The Iberian Peninsula, located in the SW of Europe, has an area close to 600 000 km2 and a population of almost 60 million, the majority of which distributed along the Atlantic and Mediterranean coastlines, except for some important conurbations such as Madrid, Seville or Zaragoza. Forests (with several pine species commonly present) are scattered through the whole territory. Mountainous areas follow the same trend, with the most elevated chains found in the northern borders (Pyrenees and Cantabria) and in the south (Sierra Nevada). Rural activities can be found almost everywhere, but are particularly important for the economy in the central plateau, where population density is scarcer. A representation of the different land uses in the target domain as represented by the WRF+CHIMERE modelling system can be found in Ratola and Jiménez-Guerrero (2015). In this study, and according to their availability, needles from Pinus pinaster, Pinus pinea, Pinus halepensis and Pinus nigra with up to 1.5 years of exposure to contamination were collected from the bottom and outer branches, placed in sealed plastic bags, kept from light and frozen until extraction. The sampling campaigns were carried out in 33 sites in Portugal and 37 in Spain, in both cases including urban, industrial and rural or remote areas. For further description of these campaigns, the reader is referred to Ratola et al. (2009, 2012).

The analytical procedure used to quantify the levels of PAHs (BaP included) in pine needles was reported previously (Ratola et al., 2009, 2012). A brief description of the methodology and of some characteristics of the pine needles from the different species can be found in the Supplement.

Given the lack of information on atmospheric concentrations of BaP in the sampling sites chosen for this study, an estimation of those values from data provided by biomonitoring studies with vegetation (coniferous needles in this case) was required. Resorting to literature, six approaches (four of them using the same main calculation method, varying only one parameter) were tried and the resulting estimated BaP concentrations were compared with the modelling experiments.

In this study, the Weather Research and Forecasting (WRF) (Skamarock et al., 2008) and the CHIMERE modelling system (Menut et al., 2013), with a resolution of 9 km for the entire Iberian Peninsula coupled to EMEP BaP emissions (Vestreng et al., 2009), was run and evaluated for the Iberian Peninsula in a simulation covering the years 2006 to 2010 on an hourly basis. This CHIMERE version has been modified to include gaseous and particulate BaP. Gas-phase degradation by OH radicals, which represents over 99 % of the degradation path for gas-phase BaP, was accounted for, with a kOH= 5.68 × 10-11 (Schwarzenbach et al., 2003). But more importantly, the oxidation of particulate BaP with ozone was also included, since the respective reaction rate is one order of magnitude higher than other degradation processes, and can be considered the only effective degradation path for particulate BaP in the atmosphere (Bieser et al., 2012). In this case, the reaction constant follows the approach of Pöschl et al. (2001): k=kmax⁡[O3]/(1+KO3[O3]), being kmax⁡= 0.015 s-1 and KO3= 2.8 × 10-13 cm3. A bias adjustment technique was applied and is referred to in the Supplement, together with a description of the modelling set-up and validation procedures (Table S2). All modelled concentrations presented in this work are bias-adjusted.

The BaP concentrations in pine needles used in this work are taken from biomonitoring campaigns previously performed in the Iberian Peninsula (Ratola et al., 2009, 2010a, b, 2012). These data were compared to the deposition over vegetal canopies as estimated by the CHIMERE transport model. The dry deposition flux in CHIMERE is directly proportional to the local concentration C of the target compound (in this case, BaP): F=-vd⋅C, where F represents the vertical dry deposition flux, the amount of material depositing to a unit surface area per unit time. The proportional constant between flux and concentration, vd, is known as the deposition velocity. The main factors governing dry deposition are the grade of the atmospheric turbulence, the chemical properties of the species, and the nature of the soil and the vegetation.

The deposition over vegetal canopies in CHIMERE for particles employs a resistance scheme (Wesely, 1989). The dry deposition velocity follows the formulation of Seinfeld and Pandis (1997): vd=(1/(ra+rb+ra⋅rb⋅vs))+vs, where ra is the aerodynamic resistance (or aerodynamic drag) and rb the resistance at the quasi-laminar sublayer. The aerodynamics resistance is calculated as the integral of the inverse of the diffusivity coefficient Kz up to the middle of the model surface layer, which can be estimated using the analytical formulae of the surface-layer similarity profiles for K (Seinfeld and Pandis, 1997) and vs stands for the sedimentation velocity. For vegetal canopies, as in our case, corrections have been implemented. These corrections are not detailed in the CHIMERE manual (http://www.lmd.polytechnique.fr/chimere/), but rather supported on the literature presented (Giorgi, 1986; Peters and Eiden, 1992; Zhang et al., 2001). For this reason, and for the sake of brevity, the same strategy is adopted here and readers are referred to those works for further details.

The model climatologies for BaP in canopy deposition and air concentration were done under the premise of constituting a base for a broad spectrum of studies within the air-vegetation interactions. In fact, a description of these simulations was mentioned previously by Ratola and Jiménez-Guerrero (2015). However, given the importance for the current study, a summary is presented here, also considering a different perspective.

Databases on the atmospheric levels of SVOCs are already available, but the existing ones (like EMEP) do not cover, for instance, the entire Iberian Peninsula for a climatologically representative period of time (apart from some isolated measurements). In terms of vegetation, the scenario is even worse, but since the presence of SVOCs in such environmental matrices (and in particular in pine needles) reflects entirely an entrapment from the atmosphere (Hwang and Wade, 2008), these measured data can be used not only to validate the model results in vegetation but also to complement the information gathered by the direct atmospheric sampling. For that purpose, six approaches to convert the concentrations found in the 70 sites where pine needles were collected into atmospheric levels were compared to the reference provided by the CTM simulations. This hypothesis is based on the fact that models represent correctly the measured atmospheric concentrations of BaP over the Iberian Peninsula, taking into account the evaluation against EMEP field measurements available. This hypothesis was forced by the lack of simultaneous samplings of vegetation and air concentrations over the target area. Therefore, we used the following methodology: (a) validate simulations with WRF+CHIMERE data against EMEP network measurements, in order to check the ability of the CTM to reproduce atmospheric concentrations over the entire Iberian Peninsula; (b) once proven that errors are acceptable and that the model shows no trend bias, we use modelled atmospheric concentrations as a consistent reference that allows us to compare various vegetation-to-air estimating methods and check which is the most suitable approach for the particular conditions of the area.

It is clear that given the numerous variables and conditions involved, the uptake processes of compounds like PAHs by matrices such as pine needles are not entirely understood (Barber et al., 2004). But the information we have so far indicates that pine needles are valid biomonitors of atmospheric loads, but also can be used to assess the performance of different methods to convert vegetation uptake levels into atmospheric concentrations. Thus, the objective is to test the response of the six vegetation-to-air approaches detailed in Sect. 2.3 through a field and/or model check in the sampling points chosen.

Results (Table 2) reveal that approach 1d is the best fit to convert the levels measured in vegetation into air concentrations, when compared to the outcome provided by the model. This approach was used by Ratola and Jiménez-Guerrero (2015) to assess differences between pine species in modelling simulations as the deposition velocity is in this case defined for an entire forest canopy and not for a given species. This general characteristic is seemingly giving this approach an advantage in terms of the vegetation-to-air calculations. The MFB ranges from -19 % for spring (MAM) to a slight overestimation during winter (DJF, +9 %), being the biases under 3 pg m-3 for all seasons. These errors are relatively low bearing in mind the diversity of the sampling sites considered in this work. Previous works have demonstrated the seasonal variability of PAHs uptake by pine needles (Hwang and Wade, 2008; Ratola et al., 2010a), with the highest levels occurring in winter and the lowest in summer. However, these differences are much more visible in the lighter PAHs (the ones in the gas-phase), given the stronger affinity of the pine needles waxy layer towards their entrapment, when compared to the particulate PAHs.

Being one of the latter, BaP in pine needles may not experience the same level of seasonal variation as in the atmosphere, even if it presents a similar trend. These seasonal differences can be much stronger in the atmosphere, due to the fluctuation of the emission rates from winter to summer. It is then not surprising that the model underestimates the atmospheric concentrations of BaP measured in the colder months and overestimates them in the warmer ones, since in this case the field values are obtained from the levels found in the pine needles. Approach 1d is also the best representation for this seasonal variability (estimated as the standard deviation between approaches and the CTM). Additionally, this approach shows the best air–vegetation relationship simulated by the model, with the rest of the methods providing unrealistic concentrations when compared to the measurements in EMEP stations and modelling results. In fact, approaches 1a and 2 tend to underestimate the modelled concentrations by a factor up to 10, yielding negative biases for all seasons. The rest of the approaches greatly overestimate the levels of BaP (by a factor of 100 in the case of 1c and 3 and of 1000 in approach 1b). These large variations are mainly caused by the difference in the deposition velocities used in each approaches 1a to 1d (from 10.8 m h-1 in 1a to 0.0039 m h-1 in 1b) and in completely different vegetation-to-air estimation strategies in approaches 2 and 3. The deposition velocity has an important role in one of the three methodologies for estimating air concentrations from vegetation (methodology which derives into approaches 1a to 1d), but it allows precisely to understand the differences that may occur when conditions are changed (different species, different locations, different times of the year in the same locations, different affecting sources, etc.).

With respect to the temporal correlation coefficients, approaches 1a to 1d present the same value (0.51), as they rely on the same calculations (only changing the deposition velocity). This is an acceptable description of the temporal variability observed in all sites. Approach 2 is not able to reproduce these time series (correlation coefficient of -0.55), but, interestingly, it is approach 3 that presents the best correlation (0.80). In this latter case, although the bias for the BaP concentrations is quite high, the r value can be related with the different uptake efficiencies pine needles show for gas-phase or particulate PAHs. The two equations suggested by Chun (2011) to relate concentrations of PAHs in needles and air separate the lighter from the heavier ones. So even if the actual concentrations are not very well described, the temporal air-needles synergies may be better projected by this approach in this particular case.

Finally, spatial correlation coefficients (which provide a simulation for the adequate representation of the BaP spatial patterns over the Iberian Peninsula) are correctly reproduced by all approaches (Table 2). The highest value is seen for winter in approach 2 (r= 0.68) and for the rest of the seasons, approaches 1a–1d present the higher correlation coefficients (from 0.67 in JJA to 0.85 in MAM). Approach 3 generally offers the lowest spatial correlation coefficients for all seasons, except in summer. The fact that the lowest r values are generally found for winter and summer (also the extremes of BaP concentrations in the environment), highlights the limitations of the model to represent these extremes.

Ideally, the air levels SVOCs are measured in the field using expensive active air sampling equipment which also require permanent power supply while operating. Thus, these devices only exist in certain parts of the world, which does not allow a proper coverage of the global presence of such contaminants, which naturally hinders the efforts of modelling estimation as well. As mentioned above, as living structures vegetation matrices have morphological, physical and chemical behaviour that depends on many parameters, even within the same species. Thus, the equations describing the air-vegetation partition suffer from these effects when a broad solution is searched for. Again in ideal terms, only a direct comparison of field campaigns and active air sampling performed in the same spots is bound to achieve some accuracy, if it includes a seasonal framework as well. In fact, the main approaches presented in this work derive from these types of combined studies. But when it is impossible to have simultaneous active air and biomonitoring sampling models can help us to assess if the assumptions we are working with are sound, if a previous validation with the field-based air concentrations were successful (as is the case in our study). Naturally, there is a concern that the uncertainty associated to all the steps involved may affect the conclusions of a study like this. Even if a detailed analysis were to be extremely complex and out of the scope of this work, the main source of uncertainty of our global process can be identified: the emission inventories for PAHs, as stated by San José et al. (2013). In general, this uncertainty was estimated to be within a factor of 2 to 5 (Berdowski et al., 1997), much larger than any other uncertainty associated with the validation process and rest of steps. For instance, EMEP individual measurements should have a precision within ±10 % and the data quality objectives for the sampling and chemical analysis set a combined uncertainty between 15 and 25 % (EMEP, 2001). Also, the analytical methodology to quantify BaP in pine needles has similar precision values (Ratola et al., 2009). The contribution of these processes to the global uncertainties would be reduced in comparison to the BaP emissions.

Information on pine needles characteristics, sampling, analytical methodology, as well as on the modelling and vegetation-to-air estimation strategies. This material is available in the Supplement free of charge via the Internet.