This study was conducted from March 2015 to February 2017 at nine sites that were chosen to explore atmospheric P deposition, spanning a transect covering urban areas (UAs), intensive agricultural areas (IAAs) and rural areas (RAs) in the southwestern Chengdu Plain (Fig. 1, Table S1, Deng et al., 2019). Urban areas, including Shangnan Street (SS), Yongkang Street (YS) and Xihe Bridge (XB) sites, are located in Chongzhou, which has 74.4 km2 of urban land and 130 000 permanent dwellers. Intensive agricultural areas, including the Caichang (CC), Liaoyuan (LY) and Qiquan (QQ) sites, covered 1.8 km2 of the agro-facility land-use type, which accounted for approximately 69.9 % of the total in nine sites. Rural areas, including Yuantong (YT), Liujie (LJ) and Huaiyuan (HY) sites, covered 13.59 km2 of forest, which accounted for approximately 96.2 % of the total in nine sites. The total area of one land-use type was calculated by adding the values in each column, as shown in Table 1, where each column indicates the area occupied by each land-use type in nine sites. More details about the study sites are shown in Table 1. The climate at the sites is subtropical monsoon humid, with monthly precipitation, ambient temperature, relative humidity and wind speed at nine sites varying from 0.6 to 238.63 mm, 5.83 to 27. ∘C, 66.0 % to 89.3 % and 0.5 to 1.80 m s-1, respectively. The meteorological data used in this study are from the Chongzhou Meteorological Bureau, Sichuan Province, China.

Both dry deposition (from gases, aerosols and particles) and wet deposition (from rain and snow) of P were monitored. In addition, three parallel collectors were used at each site to collect atmospheric wet and dry deposition to ensure three replicate data, respectively. Dry deposition was determined by the aqueous surface method (Anderson and Downing, 2006). Briefly, three precleaned glass cylinders (inner diameter × height of 10.5 cm × 14.5 cm) were used as dry collectors at each site. All the collectors were placed 1.2 m above the ground with no obstacles and tall buildings around each site. A stainless-steel net (pore size, 0.02 × 0.02 m2) was used to avoid any disturbance and pollution from birds and crops. The cylinder was filled with ultrapure water and examined if a refill was needed on a 4 or 6 h basis (4 h in summer and 6 h in other seasons) to keep the water depth at a level of about 10 cm (Wang et al., 2016). Dry-deposition sampling was conducted for 5 consecutive days at the end of the month, avoiding continuous rainfall as much as possible. Samples were collected in precleaned glass bottles with lids at 08:00 during these 5 d periods. In the case of rainfall, the lid on top of the collector was manually closed to eliminate the effect of wet deposition. At the end of sampling every month, samples collected during the 5 d were mixed and transported to the laboratory to determine total P (TP) concentrations on the same day.

A total of 5 consecutive days per month with a relative frequency of rainfall events were selected for wet-deposition collection, based on weather forecasts every month. Wet deposition was collected at the end of each rainfall event (Oladosu et al., 2017). If the volume of samples (100 mL) collected in one rainfall event was too little, samples from continuous rainfall events were pooled as one mixed sample. The duration (min) and rainfall capacity (mm) were recorded for each rainfall event. Rainfall samples collected monthly were mixed and transferred to the laboratory to determine total P (TP) concentrations on the same day.

During the sampling period, a total of 1026 deposition samples were collected, with half of the dry-deposition and half of the wet-deposition samples. Changes in sample volume and air exposure were minimized. Moreover, river water samples from the Xihe River (30∘36′02′′ N, 103∘39′57′′ E; XB) were collected to measure the P concentration. The total P in the collected samples was digested using potassium persulfate at 120∘ to convert TP to PO43- and then analyzed PO43- using ammonium molybdate using an ultraviolet spectrophotometer at 700 nm.

Monthly dry deposition (MDP) was calculated as the product of the amount of sampling fluid and the concentrations of TP in the sampling fluid. 1MDP=Cd×Vd×NS×105×5, where MDP is the dry-deposition flux of TP in month d (kg P hm-2 per month), Cd is the concentration of TP in the monthly sampling fluid (mg P L-1), Vd is the sampling fluid amount (mL), d represents each month, N is the total number of days per month (d), S is the surface area of the sampling cylinder (m2), and 5 is the sampling days per month.

Monthly wet deposition (MWP) was calculated as the product of the monthly precipitation amount and the concentrations of TP in wet precipitation. 2MWP=0.01×Ci×Pi, where MWP is the wet-deposition flux of TP in month i (kg P hm-2 per month); Ci is the concentration of TP in monthly wet precipitation, which was mixed with all samples for a month (mg P L-1); Pi is the monthly precipitation amount, all from weather stations within 2 km (mm); and i represents each month.

Monthly total deposition (MTP) is the sum of MDP and MWP. 3MTP=MDP+MWP, where MTP is the total deposition flux of TP in each month (kg P hm-2 per month), and MDP and MWP are calculated from Eqs. (1) and (2).

The land-use data (2016) used in this study were provided by the Center of Land Acquisition and Consolidation in Sichuan Province. Land-use types were divided as follows: agricultural area (paddy field, dry farm, yard and agro-facility area), built-up area (urban, town and village), road (highway and country road), forest and water (Fig. 1). Taking the sampling point as the center and extracting the land-use type area with a radius of 5 km from the center, ArcGIS 10.6 was used. Correlation analysis was used to study the covariation between the fluxes of atmospheric total, dry and wet P deposition and land-use areas.

One-way analysis of variance (ANOVA) was performed to determine the spatial and temporal variation among the three areas. Statistically significant differences were set at P<0.05. Pearson's correlation analysis with a two-tailed significance test was used to examine the relationship between the fluxes of atmospheric total, dry and wet P deposition, land-use types and meteorological factors. All analyses were conducted using SPSS^® 20.0 (SPSS Inc., Chicago, USA).

Monthly variations of atmospheric total, dry and wet P deposition fluxes at nine study sites were monitored for 24 months (Fig. 2). For wet deposition, the fluxes peaked in July 2016 (0.06–0.15 kg P hm-2 per month), and the lowest fluxes were found in February 2017 (0.00–0.00 kg P hm-2 per month) (Fig. 2a). In contrast, the highest fluxes of dry P deposition occurred in November 2015 (0.07–0.24 kg P hm-2 per month), and the lowest values occurred in April 2016 (0.01–0.02 kg P hm-2 per month) (Fig. 2b). A similar variation trend was observed in total P deposition, but it reached its lowest value in April 2015 (0.01–0.03 kg P hm-2 per month) (Fig. 2c). Additionally, the monthly contribution rates of dry P deposition to total P deposition varied from 25.0 % to 99.7 % temporally (Fig. 3). Atmospheric dry P deposition constituted more than half of the total P deposition, except in April and August 2015 and May and July 2016, in which heavy rains accounting for 20.37 % of the total precipitation were observed.

The fluxes of atmospheric wet P deposition in summer (including June, July and August in this study) are 2.5–17.1 times higher than those in other seasons (P<0.05, Fig. 4a). Conversely, the fluxes of dry P deposition and total P deposition in autumn are significantly higher than those in other seasons (by 1.4–2.9 times, P<0.05; Fig. 4b, c). Summer (June to August) is the key season for wet P deposition, while autumn (September to November) is the key season for dry P deposition and total P deposition. The study area belongs to the subtropical monsoon climate zone, with high rainfall, temperature and humidity in summer and autumn, which contribute to the emission and deposition of P. Thus, correlation analysis between three types of depositions (atmospheric wet, dry and total P deposition) and meteorological factors (precipitation, wind speed, temperature and relative humidity) was adopted. The fluxes of wet P deposition were positively correlated with precipitation (R=0.917) and temperature (R=0.574) (P<0.01), while the values of dry P deposition have a positive correlation with relative humidity (R=0.439) (P<0.01). Additionally, significant correlations were found between total P deposition fluxes and precipitation (R=0.360), relative humidity (R=0.481) (P<0.01) and temperature (R=0.294) (P<0.05) (Table S2).

The average atmospheric wet P deposition rates at the nine sites showed no significant spatial variations (Fig. 5a), whereas the dry P deposition and total P deposition were observed to have significant spatial variations across the study urban–rural transition (Fig. 5b, c) (P<0.05). Specifically, the annual fluxes of dry P deposition in CC, LY and QQ (0.76–0.84 kg hm-2 yr-1) were significantly higher than those in SS, YS and XB and YT, LJ and HY (0.32–0.49 kg P hm-2 yr-1) (P<0.05). The average rate of dry P deposition among the nine sites was 0.54±0.18 kg P hm-2 yr-1. In addition, the annual fluxes of total P deposition decreased in the order CC, LY and QQ (0.97–1.06 kg P hm-2 yr-1) > SS, YS and XB (0.61–0.71 kg P hm-2 yr-1) > YT, LJ and HY (0.50–0.55 kg P hm-2 yr-1).

To better understand the potential sources of P deposition, the correlations between monthly fluxes of P deposition and areas of land-use types were analyzed (Fig. 6). The monthly atmospheric wet P deposition fluxes were significantly positively correlated with the agro-facility areas in the five months (P<0.05) (Fig. 6a), and the monthly dry P deposition and total P deposition fluxes were significantly positively correlated with the agro-facility areas for almost the whole year (P<0.05) (Fig. 6b, c). Meanwhile, the annual fluxes of atmospheric total, dry and wet P deposition all were strongly positively correlated with the agro-facility areas (R=0.765, R=0.898 and R=0.903, P<0.05).

In addition, dry P deposition and total P deposition both had a positive correlation with the town and paddy field during almost the whole year; the town was significant in February and October, and the paddy field was significant in almost the whole year (P<0.05). Conversely, there was a negative correlation with country roads and forests during the whole year, with country roads being significant in November (P<0.05).

More atmospheric P was deposited through wet deposition in summer than in other seasons (P<0.05). This phenomenon occurred due to the fluxes of atmospheric wet P deposition having a significantly positive correlation with monthly precipitation and temperature (Table S2). In this study area, summer accounted for approximately 51.46 % of the annual precipitation, which would allow more P-containing aerosols to be scavenged in and below clouds by precipitation and deposited on the terrestrial surface, resulting in higher fluxes in summer as well. Similarly, precipitation did have a positive impact on the monthly P deposition fluxes in previous studies (Tsukuda et al., 2005; Zhu et al., 2016; Wang et al., 2018). In addition, the temperature in summer was approximately 7.44–17.19∘C higher than that in other seasons. High temperatures can decrease the stability of the atmosphere and increase the activity of P-containing aerosols, which can enlarge their contact area with the atmosphere. This causes more aerosols containing P to be adsorbed and dissolved in the air (Tipping et al., 2014).

The fluxes of dry P deposition showed varied seasonal variation with those of wet P deposition and had the highest values in autumn than in other seasons (P<0.05) (Fig. 5). In theory, the seasonal variation of dry deposition is mainly controlled by ambient concentration and dry-deposition velocity. On the one hand, high ambient concentrations could be caused by high ambient emissions. In this study, the application of P fertilizer was conducted in autumn, causing higher ambient emissions than in other seasons. On the other hand, dry-deposition velocity was more affected by meteorological factors. Previous studies found that P-containing aerosols were the main components of dry P deposition collected by the alternative surface method and were changed with relative humidity (Qi et al., 2005). In this study, the fluxes of dry P deposition were strongly correlated with relative humidity (Table S2). An increase in relative humidity would lead to an enlargement in particle size and an increase in hygroscopic growth. This growth can significantly increase the particle deposition rate (Mohan, 2016). There are two reasons as follows: on the one hand, P-containing aerosol contact areas with water droplets will be enlarged, which will cause more aerosols to deposit. On the other hand, aerosols can absorb more moisture and increase particle size (hygroscopic growth), making them deposit quickly. Overall, wet deposition was affected by precipitation and temperature. In contrast, dry deposition was influenced by relative humidity and ambient concentration.

Several studies divided P deposition into dry and wet deposition separately for monitoring, and the results demonstrated that the percentage of dry deposition was in the range of 50–85 % (Hou et al., 2012; Tipping et al., 2014). Similarly, this phenomenon was observed in this study. To explain this, first, only a fraction of P-containing aerosols was water-soluble (Herut et al., 2005; Nenes et al., 2011), causing it to likely be deposited as dry deposition. Second, it was observed that the months dominated by wet deposition all followed higher precipitation during the whole year. As discussed earlier, precipitation accelerates wet deposition. Third, the composition characteristics indicated various P sources. The fine dust from desert and soil is more likely to be transported over a long distance and deposited as wet deposition. However, it originated from intensively farmed, especially arable, soil fertilized with P and was more likely to be deposited as dry deposits (Mahowald et al., 2008; Das et al., 2011; Gross et al., 2016). The factor is that the fraction of soil lost as dust is small and likely to be enriched, thus increasing the content of P aerosols and increasing their size (Field et al., 2010). In general, the contribution of wet and dry deposition to the total deposition was impacted by the solubility of P depositions and meteorological factors.

Due to the varied mechanisms of wet- and dry-deposition processes, the fluxes of atmospheric wet and dry P deposition showed distinct spatial variation trends (Fig. 5). The annual atmospheric P dry deposition in CC, LY and QQ (0.76–0.84 kg hm-2 yr-1) was significantly higher than in other areas (0.32–0.49 kg hm-2 yr-1) (P<0.05), but the wet P deposition did not show significant spatial variation.

In general, dry P deposition is dominated by local sources, while a considerable proportion of wet P deposition comes from long-distance P-particle sources (Mahowald et al., 2008; Das et al., 2011; Gross et al., 2016). In this study, more local P aerosols were emitted into the atmosphere in CC, LY and QQ for high-intensity agricultural production, such as large-scale livestock and poultry breeding. These local P aerosols were deposited as local resources, causing a higher value of dry P deposition. Hence, to further clarify the influencing factors from multiple land-use types on P deposition, the analysis of the correlation between land-use types and P deposition was carried out as follows in this study.

Moreover, the annual total P deposition fluxes in CC, LY and QQ (0.97–1.06 kg hm-2 yr-1) were significantly higher than those in YT, LJ, and HY (0.50–0.55 kg hm-2 yr-1) (Fig. 4c; P<0.05). They were also higher than a large number of fluxes on a global scale, such as in Chinese forests (0.69 kg P hm-2 yr-1; Du et al., 2016) and a French tropical forest (0.62 kg P hm-2 yr-1; Van Langenhove et al., 2020). Additionally, here, compared with the global total P deposition rates compiled from 396 published observations during the period 1959 to 2020, including in North America (0.26 kg hm-2 yr-1), Europe (0.29 kg hm-2 yr-1), Asia (0.41 kg hm-2 yr-1), Oceania (0.19 kg hm-2 yr-1), South America (0.40 kg P hm-2 yr-1) and Africa (0.58 kg P hm-2 yr-1) (Pan et al., 2021), the fluxes in CC, LY and QQ in this study showed much higher values. To explain this phenomenon, first, the level of economic development and natural environmental conditions in different regions varied between regions. For instance, in developed regions, substantial anthropogenic P has been emitted into the atmosphere and transported to surrounding areas with the application of P fertilizer on farmlands and the combustion of fuels (Wang et al., 2015; Du et al., 2016). Additionally, this study was compared with the prior findings that were all carried out under multiple land-use types. On the one hand, lands can also be noted that nearly all measurements above refer to natural or seminatural locations. On the other hand, the land-use types at the nine sites in this study were different from each other. A previous study revealed that the sites characterized by land use in an agro-facility contributed more P deposition (3.22 kg hm-2 yr-1), which was higher than that in rural, urban and forest areas (0.20–1.07 kg hm-2 yr-1; Song et al., 2022). Furthermore, the collecting methods utilized for P deposition can also be used to explain the causes of the discrepancies between the experimental results of various studies. In this study, wet- deposition and dry-deposition samples were collected separately, while most reported measurements were based on bulk deposition, which generally ignored dry deposition (Helliwell et al., 1998; Tipping et al., 2014). Additionally, the actual rates could be underestimated with a decrease in collection frequency due to the evaporation of water that would remove the P deposition, which would also be on the wall of the cup. Therefore, differences in the method of sample collection could cause variability among the regions discussed above.

In general, several factors could contribute to the spatial variation of P deposition. As discussed before, the flux of P deposition in this study area is at a high level. Excessive P deposition poses a certain threat to the ecosystem (Wang et al., 2015). Therefore, the potential risk of P deposition in this study area cannot be ignored. More attention needs to be paid to effectively managing P inputs and cycles.

Lands show a tradeoff effect on atmospheric P deposition. On the one hand, agro-facility areas, towns and highways were positively correlated with P deposition, suggesting that those land-use types might be a source for P deposition. On the other hand, there was a negative correlation between country road, forest and P deposition, indicating that they may be sink land-use types for P deposition.

This study showed that P deposition had a strong positive correlation with the agro-facility (Fig. 6). In this study, the area of agro-facility around CC, LY and QQ was approximately 4.5 times larger than the other sites (Table 3). Commonly, agro-facility areas include land designated for livestock and poultry breeding, fertilizer plants, greenhouses with vegetable production and aquaculture (current land-use classification, GB/T 21010–2007). Meanwhile, the survey of this study found that livestock and poultry breeding and fertilizer industries dominated the land use of agro-facility, and there were few greenhouses due to the lack of light. Livestock production and manure generation could be contributors to P deposition (Ma et al., 2011; Tong et al., 2017; Zhang et al., 2019). Many previous studies reported the relationship between P deposition and agricultural production. For instance, P deposition originated from intensive agricultural management and extraction of rock phosphate in Sichuan suburban areas (Song et al., 2022). In areas with a high density of livestock husbandry in Germany, P deposition originated from agricultural emissions from livestock farming (Tipping et al., 2014). Furthermore, emissions from the phosphate fertilizer industry will cause high phosphorus concentrations in the air layer and increase total P deposition fluxes (Rodríguez et al., 2011). This study demonstrated that the land use of agro-facility acts as a main source affecting P deposition almost year-round.

In addition, the monthly fluxes of atmospheric dry P deposition had a significantly positive correlation with the town areas in February, August and October (Fig. 6b, c; P<0.05) during the Spring Festival and National Day. During these important festivals, custom fireworks could induce more harmful gases and dust, thereby increasing combustion emissions of dissolved P, dust emissions and organic P contained in bioaerosol emissions entering the atmosphere (Kanakidou et al., 2020). In addition, monthly dry P deposition and total P deposition fluxes had a positive correlation with paddy field areas during almost the whole year and were significant in the fertilizer period (September, P<0.05). This phenomenon is mainly caused by agricultural phosphorus emissions to a certain extent. Agricultural activities have intensely disturbed paddy field disturbances (Anderson and Downing, 2006). P deposition may originate from local agricultural sources (Tipping et al., 2014). A previous study reported that P deposition increased after P fertilizer application (Gao et al., 2009).

Notably, monthly fluxes of dry P deposition and total P deposition both had a negative correlation with forest and country roads (Fig. 6b, c). Firstly, a negative correlation indicates lower levels of P sources for P deposition in road and forest than in other land use such as agro-facility and agricultural areas. Secondly, it is well known that forests can absorb harmful gas, aerosols and dust particles, including P-containing aerosols, which is attributed to the porous sponge-like underlying surface, high productivity and strong microbial activity (Oladosu et al., 2017; Wang et al., 2017; Zhai et al., 2019). However, forest canopies could elevate P deposition by trapping atmospheric P in the form of dust and particulates (Zhou et al., 2018). Therefore, in this study, a negative correlation indicated that canopy P absorption was greater than the trapping of atmospheric P (Parron et al., 2011). Above all, the sink land use denotes a lower level of P sources and a higher level of P sinks than other land use. Due to similar reasons, paved country roads without hardening showed a similar correlation with P deposition.

In general, land use plays a vital role in P deposition. It was suggested that agro-facility, town and paddy fields were source land types, while forest and country roads were sink land types for P deposition. Furthermore, the key land use for P deposition is the agro-facility in a typical urban–rural transition.

The need for region P deposition control will increase in the future, with the control of N emissions and deposition. The results of this study indicate a range of feasible regional P management measures. Adjusting the land-use structure is the first step, ranging from increasing areas of forests to controlling the scale of aquaculture and livestock farming. The next step is to increase the use of ecological materials and reduce road hardening in the process of road construction. Third, to manage fertilization effectively, more attention should be given to four major fertilization factors (the 4Rs): right rate, right source, right placement and right timing. Last, a policy of prohibition and restriction on fireworks should be implemented (Hochmuth et al., 2022).