Meteorological influences on airborne pollen and spores in Johannesburg (Gauteng), South Africa

Airborne fungal spores and pollen (aerospora), synergistic with air pollution, are key triggers of allergic respiratory diseases. Effective diagnosis and treatment requires up-to-date location-specific knowledge on the temporal variability of aerospora types and levels. Johannesburg is the largest city in South Africa and has grown substantially in three decades, with changes in ground cover, population density and air pollution, yet until now, no continuous aerospora sampling has occurred. We present a daily two-year (August 2019–July 2021) aerospora assemblage for Johannesburg and explore temporal characteristics of 13 dominant aerospora in relation to daily meteorological variables (pressure, rainfall, relative humidity, temperature and wind characteristics). February–July, July–September and January-July represent high-risk periods for fungal spores [(Alternaria alternata (Fries. ex Keissler), Ascospores, Aspergillus niger (Van Tieghem), Penicillium chrysogenum (Thom), Cladosporium graminum (Corda), Epicoccum nigrum (Link), Helminthosporium solani (Durieu and Montagne) Nigrospora sphaerica (Saccardo ex. Mason), Smuts Ustilago nuda (Jensen ex. Rostrup) and Torula herbarum (Link)], trees (Cupressus, Morus and Platanus) and grass (Poaceae), respectively. Using a generalised additive model, results show that daily meteorological characteristics explained 7–32% of daily aerospora variability, with the largest effect on tree pollen. Rainfall, relative humidity and temperature influenced daily fungal spore and Poaceae counts, with moderate/low rainfall (< 20 mm), higher/mid-ranging relative humidity (~ 40–60%) and temperatures of ~ 15–20 °C associated with higher counts during high-risk periods. Rainfall predominantly influenced tree counts during high-risk periods, with higher counts occurring on low rainfall (<10 mm) days. These results update the aerospora profile of Johannesburg, South Africa, providing important information to inform allergy care.


Introduction
The atmosphere contains many types of suspended particles of organic and inorganic origin, which have a large diversity in size, shape and density, and are from diverse sources.Among these are airborne fungal spores and pollen (termed aerospora hereafter), which are inhalable particles, triggering allergic reactions including allergic rhinitis, conjunctivitis, asthma, hay fever and atopic dermatitis (Ajikah et al., 2020).Aerobiological investigations have revealed that aerospora are the most dominant, pervasive, respirable and potent sources of atmospheric allergens, giving rise to morbidity among hypersensitive individuals (Agarwal et al., 1974;Ajikah et al., 2020).Because of this, much research has been undertaken to understand factors influencing the quantities and seasonality of aerospora, where factors including geographic region, plant flowering periods and plant responses to meteorological variables, such as rainfall, relative humidity and temperature, are considered strong influential factors (Bishan et al., 2020;Grinn-Gofroń et al., 2020).However, changes in meteorological variables, resulting from anthropogenic influences on atmospheric carbon dioxide levels and enhancement of the greenhouse effect, also affect plant growth and seasonality and can subsequently cause an increase in aerospora production (Damialis et al., 2019).These changes not only affect the increase in aerospora production, but also the allergenic capacity of aerospora, with several studies indicating that the same aerospora species has more allergenic proteins if it grows in cities than in rural areas (Damialis et al., 2019).
Anthropogenically induced climate change influences the bioavailability and potency of allergenic aerospora and adjuvants in multiple ways, including changes in vegetation cover, phenology, pollination and sporulation periods and chemical modifications (Damialis et al., 2019).Changing climatic conditions and even air pollutants may skew physiological processes and the immune system towards development/ aggravation of allergies (Skjøth et al., 2021).Air pollution is high within South Africa, especially across the central western interior (i.e. the Highveld), which includes Johannesburg within the Gauteng Province, due to the large degree of coal powered plants, mining, petrochemical industries and coal-burning households (Adesina et al., 2022;Lourens et al., 2012).Within the Highveld persistent anticyclonic systems trigger subsidence of air masses, producing increased stability with low precipitation and frequent dry periods where air pollutants, including aerospora, can be trapped by inversions (Turner et al., 1995).Beyond this, as a consequence of anthropogenically induced climate change certain exotic plants which produce highly allergenic pollen (e.g.Ambrosia artemisiifolia) are spreading in South Africa, while neophytic trees (e.g.Cupressus and Platanus), which negatively impact public health due to production of high quantities of allergenic pollen, are widespread as ornamentals in Gauteng (Ajikah et al., 2020;Berman et al., 2020;Cariñanos et al., 2019).Mitigating this huge health and climatic anomaly and reversing the effects of these changes are major challenges, increasing the burden on health care systems.
Despite the abundance of aeroallergens, little attention has been given to their distribution within Johannesburg's atmosphere in almost 30 years.A new monitoring project in Johannesburg has, however, been initiated by the South African Pollen Network (SAPNET), offering the opportunity to further understand the characteristics of aerospora within Johannesburg's air (Ajikah et al., 2020).Hence, this research aims to report detailed data collected by Johannesburg's SAPNET monitoring station to define temporal characteristics of dominant aerospora within Johannesburg for August 2019-July 2021.A further aim, for the benefit of allergy sufferers and medical practitioners, is to explore the relationship/association between dominant aerospora and basic meteorological factors for this same period (i.e.August 2019-July 2021).

Study area
Johannesburg is a highly urbanised city located at ~ 26.1°S and ~ 28.1°E within the Gauteng Province, across South Africa's high lying interior plateau (i.e. the Highveld) at ~ 1700 m asl (Fig. 1a).Johannesburg forms part of the Gauteng City-Region, an urban conglomeration of cities and towns, which represent South Africa's economic hub, responsible for a third of the country's Gross Domestic Product (GCRO, 2018).
Due to its subtropical location, Johannesburg is characterised by a warm subtropical to temperate climate with cool, dry winters and warm, wet summers, classified as Cwb of the Köppen-Geiger climate types (warm temperate, winter dry, warm summer; Lennard, 2019).Johannesburg's mean annual temperature is ~ 18 °C, and mean summer (December-February) and winter (June-August) temperatures are ~ 20 °C and ~ 11 °C, respectively, while mean annual rainfall is ~ 700 mm (Landman et al., 2017).Most of Johannesburg's rain is delivered by convective weather systems during warmer, more humid October-March/ April months, having an average rainfall of ~ 600 mm (Dyson et al., 2015;Roffe et al., 2020).During these warmer months, southeasterly-to-northeasterly winds transporting moisture from the warm southwest Indian Ocean are dominant (Kruger et al., 2010).Remaining cooler months are also characterised by dominant southeasterly-to-northeasterly winds, but winds are generally weaker, and the subtropical high-pressure belt exerts a stronger influence promoting anticyclonic circulation with higher atmospheric pressure and stability (Kruger et al., 2010).

Aerospora sampling and pre-processing
In August 2019, the 7-day Hirst-type volumetric pollen and spore sampler/trap, designed in 1952 by Dr James Hirst and manufactured by Burkhard, was set up on the roof of the Richard Ward building (~ 12 m above ground at the top of a watershed) at the University of the Witwatersrand, forming part of the SAP-NET aerospora monitoring network (Fig. 1b; Ajikah et al., 2020;Berman et al., 2020); this type of sampler is a standard aerospora monitoring device (e.g.Ajikah et al., 2021;Berman, 2018;Fang et al., 2018;Rahman et al., 2020).For sampling, aerospora are continuously collected by this spore trap, using a constant air intake speed of 10 l min −1 and drum rotation of 2 mm.hr −1 (see Berman, 2018 for additional information).This sampler has a suction slit impactor, wherein air is drawn into a 14 × 2 mm orifice at a rate of 10 l.min −1 , while the head rotates into the oncoming wind utilising fins and the 10 mm × 3.7 µm pollen-trapping orifice which traps the smallest spores (Berman, 2018).
Daily aerospora samples collected for August 2019-July 2021 were used for analyses.To obtain these samples, pollen and spore grains are trapped on Melinex tape fixed on a rotating drum coated with vaseline jelly.This is then cut into seven daily segments and mounted on slides, measuring 26 × 76 mm and coverslips 24 × 50 mm, with a glycerol jelly media.Thereafter, daily counts of the number of pollen grains and spores were examined using A ZEISS Axioscope 5 microscope with a camera, at a magnification of × 400 along three full lengthwise traverses, for examination and photomicrography to count the pollen and fungal spores.Aerospora were then identified based on distinct morphological characteristics as well as published works and reference collections of fungal spore and pollen grains (e.g.Berman, 2018;Scott, 1982).
Overall, the resulting aerospora dataset for August 2019-July 2021 has ~ 20% missing data, due to South African lockdown periods relating to the Covid-19 pandemic (see supplementary information for details).Although we checked data quality, no gapfilling procedure was undertaken as the accuracy of this cannot be properly assessed given that the short data period and no reference data omits any signals of interannual variability in daily counts, which undoubtedly occur (Ordman and Etter, 1956;Berman et al., 2020;Supplementary Fig. 1).While the data gaps potentially influenced the quality of our results, it is important to highlight that with additional years of data, missing data can be imputed accurately to produce more complete records for future research and continuous monitoring.

Meteorological data and pre-processing
Meteorological data, spanning August 2019-July 2021, were obtained from the South African Weather Service (SAWS) for the Johannesburg Botanical Gardens weather station, located ~ 5 km from the spore trap (Fig. 1b).This data includes daily rainfall (mm) and daily maximum and minimum temperature (°C) records, and sub-daily records (measured at 08:00, 14:00 and 20:00) of relative humidity (%), surface air pressure (hPa), wind direction (°) and wind speed (m s −1 ).All variables had relatively complete raw datasets, with < 10% missing data.
Before analysis, raw data were scrutinised for outliers and errors, which were deleted, and cleaning was performed where necessary (Durre et al., 2010).Subsequently, we imputed all missing/deleted values, on the raw data temporal scales, using a weighted ratio considering nearby station records with a Pearson correlation coefficient of > 0.70 (Wolski et al., 2021).To produce daily values for analyses, daily maximum and minimum temperatures were averaged to quantify Fig. 1 Map of the study area depicting a Johannesburg (JHB) within South Africa's Gauteng province and Grassland biome, and b the location of the Johannesburg Botanical Gardens weather station (26.156°S 27.999°E at 1642 m asl) and the University of the Witwatersrand 7-day volumetric spore trap (26.193°S 28.030°E at 1758 m asl) sampling sites surrounded by a 30 km buffer around the spore trap (buffer size after Grinn-Gofroń et al., 2020).Biome data are sourced from the South African National Biodiversity Institute and represent the biome distribution for 2018 (Biome data from: SANBI, 2018).Land use and land cover (LULC) data are sourced from the South African Department of Forestry, Fisheries and the Environment and represent that recorded for 1 January to 31 December 2020 (LULC data from: DFFE, 2021) ◂ daily average temperature, while sub-daily records were averaged to calculate daily relative humidity, surface air pressure, wind direction and wind speed.

Statistical analyses
Statistical analyses were undertaken using daily data, considering dominant aerospora within Johannesburg's assemblage.Varying definitions exist to identify dominant taxa.As examples, Murray et al. (2008) defines dominant taxa as those contributing ≥ 0.5% of the total sample, while Bicakci et al. (2017) and Tosunoglu et al. (2018) define dominant taxa as those contributing ≥ 3% of the total sample.Here, dominant taxa are defined as those contributing ≥ 1% of Johannesburg's assemblage for August 2019-July 2021 (Table 1).
We first explored daily and monthly characteristics using descriptive statistics of the dominant aerospora types and meteorological variables.Although studies typically start analyses by defining the main aerospora season (e.g.Grinn-Gofroń et al., 2020;Tosunoglu et al., 2018), we dismiss this option because missing data would influence the season timing and length; however, from our monthly scale analyses seasonality characteristics are evident; when more data are available, we suggest that aerospora seasons are investigated using a percentile-based method (e.g.Grinn-Gofroń et al., 2020).Notably, flowering seasons are often weakly defined in southern Africa compared to temperate Europe or North America regions, as many plants have varying blooming dates, mainly determined by rainfall and temperature (Johannsmeier, 2016).
Due to non-normal data distributions (Supplementary Fig. 1, 2), we applied nonparametric statistics to model how meteorological variables influence daily aerospora counts.The R mgcv package was used to apply a generalised additive model (GAM) to model combined and individual meteorological effects (Wood, 2017).A GAM is applied as this form of multivariate regression allows for consideration of linear and nonlinear influences on the response variables (Hammouda et al., 2021).Due to large variability in daily aerospora counts, resulting in overdispersion (Supplementary Fig. 2), a negative binomial distribution with a log link is assumed for the GAM (Puig and Valero, 2006;Wood, 2017).In building a GAM for each response variable, all meteorological variables were considered as explanatory variables.Day of the year (Julian day) was also included as an explanatory variable to account for temporal variability in aerospora levels.
Before building each GAM, multicollinearity was assessed because some explanatory variables are significantly correlated (Supplementary Table 1).The variance inflation factor (VIF) quantifies the degree of multicollinearity and indicates whether multicollinearity influences an explanatory variable model coefficient (Hammouda et al., 2021;Ma et al., 2020).Thus, VIF values were quantified for each explanatory variable.Variables with VIF values closer to 1 indicate low multicollinearity and are retained in a GAM, while large VIF values > 4 are typically excluded as these indicate high multicollinearity which negatively influences the GAM stability and goodness of fit (Hammouda et al., 2021;Ma et al., 2020).VIF values are low, ranging from 1.14 (day of the year) to 1.83 (pressure), thus no explanatory variables were excluded (Supplementary Table 2).
To build each GAM, we employed a forward stepwise variable inclusion process which begins with an empty model and follows with introducing explanatory variables one at a time (Ma et al., 2020;Wood, 2017).All models were developed based on the restricted maximum likelihood (REML) standard, argued to perform well in smoothing (Wood, 2017(Wood, , 2021)).Day of the year was the first variable introduced with a thin plate spline smoothing function.Thereafter, the meteorological variables were added, one at a time, again using the thin plate spline smoothing function.The resulting baseline models for each response variable can be written as: log E y i represents the join function and y i rep- resents each response variable.DOY, P, R, RH, T, WD and WS represent the day of the year (as Julian day), pressure, rainfall, relative humidity, temperature, wind direction and wind speed, respectively.The smoothing function for each variable is represented by s().The k term for each explanatory variable denotes the number of knots required for each smoother, representing the number of joins between each spline curve (Wood, 2017).Each k term was determined throughout the model fitting, where these were left to be automatically set by the GAM, unless fitting of additional explanatory variables influenced the variable fit (Hammouda et al., 2021;Wood, 2017).Because the relationship with wind conditions, especially wind direction, is difficult to model (Hammouda et al., 2021), we additionally prepared polar plots (Carslaw and Ropkins, 2012), which depict patterns of aerospora transport to the spore trap and are useful to identify source locations (Grinn-Gofroń et al., 2020).

Results and discussion
3.1 Airborne fungal spores: temporal characteristics and relationship to meteorological variables Among the 85 airborne fungal spore and pollen (i.e.aerospora) taxa recorded for August 2019-July 2021, 23 were fungal spores, contributing 83.4% of the sample (Table 1).In order of contribution, Aspergillus/ Penicillium, Cladosporium, Ascospores, Alternaria, Epicoccum, Torula, Smuts, Nigrospora and Helminthosporium were dominant airborne fungal taxa, contributing 81% of the sample (Table 1).Within Johannesburg and surrounding areas, all of these taxa have similarly been identified among principal airborne fungal spores for two sampling campaigns collectively spanning 1950-1959 (Ordman, 1963;Ordman and Etter, 1956), and a later campaign spanning 1992-1994, which similarly demonstrated that fungal spores were most dominant in the sample (Cadman et al., 1997).
Allergenicity associated with most fungal spores is not yet well-understood, and fungal spores in aerobiological monitoring correlate poorly with sensitisation patterns.However, fungal spores, such as Alternaria and Cladosporium are typically found at high levels in aerobiological monitoring and are important allergens (Sousa et al., 2016;Grinn-Gofroñ et al., 2020).South African allergy studies, based on the skin prick test (SPT) results, conducted in Cape Town (Potter et al., 1991) and the Free State (Joubert, 2006) found Alternaria to be the most common fungal sensitiser.Also, Alternaria according to Bousquet et al. ( 2012) is described as one of the main fungal aeroallergens around the world.The Cape Town study, by Potter et al. (1991), also highlighted a high level of allergenicity associated with Epicoccum purpurascens.Among the remaining dominant fungal taxa, known allergens of Torula (Rhodotorula), for instance, include Rho m 1, Rho m 2 (Simon-Nobbe et al.,  4; Supplementary Fig. 1a-j), they may be allergenic.Nonetheless, at low levels, it is suggested that < 20% of those with fungal sensitisation would be expected to experience significant symptoms (SAPNET, 2019).
The annual behaviour of airborne spore and pollen (i.e.aerospora) types") below depicts when aerospora levels are maximal and most likely pose risk to sensitised persons (Berman et al., 2020).Monthly bar graphs and daily GAM plots depict the annual behaviour for each of the dominant taxa (Figs. 2,3,4,5).Focusing on the monthly bar graphs, because patterns are similar to the daily GAM plots, it is evident that among the airborne fungal taxa, Torula's annual behaviour was most distinct, with higher counts observed during autumn-spring months of April-October, while counts peaked during winter in June-July when daily counts almost consistently exceeded the 90th percentile level (Fig. 2i; Supplementary Fig. 1i).Alternaria and Cladosporium depicted nearly identical seasonal regimes.Observed counts were higher during late summer and autumn from February-May, and during March-April counts peaked (Fig. 2a, d; Supplementary Fig. 1a, d).While their counts dropped to relatively low levels during winter, a slight increase was observed during October-November spring months (Fig. 2a, d; Supplementary Fig. 1a, d).Ascospores, Aspergillus/ Penicillium, Nigrospora and Smuts depicted a spore season with higher monthly and daily counts during autumn-winter months of February-July, where daily counts were generally consistent during this period (Fig. 2b, c, g, h; Supplementary Fig. 1b, c, g,  h).Epicoccum and Helminthosporium demonstrated relatively shorter seasons of spore release, which were primarily concentrated during February-April (March-May) for Helminthosporium (Epicoccum) when daily counts frequently exceeded the 90th percentile level and peaked during March (April) for Helminthosporium (Epicoccum; Fig. 2e, f; Supplementary Fig. 1e, f).
Although the annual behaviours differ across the airborne fungal taxa, an overall spore production season for these taxa can be identified for Johannesburg.This spans February-July, from late summer through to autumn and early winter (Fig. 2a-i; Supplementary Fig. 3b, 4).This annual behaviour data is consistent with that depicted for Johannesburg in two previous studies (Ordman, 1963;Ordman and Etter, 1956).Thus, Johannesburg's fungal spore season generally spans the end of the wet-season into the dry-season, when daily rainfall is typically lower, pressure is relatively high due to dominant anticyclonic circulation, daily temperatures are generally mid-range, relative humidity levels are generally mid-range and prevailing southeasterly-to-northeasterly winds are relatively slow (Kruger et al., 2010;Dyson et al., 2015;Landman et al., 2017;Lennard, 2019;Roffe et al., 2020;Fig. 6;Supplementary Figs. 5, 6).
To explore how meteorological variables influence the dominant airborne fungal taxa, we consider the GAM results.These results consistently demonstrate that meteorological variables exerted a weak influence, possibly due to the short study period and the fact that many meteorological variables exerted nonlinear influences (Table 2; Figs. 3, 4).The edf values and graphical results reflect a range of nonlinear, near-linear and linear influences of the meteorological variables on daily airborne fungal spore counts (Table 2;Figs. 3,4,5).In most instances, pressure and temperature exerted nonlinear influences, while rainfall and relative humidity generally exerted near-linear influences, and wind conditions generally exerted linear influences (Table 2; Figs. 3,  4).Overall, however, meteorological variables explained only 6.8% (Smuts) to 18.0% (Epicoccum) of the variability in daily airborne fungal spore levels (Table 2).This low influence of meteorological variables is consistent with studies conducted in subtropical and temperate climate regions (e.g.Oliveira et al., 2009;Ivanovici, 2016;Berman, 2018;Antón et al., 2019;Grinn-Gofroń et al., 2020).Collectively, individual meteorological variables exerted a similar influence on the airborne fungal spores (Figs. 3, 4; Fig. 2 Average monthly sum/count of dominant aerospora taxa for August 2019-July 2021.Colours represent the taxa categories, where blue green, dark green and yellow green represent fungal spore, tree and grass types, respectively.Numbers per month represent the number of days when pollen counts exceeded the moderate category of the Burge scale category (Supplementary table 4), and when fungal spore counts exceeded the 90th percentile level (Supplementary Table 3) ◂ Supplementary Fig. 4a).However, for specific taxa, particular variables were more important, as is highlighted by the significant F-values in Table 2, where significant F-values provide confidence that respective explanatory variables influenced daily counts, while insignificant F-values for meteorological variables suggest they exerted little influence (Hammouda et al., 2021).Among the meteorological variables, rainfall, relative humidity and temperature were dominant variables, significantly influencing at least four of the dominant airborne fungal taxa (Table 2).Overall, considering that rainfall and temperature were almost always dominant explanatory variables (significant for seven of the fungal taxa), it should be highlighted that studies typically find higher fungal spore release during relatively dry, warm periods as is generally evident from our results (Figs. 2a-i, 4); this has similarly been highlighted for taxa such as Alternaria (O'Hollaren et al., 1991;Weber, 2001), Ascospores (Haskouri et al., 2016), Cladosporium, Alternaria, Epicoccum, Smuts and Torula (Crotzer and Levetin, 1996;Grinn-Gofroñ et al., 2018 and references therein).
Individually, meteorological variables exerted relatively similar influences on the fungal spore types (Figs. 3, 4).Daily counts of Alternaria, Ascospores, Aspergillus/Penicillium, Cladosporium, Epicoccum, Helminthosporium and Nigrospora were approximately similar across varying daily pressure; however, for pressures ranging from ~ 840 to 845 hPa, increased counts were generally observed (Figs. 3,4).Although this pattern is similarly evident for daily Smuts and Torula counts, the GAM relationships reflect higher daily pressures corresponding to higher daily counts (Fig. 4c, d).Notably, few studies consider air pressure in relation to fungal spore levels; however, some results similarly suggest an association between relatively higher daily fungal spore levels and higher pressure (Burch and Levetin, 2002); this is somewhat expected given that anticyclonic systems, trapping pollutants, are dominant when pressure is higher (Kruger et al., 2010;Turner et al., 1995).Despite nonlinear influences for some taxa, all fungal spores had higher daily counts for days where little/no rainfall was recorded (Figs. 3, 4), likely due to a washing out effect, where rainfall promotes the deposition of fungal spores from the atmosphere (Ivanovici, 2016).Daily fungal spore counts typically increased with higher relative humidity levels; however, for Smuts and Torula the opposite pattern is evident (Figs. 3, 4).Overall, however, most fungal spores had higher counts for mid-ranging relative humidity levels, consistent with that reported in other studies highlighting that a certain degree of dryness is required in the atmosphere for the release of fungal spores (Ivanovici, 2016).All fungal spore types exhibited near-linear to nonlinear patterns with varying daily temperatures (Figs. 3, 4).Daily counts for Alternaria, Ascospores, Aspergillus/Penicillium, Cladosporium, Epicoccum and Helminthosporium were higher for temperatures ranging from ~ 15 to 20 °C and for temperatures outside of this range, counts were relatively consistent and lower (Figs. 3, 4a, compare Grinn-Gofroñ et al., 2018).Daily Nigrospora, Smuts and Torula counts followed a relatively different pattern, depicting a general decrease with increasing temperatures (Fig. 4b, c, d).Overall, however, mid-ranging temperatures, of ~ 15-20 °C, were associated with higher airborne fungal spore counts (Figs. 3, 4; Table 2; Supplementary Fig. 4a), which is somewhat consistent with the notion that higher temperatures promote fungal spore release (Grinn-Gofroñ et al., 2018).Among the various meteorological variables considered, rainfall, relative humidity and temperature are typically the strongest influencing factors on fungal spore dispersal, and while influences are generally location-specific, the results for studies within subtropical and temperate climate regions broadly correspond to that depicted for Johannesburg (Crotzer and Levetin, 1996;O'Hollaren et al., 1991;Weber, 2001;Oliveira et al., 2009;Haskouri et al., 2016;Ivanovici, 2016;Berman, 2018;Antón et al., 2019;Grinn-Gofroñ et al., 2018, 2020).
The GAM relationships and polar plots demonstrate somewhat diverse responses of airborne fungal spores to varying wind conditions (Figs. 4,5,7).Indeed, difficulty in modelling responses to varying wind conditions is highlighted by these results as relatively little agreement exists across these for each fungal spore.Nonetheless, we can interpret some interesting response patterns.Considering higher densities of points on the GAM plots for all fungal spores, it appears that higher daily counts mostly corresponded to wind speeds < ~ 3 m s −1 (Figs. 4, 5; Table 2).Several studies considering similar fungal spores also depict that slower wind speeds are associated with higher airborne fungal spore levels (Ivanovici, 2016;Antón et al., 2019;Grinn-Gofroñ et al., 2020); however, studies also report higher fungal spore air loads in association with stronger winds (Wu et al., 2004;Grinn-Gofroñ et al., 2018;Oneto et al., 2020).Higher densities of points on the GAM plots also depict that wind directions ranging from 0 to 150° mostly coincided with higher counts (Figs. 4, 5); a result dependent on fungal spore source locations (Grinn-Gofroñ et al., 2020).As for the polar plots, these reveal relatively similar patterns of transport to the spore trap for Alternaria, Cladosporium, Epicoccum, Helminthosporium and Smuts, which suggests similar source regions generally spanning northeast and southwest of the spore trap, given that higher counts are evident for winds blowing from these directions (Figs. 1b,7a,d,e,f,h).The polar plots also depict approximately similar patterns of transport to the spore trap for Ascospores, Aspergillus/Penicillium and Nigrospora, suggesting that regions to the northeast and southeast-to-southwest of the spore trap were likely sources as winds blowing from these directions generally recorded higher daily counts (Figs.1b, 7b, c, g).For Torula, the polar plot suggests regions southwest of the spore trap were likely sources as southwesterly winds coincided with higher daily counts (Figs. 1b,7i).Although further research is needed to identify specific fungal spore source regions, considering results presented by Grinn-Gofroñ et al. (2020), it is likely that cultivated, forest, grassland, and mining and quarry areas spanning northeast-to-southwest regions of the spore trap were sources of spores captured by the spore trap (Fig. 1b).
Collectively, daily rainfall, relative humidity and temperature were the most dominant influencing factors on daily airborne fungal spore counts (Table 2; Supplementary Fig. 4a).Thus, for fungal spores, days with northeasterly-to-southwesterly originating winds (given that northeast-to-southwest regions represent predominant source regions; Fig. 7a-i), moderate/low rainfall (< 20 mm), mid-to-higher relative humidity levels (> 40%) and mid-range temperatures (~ 15-20 °C) represent potentially risky days for allergy sufferers (Figs. 3, 4; Supplementary Fig. 4a).Rainfall, relative humidity and particularly temperature are typically identified as dominant factors influencing daily fungal spore levels (e.g.Oliveira et al., 2009;Haskouri et al., 2016;Ivanovici, 2016;Berman, 2018;Antón et al., 2019;Grinn-Gofroñ et al., 2020).While days with relatively low rainfall and mid-ranging relative humidity are consistently suggested to result in higher airborne spore levels particularly due to wet deposition associated with wetter days, higher temperatures are predominantly reported as being associated with higher spore levels (Oliveira et al., 2009;Haskouri et al., 2016;Ivanovici, 2016;Berman, 2018;Antón et al., 2019;Grinn-Gofroñ et al., 2020).This slightly contradicts the relationship we identified for temperature and fungal spore levels, which can be explained by considering that Johannesburg's wetseason occurs during warmer summer months when wetter days can do coincide with higher temperatures (Supplementary Table 1).

Tree taxa: temporal characteristics and relationship to meteorological variables
We recorded 29 tree taxa, contributing 11.7% of the sample (Table 1).Platanus, Morus and Cupressus were dominant tree taxa, contributing 8.6% of the sample (Table 1).These dominant tree taxa were among the principal airborne pollen taxa within Johannesburg's, and surrounding areas, air during 1987during -1988during and 1992during -1994during (Cadman, 1990during , 1991;;Cadman et al., 1997).They are of particular concern, as they are all highly allergenic neophytes which produce high quantities of pollen over a short period (Hidalgo et al., 1999;Papia et al., 2020;Vrinceanu et al., 2021).The dominant tree taxa Platanus has been identified as a tree allergen from SPT and IgE testing panels conducted in South African allergy clinics as well as in other European cities (Joubert, 2006;van Rooyen et al., 2020).Individually, trees are suggested to be a less common cause of seasonal allergies in South Africa, but collectively their high  2 Results of the generalised additive models depicting the effect of the explanatory meteorological variables on the daily counts of the dominant taxa for Johannesburg for August 2019-July 2021.Edf represents effective degrees of freedom (edf >1 indicates nonlinear relationships, and the larger the edf value, the more nonlinear the relationship is; Ma et al., 2020), while the F-value is an approximate of the F-test which quantifies the strength of influence an explanatory variable has on the response variable (Hammouda et al., 2021).The statistically significant results are denoted with an asterisk for the 5% alpha level.levels are problematic for allergy sufferers (Ajikah et al., 2020).Moreover, there is significant crossreactivity between mulberry pollen and other members of the Moraceae family, where Morus pollen can induce asthma and allergenic rhinitis (Navarro et al., 1997;Papia et al., 2020).Furthermore, allergen Cup s 3 is identified in Cupressus sempervirens and across southern Europe, for instance, Cupressaceae are regarded as a major driver for pollinosis (Togawa et al., 2006).This major contribution of tree pollen to the aerospora in Johannesburg is among the highest compared to the other cities monitored in the SAP-NET project (Berman et al., 2020), reflecting the high concentration of trees within the Johannesburg region (Fig. 1b); those tree allergens should be added to Johannesburg's allergy testing panels and considered as important drivers for pollinosis in the city.Similar to previous studies conducted in subtropical and temperate climate regions (Berman, 2013(Berman, , 2018;;Celenk et al., 2010;Fang et al., 2018;Rahman et al., 2020;Tosunoglu et al., 2018), tree taxa in Johannesburg were characterised by very short pollen seasons of roughly two months during late winter to early spring, and outside of this season little/no counts were recorded (Fig. 2j-l; Supplementary Figs. 1 k-n,  3).Monthly and daily counts for Morus and Platanus were highest during August-September (Fig. 2k, l; Supplementary Fig. 1 k, l), while that for Cupressus was highest during July-August (Fig. 2j; Supplementary Fig. 1 k).Within Johannesburg, Cadman (1990Cadman ( , 1991) ) similarly demonstrated shorter pollen seasons for Cupressus (cypress), Morus (mulberry) and Platanus (plane), ranging from July-October, highlighting that despite interannual variability, the July-September period we identified is likely the most risky for those sensitised to tree pollen.July-September tree counts were frequently at moderate and/or high Burge scale levels (Supplementary Fig. 1j-l).At moderate levels, at least 50% of tree pollen allergy sufferers are expected to experience symptoms associated with allergenic respiratory diseases, while this level increases to at least 90% for high levels, when outdoor activities should be limited (SAPNET, 2019).
Overall, Johannesburg's July-September tree pollen season coincides with the end of its dry-season when little/no rainfall occurs, air pressure is at a midto-high range, relative humidity is low, temperatures are mid-range and prevailing southeasterly-to-northeasterly winds are flowing relatively slow-to-mid range (Kruger et al., 2010;Dyson et al., 2015;Landman et al., 2017;Lennard, 2019;Roffe et al., 2020;Fig. 6;Supplementary Figs. 5, 6).During this period, respiratory diseases are reported to peak in Johannesburg (Thabethe et al., 2021).Compared to other South African regions, Johannesburg typically has high respiratory disease occurrences, probably related to the high degree of industrialisation (Thabethe et al., 2021); this underlines the importance of exploring linkages between respiratory diseases, air pollution and aerospora levels, particularly across industrialised areas of South Africa.
Tree taxa were most strongly influenced by meteorological variables, with the explained deviance ranging from 27.0% (Cupressus and Morus) to 31.8% (Platanus; Table 2), largely the result of the very short tree pollen season.Rainfall was the only significant meteorological variable influencing all three dominant tree taxa counts (Table 2).Cupressus was most promintently present from July-August when rainfall is low (Supplementary Figs. 1k,4b), aligning with the previously recorded Cupressus flowering season for southern Africa (Johannsmeier, 2016).Temperature was the other significant meteorological variable influencing Platanus counts, while Morus counts were also significantly influenced by pressure, relative humidity and wind direction (Table 2).Both taxa depicted the highest counts in August-September before the start of the rainy season (Fig. 6b), consistent with previously reported flowering seasons (Fig. 2k, l;Johansmeier, 2016).Despite similar pollen season timing, the GAM results differed for the strongest influencing meteorological variable across the tree taxa (Table 2).For Cupressus, this was rainfall, while for Morus (Platanus) this was wind direction (temperature; Table 2).Despite these differences, the patterns of influence of the meteorological variables were approximately similar.Near-linear patterns were observed for the influence of pressure, where higher daily counts generally corresponded to higher observed pressure (Table 2; Fig. 5a-c).All tree taxa had a near-linear response of higher daily counts for days where little/no rainfall was recorded (Table 2; Fig. 5a-c).Generally linear patterns were modelled for the response to relative humidity, where all tree taxa had higher counts recorded for days with lower relative humidity levels (Table 2; Fig. 5a, b, c).Nonlinear patterns were modelled for the response to temperature (Table 2).For Cupressus, GAM results Vol.: (0123456789) reflect lower temperatures generally corresponded to higher daily counts (Table 2).Morus and Platanus counts were, however, generally consistent across varying temperatures, but counts increased for temperatures ranging between ~ 15 and 20 °C (Fig. 5a-c).Although the polar plots depict the clearest pattern for wind conditions and tree count variability (Fig. 7), the GAMs appeared to struggle with modelling the response to wind conditions (Fig. 5a-c).Nonetheless, if the high density of points is considered for wind directions ranging from ~ 0 to 100° and wind speeds ranging from ~ 0-3 m s −1 , then a similar pattern is depicted for the GAM and polar plot results .Interpreted together, these suggest northeasterly winds of ~ 2-3 m.s −1 corresponded to higher daily tree counts (Figs.5a-c, 7j-l); a finding corroborated when considering the dominant tree taxa collectively (Supplementary Fig. 4b, d).Thus, northeastern forested regions were likely tree pollen sources, however, for Cupressus evidence for a second source exists, being southwest forested regions (Figs. 1b,7j).These source areas correspond well to surrounding regions containing a high density of trees (Fig. 1b).
Studies considering meteorological variables associated with higher daily levels for tree taxa are generally consistent with our results for higher counts associated with drier, less humid, cooler and calm days during the flowering season, however, in some instances significant positive relationships are detected for temperatures (e.g.Ҫeter et al., 2012;Bicakci et al., 2017;Fang et al., 2018;Tosunoglu et al., 2018;Bishan et al., 2020).Overall, rainfall was the most dominant influencing factor across all tree taxa (Table 2), summarily suggesting that drier days (< 10 mm) are most likely to coincide with higher tree counts (Supplementary Fig. 4b); similar to the fungal spores, the relationship between Johannesburg's rainfall and temperature likely explains this discrepancy as rainfall plays a strong role in the deposition of tree pollen grains from the atmosphere (Tosunoglu et al., 2018).

Poaceae: temporal characteristics and relationship to meteorological variables
Among the pollen types of herbaceous plants analysed (2.9% of sample), Poaceae was the only dominant taxon, contributing 2.8% (Table 1).Previous studies for subtropical and temperate climate regions have similarly identified Poaceae as a dominant taxon (e.g.Bishan et al., 2020;Tosunoglu et al., 2018).Since Poaceae is a major pollen producer within the Johannesburg grassland region, it is not surprising that it was identified as a dominant taxon in this sample and previous samples from Johannesburg and surrounding regions (Cadman, 1990(Cadman, , 1991;;Cadman et al., 1997).Indeed, grasses are reported as major pollen allergens in many subtropical and temperate climate regions (D'Amato, 2007), and within South African allergy clinics, SPT test rankings identify grass pollen as a dominant aeroallergen (van Rooyen et al., 2020).Seasonally, Poaceae, depicted higher levels during January-July, when daily counts frequently exceed the moderate and high Burge scale threshold levels, with peak counts concentrated during February-April when high levels were most frequent (Fig. 2m; Supplementary Fig. 1o).This is similar to the Poaceae pollen production season of October-May detected for Johannesburg for 1987-1988(Cadman, 1990, 1991)), though our shorter season may relate to urban sprawl, with buildings replacing open spaces which previously contained grasses, among other vegetation.Overall, the Poaceae season detected here coincides with the latter portion of Johannesburg's wet-season and much of its dry-season, when daily rainfall ranges from little/no rain to relatively high rainfall levels, air pressure levels are relatively lowto-moderate, relative humidity and temperature levels are relatively moderate-to-low, and prevailing southeasterly-to-northeasterly winds range from relatively low-to-moderate speeds (Kruger et al., 2010;Dyson et al., 2015;Landman et al., 2017;Lennard, 2019;Roffe et al., 2020;Fig. 6;Supplementary Figs. 5, 6).
Overall, meteorological variables explained 15.1% of daily Poaceae variability (Table 2).All variables exerted a nonlinear or near-linear influence (Table 2; Fig. 5d).F-values reflect that rainfall, relative humidity and temperature exerted significant influences (Table 2).With the highest F-value, temperature exerted the strongest influence (Table 2).Although  2), a broad linear pattern can be interpreted, where higher daily temperatures corresponded to higher Poaceae counts, however, when temperatures exceeded ~ 20 °C, Poaceae counts decreased (Fig. 5d).This corresponds to other studies in subtropical and temperate climate regions, reflecting a clear linkage between higher temperatures and grass pollen release (Addison-Smith et al., 2021;Bishan et al., 2020).The GAM results reflect higher relative humidity levels corresponded to higher Poaceae counts in a relatively linear manner (Fig. 5d).This corresponds to southern Africa's end of summer-autumn grass flowering season, occurring approximately five weeks after plentiful rain occurs (Johannsmeier, 2016).Days recording little/no rainfall consistently corresponded to higher Poaceae counts   (Carslaw and Ropkins, 2012), depicting the relationship between daily wind conditions (direction and speed) and daily counts of the dominant taxa (Fig. 5d).Pressure and wind conditions had insignificant F-values, suggesting they exerted little influence (Table 2).Despite this, clear patterns of influence are evident.Higher pressure generally corresponded to lower counts, while pressure ranging from ~ 840 to 845 hPa corresponded to higher counts (Fig. 5d).In considering the relationship between daily wind conditions and Poaceae counts, it is notable that the polar plots and GAM results depict similar patterns.These results demonstrate that higher Poaceae counts corresponded to northerly-to-southerly winds (<150°) which were generally slower (< ~ 3-4 m.s −1 ; Figs. 5d, 7m; Table 2).Above this, the polar plot suggests northeastern and southeastern-to-southwestern grassland regions were likely Poaceae sources (Figs. 1b,7m).
Considering the dominant meteorological drivers (Table 2), we can highlight that higher Poaceae counts are most likely associated with moderately wet to drier days of < 20 mm often several weeks after rain, mid-ranging humidity levels of ~ 40-60% and mid-ranging temperatures of ~ 15-20 °C (Fig. 5d).Studies typically find similar meteorological conditions for days with higher Poaceae levels, but similar to the aforementioned dominant taxa, significant positive relationships have been identified for temperature (e.g.Ҫeter et al., 2012;Bicakci et al., 2017;Tosunoglu et al., 2018;Bishan et al., 2020).Again, the relationship between Johannesburg's temperature and rainfall likely explains this discrepancy as rainfall appears to play a strong role in washing away atmospheric Poaceae grains (Tosunoglu et al., 2018).However, this aspect requires more research as other subtropical and temperate region studies demonstrate a weaker/inconclusive linkage between grass pollen concentration and rainfall (e.g.Addison-Smith et al., 2021.

Conclusions and future research
This study provides the latest aerospora data within Johannesburg, presenting a daily two-year (August 2019-July 2021) aerospora assemblage, and exploring how dominant taxa relate to commonly monitored meteorological variables.Because South Africa is home to ~ 20 million allergy sufferers (Ajikah et al., 2020), and regional differences exist for dominant taxa and aerospora seasons (Berman et al., 2020;Grinn-Gofroń et al., 2020), this data can only inform allergy sufferers and their healthcare providers within Johannesburg to aid effective diagnosis and management of aeroallergen driven allergic respiratory diseases.
Similar to previous studies (e.g.Bishan et al., 2020;Grinn-Gofroń et al., 2020), our selected meteorological variables did not have major explanatory power to account for daily variation in the levels of dominant aerospora taxa.However, we did note that certain key meteorological variables, including rainfall and temperature, had a significant influence on levels of particular taxa.This highlights a need to consider a longer aerospora record as links to meteorological variables may then be more distinct and easier to identify.This also highlights a need to explore additional characteristics of the dominant meteorological variables (e.g.daily rainfall intensity and duration, and maximum and minimum temperatures) in relation to daily aerospora counts.To comprehensively understand which factors influence daily aerospora counts, further studies should also consider additional explanatory variables, for a GAM analysis, such as land use/cover, density and abundance of plant species, other meteorological variables (e.g.daily sunshine hours and cloudiness), and air pollution concentrations.Furthermore, we also found that certain meteorological conditions, corresponding with particular seasons within the city favoured higher counts of particular taxa.This highlights the need for the reporting of aerospora to be both more detailed, commenting on specific taxa and links to influencing meteorological factors, as well as caution in over generalisations about the influence of common meteorological variables on taxa.The lack of predictive power of meteorological variables also highlights the importance of real-time monitoring of aerospora.Further research is thus required, with additional monitoring in different parts of the city, and with collection of additional data on other potential co-variates, to allow for improved abilities to model and forecast aerospora data in South African cities.To this end, it is critical that SAPNETs database grows in data availability period and monitoring locations.In addition to exploring forecasting possibilities, the data is needed to study how anthropogenically induced climate change will influence the magnitude of aerospora production and length of

Fig. 3
Fig. 3 Scatter plots depicting the generalised additive model relationships for the influence of day of the year, pressure, rainfall, relative humidity, temperature, wind direction and wind speed on a Alternaria, b Ascospores, c Aspergillus/Penicillium, d Cladosporium and e Epicoccum.Grey-shaded areas denote 95% confidence intervals ◂

Fig. 4
Fig. 4 Scatter plots depicting the generalized additive model relationships for the influence of day of the year, pressure, rainfall, relative humidity, temperature, wind direction and wind speed on a Helminthosporium, b Nigrospora, c Smuts, and d Torula.Grey shaded areas denote 95% confidence intervals ◂

Fig. 5
Fig. 5 Scatter plots depicting the generalised additive model relationships for the influence of day of the year, pressure, rainfall, relative humidity, temperature, wind direction and wind speed on a Cupressus, b Morus, c Platanus, and d Poaceae.Grey-shaded areas denote 95% confidence intervals ◂

Fig. 6
Fig. 6 Average monthly values for August 2019 to July 2021 for a air pressure, b rainfall, c relative humidity, d temperature and e wind speed.Monthly wind direction characteristics are presented in supplementary Fig. 5

Fig. 7
Fig.7Polar plots, prepared using R's openair package(Carslaw and Ropkins, 2012), depicting the relationship between daily wind conditions (direction and speed) and daily counts of the dominant taxa

Table 1
Details of aerospora taxa contributing to Johannesburg's assemblage for August 2019 to July 2021.The top (bottom) portion of this table corresponds to the fungal spore (pollen) taxa, where bold text corresponds to the dominant aerospora taxa contributing ≥ 1% of the sample and used for further analyses Vol.: (0123456789)

Table 1
(continued) For explained deviance, values outside (inside) the brackets indicate that for the overall model (only the meteorological variables) *DOYDay of year, P