Assessing the Linkages between Tree Species Composition and Stream Water Nitrate in a Reference Watershed in Central Appalachia

: Many factors govern the ﬂow of deposited nitrogen (N) through forest ecosystems and into stream water. At the Fernow Experimental Forest in WV, stream water nitrate (NO 3 − ) export from a long-term reference watershed (WS 4) increased in approximately 1980 and has remained elevated despite more recent reductions in chronic N deposition. Long-term changes in species composition may have altered forest N demand and the retention of deposited N. In particular, the abundance and importance value of Acer saccharum have increased since the 1950s, and this species is thought to have a low afﬁnity for NO 3 − . We measured the relative uptake of NO 3 − and ammonium (NH 4+ ) by six important temperate broadleaf tree species and estimated stand uptake of total N, NO 3 − , and NH 4+ . We then used records of stream water NO 3 − and stand composition to evaluate the potential impact of changes in species composition on NO 3 − export. Surprisingly, the tree species we examined all used both mineral N forms approximately equally. Overall, the total N taken up by the stand into aboveground tissues increased from 1959 through 2001 (30.9 to 35.2 kg N ha − 1 yr − 1 ). However, changes in species composition may have altered the net supply of NO 3 − in the soil since A. saccharum is associated with high nitriﬁcation rates. Increases in A. saccharum importance value could result in an increase of 3.9 kg NO 3 − -N ha − 1 yr − 1 produced via nitriﬁcation. Thus, shifting forest species composition resulted in partially offsetting changes in NO 3 − supply and demand, with a small net increase of 1.2 kg N ha − 1 yr − 1 in NO 3 − available for leaching. Given the persistence of high stream water NO 3 − export and relatively abrupt (~9 year) change in stream water NO 3 − concentration circa 1980, patterns of NO 3 − export appear to be driven by long-term deposition with a lag in the recovery of stream water NO 3 − after more recent declines in atmospheric N input.


Introduction
The northeastern United States experienced relatively high atmospheric N deposition during the latter half of the 20th century [1,2], increasing N supply into some forested ecosystems enough that the availability of N exceeded stand N demand-a situation that can cause significant nitrate (NO 3 − ) leaching [3]. Substantial loss of NO 3 − contributes to an associated leaching of base cations, such as calcium and magnesium, which are important to plant growth [4][5][6], and may also have negative effects downstream [1]. Since the passage and subsequent amendment of the Clean Air Act, national emissions of NO x and atmospheric N deposition have steadily declined; however, the response of forested catchments is variable. Some have lower N export following national emission and deposition trends, while the levels of N export in others remain high and result in declining inorganic N retention [7][8][9]. Given the ecological implications of N export into stream water, it is important to understand what controls watershed responses to changes in N deposition through time.
Many factors (both belowground and aboveground) can affect the retention and export of N deposited into forests [10]. Below ground, soil organic matter is the largest pool of N in temperate forests and is a major sink for added N [11]. Microbial immobilization, plant uptake, mineralization, and nitrification control mineral N availability in the soil [12], and net nitrification has a large impact on N export due to the mobility of NO 3 − in soils. Above ground, stand age has a large impact on N retention, as young, aggrading stands usually retain more N due to greater N demand [10]. Even between stands of similar age, differences in species composition can lead to differences in N retention and loss [3,[13][14][15][16]. As a result, gradual changes in species composition through time could also impact watershed N retention but are more challenging to study due to the need for long-term records.
Fortunately, there are long-term records of changes in both stream-water NO 3 − (since 1970) and the composition of tree species (since 1959) in a reference watershed (WS 4) at the Fernow Experimental Forest (FEF) in the central Appalachian Mountains of West Virginia. From 1975 to 1984, there was a 435% increase (1.3 to 6.9 kg N ha −1 yr −1 ) in stream water NO 3 − export, and one assessment of 24 watersheds in the eastern United States found that WS 4 at the FEF had the lowest retention of inorganic N among those examined [17]. This relatively abrupt increase in stream water NO 3 − export along with other changes in stream water chemistry were likely symptoms of nitrogen saturation caused by long-term N deposition [18]. In addition, nearby measurements show a significant increase in the importance of A. saccharum through time [19], which is a species associated with high rates of NO 3 − production. The maintenance of high NO 3 − export from WS 4 despite a reduction in N deposition suggests that long-term changes within the watershed may be responsible, and that these changes may not be quickly reversed. Thus, long-term data sets for WS 4 afford the unique opportunity to assess the potential impact of changes in stand species composition on stream water NO 3 − loss and its potentially long-lasting effect on inorganic N retention.
Tree species composition could impact N retention due to interspecific differences in rate of total N uptake, and interspecific differences in their reliance on different forms of mineral N. Relatively slow-growing Fagus species, as well as coniferous species, tend to have lower rates of total N uptake, while other species, including A. saccharum and European Fraxinus and Tilia species, have higher rates of N uptake [20][21][22][23]. Therefore, should species with different N uptake requirements change in relative abundance, the overall stand demand for N could shift and alter watershed N retention.
Similarly, differences among species with respect to the mineral forms of N they prefer could also affect watershed N retention if the composition of tree species is altered. The relative uptake of different forms of N varies from species that rely mostly on NO 3 − [24], to species that prefer NH 4 + [25][26][27][28], to species that change their preference to match the form that is most available [29,30]. More specifically, A. saccharum trees, which are often abundant in northeastern and Appalachian deciduous forests, may have a strong preference for NH 4 + [21,[31][32][33][34]. While many other trees also preferentially take up NH 4 + , some acquire most of their N as NO 3 − [22]. Indeed, seedlings of several species found in central Appalachian forests (Fagus grandifolia, Tsuga canadensis, Quercus rubra, and Betula lenta) either take up more NO 3 − than NH 4 + [21], or grow better under NO 3 − additions [35]. Thus, both the total uptake of N and the variability in relative uptake of different mineral N forms by overstory trees could impact NO 3 − losses following shifts in stand species composition. Given the variation between species in both total N uptake and relative utilization of different mineral forms, it is interesting that the importance of A. saccharum in the FEF has increased substantially over the past century [19]. Since this species appears to strongly prefer NH 4 + , a shift towards a greater influence of A. saccharum on the overall community could partially explain the maintenance of stream water NO 3 − exhibited in FEF WS 4 despite recent reductions in N deposition, particularly if the species it replaces preferentially utilizes NO 3 − . In addition, A. saccharum in the FEF is associated with soils having higher NO 3 − production rates and higher soil water NO 3 − concentrations at the scale of individual trees, plots, and entire watersheds [15]. Thus, an increase in the relative importance of this species may not only diminish the demand for NO 3 − but also increase its supply. These combined effects indicate that shifts in species composition and stand NO 3 − utilization may contribute to the temporal trends observed in stream NO 3 − export from WS 4.
To assess whether changing tree species composition in WS 4 could reduce long-term watershed N retention, we took advantage of the relatively unique stand inventory and stream water chemistry data at the FEF by coupling these data with in situ measurements of NO 3 − versus NH 4 + preference for the dominant, overstory tree species found at this location. This combination of data was then used to estimate total N uptake and temporal changes in stand composition in order to evaluate the hypothesis that changes in species composition at this site have contributed to elevated NO 3 − export in stream water.

Study Site
The focus of this study was a long-term reference watershed and a nearby untreated stand at the FEF. The reference watershed (WS 4) is 39 ha at an average elevation of 792 m and has a southeastern aspect. The predominant soil type is a Calvin channery silt loam (loamy-skeletal, mixed, mesic Typic Dystrochrept), and the average annual precipitation is 145 cm [36]. The forest in WS 4-and the entire FEF-was heavily cut in approximately 1905-1910, and since that time the forest in WS 4 has been left uncut and untreated. WS 4 is dominated by temperate broadleaf trees, with Quercus spp., Acer spp., Liriodendron tulipifera, and Prunus serotina making up >75% of the tree stems. In this watershed, the forest canopy is closed along the drainage and there is no clear delineation separating the riparian zone from surrounding areas and no discernable difference in riparian vegetation compared to that of the surrounding slopes.
Continuous stream flow measurements for WS 4 began in 1951 [37], and weekly or bi-weekly stream water samples have been analyzed for their NO 3 − concentration since 1970 [36]. All precipitation and stream water chemistry variables were measured using EPA-approved protocols by the USDA Forest Service's Timber and Watershed Laboratory in Parsons, WV. The analyses and quality control measures are detailed by Edwards and Wood, 1993 [38]. From 1975 through 1984, NO 3 − export in stream water increased by 5.6 kg N ha −1 yr −1 (~435%); since that time, NO 3 − levels have remained elevated, with fairly regular~5-10 year oscillations ( Figure 1). Stream water NH 4 + concentrations average~0.05% of NO 3 − concentrations, and although dissolved organic N is not regularly measured in stream water at this site, one year of measurements in the 1995 show that 87% of N export is as NO 3 − ; thus, we focused on stream water NO 3 − export. Historically, the area has received high rates of N deposition ( Figure 1), with total (wet + dry) deposition estimated to be~10 kg N ha −1 year −1 from 1986 to 2002 [15].

Species Composition and Stand N Uptake
Complete inventories of all trees in WS 4, including the total number of live trees of all species in 2 inch diameter at breast height (DBH) categories, were completed by the US Forest Service in 1959, 1964, 1972, 1984, and 2001 [39]. To investigate changes in species composition, we calculated relative importance value (RIV) for each species in each inventory year as the average of its relative abundance (RA, the number of stems of that species divided by the total number of tree stems) and its relative basal area (RBA, the basal area of that species divided by the total tree basal area). We estimated the total N uptake by the trees in WS 4 as the sum of annual N storage in aboveground woody biomass and annual N return to the soil via litterfall. Complete forest inventory data  were used to estimate annual woody N storage, and since these were 100% live-tree inventories, tree death is accounted for in these measurements, and in our estimates. To determine the N concentration in aboveground woody tissue, trees greater than 8 cm in DBH were cored in 16 plots (10 m radius) spread evenly throughout WS 4 in the summer of 1998 (Christ and others 2002). Using these cores, the width of the last 5 growth rings was measured, and the wood within 1 cm of the bark was ground and analyzed for N concentration by Dumas combustion [40] using a Carlo Erba 1500 CNS elemental analyzer. The total aboveground woody biomass of each tree was estimated with FEF-specific allometric equations [41], and annual N storage was then calculated as the product of annual biomass increment and woody tissue N concentration. Using the DBH and annual N storage, a regression equation was built to estimate the annual woody N storage based on the DBH of any tree in the watershed (R 2 = 0.790): log(annual woody N storage) = −2.256 + 2.182 log(DBH) + a where a is a species-specific constant (Table A1) based on the average residual for each species (Christ and Peterjohn, unpublished data).
Total autumnal litter fall mass (~September through December) was collected annually beginning in 1988 by the US Forest Service using 25 litter traps throughout the watershed (0.7679 m 2 wooden frames with bottoms of~0.625 × 0.625 cm-opening metal mesh). A relationship between autumnal litter fall and total stand basal area was created using the total basal area measured at 13 long-term growth plots in WS 4, and the total litter fall measured in 1989, 1994, 1999, and 2009 (R 2 = 0.887). Using this relationship, we estimated total litter fall for the years of stand inventories prior to the start of the collection of litterfall data (1959, 1964, 1972, and 1984). We then estimated each species' litter N returns for all inventory years using the relationships between a tree species' RBA and the species-specific litterfall N contents at 16 plots in 1998 [13].

15 N Labeling
To avoid affecting the δ 15 N of materials in the long-term reference watershed, we used a "test area" located in a nearby untreated area of the FEF (<1 km from WS 4) to measure the relative uptake of NO 3 − versus NH 4 + . This area has a similar elevation, slope, and tree composition to WS 4, and an east-northeasterly aspect. Unlike WS 4, small (0.2-ha) plots in this portion of the FEF were harvested to selected basal areas in the 1980s. However, for this study we selected trees within an area showing no signs of harvest, and the trees selected were of similar size to those in WS 4.
At our "test area" in early July 2014, we conducted a 15 N-labeling experiment similar to one by performed by McKane et al. [42] to determine the relative uptake of NH 4 + vs. NO 3 − for 6 major tree species at the FEF: A. rubrum, A. saccharum, B. lenta, L. tulipifera, Q. rubra, and P. serotina. We used the holes in pieces of commercial peg board (625 cm 2 each, with 10 rows × 10 columns of holes spaced 2.54 cm apart) to evenly space injections of 3.5 mM 15 N as K 15 NO 3 in one area (1 mL per hole), and 3.5 mM 15 N as 15 NH 4 Cl in another area under the canopy (within~3 m of the trunk) of five mature trees of each species. The solutions were injected midday at approximately the boundary between organic and mineral soil horizons-a depth of~3 cm-using a syringe needle with four side ports. Based on the soil NH 4 + and NO 3 − concentrations, we estimate that this injection increased background N concentrations by 10% and 5%, respectively. After three hours, we harvested fine roots (<2 mm diameter) from a depth of~3 cm at each injection site, and roots from one unlabeled area under each tree to measure the natural 15 N abundance of root tissue. The sampled roots were traced as far as possible towards the target canopy tree, and we compared the morphology of the collected roots to the fine roots of nearby seedlings of the same species. All species had distinct root characteristics except the two Acer species. Thus, we selected A. saccharum and A. rubrum trees that had no nearby Acer spp. within~15 m.
All harvested roots were placed on ice and transported to the lab, where they were soaked in 1 M CaSO 4 for 1 min to remove unassimilated N from the Donnan free space [43]. They were then dried at 65 • C for 48 h and ground to a fine powder in a dental amalgamator (Henry Schein, Inc., Melville, NY, USA). Approximately 5 mg of each sample was wrapped in tin capsules and analyzed for δ 15 N via isotope ratio gas chromatography-mass spectrometry at the Central Appalachian Stable Isotope Facility that is part of the University of Maryland Center for Environmental Science Appalachian Laboratory (Frostburg, MD, USA).
We calculated root uptake of 15 N from the labeled N pool as described in Burnham and others [44]. We first converted δ 15 N values to the fraction of the heavy isotope in the sample (F) using the 15 N/ 14 N ratio in each sample (R sample ) [45]: where R std = 15 N/ 14 N ratio in atmospheric N 2 (0.0036764). Using the root tissue N content and F, we calculated the µmol 15 N g −1 root, and then estimated the rate of 15 N uptake from the 15 N-labeled pools by dividing the 15 N excess ( 15 N content of labeled-unlabeled roots from the same tree) by the exposure time (3 h). Finally, we calculated total uptake of 15 N label ( 15 NH 4 + + 15 NO 3 − ) and the percent that was taken up as NH 4 + and NO 3 − .

Data Analysis
Our overall 15 N label study design included six species, and five trees per species, with a measurement of NO 3 − vs. NH 4 + uptake associated with each tree. We used a nested ANOVA with Tukey's HSD post hoc test (α = 0.05) to determine if the percent of total N taken up as NO 3 − varied by species. The model included the effect of tree nested within species. We then performed one-tailed t-tests to determine if the contribution of NO 3 − to total uptake of N from the labeled pool was greater than 50%, which would indicate a significant preference of NO 3 − over NH 4 + . We used the error terms in our plot-level RBA vs. leaf litter N return and BA vs. woody N storage models to run a Monte Carlo simulation to estimate the uncertainty in our total stand N uptake calculations. For this simulation, we assumed errors were normally distributed and randomly sampled 100 times from the error distribution, and we report uncertainty estimates in woody N storage, litter N return, and total N uptake are reported as 95% confidence intervals.

Results
From 1959 to 2001, total stand density in WS 4 decreased 18% (from 372 to 305 trees ha −1 ) and total stand basal area increased 45% (from 24.3 to 35.2 m 2 ha −1 ). In 2001, eight species accounted for~85% of the stand composition (84.6% of stems and 85.8% of basal area): Quercus rubra, Q. prinus, Acer saccharum, A. rubrum, Liridendron tulipifera, Prunus serotina, Betula lenta, and Fagus grandifolia. Over the study period, five of these species increased in RIV, and three decreased (Figure 2). The RIVs of A. saccharum and A. rubrum increased 5.8 and 8.5%, respectively, the most of any species. While the RIV of A. saccharum increased, its relative basal area decreased slightly (1.4%) and the number of stems increased substantially (from 8.9% to 21.9%) throughout the period examined. The RIV of Q. rubra increased to a more modest degree (2.9%), with its relative basal area increasing from 22.6% to 32.3% and its relative abundance decreasing from 20.4% to 16.7% throughout the study period. The RIV of Q. prinus, B. lenta, and F. grandifolia all declined through the study period ( Figure 2). The RIV of Q. prinus fell from 6.8% to 5.6%, and the RIV of B. lenta fell from 6.9% to 3.8%. While there was only a slight decline in the RIV of F. grandifolia, from 4.1% to 3.7%, its relative basal area fell from 5.4% of the stand to 3.4%, but its relative abundance increased from 2.8% to 4.0%.
Aboveground woody N storage increased from 6.4 (6.1-6.7 95% CI) to 9.8   (Figure 3). For the second scenario, using all available estimates of NO 3 − uptake, the stand uptake of NO 3 − increased from 17.6 to 20.5 kg N ha −1 yr −1 (2.6%) and uptake of NH 4 + increased from 13.1 to 14.7 kg N ha −1 yr −1 (1.5%) from 1959 to 2001. In addition, the percent of total stand uptake of N as NO 3 − increased slightly 57.4% to 58.3% (Figure 3). Thus, in neither of the two scenarios did the observed change in the importance of A. saccharum reduce the absolute amount NO 3 − uptake, and in only one scenario was the relative amount of NO 3 − uptake reduced-but this apparent reduction was extremely small.

Discussion
Unexpectedly, the tree species we considered did not differ in their relative uptake of NH 4 + and NO 3 − and utilized significant amounts of both forms in their mineral N nutrition. This is surprising because prior studies found large differences in the relative uptake of N as NH 4 + vs. NO 3 − for temperate forest species [21,22]. Notably, in past studies, A. saccharum trees took up substantially less NO 3 − than we found using an in situ 15 N-labeling technique (Table 2) [21,[32][33][34], and it seems likely that methodological differences could account for the higher relative NO 3 − uptake in this study [23]. Most of the prior research on the form of mineral N uptake utilized seedlings [21], hydroponic techniques [27,46], or N depletion in a simulated soil solution [20,27]-techniques that do not account for some aspects of in situ soil N dynamics. Perhaps most importantly, the differential diffusional resistances of NH 4 + and NO 3 − in soils [47] are not represented in hydroponic and simulated soil solution techniques. It is possible that tree preferences for NH 4 + vs. NO 3 − are dynamic through time, particularly as the rate of N deposition changes. However, the relative contributions of NH 4 + and NO 3 − to total N deposition have not changed substantially (Figure 1), and we therefore believe that large changes in tree N form preference due to changing relative availability of the two mineral N forms is unlikely. Thus, assuming that our 15 N-labeling experiment is representative of the long-term mineral N form preference of these tree species, NO 3 − may contribute more to N nutrition of trees than previously thought due to the greater rates of transfer of NO 3 − to roots in the soil. Since the species examined did not differ in their relative contribution of NO 3 − to total N uptake, it seems unlikely that changes in stand composition contributed to the relatively rapid increase in NO 3 − export or to the long-term persistence of low N retention via a reduction in the demand by trees for NO 3 − . Furthermore, since the stand N demand may have increased over the second half of the last century, it may have contributed to the gradual and slight decrease in soil and stream water NO 3 − since the early 1980s [48]. Although a forest inventory has not conducted after 2001, there have been no major changes in the stand or significant disturbances in this time. We speculate that the reduction in stream water NO 3 − concentration circa 2010 resulted from decreasing N deposition with a significant lag after this decline in deposition started in the early-1990s. Thus, it appears that the large increase observed in NO 3 − export from WS 4 in approximately 1980 resulted from an enhanced supply of available NO 3 − via deposition, and the long-term trend in stream water NO 3 − is controlled primarily by atmospheric N inputs with a lag in recovery as inputs decline.
Although changes in stand NO 3 − demand do not seem to account for the increase in NO 3 − export in stream water, shifts in stand composition could still affect NO 3 − production in the soil and thus contribute to a lag in the recovery of stream water NO 3 − export after deposition declines. At several locations in the eastern U.S., A. saccharum trees are associated with high rates of soil net nitrification and low soil C:N ratios [16,31,32,49,50], including WS 4 and other locations in the FEF [13,15], and nitrification rates are positively associated with stream NO 3 − export [51]. The relationship between A. saccharum abundance and nitrification is driven, in part, by relatively labile litter and low N residence time [15,52]. To make an initial assessment of the potential impact of species shifts on soil NO 3 − production and stream water NO 3 − export, we used previous plot-level measurements of net nitrification potential and the relative importance and relative basal area of tree species in WS 4. We estimated that net nitrification potential increases 0.02 kg ha −1 day −1 for every 1% increase in A. saccharum importance value (R 2 = 0.45) and decreases 0.017 kg ha −1 day −1 for every 1% increase in A. rubrum importance value (R 2 = 0.13) [14]. Similarly, net nitrification potential increases 0.017 kg ha −1 day −1 for every 1% increase in A. saccharum relative basal area (R 2 = 0.20) and decreases 0.016 kg ha −1 day −1 for every 1% increase in A. rubrum relative basal area (R 2 = 0.12) When analyzed in the same manner, no other species was associated with significant changes in net nitrification potential. Since A. saccharum and A. rubrum had large changes in relative importance value and basal area from 1959 through 2001, and have opposite associations with net nitrification potential, we assessed their potential impact on soil NO 3 − supply and NO 3 − loss to stream water. To arrive at an annual estimate, we assumed that: (1) the estimated daily rate of change in net nitrification potential applied during the months of May through August; (2) only 50% of the estimated daily rate occurred during March, April, and September through November, when the rate of nitrification is lower [53]; and (3) the species change had no effect on net nitrification potential during the months of December through February, when very little nitrification takes place.
The decline in A. saccharum and increase in A. rubrum relative basal area in WS4 suggest that nitrate production via nitrification was 19.4 kg N ha −1 yr −1 lower in 2001 than in 1959. However, our plot-level data show a stronger relationship between relative importance value and nitrification potential. Furthermore, past studies have detected a strong relationship between soil NO 3 − concentration and A. saccharum abundance [31,54]. Thus, using the relationship between these species' relative importance values and nitrification potential, our initial approximation suggests that the effects of A. saccharum and A. rubrum on soil NO 3 − production from 1959 to 2001 mostly offset each other, with the negative effect of A. rubrum on nitrification causing a net decrease in the rate of NO 3 − production of 2.6 kg NO 3 − -N ha −1 yr −1 within WS 4. However, the majority of the increase observed in the importance of A. rubrum occurred in a silvicultural compartment of the watershed (compartment WS 4c) that produces very little NO 3 − in the soil, and that has very low NO 3 − concentrations in soil water collected by tension-free lysimeters [55]. Thus, it is unlikely that this region of the WS 4 contributed to the observed patterns in stream NO 3 − export. Additionally, this subcompartment contains no A. saccharum trees, so the increased importance of this species only occurred in the portions of the watershed where nitrification and soil solution NO 3 − levels are currently much higher [13,55]. Although it is unclear why A. saccharum has increased in importance at this site, we believe that this is a long-term successional change due to the decline of other subcanopy species.
Considering these known spatial patterns in NO 3 − availability, we refined our initial assessment to~86% of WS 4 by excluding compartment WS 4c where NO 3 − availability is very low. Taking this approach, we estimate that the net effect of changes in the importance of A. saccharum and A. rubrum was to increase soil NO 3 − production by 3.9 kg NO 3 − -N ha −1 yr −1 from 1959 through 2001. The long-term change in species composition resulted in a 2.7 kg N ha −1 yr −1 increase in NO 3 − demand, which mostly offsets the estimated increase in soil NO 3 − production. Thus, we estimate that a net increase of 1.2 kg NO 3 − -N ha −1 yr −1 was available for leaching into stream water. Consequently, it seems that patterns of NO 3 − export were primarily driven by long-term changes in N deposition, but changes in tree species composition may have contributed an increase in soil NO 3 − production and thus to a lag in the recovery of stream water NO 3 − export, which remained 3.5 kg N ha −1 yr −1 higher from 1992 to 2001 (~5.0 kg NO 3 − -N ha −1 yr −1 ) (Figure 1) than the export that occurred from 1970 to 1979 (~1.5 kg NO 3 − -N ha −1 yr −1 ). This first-order estimate illustrates that understanding the effect of N deposition on the temporal dynamics of stream water NO 3 − loss requires a relatively complete understanding of how changes in forest species composition can influence the balance between nutrient supply and demand. Moreover, the spatial patterning of N supply and demand within a watershed and connectivity to stream discharge and N export may also be important. We suggest that the recent reductions in atmospheric inputs of N in the eastern US may result in a delayed return of stream water NO 3 − losses to "baseline" levels in situations where a long-lasting shift in the composition of tree species changes the inherent rates of soil NO 3 − production and biotic NO 3 − demand.
Appendix A