Typhoon- and pollution-driven enhancement of reactive bromine in the mid-latitude marine boundary layer

ABSTRACT Tropospheric reactive bromine is important for atmospheric chemistry, regional air pollution, and global climate. Previous studies have reported measurements of atmospheric reactive bromine species in different environments, and proposed their main sources, e.g. sea-salt aerosol (SSA), oceanic biogenic activity, polar snow/ice, and volcanoes. Typhoons and other strong cyclonic activities (e.g. hurricanes) induce abrupt changes in different earth system processes, causing widespread destructive effects. However, the role of typhoons in regulating reactive bromine abundance and sources remains unexplored. Here, we report field observations of bromine oxide (BrO), a critical indicator of reactive bromine, on the Huaniao Island (HNI) in the East China Sea in July 2018. We observed high levels of BrO below 500 m with a daytime average of 9.7 ± 4.2 pptv and a peak value of ∼26 pptv under the influence of a typhoon. Our field measurements, supported by model simulations, suggest that the typhoon-induced drastic increase in wind speed amplifies the emission of SSA, significantly enhancing the activation of reactive bromine from SSA debromination. We also detected enhanced BrO mixing ratios under high NOx conditions (ppbv level) suggesting a potential pollution-induced mechanism of bromine release from SSA. Such elevated levels of atmospheric bromine noticeably increase ozone destruction by as much as ∼40% across the East China Sea. Considering the high frequency of cyclonic activity in the northern hemisphere, reactive bromine chemistry is expected to play a more important role than previously thought in affecting coastal air quality and atmospheric oxidation capacity. We suggest that models need to consider the hitherto overlooked typhoon- and pollution-mediated increase in reactive bromine levels when assessing the synergic effects of cyclonic activities on the earth system.


Figures S1 to S10
Table S1 to S5 Text S1.Field measurements Field measurements were performed from July 17 to 31, 2018, at an MBL site of Huaniao Island site (HNI,30.85° N,122.69°E, ~60 m a.s.l.) located in the ECS, approximately 70 km away from the nearest shore (Fig. S1) (Zhang et al., 2022).Due to the absence of industrial and traffic activities, the HNI site is an ideal background MBL environment (Li et al., 2016).Due to the typical monsoon characteristic, the anthropogenic pollutants from adjacent regions and distant inlands can be transported to the coastal areas, and even further reach the surrounding marginal seas, i.e., ECS (Nakamura et al., 2005;Pani et al., 2017).So HNI site is also reported to be a receptor of transport of continental outflow on the MBL of ECS (Guo et al., 2019;Yang et al., 2020;Wang et al., 2019).
Due to the prevailing summer monsoon climate, the south to the southeast wind is dominant at the HNI site during this campaign, as the wind rose shown in Fig. S3.Also, the dependency of NO2 VMR at the surface layer observed by MAX-DOAS on wind conditions during the daytime between 07:00 and 18:00 (LT) together with the backtrajectories of the air masses was displayed, which suggested the weak influence of continental outflow during the campaign.The typhoon "Ampil" occurred in the ECS, eventually passed along HNI and made landfall on mainland China during the campaign.
More details about the typhoon are presented in "Text S10.Typhoon 'Ampil'" following.

Text S2. Previous field measurements of BrO in MBL
As a key indicator of the reactive bromine chemistry in the troposphere, BrO has been detected over polar regions, salt lakes and volcanic plumes for its strong sources.In MBL, previous measurements of BrO were carried out in the two mid-latitude coastal locations, Mace Head (Saiz-Lopez et al., 2004) and the Canary Islands (Leser et al., 2003), and also at the equatorial mid-ocean site of Cape Verde (Read et al., 2008;Mahajan et al., 2010), etc., most of them were located in the Atlantic Ocean area.So far, the mid-latitude MBL of the Pacific Ocean area is still a vacancy for measurements.
As summarized in Table S1, BrO shows the typical daytime mean maxima around 2~3 pptv in the clean MBL (Laser et al., 2003;Saiz-Lopez et al., 2004;Keene et al., 2007;Read et al., 2008;Mahajan et al., 2010;Le Breton et al., 2017), and higher BrO was reported in a polluted coastal environment when NO2 levels exceeded 1 ppb (Mahajan et al., 2009).An extremely high BrO exceeding 10 pptv was observed during the ship cruise along the west African coast area, which might be due to the halogenated hydrocarbons emitted by tropical halophytes growing in Mauritania and organohalogens emitted from algae (Martin et al., 2009).Here, it is the first time to report the BrO abundance in MBL of mid-latitude in East Asia.Compared with previous MBL measurements, we found significantly higher daytime BrO (~9.0 pptv) than previous measurements under semi-polluted or polluted environments.
In addition to the reported BrO dataset in the main text, we have conducted short-term and segmented observations in the spring of 2018 to test the instrument performance.
As shown in Fig. S4, the spring observation data are also categorized into three conditions: (1) middle wind speed (3-6 m/s), (2) high wind speed (> 6 m/s), and (3) lower wind speed (≤3 m/s) combined with higher NO2 levels (> 2 ppbv), which corresponds to representative, post-typhoon, and polluted days, respectively, in the original manuscript.These additional results also show similar conclusions as highlighted in our manuscript: (1) the Chl-a in this region in spring 2018 is larger than those reported previously in other regions, and the representative mixing ratios of BrO in Spring 2018 are ~5 pptv, which is similar to those in July 2018; (2) larger wind speeds and SSA (AEC) lead to higher BrO levels, even when the NO2 levels are similar to the representative days; and (3) higher NO2 levels also contribute to the increase of BrO compared to the representative days.

Text S3. MAX-DOAS and short-path DOAS instrument
Due to the sensitivity to the aerosol and trace gases abundances in the lower troposphere (Platt and Stutz, 2008), the MAX-DOAS instrument was used to observe the profiles of aerosol, BrO, and other trace gases.Briefly, it consists of a receiving telescope, a spectrometer and a computer to operate the system (Zhang et al., 2018;Cheng et al., 2019;Zhang et al., 2022).The telescope collects the scattered sunlight at different lower elevation angles (i.e., 0°, 1°, 2°, 3°, 5°, 7°, 10°, 30°) and the zenith direction (90°) sequentially driven by the stepper motor, which takes about 12 min for each cycle.The received scattered sunlight was converged by the lens with a focal length of 101.6 mm onto the fibre bundle connected to the spectrometer.Afterwards, the light signal was detected by the charge-coupled device (CCD) and recorded by the spectrometer (Ocean Optics, QE65 Pro) with a wavelength range from 290 to 473 nm and a resolution of about 0.5 nm full width half maximum (FWHM).To avoid the interference of direct sunlight, the telescope was pointed to the azimuth angle of 330° from the north clockwise.The signal of the dark current was extracted automatically from background measurements taken each night.
In addition, an active DOAS with a short light path was collocated to measure surface O3 concentration (~35 m a.s.l.).The so-called short-path DOAS (SP-DOAS) consists of two telescopes with a diameter of 210 mm as transmitter and receiver, respectively, a 35 W Deuterium lamp as light source and a spectrograph with a range of 195-455 nm (Wang et al., 2015;Guo et al., 2021).The light path between the transmitter and the receiver is 180 m.The exposure time of each scan was adjusted automatically according to the light intensity and the average temporal resolution of the system was about 1 min.

Text S4. Spectral analysis and profile retrieval
The DOAS algorithm is based on the Beer-Lambert law, which describes the extinction of radiation through the atmosphere (Platt and Stutz, 2008).For MAX-DOAS measurement, the spectral analysis generates the measured SCD (slant column density), defined as the integral of the trace gas concentration along the entire optical path including the SCDs in the troposphere and the stratosphere.To eliminate the stratospheric contribution to the SCD, the zenith spectrum was adopted as the Fraunhofer reference spectrum (FRS) for the measured spectra of lower elevation angles during each measurement scanning cycle, assuming the stratospheric absorption would be at the same level within a cycle.Consequently, the SCD of the trace gas measured at each lower elevation angles is represented by the DSCD (differential SCD), which is defined as the light-path integrated trace gas concentration relative to a reference spectrum (Wagner et al., 2011).
Before spectral analysis, we used the ratio of light intensity at 330 nm to that at 390 nm served as colour index (CI) to filter the cloud conditions (Wagner et al., 2016).The QDOAS software (http://uv-vis.aeronomie.be/software/QDOAS/) was applied to analyze the measured spectra.Besides the absorption of interference trace gases, a synthetic ring spectrum (calculated by QDOAS) was included in the spectral fitting analysis.Table S4 summarizes the relevant configurations for the spectral analysis of O4 (oxygen dimer O2-O2), BrO and NO2. Figure S7 shows the example of spectral fitting for O4, BrO and NO2, which displays good performance with obvious absorption structures and low residuals.Before the profile retrieval, the threshold of RMS <1.0 × 10 −3 and the ratios of DSCD errors to DSCDs of 10% for O4 and NO2, and 50% for BrO were applied for the DSCDs to filter the unreliable data.The proportion of O4, BrO and NO2 passing the filter criteria accounted for 82.3%, 50.5% and 77.6% of the total daytime data, respectively.Due to the bad weather caused by typhoon Ampil, considerable spectral fitting results during typhoon interference days, i.e., 20 to 23 July, did not pass the filter.
Combined with the forward radiative transfer model, the measured DSCDs data set can be further retrieved to obtain the vertical profile and vertical column density (Platt and Stutz, 2008).Based on the optimal estimation method (OEM), the vertical profiles of aerosol extinction and volume mixing ratios (VMRs) of trace gases are retrieved by HEIPRO algorithm in this study (Heidelberg Profile, developed by IUP Heidelberg) (Frieß et al., 2006;2011), coupled with the SCIATRAN radiative transfer scheme (Rozanov et al., 2002) as forward model.In general, the two-step approaches are implemented in the profiling algorithm, i.e., aerosol extinction coefficient (AEC) profile is firstly retrieved from measured O4 DSCDs, which is then introduced into the radiative transfer model (RTM) as the aerosol scenario for the second step of trace gases profiles retrieval.
The vertical grid for retrieval was set to an altitude range of up to 3 km with a resolution of 100 m.The fixed a priori profiles were defined by an exponential decay function and constructed with a priori profile error covariance matrix (Sa) of 100% for aerosol and NO2, and 150% for BrO.The covariance length of 500 m for all profile inversions.The a priori extinction of aerosol and surface concentrations of NO2 and BrO, were set to 0.1 km −1 , 1.0 and 0.01 ppbv, respectively.Parameters of surface albedo of 0.1, singlescattering albedo (SSA) of 0.95 and asymmetry factor for HG phase function of 0.72 were used for the radiative transfer simulations.The wavelength of profile retrieval is 360 nm for AEC, 350 nm for BrO and 432.5 nm for NO2, respectively.And the retrieved aerosol scenario at 360 nm can be adapted for trace gas retrieval as input via the Ångström exponent within the algorithm scheme.Examples of aerosol extinction coefficient and trace gas profile retrievals are presented in Fig. S8.Afterwards, we set chi-square of 30 for O4, 8 for HCHO and 50 for NO2 as the thresholds, respectively, in order to ensure the reliability of the retrieval results.Besides, the results with the Degrees of Freedom for signal (DoF) lower than 2.5 for aerosol and 1.5 for trace gases were excluded from further discussion.After the filtering, the squares of the correlation coefficients (R 2 ) increased above 0.99 for aerosol and NO2 and 0.93 for BrO, the slopes deviate from unity decreased below 4% for all species, as shown in Fig. S8d.
For SP-DOAS measurement, the spectral analysis yields directly integrated concentrations of atmospheric trace gases along the optical path and can be further converted to the average trace gas concentration via dividing by the length of the absorption path.The SP-DOAS measured spectra are analyzed using the DOASIS spectral fitting software (IUP in Heidelberg University, Germany).The fitting windows for O3 are 262-272 nm, and absorption of SO2 (Vandaele et al., 2009), HCHO (Meller and Moortgat et al., 2000) and O2 (Bogumil et al, 2003) were considered (Gu et al., 2022).

Text S5. THAMO model simulation
Tropospheric Halogen Chemistry Model (THAMO; Mahajan et al., 2009;Tham et al., 2021) is used to show the response of BrO levels to a few critical factors (as shown in Fig. 2c), including NOx, initial inorganic bromine level, and aerosol surface area (ASA).
The observed (where available) and WRF-Chem simulated levels of O3, NO, NO2, CO, SO2, HCHO, H2O2, CH4, OH, HO2, etc. are used to set up the simulation cases, including Standard, High_NOx, High_INI, High_(INI+NOx), High_ASA, High_(ASA+NOx), and High_(INI+ASA+NOx) cases.All available data points during the whole campaign are used to derive the average diurnal variation to constrain the THAMO simulations.Please refer to Table S2 for details.
An additional case (the Constrained case) was also conducted to use all available observation datasets (including BrO) to quantify the individual contribution of various processes to the overall loss of O3 at HNI (as presented in Fig. 3a).

Text S6. WRF-Chem model simulation
In the present study, we employed a widely-used regional chemical transport model (WRF-Chem), incorporated with comprehensive bromine chemistry (Li et al., 2021a;Badia et al., 2019) ("Text S7.Reactive bromine sources and chemistry"), to investigate the potential influences of various sources on the abundance and impacts of reactive bromine species at HNI and the surrounding region (West Pacific and East Asia).We conducted eight cases (HAL, noSSA, lowORG, noANT, lrgGM, 7NOx_HNI, wthDMS, and noBr; Table S3) to distinguish the impacts of (1) HOBr/BrNO2/BrONO2 heterogeneous efficiency (HAL v.s.noSSA), (2) biogenic source of organic bromine (HAL v.s.lowORG), (3) anthropogenic emission of reactive bromine (HAL v.s.noANT), (4) NOx emission and abundance (HAL v.s.7NOx_HNI), ( 5) bromine species uptake coefficients on SSA (HAL v.s.lrgGM), (6) all reactive bromine sources and processes on the abundance of bromine and other atmospheric compositions (HAL v.s.noBr), and ( 7) DMS source and chemistry on the simulated BrO levels (HAL v.s.wthDMS).Please note that Mass et al. ( 2021) reported that the organic Br emission (airsea flux) in this region (along the coast of East Asia) could be significantly (7.75 to 37.75 times) larger than those reported by Ziska et al. (2013); therefore, in HAL, noSSA, and noANT cases, we applied a medium scaling factor (10) to the Ziska et al. (2013) organic Br emission.
The model setup follows our previous halogen modelling exercises in this region (Li et al., 2021a;2021b).Table S5 summarizes the main settings adopted in this study.In particular, we adopted the sea-salt aerosol (SSA) emission parameterization proposed by Gong et al. (1997).In this parameterization, the SSA emission intensity (particles m -2 s -1 µm -1 ) depends on the surface wind speed as shown in Eq1-2, in which u is the wind speed at 10m above sea level, r is the particle radius at 80% relative humidity.
Such dependence of SSA emission on the wind speed is exponential: for a given radius of the particle, the SSA emission intensity at a wind speed of 20 m/s is >100 times stronger than that at a wind of 5 m s -1 (20^3.41=27322v.s.5^3.41=242).Note that the SSA emission intensity in other literatures (Gong et al., 2003;Jaeglé et al., 2011) also have similar exponential dependence of SSA on wind speed.The typhoon activity significantly stirs up the movement of tropospheric air and sea-air exchange, resulting in larger amounts of SSA in the MBL.The amplified amount of SSA subsequently leads to an outburst of reactive halogen species (including bromine) to the MBL, in a similar manner to the bromine explosion events reported in the Arctic. Eq(1) = (0.38 − log())/0.650Eq(2) The loss of SSA mainly depends on the aerosol dry and wet depositions.The aerosol dry deposition is calculated online following the Wesely (1989) scheme.The in-cloud wet scavenging scheme follows the approach of Easter et al. (2004) and removes cloudborne particles using the first-order loss rate of cloud water, while the below-cloud scavenging washes out aerosols by impaction and interception and was computed in the model based on Slinn (1984) parameterization.Thus, the grid-scale wet scavenging depends on variables such as cloud water, ice, rain, snow, and graupel.Subgrid-scale wet scavenging is calculated using parameterized variables (updraft/downdraft mass flux, entrainment/detrainment rate, and precipitation rate) from the subgrid convection scheme used in the model (Grell and Devenyi, 2002).Typhoon intensity accelerates the formation of raindrops and increases SSA wet deposition.
We then compare the daytime average BrO mixing ratios in various WRF-Chem simulations, as shown in In this study, we apply the WRF-Chem model to provide a first-order estimate of bromine abundance, its potential causes, and the corresponding impacts.Further studies are desired to quantify the relative contribution of different sources to the bromine abundance during the typhoon period, to estimate the impact (or the range of the impact) of bromine species on atmospheric chemistry, and to assess the policy-relevance of halogen chemistry in air quality regulation.
HOBr/BrNO2/BrONO2 + SSA → aBr2 + bBrCl (2a+b≧1) (R1) The first-order reaction rate of HOBr/BrNO2/BrONO2 heterogeneous uptake on SSA is as follows: in which γ is the dimensionless reactive uptake coefficient, C is the molecular speed of the uptake gas species (depending on air temperature; cm s -1 ), and S is the aerosol surface area (cm 2 cm -3 ).
Reactive bromine species actively participate in tropospheric chemistry.With the presence of VOC and NOx (in polluted regions), the Br atom is transformed to BrNO2 (R2) or oxidises VOC species leading to the production of OH, O3, and NO3 (R3 -R9).
In clean and semi-polluted environments (e.g., HNI), the Br atom mostly reacts with The activation of reactive chlorine and iodine during the SSA increase is also worth to be discussed.However, the spectrometer equipped to the MAX-DOAS instrument in this study can only record the light signal between 290 and 470 nm.The spectral fitting range for reactive chlorine and iodine, like ClO and IO, is usually around 300-320 nm (Bobrowski et al., 2007) and420-450 nm (Gómez Martín et al., 2013) by utilizing passive DOAS method measuring scattered sunlight, which are not in the spectral range of the spectrometer with the best performance.So, we are not able to detect the related chlorine and iodine species, which is unfavorable to further investigate their impacts related to typhoon process based on the measurements.
In principle, enhanced release of chlorine is indeed possible from enhanced SSA concentration, considering that the acid displacement process (HNO3+SSA->HCl) is one of the major sources of reactive chlorine gases in the MBL.However, the abundance of chloride in SSA is generally treated as saturated and the limiting factor of this process is the amount of HNO3, instead of the amount of SSA.Meanwhile, the activation of bromine gases from SSA (HOBr/BrNO2/BrNO3 + SSA -> 0.65Br2 + 0.35BrCl, as used in our model) is exponentially enhanced with the increased amount of SSA.Note that this bromine-induced chlorine activation (i.e., BrCl production) is a minor term compared to the acid displacement process.Therefore, we expect that the enhancement in reactive chlorine species during the typhoon period to be smaller than that in bromine species.Future studies, particularly field observations, are recommended to further investigate the potential enhancement in chlorine activation due to typhoon activities.
The main source of reactive iodine in the MBL is the O3 deposition onto the sea surface, releasing I2 and HOI to the atmosphere.The heterogeneous process of reactive iodine involving SSA is actually a recycling one instead of a net source or sink (Li et al., 2022;Saiz-Lopez et al., 2015, 2023;Barrera et al., 2023).Therefore, we expect that the increase in SSA will not substantially affect the source of reactive iodine.Future studies are recommended to look into this aspect.

Text S9. Satellite-derived Chl-α data
In addition, chlorophyll-a (Chl-a) concentrations of the surrounding sea areas derived from satellite observation were acquired for the relevant BrO measurements, as listed in Table S1.Here, we used the 8-day composite L3 products of chlorophyll-a concentration derived from the Moderate Resolution Imaging Spectroradiometer (MODIS) on the polar-orbiting satellite TERRA platform for previous MBL measurements (data available from: https://oceancolor.gsfc.nasa.gov/l3/).For HNI measurements, chlorophyll-a concentration product of the Himawari-8 geostationary meteorological satellite was selected for higher temporal resolution (data available by request from: https://www.eorc.jaxa.jp/ptree/index.html).Chlorophyll-a concentration data were averaged for measurement locations (in-situ measurements) or centre of areas (cruise measurement) with a radius of 20 km and the corresponding measurement period.The calculated daily Chl-a data and corresponding BrO VMR are presented in

Category Options
Radiation RRTMG (Iacono et al. 2008) Land-surface model Noah land surface model (Chen and Dudhia, 2001) Microphysics scheme Morrison double-moment scheme (Morrison et al., 2008) PBL scheme Quasi-Normal Scale Elimination PBL scheme (Sukoriansky, et al. 2005) Gas chemistry MOZART (Emmons et al. 2010;Badia et al. 2019;Li et al., 2021a) Aerosol chemistry MOSAIC (Zaveri et al. 2008;Badia et al. 2019;Li et al., 2021a) Photolysis FTUV (Tie et al. 2003;Badia et al. 2019) Sea-salt aerosol emission Gong et al., 1997;Archer-Nicholls et al., 2014 Anthropogenic emission MEIC (www.meicmodel.org)Shipping emission Fan et al., 2016;Yuan et al., 2023 Biogenic emission MEGAN (Guenther et al., 2006) Biomass burning emission FINN (Wiedinmyer et al., 2011) Initial and Boundary condition CAM-Chem global model output (Li et al., 2022) Horizontal resolution 27 km Simulation period May 15, 2018 to July 31, 2018 with the first 60 days as spin-up Fig. S5.Our WRF-Chem model (HAL case) is able to reproduce the key observed feature in the three periods, particularly the rapid enhancement in BrO mixing ratio under the influence of the typhoon.The comparisons between the main case (HAL) and other sensitivity simulations suggest that (1) the simulated BrO levels are very sensitive to the bromine heterogeneous reactions on SSA (noSSA v.s.HAL); (2) organic bromine species in seawater (chlorophyll-a related biogenic emissions) have small effects on the BrO levels at HNI in the MBL during the typhoon days but have some noticeable influence during other days (lowORG v.s.HAL); (3) the impact of the currently known continental anthropogenic bromine emissions on the BrO on HNI is very limited during the simulation period (noANT v.s.HAL); (4) our sensitivity tests(HAL v.s.7NOx_HNI, HAL v.s.lrgGM, HAL v.s.wthDMS)  suggest that our main conclusion remains the same (typhoon→larger wind speed→more SSA→more bromine) regardless of the simulated levels of NOx, BrO, and DMS; and (5) the heterogeneous uptake process of HOI on SSA (producing IBr and ICl) has limited contribution to the total bromine abundance in this region under typhoon influence.Please note that noANT refers to the case without anthropogenic bromine emissions.
O3 and forms BrO, an important indicator of reactive bromine chemistry in the troposphere (R10).The reduced amount of O3 decreases the production of OH (R11-12).BrO reacts with HO2 (forming HOBr; R13) or NO2 (forming BrONO2; R14) depending on their relative abundance.HOBr photolyzes and forms Br atom and OH radical (R15); the combination of R11 and R13 transforms HO2 to OH and increases OH levels in some regions (ref).HBr, BrNO2, and BrONO2 also recycle bromine species through heterogeneous reactions (R16-

Fig. 2b in
Fig. 2b in the main text and Fig. S4 in the supplement.

Figure S1 |
Figure S1 | Location of the measurement site of Huaniao Island in the East China Sea, China together with the track of typhoon Ampil in 2018.Information on the moving position of Typhoon Ampil was obtained from the Typhoon and Marine Forecasting Center of the China Meteorological Administration (http://typhoon.nmc.cn).

Figure S2 |
Figure S2 | Time series of the daytime BrO, aerosol extinction coefficient and NO2 vertical distribution, observed by MAX-DOAS in HNI site, China.

Figure S3 |
Figure S3 | Potential source region of the air mass arriving at HNI.(a) Wind rose during the observation, (b) the dependency of surface NO2 VMR retrieved from MAX-DOAS measurements at daytime, and (c) back-trajectories of the air masses reaching HNI for each day during the measurement (generated via the MeteoInfoMap software (Wang, 2014) based on meteorological data from the Global Data Assimilation System (ftp://arlftp.arlhq.noaa.gov/pub/archives/gdas1/)).

Figure S6 |
Figure S6 | Track map of Ampil in 2018.Information on the moving position of Typhoon Ampil was obtained from the Typhoon and Marine Forecasting Center of the China Meteorological Administration (http://typhoon.nmc.cn).

Figure S7 |
Figure S7 | Examples of spectral fitting for (a) O4, (b) BrO and (c) NO2 of spectrum measured at an elevation angle of 0° at 07:02 LT on July 19, 2018.The fitted DSCDs and RMS are given in the individual subpanels.

Figure S8 |
Figure S8 | Examples of retrievals from MAX-DOAS measurements on July 28, 2018 around 09:55 LT.(a) A priori profiles and retrieved profiles; (b) The retrieval errors; (c) The average kernels with information on the degree of freedom.(d) Scatter plots of measured DSCDs (ODs) against modelled DSCDs (ODs) during the whole measurement period.The squares of the correlation coefficients (R 2 ) and the slopes derived from the linear regressions (red lines) are given in each subplot.The upper, middle and lower panels indicate the aerosol extinction, BrO and NO2 retrievals, respectively.

Figure S9 |
Figure S9 | Comparison of observed (MAX-DOAS) and simulated NO2 (profile), BrO (profile), and O3 (surface mixing ratio) at HNI.A plot with observed AEC profile and simulated PM2.5 profile is also included to show that WRF-Chem is able to reproduce the key aerosol extinction characteristics, particularly the changes from representative (REP) condition to the post-typhoon (PT) and to the polluted condition (POL).The solid blocks indicate the mean, while the error bars show the standard deviation.

Figure S10 |
Figure S10 | Measured and simulated (WRF-Chem) O3 mixing ratio in Shanghai from July 17 to July 31 in 2018.

Table S1 |
Summary of previous observed daytime BrO in MBL and corresponding chlorophyll-a (Chl-a) concentration of the surrounding sea areas.

Table S2 |
Design of THAMO box model simulations

Table S3 |
Design of WRF-Chem regional model simulations

Table S4 |
Summary of DOAS spectral fitting configuration for O4, BrO and NO2 DSCDs retrieval.

Table S5 |
Model settings and input data adopted in WRF-Chem