spore trap and identified via microscopy. Individual grass pollen grains (intact and ruptured) and fungal spores (yeasts or other) expressed as grains/spores per m³ of air.

ETSA, epidemic thunderstorm asthma.

Figure 3.

Time-resolved analysis of airborne pollen concentrations during a significant ETSA event: (a) Adhesive tape collected by the Deakin AirWATCH Burwood pollen monitoring site containing the atmospheric particles from 3 p.m. 20 November to 3 a.m. 22 November. Atmospheric pollen levels during this time period are depicted in (b) as 3-h time points, as well as averages over the entire 36-h (i.e. 12× 3-h) period. Time points flagged within the tape (i, ii, iii, iv) indicate the changes in atmospheric particles during the ETSA event. These time points are depicted in visual detail by representative light micrographs shown in Ci-iv. Specifically, image of particles trapped and collected by the Burkhard sampler at 4 p.m. AEDT on 21 November 2016 used to visualise intact grass pollen (A), Cladosporium (C) and smut (fungal) teliospores (S) (ci). Image of samples trapped and collected during the peak 6 p.m. AEDT in atmospheric particles correlating with the storm outflow. Ruptured grass pollen (double arrows ↑↑) and intact grass pollen grains (single arrow ↑) (cii). Microscope image of atmospheric particles and fungal spores collected just after 7 p.m. showing smut teliospores (S) among unidentifiable debris (ciii). Microscope image of air sample just after 8 p.m. Leptosphaeria ascospores (single arrow ↑), algal cells (green) and yeast-like conidia (double arrows ↑↑) (civ).

ETSA, epidemic thunderstorm asthma.

Dissecting the storm

Pollen monitoring is most commonly conducted over a 24-h period, which are averaged over 24 h to give a forecast in grains/m³ for allergy sufferers. This, therefore, means that the average concentration of airborne pollen and fungal spores may not reflect any sudden increases caused by sudden changes in environmental conditions; for example, strong winds prior to a thunderstorm. The average airborne pollen concentration during the 24 h that included the 21 November 2016, storm event, was classified as extreme according to Victorian classification system (i.e. 100 + intact grass pollen per m³). This included on average 210 intact grass pollen grains, 86 ruptured grass pollen grains and 212 other pollen grains per m³ of air over 24 h; a high pollen level (Figure 2), which is not unusual for mid-November. This was surprising because the majority of individuals who were affected during the ETSA event were allergic to grass pollen but the pollen concentrations were not out of the ordinary for November. Also, these data did not accurately reflect the peak in atmospheric particles observed by the EPA, nor the sudden and unprecedented rise in acute asthma across the community. Therefore, a more appropriate time-resolved analysis of both intact and ruptured grass and other pollen levels to observe changes temporally related to the storm passage was required to test consistency with the hypothesised mechanism. Using this finer temporal resolution (Figure 3), airborne pollen levels increased significantly in a step wise manner from low levels in all 3 h time blocks over the preceding 24 h, to a sudden rise in the 3 h just preceding the storm and an even higher level reached in the 3 h, which included the storm. Thereafter, pollen and spore levels were almost undetectable.

Atmospheric samples immediately prior to the storm on the 21 November between 15:00 and 18:00 AEDT included low levels of fungal spores, 207 intact grass pollen, 113 ruptured grass pollen and 92 other pollen types per m³ of air (see Figure 3(b)). This was followed by a near 250% increase in ruptured grass pollen as the outflows of the thunderstorm began at 6:18 p.m. AEDT. The 3-h period involving the thunderstorm (i.e. time point 6–9 p.m.) found 264 intact grass pollen, 279 ruptured grass pollen and total of 402 other types of pollen per m³ of air. This recorded level exceeded any previous data, including those under past extreme conditions, with numbers of this magnitude having not previously been seen in Burwood’s air sampling history. Results presented very dense and complex light micrographs for counting (see Figure 3(cii)). It should be noted here that this is the first time that such temporal scales have been undertaken in Australia and as such, may contribute to improved future technologies and methodologies for better understanding of this phenomenon.

Pollen levels were quickly reduced as rain arrived within the hour. This was visible both with the naked eye (Figure 3(a)) and microscopically (Figure 3(ciii)). The absence of larger particles and pollen is depicted in Figure 3(ciii), with smut and fungal spores dominating the atmosphere. This is consistent with these spores having arrived at the Burwood sampler towards the end of the outflow when raindrops were selectively scrubbing large particles from the atmosphere. After 8 p.m. AEDT, and through to the morning of 22 November, no further pollen grains and very few giant particles were observed (Figure 3(civ)). Very few microsphaeropsis spores were detected, and the air was dominated by Leptosphaeria ascospores and hyaline spores similar to yeast blastoconidia. These are typical wet weather fungi likely emitted from local Eucalyptus trees and grasses (Figure 2).

Time-resolved asthma presentation analysis

A statistically significant correlation between total asthma presentations and the abundance of: ruptured grass pollen (r = 0.747, p = 0.008), intact grass pollen (r = 0.686, p = 0.02) and other pollen (r = 0.733, p = 0.010) concentrations was observed after incorporating a time delay of 3 h [for a visual depiction of temporal trends, see Figure 4(a)]. When focusing on individual hospital presentations (Figure 4(b)), a variation in asthma presentations at different hospitals was evident (Figure 4(c) and (d)). Of significance, asthma presentations at the Austin hospital correlated with the notable increase in airborne pollen concentrations (ruptured grass: r = 0.854, p = 0.001; intact grass: r = 0.755, p = 0.007; other pollen: r = 0.861, p = 0.00). In addition, asthma presentations at the Royal Melbourne Hospital were also significantly correlated with the notable increase in airborne pollen concentrations (ruptured grass: r = 0.832, p = 0.001; intact grass: r = 0.811, p = 0.002; other pollen; r = 0.793, p = 0.004). However, asthma presentations at Maroondah Hospital showed no significant correlation with asthma presentations and pollen concentrations (Figure 4).

Figure 4.

Asthma presentation to emergency departments during the 21 November 2016 ETSA event. Time-resolved 3-h total asthma presentations and airborne pollen levels just prior (3–4 p.m.), during (6–9 p.m.) and after (9–12 a.m.) the thunderstorm (a). A time delay of 3-h was used for the asthma presentations, to coincide with the 3-h pollen data. Geographical location of hospitals (1: Royal Melbourne Hospital, 2: Austin Hospital, 3: Box Hill Hospital, 4: Maroondah Hospital, 5: Monash Medical Centre) and storm direction (b), and time-resolved 3-h total asthma presentations of 5 Melbourne hospitals during the 21 November ETSA event (c). Daily asthma presentations, by hospital for the 3 days prior and 6 days after the ETSA event (d).

ETSA, epidemic thunderstorm asthma.

Time-resolved weather and environmental analysis

A 3-h time-resolved analysis of the conditions for the 24 h prior to the storm, and immediately after, was performed to identify any specific conditions that would contribute to the observed pollen concentrations and resulting asthma presentations. As expected, changes in atmospheric pollen levels were associated with weather conditions (Figure 5). Specifically, a rapid increase in the atmospheric concentration of ruptured grass pollen and other pollen was associated with a rapid rise in humidity (Figure 5(a)). Other atmospheric conditions such as air temperature demonstrated a negative association with overall concentrations of airborne pollen (Figure 5(b)). In contrast, dew point temperature exhibited a positive association (Figure 5(c)) with all levels of pollen (ruptured grass: r = 0.182, p = 0.009; intact grass: r = 0.591, p = 0.043; other pollen; r = 0.740, p = 0.006). Finally, although a positive association between all pollen types and precipitation was seen during the event, this association disappeared once the event ended, with the increasing precipitation levels (Figure 5(d)) indicating that airborne pollen was being largely washed out following the thunderstorm.

Figure 5.

Time-resolved pollen and weather conditions. Time-resolved 3-h analysis of total intact grass pollen, ruptured grass pollen and other pollen alongside various weather conditions (mean ± SD), relative humidity in % (a), air temperature in °C (b), dew point in °C (c) and precipitation in mm (d).

Air pollution has been previously observed to alter pollen allergenicity. ²² Similarly, we sought to identify any associations with airborne pollen concentrations and environmental pollutants during the 21 November 2016 ETSA event. A significant increase in PM10 at the same time was observed with increase in pollen concentrations during the arrival of the storm front [ruptured grass: r = 0.798, p = 0.002; intact grass: r = 0.850, p = 0.000; other pollen: r = 0.739, p = 0.006; Figure 6(a)]. A similar trend was observed for API, as expected (Figure 6(b)). In contrast, a sudden increase in ozone levels 6 h prior to the event was then negatively associated during the event (Figure 6(c)). Finally, a positive association was observed with the concentration of ruptured grass pollen and an increase in nitrogen dioxide levels [r = 0.636, p = 0.026; Figure 6(d)].

Figure 6.

Time-resolved pollen and environmental pollutants. Time-resolved 3-h analysis of total intact grass pollen, ruptured grass pollen and other pollen alongside various environmental conditions (mean ± SD), including Pm10 in μg/m³ (a), API (b), ozone in parts per billion (ppb) (c) and nitrogen dioxide (NO₂) in ppb (d).

API, Air Pollution Index.

A comparison was performed with increase in PM10 levels to discount the possibility that the increase in asthma presentations was not due to the rapid non-specific decrease in air quality accordingly (Figure 7). Historical pollution data for the three following years including a mid-November day with the highest PM10 recording were analysed (Figure 7). This demonstrated that even when PM10 is much higher than that present during the ETSA event of 21 November 2016 (e.g. 21 November 2019, when a significant dust storm also occurred in Melbourne), this did not result in significant increases in hospital respiratory admissions (Figure 7). Consequently, the major difference in respiratory admissions can be associated with the varying loads of airborne pollen concentrations, with the levels 10 times higher during the ETSA event.

Figure 7.

Year on year comparison. For comparison, the day with the highest PM10 in November was selected for comparison to the 21 November 2016 ETSA event, to discount high PM10 as the sole contributor of asthma presentations to the emergency departments. Daily total PM10 (μg/m³), percent of ruptured or intact grass pollen, total asthma presentations as well as the presence of storm activity for the 21 November 2016 and comparative high PM10 days: 14 November 2017, 2 November 2018 and 21 November 2019. It should be noted that 232 asthma presentations on 21 November 2016 was only for that specific day. In addition, there were 449 hospital admissions in the 3 days following the 21 November ETSA event.

ETSA, epidemic thunderstorm asthma.

Factors affecting grass pollen rupture

Given that the mechanism behind allergen-induced thunderstorm asthma is thought to be related to inhalation of grass pollen allergens, we explored the weather conditions and environmental pollutants that may have bearing on the pollen rupturing ratio. That is, the proportion of grass pollen that was ruptured, where a high rupture ratio indicates higher airborne allergen loads as micron-sized particles compared with a lower pollen rupture ratio, and lower airborne allergen levels as micronic particles. The ratio of rupturing grass pollen was used to better observe how environmental parameters influence this mechanism; absolute numbers may not reveal the underlying factors that increase rupturing rates. Interestingly, temperature was significantly negatively correlated with rupture ratio (Figure 8(b)). In contrast, the rupture ratio was likely to increase significantly with an increase in humidity (Figure 8(e)), up until the arrival of rain. Rain was observed to scrub all large particles, including pollen, from the air. Dew point, precipitation and air pressure were not correlated with changes in pollen rupture (Figure 8(a), (c) and (d), respectively).

Figure 8.

Grass rupturing under specific weather conditions. Relationship between grass rupturing ratio (ruptured grass/total grass) and various weather parameters, including dew point (a), air temperature in °C (b), precipitation in mm (c), air pressure in Pa (d) and relative humidity % (e). Pearson’s (r) and significance (p) for each parameter shown, respectively.

The pollen rupture ratio was then compared with airborne contaminants, as previous research indicated that pollen wall integrity is affected by environmental pollutants. ²³ Unexpectedly, ozone concentration was found to be negatively correlated with pollen rupture (Figure 9(c)), where pollen was observed to be less likely to be ruptured in higher levels of ozone. No significant correlations were observed between the ratio of pollen rupture and PM10, API and NO₂ (Figure 9(a), (b) and (d), respectively).

Figure 9.

Grass rupturing in the presence of environmental pollutants. Relationship between grass rupturing ratio (ruptured grass/total grass) and various environmental conditions, including PM10 in μg/m³ (a), API (b), ozone in ppb (c) and nitrogen dioxide in ppb (d). Pearson’s (r) and significance (p) for each parameter shown, respectively.

API, Air Pollution Index.

Discussion

An unprecedented event

The rapid onset of the 2016 Melbourne epidemic thunderstorm asthma event, and the scale of its consequences were unprecedented. ²⁴ Emergency service calls reporting asthma attacks or breathing problems began at 6 p.m. AEDT and quickly escalated. In one 15-min period from 7 p.m. AEDT, 201 emergency ambulance calls were received, and in the 12-h period following the storm, 2332 emergency ambulance calls were answered, which was unprecedented at the time. The numbers surpassed call surges experienced during the 2009 Black Saturday bushfires. ²⁴ Also, from the hours of 7 p.m. to 8 p.m. AEDT alone, emergency departments were faced with over two times the caseload experienced over a 2-year period at the 99th centile. ²⁵ Thousands of Victorians were affected and 10 people died. ²⁶

The dramatic and sudden increase in cases of acute respiratory distress across emergency departments in Melbourne observed from 6 p.m. AEDT (21 November) to midnight (22 November) AEDT is consistent with our previous understanding regarding the mechanism of ETSA, that a large sudden spike in pollen is associated with a spike in individuals presenting to public hospitals with respiratory symptoms (Figure 4). This was captured as a result of the time-resolved 3-h analysis employed in this study. These findings emphasise the importance of gathering time-resolved data to adequately capture an ETSA event, such as the data gathered in this study (Figure 3), or as provided by automated pollen counters, when these become more widely validated and accepted.

Dissecting the storm

The scale of the events of 21 November 2016 was unprecedented. It surpassed comparison data from the three previously reported Melbourne thunderstorm asthma events of 20 November 2003, 25 November 2010 and 8 November 2011 and exceeded all previous data from prior years.^3,27 The total number of people affected by the Melbourne 2016 event was greater than all previously reported worldwide ETSA events combined.

Recent mechanistic analysis indicates that convergence lines, usually present in thunderstorm outflow, probably accumulate pollen grains in the clouds via air flows, and subsequent nucleated condensation then causes rupture by what is often termed ‘osmotic shock’, leading to release of smaller respirable allergens. ²⁸ We have further hypothesised that these particles (respirable allergens) would be released into the microdroplets of liquid that accumulates around the pollen grain. Drying downdrafts are then required to disperse these particles from the pollen shell. The precise mechanisms of such dynamic processes remain unclear, including how agglomerated cohesive starch-based granules will disperse into respirable sizes. Downdrafts would carry these released respirable allergens and pollen fragments to ground level, entrained in the outflows.

We further hypothesise that allergenic particle abundance may be concentrated at ground level by a potential vortex mechanism generated by and within the thunderstorm downdrafts and outflows, re-entraining the outflow back into the cloud base (Figure 1). In this process, strong electric fields similar to those during thunderstorms have also been observed to speed up the pollen rupture process when applied in the laboratory.^29,30 These respirable allergens could then be inhaled and penetrate into the lower airways, where they can trigger allergic asthma. Further detailed study will help confirm the validity of such hypothesised processes.

Following the 2016 event, there was an increased imperative to provide people with allergy and asthma with precise, relevant and localised information regarding the pollen concentrations in order to be appropriately prepared and to better manage their symptoms day to day.

From the collected data, it is estimated that levels over 1000/m³ ruptured grass pollen grains were present as a result of the sudden spike in grass pollen concentration (Figure 3(b)), and by applying the proposed process of pollen rupture during a thunderstorm described by Taylor and Jonsson in 2004, ¹² consequently up to millions of allergy-causing starch granules per m³ would have been combined with the storm outflow, in a process lasting close to 12 min. It is hypothesised that the impact of such pollen rupture at unprecedented pollen levels was consequently observed at Austin Hospital and Royal Melbourne Hospital, where there were strong significant correlations found between grass pollen concentrations and acute asthma presentations. This was likely due to the geographical location, with these emergency departments receiving patients from regions that experienced the poorest air quality with potentially such high concentrations of highly allergenic grass pollen micron-sized fragment particles during the storm. Also, Austin Hospital and Royal Melbourne Hospital are located closer to the CBD, which has a high population density, compared with other emergency departments analysed for this study that are further east.

While a significant correlation was not observed between grass pollen and patients admitted to Maroondah Hospital, a significant link was determined between asthma presentations and the rupturing ratio. Arguably unexpected, this could also be explained by geographical location. Maroondah Hospital is the furthest of the emergency departments we analysed from the Melbourne CBD, approximately 30 km east (Figure 4(b)). Intact pollen grains may not have been carried that extra distance by the storm flow, to reach these eastern suburbs, but as ruptured pollen and sub-pollen particulates are smaller and lighter, they could have been carried further by wind and thus able to induce severe asthma symptoms.

The data presented here, for the first time, support our understanding about these ETSA mechanisms. Specifically, that an epidemic thunderstorm asthma event is likely due to large number of people being exposed at the same time by a meso-scale climatic event resulting in short-term exposure to highly concentrated allergenic micronic particles from ruptured ryegrass pollen, by sensitised individuals.

The infrequent nature of ETSA events makes it critical to capture as much and as detailed environmental data as possible in order to test current hypothesis relating to its mechanism.

At present the forecasting of ETSA is based on total grass pollen, of which ryegrass is a component. It is a surrogate measure, as is total ruptured pollen concentration of what we understand to be the actual trigger; the allergen (i.e. Lol p 5) released from its usual intracellular containment and located on micronic particles.

Although our current pollen intelligence has improved significantly since the 2016 Melbourne ETSA event, leading to novel forecasting and dispersion machine learning models,^17,20,31 our current early warning system relies on pollen information that takes 24 h to capture, analyse and share with modelling systems for ETSA forecasting. The latency of this information does not help discern if an event is occurring due to the dynamic, proximal and rapid onset of ETSA. Furthermore, the current system detects the whole pollen, not the respirable allergens as micron-sized particles that trigger ETSA, making it a sub-optimal system. Thus, our current air surveillance systems need to be improved to detect 1 to 3 h or closer to ‘real-time’ (particularly on days of potential high ETSA risks) to detect not only whole pollen but also ruptured pollen and their highly allergenic micronic particles. By focusing on the rupturing mechanism and the actual allergens involved, we can make improvements to the performance of ETSA forecasting models to develop more accurate early warning systems.

Weather conditions

The Spring of 2016 was Victoria’s 10th wettest on record; however, rainfall varied throughout the season with record-breaking highs in September, while November being drier than average lead to uncommonly high pasture growth. ³²

High relative humidity was found to coincide with an increasing ratio of rupturing pollen grains. However, a recent study has used modelling systems to investigate this potential mechanism for pollen rupture but indicated that relative humidity alone was not adequate for pollen rupturing during the 2016 Melbourne ETSA event. Instead, a number of mechanisms likely operated in tandem with each other. Lightning strikes, electrical build up, and discharge would have incurred in conditions of low relative humidity and mechanical friction from wind gusts ³³ and interactions between these factors could be the cause of pollen rupture. Examining actual pollen and meteorological data in the current study supported that this mechanism may have occurred during the 2016 Melbourne ETSA event to induce the high rates of pollen rupture we observed (Figure 5(a)). We did not have access to lightning strike data and recommend this should be part of future studies to scrutinise and help validate our proposed pollen rupture mechanisms further, by testing all of these thunderstorm-specific environmental factors, perhaps in a simulated environment, in isolation and in different combinations. Furthermore, as mentioned earlier, there are many factors that occur during a thunderstorm event with the potential to rupture pollen grains and further research is needed to recognise the actual factors that contribute to storm-induced rupturing.

Environmental conditions

While several environmental factors prior to and during the epidemic event were extreme, these conditions may not be sufficient to account for the gravity and enormity of the thunderstorm asthma epidemic of 2016. ³⁴ For this specific reason, and to exclude high PM10 as the main trigger of increased asthma hospital presentations, a rudimentary comparison was performed identifying the potential correlations. This highlighted the unique conditions required for ETSA, excluding high PM10 as a key ETSA trigger (Figure 7).

Ozone was observed to negatively influence rupturing of airborne pollen concentrations during the thunderstorm. Previous studies have linked high ozone to increased pollen allergenicity but also suggested a potential detrimental effect on pollen wall viability.^23,35,36 Here, we present the hypothesis that elevated levels of ozone prior to the storm contribute to increasing grass allergen potency via this wall viability effect, reducing the integrity of the grass pollen and facilitating enhanced pollen rupture. This key proposition could have a significant impact as factors involved in an ETSA event and would benefit greatly from further research investigating the impact of ozone on the rupturing mechanism during ETSA events compared with standard thunderstorms.

Further, we examined EPA data for the month of November of previous six Melbourne ETSA events and have thus uncovered the occurrence of high, and often multiple, ozone spikes in the hours prior to ETSA events (Figure 10). While most of these values were less than 50 ppb and were not high enough to be considered hazardous levels, ³⁷ they were noticeably high levels compared with other readings recorded in the hours leading up to each ETSA event. This finding is significant, which justifies further research and modelling to include it as a potentially critical environmental indicator for an impending ETSA event. Thus, exposure to high ozone levels on days of high to extreme grass pollen concentrations prior to thunderstorms may be the ‘missing link’ for ETSA events, which can be used in future ETSA forecasting and warning systems and thus deserves further research to determine the relationship between ozone and pollen rupturing.

Figure 10.

Time-resolved ozone levels in previous ETSA events in Melbourne. Time-resolved 3-h (mean ± SD) ozone levels (in ppb) for the hours prior to and after the previous six recorded ETSA events [3] on 11 November 1984 (a), 8 November 1987 (b), 29 November 1989 (c), 19 November 2003 (d), 25 November 2010 (e) and 8 November 2011 (f). Arrows indicate the reported arrival of the ETSA event with sudden increase in acute asthma presentations.

ETSA, epidemic thunderstorm asthma.

Improving ETSA preparedness

The concurrence of extreme weather conditions and environmental factors gave rise to unparalleled epidemic thunderstorm asthma faced by Melbournians on 21 November 2016. The prevalence of asthma has rapidly increased since 2001 as a result of innumerable changes in living conditions. ³⁸ The Intergovernmental Panel on Climate Change projected the increase in greenhouse gases such as nitrogen dioxide, carbon dioxide and methane, which will result in a rise in global temperatures leading to recurrent severe weather phenomena. ³⁹ The effect of global warming is linked to increases in the frequency and severity of not only dangerous snow falls, heavy rains, hurricanes, tornadoes and storms but also a rise in the abundance of aeroallergens such as pollen. ³⁰

As evidence of asthma epidemics following thunderstorms continues to increase, ETSA is now more readily recognised due to increased demand on emergency departments as well as increased research into monitoring aeroallergens and developing ETSA forecast modelling. As temperatures increase globally, the warmer air carries added moisture resulting in heavy rainfalls, increasing storm events and consequently epidemic thunderstorm asthma events. ⁴⁰

Changing landscape’s changing health

The gravity of the threat of thunderstorm asthma was first recognised over 35 years ago, so why did the 2016 event have such a sizable impact? First, Victoria experienced exceptionally wet weather in winter and spring of 2016, encouraging high levels of growth. ⁴¹ A high grass pollen season was forecast ⁴² and therefore high-risk of severe allergic rhinitis.

The population of Melbourne has also greatly increased in the past 30 years by close to 2 million. ⁴³ Ongoing encroachment into land previously used as pastures to develop housing for the increasing population density, and the subsequent transformation of pastures to allergenic urban grasses, has changed management and grazing practices as well as exacerbating flowering and pollen emission.^21,44 The abrupt change from a very dry spring in 2015 to a very wet spring in 2016 altered farming practices resulting in an abundance of ungrazed grass. ³² Concurrently, urban grass was subjected to selective breeding over the past 30 years in order to improve their performance, consequently yielding very rapid grass growth during heavy periods of rainfall.

Melbourne’s particularly variable weather provides more opportunities for pollen rupture from species that originate in much more stable temperatures of the northern hemisphere. Although daily grass pollen concentrations prior to 21 November 2016 remained close to average, the days of extreme counts earlier that month are likely to have contributed to the epidemic by potentially priming sensitised individuals. ⁴⁵ These numerous changes over the past three decades prompted the wider spread of one of the most allergenic plant species across one of the most variable weather zones during a time when many people were outside and at full risk of exposure.

Recommendations going forward

In order to test the hypothesis behind thunderstorm asthma mechanism and the data presented here, further research is highlighted and needed. This article highlights the critical data provided by our time-resolved 3-h analysis of pollen data, which detailed the various environmental parameters during the thunderstorm. This was used, instead of the routine 24-h monitoring, providing 8× greater resolution. The 24-h monitoring can hide or ‘dilute’ significant factors. Such analysis has not been previously conducted during a thunderstorm asthma epidemic and thus this work provides vital information regarding not only the spike in atmospheric particles but also the sudden drop following heavy rain.

As the storm outflow only lasted approximately 12 min, 24-h averaged surveillance data would have overlooked this detail, of a rapid increase and decrease as visualised in Figures 3 and 4. This is the first time that time-resolved data of aeroallergens was available in combination with the overwhelming impact witnessed on the emergency health services, and highlights the need for improved standardised monitoring.

This includes the necessity for development and implementation of modernised ‘real-time’ pollen and spore traps allowing for improved time-resolved sampling in, preferably, 10-min increments. In addition, establishment of such ‘real-time’ spore traps in the direction of storm approach (i.e. in Victoria such storms usually originate from north-west and move across the state towards the east) would provide an early warning of approaching aeroallergens outflow. Specialist samplers could be operational during thunderstorm events to collect samples across defined aerodynamic diameters (cut-points). A series of modern analysis methods such as Polymerase Chain Reaction (PCR), next-generation high-throughput sequencing, allergen quantitation by mass spectroscopy as well as allergen-specific immunoassays could be utilised to measure respirable allergen particles. Further, laboratory-based research is also required in order to replicate and verify the hypothesised conditions that trigger pollen rupturing. Finally, forecasts, modelling and real-time estimates for both intact and ruptured grass pollen are vital. Current pollen counts do not differentiate between intact and ruptured pollen grains, and since pollen rupture is pivotal for causing exacerbations in allergic asthmatics,^7,8 more effort should be directed for counting ruptured pollen as well. Recording and utilising ruptured pollen counts at pollen monitoring stations would improve the accuracy of current modelling systems used to forecast ETSA and give a better understanding on how rupturing contributes to these epidemics.

The data correlated during this epidemic emphasise the importance of providing real-time alert information to the public. Modernised ETSA forecasting will improve the quality of life of at-risk individuals and lower the medical and socio-economic burden caused by such asthma epidemics, and ultimately save lives.

Conclusion

In conclusion, despite the tragic events, as a result of converging environmental and patient factors, the thunderstorm event of November 2016 led to establishing a new standard for emergency health responses during ETSA. We have demonstrated the value of more granular temporal analysis of environmental data and time-resolved atmospheric data also began a new wave of interest and importance of standardised pollen monitoring. Indeed, our time-resolved data analyses presented here indicate the potential critical roles that high pollen, relative humidity, temperature and ozone levels (tested in isolation or in combination) played in pollen rupture during the 2016 Melbourne ETSA event, which should be the focus of future research. The conditions faced on this fateful day highlight the importance of better understanding this phenomenon, expanding the focus of current pollen monitoring, and optimising ETSA forecasting to predict future events and protect at-risk patients.