Introduction

In south China, concentrations of fine particulate matter (PM2.5) are generally low under southerly winds, which bring clean air from the South China Sea (SCS)1,2. This is illustrated by pollution rose using measurement data from Hong Kong (Supplementary Fig. S1). The predominant winds in Hong Kong were steady southerly from March 31 to April 4, 2024 (Fig. 1a). However, visibility declined and reached the lowest level of 4.4 km overnight between April 1 and 2 (Fig. 1b). Meanwhile, the PM2.5 concentration increased and peaked at 38 µg/m³ at 7 pm on April 1 (Fig. 1b). The minimum visibility level was lower than the 1st percentile (i.e., 6.5 km), while the maximum PM2.5 level exceeded the 99th percentile (i.e., 34.5 µg/m³) calculated from historical data under the southerly winds from 2021 to 2023. Additionally, the increase in PM2.5 levels propagated northwards across Hong Kong and its adjacent Pearl River Delta (PRD) region on April 1 (Fig. 1c). The deterioration of air quality and visibility suggests the presence of an unusual source or transport mechanism that overrode the typical clean air conditions associated with southerly winds in the region.

Fig. 1: Time series of wind, visibility, and PM2.5.
Fig. 1: Time series of wind, visibility, and PM2.5.
Full size image

Wind speed and direction at Hong Kong airport from March 31 to April 4, 2024 (a). Visibility with relative humidity below 95% at the Hong Kong University of Science and Technology (HKUST) and PM2.5 concentration at Tap Mun (18 km away from HKUST) (b). Blue and red horizontal dashed lines represent the 1st percentile of visibility and 99th percentile of PM2.5 calculated from data under the southerly winds from 2021 to 2023, respectively. Spatial distributions of PM2.5 concentration (unit: µg/m³) in Hong Kong and its adjacent Pearl River Delta (PRD) region at 10 am, 2 pm, 6 pm, and 10 pm on April 1, 2024 (c).

As air pollutants can harm human health3,4, it is important to promptly attribute air pollution events to their source regions. Potential sources of air pollutants for such events include biomass burning and volcanic emissions from Southeast Asia5. The Taal volcano in the Philippines (Supplementary Fig. S2), located approximately 1000 km from Hong Kong, has experienced long-term deflation. Volcanic emissions, which consist of ash and gases like sulfur dioxide (SO2), are able to affect air quality in regions thousands of kilometers downwind of the volcano6,7,8.

Accurately tracking pollution sources for such events requires large-scale data from diverse domains. Akilan et al.9,10 used GPS data to demonstrate the impacts of volcanic eruptions around the Antarctic continent and in India9,10. With improvements in computational capability, state-of-the-art chemical transport models (CTMs) have become advanced tools increasingly used to reveal the transport of air pollutants from volcanoes11. Based on a CTM model, Filonchyk et al.12 explored the transport of air pollutants from the volcanic eruption in the Canary Islands12. Inness et al.13 enhanced the capability of another CTM to forecast pollutant transport from the 2019 Raikoke eruption by assimilating SO2 column and layer height data13. Additionally, large-scale satellite measurements have proven useful for revealing emissions and the long-range transport of air pollutants from volcanoes12,14. For instance, based on satellite measurements, Milford et al.15 evaluated the impact of the 2021 La Palma volcanic eruption on air quality15.

Most of the aforementioned studies used large-scale data to evaluate the impacts of volcanic eruptions by focusing on pollutants that are regularly monitored, such as PM2.5 and SO2. These large-scale analyses may face two challenges. First, the concentration of PM2.5 is influenced by various anthropogenic and natural sources. Therefore, an increase in PM2.5 concentration may not necessarily be related to volcanic emissions, complicating the assessment. Second, the concentration of SO2 may rapidly decrease during long-range transport as it converts to sulfate, which is a component of particulate matter.

Analysis of the concentration of tracer components in aerosols provides direct evidence of the impacts of volcanic emissions16. It is widely recognized that volcanoes are significant contributors to the sulfur (S) budget and that volcanic activities lead to increased aerosol S concentrations17. Additionally, volcanic ash contains high levels of phosphorus (P), which can promote the growth of vegetation in areas affected by ash deposition18,19. During a volcanic event in Hawaii, elevated concentrations of aerosol S and P were detected at the Hawaiian Volcano Observatory20. Therefore, detailed measurements of aerosol chemical composition, even at specific locations, can significantly aid in confirming pollution sources during events influenced by volcanoes.

Here, we aim to evaluate the impacts of volcanic emissions on air quality by using a multi-scale data system that combines large-scale data (e.g., satellite monitoring and CTM simulations) with advanced monitoring of aerosol chemical compositions at specific stations in Hong Kong. The real-time multi-scale data were used to identify the potential sources and long-range transport mechanisms responsible for the mysterious air pollution event that occurred in south China, particularly in Hong Kong, on April 1-2, 2024. This multi-scale approach exemplifies comprehensive analyses that accurately attribute air pollution to unusual emission sources, such as volcanoes. Additionally, it enhances emergency preparedness and reduces health risks in regions vulnerable to volcanic events.

Results

Transboundary transport of volcanic pollutants

According to the vertical column density of SO2 from GEMS measurements and the true-color image from MODIS aboard Aqua on March 30, 2024 (Supplementary Fig. S3a, b), the Taal volcano released a large amount of SO2 and ash into the atmosphere, which was then transported westward over the SCS. The WRF-CMAQ simulations of sulfate (SO42-) concentration (Supplementary Fig. S3c) showed a similar track of the volcanic plume. During the pollution episode on April 1, 2024, the satellite measurements show that the volcano-generated gas and aerosol were initially carried westward, then turned northward and eventually reached south China with the predominantly southerly winds over the SCS (Fig. 2a, b). During the transport of volcanic pollutants, SO2 was converted to sulfuric acid, rapidly condensing to form sulfate aerosols, leading to secondary air pollution issues21. The WRF-CMAQ simulations of the SO42- concentration show a similar track of the volcanic plume and confirm the transport of the volcanic aerosol from the Taal volcano to south China (Fig. 2c). Furthermore, the trajectory of volcanic pollutants simulated using the HYSPLIT model shows similar patterns of transboundary transport from the Philippines to Hong Kong (Supplementary Fig. S4). Lidar measurements in Hong Kong were used to identify the altitude of the volcanic pollutants (Supplementary Fig. S5). As ground-level PM2.5 concentrations increased, aerosol concentrations near the surface also rose. Additionally, the majority of volcanic pollutants were transported within the planetary boundary layer, which is approximately 500 meters high. Consequently, these volcanic pollutants significantly affected ground-level air quality in south China.

Fig. 2: Satellite measurements and CTM simulations.
Fig. 2: Satellite measurements and CTM simulations.
Full size image

Vertical column density of SO2 from GEMS measurements at 1 pm on April 1, 2024 (a). True-color image from MODIS aboard Aqua at 1 pm on April 1, 2024 (b). WRF-CMAQ simulations of the SO42- concentration at 6 pm on April 1, 2024 (c).

Evidence from aerosol chemical measurements

The aerosol chemical measurements at HKUST station in Hong Kong revealed unusually high levels of SO42-, with the maximum SO42- concentration reaching 35.2 µg/m³ at 6 pm on April 1 (Supplementary Fig. S6). The variation in SO42- also became decoupled from nitrate (NO3-), another common component in PM2.5. The percentage of SO42- within PM2.5 increased from the common levels of around 30% to a high level of 75% during the peak hours (Supplementary Fig. S7). With the increase in secondary SO42- in fine particles, the fine mode fraction (i.e., the ratio between PM2.5 and PM10) increased to a level exceeding 0.9 (Supplementary Fig. S8).

Furthermore, unusually high P levels, up to 93 ng/m³, were observed on April 1, coinciding with the elevated S concentrations (Fig. 3a). The increase in P concentration was significant, as its concentration level was often below the detection limits. Similar increases in SO42- and P concentrations were found from CDSS and Mong Kok, two additional PM2.5 chemical speciation monitoring stations in Hong Kong (Supplementary Fig. S9). To rule out the contributions from biomass burning and dust, concentrations of potassium (K), organic carbon (OC), and elemental carbon (EC), common tracers of biomass burning22,23, were analysed, along with calcium (Ca), a common tracer of dust. During the pollution episode on April 1–2, 2024, the concentrations of K and Ca remained at low levels and did not exhibit coincident variation with PM2.5 (Supplementary Fig. S10). Additionally, concentrations of OC and EC also remained at low levels (Supplementary Fig. S11). These aerosol chemical analyses provided clear evidence of the transboundary transport of the volcanic plume from the Taal volcano, as indicated by the distinctive signatures of elevated S and P levels in the aerosols in Hong Kong.

Fig. 3: Chemical measurements and chemical-wind index.
Fig. 3: Chemical measurements and chemical-wind index.
Full size image

Concentrations of S (blue line) and P (yellow line) at HKUST station from March 1 to June 30, 2024 (a). Chemical-Wind Index (CWI) in Hong Kong (b).

A chemical-wind index (CWI)

The pollution episode analyses underscore the significant impacts of wind patterns on the transboundary transport of volcanic pollutants and highlight the importance of using chemical data as direct evidence for its impacts on air quality. These analyses prompted the development of a real-time Chemical-Wind Index (CWI) that combines local chemical and wind data in Hong Kong as a tool to identify the occurrence of air pollution associated with volcanic emissions from Taal. Details of the CWI are shown in the Method Section. A positive CWI value above one suggests that it is very likely the pollution in Hong Kong is associated with the volcanic emissions from the Philippines.

The results show that the CWI reached a maximum of 73.1 at 6 pm on April 1 (Fig. 3b). Additionally, the application of this index identified three additional air pollution events from April to June 2024 (e.g., April 5, April 15-16, and June 21-22) that could be attributed to ongoing emissions from the Taal volcano. During these four pollution episodes, we found unusually high levels of S (peaking at 18.4 μg/m3, 13.6 μg/m3, 11.0 μg/m3, and 6.8 μg/m3), which significantly exceeded its 99th percentile of 4.6 μg/m3. Additionally, we found unusually high P levels (peaking at 93.1 ng/m3, 30.7 ng/m3, 42.3 ng/m3, and 38.8 ng/m3), which also significantly exceeded its 99th percentile of 5.1 ng/m3. During all these pollution episodes, increases in P levels coincided with elevated S concentrations, and the dominant winds in Hong Kong were southerly. This evidence from meteorological conditions and chemical measurements supports the transboundary transport of volcanic pollutants.

Discussions

We further extend the assessment to the past three years. The application of CWI did not detect any additional significant air pollution cases related to volcanic emissions in Hong Kong during the previous three years. This can be explained by the variations in volcanic activity and occurrence frequency of the suitable wind patterns for transboundary transport. According to the emission data (Supplementary Fig. S12a), the volcano was less active during April 2022 and September 2023. The suitable wind conditions for transboundary transport include easterly winds at Taal and southerly winds in Hong Kong. These suitable wind conditions occurred less frequently in winter due to the dominant northerly winds that are associated with the winter monsoon (Supplementary Fig. S12b). In contrast, the occurrence frequency of suitable wind conditions for transboundary transport greatly increased to 50.9% in April 2024. The increased occurrence of suitable wind conditions has led to the noticeable increase in air pollution episodes in Hong Kong that are related to the Taal volcano emissions in April 2024. Therefore, the varying impacts of volcanic emissions on air quality in south China, both in April 2024 and over the past three years, can be explained by our CWI system. These results support the robustness of the CWI system.

The analysis in this study highlights the value of integrating multi-scale real-time data capabilities to promptly attribute air pollution events to their sources. The large-scale satellite measurements and model simulations (e.g., CTM and HYSPLIT models) provide a general understanding of transboundary transport. The measurements of aerosol chemical compositions at specific locations offer direct evidence of the impacts on local air quality. The combined analysis of the aerosol chemical composition and large-scale data can help elucidate the potential impacts of unusual emission sources from regions located thousands of kilometers away.

Wind conditions in Hong Kong, the Philippines, and even the SCS can affect the transport of air pollutants. Here, the development of the CWI focuses on its application in a city impacted by volcanic emissions. Therefore, the index relies solely on local data from this city and does not require data from other regions. The development and utilization of the CWI, which integrates local chemical and wind data, provide a robust approach for specific regions to identify the influence of the transported volcanic plumes. The use of local data in the CWI facilitates its application in cities where chemical and meteorological measurements are available. This index is particularly useful for people residing in areas at risk of volcanic activities, helping to protect their public health. The results of this study support emergency preparedness and help minimize the adverse impacts on the population. Additionally, this prompt analysis demonstrates the scientific process of hypothesis development and testing in near real-time. Such capabilities enhance public understanding of transboundary air pollution through education about the scientific methods used to investigate these events.

Methods

Ground measurements

Daily emission of SO2 from the Taal volcano since September 2021 was obtained from the Philippine Institute of Volcanology and Seismology (PHIVOLCS) (https://www.phivolcs.dost.gov.ph/). According to the PHIVOLCS bulletin, the volcano became more active in 2024, with SO2 emissions reaching a peak of 18,638 tons per day on March 28, 202424. The PM2.5 data were obtained from 96 air quality monitoring stations in Hong Kong and its adjacent PRD region (Supplementary Fig. S2). Among these stations, Tap Mun represents the air quality background station located in a rural area of Hong Kong. Wind data at meteorological stations at the airport in Hong Kong and the Taal volcano in the Philippines were obtained from the Global Telecommunication System of the World Meteorological Organization. Visibility, relative humidity, and hourly PM2.5 chemical composition data were obtained from an atmospheric monitoring supersite on the campus of HKUST25,26. Additional hourly PM2.5 chemical composition data was obtained from another two stations, namely Cape D’Aguilar Supersite (black triangle) and Mong Kok (red triangle). In this online PM2.5 chemical speciation network, the concentrations of S, P, K, and Ca were measured by online X-ray fluorescence (XRF) spectrometers (Xact 625i Ambient Continuous Multi-metals Monitor, Cooper Environmental Services, USA), while SO42-, NO3-, and NH4+ were measured by the Monitor of AeRosols and GAses in ambient air (MARGA, Metrohm Applikon, The Netherlands). All of these ground-based data were automatically collected from their respective data sources and their data validation and quality assessment protocols are available in Wang et al.26. The ground measurements from March 1 to June 30, 2024, were used to explore the causes of the air pollution episode. Historical data from 2021 to 2023 were used to evaluate the regular meteorological and air pollution conditions without the impacts of volcanic emissions.

Satellite measurements

The vertical column density of SO2 data from the Geostationary Environment Monitoring Spectrometer (GEMS) were used to track the transport of volcanic gas during the pollution episode. The real-time GEMS data were obtained from the National Institute of Environmental Research (NIER) of South Korea (https://nesc.nier.go.kr). Meanwhile, real-time true-color images from the Moderate Resolution Imaging Spectroradiometer (MODIS) aboard the Aqua satellite were obtained from the NASA Worldview database (https://worldview.earthdata.nasa.gov/) and used to track the dispersion of the aerosols.

CTM simulations

The Weather Research and Forecast and Community Multiscale Air Quality (WRF-CMAQ) model was used to simulate the transport of SO42- resulting from the volcanic SO2 emission. More details of the model setup can be found in Zhang et al.27. The simulations assumed continuous emissions of SO2 at a rate of 10,000 tons per day from the Taal volcano. The simulations of SO42- concentrations were performed for two scenarios: (1) a basic scenario with anthropogenic and biomass burning emissions, and (2) a volcanic scenario with the addition of the volcanic emissions from Taal. The impacts of the volcanic emissions were obtained by evaluating the differences between the two scenarios. Additionally, the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model was used to simulate the trajectories of the volcanic pollutants.

Chemical-Wind Index (CWI)

The transboundary transport of volcanic pollutants across the SCS requires southerly wind in Hong Kong. Based on the wind data in Hong Kong, a Wind Index (WI) is defined as:

$${WI} = \left\{\begin{array}{ll}1, {if\; wind\; is\; southerly} \hfill \\ 0, {if\; wind\; is\; not\; southerly}\end{array} \right.$$
(1)

When the wind in Hong Kong was southerly (90–270°), the WI equals one. Otherwise, the WI equals zero.

Considering that S and P are important tracers for volcanic emissions, we leveraged the 99th percentiles of S and P concentrations under the southerly winds from 2021 to 2023 as the threshold to identify a pollution episode that was associated with volcanic emissions. Based on the historical measurements of aerosol chemical compositions from 2021 to 2023, the 99th percentile of S and P concentrations under the southerly winds were estimated to be 4.6 µg/m3 and 5.1 ng/m3, respectively. A Sulfur Index (SI) is then defined as:

$${SI}= \left\{\begin{array}{ll}{\frac{S}{{S}_{99}}, {if}{S} \, \ge \, {S}_{99}} \\ 0, \hfill {if}{S} \, < \, {S}_{99} \hfill \end{array}\right.$$
(2)

When the S concentration exceeds its 99th percentile, the SI equals the ratio of the actual S concentration to its 99th percentile. Otherwise, the SI equals zero. Similarly, a Phosphorus Index (PI) is defined as:

$${PI}= \left\{\begin{array}{ll}{\frac{P}{{P}_{99}}, {if}{P} \, \ge \, {P}_{99}} \\ 0, \hfill {if{}P} \, < \, {P}_{99} \hfill \end{array}\right.$$
(3)

When the P concentration exceeds its 99th percentile, the PI equals the ratio of the actual P concentration to its 99th percentile. Otherwise, the PI equals zero. Subsequently, the CWI is defined as a product of SI, PI, and WI:

$${CWI}={SI}\,{{\cdot }}\,{PI}\,{{\cdot }}\,{WI}$$
(4)

A positive CWI value above one indicates that it is very likely the pollution in Hong Kong is associated with the volcanic plume originating from the Philippines.