The data used in this study were obtained from the National Climatology Observatory (32.44°N, 116.79°E; 26.8 m above sea level) in Shouxian County, Anhui Province, China. The site is located on the south bank of the Huaihe River and is influenced by the subtropical monsoon climate (Figure 1). The annual mean T_a in 2015–2018 was 16.4°C and the annual precipitation was 1116 mm (Duan et al., 2021). During the measurement period (March–May 2016), winter wheat was grown around the site. A 2.5 m eddy covariance flux tower, equipped with a three-dimensional sonic anemometer (CSAT3A, Campbell Scientific Incorporation, United States) and a CO₂/H₂O open-path gas analyzer (EC150, Campbell Scientific Incorporation, United States), was installed at the center of the wheat field to continuously monitor the surface turbulent flux. In addition, a net all-wave radiometer (CNR4, Kipp & Zonen Inc., Netherlands) was installed at 1.5 m a.g.l. (meters above ground level). The raw 10-Hz EC eddy covariance data were averaged over 60-min periods following the methods described in Duan et al. (2022). Additional sensors measuring coarse PM (PM₁₀) and fine PM (PM_2.5) (GRIMM180) were installed on top of a 3 m-high building, 100 m northwest of the flux tower. The PM concentration, radiation, and meteorological data were excluded according to the criterion of X t < X ¯ − 4 σ or X t > X ¯ + 4 σ , where X t is the observation value, X ¯ denotes the means over the interval, and σ is the standard deviation. Eddy fluxes of sensible heat and latent heat were calculated as: H s = ρ ¯ C p ω ′ T ′ ¯ , L E = L ρ ¯ ω ′ q ′ ¯ , where ρ ¯ , C p and L are the density of air (kg m⁻³), the specific heat of air (J kg⁻¹ K⁻¹), and the latent heat of vaporization (J kg⁻¹), respectively. ω ′ , T ′ and q ′ are the fluctuations in the vertical wind component (m s⁻¹), air temperature (K) and specific humidity, respectively (Gao et al., 2009). In addition, data collected on rainy days were discarded to minimize the effects of clouds and precipitation. More detailed information about the instruments can be found in Table 1.

Based on the environmental air quality criteria proposed by the Ministry of Environmental Protection of China, the days on which pollution events occurred during the measurement period were categorized into three levels: 1) background days, 2) clean days, and 3) polluted days. The selected period covered 1 March–31 May, during which, as illustrated in Figure 2, pollution episodes occurred frequently, with some episodes (e.g., on 2 March) having daily maximum values of PM₁₀ and PM_2.5 that reached as much as 414 and 320 μg m⁻³. Background days represented the arithmetic means of variables during the measurement period apart from on days when precipitation events occurred and the 2 days for either side of those events. Daily mean PM₁₀ concentrations higher than 150 μg m⁻³ for more than 48 consecutive hours and hourly PM_2.5 concentrations higher than 100 μg m⁻³ were regarded as polluted episodes. To reduce the impacts of solar altitude and seasonal changes, four polluted cases (i.e., 2–8 March, 13–15 March, 17–20 March, and 26–27 March) in the same month were selected. In addition, clean days were defined as when the daily mean PM₁₀ was lower than 150 μg m⁻³ within 5 days before and after each polluted episode. Ultimately, 20 background days (background episodes, BEs), 6 clean days (clean episodes, CEs), and 6 polluted days (polluted episodes, PEs) were obtained during 3 months (Table 2). The daily mean PM₁₀ (PM_2.5) concentrations for CEs, BEs, and PEs were 104, 124, and 234 μg m⁻³ (68, 91. and 184 μg m⁻³), respectively. Moreover, four dust cases (i.e., 8–14 March, 18–23 March, 5–11 May, and 10–15 May) were identified in the study period from the dust records of weather phenomena at the weather station, in which the ratios of PM₁₀ and PM_2.5 are combined and precipitation events are excluded.

As shown in Figure 3A, the wind speed and variance of vertical wind speed (σ_w ²) had consistent diurnal variations under different pollution conditions. The wind speed fluctuated between 1.5 m s⁻¹ during 00:00–08:00 LST (local standard time), and then increased sharply to a relatively high value (about 2.5–3.5 m s⁻¹) at noon. After that, the wind speed began to decrease and kept at a stable value of about 1.5 m s⁻¹ during 18:00–24:00 LST. The wind speed and σ_w ² were markedly different on background, clean, and polluted days. For example, the wind speed on clean days was broadly higher than that on polluted days, especially in the middle part of the day (13:00 LST; difference could be up to 1 m s⁻¹), which was favorable for atmospheric diffusion. High-PM concentrations were mostly found under low wind speeds (about 1.7 m s⁻¹) from the east or southeast—around 20% lower than during CEs—indicating that local sources contribute to elevated loadings of PM₁₀ at the observatory (Figures 3B, C).

The PM concentration increased with increasing of T_a (Figure 4). High T_a was induced by southeasterly winds during PEs, which would bring warm and moist air to Shouxian. In addition, the radiative warming effect due to the absorption/reflection of longwave radiation by air pollutants can lead to a high T_a (Cao et al., 2016). The daily mean RH during PEs (40%) and BEs (23%) was much higher than during CEs, indicating that high RH was conducive to the growth of aerosol hygrosorption (Peter, 1988; Hara et al., 2018).

Aerosols modulate the climate by absorbing and scattering solar radiation to alter the energy budget of the earth–atmosphere system (Wang et al., 2019). Figure 5 shows the diurnal variations in surface radiation (Rn), sensible heat flux (Hs), and latent heat flux (LE) under the PEs, CEs, and BEs in Shouxian from March to May 2016. The downward shortwave radiation (DSR) and upward shortwave radiation (USR) during PEs were much lower than during CEs (Figure 5A). The DSR on PEs decreased by 15–287 W m⁻² during the daytime (07:00–18:00 LST); the attenuation rate was 28%–84% and the daily average value was 123 W m⁻². Meanwhile, the USR during PEs decreased by 4–52 W m⁻², with an attenuation rate of 27%–80% during the daytime. In order to eliminate the interference of cloud cover, some typical polluted days with few or no clouds were selected (Tables 3, 4). The DSR and USR during PEs with few or no clouds were also lower than those during CEs; the DSR decreased by 6–75 W m⁻² during the daytime (07:00–18:00 LST), with an attenuation rate of 1%–11% (Figure 6A). The downward longwave radiation (DLR) and upward longwave radiation (ULR) showed unimodal diurnal variation. The peak of ULR was delayed by 30 min–1 h compared with the peak of DLR, reflecting the response time of the underlying surface of farmland in Shouxian County to solar radiation heating during the diurnal cycle (Figures 5B, 6B). The DLR and ULR during PEs were higher than those during CEs, with daily average differences of 13 and 4 W m⁻² between PEs and CEs, respectively (Figure 5B).

Under the three types of weather conditions, the trend of Rn was consistent with that of shortwave radiation (both DSR and USR), indicating that shortwave radiation was the main factor affecting Rn (Figures 5C, 6C). The Rn during the day was significantly lower during PEs than CEs with clouds (few clouds or clear days), with a peak difference of more than 100 W m⁻² (61 W m⁻²), approaching 32% (12.5%) and a mean daily difference of 115 W m⁻² (20 W m⁻²) (Figures 5C, 6C). In contrast, the difference in Rn between different weather conditions at night was very small, and the Rn during PEs was slightly higher than during CEs (Figures 5C, 6C). In the daytime, aerosols weakened the solar radiation reaching the ground and reduced the effective radiation at the surface. Under the combined effect of the two, the Rn was reduced, thus weakening the energy of the earth–atmosphere system and causing the ground to cool. Longwave radiation played a dominant role at night, and the existence of aerosols reflected the longwave radiation emitted by the ground and the lower atmosphere back to the surface, increasing the DLR and playing a role in thermal insulation (Figures 5B, 6B).

The diurnal variations of Hs and LE were similar to that of Rn, showing a consistent unimodal pattern (Figures 5D, 6D). The diurnal variation of Hs was comparable during BEs, PEs, and CEs with clouds (few clouds or clear days), with daytime mean values of 36, 33, and 39 W m⁻² (36, 36, and 39 W m⁻²), respectively. LE dominated the Rn during all three types of weather conditions with clouds (few clouds or clear days), with the daytime mean LE/Rn being 32%, 21%, and 28% (32%, 24%, and 28%) for BEs, PEs, and CEs, respectively. The LE during PEs with few clouds also increased by 13 W m⁻² (Figures 5D, 6D). In the daytime, the mean Bowen ratio (β = Hs/LE) was about 0.36, 0.83, and 0.44 (0.36, 0.55, and 0.44) for BEs, PEs, and CEs with much cloud (few clouds or clear days) (Figures 5D, 6D). Therefore, through comparative analysis, the reduction in aerosols changed the distribution of Rn in Hs and LE, and the effect of aerosols on LE was greater than that of Hs. However, in the urban areas of the Yangtze River Delta, the situation is on the contrary. Hs dominates Rn and aerosols had deeper influence on Hs than that on LE (Ling and Han, 2019).

The diurnal variation of LE was more obvious than that of Hs (Figures 5D, 6D). Since the underlying surface of this area is farmland, crops transpire and vegetation interacts with the atmosphere, so there is a high rate of water vapor exchange. The LE was obviously larger than the Hs (Figures 5D, 6D), which indicated that the heat exchange in this area was mainly in the form of LE, and the heat exchange by moisture was stronger than the heat exchange caused by the change in surface temperature between land and atmosphere. This was because aerosols during PEs in Shouxian reduced the transpiration rate of crops, so the LE during PEs was much less than that during CEs (Wang et al., 2021). Meanwhile, the Hs had a similar range of variation and the same trend under the three types of weather conditions, so the influence of aerosols on LE during PEs was greater than on Hs (Figures 5D, 6D).

In order to deepen the level of understanding regarding the impact of transported dust pollution on the surface energy balance over farmland in eastern China, four cases (i.e., 7–8 May, 10–11 March, 20–21 March, and 12–13 March) were selected. Here, we focus on the evolution of the atmospheric boundary layer structure during a dust pollution period (7–8 May 2016).

There are differences in composition and size among different aerosol types. In terms of composition, absorptive and scattering aerosols differ in their effects on radiation and energy balance. For example, absorbing aerosol dominated by black carbon has strong absorbing properties, which induce both prominent surface cooling and near-surface warming. Scattering aerosols are dominated by sulfate, which has a strong scattering capacity and cools the surface to result in a weaker near-surface warming (Gu et al., 2006; Ding et al., 2016; Huang et al., 2018; Zhu et al., 2018; Yang et al., 2020b).

In addition, the size of aerosol particles also affects the radiation budget (Tegen and Lacis, 1996; Murphy et al., 2021), but there is no way to quantitatively distinguish the corresponding relationship between radiation and particle size when measuring radiation. Therefore, mainly, the differences in radiation and energy balance are compared by taking the PM_2.5 as the primary pollutant under the eastern region’s background and the PM₁₀ transported from a distance. However, the meteorological factors under the two different backgrounds also bear some differences, so they too will bring some errors to the contributions of different aerosol species to the effect on the surface energy balance over farmland. Sensitivity tests with more precise observations and numerical simulations will be conducted in the future.