Using a coupled LES-aerosol radiation model to investigate urban haze: Sensitivity to aerosol loading and meteorological conditions

The aerosol-radiation-meteorological feedback loop is the process by which aerosols interact with solar radiation to influence boundary layer meteorology. Through this feedback, aerosols cause cooling of the surface, resulting in reduced buoyant turbulence, enhanced atmospheric stratification and suppressed boundary layer growth. These changes in meteorology result in the accumulation of aerosols in a shallow boundary layer, which can enhance the extent of aerosol-radiation interactions. The feedback effect is thought to be important during periods of high aerosol concentrations, for example during urban 5 haze. However, direct quantification and isolation of the factors and processes affecting the feedback loop has thus far been limited to observations and low resolution modelling studies. The coupled LES-aerosol model, UCLALES-SALSA, allows for direct interpretation on the sensitivity of boundary layer dynamics to aerosol perturbations. In this work, UCLALES-SALSA has for the first time been explicitly set up to model the urban environment, including addition of an anthropogenic heat flux and treatment of heat storage terms, to examine the sensitivity of meteorology to the newly coupled aerosol-radiation scheme. 10 We find that: a) Sensitivity of boundary layer dynamics in the model to initial meteorological conditions is extremely high, b) Simulations with high aerosol loading (220 μg/m) compared to low aerosol loading (55 μg/m) cause overall surface cooling and a reduction in sensible heat flux, turbulent kinetic energy and planetary boundary layer height for all three days examined and c) Initial meteorological conditions impact the vertical distribution of aerosols throughout the day.

In addition to the unfavourable meteorological conditions; heavy emissions and regional transport of pollutants into Beijing 25 cause high concentrations of urban aerosol particles to accumulate. These particles can either scatter or absorb solar radiation, depending on their composition. Observations predominantly show that aerosol particles cause net cooling at the surface and warming in the upper atmosphere. This consequently alters the thermal profile of the atmosphere, reducing turbulence due to buoyancy. Reduced turbulent mixing suppresses boundary layer development during the day, minimises the vertical distribution of pollutants and increases surface aerosol concentrations. Furthermore, reduction in planetary boundary layer 30 (PBL) height due to the feedback effect also increases water vapour concentrations which can result in enhanced aqueous heterogeneous reactions, thus increasing the rate of secondary aerosol formation. If the aerosol particles are hygroscopic, increased water vapour concentrations will also cause particle growth, resulting in stronger aerosol-radiation interactions. This positive feedback loop between aerosols, radiation and meteorology can lead to sustained periods of stagnation and has been found to enhance pollution events ( Figure 1) Luan et al., 2018;Petäjä et al., 2016). Aerosol composition and size are the main factors impacting an aerosol particle's single scattering albedo thus impacting the extent by which it will interact with radiation. Most aerosol particles predominately scatter radiation and thus have an overall cooling effect, stabilising the boundary layer and allowing for further accumulation of aerosol particles. However, black carbon (BC), an absorbing aerosol which can contribute up to 10 % of PM in Beijing  has the potential to have the opposite effect, through warming of the lower atmosphere, which promotes buoyancy and destabilises the boundary layer. 40 However, depending on the vertical distribution of the BC layer, BC can also enhance stratification through causing warming in the upper PBL Zhong et al., 2018a;Ding et al., 2016).
Research examining the feedback effect on Beijing haze episodes has thus far relied upon observations or regional modelling studies. Liu et al. (2018b), Zhong et al. (2018b), Gao et al. (2015) and Wu et al. (2019) performed model simulations of pollution episodes using the Weather Research and Forecasting model with added chemistry (WRF-CHEM) to examine the feedback 45 effect. Their results all confirm that aerosol-radiation interactions, aerosol hygroscopic growth and aqueous heterogeneous reactions all factor in the suppression of boundary layer development and result in increased surface PM 2.5 concentrations during polluted episodes in the North China Plain. Gao et al. (2015) suggests that aerosol-radiation interactions decrease temperature and shortwave (SW) radiation at the surface while increasing them aloft (925 hPa). Examining the feedback from a quantitative perspective, Wu et al. (2019) found that when PM 2.5 increased from 50 to 200 µg/m 3 , maximum average boundary layer 50 height decreased from 700 to 400 m. Furthermore, Zhong et al. (2019a) suggested that threshold PM 2.5 concentrations of 75 -100 µg/m 3 exist in Beijing, above which the feedback effect is increasingly important and leads to aerosol accumulation and exacerbation of pollution episodes.
Observational studies also show a link between aerosol concentrations and boundary layer meteorology. Zou et al. (2017) studied the impact of high aerosol concentrations (PM 2.5 > 75 µg/m 3 ) on Beijing meteorology over a year long period. Their 55 results demonstrate that the aerosol impact on meteorology was different depending on the season, with particularly large reductions in sensible heat flux (SHF), PBL height and surface SW radiation reported in autumn and winter  used the same PM 2.5 threshold to estimate the impact of high aerosol concentrations on observed meteorological data over a one month period where haze episodes occurred every 4-7 days. Comparing high and low aerosol periods they found that on average surface SW radiation was 36 % lower and daily maximum PBL height was reduced from 1.3 km to 0.6 km.

60
Despite an increase in research in this area, quantification of aerosol perturbations on boundary layer meteorology is still uncertain. In WRF-CHEM, results are strongly dependent on the boundary layer scheme or parameterisation employed throughout the simulations, while observations of this effect, although useful, only show links between the phenomena without being able to quantify the processes or separate factors. High resolution sensitivity studies which allow for direct analysis of boundary layer meteorology are therefore needed to be able to assimilate the major contributions to haze events. aerosol loading on boundary layer dynamics. Model description and details of set up for an urban environment are outlined in section 2, section 3 describes the experimental set up for cases 1, 2 and 3, section 4 shows results of the simulations and section 75 5 discusses the results, including sensitivity of UCLALES-SALSA to: 5.1 -Meteorological conditions, 5.2 -Aerosol loading and 5.3-Aerosol vertical profiles.

Model Description
The model used in this work is UCLALES-SALSA. A comprehensive description of the model and its previous uses can be found in the paper by Tonttila et al. (2017). The version used here can be downloaded at https://www.github.com/UCLALES-80 SALSA. A description of the model set up, validation and sensitivity to parameters are described below.

UCLALES
UCLALES is a large eddy simulation which has mainly been used in idealised cloud and fog studies. It is based on the Smagorinsky-Lilly subgrid model and solves the Ogura-Phillips anelastic equations with an Asselin filter. Boundary conditions are doubly periodic in the horizontal and fixed in the vertical. Momentum variables are advected with leapfrog time stepping 85 and scalar variables through forward time stepping. In the standard model a two-moment warm rain microphysical scheme is used, the vertical is spanned by a stretchable grid and a sponge layer is applied at the domain top to prevent gravity waves being released into the boundary (Stevens et al., 2003(Stevens et al., , 2005Tonttila et al., 2017). The surface scheme explicitly calculates sensible (SHF) and latent heat (LHF) fluxes at each time step and is based on a coupled soil moisture and surface temperature scheme by Ács et al. (1991) (Eq.1, 2 and 3).
Where ρ is air density, C p is specific heat capacity of dry air, T g and T a are surface and air temperature respectively, γ is the psychrometric constant, f h is a dimensionless function related to water volume fraction and takes the value 0.267 in our 95 case. e s (T g ) is saturation vapour pressure at surface temperature (T g ) and e s is water vapour at 2 m height. r surf is the surface resistance to bare soil and is related to surface friction velocity (u*). r a is atmospheric resistance to water vapour and heat and is dependent on atmospheric stability (Ács et al., 1991).
ω is angular frequency (s −1 ) andT (K) is the average daily temperature in the 2 cm soil layer. The resulting parameters as well as the overall radiation are utilised in the surface energy balance scheme detailed in (Eq. 4) where Q * is net all wave radiation. mixed. In terms of aerosol processes-coagulation and water vapour condensation are switched on, while nucleation, aerosol deposition, emissions and semi-volatile condensation are not considered here for simplicity but may be considered in future work.

UCLALES-SALSA
UCLALES-SALSA couples the UCLALES with SALSA and is primarily described in the paper by Tonttila et al. (2017). The 115 version of UCLALES-SALSA here is a fully coupled radiation-dynamical model, whereby the aerosol-radiation interactions in SALSA are fully coupled with the four stream radiative solver in UCLALES which feeds back on boundary layer turbulence.
This is the first time that aerosol-radiation interactions have been dynamically coupled to UCLALES and the work outlined here examines the sensitivity of aerosol loading on these interactions and feedback.

120
The solution for radiative transfer in UCLALES is based on the 4-stream method integrating over 6 shortwave bands and 12 longwave bands according to Fu and Liou (1993). In this work, the scheme has been adapted to account for the sectional size distribution of the atmospheric aerosol. To this end we use pre-compiled look-up tables of the aerosol extinction cross-section, asymmetry parameter and single scattering albedo, which are given as a function of the size parameter (particle diameter divided by wavelength) and the real and imaginary parts of the refractive index. For a given aerosol constituent, the refractive 125 indices are catalogued at specific wavelengths. Nearest-neighbour interpolation is used to find the values closest to the centres of the wavelength bands used by the radiation solver. Assuming a perfect internal mixture of all aerosol constituents within one aerosol size section, the refractive index in that size section is then calculated as a volume-weighted average of its constituents.
This yields the optical thickness, single scattering albedo and phase function parameters weighted by the actual particle size distribution resolved by SALSA, which are then taken to the 4-stream integration. (Fu and Liou, 1993) 130

Set up in an urban environment
In the past few decades, rapid urbanisation has transformed the landscape in Beijing, creating a microclimate which can be represented by its own distinct physics. Part of this is the Urban Heat Island (UHI), which refers to the phenomenon where a city is significantly warmer than its surrounding areas. This is a result of: increased SW radiation absorption, decreased longwave (LW) radiation loss, decreased turbulent transport, increased heat storage and anthropogenic heat sources. Furthermore, urban 135 environments often consist of mainly impervious surfaces, and therefore the urban heat island is also often characterised by low latent heat and comparatively higher sensible heat fluxes (Oke, 1982;Tong et al., 2017;Yang et al., 2016;Ikeda et al., 2012). To set up UCLALES for an urban environment, alterations to the surface energy balance equation (3) were performed.
Studies by Oke (1982) outline two terms which can be used to represent the presence of the urban heat island. The first is the alteration to ∆Q s or the heat storage term which alters the rate of surface absorption and re-release of heat. In an 140 urban environment, typically the surface has higher surface heat capacity (C h ), water fraction, soil hygroscopicity and lower thermal conductivity (λ) compared to rural environments. This subsequently feeds back on the surface temperature and heat fluxes (Eq.4) The second term is an additional anthropogenic heat flux (Q f ), which accounts for all activities which result in additional heat in a city. This can be split into heat from: buildings, industry, transport and human metabolism. Estimates of the anthropogenic heat flux are difficult to perform and have not been done in wintertime Beijing, although a recent study gives 145 anthropogenic heat estimates for the summertime, which have a mean midday value of 67.2 W/m 2 (Dou et al., 2019). The anthropogenic heat flux has a distinct diurnal profile, attuned to anthropogenic activities within a given city. It is high in the daytime and decreases at night. The additional term is included in the surface energy balance scheme for an urban environment as described in equation 5 (Grimmond and Oke, 1999;Hu et al., 2012;Schwarz et al., 2011;Xie et al., 2016;Yang et al., 2016).
In order to set up UCLALES-SALSA for an urban environment, alterations to the heat storage term and a simplistic additional anthropogenic heat flux were included in the surface scheme and sensitivity studies were performed for a non polluted  Of all surface parameters altered, the largest sensitivity the model showed was to volumetric heat capacity (C h ). Increasing this term decreased maximum SHF, noticeably delayed nocturnal radiative cooling and slightly lowered the temperature through the profile (Figure 2). This is due to slower release of outgoing radiation, which is stored for longer in urban surfaces.
165 Figure 2 shows the sensitivity to varying surface volumetric heat capacity (J m −3 K −1 ) between the initial value (2x10 6 ) and chosen value (7x10 6 ). Higher volumetric heat capacity of the surface causes delayed nocturnal cooling, resulting in higher sensible and latent heat flux in the evening. The surface urban energy balance is also affected by an anthropogenic heat flux which varies seasonally and spatially. A diurnal anthropogenic heat flux which peaks at 70 W/m 2 during the daytime and remains around 20 W/m 2 in the evening was included in a further simulation. Inclusion of a diurnal Q f profile increased overall 170 temperatures as well as latent and sensible heat fluxes (Figure 2).
This sensitivity work provides the setup for UCLALES-SALSA in an urban environment and this is utilised for the remainder of results presented below which all include a diurnal Q f profile and heat capacity (C h ) set at 7x10 6 Jm −3 K − 1 , which is a value typical of concrete (Takebayashi and Moriyama, 2012). The scope for variation of surface parameters within UCLALES is extremely high, therefore we recognise that within the model framework there is a strong dependence on parameters such as 175 temperature, roughness, heat capacity, albedo and soil moisture. It is also likely that due to the simple homogeneous surface scheme used, some features of the urban environment that are observed cannot be replicated in the chosen model framework.
Although the effect of these features is important to understand, the purpose of this paper is to examine the suitability of using Aerosol size distribution parameters and volume fraction of aerosol components were the same for all simulations, detailed in 200 tables 1 and 2. In all cases, BC can be considered to be the primary absorbing aerosol, with SO 4 strongly scattering and OC predominantly scattering with a small absorbing component, while both NO 3 and NH 4 do not directly interact with radiation in these cases Aerosol growth is considered through coagulation and condensation of water vapour but for simplicity semi-volatile condensation and dry deposition is switched off and no additional emissions are considered.

Results
The results highlighted in this section aim to test the sensitivity of the newly coupled aerosol-radiation scheme in UCLALES-SALSA to aerosol loading, using meteorological conditions, urban characteristics and simplified aerosol conditions, associated with Beijing haze episodes. Case 1 shows boundary layer development for 24/11, 25/11 and 26/11 with no aerosols, case 2 examines the effect of high and low aerosol loading for each of the days and case 3 focuses on the impact of varying aerosol 210 vertical profiles.

Case 1-No Aerosols
Simulations in case 1 examine the development of boundary layer dynamics for 24/11, 25/11 and 26/11, without aerosolradiation interactions. All 3 days are initialised with different meteorological vertical profiles, taken from ECMWF profiles.
On 25/11 there is a strong temperature inversion throughout the whole profile, while on 26/11 there is strong vertical wind 215 shear, higher surface humidity and strong stability in the lowest 300 m (Figure 3). Strong vertical wind shear causes mechanical turbulence, while a strong temperature inversion in the morning can suppress boundary layer development through reducing buoyancy. Figure 3 (right) shows development of SHF, PBL height and total turbulent kinetic energy (TKE) for the three simulated days with different meteorological conditions initialised in the morning. SHF is similar in magnitude for all 3 days, while TKE and simulated PBL height is significantly lower for the 25/11 220 simulation. A well mixed, turbulent boundary layer forms quickly on 24/11, however, on 25/11 a shallow, weakly turbulent boundary layer remains throughout the day and on 26/11 a turbulent boundary layer is much slower to develop (Figure 3). The changing conditions used here are typical for a Beijing haze episode and show that even without the consideration of aerosols, meteorological conditions can largely affect the diurnal development of boundary layer dynamics.

225
Case 1 shows that simulated boundary layer dynamics are impacted by initial meteorological conditions. In case 2, the sensitivity of boundary layer dynamics to aerosol loading is examined, where aerosol mixing ratios were constant throughout the profile ( Figure 5). Table 3

235
To assess the sensitivity of the model to a varied aerosol vertical profile, case 3 uses the same set up as case 2 but varies the aerosol mass mixing ratio with altitude, as shown in figure 6. This is to assess the impact of high aerosol concentrations aloft in case 2 simulations which may magnify the aerosol-radiation effect, due to higher total loading increasing the total column aerosol optical depth (AOD). In case 3 simulations, total aerosol mass loading throughout the column is ∼ 22 % less than for case 2 simulations for both high and low aerosol simulations. The aerosol profile was chosen so that at the first time 240 step, aerosol mass mixing ratio at the surface was the same as those with a constant profile and decreased above the PBL in accordance with the potential temperature profiles, while composition and size remained constant throughout ( Figure 6).   The results highlighted above show the use of a novel coupled LES-aerosol radiation model to investigate haze in the urban environment of Beijing. Simulated sensitivity to urban surface parameters is high and these will be different for other urban 255 locations. It is therefore necessary to evaluate and tune these parameters to observations in specific environments in order to use an LES model to fully explore boundary layer dynamic sensitivities. Aerosol-radiation interactions were tested for the first time in the model framework and showed that sensitivity of boundary layer meteorology and turbulence to aerosol loading was strong while also being dependent on initial meteorological conditions.

260
Case 1 identifies the importance of meteorological conditions on boundary layer dynamics throughout the day. Many observations in Beijing found that meteorological conditions are a main driver on both the onset and longevity of haze. Large scale synoptic conditions such as southerly winds and low pressure often preempt pollution episodes which tend to occur every 4-7 days in Beijing wintertime Wang et al., 2019). These conditions are associated with the beginning of 'haze' as the switch in meteorological conditions from strong northwesterly to southerly winds advects pollution from surrounding 265 provinces into Beijing. This change is also associated with a low pressure field within the city, where stagnant air becomes trapped and the dispersion of pollutants is inhibited (Gao et al., 2016).
The initial meteorological profiles for the simulations on 24/11 are taken prior to the onset of the haze and are associated with clean conditions. This is likely the reason for the quick turbulent boundary layer development along with high TKE and SHF throughout the day. Observations show that aerosol concentrations begin to build up around midday on 24/11 and 270 remain constant until the afternoon of 25/11 when concentrations build up rapidly, peaking overnight on 25/11 and remaining high until the afternoon of 26/11. Therefore, the initial conditions used in the simulation of 25/11 will have been affected by aerosol-radiation interactions of the previous evening.This explains the strong temperature inversion in the morning and results in a shallow turbulent boundary layer forming in these simulations, with lower turbulent kinetic energy compared to the 24/11 simulation ( Figure 3).

Sensitivity to aerosol loading
Aerosol-radiation interactions cause a reduction in SHF, surface SW radiation and TKE resulting in a reduction in the daily maximum PBL height for all three days examined. High aerosol concentrations enhance this effect due to an increased number of aerosols interacting with radiation.This leads to a reduction in maximum SHF of 44, 33 and 45 W/m 2 for high compared to low aerosol loading simulations on 24/11, 25/11 and 26/11 respectively. However, results from case 2 show a variation in the 280 magnitude of the aerosol-radiation effect with a larger impact on maximum PBL height for high aerosol simulations on 26/11 compared to 24/11 and 25/11 (Table 3). Including high aerosols on 26/11 causes > 1 o C of daytime cooling in the lowest 300m compared to 0.3 o C of cooling on 24/11 ( Figure 6). The larger degree of cooling on 26/11 leads to a larger reduction in buoyant turbulence and prevents the full growth of a deeply turbulent boundary layer to a larger extent on 26/11. mentum to the surface (Jacobson and Kaufman, 2006). This can reduce wind speeds at lower altitudes and thus decrease wind shear and the shear component of TKE. In case 2, high aerosol loadings reduce surface wind speeds, wind shear and surface frictional velocity (u*) for all 3 days, with a greater reduction on 26/11 compared to 24/11 and 25/11. High aerosol loading also causes a reduction in the variance of vertical velocity (σ 2 w ), which can be considered a measure of turbulence (Stull, 2015). On both 24/11 and 26/11, simulations with high aerosol loading caused a reduction in the magnitude of σ 2 w particularly between 290 500-800 m. On 24/11, the decrease at the surface at σ 2 w is ∼ 40 %, while on 26/11 the reduction is 75 %. In the case of 26/11, this is accompanied by increased values of σ 2 w in the upper layers close to model top, which results in two turbulent layers forming separated by a stable layer. (Figure 8).
In these simulations, aerosol profiles are constant through the column and high aerosol concentrations aloft Figure 7 shows that high aerosol throughout the column causes warming in the upper layers and cooling in the lower layers, which causes 295 strong stability throughout the profile. In reality, aerosols tend to be concentrated closer to the surface and within the boundary layer, although occasionally in Beijing regional transport can lead to higher aerosol concentrations aloft. Therefore, case 3 investigated the effect of limiting pollution to the surface by including aerosol vertical varying profiles.

Vertical profiles
Case 3 examined the impact of meteorological feedback on aerosol vertical mixing for high and low aerosol loading simulations 300 by including aerosol vertical profiles on 24/11 and 26/11. Simulations with a varied vertical aerosol profile had the same aerosol concentrations at the surface but reduced concentrations at higher altitudes ( Figure 6). This resulted in a small increase in maximum SHF (∼ 7 W/m 2 ) on both 24/11 and 26/11. For 26/11, limiting high aerosol loading to the surface results in an afternoon increase in turbulence up to 500 m. Furthermore, The effect of high aerosols throughout the column (case 2) resulted in a highly turbulent layer at model top and a large reduction in surface wind speed on 26/11 ( Figure 8). As this turbulent layer 305 is significantly reduced with lower aerosols aloft, this effect may be considered to be an artefact of aerosol loading at high altitudes which is not often observed in poor air quality events during wintertime in China. However, overall the contribution of the shear term to turbulence is minimal compared to the buoyancy term, which is greatly reduced by high aerosol loading in both case 2 and case 3. The high aerosol loading in case 2 has a larger effect on boundary layer development than the effect of a varying the aerosol vertical profile in case 3. Therefore, we can consider that the change in the thermal profile of the 310 atmosphere, due to high concentrations of aerosols increasing aerosol-radiation interactions, to be the prominent cause of the reduction in SHF and PBL height ( Table 3).
The large degree of cooling on 26/11 compared to 24/11 is due to the effects of initial meteorology feeding back on aerosolradiation interactions. Figure 7 shows potential temperature and aerosol number mixing ratio vertical profiles for each case (after 9 hours of simulation) under high aerosol loading at a) the surface only and b) throughout the profile. After 9 hours 315 of simulation (5pm LST) surface aerosol concentrations on 26/11 are higher than on 24/11. This is due to aerosol-radiation interactions and initial meteorological conditions on 26/11 resulting in a shallower PBL (Table 3). This shows the ability of UCLALES-SALSA to simulate the aerosol-radiation-meteorological feedback loop and that the feedback effect can have a significant impact on aerosol surface concentrations, which will consequently feedback further on atmospheric stability.
6 Conclusions 320 UCLALES-SALSA was set up to model an urban environment for the first time, in order to investigate the impact of aerosolradiation interactions on urban haze. During set up, sensitivity to urban surface parameters was shown to be high, and accounted for the slower release of heat throughout the day as observed in urban Beijing. Inclusion of a diurnal anthropogenic heat flux in simulations resulted in a warmer environment typical of an urban heat island. Given the sensitivity to such parameters, accurate measurements of these properties can be considered paramount in order to improve modelling of the urban environment.

325
Turbulent motion throughout the day in each simulation is further impacted by initial meteorological profiles. Conditions associated with clean periods in Beijing allow for the development of a highly turbulent boundary layer, while strong morning temperature inversions prevent the growth of a turbulent boundary layer throughout the day. Aerosol-radiation interactions in all cases decreases SHF, TKE and PBL height, as well as causing cooling at the surface and reducing surface wind speeds. All simulations also show large sensitivity to aerosol loading, with more than a third reduction in SHF due to high aerosol loading 330 in all simulations. Through comparing simulations with and without aerosol vertical profiles (case 3) we observe that on 26/11 the simulated development of a turbulent boundary layer in the afternoon is impacted by high aerosol loading aloft (case 2) This is due to aerosols at high altitudes reducing mechanical shear as well as the reduction in buoyancy. However, overall the effect of including a vertical aerosol profile is minimal compared to the effect of overall aerosol loading which suggests a higher effect of surface aerosols.

335
The sensitivity work outlined above aims to isolate the aerosol and dynamical effects on pollution episodes through using a specific period with varying meteorological conditions and simplified aerosol conditions. LES models are limited in their ability to represent changing synoptic conditions without additionally forcing or nudging simulated profiles with mesoscale model results or through observations. However, these simulations do show the sensitivity to and importance of meteorological conditions on the development of boundary layer turbulence in Beijing. As well as assessing the importance of aerosol loading 340 on the aerosol-meteorology feedback loop and the impact on PBL turbulent statistics. The aerosol feedback loop is thought to have the largest impact on haze episodes during the cumulative and dissipation stages of the pollution episode. Future work will focus particularly on these stages and the impact of aerosol-radiation-meteorology interactions. As aerosol optical properties play an important role in the feedback, future work will also take advantage of the SALSA framework to vary aerosol optical properties in a case study of Beijing haze.
Acknowledgements. Jessica Slater is fully funded by the National Centre for Atmospheric Science (NCAS). This work was carried out as part of AIRPRO (NE/N00695X/1) for which Jessica Slater, Hugh Coe and Gordon McFiggans also acknowledge funding. Sami Romakkaniemi and Juha Tonttila are supported by the Academy of Finland (projects 283031 and 309127) Model simulations were carried out on the ARCHER UK National Supercomputing Service (http://www.archer.ac.uk). We gratefully acknowledge Yele Sun's group and Pingqing Fu's group at IAP for aerosol composition data and tower meteorological data respectively, as well as Zifa Wang's group at Peking Univeristy for 355 aerosol size data and Eiko Nemitz at CEH Edinburgh for heat flux data.