Substantial contribution of tree canopy nitrifiers to nitrogen fluxes in European forests

Human activities have greatly increased the reactive nitrogen in the biosphere, thus profoundly altering global nitrogen cycling. The large increase in nitrogen deposition over the past few decades has led to eutrophication in natural ecosystems, with negative effects on forest health and biodiversity. Recent studies, however, have r ep or ted o li go tr op hi cation in forest ecosystems, constraining their capacity as carbon sinks. Here we demonstrate the widespread biological transformation of atmospheric reactive nitrogen in the canopies of European forests by combining nitrogen deposition quantification with measurements of the stable isotopes in nitrate and molecular analyses across ten forests through August–October 2016. We estimate that up to 80% of the nitrate reaching the soil via throughfall was derived from canopy nitrification, equivalent to a flux of up to 5.76 kg N ha −1 yr −1 . We also document the presence of autotrophic nitrifiers on foliar surfaces throughout European forests. Canopy nitrification thus consumes deposited ammonium and increases nitrate inputs to the soil. The results of this study highlight widespread canopy nitrification in European forests and its important contribution to forest nitrogen cycling

Human activities have greatly increased the reactive nitrogen in the biosphere, thus profoundly altering global nitrogen cycling.The large increase in nitrogen deposition over the past few decades has led to eutrophication in natural ecosystems, with negative effects on forest health and biodiversity.Recent studies, however, have r ep or ted o li go tr op hi cation in forest ecosystems, constraining their capacity as carbon sinks.Here we demonstrate the widespread biological transformation of atmospheric reactive nitrogen in the canopies of European forests by combining nitrogen deposition quantification with measurements of the stable isotopes in nitrate and molecular analyses across ten forests through August-October 2016.We estimate that up to 80% of the nitrate reaching the soil via throughfall was derived from canopy nitrification, equivalent to a flux of up to 5.76 kg N ha −1 yr −1 .We also document the presence of autotrophic nitrifiers on foliar surfaces throughout European forests.Canopy nitrification thus consumes deposited ammonium and increases nitrate inputs to the soil.The results of this study highlight widespread canopy nitrification in European forests and its important contribution to forest nitrogen cycling.
Nitrogen is one of the main elements on Earth.It links the atmosphere with terrestrial and aquatic ecosystems via biogeochemical processes.Reactive nitrogen in the atmosphere has doubled over the last century as a result of the combustion of fossil fuels and the intensive use of fertilizers in food production to sustain global population growth 1 , extending planetary boundaries for nitrogen fluxes beyond a safe limit 2 .Elevated rates of nitrogen deposition are one of the drivers of the increased strength of forest carbon sinks 3,4 , particularly in nitrogen-limited ecosystems in temperate and boreal biomes, thus contributing to the mitigation of climate change.The global increase in nitrogen deposition, however, generates cascading effects and contributes to the acceleration of nitrogen cycling 5 , which has important implications for both natural ecosystems functioning and human health [6][7][8] .Tracing and quantifying the fate of atmospheric nitrogen when moving from the atmosphere to the soil through forest canopies is paramount to quantifying its effects on terrestrial carbon and nitrogen Article https://doi.org/10.1038/s41561-023-01364-3canopies and aqueous solutions) 27,28 .Investigations of canopy nitrogen retention or release have mostly focused on the passive role of foliar surfaces to retain nitrogen compounds from the atmosphere (particularly the dry forms), which affects the input of these compounds to the soil.However, next-generation sequencing analyses have begun to unravel the highly diverse microbial communities living in forest canopies 29,30 and to identify, in situ, the taxa contributing to processes that were previously considered to occur only in soil 31,32 .Previous studies have found that nitrate in TF water represents not only a flux from atmospheric deposition, but also from biological nitrification in tree canopies 33−34 .Whether microbes transform nitrogen in forest canopies across a wide range of forest types, climatic conditions and levels of atmospheric deposition, and how they contribute to NO 3 − fluxes reaching the soil via TF, have not yet been determined.In this Article we examine the importance of microbially driven canopy nitrification in the context of the ongoing debate surrounding the trajectories of apparent ecosystem nitrogen saturation and tree oligotrophication.We combine isotopic and genetic analyses to estimate canopy nitrification across European forests, to identify microorganisms potentially involved in this process, and to quantify the contribution of gross canopy nitrification (GCN) to nitrogen fluxes.Ten sites included in the European International Co-operative Programme on Assessment and Monitoring of Air Pollution Effects on Forests (ICP Forests) were chosen that include the two most dominant tree species in Europe (Fagus sylvatica L. and Pinus sylvestris L., hereafter F. sylvatica and P. sylvestris, respectively).Forests were selected along wide gradients of climate and wet deposition of inorganic nitrogen (Fig. 1 and Extended Data Table 1), with levels of nitrogen deposition below and above the empirical CLs of 10-15 kg ha −1 yr −1 (ref.35).We consider fluxes of BD and TF nitrogen deposition measured in 2016 across the investigated sites to determine whether tree canopies act as apparent sinks (retention) or apparent sources (release) of atmospheric reactive nitrogen compounds, focusing on NH 4 + -N, NO 3 − -N and total nitrogen (TN, including dissolved organic nitrogen).We use isotopic tracers, particularly Δ 17 O (defined as Δ 17 O = δ 17 O − 0.52 × δ 18 O) in NO 3 − , in a two-end-member isotope model, to estimate the atmospheric (f Atm ) and biological (f Bio ) fractions contributing to NO 3 − recovered in TF water.Δ 17 O is a robust tool for quantifying the contributions of the two sources of NO 3 − directly reaching the TF (atmospheric deposition and canopy nitrification), due to the distinctive Δ 17 O values of cycling.This is particularly important for nitrogen in the form of nitrate (NO 3 − ), a highly mobile compound that can be readily lost by leaching and denitrification when its availability exceeds plant and microbial demands 9 .Nitrate concentration in soil solutions is a key indicator of nitrogen saturation in forest ecosystems 9 , setting the threshold for the critical nitrogen loads (CLs) beyond which ecosystem processes and plant community compositions can shift [10][11][12] .Indeed, the seminal hypothesis of nitrogen saturation 9 suggests that nitrogen deposition accumulates in soil nitrogen pools and progressively increases mineralization and nitrification, eventually leading to an excess of nitrogen availability compared to the demand from plants, as well as to elevated NO 3 − leaching and the emission of trace gases (particularly N 2 O).Negative effects of nitrogen saturation on soil processes and tree health have been reported in soil manipulation studies 13,14 , often following the application of nitrogen doses much higher than ambient nitrogen deposition, thus exposing ecosystems to larger changes in atmospheric nitrogen input than those projected in all scenarios of the IPCC Representative Concentration Pathway 15 .Field studies across Europe and North America have confirmed an increase in NO 3 − leaching at sites exceeding nitrogen depositions of 10 kg ha −1 yr −1 , but there is no clear evidence that this pathway of nitrogen loss (and hence saturation) increases over time under increasing nitrogen deposition 16 .In contrast, recent analyses have identified pervasive nitrogen limitations in forests ('too little of a good thing' or oligotrophication) 17,18 , which can constrain the expected CO 2 fertilization effect [19][20][21] .
Atmospheric nitrogen undergoes complex transformations in its trajectory from the atmosphere to the soil, including interactions with forest canopies and the associated microbial communities 22,23 .Evidence since the 1980s has shown that fluxes in precipitation are substantially affected when passing through forest canopies 23 .These studies have led to the quantification of fluxes underneath tree canopies (so-called throughfall, TF) to better estimate the total input of nitrogen from atmospheric deposition (including dry reactive nitrogen compounds) and the proportion eventually reaching the soil.The differences between TF and open-field deposition (either wet deposition or, as in our study, bulk deposition, BD) of nitrogen in the form of ammonium (NH 4 + -N) and NO 3 − -N have been extensively used to determine whether tree canopies act as sinks (TF − BD < 0, canopy retention) or sources (TF − BD > 0, canopy release) of nitrogen compounds (including deposited dry compounds [24][25][26] , as well as ionic exchanges between  ) 36 (Methods).Finally, we document the presence in the canopy of microorganisms and key bacterial and archaeal genes that drive nitrification, using 16S rRNA gene metabarcoding and gene-specific quantitative polymerase chain reaction (qPCR) for genes that encode for ammonia monooxygenase subunit A (amoA) and nitrite oxidoreductase subunit B (nxrB), catalysing the oxidation of ammonia to nitrite and of nitrite to nitrate, respectively (Methods and Supplementary Information).

Biological transformations in forest canopies
At the ten sites, deposition of TN ranged from 3.4 to 15.8 kg ha −1 yr −1 for BD and from 1.5 to 25.4 kg ha −1 yr −1 for TF (Extended Data Table 1).Nitrogen fluxes were higher in TF than BD at most sites (Fig. 1), although the difference was significantly greater than zero only for NO 3 − -N and TN (Extended Data Fig. 1 and Extended Data Table 2).The canopies consistently retained all forms of nitrogen at the forest site in Sweden with the lowest nitrogen deposition, particularly when looking at the long-term changes in nitrogen fluxes (Extended Data Table 2).
Canopy release (TF > BD) is normally attributed to dry reactive nitrogen compounds previously deposited onto tree canopies being washed out during precipitation [24][25][26] .However, the isotopic tracers (δ 15 N, δ 18 O and Δ 17 O) indicated that NO 3 − in TF was a mixture of atmospheric and biological sources (Extended Data Fig. 2 and Extended Data Tables 3 and 4).We used a mass-balance approach based on Δ 17 O (Methods) to determine that 68 ± 25% (s.d.) of the total NO 3 − in TF was from atmospheric deposition (f Atm ), and 32 ± 25% of the NO 3 − reaching the soil via TF was from canopy nitrification (f Bio ) (Fig. 2a), ranging across the gradient from 4% (f Bio = 0.04) to 80% (f Bio = 0.8).f Bio was 0.37 ± 0.30 and 0.25 ± 0.18 for F. sylvatica and P. sylvestris, respectively (Fig. 2b), within the ranges reported in previous studies conducted on the same species in the UK 33 and in a Quercus ilex forest in Spain 34 .Increases in f Bio were associated with increases in TN (Extended Data Fig. 3 and Extended Data Table 5), suggesting that this process could be substrate-limited at lower inputs of atmospheric nitrogen.
Site-specific f Bio was then used to partition TF NO 3 − -N fluxes into two sources: atmospheric deposition and biological transformation (hereon described as gross canopy nitrification, GCN).GCN contributed to adding between 0.40 and 4.97 kg N ha −1 yr −1 for F. sylvatica and between 0.21 and 3.23 kg N ha −1 yr −1 for P. sylvestris to the NO 3 − -N derived from atmospheric deposition (Fig. 3a; Extended Data Table provides the full range of values), with a mean (±1 s.d., but adjusted for cross-site differences in TN deposition) of 1.77 ± 1.38 kg N ha −1 yr −1 .By comparison, the real (after correcting for GCN) contribution of atmospheric deposition to the TF NO 3 − -N fluxes averaged at an equivalent of 3.26 ± 1.13 kg N ha −1 yr −1 .Correction of the TF fluxes for the occurrence of GCN changed the balance of several forest canopies from being sources to apparent sinks of atmospheric NO 3 − -N (Fig. 3b), because the apparent release of NO 3 − -N from the canopy was often due to nitrification.

Microbial nitrifiers in the phyllosphere
The identification and quantification of microbial nitrifiers on foliar surfaces were supported by two independent DNA-based approaches: 16S rRNA gene metabarcoding and qPCR for functional genes.The analysis of 16S rRNA sequences confirmed the presence on foliar surfaces of previously known bacterial and archaeal autotrophic ammonia oxidizers (Nitrosospira and Nitrospira, and Nitrososphaeraceae, respectively) and bacterial autotrophic nitrite oxidizers (Nitrospira, Nitrobacter and Burkholderia insulsa) 37 .The functional genes amoA and nxrB were detected in the phyllosphere (Fig. 4 and Extended Data Fig. 4).Archaeal amoA was more abundant (P = 0.05) than bacterial amoA, as previously observed in Q. ilex in the Mediterranean region 34 , whereas only archaeal nitrifiers were present in the phyllosphere of Criptomeria japonica 38 .The median difference between TF and BD for the functional genes involved in nitrification was significantly less than zero only for archaeal nitrifiers (V from a Wilcoxon test = 20, P < 0.01, Supplementary Fig. 1), suggesting that they could be more efficiently retained by tree canopies.
Microbial nitrifiers were also detected at sites where atmospheric nitrogen was retained in the canopy (Extended Data Table and Extended Data Fig. 4), including three additional sites in Finland (Extended Data Table 1) that were added to the survey later, but for which estimates of f Bio were not available (Methods).This finding does not necessarily indicate that canopy nitrification did not occur at these sites, but that NO 3 − -N is directly taken up by canopies 39 or that canopy nitrification is too low to be detected with our approach or, lastly, that the produced nitrate is consumed through denitrification.Indeed, microbial genes that encode enzymes involved in denitrification as collectors (see Methods for more details).The Swedish site was not included in the isotope analyses, because canopy retention was observed but no canopy nitrification was detected (Extended Data Fig. 1 and Extended Data Tables 2 and  4).The dashed vertical grey line in b separates sites with low (left) and high (right) nitrogen deposition, identified as explained in the main text and Methods.

Article
https://doi.org/10.1038/s41561-023-01364-3well as N 2 fixation were quantified on foliar surfaces (Supplementary Fig. 2), suggesting that nitrification may not be the only process occurring in the tree canopies of the investigated forests, as already observed in previous studies 31,32,40 .Interestingly, N 2 O emissions from tree canopies on a boreal forest were mostly associated with transport of N 2 O (derived from denitrification in the soil) through the transpiration stream 40 .Molecular data from this study suggest that the process could potentially occur in the canopies too, although hypoxic-to-anoxic conditions are mostly required to promote denitrification, and these conditions may not be prevalent in the phyllosphere.

Fagus sylvatica Pinus sylvestris
Atmosphere GCN -N fluxes (both in kg N ha −1 yr −1 ), obtained using a mass-balance approach, is summarized in the cartoon on the right side of the panel.The range of values (minimum and maximum, all values are reported in Extended Data Tables 2 and  4) for the BD and TF fluxes and the Δ 17 O in nitrate in the different components (BD, TF and nitrification) are also reported in the side boxes.For details on the mass-balance approach based on Δ 17 O in nitrate to partition between GCN and atmospheric nitrate, see the Methods.The quantification of the main functional genes that encode for ammonia monooxygenase subunit A and nitrite oxidoreductase subunit B of bacteria (amoA AOB and nxrB) and archaea (amoA

Ecological importance of canopy nitrification
Two lines of evidence support the argument that GCN is a key process in nitrogen cycling in forests.First, GCN, being processed in tree canopies, represents an input of dissolved unutilized NO 3 − that bypasses biological soil nitrification, thus contributing directly to potential pathway losses (leaching and denitrification).Field studies have reported lower net soil nitrification and NH 4 + availability under pine and oak canopy covers compared to open areas, which was explained by a lower competition for NH 4 + substrates between plants and nitrifiers in the former case 41 .Yet, our data indicate that GCN could contribute to reducing the input of atmospheric NH 4 + to the soil.A data synthesis carried out in US forests showed that net soil nitrification can contribute to produce an average of 35 kg NO 3 − -N ha −1 yr −1 , ranging from 0 (at the Harvard hardwood forest in Massachusetts) to 114.9 kg N ha −1 yr −1 (at the Fernow N-saturated mature forest in West Virginia) 41,42 .Our GCN estimates were within the lower range of estimates of net soil nitrification from the above-mentioned meta-analysis.Most soil nitrification, however, represents internal cycling, as opposed to GCN, which represents a small (4.7% of total NO 3 − -N from canopy plus mean soil nitrification, that is, 1.77/(1.77+ 35) × 100) but cumulative addition of NO 3 − to the ecosystem.Nevertheless, additional research, including both phyllospheric and soil nitrogen transformations, is needed to ascertain the absolute contribution of GCN to ecosystem nitrogen cycling.
Second, the ecological relevance of tree canopies in nutrient cycling will be underestimated if biological transformations by phyllospheric microbiota are neglected.Indeed, most forest canopies were NO 3 − -N sinks rather than sources (for atmospheric deposition) when we accounted for the contribution of GCN to the measured TF NO 3 − -N fluxes (Fig. 3b).The actual mean canopy retention (that is, the difference between TF and BD for NO 3 − -N after accounting for GCN) was equivalent to −1.03 ± 1.78 kg N ha −1 yr −1 , ranging from −4.76 to 1.33 kg N ha −1 yr −1 .The NO 3 − -N sink activity of tree canopies increased significantly with increasing TN deposition (Extended Data Fig. 5), indicating that this ecological process is relevant at low levels of nitrogen deposition 39,43 but can also contribute to tree nutritional status at higher levels of atmospheric nitrogen inputs 44 .
Our results force us to reconsider the current models used to estimate dry deposition.The canopy budget model 27 is based on the assumption that canopy exchange is the main process governing the uptake and release of ions when wet atmospheric deposition reaches forest canopies.For inorganic nitrogen, the uptake of NH 4 + and/or H + by tree canopies is assumed to be balanced by the release of cations such as K + , Ca 2+ and Mg 2+ , whereas canopy NO 3 − uptake is proportional to NH 4 + uptake 28 .The lower NH 4 + -N fluxes in TF than BD would thus commonly be accounted for by canopy uptake 24 , but we show that they could be due to transformations by epiphytic nitrifiers that bypass the trees.Furthermore, Δ 17 O provided a robust estimate of the actual atmospheric NO 3 − -N deposition, suggesting that attributing NO 3 − fluxes in TF only to the seldom-quantified fraction of dry deposition instead of GCN leads to overestimating atmospheric deposition in the oxidized form.We addressed possible limitations associated with the isotope approach considered in this study, which did not quantify nitrification directly on leaf surfaces but on collected forest water samples.One important limitation may be related to biological transformations occurring in water samples before their collection and/or during storage in the laboratory.However, a methodological test (Supplementary Text 1) was carried out at selected ICP Forests sites, where TF water samples were collected after the precipitation event and then subjected to different treatments (filtered and unfiltered water) and temperature levels (to simulate field conditions) for up to 85 days after collection.The results showed only small changes in NO 3 − concentrations and pH over time for different temperatures (Supplementary Tables 1 and 2 and Supplementary Figs. 3 and 4).The methodological test allowed the definition of a protocol for a robust estimate of water collection and storage across all ICP Forest sites, as also employed at sites included in this study.
Our findings are relevant to the debate on widespread ecosystem nitrogen saturation versus oligotrophication 18 .The GCN flux we have shown could reduce the potential inputs of NH 4 + to the soil and hence its preferential uptake (over NO 3 − ) mediated by mycorrhizal associations for the investigated species, as reported in previous studies 45,46 .We hypothesize that lower NH 4 + availability could enhance tree-microbes competition and tree nitrogen demand under increasing levels of atmospheric CO 2 (refs.18,47).However, this hypothesis should be tested in future studies, where the relevance of the GCN flux should be included in the well-investigated framework of inorganic and organic nitrogen uptake strategies and plant-microbe interactions in the rhizosphere [48][49][50] .GCN can potentially contribute to increasing the highly soluble NO 3 − in the soil, and hence its loss via leaching and denitrification 42 .Studies at the watershed scale have shown greater export of microbiologically derived NO 3 − (attributed mostly to soil nitrification) with increasing nitrogen deposition in the case of nitrogen saturation 51,52 .More studies are needed, however, to clarify the origin of the microbiologically derived NO 3 − , adding potential biological activities that are unseen in tree canopies.
The fact that this chronic net canopy retention of NO 3 − is more efficient under high atmospheric nitrogen loads also helps to explain why experimental nitrogen manipulations via soil fertilization generally underestimate the response of the tree carbon cycle to nitrogen availability under those same deposition conditions 53 .More realistic nitrogen manipulation experiments are needed 54 if a more comprehensive understanding of the coupling between nitrogen and carbon cycling is to be achieved.Finally, our results highlight the need to integrate isotopic tracers to improve the current approaches used to estimate nitrogen deposition, to provide a more robust assessment of the CLs for policies aiming at limiting nitrogen pollution.

Investigated sites and quantification of atmospheric nitrogen deposition
Ten forested sites within the Level II ICP Forests network (http:// icp-forests.net/) 55composed of the two most common European tree species (Fagus sylvatica L. and Pinus sylvestris L.) were selected to span a range of climates (Fig. 1a) 56 and nitrogen-deposition conditions (Extended Data Table 1).Specifically, sites were considered at low or high levels of nitrogen deposition if the sum of the NH 4 + -N and NO 3 − -N fluxes in TF was below or above the CL, respectively.The empirical CL is 10-15 kg N ha −1 yr −1 according to the summary table 9.2 in ref. 35.Atmospheric deposition, foliar nutrients and tree growth, among other parameters, have been continuously measured at sites in the ICP Forests monitoring network since 1995.The flux of atmospheric nitrogen deposition at these sites was quantified by measuring NH 4 + and NO 3 − concentrations in water collected in open fields outside the forests (BD) and beneath tree canopies (TF).At each site, between one and four collectors were used for BD, and 10-16 TF collectors were placed within the monitored plots, following either a systematic or random distribution, to capture the spatial variability of the canopy cover.In addition to the collectors for BD, wet-only deposition was determined at one site (in Belgium) by employing collectors whose lid automatically opened at the start of rain and closed 10-15 min after the rain ended, to avoid collecting ions from dry deposition.Bottles in which water samples were collected were placed in PVC pipes to minimize algal growth and biological transformations in the samples (Supplementary Text 1).The samples were collected and processed following a protocol harmonized across all sites, as described in the ICP Forests manual 57 .Briefly, the samples were normally collected weekly (once a week), biweekly (once every 2 weeks) or monthly (depending on the site and precipitation).An aliquot of the total volume of BD and TF water collected, between 100 and 200 ml (depending on the site), was filtered through acetate cellulose membranes with a pore size of 0.45 μm and used for measuring  N) and organic nitrogen.The total deposition of nitrogen onto forests is assumed to be typically a factor of one to two times higher than TF fluxes 59 due to the deposition of dry reactive nitrogen compounds (particulate and gaseous) onto tree canopies.The deposition of these reactive compounds, washed out during precipitation, is larger in forests than in the open field because of the larger surface area and resistance of canopies compared with open field receptors.However, here we only focused on measured water fluxes below the forest canopy and outside the forest, as dry deposition is not extensively and separately measured across all the ICP Forests sites.

Collection of foliar and water samples
BD and TF water were collected on two or three sampling dates during summer of 2016 at each of the ten sites (Supplementary Table 3).The sampling was originally planned for August and September, but it was extended from July to October 2016 due to the low precipitation at most of the sites in August.All water samples were transported to the laboratory on the day of collection and stored at 2-4 °C in the dark until processing, as described earlier.The remaining water collected on a sampling date was filtered through sterilized mixed cellulose membrane filters with a pore size of 0.22 μm (S-PakTM Membranes, Merck) to collect microorganisms for the DNA-based molecular analyses (Supplementary Text 2), which were stored at −20 °C until DNA extraction.At each site, 1-3 and 10-12 filters were used to filter the BD and TF water samples, respectively.The filtered water was stored at 2-4 °C in the dark until preparation for isotopic analyses.At each site, foliar samples were collected from five trees where TF collectors were located, chosen among those already considered for nutrient analysis in the ICP Forests network.Three additional sites in Finland were added later to increase the number of sites at the lower range of nitrogen deposition (that is, <5 kg ha −1 yr −1 , Supplementary Table 1).Three shoots from each tree were sampled in the upper, middle and lower third of the canopy.Professional tree climbers collected the samples in August 2016 (except in Sweden and Finland, where the samples were collected in October 2016 and August 2017, respectively).To avoid contact between the foliage and the ground (and possible contamination with soil microbes), shoots were sampled from the canopy and directly placed in labelled sterilized bags, which were sealed when the tree climber was still in the canopy.The sealed bags were then dropped to the forest floor and immediately placed in a box containing dry ice.The foliar samples were stored in the laboratory at −20 °C until the extraction of microbial DNA.

Analyses of stable isotopes in nitrate and partitioning between atmospheric and biological sources for TF fluxes
The site-composite samples of BD and TF water from each sampling date were used for the measurements of the stable isotopes of NO 3 − .The samples were conveyed through an anion-exchange resin (AG 1-X8, analytical grade, 20-50 mesh, chloride form) to retain NO 3 − , which was then eluted from the resin using 20 ml of 1 M KBr and processed using bacterial reduction and thermal decomposition to measure the 18 O/ 16 O, 17 O/ 16 O and 15 N/ 14 N isotopic ratios, as described in ref. 34.The oxygen and nitrogen isotopic ratios in the samples were expressed in parts per thousand (‰) relative to international standards (Vienna Standard Mean Ocean Water (VSMOW)) and atmospheric N 2 , respectively, to obtain the oxygen (δ 18 O and δ 17 O) and nitrogen (δ 15 N) isotopic compositions.The long-term analytical precisions were ±0.2‰ (δ 18 O and δ 15 N) and ±0.5‰ (δ 17 O).The isotopic composition values are presented in Extended Data Tables 3 and 4.
The excess 17 O in atmospheric NO 3 − , that is, Δ 17 O (expressed as Δ 17 O = δ 17 O − 0.52 × δ 18 O), unequivocally estimates the proportion of NO 3 − derived from atmospheric deposition, because it is determined by mass-independent fractionations during the photochemical production of NO 3 − in the atmosphere 36 .The fraction of NO 3 − from biological transformation (that is, nitrification, f Bio ) in the canopy relative to the fraction from atmospheric deposition (f Atm ) was assessed using a mass-balance approach based on Δ 17 O values 36 measured in BD and TF, as described elsewhere 33,34,60 .Briefly, we started with where Δ 17  , respectively, and f Bio and f Atm are the two unknown NO 3 − fractions from the two sources (biological and atmospheric), whose sum should be 1: f Atm included the fractions of both the wet (f Wet ) and dry (f Dry ) NO 3 − depositions washed from the canopy and the net of the fractions retained in the phyllosphere and/or taken up by leaves (f U ), that is, which can then be used in equation ( 2) to obtain f Bio as https://doi.org/10.1038/s41561-023-01364-3 f Bio and f Atm were then used to calculate the NO 3 − -N flux derived from canopy nitrification (GCN) and the actual atmospheric NO 3 − -N (after accounting for GCN) using an approach similar to that applied at the watershed scale 52 (Extended Data Table 4), that is, by multiplying the measured NO 3 − -N TF flux by f Bio and f Atm .The fact that dry nitrogen deposition was not included in the study does not affect the estimate of GCN, as the latter is calculated from TF fluxes.

Genetic analyses
For the phyllosphere, of the five trees from which shoots (n = 3) were originally collected, only three were used for the analysis of phyllospheric microbiota, and the remaining two were used when foliar samples did not look healthy (for example, in the Collelongo site in Italy) and/or when not enough microbial DNA (for example, in the forest site in Sweden) could be collected.Epiphytic microbial DNA was obtained from 5-6 g (for F. sylvatica) and 8-10 g (for P. sylvestris) of foliage randomly collected from each of the three shoots sampled per tree and placed (as a composite sample for each tree) in sterile 50-ml Falcon tubes.Thirty-five millilitres of 1:50 diluted Redford buffer wash solution (1 M Tris•HCl, 0.5 M Na ethylenediaminetetraacetic acid (EDTA) and 1.2% cetyltrimethylammonium bromide (CTAB) 35 ) was added to each tube, which was then stirred for 5 min.The washing solution was transferred to a second sterile 50-ml tube and centrifuged at 3,000g for 30 min.The pellet was transferred to 2-ml MO BIO PowerSoil bead-beating tubes for DNA extraction, which was conducted following the manufacturer's instructions (DNeasy PowerSoil Kit, Qiagen, Benelux BV; previously the PowerSoil DNA isolation kit from Mo Bio laboratories).The microbial DNA was quantified using a NanoDrop spectrophotometer (Thermo Fisher Scientific).For the water samples, filters for the genetic analyses were cut into pieces and transferred to 2-ml MO BIO PowerSoil bead-beating tubes.Microbial DNA was extracted from all samples using an MO BIO PowerSoil DNA isolation kit (MO BIO) following the manufacturer's instructions, as described for the phyllosphere.
Aliquots of microbial DNA, obtained as described above, were used to (1) prepare amplicon libraries for Illumina 16S rRNA gene sequencing and (2) quantify the functional genes involved in nitrification using qPCR analyses (expressed as the number of copies of the gene per nanogram of microbial DNA extracted from a sample), that is, genes that encode for ammonia monooxygenase subunit A (amoA) and nitrite oxidoreductase subunit B (nxrB) catalysing the oxidation of ammonia to nitrite and nitrite to nitrate, respectively.More details for those analyses are provided in Supplementary Text 2.

Statistical analyses
We used t-tests to assess whether the mean of the difference between TF and BD for nitrogen deposition in the form of NO 3 − -N and TN differed from zero.The data for determining the difference between TF and BD for NH 4 + -N and the microbial functional genes (amoA AOA, amoA AOB and nxrB) were not normally distributed, so we used the non-parametric Wilcoxon-Mann-Whitney test to test the null hypothesis that the median of the difference between BD and TF differed significantly from zero.Simple regression analyses were conducted to identify the relationships between f Bio or GCN and the data for N deposition.All statistical analyses were conducted in R 62 using the following packages: 'stats' 62 for the t-and Wilcoxon-Mann-Whitney, 'car' 63 for linear regression analyses and 'ggplot2' 64 for the figures.Finally, the package 'rnaturalearth' 65 was used for producing the map in Fig. 1a.https://doi.org/10.1038/s41561-023-01364-3Y. De Bodt from INBO (Belgium) during fieldwork and filtering of the water.This study greatly benefited from the large efforts of the site PIs and collaborators coordinating long-term monitoring within the ICP Forests network.The measurements within The Swedish Throughfall Monitoring Network (SWETHRO) are mainly funded by various regional air quality protection associations, county administrative boards and the Swedish Environmental Protection Agency.We acknowledge the E-OBS dataset from the EU-FP6 project UERRA (https://www.uerra.eu).Note that the figure reports raw measurements for each sampling date and that the accurate partitioning of the fractions of biologically derived NO 3− requires examining the ratios of the individual BD-TF pairs in the mass balance approach (see Methods).

Extended Data
Extended Data Fig. 3 | Relationship between biological fraction of nitrate and atmospheric deposition.Relationship between the fraction of nitrate derived from biological transformation in tree canopies (f Bio ) and the annual bulk deposition (BD) of total nitrogen (TN, that is, inorganic + dissolved organic nitrogen deposition) in 2016.We plot f Bio values obtained from each sampling period (Supplementary Table 3) as listed in the Extended Data Table 4, while annual values for TN in BD were considered (as reported in the Extended Data Table 1).See Extended Data Table 5 for the statistical results of the linear regression analysis.

Fig. 1 |
Fig. 1 | Investigated sites and wet deposition fluxes of inorganic nitrogen.a, Map showing the forested sites included in this study.b, Difference between TF and BD for the nitrogen fluxes NH 4 + -N, NO 3 − -N and TN.Means and 95% confidence intervals for nitrogen forms across the sites are indicated in black (n = 10, data are reported in Extended Data Table2).Orange and green rectangles

Fig. 2 |
Fig. 2 | Atmospherically and biologically derived nitrate fractions.a, Fraction of NO 3 − derived from atmospheric deposition and biological nitrification in tree canopies for the two species across the investigated sites.Means and s.d.values are shown in black.b, Fraction of NO 3 − from canopy nitrification (f Bio ) measured at each site for each sampling date.The calculation of f Bio for Brasschaat in Belgium was obtained from Δ 17 O in NO 3 − collected in BD and wet-only precipitation (wo)

4 +-N and NO 3 −- 4 + 4 +
NH 4  +  and NO 3 − concentrations (in mg l −1 ), which were then used to quantify the monthly fluxes of deposition of NH N based on the volume of water collected25,58,59 .Annual BD and TF fluxes were obtained as sums of monthly fluxes and are expressed as NH TN, with TN obtained as the sum of both inorganic forms of nitrogen (NH O TF , Δ 17 O Bio and Δ 17 O Atm are the values of Δ 17 O in TF water NO 3 − , nitrification-derived NO 3 − and atmosphere-derived NO 3 −

Fig. 1 |+-N, NO 3 −
Nitrogen fluxes expressed as difference between TF and BD across the investigated sites.Differences between throughfall (TF) and bulk deposition (BD) for each form of nitrogen compounds measured in 2016 at the 10 sites in the study.Different shapes indicate sites at low or high levels of N deposition, if the sum of NH 4 -N fluxes in TF was below or above the critical nitrogen loads (CL), respectively.Empirical CL is 10-15 kg N ha −1 y −1 according to Bobbink et al. 35 (Table 9.2).Extended Data Fig. 2 | Stable oxigen and nitrogen isotope compositions of nitrate in forest water samples.a,b, Dual isotope plots showing the distribution of δ 18 O vs. δ 15 N (a) and ∆ 17 O vs. δ 15 N (b) measured on NO 3 − in forest water samples collected on two or three sampling dates, reported in Supplementary

Data Fig. 4 |
Abundance of the microbial functional genes involved in the canopy nitrification.Abundance of the archaeal and bacterial genes that encode for ammonia monooxygenase subunit A (amoA AOA and amoA AOB, respectively) and the bacterial gene encoding for nitrite oxidoreductase (nxrB).Abundance (y-axis) is expressed as copy number of a given gene relative to nanograms of extracted DNA from foliar surfaces.All values were log-10 transformed.The sites on the x-axis follow a latitudinal order from north to south.Number of observations (n) for each functional gene is: n=3 at each site, except for amoA AOB in Bertiz, Haguenau, Savettijarvi (n = 1) and Brasschaat and Kivalo (n = 2), for amoA AOA in Kivalo and Hoeilaart (n = 2) and for nxrB.The lower number of observations at those sites were either related to low microbial DNA extracted from the phyllosphere or to copy number lower than the minimum values of 100.Extended Data Fig. 5 | Relationship between TF-BD difference (after correction for canopy nitrification) and nitrogen deposition.Relationship between the difference between throughfall (TF) and bulk deposition (BD) of nitrate (NO 3 − -N) after correction for gross canopy nitrification and bulk deposition of total nitrogen (TN).Linear regression indicated significant relationships between the two variables (slope: −0.4 ± 0.14, R 2 = 0.53, p = 0.03), which was stronger when the interaction term TN × Species was also included (R 2 = 0.92, p < 0.01).Extended Data Table 3 | Concentrations of NO 3− in the BD and TF water collected on different sampling dates.We also report δ15 N and δ 18 O for each sampling date.n/a indicates that no values were measured for a sampling date due to very low NO 3 − concentrations and/or instrument failure.[1], [2] and [3] indicate the sampling dates detailed in the Supplementary Table 3 Extended Data Table 4 | ∆ 17 O values of NO 3 − in bulk deposition (BD) and throughfall (TF) for each sampling date (indicated as [1];[2];[3], see Supplementary Table

3 − 4 +
) and the NO 3 − fraction derived from atmospheric and biological transformation (f Atm and f Bio , respectively) obtained using the mass-balance approach based on ∆ 17 O-NO 3 − in TF and BD (see Methods in the main text).Finally, we also show the partitioning between Biological and Atmospheric NO 3 − TF flux (mean and standard deviation, SD at each site), which was obtained by multiplying the total NO 3 − TF flux by f Bio and f Atm for each sampling date.n/a indicates that no values were measured for a sampling date due to very low NO 3 − concentrations and/or instrument failure.The data are shown in Figs.2-4.Note that occasional values >1 and <0, for example Höka and the sampling dates for Thetford, Brasschaat, Haguenau, Bertiz, and Collelongo (see Supplementary Text 1), were not included in the calculations of NO 3 − flux.Brasschaat* refers to isotopic measurements in NO 3 − obtained from wet-only collectors, in addition to the NO obtained from the traditional approach common at all sites and described in the Methods) Extended Data Table 5 | Results from the simple linear regression analyses for the relationships between f Bio or gross canopy nitrification (GCN) and bulk deposition (BD) of total nitrogen (TN BD ) and of NH

Table 2 )
. Orange and green rectangles to the right indicate, respectively, the range of apparent release (source) and apparent retention (sink) of nitrogen compounds by forest canopies.The daily updated E-OBS dataset for temperature (1950-2021) from the Copernicus Climate Change Service was used to produce the map (https://surfobs.climate.copernicus.eu/dataaccess/access_eobs.php).

Table 3 and
Extended Data Tables 3, 4. The blue and orange boxes indicate the published ranges of isotopic values for atmospherically and biologically derived NO 3 −, respectively, reported by Kendall et al.