Quantitative study on nitrogen deposition and canopy retention in Mediterranean evergreen forests

To assess the impact of nitrogen (N) pollutants on forest ecosystems, the role of the interactions in the canopy needs to be understood. A great number of studies have addressed this issue in heavily N-polluted regions in north and central Europe. Much less information is available for the Iberian Peninsula, and yet this region is home to mountain forests and alpine grasslands that may be at risk due to excessive N deposition. To establish the basis for ecology-based policies, there is a need to better understand the forest response to this atmospheric impact. To fill this gap, in this study, we measured N deposition (as bulk, wet, and throughfall fluxes of dissolved inorganic nitrogen) and air N gas concentrations from 2011 to 2013 at four Spanish holm oak (Quercus ilex) forests located in different pollution environments. One site was in an area of intensive agriculture, two sites were influenced by big cities (Madrid and Barcelona, respectively), and one site was in a rural mountain environment 40 km north of Barcelona. Wet deposition ranged between 0.54 and 3.8 kg N ha−1 year−1 for ammonium (NH4+)-N and between 0.65 and 2.1 kg N ha−1 year−1 for nitrate (NO3−)-N, with the lowest deposition at the Madrid site for both components. Dry deposition was evaluated with three different approaches: (1) a canopy budget model based in throughfall measurements, (2) a branch washing method, and (3) inferential calculations. Taking the average dry deposition from these methods, dry deposition represented 51–67% (reduced N) and 72–75% (oxidized N) of total N deposition. Canopies retained both NH4+-N and NO3-N, with a higher retention at the agricultural and rural sites (50–60%) than at sites located close to big cities (20–35%, though more uncertainty was found for the site near Madrid), thereby highlighting the role of the forest canopy in processing N pollutant emissions.


Introduction
Quantifying nitrogen (N) atmospheric deposition to forests is a key issue to understand nutrient availability for forest growth and to assess the forests status regarding excess N deposition (Johnson and Lindberg 2013).Even though N deposition is of concern for many ecosystem types, forests probably receive larger deposition loads, mainly due to their greater aerodynamic roughness that favors the capture of gases and fine particles (Gallagher et al. 1997).Therefore, high N deposition affects forest ecosystem compartments comprising vegetation, soil, soil water and the animal, fungi and microbial biota (Sutton et al. 2011).
Wet deposition (WD) or bulk deposition (BD), which includes part of coarse particle fallout, can be quite straightforwardly quantified with wet or bulk collectors.However, for dry deposition (DD) no standard method exists, and various approaches have been used for its determination (Hanson and Lindberg 1991).Dry deposition to foliar surfaces refers to the transfer of gas and particulate species between the atmosphere and vegetation surfaces in the absence of precipitation.Several processes control dry deposition, such as ambient gas and aerosol concentrations, physicochemical characteristics of the species of interest, canopy characteristics and the site prevailing meteorology (Hanson and Lindberg 1991).For N species, the deposition behavior of N gases can be separated in two main groups: 1) highly reactive and water soluble gases (HNO3 and NH3) that are readily deposited on leaf surfaces , and 2) less soluble gases that diffuse through the stomata (NO, NO2 and partly NH3, Hosker and Lindberg 1982).Also, HNO3 can be transported via cuticular uptake (Padgett et al. 2009).Gaseous HNO3 and NH3 can be incorporated into atmospheric particles, mostly through reactions with sulfate and nitrates to form fine particles, but they can also react with soil dust and sea salts to form coarse nitrate particles (Querol et al. 1998;Pey et al. 2009).These N-containing particles (pNO3 -and pNH4 + ) may be incorporated into cloud water and be deposited via wet deposition, and also they can deposit via dry deposition to the leaf surfaces (Hanson and Lindberg 1991).
To estimate dry N deposition to forests, micrometeorological methodologies, such as eddy correlation and the aerodynamic gradient method have been widely used.These methods are economically costly and cannot be applied to sites with complex topography (Hicks et al. 1991).
To overcome these drawbacks, approaches based on recovering accumulated deposition on deposition surfaces have been developed, e.g the throughfall and branch washing methods.
Throughfall measurements (collection of precipitation water that has passed through the canopy) have been frequently used and net throughfall fluxes have been used as indicators of dry deposition (De Schriever et al. 2007).However, nitrogen compounds can experience exchanges and transformations in the canopy that need to be taken into account when determining N dry deposition (Parker 1983;Hanson and Lindberg 1991).This is particularly true for sites in low pollution environments with moderate N deposition loads where canopy N retention and transformation by canopy epiphytes and microorganisms may have a higher relative contribution (Guerrieri et al. 2014).Branch rinsing techniques have also been widely used to recover deposited N compounds from foliar surfaces (Bytnerowicz et al. 1987(Bytnerowicz et al. , 2015;;García-Gómez 2016).
Models indicate that western Europe may be particularly affected by high N deposition in 2030 under current legislation scenarios (Dentener et al. 2006).In fact, empirical N critical loads set for the protection of terrestrial habitats under the Convention on Long-Range Transboundary Air Pollution (CLRTAP) are being currently exceeded in some habitats of Community interest of the Spanish Natura 2000 network (García-Gómez et al. 2014).N deposition estimated with the EMEP and CHIMERE models indicated that a surface of 3785 km 2 (modeled with the EMEP model) and 1441 km 2 (modeled with CHIMERE) corresponding to habitats of the Annex 1 of the Habitats Directive received N deposition that exceeded the habitat critical loads (García-

Gómez et al. 2014).
Other studies indicate N enrichment in forest ecosystems in Spain, such as the observed increase of N content in herbarium bryophytes collected in the 20th century (Peñuelas and Filella 2001), the increase of nitrophilous species in natural areas from the Spanish Natura 2000 network (Ariño et al. 2000) and the increased streamwater nitrate concentrations in headwater streams (Avila and Rodà 2012).On the other hand, N deposition has been related to acidification, with implications on plant nutrition and soil microbial community structure in pine forests in central Spain (Ochoa-Hueso et al. 2014).
Data on N deposition and the contribution of DD to total N deposition is rather scarce in Spain.
In a study of five rural localities in NE Spain, wet N deposition ranged between 4 and 7 kg N ha - 1 y -1 and total N deposition was in the range of 12-19 kg N ha -1 y -1 with dry deposition accounting between 50 to 70% of total N deposition (Avila et al. 2010).In central-western Spain (Salamanca region), wet deposition ranged between 3-5 kg N ha -1 y -1 , but dry deposition (estimated with the regression method of Lovett and Lindberg 1984) was only 0.8 and 1.5 kg N ha -1 y -1 and made a lower contribution to total N deposition amounts (25-45%; Moreno et al. 2001).
Recently, research has been carried out in Spain to describe N deposition in holm oak forests and major advances have been done in the quantification of dissolved organic nitrogen (DON) deposition (Izquieta-Rojano et al. 2016), in testing methods for wet and throughfall deposition sampling (García-Gómez et al. 2016b), in analysing the effect of forests to improve air quality (García-Gómez et al. 2016a) and modelling N deposition at a Spanish scale (García-Gómez et al.

2014).
In this paper, we will provide information on total N deposition fluxes and assess the role of dry deposition and canopy uptake based on wet deposition, throughfall and ambient gas measurements from 4 holm oak forests under different pollution environments in the Iberian Peninsula.Dry deposition is evaluated with three different approaches (a canopy budget model based in throughfall, branch washings and inferential calculations).The range of the obtained dry deposition values is used to provide a tentative total deposition estimate and to evaluate the scope of canopy uptake in these forests.

Locations and experimental sites
The study was conducted at 4 holm-oak forests (Quercus ilex L.) in the north, center and northeast of the Iberian Peninsula (Fig. 1).Two sites were located in Catalonia in NE Spain near Barcelona (La Castanya and Can Balasc, LC and CB respectively), one in Madrid (Tres Cantos, TC) and another site in Navarra, North Spain (Carrascal, CA).The main characteristics of the sampling sites are shown in Table 1.
The LC site (41º46'N, 2º21'E, 696 m.a.s.l.) is located in the Montseny Mountains, 40 km to the NNE of Barcelona.This site is considered as a rural background station with some influence of pollution from the metropolitan area of Barcelona.Vegetation at LC consists of a dense and closed canopy forest dominated by holm-oak (Quercus ilex L.) trees.Lithology at this area is composed by schists and granodiorites.Climate is Mediterranean, with a clear seasonal cycle with lower precipitation in summer and winter.
The CB site (41º25'N, 2º04'Eº, 255 m.a.s.l.) is located in the Collserola Natural Park, a protected area lying to the west of the Barcelona Metropolitan Area (3.5 million inhabitants).The plot lies at 4 km linear distance from Barcelona outskirts.A moderate to heavy traffic highway (C-16) runs about 150 m from the study plot, and it is affected by industrial emissions from the Baix Llobregat area (García-Gómez et al. 2016a).Vegetation at CB is characterized by a continuous cover of holm-oak (Quercus ilex L.) mixed with Quercus humilis Mill.Lithology consists of shales and slates with granitic outcrops.Climate is Mediterranean.
The CA site (42º39'N, 1º38'W, 645 m.a.s.l.) is situated at the foot of the Alaitz-Izco hills, in central Navarra.The nearest larger city, Pamplona (~200 000 inhabitants) is 15 km to the North.The site is about 50 m distant from a moderate to heavy traffic highway (AP-15) and is surrounded by fields of irrigated and fertilized cereal that have been found to influence N organic and inorganic inputs to this site (Izquieta-Rojano et al. 2016).An opencast limestone quarry is located approximately 2 km to the north.The forest comprises mostly Quercus ilex L. trees with scattered Quercus faginea Lam. and Quercus humilis Mill.individuals.The site lies on calcareous soils.The climate at CA is Mediterranean continental with oceanic influence from the Atlantic sea.
The TC site (40º35'N, 3º43'W, 705 m.a.s.l.) is located 9 km NE from Madrid outskirts (3.2 million inhabitants).The site lies in the north-eastern border of the holm-oak forest of El Pardo, which extends over an area of 170 km 2 and is a protected area.Vegetation was historically managed as a traditional 'dehesa', a savannah-type managed formation of low density isolated trees.The low level of management during the last decades has allowed the vegetation to grow as an open low density forest with an understory of shrubs and grasslands.Lithology is composed by sandy arkoses sediments from granites and gneisses.A moderate to high traffic intensity highway (M-607) is ~ 2 km distant from the monitoring site.The climate is continental Mediterranean, characterized by long dry periods and a more contrasted seasonality than the typical Mediterranean climate.

Field sampling and bulk deposition and throughfall chemical analysis
In every location, an open-field (for bulk deposition, BD) and a below-canopy plot (for throughfall, TF) were instrumented.The same model of sampler was used for bulk and throughfall deposition collection at all sites, composed of an ISO-standardized funnel (Norwegian Institute for Air Research, NILU) with a 314 cm 2 horizontal interception surface, connected to a polypropylene 2 L bottle.A bug sieve was placed at the funnel neck to prevent leaves and other materials from entering into the bottle.The upper edge of the funnel was equipped with an external ring to prevent contamination from bird droppings.The rim of all funnels stood approximately at 1.5 m above ground level.For bulk sampling, two collectors were used per site at LC and CB, and 4 at CA and TC.For throughfall sampling, 12 collectors were used at all sites; they were randomly located in a forest plot of 30*30 m at LC, CB and CA.At the dehesa-like forest of TC, the collectors were randomly placed in different orientations under dominant trees.Wet deposition (WD) was also measured at LC and TC in the open-field plot, by means of an automatic Andersen sampler (ESM Andersen instruments, G78-1001) consisting on a wet and a dry bucket covered with a moving lid that covers the wet collector in dry periods and moves to open WD the collector at the onset of rain.All funnels and WD buckets were thoroughly cleaned in the field with deionized water after each sampling.Bulk and throughfall sampling bottles were retrieved and replaced by clean ones at each site.Field blanks (recovered distilled water after rinsing the funnels and buckets in the field) were periodically obtained and analyzed.
Sampling took place from June 2011 to June 2013 in a weekly schedule or biweekly in case of rainless weeks.All collected samples were filtered with 0.45 m size pore membrane filters of cellulose (Millipore) and frozen until analysis.Ammonium (NH4 + ) and nitrate (NO3 -) were determined by ion chromatography at all sites.Analytical accuracy was checked with internal control samples of known concentrations, with differences being lower than 10%.In addition, all major anions and cations in the precipitation and throughfall samples were analyzed by ion chromatography (Dionex, Sunnyvale, USA) and an accuracy check for analytical quality was applied based in recommendations of the ICP-Forests manual (2010).The balance of the sum of cations and anions, and the calculated conductivity related to the measured one was also scrutinized and outliers (>10%) were discarded (Izquieta-Rojano et al. 2016).Detection limit for NO3 -and NH4 + was 1.5 eq L -1 .
Precipitation amount has been found to vary depending on the device employed for measurement (Erisman et al. 1994).In this study, precipitation and throughfall amounts were obtained from the water volume collected in bulk collectors, divided by the collector exposed surface, and expressed as Lm -2 .To ascertain the accuracy of these measurements, we compared the water depths recorded by 4 different sampling devices deployed in parallel from August 2011 to June 2013 at the LC site: 1) a wet Andersen collector, 2) two replicated bulk collector buckets, 3) a Hellmann standard rain gauge and 4) a Campbell tipping bucket rain gauge.An ANOVA analysis performed on log-transformed weekly data indicated non-significant differences between these measurement methods (p=0.76).

Gas and particulates sampling and analysis
Atmospheric concentrations of ammonia (NH3), nitrogen dioxide (NO2) and nitric acid vapor (HNO3) were monitored from February 2011 to February 2013 using passive samplers.A full description of the sampling is given in García-Gómez et al. (2016a).Although an open-field and a below-canopy plot were installed in each plot, here we will only consider open-field measurements.The open plots were >500m distant from the forest edges to achieve a proper exposure to ambient concentrations.
Two replicate passive samplers per gaseous species were exposed during two-week periods at 2 m height in each plot.In parallel, unexposed samplers were used as blanks for each site, period and type of sampler.After collection, all samples were kept refrigerated (4º C) in darkness until analysis.Tube-type samplers (Radiello®) were used to measure atmospheric concentrations of NH3 and NO2.Tubes were extracted according to Radiello's specifications (Fondazione Salvatore Maugeri, 2006).Atmospheric concentrations of HNO3 were measured by means of badge-type samplers manufactured following Bytnerowicz et al. (2005).In CA, Passam® passive samplers and methods were employed during the second year for monitoring NO2 after checking their comparability with Radiello®.
Particulate matter with diameter up to 10 µm (PM10) was collected with 150 mm quartz microfibre filters (2500 QAO-UP, Pall Life Sciences) using high volume samplers installed in open-field plots of TC, CA and LC sites (Digitel ® DH80 in LC -MSY monitoring station; MCV ® CAV-A/mb in TC and CA).Samples were collected from February 2012 to February 2013 once a week, using a flow of 30 m 3 h -1 during 24-h periods.The day of the week for PM10 collection changed weekly.
The concentration was gravimetrically determined and NO3 -and NH4 + were water-extracted and analyzed by ion chromatography.For statistical comparison with gaseous pollutant concentrations, PM10 data were grouped and averaged in accordance to passive sampling.

Data handling and statistical analysis
Annual BD and TF mean concentrations were calculated as volume-weighted means (VWM, expressed as µeq L -1 ).Annual BD and TF fluxes were obtained as the product of their respective VWM by the annual precipitation or throughfall volume and are expressed as kg N ha -1 y -1 .
The Kruskall-Wallis was applied to explore differences in rainfall amount or N compounds and the Wilcoxon signed-rank test was used to determine differences between site pairs.

Dry deposition estimation
In this work, an estimation of dry deposition fluxes is proposed based on three model approaches: 1) canopy budget model (CBM), 2) branch surface washings (BW), and 3) inferential model with Vds obtained from references in forest studies (IM).

Canopy budget model
A complete description of this model is given elsewhere (Drraijers and Erisman 1995, Balestrini and Tagliaferri 2001, Staelens et al. 2008, ICP-Forest Manual 2010, Adriaenssens et al. 2012, Drapelova 2013) and here we will give a brief summary.The model is based on the balance: nTF = TF -WD = DD + CE (eq. 1) where nTF stands for net throughfall, TF for throughfall, WD for Wet Deposition, DD for Dry Deposition and CE for Canopy Exchange.Canopy exchange can be positive and then it is attributed to leaching of ions from the leaf pool (canopy leaching, CL) or be negative and then it is attributed to the uptake/transformation of the deposited ions (canopy uptake, CU).
The aim of the CBM is to distinguish and make an apportionment of DD and CE fluxes.To this purpose the filtering approach proposed by Ulrich (1983) is generally used.This considers that some ions do not interact with the canopy and then their enrichment in nTF is solely due to DD.
Here we have used Na as reference ion.Other ions in aerosols (e.g base cations, SO4 -2 , Cl -) are considered to behave as the Na-containing aerosols and, therefore, are considered to deposit at similar rates as the reference ion.Nitrogen compounds, which in our sites are mostly deposited as gases (García-Gómez et al. 2016a) do not comply with the above assumptions and another approach has to be taken: N exchange is determined first and then DD is derived from equation 1.It has been proposed (Balestrini and Tagliaferri 2001;Staelens et al. 2008) that the NH4 + canopy uptake (CU) flux can be estimated by considering that its canopy uptake equals the canopy leaching of base cations (the sum of leaching of Ca 2+ , Mg 2+ and K + ) once corrected by the sum of all leached anions (Staelens et al. 2008;Zhang et al. 2006).Several studies have only taken into account weak acid leaching (Adrieanssens et al. 2012;Balestrini and Tagliaferri 2001;Thimonier et al. 2005), but since our data also suggested Cl -leaching (Aguillaume et al. 2017), it was also included in the sum of leached anions.Besides, experimental and field work has shown that NO3 -can also be retained by the canopies (Harrison et al. 2000;Stachurski and Zimka 2002;Fenn et al. 2013).It has been proposed that the CU of NH4 + + NO3 -can be calculated based on the TF fluxes of both, distributing their relative CU weight by using an efficiency factor of NH4 + vs. NO3 -uptake (xNH4= moles of NH4 + taken up for each NO3 -mol) (De Vries et al. 2003;Staelens et al. 2008).
In the CBM method, the use of WD provides more accurate DD estimate than BD, since the later includes a fraction of DD (coarse particle DD).Since WD was not sampled at CB and CA their BD value was corrected by the ratio WD/BD from LC and TC (0.76 and 0.69 for NH4 + and 0.65 and 0.67 for NO3 -at LC and TC respectively).Since the ratio differences between sites were small, we used the averaged WD/BD of the two sites (0.72 and 0.65).
In the CBM, calculations are made on an equivalent basis but we express results in kg N ha -1 y -1 for comparison with the other methods.
The NH4 + and NO3 -dry deposition values estimated with this method depend on the DDs of base cations which are estimated with the filtering method.The accuracy of the model is affected by analytical errors in base cations and anions.To minimize biases, the analytical accuracy was scrutinized in all WD and TF samples with the protocol of the ICP-Forest Manual ( 2010), and values differing by >10% of the charge and the conductivity balances were discarded.Another source of uncertainty in the CBM is the efficiency factor of NH4 + vs. NO3 -uptake.We have used here a value of xNH4=6, which is backed up by experimental work in holm oak saplings (Uscola et al. 2014) and also is the one proposed at an European scale (de Vries et al. 2003).

Branch washings
At each site and for rain-free periods of > 7 days distributed along the year in the period June 2011 to June 2013 (Table 2), deposition measurements were made by washing selected holm oak branches: one branch of 20 cm length was cut from the top of 10 selected trees at each site.
The branch tips were sealed with Parafilm and carried to the laboratory in sealed plastic bags where they were washed for 3 min with 200 mL distilled water.Bag blanks were also obtained.
The branch exposure was considered to begin at the end of the previous rain producing throughfall.Linear regressions between precipitation and throughfall have been generally used to describe the canopy storage capacity (Zinke 1967).These regressions were explored for the study sites and indicated a storage capacity (in mm) of 2.8 for LC, 1.5 for CB and CA, and 0.9 for TC (with correlation coefficients of 0.98-0.99).Therefore, we considered that rainfalls greater than these quantities were an adequate starting point.In general, previous rainfall was well above the storage amounts (except for two occasions at TC of 1-1.5 mm), as shown in Table 2. Thus, we considered these previous rain amounts to be sufficient to wash previous dry deposition, though we are conscious that deposition obtained from branch washings is probably an overestimation due to the fact that evaporation of intercepted rainfall would leave an ionic residue from the previous rain event.To overcome this pitfall, branches would need to have been washed with distilled water at the onset of each sampling period, but this was not possible at our study sites for technical, logistic and economic reasons.We are confident that the exposure periods were long enough at all sites to minimise the contribution of the residual previous deposition.In fact, at TC, the exposure time was longer than one month for the periods of low antecedent precipitation (Table 2).
After washing, branches were air-dried and the leaf surface of each branch was obtained from Li-Cor 3100 area-meter measurements.The washing solutions were analysed by ion chromatography (Dionex, Sunnyvale, USA) with the same quality controls as reported above.
The N deposition flux to branches was calculated as the product of the NH4 + or NO3 - concentrations in the washing solutions (corrected for blanks) by the volume used (200mL) and divided by the exposure duration (in days) and the projected leaf area (in cm 2 ) to obtain the daily surface flux deposition to branches.To extrapolate to fluxes to canopy and year, the daily flux to branches was multiplied by each site's LAI and days for the year, and is expressed as kg N ha -1 y -1 .

Inferential model
The inferential method is based on the assumed steady-state relationship: where the dry deposition flux (Fa) is a product of the dry deposition velocity (Vd) and the concentration (Ca) of the considered air pollutant (a).It involves the measurements of pollutant air concentrations and modeled Vds.In our sites, N gas and particle atmospheric measurements were available for the period of the study (only 2012-2013 for particle measurements).
However, models for Vd need data of meteorological variables taken at high frequency (Wesely and Hicks 2000), which were not available at the sites.We provide here a preliminary analysis of DD fluxes based on a compilation of Vd values from literature reports from forests studies that have applied the inferential method (Tables 3 and 4).
The different methods applied for the DD estimation are based in different approaches, and each of them has its own particularities and drawbacks, which are briefly examined here.The CBM considers ion exchanges of the N compounds at the leaf surfaces and is based on the equilibrium of charges between all ions reaching the canopy.This method was developed to overcome the difficulties in interpreting TF results derived from the fact that TF includes both deposition and exchanged ions from the canopy.The exchange processes comprise ion diffusion or exchange between the water layer covering the leaves and the apoplast (Bytnerowicz et al. 2015;Padgett et al. 2009).Stomatal and cuticular uptake of some N gases (e.g.HNO3, NH3, NO2) can also occur and modify TF fluxes if they are dissolved in the leaf surface or within stomata (Draaijers et al. 1997;Gessler et al. 2002;Bytnerowicz et al. 2015).On the other hand, the CBM method needs to take into account an efficiency factor of NH4 + vs. NO3 -uptake (xNH4= moles of NH4 + taken up for each NO3 -mol).To account for this, we took advantage of an experimental work of N uptake with holm oaks to better attune this value to the studied species (Uscola et al. 2014).Interestingly, the value obtained (xNH4 =6) was the same to that proposed for European forests (de Vries et al. 2003).Although the CBM approach has the important drawback that analytical errors will propagate through the enchained calculations, it is a method widely in use (Thimmonier et al. 2005;Balestrini et al. 2007;Staelens et al. 2008;Adriaenssens et al. 2012;Drapelova 2013) and it is more appropriate to describe the ongoing canopy processes than the assumption that nTF is equivalent to DD for N compounds.
The branch washing method is also a direct method for measurement of dry deposited material to leaf surfaces.By excluding wet episodes, this approach expects to reduce cuticular exchanges, which are favored by the dissolution of compounds to water films, though some uptake or transformation of the deposited chemical species may also occur (Hanson and Lindberg 1991).
This method is similar to TF measurements, the most important difference being its more episodic sampling nature (4-5 periods during the year vs. weekly/biweekly sampling for throughfall) and the avoidance of wet deposition.
The inferential method applied here relied on measured air concentrations and Vd from the bibliography.While differences between sites in Vd are expected due to variation of meteorological and canopy structure factors, the Vds of the different constituents were quite consistent over the range of forests surveyed (Table 3), so we considered the average Vd from the compiled data in Table 3 for flux calculations presented in Table 4.The inferential method is the sole method of those examined that allows to differentiate gas and particle deposition since both TF and BW give the sum of N oxidized and N reduced compounds in the form of NH4 + -N and NO3 --N fluxes.For comparison with TF and BW, inferential calculations have considered the sums of HNO3 + pNO3 -(N oxidized), NH3 + pNH4 + (N reduced), and as such are presented in Table 4.

Water fluxes
Rainfall amount differed markedly between sites, with TC being the driest location (Fig. 2).
Precipitation was very variable between years, particularly at CA and TC where the second year experienced a wetter spring and winter.All sites, except CA, showed a seasonal pattern with spring and autumn receiving significantly higher precipitation than winter and summer (Fig. 2).
At CA, higher precipitation in the 2012-2013 winter resulted in the average winter precipitation not significantly differing from that of spring and autumn.precipitation from the Atlantic.The frequency of these fronts is higher in winter and spring, thus this site differs in the seasonal precipitation from the other sites which present dry winters (Fig.

2).
Throughfall was highly correlated with rainfall (r 2 = 0.97; 0.98, 0.96 and 0.73 for LC, CB, CA and TC respectively).Similarly to rainfall, significant differences with higher throughfall in the wet spring and autumn seasons were found at all sites, except at CA.

The difference between precipitation in the open (BD collectors) and throughfall (TF) indicates
the water quantity intercepted by the canopies (In).The lowest interception was at TC (Table 5), and this is attributed to the open structure of this site (Table 1) that will allow for direct passage of rainfall to the soil thus avoiding evaporation on the canopy.In a revision of rainfall partitioning in Mediterranean forests and shrubs, lower interception was found for forests with lower leaf area index, basal area, and height (Llorens and Domingo 2007), though no relationship with tree density was observed.

Nitrogen fluxes in wet deposition and throughfall
Annual wet deposition and throughfall fluxes for NH4 + -N and NO3 --N and the sum of inorganic N (DIN) are shown in Table 6, where the percent contribution of NO3 --N to DIN is also indicated.
Wet deposition ranged between 0.54 and 3.8 kg N ha -1 y -1 for NH4 + -N and between 0.65 and 2.1 kg N ha -1 y -1 for NO3 --N , with the lowest deposition being at TC for both components, and the highest at CA and LC for NH4 + -N and NO3 --N, respectively.The high NH4 + -N deposition at CA can be attributed to the intensive agriculture activities in surrounding fields where ammonium nitrate and urea fertilizers are regularly applied in winter and spring (Izquieta-Rojano et al. 2016).Wet deposition at LC can be compared to previous wet deposition measurements in 2002-2003 and 2009-2010 from the work of Izquierdo and Avila (2012).In the period of the present study, NH4 + -N was 33% and 40% lower respectively than these previous periods and NO3 --N was 13% and 25% lower.An analysis of the trends in atmospheric deposition at the LC site for the last 30 years only found a significant declining trend for NO3 -concentrations, not for N fluxes in both forms (Aguillaume et al. 2016).This indicates that the study period had a particularly low wet N deposition, probably a result of low NH4 + and NO3 -concentrations and lower precipitation than the compared periods.Of the two sites located in NE Spain, and contrary to expectations, the semi-urban site close to Barcelona (CB) had lower wet deposition of N compounds than the more remote site LC.Part of this difference may stem from lower precipitation at CB (Table 5), since the difference in VWM concentration in WD between these sites was small (WVM of 17 and 14 eq L -1 for NH4 + and 16.1 and 15.9 eq L -1 for NO3 -at LC and CB respectively).This result indicates that the LC site, which has been taken as a rural background station, was also affected by urban and industrial pollution from the Barcelona metropolitan area, as also found for aerosols (Pey et al. 2008;Pérez et al. 2009) and HNO3 gases (García-Gómez et al. 2016a).In fact, Aguillaume et al. (2016) showed that NO3 -concentrations in bulk deposition at this site was mainly explained (r 2 =0.85) by NOx air concentrations in Barcelona city center, national NO2 Spanish emissions and the amount of precipitation.
The site in central Spain had the lowest wet deposition inputs owing to the combination of low precipitation at this site and lower rain concentrations due the predominant air mass fluxes coming from low polluted areas in the west and the Atlantic Ocean (Salvador et al. 2011).
DIN wet deposition ranged between 1.2 kg N ha -1 y -1 at TC and 5.8 kg N ha -1 y -1 at CA, and the northeastern sites had an intermediate value of 3-4 kg N ha -1 y -1 (Table 6).NH4 + -N and NO3 --N showed a similar contribution to DIN, except at the agriculture-affected CA site, where NH4 + -N was dominant (Table 6).
Throughfall NO3 --N deposition was higher (range 1.8 and 5.4 kg N ha -1 y -1 ) than NH4 + -N (range 0.5 and 3.1 kg N ha -1 y -1 , Table 6).Similarly to WD, the lowest TF deposition for both N components was found at TC, while CA and CB showed the highest NH4-N TF and NO3-N TF fluxes, respectively.The DIN flux that reached the soil varied between 2 and 7.4 kg N ha -1 y -1 in the studied sites, with the lowest N input in TC.In a study of ICP-Forest plots at a European scale, elevated nitrate concentrations in seepage water were found over a threshold of 7 kg N ha -1 y -1 in DIN TF input (De Schrijver et al. 2007).In our study, DIN TF inputs at the site with high agricultural influence (CA) and the one close to Barcelona (CB) were above the proposed threshold value, and might be the more susceptible to soil solution N enrichment.Consistently with the above study, the LC site which receives a TF DIN input lower than the proposed threshold retains N the catchment scale, although the ratio N export/N input was found to increase in recent years (Aguillaume et al. 2016).
Forest canopies play a significant role in altering deposition of N compounds, either because of its filtering effect to capture dry deposition or because of their capacity to retain, take up or transform N species (Sparks 2009).Net throughfall fluxes (nTF), the difference between TF minus BD, indicate the net contribution of the canopy to below canopy fluxes.For inorganic N, lixiviation may be negligible (Rodrigo and Avila 2002), therefore positive nTF fluxes indicate that dry deposition is higher than canopy retention, while negative values indicate that the canopy retains more than the dry deposited amounts.
Net throughfall was positive for NO3 --N but negative for NH4 + -N at LC and CA (Table 6), suggesting that reduced N was more efficiently retained in the canopy than the oxidized forms of N, as it has been shown with 15 N labeled rain experiments (Boyce et al. 1996), by surrogate surface washings (Ignatova and Dambrine 2000), and as is suggested by differences in N gas concentrations in the open and below the forest at the study sites (García-Gómez et al. 2016a).
A negative NH4 + -N nTF flux was found at the agriculture site receiving the highest NH4 + -N wet deposition, thereby suggesting a strong ability of this holm oak forest to retain high N inputs.A similar nTF value was found at LC, though this site received 40% less NH4 + -N inputs (Table 6).

Dry deposition estimation
In this work, three approaches have been used to derive DD: 1) canopy budget model (CBM), 2) branch washing (BW), and 3) inferential model with Vds obtained from forest studies (IM, Table 4).The resulting estimates from the three methods are shown in Table 7.
The different approaches show fairly consistent estimates given the various assumptions in the different methods.Relative differences between methods were more pronounced in TC, the site with lower DD: differences between the lowest and highest values were approximately 70-80%.
However, in absolute terms, these differences were of 1 kg N ha -1 y -1 for NH4 + -N and 3 kg N ha - 1 y -1 for NO3 --N, similar to differences at the other sites (except for NO3 --N at CA).Estimated DD values with the three approaches matched better for N reduced than for N oxidized deposition (Table 7).For N oxidized deposition, a good match was observed between CBM and IM methods, but BW estimates were about double (CA, TC) or 40% greater (LC) than the other estimates.
When considering DIN dry deposition from the applied methods, the range of estimates was of 6 to 11 kg N ha -1 y -1 for the NE Spain sites, 8-14 kg N ha -1 y -1 for CA and 1.5 to 6 kg N ha -1 y -1 for TC (Table 7; Fig. 3).The range of variation of these estimates (1.5 to 4) is similar to that reported in a study that compared 4 inferential models in a network of 55 monitoring sites in Europe, in which between-model differences were of a factor 2-3 (Flechard et al. 2011).
Considering the averages between minimum and maximum estimates (Table 8), the contribution of DD to TD was of 51 to 67% for reduced-N and 72 to 75% of oxidized-N compounds.DIN dry deposition contributed between 65 to 71% to total DIN deposition, indicating the importance of taking into account the dry deposition flux when tackling with the effects of N deposition to ecosystems.
In various locations along the Levantine coast of Spain, dry deposition percentages were similar to the values in this study: 58% for N reduced and 60% for N oxidized forms (Avila and Rodà 2012).However, in oak (Quercus pyrenaica) forests in subhumid western Spain, the DD contribution was lower (10-20% and 30-40% for reduced and oxidized-N respectively (Moreno et al. 2001); though differences in procedure also may have a role since DD was estimated with the regression method of Lovett and Lindberg (1984).
In the USA, a recent study based in 37 localities has reported a substantial decline in oxidized-N emissions that leads to an ammonium-dominated atmospheric composition.Under these conditions, dry deposition of NH3 has been found to play a key role in N deposition, contributing 19-65% of total deposition (Li et al. 2016).In agricultural and rural locations in northern China, reduced-N contributed similarly (28-60%) while oxidized-N represented only 13-30% of total N deposition (Pan et al. 2012).In contrast, in our sites DD of oxidized compounds was the dominating deposition flux (Table 8).This agrees with the fact that NOx emissions in Spain are about triple of NH3 emissions and only started to decline since 2005 (Aguillaume et al. 2016).

Canopy uptake
The ranges of canopy uptake (calculated as TD minus TF) for the different N compounds are shown in Table 9 and Fig. 3.It is seen that N is retained either in the oxidized or reduced forms: the values were similar in both N forms at LC, but were higher in the NH4 + -N form at CB and CA, and in the NO3 --N form at TC.Many findings derived from labeled 15 N experiments have shown retention and stomatal uptake and transformation of dissolved and gaseous N species on foliage (Garten and Hanson 1990;Gaige et al. 2007).Microbial transformations of N deposition can also alter the N forms, transforming inorganic N to organic forms (Cape et al. 2001, Neff et al. 2002) that may explain part of the inorganic N reduction.On the other hand, nitrification in the canopy has been shown to be of significance in beech forests (Guerrieri et al. 2015), a process that may also account for part of the NH4 + "retention" in the canopy Canopy uptake was highly correlated with wet deposition (r 2 =0.992; p<0.001) and total DIN inputs (Fig. 4), indicating that these forest canopies have not reached a limit in their capacity to take up N from the atmosphere in the wet form and the sum of wet and dry (assuming negligible changes due to DON formation).The agricultural site, receiving the highest deposition fluxes (14-20 kg N ha -1 y -1 ) also presented the highest NCU values (7-12 kg N ha -1 y -1 Table 9, Fig. 4), most of this uptake (60%) being in the reduced N form.
The results of this study indicate that the holm oaks canopies can retain an important part of the incoming N deposition, thus reducing the direct impact of N deposition to soils.The total N inputs to these forests (assuming a range of dry deposition estimates obtained with three different approaches) were between 3 and 20 kg N ha -1 y -1 .Recent studies in these sites indicate that DON would add around 3 kg N ha -1 y -1 in bulk deposition (Izquieta-Rojano et al 2016).
Therefore, the total N input to these holm oak forests can be framed in 20-23 kg N ha -1 y -1 exceeding the critical loads values proposed for sclerophylous forests (15-17 kg N ha -1 y -1 , Bobbink et al. 2010) except at a lower impacted site in central Spain.The long term effects of these continued N inputs and their evolution as N emissions change in recent years has not been yet fully addressed and may deserve attention given its potential impact on soil chemistry, water quality, forest functioning and plant biodiversity.

Conclusions
Atmospheric N deposition to 4 sites in Spain (one affected by an agricultural environment, two by big cities and one as rural background) was determined, distinguishing the wet and dry deposition pathways.To estimate DD, three different methods were applied and compared: a canopy budget model, a branch washing method and the inferential method with Vds obtained from bibliographical references of forest studies.Higher consistency between methods was found for reduced N tan for oxidized N. The branch washing method tended to produce the highest estimates.The site receiving the lowest dry deposition presented the highest relative differences between minimum and maximum estimates, but in absolute terms, differences were similar to the other sites.Taking the average DD from the various methods, DD represented 51-67% (reduced N) and 72-75% (oxidized N) of total reduced and oxidized N deposition.The canopies retained both NH4 + -N and NO3 --N, with the agricultural site and the urban site close to Barcelona retaining more in the reduced than the oxidized form.A very good correlation (r=0.92 and 0.99) between N deposition and canopy uptake indicated that holm oak forests in Spain retain N deposition inputs up to 17.5 kg N ha -1 y -1 .The uptake efficiency (N taken up in the canopy related to N deposition) was higher at the agricultural and rural sites (50-60%) compared to the site close to Barcelona (20-35%), while for Madrid, great differences in DD estimation precluded this analysis.This result points to a decreasing N removal capacity in the canopies of peri-urban forests that may lead to higher N impacts to the soil and soil waters in the future.) TD Reduced N (kg ha -1 y -1 ) Differences in seasonal and total rainfall in the study sites are explained by the climatic characteristics of the Iberian Peninsula and are in accordance with the precipitation pattern of the Mediterranean climate in this region characterized by wet springs and autumns (Rodriguez-Puebla et al. 1998).TC, located at the center of the Iberian Peninsula is under a continental Mediterranean climate, drier and colder than at the coastal Mediterranean region.The northern CA site is affected by the passage of low pressure fronts from the north-northwest that brings

Fig 4 .
Fig 4. Relationship between nitrogen canopy uptake (NCU) and total deposition of reduced N. 85 oxidized N and DIN.Linear regressions are indicated for significant correlations (P<0.01)86

Table 1 .
Study site characteristics, climatic features, forest stand parameters, atmospheric information and air quality at the study sites.Climate and pollutant data are mean values for the study period.

Table 2 .
Exposure period for the branch washing experiment at the study sites.Final date 825 corresponds to the date of branch washing.Precipitation during 1 to 2 days previous to the 826 onset of the experiment is also indicated.

Table 3 .
Compilaton of deposition velocities (Vds in cm sec -1 ) from dry deposition studies in forests, with specification of forest type, method and study period.

Table 4 .
Annual dry deposition fluxes for N gaseous compounds (in kg ha -1 y -1 ), calculated by 2 the inferential method considering the average Vd values in Table3for the different N gases.

Table 5 .
Spatial patterns: Basic statistics of water amount in bulk precipitation (BP).throughfall (TF) and Interception (In= BP-TF) in L m -2 per period.Number of observations = 49 for LC.41 for CB 58 for CA and 50 for TC during the period June 2011 to June 2013.Kruskal -Wallis test indicated significant differences (P<0.001) for all the variables.Differences between site pairs by means of a Mann-Whitney test are indicated with letters.

Table 6 .
Annual wet deposition and throughfall fluxes (in kg ha -1 y -1 ) at the study sites (period June 2011 to June 2013).Percent contribution of oxidized N to DIN is also indicated.WD = wet deposition; TF = throughfall; nTF= net throughfall.

Table 7 .
Estimated dry deposition with the different methods (in kg ha -1 y -1 ).CBM= Canopy budget model, BW= Branch washing, IM = Inferential model at the 4 studied sites.