Interactions between leaf nitrogen status and longevity in relation to N cycling in three contrasting European forest canopies

. Seasonal and spatial variations in foliar nitrogen (N) parameters were investigated in three European forests with different tree species, viz. beech ( Fagus sylvatica L.), Douglas ﬁr ( Pseudotsuga menziesii (Mirb.) Franco) and Scots pine ( Pinus sylvestris L.) growing in Denmark, the Netherlands and Finland, respectively. The objectives were to investigate the distribution of N pools within the canopies of the different forests and to relate this distribution to factors and plant strategies controlling leaf development throughout the seasonal course of a vegetation period. Leaf N pools generally showed much higher seasonal and vertical variability in beech than in the coniferous canopies. However, also the two coniferous tree species behaved very differently with respect to peak summer canopy N content and N retranslocation efﬁciency, showing that generalisations on tree internal vs. ecosystem internal N cycling cannot be made on the basis of the leaf duration alone. During phases of intensive N turnover in spring and autumn, the NH + 4 concentration in beech leaves rose considerably, while fully developed green low bulk serice, Cabari: Cabari, F.s.: Fagus sylvatica, I.a.: Ilex aquifolium, L.d.: Larix decidua, L.h.: Licania heteromorpha, O.c.: Ocotea costulata, P.h.: Pinus halepensis, P.pa.: Pinus pinea, P.pr.: Pinus pinaster, P.r.: Pinus resinosa, P.s.: Pinus sylvestris, P.s.: Pinus strobus, Pic.a.: Picea abies, Pr.sp.: Protium sp., Ps.m.: Pseudotsuga menziesii, Q.c.: Quercus coccifera, Q.ro.: Quercus rotundifo-lia, Q.ru.: Quercus rubra, Q.s.: Quercus suber, T.b.: Taxus baccata, and V.j.: Vismia japurensis. The canopy renewal time for P.s. needles in N¨asholm (1994) was assumed to be the P.s. in Escudero and Mediavilla (2003).


Introduction
Forests cover 44 % of the land area in Europe (FAO, 2006). Nitrogen (N) fluxes and turnover in forests are therefore relevant for biogeochemical cycling of N in Europe . Forest ecosystems are generally considered to be adapted to N limitation rather than to N excess (Vitousek and Howarth, 1991;Rennenberg et al., 1998;Aerts, 1996;Xia and Wan, 2008). Due to deposition of atmospheric NH 3 , the critical N load for European forest ecosystems is often exceeded Pinho et al., 2012). This may in turn lead to elevated concentrations of NH + 4 and organic N compounds in the trees and to NH 3 emission events (Rennenberg et al., 1998;Fowler et al., 2009)

. The physiological
Published by Copernicus Publications on behalf of the European Geosciences Union. drivers controlling the direction and magnitude of NH 3 exchange between trees and the atmosphere are not well established. For this reason, more information about seasonal changes in plant N status and turnover is required to improve modelling of NH 3 exchange over forest ecosystems (Massad et al., 2008(Massad et al., , 2010. Even under conditions with high atmospheric N deposition loads to forests, internal N fluxes between the trees and the soil (ecosystem internal cycling) and within trees (tree internal cycling) are generally larger than the external N fluxes associated with leaching and gaseous emissions (Gundersen et al., 1998;Nadelhoffer et al., 1999). Ecosystem internal N cycling takes place via processes such as litter production, mineralisation and root uptake, while tree internal N cycling occurs via assimilation, re-translocation and storage. By these processes, large amounts of N are kept plant available in the system and are being moderately protected against immobilisation in stabile soil organic compounds or losses via leaching and gaseous emission. Nitrogen in the ecosystem internal N cycle is more prone to become lost than N in the tree internal cycle, although foliar N may contribute to NH 3 emission in fertile systems under conditions with low atmospheric NH 3 concentrations (Wang et al., 2011).
Internal N re-translocation and cycling patterns in trees are seasonally programmed and closely linked to tree phenology. For instance, in deciduous trees, leaf N tends to be translocated into the woody roots and/or the trunks before leaf fall, to support tree growth in the following spring. In evergreen trees, N stored in previous-years leaves can nourish new leaf growth and, especially in most conifers, the youngest age class of needles are the main sites of N storage (Millard and Grelet, 2010). In N-limited forests, trees face a relatively high selection pressure to balance the N economy with acquisition, retention and remobilisation (Pearson et al., 2002). Knowledge about interaction between foliar N status and N re-translocation efficiency in tree species differing in leaf longevity is still limited.
It can be hypothesized that the vertical N distribution within deciduous canopies is more closely related to the actual needs for photosynthesis than is the case for longer-lived conifer needles and that the proportion of tree internal N retranslocation in general will be larger in deciduous canopies than in coniferous canopies. Along with this it will be expected that N pollution may lead to an opening of the N cycle, i.e. increase the importance of ecosystem internal over tree internal N cycling with the consequence of higher probability of N losses via leaching and gaseous emissions. On this background, the objective of the present work was to investigate the distribution of N pools within the canopies of different forests and relate this distribution to factors and plant strategies controlling leaf development throughout the seasonal course of a vegetation period. The experimental sites included a deciduous and a coniferous forest situated in strong and medium N polluted areas in temperate Europe (the Netherlands and Denmark, respectively) together with a conifer forest in Central Finland with low N pollution. Different N parameters such as chlorophyll, N concentration per unit dry matter, C/N ratio, soluble NH + 4 , H + , and the N content per unit projected leaf area were measured in order to obtain information on the relation to CO 2 assimilation, tree and ecosystem internal N cycling, and potential for NH 3 exchange with the atmosphere.

Site description and sampling of plant material
The experimental sites included a Danish beech forest, a Dutch Douglas fir forest and a Finnish Scots pine forest. Details of the three localities are given in Table 1. The mean annual air temperatures in the experimental years were 9.2 and 8.6 • C for the beech forest (2008-2009); 10.2, 10.2 and 8.7 • C for the fir forest (2008)(2009)(2010); and 3.2 and 3.9 • C for the pine forest (2008)(2009) (Fig. 1). The corresponding precipitation values were 794 and 703 mm; 825, 735 and 780 mm; and 903 and 496 mm (Fig. 1).
It is important to note that the pollution with atmospheric N components differed substantially among the sites. The average atmospheric NH 3 concentration at the Dutch Douglas fir site was much higher (3.9 µg NH 3 -N m −3 ) than at the Danish beech forest site (1.3 µg NH 3 -N m −3 ) and at the Finnish pine site (0.1 µg NH 3 -N m −3 ). The NH 3 concentrations were measured with denuders at the sites and are annual averages from monthly integrating samples (Flechard et al., 2011). The dry deposition of total inorganic N via gasses and particles was estimated to be 30, 15 and 4 kg N ha −1 yr −1 for the forest sites in the Netherlands, Denmark and Finland, respectively (Flechard et al., 2011).
Leaves and needles were collected once or twice per month in 2008 and 2009 from the top (representing sun branches) and base (shade branches) of the crowns of 4 sample trees. In addition, monthly samples were taken in the fir forest in 2010. The needle samples were collected to obtain a representative mixture of different age classes. All plant samples were quickly weighed and frozen in liquid nitrogen upon sampling. Representative subsamples were lyophilized or oven dried at 70 • C.

Leaf analyses
Chlorophyll a (Chl a) and chlorophyll b (Chl b) concentrations in the leaves and needles were measured because these parameters are useful indicators of the physiological activity and degree of senescence. Extraction of chlorophyll was carried out by use of methanol for beech leaves, while dimethyl sulfoxide (DMSO) was used for coniferous needles. For analysis of bulk tissue NH + 4 concentrations and pH, frozen leaves were ground in 10 mM formic acid or milli-Q water, respectively. NH + 4 was analyzed using fluorometric detection after derivatization with o-pthaldialdehyde by FLUOstar (BMG LABTECH, Germany). For bulk tissue pH measurements, 0.2 g of leaf tissue was homogenized in a mortar on ice using 2 ml of deionized water with a little sand. The homogenate was centrifuged at 14 000 g (4 • C) for 10 min, and pH in the supernatant was measured with a microelectrode (Metrohm, Herisau, Switzerland). The NH 3 emission potential of bulk tissue extract was estimated based on the parameter , which is defined as where [NH + 4 ] and [H + ] are the NH + 4 concentration and the proton concentration ([H + ] = 10 −pH ) in the extracts, respectively.

Calculation of canopy N content and N re-translocation from senescing leaves and needles
The canopy N content (N c ), i.e. the N content of the leaves or needles per unit ground area, was calculated as the product of the measured leaf area index (I LAI ) and the average N content per unit projected leaf area (N l ) using measured N concentrations per unit leaf dry matter (N conc ) and specific leaf areas (A SLA ) from beech leaves (30 and 15 m 2 kg −1 for leaves in canopy top and base, respectively) and Douglas fir needles (79 and 54 m 2 kg −1 for canopy top and base, respectively), i.e.
For the boreal pine stand, N c was calculated as the product of average needle N concentration (N conc ; mass N per unit dry mass) obtained in the present study and estimated canopy needle dry masses (M c , i.e. 612 g m −2 ) that were derived from biometric relationships based on measured trunk diameters at 1.3 m and tree heights as input (Repola, 2009): The N re-translocation, R N , in beech was estimated as the product of the measured I LAI and the difference in N content per unit leaf area in summer and that of fallen leaves: In the conifer stands, the fallen needles were separated from the rest of the litter. The annual dry mass of green and active needles that were shed later in a year was estimated as the ratio of the measured amount of needle litter production (L) and a published mass loss factor (r M ) between green and brown needles (1/1.49, Helmisaari, 1992). This needle mass was then multiplied with the differences in N concentration between green and brown needles. From mass balance equations and L, the mass loss factor for the Dutch fir forest was estimated to be 1/1.48, i.e. very similar to that of pine.
In the above calculations it was assumed that the average leaf N contents were sufficiently well approximated by the two layers which were investigated in this study. In summer 2005, the vertical N l concentration profile in the experimental beech forest was investigated in relation to the cumulated leaf area above the sampled leaves. Based on this, a more complete calculation of N c was performed. The difference between the multilayer estimate and the one using the topmost and the lowest layer as predictor for the average N l amounted to only 1 %. The data of Meir et al. (2002) that included vertical N distributions in canopies from 5 different forest tree species were with one exception, oak, linearly related with height. Unfortunately Meir et al. (2002) did not mention the vertical leaf area distribution, and thus a similar quantitative test as above could not be performed.

Chlorophyll in leaves and needles
Both the chlorophyll concentration and the Chl a/Chl b ratio in beech increased steeply ( Fig. 2-A1 and A2) during the first half of the vegetation period (May to August), far beyond the leaf expansion phase, which lasted only until June. Senescence started at the end of September and was accompanied by a strong decrease in chlorophyll concentration and a subsequent sharp reduction of the Chl a/Chl b ratio. (Fig. 2-A2).
The chlorophyll concentration as well as the Chl a/Chl b in the evergreen trees remained constant throughout the seasons ( Fig. 2-B1, B2 and C1, C2). In general, the chlorophyll concentration at the canopy base was larger than that at the canopy top in both deciduous and coniferous trees.

Total nitrogen and carbon concentrations
The seasonal pattern of foliar dry matter N concentration differed among the two leaf habits ( Fig. 3-A1, B1 and C1). The highest N concentrations were observed in beech leaves, peaking at about 3.5 % in very young leaves and thereupon steadily declining down to around 1 % in senescent leaves. Beech leaf senescence started earlier and the reduction in N concentration progressed faster in leaves in the top of the canopy, i.e. sun crown leaves, compared to leaves at the base, i.e. shade crown leaves ( Fig. 3-A1). The N concentrations in Douglas fir needles did not show any seasonal trend but varied randomly between 1.5 and 2.2 %. In the Scots pine needles, the N concentration increased steadily from around 1 % to 1.3 % over the summer period and thereafter gradually declined over the following winter and spring. The beech and Scots pine canopies had slightly larger N concentrations at the canopy base compared to the top, while the Douglas fir trees had similar N concentrations in both canopy layers. The midsummer N concentration was largest in beech leaves (2.5 %) compared to Douglas fir (2.0 %) and the boreal pine (1.5 %). The mass ratio between C and N was relatively low, around 11, in very young beech leaves, but increased dramatically to above 40 during senescence, reflecting N re-translocation from the senescing beech leaves into the tree ( Fig. 3-A2). Douglas fir needles had a relatively constant C/N around 25 (Fig. 3-B2, Table 2), while Scots pine needles during most of the season had values above 40 ( Fig. 3-C2), i.e. more than 10 % higher than the C/N ratio in the beech leaves that were shed from the trees in the autumn. Beech leaves had the lowest N content per unit surface area and showed a progressive decline compared to needles of fir and pine, which showed little seasonal variation ( Fig. 3-A3, B3 and C3). The vertical variation in beech leaves was strongest with shade leaves containing only 50 % of the N per unit leaf area compared to the sun crown leaves. Douglas fir shade needles contained 2.3 g N m −2 , which was 32 % less than the sun crown needles with approximately 3.4 g N m −2 . In contrast, the N content in pine needles (about 3.3 g N m −2 ) was similar at both heights within the canopy.

Bulk tissue NH + 4 and ratio between NH + 4 and H + ( value)
Surprisingly high bulk tissue NH + 4 concentrations, 10 mM in the top canopy leaves, were measured in very young beech leaves sampled in the canopy top just after bud break. In fully developed green leaves, bulk tissue NH + 4 levels were relatively low (around 0.5-1.0 mM) but increased during senescence ( Fig. 4-A1). Tissue NH + 4 concentrations in Douglas fir needles were on average around 1.3 mM in base needles and 0.8 mM in top needles, except for the first summer period, where lower values were recorded (Fig. 4-B1). Scots pine needles had NH + 4 concentrations around 0.4 mM, increasing to about 0.6 mM during the summer period, but showed smaller seasonal variation than the other two species (Fig. 4-C1).
Bulk tissue (Fig. 4-A3, B3 and C3), i.e. the ratio between tissue concentrations of NH + 4 ( Fig. 4-A1, B1 and C1) and H + (Fig. 4-A2, B2 and C2), was calculated as a simple indicator for comparison of NH 3 exchange potential among the different tree species. The seasonal variation of values in beech leaves paralleled that of bulk tissue H + rather than that of NH + 4 concentration. Consequently, due to low pH in the young beech leaves, values also turned relatively low . Beech values peaked at around 400-500 during late summer, but decreased again to around 200 during 6.2 24 7.0 N flux in total aboveground litter production, F e (g N m −2 yr −1 ) 7.3 a 6.8 1.6 N flux in foliage litter production (g N m −2 yr −1 ) 1.8 5.8 0.8 N re-translocation by the trees, F i (g N m −2 yr −1) 4.4 3.4 1.9 N re-translocation efficiency, η r (% yr −1 ) b 70 37 71 N re-translocation by the trees per unit ground area (% yr −1 relative to N c ) 70 14 27 N flux in the production of new leaves (% yr −1 relative to N c ) 100 38 38 N uptake for the production of new leaves (% yr −1 relative to N c ) 30 24 11 a 2009 was a mast year. Fluxes were 36 % and 60 % higher than in normal years for C and N, respectively. b N re-translocation relative to the N content of fallen leaves when they were green in the summer. senescence ( Fig. 4-A3). This decline occurred in spite of increasing NH + 4 concentrations due to a substantial decrease in bulk tissue pH (Fig. 4-A2).
The coniferous trees exhibited much lower values than beech (Fig. 4). Douglas fir showed clear differences in values between needles in the top and at the base of the canopy. The largest values, fluctuating between 20 and 120, were recorded in the base needles due to higher NH + 4 concentrations and lower H + (higher pH). In the top layer, in Douglas fir needles consistently remained very low (<10 in most cases). Scots pine showed only moderate differences in values between needles in the top or at the base of the canopy (Fig. 4-A3). In both layers, the values reached a minimum of around 10 during late summer and increased to around 40 during the winter and spring.

Canopy nitrogen content, N re-translocation and foliar litter production
The foliage N concentration data obtained in this study were combined with field data from the investigated forest sites in order to obtain information on N cycling in the tree stands and in the ecosystems (see Sect. 2.3). Table 2 gives an overview of how the C/N ratio differed between the green and fully active leaves/needles and the brown senescent ones. The change in foliar N content was used to calculate leaf N stocks and fluxes on ecosystem scale. During senescence the C/N ratio increased by 24 % to 140 % of the summer value. The reduction in N concentration is primarily caused by tree internal N re-translocation, although other processes may potentially contribute (see further discussion in Sect. 4.3). Very large differences between the sites were shown in the values for peak summer canopy N content (N c , Table 2). The N content of the temperate Douglas fir canopy was 3-4 times larger than that in the two other forests, which had very similar N c values despite their different leaf habits ( Table 2).
The ratio between tree internal N cycling, i.e. N retranslocation (F i , Table 2) and ecosystem internal N cycling, i.e. the N flux in total aboveground litter (F e , Table 2), was calculated as 60 %, 50 % and 119 % for the beech, the fir and the pine forests, respectively. In the beech forest, intensive fruit production increased the N flux in the litter by 60 % compared to normal years. In such normal years, F i /F e would most probably be higher. The fraction F i /F e depended on the combined effects of the amount and longevity of leaves in relation to the rate of litter production from structural and reproductive plant material. All these factors differed across the investigated forests. The forest with the lowest N status had the highest F i /F e .
The absolute amount of N re-translocation (F i , Table 2) was similar for the two temperate sites (4.4 and 3.4 g N m −2 yr −1 ), despite their contrasting leaf habit and N c , while the boreal site re-translocated much less (1.9 g N m −2 yr −1 ). In relative terms, this picture changed completely. The fraction of N re-translocation relative to the initial N content in the green leaves, or in other words the N re-translocation efficiency (η r ), was lowest in the Douglas fir stand (37 %) but similar for the temperate beech (70 %) and the boreal pine stand (71 %). The contrasts between the sites became even more obvious when N re-translocation was related to N c , i.e. the peak summer canopy N content, showing values of 70, 14 and 27 % for beech, Douglas fir and Scots pine, respectively (Table 2). Assuming steady state of N c in the time scale of a few years, the proportion of N c that was invested in new leaves during year 2009 was estimated to be 100, 38 and 38 % in beech, Douglas fir and Scots pine, respectively (Table 2)  definition strictly related to the N c and the average longevity of leaves or needles. A proportion of this investment is covered by formerly re-translocated N, but an additional amount needs to be taken from either external N uptake or tree internal reorganisation of N reserves. The resulting additional N uptake, which is numerically identical to the N flux in foliage litter production, ranged between 11 % and 30 % of N c with beech having the highest and pine the lowest value (Table 2).

Differences in seasonal and spatial variation of foliar N concentrations in tree canopies across the investigated forest stands
Foliar N concentrations in beech trees decreased during the spring as the leaves expanded and increased in dry matter weight over only a few weeks. During this phase, the leaf N was diluted by the build-up of C rich leaf structural material. Thereafter, the N concentrations remained constant until the onset of senescence, when N became remobilised from the leaves to perennial woody parts, leading to a rapid decrease in both the N concentrations and chlorophyll concentrations (Figs. 3-A1 and 2-A1; see also Bauer et al., 1997;Millard and Grelet, 2010). The intensive N re-translocation is necessary to conserve N for leaf establishment in spring independently of soil N availability (Santa- Regina and Tarazona, 2001). The N re-translocation efficiency was about 55 % and 76 % for leaves at the canopy base and top of the beech trees, respectively. Similar values have been observed for beech trees at different sites in the Netherlands (Staaf, 1982) and in Japan (Yasumura et al., 2005). Beech leaves at the top and the base of the canopy started the growing season with similar dry matter N concentrations, whereupon the base leaves attained higher N concentrations during summer and autumn, reflecting higher chlorophyll concentrations ( Fig. 2-A1). Increasing chlorophyll concentrations and Chl a/Chl b ratio in beech trees during the mature green leaf period in June-August (Fig. 2-A1) reflects physiological acclimatization to increased irradiation. Chlorophyll a participates directly in the light harvesting reactions of photosynthesis, while chlorophyll b is an accessory antenna pigment acting indirectly in photosynthesis by transferring the absorbed energy to chlorophyll a. As the irradiation increases less antenna pigment is required, leading to increasing Chl a/Chl b ratio. Conversely, during senescence when chlorophyll is degraded and N remobilised, the core components of the photosystem seem to be more rapidly degraded leading to a sharp reduction of the Chl a/Chl b ratio (Fig. 2).
The evergreen coniferous trees showed much smaller seasonal changes in N pools, including N concentration, tissue NH + 4 and chlorophyll concentrations compared to the deciduous beech trees (Figs. 2, 3 and 4). The analysed needle samples included both current year needles and needles from previous years, which may have dampened seasonal changes in N pools in our observations. Minor differences between N concentrations in different age classes of Douglas fir have previously been observed at the Dutch site, older needles having slightly higher N values than young needles, as was also the case for sunlit compared to shade needles (Evers et al., 1991). Slightly higher N concentrations in older needles compared to current year needles was later confirmed by Boxman et al. (1995) and Koopmans et al. (1997). Similar differences between N concentration in young and old needles also occurred in other conifer species (Chapin and Kedrowski, 1983;Hatcher, 1990;Gebauer and Schulze, 1991;Thomas and Mead, 1992). These changes in N concentration reflect that N stored in older needles becomes remobilised to newly emerging and regrowing old needles (Millard and Grelet, 2010).
Evergreen conifers in boreal forests such as the Finnish site in the present work have evolved a conservative mechanism of nutrient turnover in order to cope with the relatively infertile soil conditions (Chabot and Hicks, 1982;Givnish, 2002). Their needles are persistent and serve as significant storage organs for carbohydrate and nutrients such as N. This storage reservoir can effectively buffer plant requirements when nutrients are not available for uptake from the soil under adverse conditions. Storage forms of N, e.g. vegetative storage proteins, are under these conditions converted to more metabolically active forms, thereby maintaining an overall relatively stable N status (Fig. 3-B1, see also Chabot and Hicks, 1982;Millard and Grelet, 2010). The retranslocation of N in Scots pine needles in Finland has been estimated to be around 73 % (Helmisaari, 1992). Across a number of evergreen shrubs and trees, the N re-translocation efficiency was on average reported to be 50 % (Aerts, 1996;Santa-Regina and Tarazona, 2001;see Fig. 5).
The higher plasticity of the N content and chlorophyll concentration in beech, even after full leaf development, compared to the coniferous trees supported our initial hypothesis that the short-lived leaves in deciduous canopies are more flexible in adjusting their functioning according to the seasonally changing weather conditions. The higher degree of shade modification in beech leaves (50 % reduced N content per unit area in shade leaves) as compared to the coniferous canopies shows that beech trees can account for vertical gradients in radiation comparably closely. There were also differences between the evergreen canopies: the fir canopy showed significant shade modification (30 %), the pine forest did not. It is obvious that the higher I LAI of the fir canopy leads to stronger extinction of photosynthetically active radiation as compared to the pine forest and, consequently, shade modification is necessary to maintain productive needles in the dark canopy zones. With regard to shade modification, the leaf habit is important for the leaf structural development. In deciduous canopies, leaves can develop according to the local conditions which may vary little over the course of a  Näsholm et al. (1994) was assumed to be the same as that for P.s. in Escudero & Mediavilla (2003 Näsholm (1994) was assumed to be the same as that for P .s. in Escudero and Mediavilla (2003).
year. In conifer canopies, leaves live longer and thus their relative position in the crown can change due to growth of new branches in the top. Nitrogen economy can lead to transient physiological modification and to transient shading, but needle structures stay more or less the same. This might set limits on shade modification in dense growing conifer canopies compared to deciduous canopies.

Seasonal variations in bulk tissue
A useful temperature-normalized parameter for evaluation and modelling of NH 3 exchange between plant leaves and the atmosphere is the apoplastic value ( apo ), which is equal to the ratio between [NH + 4 ] and [H + ] in the leaf apoplastic solution Mattsson and Schjoerring, 2002;Massad et al., 2010;Zhang et al., 2010). In the present work it was not possible to obtain apoplastic parameters. Instead, the ratio between bulk tissue concentrations of NH + 4 and H + was measured. There are indications that bulk tissue values may be important bioindicators for N turnover processes, particularly during incipient senescence Biogeosciences, 10, 999-1011, 2013 www.biogeosciences.net/10/999/2013/ where the tissue concentration of NH + 4 may increase in parallel in both leaf tissue and apoplastic solution (Mattsson and Schjoerring, 2003). There is also evidence that bulk leaf litter NH + 4 and H + concentrations in senescent tissues may be more important for the NH 3 exchange than the corresponding apoplastic concentrations . Bulk tissue values therefore still have some relevance for characterization of nitrogen pools and their relation to overall nitrogen dynamics during the growth season.
The largest bulk tissue values ( bulk ) were 500 in beech leaves, while in most cases bulk values in Douglas fir and Scots pine needles remained below 75 and 40, respectively (Fig. 4). Obviously, the relative differences between the three canopies were not directly related to the atmospheric NH 3 concentration or the dry deposition of total inorganic N at the three sites, which on average were 3.9, 1.3 and 0.1 µg NH 3 -N m −3 or 30, 15 and 4 kg N ha −1 yr −1 at the Dutch Douglas fir site, the Danish beech forest site and the Finish Scots pine site, respectively (see Sect. 2.1). Such a relationship would not be anticipated because tissue NH + 4 relations are affected by a range of species-dependent physiological processes involved in N uptake and N turnover in the leaves . Rather than differences in atmospheric concentrations, the bulk values therefore reflect foliage N status, in agreement with that beech leaves had the highest N concentration per unit leaf mass, followed by Douglas fir and Scots pine (Fig. 3). The bulk values measured in the three tree species were in all cases much lower than those measured previously in grass species (Kruit et al., 2007;Mattsson et al., 2009).
The question is to what an extent bulk can be used as a proxy for the NH 3 exchange potential of plant leaves. For the coniferous species, the bulk values were considerably lower than previously reported apo values as reviewed by Zhang et al. (2010). A clear seasonal trend in bulk values was not observed for the coniferous species in the present work (Fig. 4). In contrast, apo values in Scots pine needles in a mixed coniferous forest subjected to high N deposition were relatively high (3300) during spring and declined to 1375 during the summer/autumn (Neirynck and Ceulemans, 2008). The difference in seasonal variations between bulk and apo may reflect that NH + 4 and H + concentrations are better regulated in the bulk tissue than in the apoplastic solution. In beech leaves, bulk was very low during leaf development, increased during the mature green leaf period, and declined again during leaf senescence (Fig. 4). This pattern did not match the results derived from the NH 3 flux measurement, showing the lowest NH 3 emission during the green leaf period (Wang et al., 2011). Further gas exchange measurement and bioassays of the apoplastic solution in coniferous needles are needed to evaluate the possibility of substituting apo with the simpler bulk value. The strong seasonality of the NH + 4 concentration in beech leaves points to a surprising relationship between bulk values and N re-translocation. In phases when leaves were ei-ther expanding or started to senesce the NH + 4 concentration increased in response to the biochemical reactions involved in the N turnover ( Fig. 4; Schjoerring et al., 2000;Britto and Kronzucker, 2002). However, the bulk values were even lower during these periods, because the bulk tissue pH values decreased rapidly (Fig. 4). This pH drop can be interpreted as a so far unconsidered physiological N conservation mechanism in beech leaves counteracting gaseous NH 3 losses. The discussion of the relationships between cellular N transformations and proton production/consumption is beyond the scope of this article.

Other effects than N re-translocation that might have been contributed to decreasing N contents in leaves and needles
In agreement with many other studies, we interpreted the strong reduction in leaf N content as N re-translocation (May and Killingbeck, 1992;Killingbeck, 1996;, while we neglected NH 3 emission and N leaching from senescing leaves and needles as factors involved in N content reduction. As explained above, the contribution of NH 3 emission to the N loss in the leaves and needles during senescence could be excluded, because the measured bulk tissue values were too small to build up a sufficiently high NH 3 partial pressure in the leaves. Possible leaching losses from fallen leaves or needles were minimised by fast sampling and analysis, while leaching losses from senescing leaves or needles in the canopies were not investigated. If leaching was a substantial process, it would cause a strong N enrichment in the rain underneath the canopy. However, such enrichment was not recorded in e.g. the beech forest where canopy throughfall measurements in 1998-1999 showed that the annual sum of throughfall N on average was 0.66 g N m −2 with monthly rates during September and October ranging between 0.02 and 0.25 g N m −2 (Beier et al., 2001). This amount of N only constituted approx. 6 % of the estimated N re-translocation measured over one to two months in our study. We cannot entirely exclude effects of N leaching or NH3 volatilisation from senescing leaves, but given these considerations the contribution would very likely be small.

Relationships between nitrogen cycling, leaf habit and N status of the ecosystem
We expected that producing every year an entirely new canopy or not should have large implications for N cycling in forest trees and ecosystems and thus expected that the leaf habit, deciduous vs. evergreen, would be a major determinant for the internal N turnover in forest ecosystems. The aboveground litter production is a dominant component of the ecosystem internal N cycle, and as long as N is cycling within the trees it is protected against losses via leaching (NO − soil organic matter compounds). Trees can and do control how much N that will cycle in the ecosystem and how much that will be stored within the trees. Our results show that the investigated deciduous forest supplied the ecosystem internal N cycle with only 1.8 g N m −2 yr −1 through leaf litter production compared to 5.8 g N m −2 yr −1 in the temperate evergreen Douglas fir forest (Table 2), pointing to strong differences between trees with different leaf habit. However, our results also show that evergreen is not evergreen, because the N flux in needle litter in the boreal pine forest, 0.8 g N m −2 yr −1 (Table 2), only constituted 14 % of that of the temperate Douglas fir stand. The key difference between the two coniferous forest canopies was the canopy summer N content (N c ) which was 3.4 times larger in the temperate Douglas fir stand than in the boreal pine stand ( Table 2). The N re-translocation per unit leaf area of the pine trees reached the same efficiency as that of the deciduous forest (ca. 70 %), whereas the Douglas fir stand had a re-translocation efficiency of only 37 %, meaning that it relocated only 53 % as efficiently as the two other species (Table 2). Inclusion of data on the N re-translocation efficiency from previously published studies (Staaf, 1982;Son and Gover, 1991;Näsholm et al., 1994;Reich et al., 1995;Escudero and Mediavilla, 2003) revealed that the longer the needles stayed green, the lower the N retranslocation efficiency (η r ) was (Fig. 5). This can be interpreted as a functional adaptation against N losses in trees with short-living leaves or needles, i.e. frequent need for canopy renewal. Despite large variability especially between temperate vs. Mediterranean deciduous trees, the data still showed a general significant linear decrease in η r with increasing canopy renewal time (7 % less N re-translocated per year the canopy stayed longer; Fig. 5). A simple distinction between deciduous and evergreen mode is thus not sufficient to generalise N re-translocation efficiencies. The leaf habit is one factor among others. The two coniferous sites showed that the canopy needle mass and its N content must be taken into account together with differences in the way the evergreen forests renew their canopies and save N via retranslocation. The results from the three forest sites suggest that some of these processes are not independent from each other; the efficiency with which the N is re-translocated before leaf fall (η r ) and the time needed to renew the canopy, mainly determined by leaf longevity, were related. Because of species specific differences, we recommend investigating the relationship between η r and canopy renewal time with a larger data set of forests and possibly by inclusion of other natural and semi-natural systems. Another explanation for the observed differences in N retranslocation efficiencies could be that the high N availability at the Douglas fir site might have led to the reduced N re-translocation, because, for example, tree internal buffers for N reserves were already filled. The canopy N content reflects the N status of the system. Higher air concentrations of reactive N compounds (e.g. NH 3 ) and N deposition loads together with a more productive climate enable the establish-ment of a larger and more fertile canopy in the Douglas fir stand. The effective N re-translocation in the boreal pine forest supports a possible distinction between N scarce and N rich evergreen ecosystems, leading to a more strict N economy, i. e. reduced ecosystem internal N cycling as our initial hypothesis stated.
Summarising the results from this study, we show that forests can realise very different mechanisms to keep N losses in foliage litter production low: (i) increased N retranslocation efficiency, (ii) increased leaf longevity, and (iii) decreased canopy nitrogen content. Despite having the shortest leaf longevity and highest leaf N concentration per unit dry matter, the beech canopy reduced the N losses via leaf litter production with very efficient N re-translocation and established only a small amount of leaf mass with a small N content per unit ground. Compared to the beech stand, the N scarce boreal pine forest carried an over twice as high foliage mass, which was renewed on average every 2.7 yr with a very low N concentration per unit dry matter and a low canopy N content. Together with equally efficient N retranslocation, the N losses to the soil via foliage litter production were lowest in the boreal pine forest. The temperate N-rich Douglas fir stand had a fairly high foliar N concentration, established the by far highest foliage mass and canopy N content, and had the lowest N re-translocation efficiency. The only mechanism that kept the foliage N in the Douglas fir trees was the longevity of the needles Mechanistic models that aim to represent the dynamics of carbon and N pools and budgets need to include these factors and their relative contribution to conserve the N in forest trees. In general, the N dynamics modules require information about the N supply in the soil, N uptake demand of the trees, N losses and also the N re-translocation before litter fall, e.g. Jansson and Karlberg (2004) and Zhang et al. (2006), where η r , i.e. the N re-translocation efficiency, is a necessary input parameter. For realistic modelling, the canopy specific differences and possible interactions between N status and N re-translocation need to be taken into account. The indicated relationship between η r and canopy renewal time might help to describe systematic variation in η r across forests.

Conclusions
Leaf N pools showed clear seasonal changes in beech trees while remaining relatively constant in the evergreen canopies, i.e. Douglas fir and Scots pine. The ratio between bulk tissue concentrations of NH + 4 and H + ) ( bulk ) did not directly relate to the atmospheric NH 3 concentrations or N deposition loads at the three sites, but it reflected plant N status. This is in agreement with the fact that beech leaves had the highest N concentration, followed by Douglas fir and Scots pine. Irrespectively of the leaf habit, all investigated trees retained the majority of the canopy foliage N in Biogeosciences, 10, 999-1011, 2013 www.biogeosciences.net/10/999/2013/ the trees. This was done through effective N re-translocation (beech), higher foliage longevity (fir) or both (boreal pine forest). A simple generalisation of the mechanisms of the tree internal N cycling based on the leaf habit could not be made; the differences between the two investigated evergreen forests were as large as those measured relative to the deciduous canopy. Efficient N re-translocation is not restricted to deciduous canopies but can also be found in the N-limited forests, which released the lowest amount of N to the soil via foliage litter and to ecosystem internal cycling in this study. Apart from the efficiency of N re-translocation, the amount of N in the canopy and the fraction of annual leaf renewal are important determinants for the N turnover in trees and the contribution of N uptake from and release to the ecosystem internal N cycle. Our results indicate that there may be a general relationship of decreasing N re-translocation efficiency with the time needed for canopy renewal, i.e. leaves that live longer re-translocate relatively less N during senescence. In the stand with the highest atmospheric N deposition, i.e. the Douglas fir stand, the canopy N content and the flux rate of N in foliage litter production was largest, leading to higher turnover in the ecosystem N cycle and implying higher risks of N losses via leaching and gaseous emissions.