The measurements were carried out in Aarhus (56.13° N, 10.22° E) from 3 March to 11 April 2023. Aarhus is a coastal city, the second largest city in Denmark, and has approximately 360 000 inhabitants and an area of 468 km². As shown in Fig. , the instruments were placed in the house of a Kayak club (“Kajakklubben Viking”), located on a beach along the southeastern coastline of Aarhus. The coastline is bordered by a forest area, including a deer park, which is a habitat for a herd of deer and other wildlife. Outside the forest, there are main roads, residential houses, and agricultural fields (to the southwest). In addition, Aarhus Port, one of Denmark's largest container ports, is at a distance of 3.83 km from the sampling site (to the north). Thus, the measurement site is potentially influenced by clean marine, forest, and anthropogenic emissions from residential activities, as well as traffic both from the continent (i.e. vehicles) and the sea (marine shipping).

Wind, radiation, and relative humidity (RH) data were obtained from the Danish Meteorological Institute (DMI). The closest DMI measurement site is located in the south of Aarhus (56.1° N, 10.23° E), 7 km from the Kayak Club. PM₁₀ (particulate matter with a diameter of less than 10 µm), nitrous oxide (NO_x), and carbon monoxide (CO) were monitored by the Danish Center for Environment and Energy (DCE), from a station about 5 km from the Kayak Club, located at Banegårdsgade (56.15° N, 10.20° E), close to the railway station characterized by heavy traffic .

A nephelometer (Aurora 3000, Ecotech), an aethalometer (AE33, Aerosol Magee Scientific), and a white-light optical particle spectrometer (Welas^® 2300, PROMO 3000, PALAS) were used to obtain the scattering and absorption coefficients, as well as the number and surface particle size distributions of the aerosols. The three instruments were placed inside “Kajakklubben Viking” and shared a stainless steel inlet via a three-way splitter to separate the airflow. The inlet was installed through the wall, parallel to the ground, at a height of around 4 m above the beach. A steel net covered the end of the inlet to prevent droplets and bugs from entering the inlet. The sampled air was dried due to the temperature difference between outdoors and inside, where the instruments were located. No additional dryer was employed. A schematic is presented in Fig. S1 in the Supplement.

The nephelometer measured the integrating intensity of total light scattering and backscattering from aerosols and gas molecules at three wavelengths (λ = 450, 525, and 635 nm), with a resolution of 1 min. The data were corrected for angular truncation errors using the “no-cut” correction according to . The nephelometer operated at a flow rate of 5.0 L min⁻¹. Both total scattering (angles between 10 and 171°) and backscattering (angles between 90 to 171°) were measured. A full calibration, including zero measurement and span gas calibration with carbon dioxide (CO₂) (99.9 % purity), was performed prior to and after the measurements. As the results coincided within the measurement uncertainty, no additional correction was applied. The nephelometer was equipped with an RH sensor in the measurement cell and two temperature sensors, one in the inlet and the other one in the measurement cell.

The aethalometer measured the attenuation of light through a filter tape progressively loaded with atmospheric aerosol particles, providing a “filter-light-attenuated signal” at seven wavelengths (λ = 370, 470, 520, 590, 660, 880, and 950 nm) at 1 min time resolution and a flow rate of 5 L min⁻¹ . The absorption coefficients were calculated using Eq. () by its relation to the equivalent black-carbon (eBC) mass concentrations, where MAC is the mass absorption cross-section : 1σabs(λ)=eBC(λ)⋅MAC(λ).

MAC values for the different wavelengths were used as suggested in the manual (e.g. MAC = 7.77 m² g⁻¹ for λ = 880 nm). This value at 880 nm was previously employed for aerosols measured in southern Sweden . Black-carbon source apportionment, to subdivide the fractions from fossil fuel combustion and biomass burning, was performed according to , where the portion from biomass burning is referred to as BB%. This model is based on the assumption that the absorption signal is due to only two absorbing aerosol ensembles, one with a low wavelength dependence (fossil fuel combustion aerosols) and one with a high wavelength dependence (biomass burning aerosols), and that no other types of aerosols influence the absorption signal. A detailed explanation of how these sources can be subdivided is presented in the Supplement. There is a multiple-scattering effect caused by light scattering between the particles and the filter fibres. This depends on both the type of filter and the single-scattering albedo of the aerosol. Therefore, we followed the approach by including the single-scattering albedo of the aerosols to improve the correction factor of 1.57 for TFE-coated glass filter (M8020) tape for the aethalometer. Using a maximum value of 0.025 for the cross-sensitivity of absorption to scattering, as stated in , we find that the maximum total correction factors for both scattering from the aerosol and the filter fibres were 1.76, 1.77, and 1.79 for λ = 470, 520, and 660 nm. This results in a maximum relative error of the absorption coefficients due to the scattering of the particles of up to 12.25 %, 12.99 %, and 14.47 % for λ = 470, 520, and 660 nm, respectively. This indicates that the scattering by the aerosols in our specific case did not have a significant effect on the measured absorption signal.

The white-light optical particle spectrometer WELAS was used to measure the number and surface size distributions of particles with optical diameters from approximately 0.2 to 10 µm (in PSL (polystyrene latex)-equivalent particle diameters). The presented total particle number concentrations are thus only representative for this size range, only partly capturing the nucleation and Aitken mode. A flow rate of 5 L min⁻¹ was used. The WELAS sensors were calibrated with MonoDust 1500 as provided by PALAS prior to the measurements.

Prior to data processing, all raw data were cleaned for instances with RH above 40 % (0.78 % of the total dataset), tubing disconnection, and tape advancement of the aethalometer.

A schematic illustrating the tubing used for the inlet and instrumental setup at the measurement site is presented in Fig. S1. To assess losses of particles in the sampling probe and inlet tubing, we calculated the overall inlet efficiency as a function of particle diameter using an open-source particle loss calculator . The calculation accounts for particle losses due to Brownian diffusion, gravitational settling, and inertial impaction in straight tubing and bends. This was done for each instrument by inputting length, diameter, angle of inclination, angle of curvature, and sample flow for each tubing piece between the instrument and the sampling inlet. For all three instruments, we found decreasing inlet efficiencies with increasing particle size. The nephelometer and aethalometer have a 50 % inlet efficiency for a particle diameter of 8 µm, whereas the WELAS has a 50 % inlet efficiency for a particle diameter of approximately 6.5 µm. We also calculated the sampling efficiency of the sampling probe at five different wind speeds: 0, 2.5, 5, 7.5, and 10 m s⁻¹. At wind speeds of 0 and 2.5 m s⁻¹, there is a depletion in the sampling efficiency of larger particles, whereas, at wind speeds of 7.5 and 10 m s⁻¹, there is an enrichment in the sampling efficiencies of larger particles. Efficiency curves are presented in Fig. S2 and tubing parameter tables are presented in Tables S1–S3. The results were not used to correct the data.

The single-scattering albedo (SSA), absorption Ångström exponent (AAE), scattering Ångström exponent (SAE), spectral variation of SAE (ΔSAE), backscattering ratio (b), and asymmetry parameter (g) were derived from the scattering and absorption coefficients measured by the nephelometer and aethalometer.

SSA was calculated using Eq. () as the ratio of the scattering coefficient (σsca) to the extinction coefficient, which is the sum of the scattering coefficient and absorption coefficient (σabs). SSA values at wavelengths (λ) of 470, 520, and 660 nm were calculated; thus, the scattering coefficients were recalculated for these three wavelengths based on a power-law fitting . 2SSAλ=σsca(λ)σsca(λ)+σabs(λ)

Typical values of SSA for sea salt and sulfate aerosol are expected to be close to 1 at λ = 550 nm , while fresh BC is highly absorbing and has been shown to have SSA values close to 0.3 at λ = 554 nm .

SAE and AAE reflect the spectral variability of light scattering and absorption and were calculated using Eqs. () and (). SAE was calculated for the nephelometer wavelength pair (450 and 635 nm) and additionally for the wavelength pair of 450 and 550 nm to compare to . Values that were not directly measured were recalculated based on power-law fitting. 3SAEλ1,λ2=-ln⁡σsca(λ1)σsca(λ2)ln⁡λ1λ24AAEλ1,λ2=-ln⁡σabs(λ1)σabs(λ2)ln⁡λ1λ2

SAE is inversely related to particle size. Coarse-mode aerosol particles, such as dust, pollen, and sea salt, have been reported to show SAE values of 0.5 . Small particles, such as carbonaceous particles from fossil fuel combustion or biomass burning, typically show SAE values greater than 1.5 . AAE is related to the chemical composition of the particles. BC (emitted during both fossil fuel and biomass combustion) is known to absorb radiation equally well at all wavelengths and therefore exhibits AAE values close to 1 . Dust (e.g. from soil or volcanic eruptions) and brown carbon (BrC, from biomass burning), however, exhibit wavelength-dependent absorption, with stronger absorption in the ultraviolet and blue spectral regions , corresponding to AAE values greater than 1.5. Finally, the Ångström matrix describes the relationship between AAE and SAE values in a bivariate plot, effectively subdividing the AAE–SAE space into regions that are representative of different particle types and sources . Each point in the matrix corresponds to a pair of AAE and SAE values, and clustering of points in specific regions can be used to identify dominant aerosol types (e.g. large coarse particles, carbonaceous particles, or mixed particles). For a visualization of the full dataset, see Fig. S3.

ΔSAE describes the spectral variation of SAE, calculated as follows : 5ΔSAE=SAE(λ1,λ2)-SAE(λ2,λ3).

Here, λ1<λ2<λ3. While SAE is related to the mean size and relative concentrations of the accumulation and coarse mode aerosol, ΔSAE provides information on the contribution of fine- and coarse-mode particles, leading to a better understanding of the SAE values . A negative value indicates a large contribution from a single, fine-mode particle, while a positive value indicates a contribution from both the fine and coarse modes, where higher values represent a larger contribution of the coarse-mode particles.

The backscattering ratio b is defined as the ratio of backscattering coefficients (σbsca,λ) to total scattering coefficients (σsca,λ): 6bλ=σbsca(λ)σsca(λ). b values range from 0 to 1, where a high fraction of larger particles results in a smaller value of b, as large particles (size larger than wavelength) predominantly scatter radiation in the forward direction .

b can also be used to estimate the asymmetry parameter g . g is the cosine-weighted mean of the phase function, which describes the angular distribution of scattered radiation. It ranges from -1 to 1, representing total backscattering and total forward scattering, respectively, whereas 0 represents an equal balance between forward and backward scattering. It is a crucial parameter in radiative transfer models, climate models, and general circulation models used to estimate the direct radiative effect of aerosols . g was calculated using Eq. (): 7g=-7.143889⋅bλ3+7.464439⋅bλ2-3.96356⋅bλ+0.9893.

The origins of the air masses arriving at the receptor site, where sampling instruments were placed, were simulated with the Lagrangian particle dispersion model FLEXPART version 10.4 . The model emits computational particles at the surface that are tracked backward in time by the ambient flow, as defined by meteorological fields that are used as input into the model. In the present study, the meteorological input was based on hourly reanalysis meteorological fields (ERA5) from the European Centre for Medium-Range Weather Forecasts (ECMWF) with 137 vertical levels (up to approximately 80 km) and a horizontal resolution of 0.5° × 0.5° . Simulations extended over 30 d backward in time, a sufficient period to capture more than 98 % of most BC emissions arriving at the receptor, given a typical BC lifetime of 1 week . FLEXPART includes gravitational settling for spherical particles, dry and wet deposition of aerosols , turbulence , unresolved mesoscale motions , and deep convection . The model output consists of a footprint emission sensitivity that expresses the probability of any emission occurring in any grid cell to arrive at the receptor site and can be converted into model concentration when coupled with gridded emissions.

Anthropogenic emissions were adopted from the latest version (v6b) of the ECLIPSE (Evaluating the CLimate and Air Quality ImPacts of ShortlivEd Pollutants) dataset, an upgraded version of the previous version . The inventory includes emissions from industrial combustion (IND) and from the energy production sector (ENE); residential and commercial emissions (DOM); and emissions from the waste treatment and disposal sector (WST), transportation (TRA), shipping activities (SHP), and gas flaring emissions (FLR). The methodology for obtaining emissions from FLR, specifically over the Russian territories, has been improved in ECLIPSEv6 . Biomass burning (BB) from wildfires was adopted from the Copernicus Global Fire Assimilated System (CAMS-GFAS) , a product that provides the estimates of the injection altitude of fire emissions that are crucial for accurate simulations of BB dispersion.

Figure  presents an overview of the aerosol optical properties during the measurement period, where each data point represents a mean of 10 min, while Table  presents numerical values of statistical parameters. Figure a shows scattering coefficients (σsca) at three different wavelengths. Two periods with particularly high σsca are visible, the first being from 18 to 20 March and the second being from 30 to 31 March. Apart from these two periods, the values of σsca were below 100 Mm⁻¹. The median σsca values at 450, 525, and 635 nm were found to be 24.42, 20.24, and 17.29 Mm⁻¹, respectively. In addition, the σsca at 450 nm was generally higher than at 525 and 635 nm, indicating that fine-mode particles contributed significantly to the aerosols. The period with the highest absorption coefficients (σabs) was observed from 17 to 20 March (Fig. b). Many sharp peaks were only visible in σabs and were absent in σsca, indicating the presence of strongly absorbing particles for short periods of time. These were likely emitted by traffic or biomass burning near the site. Therefore, they are not considered to be representative of the regional aerosols. The median σabs values at 470, 520, and 660 nm were 3.61, 3.06, and 2.20 Mm⁻¹, respectively. In general, the absorption coefficients decreased with increasing wavelengths.

Compared to Vavihill, a background site in southern Sweden, median σsca values at 525 nm in this study (20.24 Mm⁻¹) were lower than in the Swedish study, where medians amounted to 31.4 Mm⁻¹ for a similar period of the year (1 March–1 April 2011). Also, median σabs values at 520 nm from this study (3.06 Mm⁻¹) were slightly lower compared to medians measured at Vavihill, amounting to 3.7 Mm⁻¹ at 520 nm, for the same season . A follow-up study in Vavihill in April 2013 reported a monthly mean of approximately 5 Mm⁻¹ at 520 nm , which is similar to our mean values (4.53 ± 4.43 Mm⁻¹). Compared to a coastal, Mediterranean city (Lamezia Terme, Italy), characterized by anthropogenic pollution emissions related to traffic and agriculture, the σsca values at 550 nm during the cold season (October to March from 2015 to 2018) were found to be higher than those from this study, with median values of 30.93 Mm⁻¹ .

A median eBC value of 231.3 ng m⁻³ was found during the whole measurement period at Aarhus Bay, as calculated from the σabs value at 880 nm. Furthermore, the median value of BB%, characterizing the fraction of biomass burning to the eBC load, was 22.4 %. This indicates that biomass burning also contributed significantly to the absorbing aerosols at Aarhus Bay. eBC values at Vavihill, as calculated from presented σabs values at 880 nm during April 2013 (monthly means), were found to be approximately 380 ng m⁻³ , which is in the same range as our measurements at Aarhus Bay (Aarhus: mean of 317.53 ± 269.97 ng m⁻³). eBC values at Hyltemossa Research Station, a rural station about 65 km northeast of Malmö–Copenhagen, located in southern Sweden, were comparable with a median of 227.5 ng m⁻³ during March 2018 . Interestingly, the median eBC concentration at Lamezia Terme during the cold season was 490 ng m⁻³ (mean: 690 ± 0 ng m⁻³) and thus was significantly higher than the median and mean values in Aarhus.

Figure c depicts the single-scattering albedo (SSA) at 520 nm, which ranged from 0.28 to 0.99, with a median of 0.87. Oceanic aerosols are generally expected to have high SSA values close to 1 at wavelengths from 440 to 1020 nm, while urban aerosols and biomass burning aerosols are typically described by lower values . Median SSA values at 520 nm in Aarhus (this study) and Vavihill were very similar (0.87 in Aarhus vs. 0.89 in Vavihill) . The median SSA value at Lamezia Terme using a wavelength of 637 nm during the cold season was 0.78 (min: 0.2; max: 0.99) , which is slightly lower than the values from this study (Aarhus median: 0.84 at 660 nm).

The median AAE_470/660 nm throughout the campaign was 1.39 (mean: 1.37 ± 0.26), which is well comparable to previously reported values for aerosols from fossil fuel combustion of 1 to 1.5 . The median SAE_450/550 nm was 1.32, with a mean and standard deviation of 1.15 ± 0.74, indicating more variability in the wavelength dependence of the scattering coefficient. This variability can typically be explained by a change in the dominant aerosol particle mode as larger particles are described by smaller SAE values and vice versa. SAE_440/675 nm values for aerosols from fossil fuel combustion were previously reported to be larger than 1.2 , indicating smaller particles, while SAE_440/675 nm values of dust and SAE_467/660 nm values of sea salt were reported to be less than 0.5 , indicating larger particles. This suggests that both fine- and coarse-mode particles occurred during this Aarhus Bay campaign, which is also in agreement with the median value of ΔSAE of 0.19, suggesting the presence of both modes . Few other studies exist in which similar data for coastal environments are reported: a monthly mean AAE_370/950 nm for April 2013 of approximately 1.3 was found in Vavihill (Sweden) , in line with our data, while reported a median SAE_450/700 nm of 2.14 during the cold season (October to March from 2015 to 2018) in the coastal city of Lamezia Terme (Italy). The substantially lower SAE value from this study indicates the presence of more large particles in Aarhus Bay compared to the Mediterranean coastal city.

The median backscatter ratios b at 450, 525, and 635 nm were 0.12, 0.13, and 0.13. As larger particles scatter predominately in the forward direction, such small b values support the SAE and ΔSAE results, highlighting the presence of, in general, larger particles. These values are in line with median values measured in Lamezia Terme during the cold season (October to March from 2015 to 2018) of 0.10, 0.11, and 0.12 at 450, 550, and 700 nm , respectively. The median values of the asymmetry parameter g in Aarhus were 0.60, 0.58, and 0.58 at 450, 525, and 635 nm, respectively, which is comparable to median values at similar wavelengths in Lamezia Terme during the cold season, found to be 0.67 (450 nm), 0.63 (550 nm), and 0.62 (700 nm) .

Figure  depicts time series of the total number concentration (NTOT; Fig. a), number size distribution (dN/dlog⁡Dp; Fig. b), and surface area size distribution (dS/dlog⁡Dp; Fig. c) measured at Aarhus Bay. Both the total number concentration and size distributions are depicted as hourly means. The size distributions are plotted for particles above 300 nm. The changes in NTOT follow the same trend as was observed for σsca and σabs; i.e. the periods that were found to have high σsca and σabs also had high NTOT, as seen in Fig. a. From 17 to 20 March, the highest total particle number concentration was observed, with a maximum concentration of ∼ 550 cm⁻³. Elevated total particle number concentrations were also detected from 9 to 11 March and from 29 to 31 March. The median total particle number concentration throughout the whole period was 31.48 cm⁻³ (mean: 58.04 ± 70.15 cm⁻³). Number size distributions highlight particles between the detection limit of the WELAS and approximately 1 µm, and surface size distributions more clearly depict the contribution of particles up to 7 µm. When comparing the evolution of NTOT with the number and surface size distributions (Fig. b and c), it is evident that the period with the highest particle number concentration coincided with the number and surface size distributions dominated by smaller particle sizes, indicating an enhanced abundance of small particles during this period. Generally, when the number and surface size distributions showed a mode above 0.5 µm in particle diameter, the total number concentration remained below 200 cm⁻³ and was typically less than 100 cm⁻³.

Figure  illustrates the median diurnal variation of σsca, σabs, and eBC concentrations during weekdays (Fig. a, c, and e) and weekends (Fig. b, d, and f). Median values were calculated excluding the data from 17 to 19 March, which showed the highest concentrations.

The absorption coefficient σabs shows clear diurnal patterns on both weekdays and weekends, with two major peaks, one in the morning and one during the evening/night, highlighting the typical rush hours. The morning peak is generally around 05:00–09:00 LT (local time), while the evening peak typically lasts from 17:00 to 23:00 LT. The morning peak appeared to be earlier and sharper during weekdays than during weekends, coinciding with expected traffic patterns during weekdays and weekends, respectively. The evening peak was slightly lower during weekdays compared to during weekends (Fig. c and d). The scattering signal did not show a clear pattern during weekdays; however, two distinct peaks could be observed during the weekend (Fig. a and b). Furthermore, the difference between the three wavelengths was larger on weekdays than on weekends. During weekends, σsca values at different wavelengths were very similar between 10:00 and 18:00 LT, indicating that both smaller and larger particles were present. During weekdays, however, σsca at 450 nm was clearly higher in the same time period, indicating a larger fraction of small particles.

Overall, the differing diurnal behaviours in scattering and absorption indicate that non-absorbing or weakly absorbing aerosol components contribute substantially to the observed scattering signal, particularly during weekdays. Figure S4a and b show the temporal evolution of the diurnal pattern of scattering and absorption coefficients, highlighting that, except for the last week, the daily patterns were quite comparable throughout the course of the campaign.

The contribution from combustion of fossil fuels was calculated by subtracting the contribution from biomass burning from the total eBC. Figure e and f present eBC values subdivided into fossil fuel (black) and biomass burning (brown) contributions. Fossil fuel contributed the most to eBC concentrations, above 70 % during both weekdays and weekends. BB% varied from 17 % to 26 % during weekdays and from 13 % to 29 % during weekends. During weekdays, a prominent peak during rush hour was observed from around 05:00 to 08:00 LT, underlining the important influence of traffic emissions at the site. In particular, the increase on weekdays was larger than that on weekends. During the evening, an increase in both fossil fuel and biomass burning contributions was found on weekends and weekdays. Notably, the contribution from biomass burning was higher during weekends, where it increased from 20 % to 29 % between 18:00 and 23:00 LT. This could arise from the increased use of domestic fireplaces during weekends and bonfires.

In addition to human activities such as traffic and biomass burning, changes in the planetary boundary layer (PBL) height may impact daily patterns. A high PBL typically leads to the dilution of surface emissions and, in general, lower concentrations, while a lower PBL hinders dispersion, thus favouring higher concentrations. Solar radiation is the major driver for PBL height and can be used as a proxy. Figure S5 shows the overall radiation time per day, and Fig. S6 illustrates the diurnal variation of solar radiation during the campaign. The typical radiation pattern showed an increase from around 07:00 to 08:00 LT and a decrease around 19:00 to 21:00 LT, where the decrease shifts to later hours at the end of the campaign. This coincides well with the time the morning peaks start to decrease and the time the evening peaks start to increase.

Figure  highlights a very good agreement between the temporal evolution of total eBC, as measured by the aethalometer, and BC, as modelled by FLEXPART analysis. FLEXPART separates the contribution from black carbon into eight different sources (see Fig. S7 for the time series of the individual sources). On average, the main BC sources contributing to the air masses in Aarhus Bay during the campaign were DOM (∼ 49 %), BB (∼ 26 %), and TRA (∼ 19 %), while minor contributions from IND (∼ 3 %), FLR (∼ 2 %), SHP and WSD (∼ 1 %), and ENE (∼ 0 %) were found.

An overall comparison to FLEXPART data yielded a mean fractional bias (MFB) of -51.9 % and a root mean squared error (RMSE) of 221 ng m⁻³. While several previous studies comparing FLEXPART BC concentrations with observations in background stations found good agreement , a comparison in urban areas is more challenging due to the difficulties that arise when local emissions have to be considered. The superposition of transported and emitted BC makes a direct comparison with FLEXPART data difficult, leading to more smoothed modelled lines as compared to the observations. Local pollution is mainly characterized by sharp BC peaks, as seen in Fig. , and the lack of these in the modelled BC indicates missing sources in the emission inventories, thus leading to an underestimation by the model. Similar results were also found in the cases of urban Vilnius (Lithuania) and Warsaw (Poland) . The BC trend in Aarhus Bay was, however, captured very well by FLEXPART.

We selected three special cases from the entire measurement period, characterized by different air mass types.

Case 1: local pollution. This period lasted 48 h, from 18:10 LT on 17 March to 18:20 LT on 19 March. It was characterized by high scattering and absorption coefficients, reaching the highest values during the campaign. The measured wind speed (1.48 ± 1.72 m s⁻¹; median: 3.10 m s⁻¹) and radiation were relatively low during this period, as recorded at the DMI station, providing favourable conditions for particle accumulation. The start time was determined to be when σsca,525 and σabs,520 were higher than their campaign means and when SSA520 was lower. The end time was set to be when σabs,520 decreased below the campaign mean.

Figure S8 presents the footprint analysis for each of the selected cases, while Fig. S9 shows CO, PM₁₀, and NO_x data from the DCE station, highlighting the case periods. Figure S10 shows the time series of wind speed and wind direction during the entire measurement period.