Isotopic evidence for the occurrence of biological nitrification and nitrogen deposition processing in forest canopies

Site description and sampling

Two Scots pine (Pinus sylvestris L) and two beech (Fagus sylvatica L.) stands were studied. The pine stands were within the UK Forest Monitoring network (http://www.forestry.gov.uk/fr/INFD-67MEVC; Vanguelova et al., 2010), which is part of the ICP European Forest Network. The two beech stands are part of long-term experiments on monitoring of the effects of N_dep on forest and soil biochemical cycling in the United Kingdom (Vanguelova & Pitman, 2009, 2011). Two forests, one for each tree species, were situated at Alice Holt and Rogate (6 km apart) in South East England and the remaining two at Thetford (<8 km apart), east England. They were chosen on the basis of similarity in stand (age, density and management history), climate and soil conditions, but at contrasting levels of ambient N_dep (Table 1). In particular, the pine and beech stands at Thetford are subjected to higher background levels of N_dep (13 and 19 kg N ha⁻¹ yr⁻¹, respectively) compared to forest stands at Alice Holt and Rogate (9–10 kg N ha⁻¹ yr⁻¹) (Table 1). Thetford in East Anglia is known to be among the areas with highest atmospheric N inputs in the United Kingdom (Vanguelova et al., 2010; RoTAP, 2012), mostly in the reduced form, coming mainly from the intensive livestock farms (in particular pigs and chickens). Therefore, the two forest stands in Thetford will be referred to as HN (high nitrogen) and the forests in Rogate and Alice Holt as LN (low nitrogen) sites. Rainfall (RF) and throughfall (TF) sampling and analysis have been carried at the sites over a number of years by means of two bulk RF collectors and ten TF collectors per site. Sampling and analytical procedures followed the level II protocols described in detail in the ICP Forests (2010) manual. In this study, only samples collected biweekly during the 2011 growing season, from June until November, were considered.

Chemical and isotope analyses of water samples

After collection, RF and TF water samples were filtered through a 0.45-μm membrane filter and then analysed for NH₄-N, colorimetrically, and total N by Carbon analyser (Shimadzu 5000, Osaka, Japan; Dionex UK Ltd, Surrey, UK) and for NO₃-N by ion chromatography (Dionex DX-500; Dionex UK Ltd). DON was calculated from the difference between measured total and inorganic nitrogen forms. Chemical analyses were carried out on water samples collected from each of the RF and TF collectors. The RF and TF elemental fluxes were calculated using measured water volumes at the sites and measured elemental concentrations. Dry N_dep values were estimated as the difference between TF and RF for each of the N forms according to European ICP forest monitoring manual, which assumed zero canopy exchange (ICP Forests, 2010) (Tables 1 and 2). To check this assumption, we compared values measured at our sites with the 5 × 5 km grid modelled N_dep data set for the United Kingdom, as used in the RoTAP (2012) review. The estimate included wet and dry NH_x-N (NH₄, NH₃) and NO_y-N (NO₂, NO₃, HNO₃) deposition, modelled with FRAME upon 2005 emissions data (RoTAP, 2012 – chapter 4).

A subsample of the water analysed for ion concentrations was used for stable isotope measurements. Based on measured concentrations, we worked out the volume of water needed to obtain NH₄-N and NO₃-N concentrations >0.5 mg. For this reason, we combined water collected from June until August and then from September until November and we considered (on average between the two time windows considered) 1.5 l for RF and 1 l for TF in the case of forests at HN, while 2 l for RF and 3–4 l for TF in the case of forests at the LN. Pooling was also necessary for RF water samples collected at the two LN and the two HN sites because not enough volume of water was available for each of the two forests at the LN. We assumed that pooling RF water samples within each level of N_dep was not likely to have an impact on the characterization of the isotopic signature of the atmospheric N, due to similar atmospheric N input and source. Indeed, no significant differences were found in the amount of NO₃-N and NH₄-N in RF at either of the two sites, except at Thetford where the NH₄-N was significantly (P < 0.05) higher in the beech relative to the pine stand at the HN site. This was likely the result of the beech site being located only a few 100 m away from a chicken farm that generates NH₃ concentrations as high as ~73 μg m³ (Vanguelova & Pitman, 2009, 2011).

Each RF and TF sample was composited as described above and then passed through cation and anion exchange resins. Ammonium from the cation resin was eluted with hydrochloric acid and converted to ammonium sulphate on a quartz filter paper using an alkaline diffusion method (Heaton, 2001). Nitrate from the anion resin was eluted with hydrobromic acid and processed to silver nitrate (Chang et al., 1999; Heaton et al., 2004). The ¹⁵N/¹⁴N ratios of the ammonium sulphate and the silver nitrate were analysed by combustion in a Flash EA online to a Delta Plus XL mass spectrometer (ThermoFinnigan, Bremen, Germany), with δ¹⁵N values vs. air (atmospheric N₂) calculated by comparison with standards calibrated against IAEA N 1 and N 2 assuming these had values of +0.4 and +20.3‰, respectively. ¹⁸O/¹⁶O ratios of the silver nitrate were analysed by thermal conversion to CO gas at 1400°C in a TC-EA online to a Delta Plus XL mass spectrometer (ThermoFinnigan, Bremen, Germany), with δ¹⁸O values calculated vs. SMOW by comparison with IAEA-NO₃ assuming it had a value of +25.6‰. Analytical precisions (1 SD) were typically <0.3‰ for δ¹⁵N and <0.6‰ for δ¹⁸O. Finally, a subsample of the composite RF and TF water as described above was used for δ¹⁷O measurements by Delta V Plus ratio mass spectrometer. The NO₃⁻ was converted to O₂ and N₂ using the denitrifier method (Casciotti et al., 2002; Kaiser et al., 2007). Analytical precisions (1 SD) for Δ¹⁷O were <1.0‰ based on replicate analysis of the reference material USGS35.

Statistical analyses

Concentrations of NH₄-N and NO₃-N were log-transformed to account for non-normality and variance heterogeneity, as assessed through Shapiro and Levene test, respectively. Independent sample t-tests were employed to test for differences between deposition levels (e.g. HN and LN) and water samples (i.e. RF and TF) for NH₄-N and NO₃-N, while, within each water sample, differences between concentrations of different compounds were tested through paired-samples t-tests (t). The nonparametric Wilcoxon test (W) was employed when log-transformed data did not conform to a normal distribution. Given the small sample size available for the isotopic data (i.e. n = 2 for RF and n = 4 for TF per level of N_dep, as a result of pooling the water samples collected from June until August and then September until November), we calculated the difference in isotopic fractionation between TF and RF without separating beech and pine stands and used a t-test to verify the significance of the difference between LN and HN stands. The level of significance of all statistical tests was set as P ≤ 0.05. R project statistical computing (vers. 3.0.2; R Core Team, 2014) was used for all the analyses.

Mass balance calculations based on Δ¹⁷O and δ¹⁸O

To assess the proportions of atmospheric vs. biologically derived NO₃⁻ collected in the TF underneath tree canopies at the HN site (i.e. Scots pine and beech forests), we considered two independent methods as described in Riha et al. (2014). The methods were only employed at the HN sites as our data suggested little to none canopy processing at the LN sites (cf., 3 section, Fig. 1). The first one is a mass balance approach based on the use of Δ¹⁷O in the following equation:

(1)

where Δ¹⁷O_Tf is the measured isotopic composition of NO₃⁻ in TF, while Δ¹⁷O_Bio and Δ¹⁷O_Atm indicate the isotopic signatures of the biologically and atmospherically derived NO₃⁻, respectively. The f_Bio and f_Atm are the two unknown NO₃⁻ flux fractions from the two different sources, the sum of which is 1. The f_Atm included the fractions of both the wet (f_wet) and the dry (f_dry) NO₃⁻ deposition washed out from the canopy, net of the fraction retained and/or taken up by the canopies (f_U), that is f_Atm = f_Wet + f_Dry − f_U. Assuming that Δ¹⁷O_Bio = 0 (Michalski et al., 2003) and that Δ¹⁷O in RF reflected both wet and dry N_dep, Eqn 1 can be reduced to:

(2)

(3)

The assumption of similar Δ¹⁷O values for wet and dry N_dep stems from the fact that Δ¹⁷O in atmospherically derived nitrate is mostly determined by photochemical oxidation of NO_x by tropospheric ozone (Michalski et al., 2011), not the phase (gaseous, solid, or liquid) into which it is partitioned. Measurements of aerosol nitrate and rain NO₃⁻ collected during the same season do not have significant differences in Δ¹⁷O values (Michalski et al., 2011; Riha, 2013). Hence it is not related to specific point emission sources and it is not affected by mass-dependent isotope fractionations, which, in fact play a significant role in the case of the other two isotope ratios, that is ¹⁸O/¹⁶O and particularly ¹⁵N/¹⁴N.

The second method considered the δ¹⁸O measured in NO₃⁻ based on the following equation:

(4)

where δ¹⁸O_Tf, δ¹⁸O_Atm and δ¹⁸O_Nitr are the oxygen isotopic signatures of the NO₃⁻ in TF, atmospheric deposition (combined RF and dry N_dep) and produced from nitrification, respectively. δ¹⁸O of NO₃⁻ derived from nitrification was calculated by considering that two oxygen atoms in the formed NO₃⁻ were derived from atmospheric water (i.e. RF) and one from atmospheric O₂ as described in the following equation (Mayer et al., 2001):

(5)

Assuming negligible the isotope fractionation during water (ɛ_Rf) and O₂ () incorporation (Mayer et al., 2001), δ¹⁸O of NO₃⁻ from nitrification was obtained from δ¹⁸O of atmospheric O₂ (δ¹⁸O_Atm = 23.9‰, Barkan & Luz, 2005) and the oxygen isotopic signature of the RF. We have assumed this latter to have values of about −5.5‰ (for June–August) and −8.5‰ (for September–November), based on the weighted mean δ¹⁸O values for June–August 2011 and September–November 2011 rainfall at a site near Oxford in the United Kingdom (W.G. Darling, personal communication).

Concentrations of NH₄-N, NO₃-N and DON in RF and TF

The concentration of N compounds varied between LN and HN sites and between RF and TF. At the two LN forests the concentrations of ions in RF were not significantly different (Fig. 1a, b) and the RF and TF had similar NH₄-N and NO₃-N concentrations (Scots pine: t = 1.78, 7.97 and P = 0.11, 0.56, respectively; beech: W = 163, 125 and P = 0.73, 0.48, respectively). In contrast, at the HN forests, the NH₄-N and NO₃-N concentrations were significantly higher in TF compared to RF, for both Scots pine (t = 6.42, 6.26, respectively; all P < 0.001) and beech (W = 265, 250, respectively; all P < 0.001) (Fig. 1a, b). Ion concentrations in both RF and TF were significantly higher at the HN than at LN sites, with the exception of RF in the beech stands, which had similar NO₃-N concentrations. Concentrations of DON in both RF and TF were significantly (RF: W = 32, P < 0.05; TF: W = 34, P < 0.01) higher for the beech stand (Fig. 1c) at the HN compared to LN site. By contrast, Scots pine (Fig. 1c) subjected to different atmospheric N loads from the atmosphere showed similar values of DON concentrations in both RF and TF. However, DON concentrations in RF did not show a significant difference when comparing beech and Scots pine stand at the LN site, while concentrations were slightly higher (W = 35, P = 0.05) at the beech compared to the Scots pine stand at the HN site. Concentrations of DON in TF were similar at the two LN forests, while they were higher (W = 33; P < 0.05) at the beech than at the Scots pine stand at the HN site (Fig. 1c).

Extrapolation of our seasonal measurements over time and model validation of estimated dry N_dep fluxes

The mean of total N fluxes during the 6 months we considered in this study (i.e. June–November 2011) is reported in Table 2. TF-N fluxes were higher than RF fluxes at the two forests at the HN, with particular reference to the NH₄-N at the beech site. By contrast, at the LN site, RF N fluxes were higher than TF-N fluxes for both species. An independent estimate of the dry N_dep at our sites can be obtained using the modelling approach outlined in RoTAP (2012). Figure S1 (in Supporting Information) shows a comparison of the measured wet N and estimated dry N fluxes (i.e. as a difference between TF and RF fluxes) at the two level of N_dep and shown in Table 1, with the fluxes of wet and dry N_dep obtained from the 5 × 5 km grid UK map, based on modelled N_dep with FRAME upon 2005 N emission data (RoTAP, 2012). Interestingly, a reasonably good agreement was found between the on-site measurements and the modelled values of wet N_dep. However, the fluxes of dry N_dep predicted using the RoTAP modelling approach were much higher than those estimated as the difference between TF and RF fluxes at our sites (Fig. S1).

Values of δ ¹⁵N-NH₄

Values of δ¹⁵N-NH₄⁺ in RF (Fig. 2a) ranged from positive at the HN site (+1.49 ± 3.5‰) to very negative at the LN site (−9.14 ± 0.2‰). Due to the limited number of RF measurements (i.e. n = 2 per level of N_dep), statistical analyses of isotope data were performed per level of N_dep, combining data for both tree species and focussing on the differences between RF and TF. However, TF values measured separately for beech and Scots pine are presented in Fig. 2a, to show the species-specific changes in the isotope compositions in N compounds collected below the canopies. More positive values were measured for δ¹⁵N-NH₄⁺ in TF compared to RF at both HN (t = −2.85, P < 0.05) and LN (t = −15.16, P < 0.001) sites. The TF–RF difference for δ¹⁵N in NH₄⁺ was much higher (t = −2.65, P < 0.05) at the LN compared to the HN site (Fig. 2b).

Values of δ ¹⁵N, δ ¹⁸O and Δ ¹⁷O-NO₃

The δ¹⁵N in NO₃⁻ of RF (Fig. 3a) showed similar negative values at the HN (−3.4 ± 1.4‰) and LN sites (−2.8 ± 1.7‰). Albeit lower, the δ¹⁵N-NO₃⁻ values in TF at the HN site (diff = −4.9‰ ± 3.4) were only slightly different (t = −1.72, P = 0.06) compared to the LN sites (diff = +1.1 ± 0.54‰) (Fig. 3d). Despite differences between RF and TF for δ¹⁵N in NO₃⁻ not being significant within each level of N_dep, δ¹⁵N in NO₃⁻ showed more negative values in TF than in RF at the HN site at the Scots pine stand (Fig. 3a).

The δ¹⁸O in NO₃⁻ of RF showed similar values at the two different levels of N_dep, that is LN = 63.9 ± 0.88‰ and HN = 64.1 ± 3.2‰ (Fig. 3b). Within each level of N_dep, δ¹⁸O values did not significantly differ between RF and TF. However, similarly to δ¹⁵N, we observed lower δ¹⁸O in TF compared to RF in the case of the Scots pine at the HN. A significant contrast (t = −2.34, P < 0.05) was found in the difference between the δ¹⁸O values of NO₃⁻ in TF compared with RF across levels of N_dep (Fig. 3e), with lower δ¹⁸O-NO₃⁻ values at HN than at LN site.

Δ¹⁷O values measured in RF at our sites ranged from 23.14 (±0.58)‰ at the LN site to 25.53 (±0.76)‰ at the HN site. A significant difference was found in the Δ¹⁷O of NO₃⁻ in the TF vs. RF at the HN sites (W = 16, P < 0.05), but not at the LN sites. Within individual species, it is worth pointing out that beech showed lower Δ¹⁷O values than Scots pine (Fig. 3c). When we considered the difference between RF and TF, Δ¹⁷O values in NO₃⁻ had lower values on average at the HN sites (t = −1.86, P = 0.05) than at LN sites (Fig. 3f), but the difference was not significant.

Combined plots for the three isotopic species of NO₃⁻ at the Scots pine and beech sites are given in Fig. 4 as trajectories of change from RF to TF values, to emphasize the consequences of canopy processing for the three tracers, with particular references to forests at HN levels. For Scots pine (Fig. 4a, b), only in the case of HN sites did δ¹⁵N, δ¹⁸O and to a less extent Δ¹⁷O values in TF diverge from those measured in RF. For beech (Fig. 4c, d), distinct changes in δ¹⁸O vs. δ¹⁵N were not observed and, only in the case of HN site, did Δ¹⁷O become substantially lower from RF to TF.

Assessing the source of NO₃⁻ in the TF at the sites with high atmospheric N loads

Two mixing models, partitioning fluxes based on either Δ¹⁷O or δ¹⁸O, were used to estimate the relative contributions of atmospheric vs. nitrification-derived NO₃⁻ collected underneath tree canopies. Using the two-end-member mixing model with the Δ¹⁷O [Eqns 2 and 3 in the 2] values measured in TF and RF (Table S1), the fractions of NO₃⁻ in TF coming from nitrification (f_bio) ranged from 0.17 for the Scots pine up to 0.59 for the beech (i.e. 17–59%) at the two HN sites (Fig. 5a). Most of the NO₃⁻ collected in the TF at the Scots pine stand derived from the atmosphere (mean of f_Atm = 0.83 ± 0.002), with only a minor contribution from nitrification (mean of f_Bio = 0.17 ± 0.002). By contrast, biologically derived NO₃⁻ seemed to be the dominant fraction of the NO₃⁻ in TF of the beech stand (f_Bio = 0.59 ± 0.03), at least for the time period considered in this study (Fig. 5a).

When using the mixing model based on δ¹⁸O partitioning [Eqns 4 and 5 in the 2], a higher fraction of NO₃⁻ in TF was estimated to derive from the atmosphere (Scots pine: f_Atm = 0.62 ± 0.07; beech: f_Atm = 0.90 ± 0.09) than from nitrification (Scots pine: f_Bio = 0.38 ± 0.07; beech: f_Bio = 0.10 ± 0.09) (Fig. 5b). The two approaches were more consistent for the Scots pine, while they did lead to opposite results in the case of beech. Averaging across the two methods, the proportion of the biologically derived nitrification was 27% for Scots pine (range of 17–38%) and 34% for beech (range of 10–59%).

Atmospheric N and its isotopic signatures at the contrasting N_dep levels

Both beech and Scots pine forests at HN sites were subjected to air masses with high NH₃-N/NH₄-N concentrations and had higher NH₄-N deposition relative to the LN sites. The HN beech site, which is right next to an intensive chicken farm, is trapping the farm's NH₄/NH₃ emissions along a very distinct 200-m-long N gradient where concentrations decrease to levels similar to those in the nearby Scots pine stand (Vanguelova & Pitman, 2009). This is showed by the higher NH₄-N concentrations in RF at the beech than at the Scots pine stand at the HN, while no difference was found for NO₃-N concentrations (Fig. 1). Fluxes relative to the 2011 growing season indicated that at the beech stand, NH₄-N is the dominant component of wet N_dep, while NH₄-N and NO₃-N contributed almost similarly to wet deposition at the Scots pine (Table 2). These results are in line with the data from long-term monitoring within the ICP forest network (Table 1), which showed that Thetford is among the sites receiving the highest N_dep in the United Kingdom (Vanguelova et al., 2010; RoTAP, 2012), mostly in the reduced form, coming mainly from the intensive livestock farms (in particular pigs and chickens). Records over more than 10 years also suggest that the overall total N_dep at the Thetford pine site has decreased over time, because of reductions in wet (in both forms NH₄-N and NO₃-N) rather than dry deposition (Vanguelova et al., 2010), confirming the national trend (RoTAP, 2012).

The relative contributions of dry vs. wet N_dep at the site-level were broadly in agreement with modelled deposition rates obtained at the 5 × 5 km scale (RoTAP, 2012, cf., Fig. S1). For example, the modelled data suggested similar values for the total (wet plus dry) oxidized N forms (NO₃-N, NO₂-N and HNO₃-N) at the HN vs. LN sites, which is consistent with a similar impart of traffic-derived emissions at these sites. The large difference in dry NH_x deposition between LN and HN is also consistent with the effects of the numerous pig and chicken farms at HN. In addition, the rates of modelled NO_x and NH_x wet deposition were similar to the long-term measurements (Table 1) in RF at both the HN and LN sites. However, the modelled values of dry NO_x and NH_x deposition at both the HN and LN sites were substantially higher compared to the estimated dry deposition as difference between RF and TF. While this suggests significant canopy uptake at the HN sites, it must be remembered that the model estimates also include NH₃-N together with NH₄-N deposition, and NO₂-N-HNO₃-N together with NO₃-N, which were not directly account for in either the data previously published and reported in Table 1 or the current study. In addition, it is possible that the 5 × 5 km model of RoTAP (2012) fails to capture the small-scale variability in N_dep, especially in dry deposited NH₃.

Isotopic signatures measured in NO₃⁻ and NH₄⁺ in RF (Figs 2 and 3a, b) at our sites were in the same range found in previous analyses of monthly rainfall samples from several sites in the United Kingdom (Heaton et al., 1997; Curtis et al., 2012; T.H.E. Heaton, unpublished data; Table 3). Overall, δ¹⁵N values in NH₄⁺ measured across the United Kingdom ranged from negative to slightly positive values (−12.6‰ to +2.8‰), with a mean of −4.3‰. The positive values observed at the Thetford sites are likely reflecting the contribution of NH₄/NH₃ emissions coming from the intensive chicken farms. Indeed, Heaton et al. (1997) reported that the δ¹⁵N value of TF ammonium in part of a Scots pine plantation artificially fumigated with ammonia gas was 17‰ higher than the value for TF in the nonfumigated part of the plantation. Moreover, in a recent study, Yeatman et al. (2001) measured δ¹⁵N values of +13.5‰ in aerosol NH₄⁺ sampled near chicken, cow and pig livestock enterprises, and positive δ¹⁵N values in bulk precipitation were also reported by Emmett et al. (1998) for two conifer stands near livestock feed lots in the Netherland.

The δ¹⁵N values of NO₃⁻ were similar to those reported in the study by Heaton et al. (1997). However, a high range of values was measured across the United Kingdom (−8.2‰ to +4.3‰) (Table 3), with a mean δ¹⁵N-NO₃ values of −2‰. A similar range of δ¹⁵N values in NO₃⁻ from −11 ‰ to +3.5 ‰ was reported in studies across the USA (Kendall, 1998; Elliott et al., 2007; Kendall et al., 2007), while Tobari et al. (2010) measured δ¹⁵N values in bulk precipitation ranging from −7 to +15.4‰ across different watersheds in Japan. Moreover, a number of studies in the literature used δ¹⁵N to assess the anthropogenic NO_x source. For instance, very negative (−13‰ to −2‰) δ¹⁵N-NO_x values were reported in the case of emissions coming from traffic, while positive values (between 4 and 16‰) were measured for emissions from coal-fired power plants (Heaton, 1990). Similar values of δ¹⁵N-NO₃ in RF at HN and LN sites in our study suggest a similar anthropogenic NO_x source, most likely emissions coming from local road traffic, consistent also with the absolute concentrations measured in RF at both HN and LN. This is confirmed also by the similar values we measured for δ¹⁸O-NO₃ in RF, irrespective of site. Moreover, Δ¹⁷O in RF at the HN was 2‰ higher than at the LN sites, possibly suggesting that NO_x went through slightly different oxidation processes (Michalski et al., 2003). Δ¹⁷O values measured at our sites (ranging from 22 to 26‰) were similar to those reported by Costa et al. (2011) for NO₃⁻ in rain samples (23.1 ± 1.8‰) collected in Michigan and by Michalski et al. (2004) in aerosol (26 ± 3‰) sampled in Southern California.

Processes affecting throughfall N at contrasting N_dep levels: canopy retention, dry N_dep and biological transformation

Our data showed that at the LN TF-N fluxes were lower than RF N fluxes (Table 2), suggesting that most of the atmospheric N was retained by tree canopies, as observed also in other studies (Lindberg et al., 1986; De Schrijver et al., 2004; Staelens et al., 2007; Fenn et al., 2013; Ferretti et al., 2014; Houle et al., 2015). Epiphytic lichens, fungi and micro-organisms on the canopy may contribute to the higher N retention and subsequent processing at the LN sites, a possibility supported by the significant increase in DON concentrations in TF (Fig. 1c; Table 2), as also reported in other studies (Woods et al., 2012).

By contrast, at the HN sites, NH₄-N and NO₃-N concentrations and fluxes were higher in TF than in RF, irrespective of tree species (Fig. 1 and Table 2). We also found higher NH₄-N in TF underneath beech than Scots pine, with the former receiving higher NH_x-N atmospheric inputs than the latter, while both NO₃-N and NH₄-N TF fluxes increased underneath the Scots pine. These last results are in line with previous studies in the literature (Lovett & Lindberg, 1993; Fenn et al., 2000; Vanguelova et al., 2010; De Vries et al., 2014), and they suggest that in areas with high dry N_dep, canopy filtering and rain washing will contribute to increasing the N inputs to TF and hence to the soils, compared to areas subjected to low atmospheric N loads, in particular dry N_dep. Nevertheless, the different proportion of the N compounds in TF underneath the two forests could also be related to species-specific canopy N retention, which, however, is difficult to quantify by looking only at the difference between TF and RF.

The more positive values for δ¹⁵N in NH₄⁺ collected in TF are consistent with the dry NH_dep washed-off the canopies and contributing to increasing NH₄-N in TF at the Thetford site (Figs 1 and 2). Indeed, the δ¹⁵N values of NH₄⁺ in dry deposition tend to be higher than those measured in bulk precipitation (Heaton et al., 1997), suggesting that a fraction of the measured TF originated from dry N_dep. Interestingly, while the NH₄-N concentration did not vary significantly from RF to TF and the N fluxes were lower in TF vs. RF at the LN forests, a fingerprint of dry N_dep was still detected by the ¹⁵N enrichment in NH₄⁺ underneath the canopies.

The higher NO₃-N in TF at the HN sites for both Scots pine and beech could in principle result from a combination of dry deposition and canopy nitrification processes. As in the case of NH₄⁺, higher values of δ¹⁵N of NO₃⁻ in TF compared to RF could be expected (Heaton et al., 1997), but were not found at these sites (Fig. 3a). Nitrification of NH₄⁺ leads to the production of ¹⁵N-depleted NO₃⁻ leaving behind more ¹⁵N-enriched NH₄⁺ (Högberg, 1997). Indeed, we measured more negative (but not significantly so) δ¹⁵N-NO₃⁻ in the TF–RF differentials at the HN compared to the LN site. The ¹⁵N depletion of NO₃⁻ in TF was particularly detected for Scots pine, but not for beech, at the HN site (Figs 3a and 4a). A decrease in δ¹⁵N in NO₃⁻ from RF to TF was reported in studies in a spruce forest in Germany by Sah & Brumme (2003) and in a montane rainforest in Ecuador by Schwarz et al. (2011), explained in both cases by isotope fractionation during nitrification of NH₄⁺ to NO₃⁻ in the canopy leaves. However, none of these previous studies could unequivocally attribute the shifts in ¹⁵N-NO₃ to biological NH₄⁺ nitrification. In this study, evidence of nitrification occurring within the canopy was clearly provided using two independent methods, based on Δ¹⁷O and δ¹⁸O. Our results showed that although atmospheric NO₃⁻ was the dominant source of NO₃⁻ in TF at the Scots pine stand, a considerable proportion (varying between 17 and 38%, depending on which isotope was employed for the mass balance) derived from biological nitrification. The two approaches broadly agreed, but Δ¹⁷O lead to higher f_Atm estimates than those obtained by δ¹⁸O (Fig. 5). Significantly, both methods detected the contribution of biologically derived NO₃⁻.

Interestingly, similar values of δ¹⁵N-NO₃ in TF and RF did not provide a clear signal of canopy transformation for the Beech at the HN (Fig. 4c). In contrast, the mass balance approach using Δ¹⁷O and δ¹⁸O proved that biological activity contributed to higher NO₃⁻ underneath beech canopies, with quite different estimate of f_Bio although. Indeed, based on Δ¹⁷O, biologically derived NO₃ was as much as from atmospherically derived NO₃ (Fig. 5a), whereas the mixing model based on δ¹⁸O estimated that 90% of the NO₃⁻ in TF derived from the atmosphere and only a small fraction from nitrification (Fig. 5b).

Moreover, the significant increase in DON concentrations in TF at both Scots pine and beech sites provide evidence of transformation of dissolved inorganic N to DON within tree canopies (Gaige et al., 2007). Higher DON concentrations in TF can be related to leaching from leaves and needles and/or release by bacterial epiphytes in the phyllosphere (Müller et al., 2004).

The fact that only in the case of the Scots pine, we found consistency between δ¹⁵N-NO₃⁻ and the mixing model approaches based on δ¹⁸O and Δ¹⁷O could be partially related to differences between species in the canopy structure and phenology. Conifers are more efficient in scavenging aerosol and atmospheric deposition than broadleaf species (Augusto et al., 2002; De Schrijver et al., 2007) due to the greater canopy surface area and roughness. Furthermore, conifer evergreen phenology implies a higher canopy retention capacity than in deciduous species (De Schrijver et al., 2000), as in the case of Scots pine, whose needles can remain in the canopy for 2–3 years. This means that atmospheric N deposited onto tree canopies and cumulated over multiple growing years and not taken up by needles could undergo several biological transformations, which imply isotope fractionations leading to a distinct isotopic signature between the atmospheric N source and the final produced N specimen. However, assessing the differences between two forests at HN for canopy N transformation goes beyond the aim of this study, due to low replicates per species. Nevertheless, our results certainly shade light on species-specific dynamic of biological activity in tree canopies, which deserves further investigation.

Effectiveness of the two mass balance approaches based on δ¹⁸O and Δ¹⁷O

The use of Δ¹⁷O in nitrate was successfully applied to assess the contribution of atmospheric vs. microbiologically derived NO₃⁻ in a forest catchment (Costa et al., 2011) and lately, in combination with δ¹⁸O, in an urban environment (Riha et al., 2014). Both studies looked at the isotopic composition in N specimens in the run-off water vs. RF. Processes occurring in the soil, with particular reference to nitrification, and isotope fractionations associated with them, are very well described (see among the others, Högberg, 1997). While canopy N retention and transformation are widely acknowledged as important pathways for trees to acquire N (Sparks, 2009; Pennisi, 2015), the underlying mechanisms are still not understood. This is particularly true for the isotope fractionations, which may occur during nitrification in the canopies or N uptake. One potential limitation of the mixing model based on δ¹⁸O is in the estimation of the δ¹⁸O_Nitr (see 2). First, precipitation intercepted by tree canopies might be subjected to evaporation, which, in turn affects the δ¹⁸O of the precipitation-derived water available for nitrification (e.g. water remaining on the canopy might be more ¹⁸O-enriched than precipitation itself). Second, the assumptions underlying the use of Eqn 5 may not be always valid. In some environments, oxygen isotope exchange between a nitrification intermediary, nitrite and water may invalidate the two-third and one-third proportions of Eqn 5. In addition, the possible influence of an equilibrium isotope fractionation during NO₂⁻ and H₂O exchange at the enzyme (+14‰) and an inverse kinetic isotope effect during NO₂⁻oxidation into NO₃⁻ have also been proposed and would lead to higher δ¹⁸O values than those predicted by the simple isotope mass balance model (Casciotti et al., 2010, 2011; Snider et al., 2010; Buchwald et al., 2012). Third, N deposited onto canopies could be more reactive and subject to further transformations before being processed within the canopies or washed-off (e.g. NH₃ volatilization, NO reaction with ozone or denitrification). Thus, improper calculation of δ¹⁸O_Nitr might affect the estimation of f_Atm and f_Bio. While mass-dependent isotope fractionations related to NO₃⁻ transformation can significantly affect the δ¹⁸O, they have no effect on Δ¹⁷O. For this reason, using Δ¹⁷O seems a more robust approach, leading to a better estimate of f_Atm vs. f_Bio (Michalski et al., 2003). However, more studies are needed to assess the Δ¹⁷O of wet vs. dry N_dep and how they change over time.

Synthesis

Our results partially confirmed the initial hypotheses (1) that at the LN sites, ion concentrations in TF and their respective isotopic signatures reflected the input of atmospheric N as derived from RF. However, isotope data revealed that even with a low atmospheric N load, canopies played an important role in intercepting and retaining dry N_dep (with particular reference to the reduced N form), which represents an additional (but often overlooked) N source relative to wet N_dep as assessed through RF. Differences in the RF and TF fluxes together with an increase in TF DON concentrations provided evidence of canopy N retention and possible uptake. At the HN sites, the passing of atmospheric N through canopies affected both ion concentrations and their isotopic signature (which confirmed our hypothesis 2). The occurrence of dry deposition explained the higher NH₄-N concentrations and ¹⁵N enrichment in NH₄⁺ measured below the canopy in TF water vs. RF. As for the higher NO₃-N in TF vs. RF, the isotopes δ¹⁵N and δ¹⁸O could not provide clear indications of its origin, even though for Scots pine δ¹⁵N-NO₃⁻ provided some indications of biologically derived NO₃⁻. The unambiguous response came, however, from Δ¹⁷O, which allowed to detect that a consistent fraction of the NO₃⁻ recovered underneath the canopies derived from biological nitrification, with an especially large magnitude at the beech stand (where the other isotopes, particularly δ¹⁸O, failed to provide conclusive evidence).

We acknowledge that the conclusions of this study rely on a limited number of isotope measurements at each site and a limited selection of forest stands, which did not allow detailed investigations of the tree species-specific pattern of canopy N transformations. However, by combining multiple isotopes, the study identified canopy processing of atmospheric deposition (and especially canopy biological nitrification) as a major process that should not be neglected and needs further exploration. This has important implications for policy-related emission abatement strategies, which aim to manage forests and landscape not only for enhancing C-sequestration, but also for atmospheric N capture.