Introduction

Under many natural conditions, the supply of nitrogen is limiting plant production in large areas of temperate and boreal forests (e.g., Tamm 1991). In parts of Europe and N. America, nitrogen deposition is now slowly removing this limitation (Galloway et al. 2008). Consequently, forests become N saturated as the supply of N exceeds the combined plant and microbial demand for N (e.g., Aber et al. 1998). When this occurs, leaching losses of nitrate become high, as do losses of nitrous oxide, a potent greenhouse gas.

Processes associated with N loss, e.g., nitrification and denitrification, discriminate against the heavier N isotope, 15N (Shearer and Kohl 1986; Högberg 1997). Thus, the N lost is isotopically light, leaving the remaining ecosystem enriched in 15N. As a consequence the 15N abundance may be a useful indicator of N losses from forests (Vitousek et al. 1989; Högberg 1990). Indeed, Högberg and Johannisson (1993) reported a correlation between the fractional loss of added N and an increase in δ15N over time in tree foliage in a long-term N-loading experiment. Likewise, enrichment in 15N of tree foliage has been found in field surveys of forests subject to high N loads in both Europe and N. America (e.g., Högberg et al. 1996; Emmett et al. 1998; Pardo et al. 2007). Högberg (1991) and Högberg and Johannisson (1993) also found that trees in plots to which urea, CO(NH2)2, had been applied became more enriched in 15N than did trees in plots to which NH4NO3 had been added. They ascribed this greater isotope effect to ammonia volatilization and higher rates of nitrification in the urea plots. The enrichment in 15N was thought to first occur in labile pools of N in the soil, and, therefore, soon also in the foliage of trees, while the larger pool of less labile N in the soil should change its isotopic composition more slowly (Johannisson and Högberg 1994).

In addition to these differences in overall ecosystem δ15N there are also distinct patterns of the N isotope ratio within ecosystems. In forests dominated by ectomycorrhizal (ECM) tree species, which are typical of strongly N-limited temperate and boreal forests (Smith and Read 2008), there is commonly an increase in δ15N with increasing soil depth (Riga et al. 1971; Nadelhoffer and Fry 1988; Melillo et al. 1989; Gebauer and Schulze 1991; Högberg et al. 1996; Billings and Richter 2006; Sah et al. 2006; Lindahl et al. 2007; Hobbie and Ouimette 2009; Wallander et al. 2009). The classical explanation of how this increase develops is that there is a fractionation against the heavier N isotope during plant N uptake, which leaves the remaining soil N enriched in 15N, while subsequently litter depleted in 15N becomes deposited on the forest floor (Nadelhoffer and Fry 1988). After the discovery that ECM fungi are highly enriched in 15N relative to their host plants (e.g., Gebauer and Dietrich 1993; Handley et al. 1996; Högberg et al. 1996), we proposed the additional explanation that ECM fungal material may be a precursor of the biologically more stable N in the lower horizons (Högberg et al. 1996). The high δ15N of ECM fungi has been attributed to isotopic fractionation during transaminations taking place in the fungi when N is transported from the soil, through the hyphae and into the plants (Hobbie and Ouimette 2009; Hobbie and Högberg 2012). This leaves the fungal material enriched in 15N, while the N passed on to the host plant is depleted in 15N relative to the N source (Högberg et al. 1999). Importantly, the δ15N of ECM plants is variable dependent on variations in the efficiency of transfer of N through the fungal symbiont along this pathway (Hobbie and Hobbie 2006).

Moreover, we found that in forests subject to high rates of N deposition or experimental N loading the uppermost soil horizon could be isotopically heavier than the lower horizons (Högberg et al. 1996), i.e. a pattern contrary to that found under natural N-limited conditions. We assumed that this reflected N losses from the plant available N pools through pathways (nitrification followed by denitrification or leaching of nitrate) leading to isotopic enrichment of the remaining plant available N.

Since the early exploratory phase, many studies have confirmed the observation of a higher δ15N in ECM fungal N as compared to the N in their host plants (e.g., Taylor et al. 1997, 2003; Högberg et al. 1999; Hobbie et al. 2000; Henn and Chapella 2001; Hobbie and Hobbie 2006; Kohzu et al. 1999; Trudell et al. 2004). The difference of around 10 ‰ in N-limited forests (Taylor et al. 2003) coincides with the difference between the uppermost and lower soil horizons (e.g., Högberg et al. 1996; Lindahl et al. 2007; Hobbie and Ouimette 2009). Recently, Lindahl et al. (2007) studied the relation between the species composition of the fungal community and the variation in δ15N down the soil profile in a boreal forest. They found that there was no change in δ15N during the initial stage of litter decomposition representing a decrease in C:N ratio from c. 130 to 45, during which saprotrophic fungi dominated the fungal flora. Further down, ECM fungi became dominant and simultaneously the δ15N of the organic material increased from −4.5 to +6.0 ‰, indicating uptake of N by ECM trees. This supports the hypothesis that uptake of N through ECM fungi drives the development of the 15N enrichment with increasing soil depth (Högberg et al. 1996). Indeed, in N-loaded forests, in which the biomass of ECM fungi was reduced (Högberg et al. 2007), there was a high δ15N in the foliage and the uppermost soil horizon, but lower δ15N in both 6–15 years after the N load was terminated, which was also associated with recovery of the ECM fungi (Högberg et al. 2011).

Thus, according to the above there may be two major reasons why plant N in N-loaded forests becomes enriched in 15N. These are (1) losses of N associated with processes fractionating the stable N isotopes, and (2) changes in the function of ECM fungi, especially their role in redistributing N isotopes in the ecosystem. The two are likely interrelated. However, while most papers in the past have focussed on (1), a recent study of effects of high additions of NH4NO3 provided little evidence of N losses as cause of high foliar δ15N, which was primarily attributed to change in the function of ECM symbiosis in redistributing N isotopes in the ecosystem, i.e. (2)(Högberg et al. 2011). Here, we use the urea treatments in the same long-term field experiment, in which case previous studies have shown a higher δ15N in foliage than in plots to which was NH4NO3 added, to test the hypothesis that the high δ15N in foliage of trees in the high N urea plots is the result of a change in the soil N isotope mass balance rather than a result of a change in the role of ECM symbiosis in the redistribution of N isotopes in the ecosystem.

Materials and methods

Experiment E55 at Norrliden is located in a Pinus sylvestris L. forest in northern Sweden (64°N). The forest was planted in 1953 after a prescribed burning in 1952. It is located on gently sloping glacial till, and the soil type is an Orthic Podzol; detailed descriptions of the site, stand, soil, and the experiment, which was established in 1970, are given by Tamm et al. (1999) and Högberg et al. (2006). We use data from the part of E55 involving additions of NH4NO3 or as urea. The urea plots are situated immediately to the east and up-slope relative to the NH4NO3 plots. The plots are quadratic, 30 m by 30 m each, and there are three replicate plots of each treatment. These are N0, N1, N2, and N3. N0 is the control, to which no N has been added experimentally; the N deposition in this area is around 3 kg N ha−1 yr−1. The N additions to N1–N3 started in 1971 at high levels, but were reduced in two steps (Table 1), leading to average N addition rates per year of 36 kg N ha−1 in N1, 72 kg N ha−1 in N2 and 108 kg N ha−1 in N3 through the period 1971–1990. The δ15N of fertilizer N derived from N2 is usually close to 0 ‰ (Högberg 1997); we have asked the one supplier for samples to analyze, but such material has not been archived. Data on KCl-extractable soil NO3-N from 1988 reported by Aronsson et al. (1999) showed no differences between the two forms of N added; comparing urea with NH4NO3 plots, the amounts of NO3-N kg−1 ha−1 in the humus layer were 6.0 versus 1.0, 19.5 versus 14.9 and 33.6 versus 27.5, in the N1, N2 and N3 plots, respectively.

Table 1 Nitrogen addition rates (kg N ha−1 yr−1) in the experiment E55, Norrliden, Sweden, over the time period 1971–1990 (there were no additions in 1991 or before the soil sampling in 1992)
Full size table

Every year samples of current needles have been taken from 10 trees per plot (located within the inner 400 m2 of the plot) and pooled into one sample per plot. The leaf samples were dried (70 °C, 24 h) and archived. Previously, we have reported analytical data on %N and δ15N of current needles in the urea plots from every fifth year from 1970 through to 1990 (Högberg 1991). Here, we report data on these variables for every year for the same period and the treatment effects on N content and δ15N of the mor-layer soil as sampled in 1992. Data from the NH4NO3 plots were reported by Högberg et al. (2011), and are shown here to provide a comparison to the data from the urea plots. Samples of the mor-layer were taken, using a 10 cm diameter auger from five locations within the central 400 m2 of each plot in 1992. Note that the N treatments were all suspended in 1991, and that the sampling in 1992 was conducted before the N additions to the treatments N1 and N2 that year (N3 was terminated in 1990). The mor-layer was divided into its three characteristic horizons, the S-layer (surface horizon comprising litter mixed with mosses and lichens; this horizon contains no plant roots or active ECM mycelium, see Lindahl et al. 2007), the F-layer (so called fermentation-layer containing identifiable fragments of decomposing litter and an abundance of plant roots and fungal mycelium, especially of ECM fungi), and the H-layer (which also contains abundant roots and ECM mycelium, but in which the organic matter is of an amorphous character). The mean ecosystem age, the time since the C was incorporated through photosynthesis, was estimated at 4–6, 11–15 and 27–47 years, for the S-, F- and H-layer, respectively (Franklin et al. 2003). In this forest, the pine trees retain around five age-classes of needles. Thus, the organic matter in the S- and F-layers was composed of organic matter produced during the period of the experiment studied here, while the H-layer contains older material. The N added has contributed to all horizons. Each sample from each soil horizon was weighed separately after drying (70 °C, 24 h).

Dried and finely ground samples of tree needles and soil were analysed for %N and δ15N on an elemental analyser coupled on-line to an isotope ratio mass spectrometer (an ANCA system coupled to a 20–20 IRMS, Europa Scientific Ltd., Crewe, England). We calculated the weighted average δ15N for each soil horizon, and for the entire mor-layer, using the data on soil mass, %N and δ15N. We used ANOVA to test for differences between treatments.

Results

Needle N concentration and amounts of N in the mor-layer

The N concentration of needles rose rapidly in N-loading treatments above the average level 1.2–1.3 % in the control, N0, treatment (Fig. 1a). The highest percentage of N, around 2.4 % in N3, occurred at an early stage of the experiment, and corresponded with the higher N addition rates through this stage (Table 1). There were no differences in  %N of needles as a result of applying the two different forms of N; only the dose mattered (Fig. 2). The total N content of the mor-layer roughly doubled after the N additions irrespective of dose or form of N applied (Table 2). The increases in N content of the mor-layers in the urea plots corresponded to a retention of 38, 28 and 16 % of the N added in N1, N2 and N3, respectively.

Fig. 1
figure 1

Concentration of N (upper panel) and δ15N (lower panel) of needles of P. sylvestris trees in the urea plots, 1970–1990, experiment 55, Norrliden, Sweden. Data are plot means. Treatments are described in Table 1. Th, forest thinning

Full size image
Fig. 2
figure 2

Concentration of N in needles in P. sylvestris trees in plots treated with NH4NO3 and urea, 1970–1990, experiment 55, Norrliden, Sweden. Data are plot means for the same year and treatment. Data for the NH4NO3 plots are from Högberg et al. (2011). The broken line shows y = x. Treatments are described in Table 1

Full size image
Table 2 Amount of N (in kg ha−1) in the organic mor-layer in the NH4NO3 and urea treatments in the experiment E55, Norrliden, Sweden in 1992
Full size table

δ15N of needles

An initial decline in δ15N in N1–N3 indicated that the added N was isotopically lighter than endogenous available N (Fig. 1b). The δ15N of needles in the N-loading treatments started to exceed that in the N0 after 3, 4 and 14 years of N-loading in N3, N2 and N1, respectively. The changes in δ15N were larger in N2 and N3 urea plots than in the corresponding NH4NO3 treatments (Högberg et al. 2011, Fig. 3), with increases of up to +6 ‰ in N3, and a total difference between N3 and N0 of 10 ‰ in 1990. In the N0 plots, the δ15N of needles fell about 4 ‰ in 20 years, i.e. faster than the 2 ‰ in the control plots among the NH4NO3 treatments (Högberg et al. 2011).

Fig. 3
figure 3

Natural abundance of 15N in tree foliage in plots treated with NH4NO3 and urea, 1970–1990, experiment 55, Norrliden, Sweden. Data are plot means for the same year and treatment. Data for the NH4NO3 plots are from Högberg et al. (2011). The broken line shows y = x. Treatments are described in Table 1

Full size image

δ15N of organic soil horizons

The data on δ15N of the mor-layers from 1992 showed larger differences among urea treatments than among the NH4NO3 treatments, corresponding to the greater above-mentioned differences in the needles (Table 3). The weighted average δ15N of the entire mor-layers in the urea plots were −0.8 ± 0.4, −0.7 ± 0.3, 1.4 ± 0.1 and 2.7 ± 0.1 for N0, N1, N2 and N3, respectively. Note that while the amounts of N retained in the mor-layers were similar for the two forms of N added (Table 2), the δ15N differed substantially, also with regard to the control plots in the two N treatment blocks; we are unable to explain the difference between control plots, and have chosen to focus on potential differences within the two N form treatments in our statistical analysis. Note also that the lack of effect in the N1 treatment indicates that the δ15N of the urea fertilizer must have been close to that of endogenous soil N (Table 3; Fig. 4). In both N0 and N1 δ15N increased with increasing soil depth, while the reverse occurred in N2 and N3 (Fig. 4).

Table 3 Weighted (by mass N) average δ15N for the organic mor-layer in the NH4NO3 and urea treatments in the experiment E55, Norrliden, Sweden in 1992
Full size table
Fig. 4
figure 4

The δ15N in soil horizons in the urea plots in 1992, experiment 55, Norrliden, Sweden. Data are plot mean ± SE Treatments are described in Table 1

Full size image

Discussion

There was no difference in %N of needles between the urea and the NH4NO3 plots (Fig. 2), nor was there any difference in accumulation of N in the mor-layer that was related to the form of N applied (Table 2). Hence, we assume in the following that the plots in the urea treatments were as N-limited or N-saturated as the corresponding plots treated with NH4NO3. Furthermore, data on KCl-extractable NO3 − from the soil in N treated plots in 1988 (Tamm et al. 1999; Aronsson et al. 1999; the data are given in the “Materials and methods” of this paper) showed no differences between the urea and the NH4NO3 plots (two-way ANOVA with form of N and N dose as factors). This is especially interesting, since in the case of NH4NO3 half of the N was added as NO3 −, while NO3 − in the urea plots must have been produced by nitrifying organisms (Popovic 1985) through a process known to leave the remaining NH4 + enriched in 15N (Högberg 1997).

In the NH4NO3 treated plots we have previously demonstrated that the major cause of the 15N enrichment of tree needles in the high N treatments, in particular, was a loss of the role of ECM fungi in tree N uptake, and thus in redistributing the N isotopes in the ecosystem (Högberg et al. 2011). Nitrogen losses occurred in the high N treatments, but were presumably mainly through leaching (Johannisson et al. 1999), i.e. mass flow, which does not cause isotope fractionation. We previously also observed that the δ15N of the new needles decreased after the termination of the N3 treatment, in which case also DNA sequencing and studies of fatty acid biomarkers also revealed a recovery of ECM fungi. Thus, we (Högberg et al. 2011) observed that the δ15N in the soil profile became similar to that in the control plot and also found no evidence that the N isotope mass balance had changed; we only saw that the role of ECM symbiosis in redistributing the N isotopes in the ecosystem was affected by the N treatments.

In case of the urea plots studied here, the N mass balance appeared to be roughly the same as for the corresponding NH4NO3 N dose treatments (Table 2), but the isotope effect of N treatments was clearly larger than in the NH4NO3 N2 and N3 treatments (Table 3; Fig. 3). A small effect could be attributed to volatilization of NH3 directly from the surfaces of the urea fertilizer pellets (Nömmik 1973); we assume that the elevated partial pressures of NH3 in air around individual urea fertilizer pellets do not overlap, and hence that this process is not dependent on amount of N applied. Nor does the acidity of the soil matter, since the volatilization occurs before the pellets are dissolved by rain or dew, and thus before the added N enters the soil. The larger effect on δ15N of foliage seen in the N2 and N3 treatments must, therefore, be explained by the other processes, e.g., enrichment of the remaining pool of NH4 + through isotope fractionation during nitrification followed by leaching or denitrification.

Our data from the NH4NO3 treatment (Högberg et al. 2011) provided clear evidence that the high δ15N of foliage in N-loaded forests may not necessarily involve increased N losses from the system, but can be a result of internal redistribution of N isotopes. It may, thus, be important to repeat that the losses of N from the urea and NH4NO3 treatments were similar, which suggests that the losses from the latter were not associated with N isotope fractionation, as expected if leaching of N added as NO3 − is the major agent of loss. However, as shown in the urea treatment, in which case isotopic fractionation during nitrification precedes leaching of NO3 −, the N isotope mass balance cannot be ignored. A good example of this is the very high δ15N of soils at seabird rookeries, where ammonia volatilization and nitrification are important (e.g., Mizutani et al. 1986). With regard to N deposition on forests, which mostly is composed of both NH4 + and NO3 −, i.e. should be compared with our NH4NO3 treatment, we propose that internal N isotope redistribution is the more important driver of early changes in δ15N of tree foliage, after the initial change towards the δ15N of the added N (Högberg et al. 2011, Fig. 1 here).

Global surveys show a smaller difference in δ15N between trees and soil in tropical as compared to temperate and boreal forests, with plant foliage being particularly low in δ15N relative to soil at the higher latitudes (e.g., Martinelli et al. 1999; Craine et al. 2009). In boreal ECM forests the N cycle is tightly closed, while in tropical rainforests, where trees with arbuscular mycorrhizal symbiosis are more common, the higher δ15N of foliage (Martinelli et al. 1999) is an indicator of a more open N cycle, especially with regard to the important role of denitrification in N losses from these ecosystems (e.g., Houlton et al. 2006). We propose that the pivotal role of ECM symbiosis, as shown in the previous study of the NH4NO3 plots (Högberg et al. 2011) and elsewhere in affecting the distribution of isotopes of N in ecosystems (e.g., Craine et al. 2009; Hobbie and Ouimette 2009; Hobbie and Högberg 2012), needs to be taken into account if δ15N is used to study N losses from forest ecosystems.