Footprints of El Niño La Niña on the evolution of particulate matter over subtropical Island Taiwan

Footprints of El Niño La Niña on the evolution of particulate matter over subtropical Island Taiwan | npj Climate and Atmospheric Science

Abstract

Particulate matter (PM) pollution has become a major problem worldwide, with significant adverse health impacts, but the climatology of annual variations in and the effect of climate change on PM levels are still not comprehensively evaluated. Here we show that the vertical motions of the East Asian winter and summer monsoons regulate the annual variation in PM over Taiwan. The PM pollution season starts and ends earlier during El Niño episodes than during La Niña episodes. Furthermore, the onset, active, break, revival, and retreat phases of the PM pollution lifecycle are innovatively defined. Our results demonstrate that the annual seasonal cycle dominates wintertime PM pollution climate development during the active phase with a minor modulation by El Niño and La Niña episodes, whereas October and March, the seasonal transition periods, are significantly modulated by these two episodes. The findings suggest a new dimension of PM pollution research—the lifecycle evolution effect.

Introduction

Particulate matter (PM; particles with aerodynamic diameters of ≤10 and ≤2.5 μm [PM10 and PM2.5, respectively]) pollution has become a major problem worldwide and is associated with adverse health outcomes, including respiratory and cardiovascular diseases. PM transport by airflow, compared with diffusion, can enable it to more efficiently travel large distances in the atmosphere and spread over a large area. Furthermore, air vertical motion helps PM to either scavenge in the atmosphere or accumulate over surfaces 1, 2, 3, 4. No study has comprehensively evaluated the climatology of annual variations in PM level, its relationship with large-scale circulation, and the effect of climate change on PM level. Although the World Health Organization (WHO) 5 provides standard guidelines for both annual and 24-h average PM levels, a uniform method to define the PM pollution season has not yet been developed. Analyzing local pollution (LP) events in Taiwan, numerous studies have consistently reported that the contribution of long-range transport (LRT) sources from the Asian continent to PM concentrations for LP episodes occurs annually from autumn to the following spring when the northeasterly monsoon persists 6, 7, 8, 9, 10. The LRT is characterized by low temperature and relative humidity (RH), weak precipitation, and steady northeasterly high wind speed, implying PM dilution through plume dispersion, mixing with drier air, inverse aerosol hygroscopic growth, and less wash-out effect 11, 12, 13. The major LP emission sources in Taiwan are industrial processes, road traffic, and construction activities, and the temporal and spatial distributions of emission sources are archived in the Taiwan Emission Inventory Data System version 11.1 (TEDS 11.1) 14. Among these studies, the PM pollution season has no coherent period. Therefore, a simple method to define the PM pollution season (Methods) is first proposed in this study, and then it is feasible to further investigate the effect of climate change, particularly that caused by different El Niño–Southern Oscillation (ENSO) episodes, on the evolution of PM during a year.

On the basis of the effect of annual variations in climatic factors caused by the alternation of the northeast winter and southwest summer monsoons in East Asia, Taiwan is naturally divided into four weather/climate regions 15, 16 by two major mountain ranges: the Xueshan Range (XR) in the northwest and the north–south-oriented Central Mountain Range (CMR) (Fig. 1a). These four weather/climate regions, namely, northern Taiwan (N), the west central to southern plains (WCS), the northeastern plain (NE), and the southeastern plain (SE) along the valley next to the Coastal Range, have distinct air pollution characteristics. Due to the blocking effect of the XR and CMR, the northeasterly wind speeds in the WCS are relatively weaker than those in N, so the overall estimation between LRT and LP contributions are approximately 60:40 in N and 40:60 in WCS 12, 13 under the ordinary LRT situation, indicating that local emissions dominated in the WCS. The ratios could be 30:70 in N and 25:75 in WCS when the northeast wind becomes much weaker. Furthermore, combining the local emissions redistributed by the atmospheric boundary layer (ABL)-level lee-side vortices with the LRT pollutants on the surface 17, the WCS becomes an air pollution entity. According to the 27-year (1994–2020) mean statistics (Supplementary Fig. 1a), a high PM10 level over the WCS plains with less rainfall and a relatively drier strong wind speed environment (Supplementary Fig. 1b), implying both LRT and LP situations as previously described, is observed from fall until the following spring. Even in the presence of high precipitation and relative humidity with more wash-out effect over N, the PM10 level continually increases from October to the following April; the average PM10 level in N is lower than that in the WCS by 34 μg∙m−3 during this period (the average level is 84 μg∙m−3 in the WCS and 50 μg∙m−3 in N). In contrast, both the SE and NE plains have cleaner air throughout the year. To fully understand PM evolution throughout a year and distinguish the possible impact of climate change, PM observations should be performed from July 1 to June 30 in the following year to observe continual changes in PM levels. The time-station diagram (Fig. 1c). In addition to the regional weather/climate and orography effect, the presence of numerous science and industrial parks in the WCS render it the most polluted place in Taiwan. This study particularly focuses on the WCS plains of Taiwan.

a The topography of Taiwan shaded with a brown scale in meters showing terrain elevations. The two major mountain ranges, Xueshan Range and Central Mountain Range, are denoted by dashed ice blue and dashed gold lines, respectively. Locations of Taiwan’s Environmental Protection Administration air quality monitoring stations in the four naturally divided climate regions are marked with blue dots over northern Taiwan (N), red dots over the west central to southern (WCS) plains, light green dots over the northeastern (NE) plain, and dark green dots over the southeastern (SE) plain. b Time-station diagram for the 26-year (1994/07–2020/06) daily climatology of PM10 levels. c Time series of the 26-year (1994/07–2020/06) monthly PM10 levels, temperature, relative humidity, wind speed, and rainfall histogram for the four climate regions.

Results

Lifecycle of the PM pollution season

Following the procedures described in the Methods, we objectively determined both the PM10 onset date (PMOD) and PM10 retreat date (PMRD) for individual years during 1994–2020 (Supplementary Fig. 2 and Supplementary Table 1), and they exhibited clear interannual variations. The average PMOD (PMRD) for the 26 years was October 7 (May 4), with a standard deviation of 13 (14) days. The earliest onset (retreat) date was September 17 (April 7) in 2012 (1998), and the latest onset (retreat) date was November 8 (June 8) in 2010 (2003). With the use of the mean PM10 level for each year as a reference by adjusting the mean value to 0 (or simply taking out the yearly mean for each year), the 26-year mean climatology data of the annual PM10 index marked with the average PMOD and PMRD are illustrated in Fig. 2a, whereas the annual mean time series data are depicted in Fig. 2e. Note that this procedure is approximate to detrending the decreasing tendency of the long-term PM10 level, particularly from 80 μg∙m−3 in 2005/2006 to 40 μg∙m−3 in 2019/2020, representing a decrease of 2.42 μg∙m−3 per year. Despite a synchronous decrease in the standard deviation of the PM10 level each year (Fig. 2).

a Daily climatology of the PM10 level (dark red line) superimposed with areal mean large-scale vertical motion omega (histogram in pink denotes upward motion and blue indicates downward motion) for the 1994/07–2020/06 period. The correlation coefficient (γ) between PM10 and vertical motions is shown at the upper-right corner of the panel. PMOD and PMRD are marked with vertical green dashed lines with corresponding dates. The monthly PM10 level is indicated by orange digits during the pollution season (October to April), whereas bold orange digits represent the active (November to January) and revival (February to April) phases. Same as in a except for b El Niño, c La Niña, and d neutral composites. e Time series of the annual mean PM10 (PM2.5) concentrations over the WCS plains from 1994/1995 to 2019/2020 (2005/2006 to 2019/2020) associated with one standard deviation for each year (in shadings). The significant decreasing trend (P = 0.0000 for both PM10 and PM2.5) from 2005/2006 to 2019/2020 is depicted by a blue dashed line. f Daily evolution of composite PM10 from 10 days before to 10 days after LNYD (vertical yellow dashed line) for each individual year (thin dashed lines) and 26-year climate mean (red solid line). g Composite relationship between PMOD anomalies (on the ordinate) and the ONI in October (on the abscissa) from 1994 to 2020. The statistically significant linear regression is above the 99% confidence level. h ONI from July to June for individual years from 1994/07 to 2020/06. El Niño (La Niña) years are depicted in red (blue) lines and neutral years in green, whereas the average of each episode is shown in thick solid lines.

The mechanisms underlying the occurrence of PMOD and PMRD remain unclear. To examine precipitous temporal development during these two transition periods, we evaluated composite wind flow and divergent circulation at 925 hPa superimposed with precipitation by using the mean climate data of the 26 years with their center date coinciding with PMOD (PMRD) for both 10 days before and after the PMOD (PMRD) together with their differences; the results are depicted in Fig. 3a(3b). The surface temperature over northeastern China continually decreases following the seasonal shift of southward-migrating overhead solar heating during the transition period of PM onset, thus intensifying the anticyclone with cooler, drier air over northeastern China. This phenomenon results in the occurrence of stronger northeasterly and divergent winds over this region after the PMOD (Fig. 13 for the PMOD. Additionally, anomalous anticyclonic circulation together with an anomalous divergence center is observed over Taiwan (difference panel in Fig. 3a). Physically, cooler, drier, and stable subsiding air present over the anomalous anticyclone around Taiwan suppresses the vertical development of the boundary mixing layer and significantly increases the levels of local air pollutants, leading to the occurrence of PMOD. The PM pollution season lasts for approximately 7 months (October – April) from PMOD to immediately before PMRD. The environment required to maintain such a long pollution season is described in Fig. 3c. The tropical Southeast Asian high over the east of the Philippine Sea in the western Pacific Ocean at 200 hPa extends downward at 700 hPa to block the southward advection of cold air 18 but facilitates the flow of cold dry air underneath this lower troposphere, resulting in cold-surge-type anticyclonic circulation at 925 hPa, which is evidently depicted by the local Hadley cell circulation in terms of the (vd, −ω) cross section 19(Fig. 3c). The vertical structure of the winter-type large-scale environment indicates that the surface anticyclone associated with divergent circulation at 925 hPa exhibits a vertical northwestward tilt toward the 200-hPa ridge along with a convergent center in the upper troposphere 20. The local east–west cell circulation in terms of the (ud, −ω) cross section also depicts the descending motion of the upper west troposphere flowing down eastward toward western Taiwan (west of 121°E). The stable, cold, and dry air together with the descending motion provides favorable conditions for the accumulation of surface air pollutants over western Taiwan, whereas the orography-induced upward motion along the eastern coastal zone (east of 121°E) helps to maintain a relatively unpolluted environment in that area. In contrast, the surface temperature over southeastern China persistently increases and gradually results in the development of an anomalous low-level cyclonic center northeastward extending through southern Japan and covering Taiwan during the PM retreat transition period (difference panel in Fig. 3b). Moreover, an anomalous convergent center forms, bringing warmer moist air from the south and generating precipitation due to ascending motions, thus reducing air pollutant levels over Taiwan and leading to the occurrence of PMRD. Subsequently, the Asian summer monsoon environment, exhibiting a vertical phase reversal structure between the upper-level anticyclonic Tibetan high associated with a divergent center and the lower-level cyclonic continental low associated with a convergent center 21, is well established (Fig. 3d). The upward branches from both the local Hadley and east–west cell circulations are coupled with convergent circulation at a lower level, bringing in warm moist air to produce abundant rainfall, and divergent circulation at an upper level, maintaining strong upward motion. This favorable environmental setting during northern summer results in the washing out or upward transport of air pollutants, resulting in better surface air quality in Taiwan from May to September. Thus, the downward motion of the winter monsoon and the upward motion of the summer monsoon provide the basic environment that regulates annual PM variations over Taiwan (the high correlation coefficient between the PM10 level and vertical velocity omega was 0.75), facilitating the differentiation of the pollution season from the clean-air season throughout the year (Fig. 2a).

a Composite charts of total wind vectors (V) associated with streamlines and divergent wind vectors associated with velocity potential (χ) at 925 hPa superimposed with precipitation during 10 days before (left panel) and 10 days after (central panel) the PM onset date (PMOD), including the PMOD and their differences (After minus Before: right panel); shaded areas and black vectors in difference panels denote a significant difference above the 95% confidence level. The contour interval of χ is 1 × 106 m2 s−1, and Δχ is 2 × 105 m2 s−1. b Same as in a except for the PM retreat date (PMRD). c 26-year (1994/1995–2019/2020) climatology charts of total wind vectors associated with streamlines at 200 hPa (upper-left panel), total wind vectors associated with streamlines superimposed with precipitation at 925 hPa (central left panel), local Hadley circulation along 119°E–123°E (lower-left panel), divergent wind vectors associated with velocity potential at 200 hPa (upper-right panel), divergent wind vectors associated with velocity potential superimposed with precipitation at 925 hPa (central right panel), and local east–west cell circulation along 21.5°N–25.5°N (lower-right panel) during the PM pollution season from October to the next April. The contour interval of χ is 1 × 106 m2 s−1. The symbol denotes the location of Taiwan at 23.5°N in the local Hadley circulation and 121°E in the local east–west cell circulation. d. Same as in c except for the PM clean-air season from May to September.

Compared with the values for October to April (Fig. 2a), the monthly mean PM10 levels in November and December were higher, at 23.75 and 23.30 μg∙m−3, respectively, and that in March was 17.79 μg∙m−3. The observation of a lower PM10 level between the end of January and early February can be ascribed to the average Lunar New Year date (LNYD) being February 5 for the 26 years, with a standard deviation of nine days. We drew a composite diagram by using the mean climate data of the 26 years with their center date coinciding with the LNYD and extending 10 days before and after the LNYD (Fig. 2f). Our findings indicate that the PM10 level begins to decrease from Lunar New Year’s Eve and reaches the yearly mean reference level on the LNYD. This decreasing trend is observed for another two days and then starts to restore immediately. Traditionally, people celebrate the Lunar New Year beginning on New Year’s Eve, with government and the private sector closing down for at least four days for the holiday break. This, consequently, reduces air pollutant levels and clean air, thus benefitting society. To the best of our knowledge, no study has attempted to comprehensively depict the PM pollution season. This study innovatively defines the five phases of the pollution lifecycle: onset (PMOD), active (November to January, NDJ), break (between the end of January and early February), revival (February to April, FMA), and retreat (PMRD); these phases are similar to the well-defined Indian and East Asian summer monsoon lifecycles 22, 23, 24. Interestingly, the most severe pollution events occur during the NDJ active phases 11, 12, 13. The East Asian summer monsoon occurs during the rainy season from May to September, resulting in clean air over Taiwan, whereas the PM pollution season occurs immediately from October to April with a dry and polluted environment during the East Asian winter monsoon period; the aforementioned two different five-phase features are consecutively noted throughout the year.

Effect of ENSO episodes

Because of the wide distribution of the PMOD and PMRD, their specific characteristics should be investigated. We focused on the month of October to examine the interannual variation in the PMOD because of its critical role in differentiating the PMOD between the early- and late-onset years given that most of the PMODs (15 out of 26 years) occurred in October in addition to the climatology on October 7. We elucidated the mechanism underlying the interannual variation in the PMOD and noted a significantly linear and negative correlation between the PMOD and Oceanic Niño Index (ONI) in October (Fig. 2g; P < 0.001; γ = −0.553), indicating that the PMOD tends to occur earlier during El Niño years but later during La Niña years. Furthermore, we compared the time series of the ONI among all the years (Fig. 2h) and divided them into three groups based on the ONI values in October: El Niño (94/95, 97/98, 02/03, 04/05, 06/07, 09/10, 14/15, 15/16, and 18/19 with sea surface temperature [SST] ≥0.5 °C), La Niña (95/96, 98/99, 99/00, 00/01, 07/08, 10/11, 11/12, 16/17, and 17/18 with SST ≤ −0.5 °C), and normal (other) episodes. Similar to the climatology of the annual PM10 index (Fig. 2a), we averaged the original PM10 indices and their respective PMODs and PMRDs for all the selected years in the aforementioned three groups (Fig. 2b–d). As evident in the statistical data presented in Supplementary Table 1 and Fig. 2b–d, the PM pollution season, including both the PMOD and PMRD, occurred earlier during El Niño episodes than during La Niña episodes, whereas normal episodes occurred between these two episodes. Except for only a one-day difference in the PMRD, the averaged PM pollution season for normal years was identical to that observed in the 26 years, indicating that the 26-year mean climate data might reach the climatic norm. Furthermore, a significant 20-day difference in the average PMOD was noted between El Niño (September 28) and La Niña (October 17) episodes (P = 0.002). However, a nonsignificant six-day difference in the PMRD was observed between them (P = 0.47). Due to the large-scale environment evolving toward a similar neutral stage around May for every episode (Fig. 2h), the separation between PMRDs became much less compared with the apparently disjointed PMODs in October. In addition, the five phases of the PM pollution lifecycle were noted in all three episodes (Fig. 2b–d). Furthermore, compared with the 26-year climatology findings (Table 1), the PM10 level significantly increased by 44% (decreased by 47%) in October after the PMOD and slightly increased (decreased) during the active NDJ in the El Niño (La Niña) episode; however, its level decreased by 24–60% (increased by 19–72%) during the revival FMA. The comparison between La Niña and El Niño episodes demonstrated similar findings, indicating that the annual seasonal cycle dominates wintertime PM pollution climate development during the NDJ active phase with a minor ENSO modulation. However, during the seasonal transition period, the FMA revival phase is significantly affected by ENSO. Thus, the slightly lower PM10 level (<6%) during the La Niña NDJ is attributable to the effect of a mild outflow pollutant transport anomaly (Supplementary Fig. 3), whereas the significantly higher PM10 level (>92%), which is nearly double that observed during El Niño, in the FMA under relatively dry conditions is associated with anomalous surface convergent flow, which facilitates the accumulation of PM10.

We explored the mechanism underlying a marked delay in the PMOD during La Niña episodes. Figure 4 illustrates the average large-scale horizontal and vertical cell circulations for the three episodes as well as differences between La Niña and El Niño between September 28 and October 17, the period between the average El Niño and La Niña PMODs. Following the seasonal march during this period, the upper-level anticyclonic Tibetan high at 200 hPa retreated southeastward to Myanmar during all three episodes (Fig. 4a–c). The La Niña episode maintained the oceanic trough over the northwestern Pacific between 140°E and 160°E, but the other two episodes resulted in a split anticyclone, which eventually developed into a tropical Southeast Asian high (Fig. 3c) with the progression of the season. For the lower level at 925 hPa, an anticyclone with relatively cool and dry air over northeastern China was noted for the El Niño and normal years; however, strong moist easterly winds over the northwestern Pacific prevailed over this region during the La Niña episode. The local Hadley circulation depicted by the vertical cross-section of (vd, −ω) was well developed in all three episodes, except that it intensified in La Niña. The lower-level convergent center at 925 hPa and the upper-level divergent center at 200 hPa occurred in response to warmer tropical Pacific SST anomalies over the west of the tropical western Pacific between the South China Sea and the Philippines during La Niña 25 and over the dateline during El Niño but over the east of the Philippines in normal years. This phenomenon not only strengthened (weakened) the ascending motions of the local Hadley circulation but also induced upward (downward) motion branches over 120°E in the local east–west cell circulation, depicted by the vertical cross section of (ud, −ω), during La Niña (El Niño). Thus, the differences between La Niña and El Niño (Fig. 4d) imply that both large-scale horizontal and vertical circulation observed in summer are sustained through mid-October (compared with the Clean-air season in Fig. 3d), thus maintaining good air quality on the surface and causing a delay in the PMOD before the strengthening of cool fall weather during La Niña development phases.

a Composite charts of total wind vectors associated with streamlines at 200 hPa (upper-left panel), total wind vectors associated with streamlines superimposed with precipitation at 925 hPa (central left panel), local Hadley circulation along 119°E–123°E (lower-left panel), divergent wind vectors associated with velocity potential at 200 hPa (upper-right panel), divergent wind vectors associated with velocity potential superimposed with precipitation at 925 hPa (central right panel), and local east–west cell circulation along 21.5°N–25.5°N (lower-right panel) during September 28 to October 17, the period between average El Niño and La Niña PM onset dates for the El Niño episode. The contour interval of χ is 1 × 106 m2 s−1. The symbol denotes the location of Taiwan at 23.5°N in the local Hadley circulation and 121°E in the local east–west cell circulation. Same as in a except for b. La Niña episode, c neutral episode, d difference between La Niña and El Niño episodes; shaded areas and black vectors denote a significant difference above the 95% confidence level. The contour intervals of Δχ are 1 × 106 m2 s−1 at 200 hPa and 2 × 105 m2 s−1 at 925 hPa.

We compared the four pollution climate regions between La Niña and El Niño episodes (Supplementary Fig. 4). The high correlation coefficients between the WCS plains and the other three climate regions for both temperature and relative humidity imply that the entire island experiences the same large-scale circulation environment. However, moderate correlation coefficients for PM10, rainfall, and wind speed are attributed to the orographic interaction effect. Despite the complicated geographic structure, the coherent reduction in the PM10 level in all four climate regions together with the large-scale vertical motion omega all passed the 99% confidence level in October, which indicates a seasonal transition period from summer to winter. These findings confirm the significant effect of large-scale ENSO events on PM evolution in Taiwan.

Discussion

In this study, we determined that the downward motion of the large-scale East Asian winter monsoon and the upward motion of the East Asian summer monsoon provide the basic environment that regulates the annual variation in PM over Taiwan based on 26 years (1994/07–2020/06) of near-climatic norm climatology data from PM10 observations. The PM pollution season starts and ends earlier during El Niño episodes than during La Niña episodes—in particular, a significant 20-day difference is evident between their starting dates. The mechanism underlying this marked delay in the PMOD during La Niña episodes is attributed to the summertime large-scale circulation sustained through mid-October that maintains good air quality on the surface and causes a delay in the PMOD. We explored the innovative onset, active, break, revival, and retreat phases of the PM pollution lifecycle to determine the effect of El Niño and La Niña episodes on each phase. Our results demonstrate that the annual seasonal cycle dominates wintertime PM pollution climate development during the NDJ active phase with a minor modulation by these two episodes, whereas the FMA revival phase, during the seasonal transition period, is significantly affected by ENSO.

Regarding the 15-year (2005/07–2020/06) fine PM2.5 archived observations, the results (Fig. 5, Supplementary Table 1, and Supplementary Figs. 1a and 5) are similar to those for PM10. Because of the smaller sample size, the five phases of the PM2.5 pollution lifecycle appeared to undergo an adjustment stage, except during the La Niña episode. However, the difference in the PMOD between El Niño and La Niña episodes was also significant (P = 0.038).

a Daily climatology of the PM2.5 level (orange line) superimposed with areal mean large-scale vertical motion omega (histogram in pink denoting upward motions and blue indicating downward motions) over the period 2005/07–2020/06. The correlation coefficient (γ) between the PM2.5 level and vertical motion is shown at the upper-right corner of the panel. The PM onset date (PMOD) and PM retreat date are marked with vertical green dashed lines with corresponding dates. The monthly PM2.5 level is indicated by red digits during the pollution season (October to April), while the active (November to January) and revival (February to April) phases are indicated by bold red digits. Same as in a except for b El Niño, c La Niña, and d neutral composites. e Composite relationship between PMOD anomalies (on the ordinate) and the October ONI (on the abscissa) from 2005 to 2020. The statistically significant linear regression is above the 99% confidence level. f Time-station diagram for the 15-year (2005/07–2020/06) daily climatology of the PM2.5 level.

In addition to offering a new perspective on PM variation, the findings of this study also suggest a new dimension of PM pollution research—the lifecycle evolution effect. Previous studies have often focused on the influence of ENSO on the interannual variation in PM pollution, such as local PM10 interannual variation in Malasia 26 and PM2.5 pollution over China 27, 28. No efforts have been made to explore the modulation of PM pollution in terms of the lifecycle evolution effect by ENSO. This study was prompted by the need for a better seasonal prediction of PM pollution levels. To advance the quantitative assessment of the PM pollution level more accurately in the seasonal outlook projection, understanding the year-to-year variability effect on the PM pollution lifecycle provides basic guidance for administrative tasks, and the new findings of this study serve this purpose. Additionally, most of the South and Southeast Asian countries, which reside in the same large-scale Asian winter and summer monsoon systems as Taiwan, also experience similar heavily polluted environmental challenges. Thus, the modulation of the PM pollution lifecycle by ENSO among these countries deserves further in-depth study.

Methods

Observations and reanalysis data

To investigate the spatiotemporal characteristics of PM relevant to climate over Taiwan, a long-term PM10 dataset covering 1994–2020 and a PM2.5 dataset covering only 2005–2020 were acquired from Taiwan’s Environmental Protection Administration (EPA). We used this dataset to construct PM indices and examined PM variability in Taiwan. After the exclusion of some monitoring stations in the mountain area and those particularly intended as traffic and industrial probes, data from 57 general stations were used in this study. During these 27 years, two extreme incidences occurred: aeolian dust on November 2, 2009, and a dust storm on March 21–23, 2010. To prevent the distortion of annual variation features being induced by these extreme outliers, we applied linear interpolation to replace these dates. For the long-term PM climatology examination, the 27-year PM10 and 16-year PM2.5 daily climatology data for each station, arranged from north to south and from western to eastern Taiwan, are illustrated in time-station diagrams (Supplementary Fig. 1a). Four PM climate groups were observed for both PM10 and PM2.5. Data on monthly rainfall, temperature, relative humidity, and wind speed for the period from 1994 to 2020 were obtained from Taiwan Central Weather Bureau (CWB) surface stations for use in the statistical analysis of differences among the four climate regions. The monthly PM10 level and four meteorological variables, averaged from their corresponding observation stations in the respective four climate regions in terms of the time series delineated in Supplementary Fig. 1b, verify these four PM pollution climate regions in Taiwan. Moreover, the findings of the comparison between El Niño and La Niña episodes for these four climate regions and those of the comparison between the WCS plains and the other three regions are illustrated in Supplementary Fig. 4.

To depict large-scale atmospheric circulation associated with PM onset/retreat, PM pollution/clean-air seasons, and different climate episodes during the PM onset period, we employed the daily ERA5 reanalysis data 29 with a resolution of 0.25° × 0.25°. The divergent winds (ud, vd) presented in Figs. 3 and 4 were evaluated using the horizontal gradients of velocity potential (χ), which may be computationally obtained by solving Poisson’s equation for the global data 30. To construct the local Hadley circulation, vd and vertical velocity omega (ω) directly obtained from ERA5 were averaged along 119°E–123°E for every level from 1000 to 200 hPa. For local east–west cell circulation, ud and ω were averaged along 21.5°N–25.5°N. Both the local Hadley and east–west cell circulations are depicted in Figs. 3 and 4. In addition, the areal mean ω (119°E–123°E, 21.5°N–25.5°N) covering the entire island, vertically averaged from 600 to 400 hPa, was used to estimate the correlation between various climatic PM10 time series (Figs. 2 and 5). Regarding rainfall, the daily Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks-Climate Data Record (PERSIANN-CDR) 31 on a 0.25° latitude–longitude grid was generated from long-term multisatellite high-resolution observations spanning 1983–2020. This dataset was used to examine the relationship between PM variability in Taiwan and the surrounding activity embedded in the large-scale environment (Figs. 3 and 4).

The historical ONI, including the classified El Niño and La Niña episodes, provided by the Climate Prediction Center (CPC) of the National Centers for Environmental Prediction (NCEP) of the National Oceanic and Atmospheric Administration (NOAA) was adopted as a measure of the ENSO. For consistency, the ONI was rearranged from July to the next June as an individual year from 1994 to 2020 (Fig. 2h). If the ONI in October was ≥0.5 °C (≤−0.5 °C), the year was categorized as an El Niño (La Niña) year; others were considered normal years.

Definition of PM onset and retreat dates

We used only the rainfall parameter to successfully determine both the onset and retreat dates for the rainy season over the Central Highlands of Vietnam 32. To investigate the features of the active PM pollution season, we also propose to determine the starting and ending dates for the season by using only the PM parameter. We first define two terms: the PM onset date (PMOD) and PM retreat date (PMRD), designated as the PM pollution season starting date and ending date, respectively. Second, the PM levels over WCS Taiwan show a significant reduction for both PM10 and PM2.5 from the WHO recommended annual mean interim target 1 (70 μg∙m−3 for PM10 and 35 μg∙m−3 for PM2.5) in 2005 approaching target 3 (30 μg∙m−3 for PM10 and 15 μg∙m−3 for PM2.5) in 2020 (Fig. 2e). Therefore, it is legitimate to use the annual mean PM level for each year as a reference to determine the pollution status, either polluted or less polluted, during a year. To reflect the apparent year-to-year change in PM level, many sensitivity tests are performed to eventually obtain the optimal criteria and procedures for detecting the PMOD and PMRD for each individual year as follows:

1.
PMOD: The daily PM level is larger than the yearly mean value (PMYRM) for five consecutive days. The daily PM level of at least 15 days should be larger than the PMYRM within 20 consecutive days after the PMOD.
2.
PMRD: The same procedures were used as those for the PMOD to determine the PMRD, but the criteria were set to be less than the PMYRM.

After the PMOD (PMRD) is determined, we are able to distinguish whether the PM pollution season of one individual year starts (ends) earlier than others and vice versa. Moreover, it is feasible to investigate whether the possible mechanisms underlying the occurrence of PMOD and PMRD are common among all years, as well as the effect of climate change on the evolution of PM during each active pollution season. For the PM10 time series, the concentrations averaged from the air quality monitoring stations over WCS Taiwan, marked with the PMOD and PMRD for individual years, during 1994–2020 are displayed in Supplementary Fig. 2, and the PM2.5 time series for 2005–2020 are exhibited in Supplementary Fig. 5.

Statistical analyses

Statistical analyses were performed using the two-tailed Student’s t test. The Mann–Kendall test 33 and Sen’s slope estimator 34 were used to detect the trend of the annual mean PM concentration time series (Fig. 2e). The statistical significance of the linear relationship in simple linear regression between two variables, PMOD and October ONI in Figs. 2g and 5e, was examined using a t test for the slope of the regression line. The t test for the difference between two population means was performed to compare between two episodes in Table 1 and Supplementary Table

Data availability

EPA PM data are available at https://airtw.epa.gov.tw/CHT/Query/His_Data.aspx. CWB monthly meteorological data can be obtained from the CWB observation data inquire system at https://e-service.cwb.gov.tw/HistoryDataQuery. ERA5 reanalysis data are available at the Climate Data Store at https://cds.climate.copernicus.eu. PERSIANN-CDR data are available at https://www.ncei.noaa.gov/data/precipitation-persiann/access/. NOAA CPC ONI index data are available at https://origin.cpc.ncep.noaa.gov/products/analysis_monitoring/ensostuff….

Code availability

The code used in this study is available from the corresponding author on reasonable request.
Mamun, A. A., Cheng, I., Zhang, L., Dabek-Zlotorzynska, E. & Charland, J.-P. Overview of size distribution, concentration, and dry deposition of airborne particulate elements measured worldwide. Environ. Rev. 28, 77–88 (2020).
Cheng, I., Mamun, A. A. & Zhang, L. A synthesis review on atmospheric wet deposition of particulate elements: scavenging ratios, solubility and flux measurements. Environ. Rev. 29, 340–353 (2021).
Yen, M.-C. et al. Climate and weather characteristics in association with the active fires in northern Southeast Asia and spring air pollution in Taiwan during 2010 7-SEAS/Dongsha Experiment. Atmos. Environ. 78, 35–50 (2013).
Lin, C.-C. et al. Influences of the Long-range Transport of Biomass-burning Pollutants on surface air quality during 7-SEAS Field Campaigns. Aerosol Air Qual. Res. 17, 2595–2607 (2017).
World Health Organization, 2021: WHO global air quality guidelines for particulate matter, ozone, nitrogen dioxide, sulfur dioxide and carbon monoxide. (WHO press, 2021).
Lin, C.-Y. et al. Long-range transport of Asian dust and air pollutants to Taiwan. Terr. Atmos. Ocean Sci. 15, 759–784 (2004).
Chuang, M. T. et al. The effects of synoptical weather pattern and complex terrain on the formation of aerosol events in the greater Taipei area. Sci. Total Environ. 399, 128–146 (2008).
Junker, C., Wang, J.-L. & Lee, C.-T. Evaluation of the effect of long-range transport of air pollutants on coastal atmospheric monitoring sites in and around Taiwan. Atmos. Environ. 43, 3374–3384 (2009).
Hsu, C. H. & Cheng, F. Y. Synoptic weather patterns and associated air pollution in Taiwan. Aerosol Air Qual. Res. 19, 1139–1151 (2019).
Su, S.-H., Chang, C.-W. & Chen, W.-T. The Temporal Evolution of PM2.5 Pollution Events in Taiwan: Clustering and the Association with Synoptic Weather. Atmosphere 11, 1265 (2020).
Wang, S.-H., Hung, W.-T., Chang, S.-C. & Yen, M.-C. Transport characteristics of Chinese haze over northern Taiwan in winter, 2005–2014. Atmos. Environ. 126, 76–86 (2016).
Lin, C. Y. et al. Long-range transport of aerosols and their impact on the air quality of Taiwan. Atmos. Environ. 39, 6066–6076 (2005).
Chuang, M. T., Lee, C. T. & Hsu, H. C. Quantifying PM2.5 from long-range transport and local pollution in Taiwan during winter monsoon: an efficient estimation method. J. Environ. Manag. 227, 10–22 (2018).
Environmental Protection Administration in Taiwan. Taiwan Emission Data System Version 11.1 (TEDS 11.1) (in Chinese) available via,
https://air.epa.gov.tw/EnvTopics/AirQuality_6.aspx(2022).
Chen, T.-C., Yen, M.-C., Hsieh, J.-C. & Arritt, R. W. Diurnal and Seasonal Variations of the rainfall measured by the automatic rainfall and meteorological telemetry system in Taiwan. Bull. Am. Meteor. Soc. 80, 2299–2312 (1999).
Yen, M.-C. & Chen, T.-C. Seasonal variation of the rainfall over Taiwan. Int. J. Climatol. 20, 803–809 (2000).
Lai, H. C. & Lin, M. C. Characteristics of the upstream flow patterns during PM2.5 pollution events over a complex island topography. Atmos. Environ. 227, 117418 (2020).
Chen, T.-C., Yen, M.-C., Huang, W.-R. & Gallus, W. A. An East-Asian cold surge: Case study. Mon. Wea. Rev. 130, 2271–2290 (2002).
Yen, M.-C. & Chen, T.-C. A revisit to the East-Asian Cold Surge. J. Meteorol. Soc. Jpn. 80, 1115–1128 (2002).
Chen, T.-C. The Structure and Maintenance of Stationary Waves in the Winter Northern Hemisphere. J. Atmos. Sci. 62, 3637–3660 (2005).
Chen, T.-C. Maintenance of summer monsoon circulations: A planetary-scale perspective. J. Clim. 16, 2022–2037 (2003).
Krishnamurti, T. N. & Subrahmanyam, D. The 30-50 day mode at 850 mb during MONEX. J. Atmos. Sci. 39, 2088–2095 (1982).
Chen, T.-C. & Yen, M.-C. The 40–50 day oscillation of the low-level monsoon circulation over the Indian Ocean. Mon. Wea. Rev. 114, 2550–2570 (1986).
Chen, T.-C., Wang, S.-Y., Huang, W.-R. & Yen, M.-C. Variation of the East-Asian summer monsoon rainfall. J. Clim. 17, 744–762 (2004).
Yen, M.-C. et al. Interannual variation of the fall rainfall in central Vietnam. J. Meteor. Soc. Jpn. 89A, 259–270 (2011).
Latif, M. T. et al. Impact of regional haze towards air quality in Malaysia: a review. Atmos. Environ. 177, 28–44 (2018).
Xie, B., Yang, Y., Wang, P. & Liao, H. Impacts of ENSO on wintertime PM2.5 pollution over China during 2014–2021. Atmos. Ocean. Sci. Lett. 15, 100189 (2022).
Zeng, L. et al. Intensified modulation of winter aerosol pollution in China by El Nino with short duration. Atmos. Chem. Phys. 21, 10745–10761 (2021).
Hersbach, H. & Dee, D. ERA5 reanalysis is in production. ECMWF Newsl. 147, 7 (2016).
Chen, T.-C. & Yen, M.-C. A study of the diabatic heating associated with the Madden-Julian Oscillation. J. Geophys. Res. 96, 13163–13177 (1991).
Ashouri, H. et al. PERSIANN-CDR: Daily precipitation climate data record from multisatellite observations for hydrological and climate studies. Bull. Am. Meteor. Soc. 96, 69–83 (2015).
Bui-Manh, H. et al. The second rainy stage onset in the Central Highlands of Vietnam. Geophys. Res. Lett. 48, e2021GL093107 (2021).
Kendall, M. G. Rank Correlation Methods. 4th edition. (Charles Griffin, London, 1975).
Sen, P. K. Estimates of the regression coefficient based on Kendall’s tau. J. Am. Stat. Assoc. 63, 1379–1389 (1968).

Acknowledgements

This study is supported by the Ministry of Science and Technology, Taiwan, under grant MOST-111-2111-M-008-024, and by the Environmental Protection Administration, Taiwan. Data archives and typing support provided by Ms. M-.C.Y. are highly appreciated.

Ethics declarations

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.

About this article

Cite this article

Fu, YT., Yen, MC., Lin, NH. et al. Footprints of El Niño La Niña on the evolution of particulate matter over subtropical Island Taiwan. npj Clim Atmos Sci 6, 42 (2023). https://doi.org/10.1038/s41612-023-00383-6
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41612-023-00383-6