Observed Interactions Among Haze, Fog and Atmospheric Boundary Layer during a Haze-fog Episode in the Yangtze River Delta Region, Eastern China

A severe haze-fog episode occurred in the Yangtze River Delta region of eastern China during 22–30 November, 2018. In this period, the PM2.5 mass concentration and meteorological parameters at the surface were collected at the Station for Observing Regional Processes of the Earth System site in Nanjing. The vertical distributions of PM2.5, humidity and potential temperature below 500 m were observed simultaneously by an unmanned aerial vehicle, and the profile of potential temperature at 1400 local standard time on each day was also observed by radiosonde at the same site. During the first four days, the PM2.5 mass concentration increased, the maximum convective planetary boundary layer height (CBLH) decreased, and the air humidity increased. These are favorable conditions for fog formation. In the latter five days, fog formed on four days, with a lowering of the CBLH and a further increase in PM2.5 mass concentration. We found that the fog top cooling induced a potential temperature jump (i.e., sharp increase of potential temperature) with much warmer temperatures above the cloud top cooling and that this particular thermal structure was maintained until the end of the fog period, which significantly suppressed the daytime development of the planetary boundary layer after fog dissipation. The fog-induced reduction of the CBLH further increased the PM2.5 mass concentration. We also found that the wet deposition of fog on PM2.5 was negligible. The scavenging effect of fog on aerosols only acts during a fog period. When the fog dissipates, the aerosols are liberated from the fog droplets to the atmosphere.


A C C E P T E D M A N U S C R I P T
ascending process of the UAV were used in this study. The UAV measurement times are listed in 132 Table 1. 133 The CBLH can be derived from radiosonde data, as well as the UAV observation data (when the 134 CBLH is below 500 m). The radiosonde-derived CBLH reveals a good agreement with other 135 methods, for example, the CBLHs derived from radiosonde and lidar backscatter measurements 136 coincide within ±200 m (Hennemuth and Lammert, 2006). In this study, when there was no fog, 137 we determined the CBLH as the height at which the vertical gradient of potential temperature had 138 its maximum value (Stull, 1988;Batchvarova and Gryning, 1991) and determined the nighttime 139 planetary boundary layer height (PBLH) as the height of the top of the near-surface inversion layer 140 (Stull, 1988). However, when fog existed, we define the CBLH as the height of the top of the fog 141 layer, above which an elevated inversion or stable layer could be identified or where air moisture 142 decreased significantly (Seibert et al., 2000). fog occurrence are shaded. A relative humidity value exceeding 90% was used to distinguish light 151 fog from haze (Schichtel et al., 2001;Doyle and Dorling, 2002). We used this criterion to separate 152 fog from haze and determine the onset time of fog. In the nine days, the WS was often less than 2

A C C E P T E D M A N U S C R I P T
fog episodes. The near surface air temperature had diurnal variation between 6 °C (279 K) and 156 20 °C (293 K). The minimum temperature at night was not very low. 157 The first pollution phase of the observation extended from 22 to 26 November (Period 1). The 158 PM2.5 mass concentration increased from 50 μg m -3 to approximately 150 μg m -3 . There was no 159 fog in Period 1, and the RH varied diurnally, having lower values in the daytime and higher values 160 in the nighttime due to the diurnal change of temperature. However, q increased steadily from 4 g 161 kg -1 to approximately 7 g kg -1 during these four days (an obvious increasing trend of RH was also 162 observed), suggesting that air moisture accumulated during this period. This is a favorable 163 condition for fog formation. The observed solar radiation (SR) and the reference solar radiation 164 (Ref_SR) are shown in Fig.1. The reference solar radiation was determined from the measured solar 165 radiation on 9 November, when Nanjing was cloud-free and had relatively low PM2.5 (<50 μg m -166 3 ). The SR curves were not smooth, indicating that there was intermittent cloud during the daytime 167 during Period 1. At noon on 5 November, the maximum SR was lower than 600 W m -2 , which was Period 2, during which fog occurred, extended from 26 November to 1 December. As the fog 172 occurred during the night of 26 November, the RH remained high (>90%), the SR was blocked by 173 the fog, and the CBLH was very low. At noon on 27 November, q decreased to 4 g kg -1 (a relatively 174 low level of air moisture), and PM2.5 mass concentration fell to about 50 μg m -3 . The likely reason 175 for these changes is that precipitation occurred during the night before and removed some water 176 vapor and aerosols (wet deposition) from the near-surface atmosphere, although the precipitation

A C C E P T E D M A N U S C R I P T
μg m -3 , and q increased to 8 g kg -1 . Fog continued to occur as the result of the accumulation of air 180 moisture. The maximum solar radiation decreased to 400 W m -2 on 30 November, which is about 181 half the value of the Ref_SR, indicating that the "blocking effect" of the fog and pollution was 182 strong. The maximum CBLHs at 1400 LST in the fog days were systematically lower than those 183 in the no-fog days (Fig. 2). These results suggest that a lower CBLH is favorable for the 184 accumulation of air pollutants, as well as air humidity, in the lower atmosphere. 185 186

CBLHs on fog and no-fog days 187
On 29 November, there was no fog before sunrise. Therefore, the CBLH developed immediately 188 after sunrise, which is similar to the situation in Period 1 (no-fog days). Here, we attribute 29 189 November as a no-fog day because there was no fog in the daytime and during the whole night 190 before. However, on 26-28 and 30 November, fog persisted after sunrise (the fog began at sunrise 191 on 28 November), and it is inevitable that the development of the CBLH was influenced by the fog 192 in these days. We refer to these four days as fog days. Fig. 2 shows the maximum CBLH (at 1400 193 LST) on each day during 22-30 November, which is plotted against the daytime mean PM2.5 mass 194 concentration. The maximum CBLHs on the no-fog days were significantly higher than those in 195 the fog days, suggesting that development of the CBL was suppressed by fog. As shown in Fig. 1, 196 the fog lasted into the later morning and even the afternoon, which possibly delayed the 197 development of the CBL and consequently reduced the maximum CBLH. A lower CBLH is 198 favorable for the accumulation of air pollutants in the CBL, which can lead to heavy air pollution. 199 From 27 to 30 November, the PM2.5 mass concentration accumulated and increased from 50 μg m -200

A C C E P T E D M A N U S C R I P T
and consequently deceased the CBLH ( surface heating began after fog dissipation. On one hand, the maximum CBLH was influenced by 207 the dissipation time of fog. When the fog dissipation was later in the morning, development of the 208 CBL was later and thus the maximum CBLH was lower (if the fog did not last into the afternoon, 209 as on 26 November, the CBL did not develop, and the CBLH was almost the same as the depth of 210 the fog layer). On the other hand, it can be expected that the shadowing effect still functioned in 211 the delayed development process of the CBL, which may have decreased the maximum CBLH 212 further. Our observational results on the fog days revealed a decreasing trend of maximum CBLH 213 with increasing PM2.5 mass concentration, which seems to support this argument. 214 When neglecting large-scale synoptic forcing, daytime growth of the CBL is driven mainly by 215 surface heating and upper-CBL entrainment (Stull, 1988;Sühring et al., 2014). In a simplified 216 mixed-layer model, the CBLH can be predicted when the surface heat flux and the profile of 217 potential temperature are known (Stull, 1988 is the lapse rate of potential temperature in the free

A C C E P T E D M A N U S C R I P T
atmosphere above the CBL, and is a parameter that is a constant of approximately 0.9 for a 228 purely buoyancy-driven CBL (Deardorff et al., 1980;Fedorovich et al., 2004;Sun, 2009). Eq. (1)  229 shows that the growth rate of the CBLH is inversely proportional to the lapse rate of potential 230 temperature in the free atmosphere above the CBL. Therefore the profile of potential temperature 231 plays an important role in the CBL development. 232 To understand the effects of haze and fog on the thermal structure of the CBL, we compared the 233 profiles of potential temperature at 0800 LST and 1400 LST on each day and at the end of the fog 234 period during 22-30 November, 2018. These profiles are plotted in Fig. 3. As shown in the upper 235 panels, on the no-fog days at 0800 LST, the near-surface atmosphere became more stable when the 236 PM2.5 mean mass concentration increased to a high level, whereas the profiles of potential 237 temperature at 1400 LST presented a decreasing trend of maximum CBLH. This situation was more 238 evident on 29 November, a heavy air pollution day, during which the lapse rate of potential 239 temperature in the near-surface layer at 0800 LST was significantly larger and the CBLH at 1400 240 was significantly lower than those of the previous no-fog days. A larger lapse rate of potential 241 temperature in the near-surface layer at 0800 LST means that after sunrise, the CBL develops in a 242 more stably stratified background atmosphere and the growth rate of the CBLH is suppressed. The 243 result that the near-surface atmosphere at the end of night becomes more stable on a heavier air 244 pollution day is consistent with the long-term observations presented in Zou et al. (2017). Therefore, 245 air pollution can impact the CBLH by different processes. On one hand, a high level of aerosol 246 loading in the PBL allows the formation of a more stable nocturnal PBL, which suppresses the 247 growth rate of the PBLH in the morning. On the other hand, high PM2.5 mass concentration 248 increases the shadowing effect of aerosols on solar radiation and reduces the surface sensible heat 249 flux, which leads to a lower CBLH. The heating effect of aerosols also influences the development

A C C E P T E D M A N U S C R I P T
is stronger in the upper part of the CBL ( Barbaro et al., 2013;Ding et al., 2016b). To the best of 252 our knowledge, the stabilized upper part of the CBL can suppress the overshooting of thermals and 253 weaken the entrainment process, which means that the entrainment flux ratio is reduced. 254 According to Eq. (1), a smaller leads to a smaller growth rate of the CBLH. 255 The lower panels in Fig. 3  layer. At about 1200 LST, the fog dissipated, but the potential temperature jump at that height still 267 existed. The fog layer depth was shallow and the potential temperature jump became lower at this 268 moment. Additionally, the surface sensible heat flux increased rapidly to about 200 W m -3 after fog 269 dissipation (Fig. 1). Thus, in the following time, the surface heating was relatively strong, and the 270 potential temperature jump could be consumed easily. The potential temperature profile (the red 271 line) shows that at 1400 LST the CBLH reached a relatively high altitude of about 800 m. A similar 272 situation occurred on 28 and 30 November. On these two days, at 0800 LST, there was a strong 273 capping inversion layer above the fog layer, and at the time of fog dissipation, a step-like potential

A C C E P T E D M A N U S C R I P T
on these two days was thicker than that on 27 November, consumption of the potential temperature 276 jump at the top of the fog layer was more difficult. The CBLH remained unchanged until the 277 potential temperature jump was consumed. This situation is similar to that in many large-eddy 278 simulation studies of the CBL, in which a large potential temperature jump is set at a chosen height 279 so that the top of the mixed layer is approximately fixed during the simulation. To our knowledge, 280 Eq. (1) is not suitable for this situation because the potential temperature jump cuts off the 281 interaction between the mixed layer and the free atmosphere. Here, we distinguish the potential 282 temperature jump from the potential temperature difference across the entrainment zone for a CBL 283 that is developing in the equilibrium state. The former corresponds to a step-like shaped potential 284 temperature profile, but the latter corresponds to a smoothly curved potential temperature profile. 285 Therefore, Eq. (1) is only applicable when the CBL reaches an equilibrium state, in which the 286 interaction between the mixed layer and the free atmosphere forms a stable interfacial layer, i.e., 287 the entrainment zone, or the so-called capping inversion layer. 288 The potential temperature profile at 1400 LST on 28 November shows that the step-like shape 289 of the profile disappeared but the CBLH changed very little, suggesting that the potential 290 temperature jump was consumed just before this time. The potential temperature profile at 1400 291 LST on 30 November shows that the CBLH increased only about 150 m from the moment of fog 292 dissipation, implying that most of the time between 1030 LST and 1400 LST was used to consume 293 the potential temperature jump. Therefore, our observations indicate that the potential temperature 294 jump formed at the top of the fog layer delays CBL development after fog dissipation because a 295 certain amount of time is needed for surface heating to increase the air temperature in the whole 296 mixed layer so that the potential temperature jump can be eliminated and then the CBL begins to and begin to convectively mix the fog layer. The fog becomes more uniform in the vertical with a 303 well-defined top edge. And this sharp top concentrates the radiative divergence closer to that region, 304 which reinforces vertical mixing in the fog layer (Stull, 1988). Therefore, the well-mixed fog 305 appears uniform in the vertical direction and has a well-defined top edge, which forms a sharp 306 potential temperature jump at the top of the fog layer. Our observations show that the step-like 307 jump corresponds to 3-5 K temperature increase over a thin layer with a depth of only about 20 m. 308

A C C E P T E D M A N U S C R I P T
Such a large lapse rate cannot be resolved by weather and climate models, and the growth rate of 309 the CBLH after fog dissipation may be overestimated. 310 To determine the exact strength of the potential temperature jump at the top of the fog layer, we 311 calculated the lapse rate in the thin layer with a depth of 20 m from each temperature profile 312 measured by UAV. The results are shown in Fig. 4 and are denoted by green stars. On 26 November, 313 the potential temperature jump was relatively small (as shown in Fig. 3), but the value of the lapse 314 rate at the top of the fog layer was larger than 50 K km -1 and could reach 100 K km -1 . In the 315 subsequent three fog days, the potential temperature jump was relatively strong, and the lapse rate 316 at the top of the fog layer was typically larger than 100 K km -1 and could reach 250 K km -1 . 317 Observations showed that a very strong stable layer formed in a thin layer at the fog top, although 318 the lapse rate in this layer varies with time and the mean strength was different in different cases. 319 In high-resolution numerical models, the vertical grid distance is about 100 m. To determine the 320 strength of stratification in a layer with a depth of 100 m at the fog layer top, we also calculated 321 the mean lapse rate of potential temperature in this layer. The results are plotted as blue stars in Fig.

A C C E P T E D M A N U S C R I P T
three fog days. This large lapse rate is maintained until the end of the fog period. This means that 324 the CBL develops in a much more stably stratified background atmosphere after fog dissipation 325 and that the growth rate of the CBLH is significantly suppressed. In this situation, the overshooting 326 of rising thermals from the mixed layer is inhibited by the strongly stabilized layer above the top 327 of the mixed layer, which eliminates entrainment. Therefore, we can use the encroachment model 328 (Boers et al. 1984) to estimate CBLH growth rate. The model can be described as 329 where is the lapse rate of potential temperature immediately above the 330 mixed layer. This model neglects entrainment, implying that the mixed layer top erodes into the 331 strongly stable layer due to the increased air temperature in the mixed layer. Assuming that = 332 50 K km -1 (considering the situation in the later three fog days), ′ ′ = 0.15 K m s -1 (the 333 corresponding surface heat flux is 180 W m -2 , which was estimated according to our measurements 334 at noontime), and with an initial CBLH of 350 m at 1030 LST (according to observed fog layer 335 depth), the calculation showed that about 3.7 h is required for the CBLH to increase to 450 m. This 336 estimation agrees with the observations in the latter two fog days. Our observations on 28 and 30 337 November indicated that about 3.5 h is required for the mixed layer top to penetrate this stable layer. 338 Therefore, fog can significantly suppress the CBL development by inducing a strongly stabilized 339 layer at the top of the fog layer. 340 341

Vertical distribution of PM 2.5 during fog days 342
Fog droplets interact with aerosol particles and soluble gases in the atmosphere. Thus, fogs can 343 affect pollutant formation, transformation, and removal. In this study, we observed the vertical 344 distribution of PM2.5 mass concentration in the lower atmosphere both during fog periods and after

A C C E P T E D M A N U S C R I P T
LST revealed that the fog layer top reached about 300 m AGL. Meanwhile, the profiles of PM2.5 348 mass concentration at the two times showed that in the upper part of the fog layer, the PM2.5 mass 349 concentration in the atmosphere was very small (only about 25 μg m -3 ), although it increased with 350 proximity to ground level. However, at 1430 LST, the profiles show that the PM2.5 mass 351 concentration in the fog layer increased to about 100 μg m -3 while the relative humidity in this 352 layer decreased to about 90%, suggesting that the fog still existed at this time but that the 353 evaporation of fog droplets released a lot of PM2.5 back to the atmosphere. At 1530 LST, the fog 354 dissipated, evaporation ended, and the PM2.5 mass concentration increased further to about 120 μg 355 m -3 , which is almost the same as the level before fog formation, implying that almost all of the 356 PM2.5 in the fog droplets was returned to the atmosphere. Fig. 5 shows that the same process 357 repeated in the following fog days. We note that on 27 November, at the time of fog dissipation, 358 the PM2.5 mass concentration was reduced to a lower level in comparison with the value before fog 359 formation, which was evidently the result of the light precipitation that occurred during the fog 360 period (Fig. 1). However, on the no-precipitation fog days, the PM2. the scope of this study. We will investigate this problem based on observations in a future research.

A C C E P T E D M A N U S C R I P T
2018. In this study, we focused on the interactions among haze, fog, and atmospheric boundary 374 layer during the haze-fog episode based on observations at the SORPES site in Nanjing, which is 375 located in the northeastern part of the YRD region. The observations showed that the PM2.5 mass 376 concentration increased from about 50 μg m -3 to more than 200 μg m -3 during this period, while 377 the maximum CBLH decreased from about 1500 to 500 m or even lower. During Period 1 (22-26 378 November), i.e., on the no-fog days, PM2.5 accumulated due to lower wind speed, and increased air 379 humidity. During Period 2 (26-30 November), fog occurred almost every day, the PM2.5 mass 380 concentration increased, and the maximum CBLHs on the fog days were significantly lower than 381 those on the no-fog days. These observational results suggest that the interactions among haze, fog 382 and atmospheric boundary layer enhance air pollution. On one hand, the increase of aerosol loading 383 reduces the CBLH by decreasing the daytime surface heating and stabilizing the atmospheric 384 boundary layer, and the aerosol-induced reduction of the CBLH limits the aerosols and air moisture 385 to a smaller space and results in higher PM2.5 mass concentration and air humidity. This is a well-386 known interaction process on the no-fog days. On the other hand, increased aerosol loading and air 387 humidity provide favorable conditions for fog formation, and the CBLH significantly reduces after 388 fog dissipation, which can further increase PM2.5 mass concentration and lead to heavy air pollution. 389 Observations showed that the maximum CBLHs on the fog days were much lower than those on 390 the no-fog days, suggesting that the occurrence of fog significantly suppressed the development of 391 the CBL. The CBL develops after sunrise on no-fog days but develops after fog dissipation on fog 392 days. Therefore, fog delays the development of the CBL. However, the existence of fog results in 393 a step-like potential temperature jump at the top of the fog layer because of longwave radiative 394 cooling and maintains it until the time of fog dissipation. Thereafter, the CBL first increases the air

A C C E P T E D M A N U S C R I P T
temperature jump. Then, it develops freely in the equilibrium state. This process further delays the 397 development of the CBL, which leads to a much lower CBLH and consequently, a high level of 398 PM2.5 mass concentration on fog days.    are referred as the no-fog days. 626 627