A New Conceptual Model of Nitrogen Saturation Based on
Transcript Of A New Conceptual Model of Nitrogen Saturation Based on
Ecosystems (2011) 14: 615–631 DOI: 10.1007/s10021-011-9432-z
Ó 2011 Springer Science+Business Media, LLC
A New Conceptual Model of Nitrogen Saturation Based on Experimental Nitrogen Addition to an Oak Forest
Gary M. Lovett1* and Christine L. Goodale2
1Cary Institute of Ecosystem Studies, Box AB, Millbrook, New York 12545, USA; 2Department of Ecology & Evolutionary Biology, Cornell University, E215 Corson Hall, Ithaca, New York 14853, USA
ABSTRACT
The dominant conceptual model of nitrogen (N) saturation in forests predicts the temporal patterns of key N cycling indicators as an initially N-limited forest is progressively enriched in N. We present the results from a long-term N addition experiment in an oak forest in southeastern New York State, USA, which do not conform to the predictions of the conceptual model in several ways. In contrast to the predictions of the conceptual model, the foliar N concentrations in the N-treated stands of our study increased to about 20% above the levels in the control stands and then remained essentially constant, and nitrogen leaching from the treated stands increased almost immediately after the start of the experiment, prior to the onset of elevated nitrification. Concentrations of N in soil solution of the Ntreated stands peaked at over 150-fold greater than the concentrations in the control stands. There were no significant changes in potential net N mineralization. Tree mortality increased in the treated stands, but the tree mortality did not appear to be the primary cause of the excess nitrate leaching. Based on these results and those of other recent studies, we present a new conceptual model of the N saturation process focused on the mass balance of N rather than the temporal dynamics of N cycling indicators. The
mass balance is characterized by inputs of N from atmospheric deposition and fertilization, internal sinks in the vegetation and soils, and outputs to leaching and gaseous losses. The key points of the conceptual model are (1) added N can flow simultaneously to all sinks and losses in the system, (2) the fate of the added N and the temporal patterns of flow of N depend on the strength of the sinks and the factors that control them, and (3) the movement of N to the various sinks determines how N saturation is manifested in the ecosystem. We distinguish capacity N saturation, in which the sinks in the vegetation and soil are zero or negative, from kinetic N saturation, in which the sinks are positive but lower than the N input rate. The sink strengths in the vegetation and soil have two components, one due to carbon (C) accumulation in the system and the other due to change in the stoichiometry (C:N ratio) of the pool. Further work quantifying the magnitudes and controlling factors for the N sinks will allow better prediction of the dynamics of N saturation in different types of forested ecosystems.
Key words: nitrogen; saturation; forest; leaching; oak; nitrification; nitrogen mineralization; foliar nitrogen; fertilization.
Received 29 April 2010; accepted 6 February 2011; published online 19 March 2011
Author Contributions: Lovett designed and executed the field study. Lovett and Goodale collaborated on the development of the conceptual model and the preparation of the paper.
*Corresponding author; e-mail: [email protected]
INTRODUCTION
Elevated levels of atmospheric nitrogen (N) deposition are a problem throughout the developed world, and are increasingly becoming a problem in
615
616 G. M. Lovett and C. L. Goodale
developing countries (Galloway and Cowling 2002). Atmospheric N compounds emanating from both fossil fuel combustion and agricultural activities are transported downwind and may be deposited on ecosystems far from the source of emission (Vitousek and others 1997). This excess N deposition may have a wide range of effects on ecosystems, including: fertilization effects, nutrient imbalances, accumulation of N in plants, soils and sediments, enhanced losses of gaseous nitrogen oxides, and acidification of soils and surface waters caused by leaching of nitrate (Fenn and others 1998; Driscoll and others 2003).
Nitrogen saturation is the term used to describe the changes that occur in a terrestrial ecosystem as it is continually exposed to elevated levels of N deposition (Agren and Bosatta 1988; Aber and others 1989). The dominant conceptual model of N saturation was proposed by Aber and colleagues (Aber and others 1989, 1998) as a hypothetical series of changes that would occur as an N-limited ecosystem accumulates deposited N (Figure 1). In this model, the ecosystem is viewed as progressing through a series of stages of N saturation, in which stage 0 is a system that is strongly N limited, stage 1 is characterized by changes indicating increasing N sufficiency, stage 2 represents the initial symptoms of N saturation such as elevated N leaching, and stage 3 represents N-induced decline in which the productivity of the forest decreases and nitrate leaching accelerates (Aber and others 1998). As this sequence progresses, enhanced N deposition makes N available to N-limited plants, which enriches plant tissues and litter, the litter N is transferred to soil organic matter, stimulating N mineralization and nitrification, and eventually resulting in elevated nitrate leaching from the ecosystem. The authors hypothesize a strong, continual increase in
Figure 1. Aber and others’ (1998) hypothesized patterns of response of forest ecosystem properties to continuing N additions. Used with permission from Aber and others (1998).
foliar N concentration through the progression of these stages, leading to an increase in net primary productivity (NPP) of the N-limited system, an initial increase and then a decrease in N mineralization, and decreasing ratios of calcium:aluminum and magnesium:nitrogen in the foliage. In stage 3, the system undergoes a decline in NPP and an increase in nitrate leaching. Though not explicitly diagrammed in Figure 1, Aber and others (1998) recognized that the major sink for added N in forest ecosystems is in the soil, and they described several potential mechanisms by which N could be incorporated in soil organic matter.
The fate of N added to an ecosystem is constrained by the N mass balance as expressed in the following equation
I À dNv=dt À dNs=dt ¼ L þ G
ð1Þ
where I = the N input rate to the system (deposition plus N amendments), dNv/dt is the vegetation sink rate (the net rate of incorporation of N into vegetation), dNs/dt is the soil sink rate (the net rate of incorporation of N into detritus and soil), L is the rate of leaching loss, and G is the rate of gaseous loss of N.
It is important to note that the mass balance in equation (1) is expressed in terms of rates, and that losses from the ecosystem can occur when the rate of N addition exceeds the rate at which N can be incorporated into the vegetation and soil sinks, even if the capacity of those sinks is not saturated. This suggests that we should distinguish kinetic N saturation, in which the rate of N input exceeds the vegetation and soil N net sink rates, from capacity N saturation, in which N has accumulated in the vegetation and soils to the point that further net N retention in these sinks does not occur. In kinetic N saturation, the vegetation and soil sinks are still active, though N losses will occur simultaneously with N retention. In capacity saturation, N retention in the vegetation and soils is negligible and system inputs should equal outputs. Both types of saturation are related to the accumulation of N in organic matter in the ecosystem. Capacity saturation occurs when the accumulation of N in the ecosystem is zero (or negative, in the case of a deteriorating system), and kinetic saturation occurs when the accumulation rate is positive but less than the N input rate. Capacity saturation is rarely observed in real forests unless disturbance reduces the vegetation sink rate or turns the vegetation into a net source of N within the system.
Ecosystem sinks for N are associated with a specific time frame, either implicitly or explicitly, and
A New Conceptual Model of Nitrogen Saturation 617
the time frame determines which processes are important. For instance, incorporation of N into microbial biomass which then dies and is remineralized over several days, may represent a sink or source of N on an hourly or daily time scale, but does not contribute to N sinks evaluated over annual time scales. For the purposes of this paper, we define the time scales of interest for the vegetation and soil sinks in equation (1) to be on the order of years to decades or longer, because this scale represents a policy-relevant time frame for ecosystem sequestration of pollution-derived N.
The responses of forests to elevated rates of N deposition can be evaluated by several approaches, each of which has strengths and weaknesses. The most straightforward approach is long-term observation of forests subjected to N deposition (for example, Peterjohn and others 1996). However, this requires decades of measurements and may be confounded by simultaneous changes in other factors affecting the forests, such as climate or disturbance. A second approach is comparative or gradient studies, in which ecosystem properties are measured in ecosystems receiving different rates of N deposition (for example, McNulty and others 1991; Lovett and Rueth 1999; Boggs and others 2005; Thomas and others 2010). This eliminates the difficulties of long-term studies but introduces uncertainty caused by potential differences in the ecosystems caused by factors other than the rate of N deposition. In gradient studies, inferring the temporal sequence requires accepting the assumptions of the space-for-time substitution, that is, that processes that occur first in a temporal sequence also occur first along a gradient from low to high N deposition. A third approach is experimental studies, in which the rates of N deposition are enhanced by adding N to experimental plots (for example, Magill and others 1996, 2004; Wallace and others 2007). Careful pairing of experimental and control plots can minimize the influence of factors other than the level of N addition. However, because the experimental N addition rate is higher than ambient rates, these experiments run the risk of altering the processes controlling the fate of N in the ecosystem. In particular, experiments enhance kinetic saturation by increasing N input rates without necessarily increasing the N sink strengths in the ecosystem. These approaches can also be used in combination, for example, long-term experiments and experiments done across gradients.
The Aber and others’ (1998) conceptual diagram (Figure 1) focuses on which ecosystem properties are expected to respond to elevated N deposition, and the temporal sequence in which they respond.
All three types of studies can be used to evaluate this conceptual model. Aber and others compared their conceptual model with results from several experimental studies in the eastern U.S. including studies in Massachusetts (Magill and others 2004), Vermont (McNulty and others 1996) and Maine (Magill and others 1996) as well as a gradient study in spruce-fir forests across northern New York and New England (McNulty and others 1991). In general the experimental responses agreed with the hypothesized trends, if one assumes that forest ecosystems start at different places along the N saturation continuum and may proceed at different rates, depending on the degree of initial N limitation induced mainly by past land use (Aber and others 1998).
However, not all studies have supported the hypotheses diagrammed in Figure 1. For example, in experimental studies in Europe at sites in which N was either experimentally added or removed, nitrogen leaching was often the first variable to respond to the treatment, rather than the last as suggested by Figure 1 (Emmett 2007). Recent studies across N deposition gradients in the eastern and midwestern U.S. have shown variable concordance with the hypotheses illustrated in Figure 1. Tree growth increases with N deposition for some tree species but not others (Thomas and others 2010). Foliar nitrogen sometimes increases along an N deposition gradient (Pardo and others 2006; McNeil and others 2007) and sometimes does not (Aber and others 2003). Likewise, net N mineralization and nitrification sometimes increase along a nitrogen deposition gradient (McNulty and others 1991; Aber and others 2003) and sometimes do not (Lovett and Rueth 1999). Across the Northeastern U.S., N leaching tends to be low in watersheds that receive less than about 8 kg N ha-1 y-1 of atmospheric N deposition (Aber and others 2003). For deposition rates above 8 kg N ha-1 y-1, some sites show N leaching and some do not (Aber and others 2003). Whether these watersheds leach N or not is controlled in part by tree species composition, soil C:N ratio, and history of disturbance (Goodale and others 2000; Lovett and others 2002). It is often difficult to determine if the absence of trends in these gradient studies results from a true lack of response in the system, or an inability to discern the signal within the noise of spatial variation inherent in gradient studies. Overall, however, the literature suggests substantial variation in the degree to which forest ecosystems conform to the hypotheses expressed in Figure 1.
In this study, we tested the hypothetical series of changes proposed by Aber and others (1998) with
618 G. M. Lovett and C. L. Goodale
an N addition experiment in a mixed-oak forest in southeastern New York State, USA. The research site had similar vegetation to the Harvard Forest (Massachusetts) oak site used as one of the test sites for the conceptual model by Aber and others (1998) but was in a ridgetop rather than a valley location. Our goal was to determine if the temporal sequence of changes in key N cycling variables matched those illustrated in Figure 1, and if not, to propose a different conceptual model in better agreement with the observations.
METHODS
Research site
The research site is located at the Cary Institute of Ecosystem Studies (41.797°N, 73.734°W) in the Hudson Valley of southeastern New York State, USA. The forest is an upland mixed-oak woods dominated by red oak (Quercus rubra L.), chestnut oak (Quercus prinus L.) and several species of hickory (Carya sp.) regenerated from selective cutting associated with use as a woodlot early in the twentieth century (Glitzenstein and others 1990). The annual average temperature at the site is 9.6°C and precipitation averages 111 cm y-1. The bedrock is mostly shale and slate and soils are predominantly thin and well-drained silt loams of the Nassau and Woodlawn series (Glitzenstein and others 1990) classified as lithic dystrudepts. Mean annual precipitation pH at the site ranged from 4.0 to 4.4 between 1984 and 2004 (Kelly and others 2002). Ambient inorganic N deposition (wet + dry) in the period of study (1996–2006) averaged approximately 9 kg N ha-1 y-1. Wet deposition was collected continuously with a wet-only precipitation collector, and dry deposition was estimated from continuous filter-pack measurements of air concentrations of HNO3 vapor and particulate NO3- and NH4+, applying a dry deposition model parameterized with meteorological data at the site (Kelly and others 2002).
We chose 6 pairs of plots, each 20 m in diameter. The two plots of each pair are within 40 m of each other. Pre-treatment measurements were made in July–August 1996, and the N addition began in November 1996. One plot of each pair was treated with granular NH4NO3 fertilizer at a rate equivalent of 100 kg N ha-1 y-1 from 1996 to 1999, and then the N addition rate was decreased to 50 kg N ha-1 y-1 from 2000 to 2006. This annual application was divided into 4 equal doses each year. Fertilizer was applied to the forest floor; although application to the canopy would have
been more realistic, it would be technically much more difficult, and prior data from these forests indicate that canopy uptake of ambient N deposition is negligible (Lovett and others 2000a).
Sample Collection and Processing
Foliar chemistry was monitored on full-sun leaves of the dominant oak trees in the plot collected with a shotgun annually in late July or early August. Potential net nitrogen mineralization and nitrification were measured annually or biennially in laboratory assays using a 14-day aerobic incubation following the methods detailed in Lovett and Rueth (1999). Soils for these assays were collected in July using 8 cores per plot (2.5-cm diameter, 10-cm deep) composited into 4 samples per plot, separated into 0–3 and 3–10 cm depth increments. The top 3 cm corresponds to the average depth of the forest floor at the sites; thus the 0–3 cm sample approximates the organic horizon and the 3–10 cm sample represents the top 7 cm of the mineral soil (typically a B horizon). We separated the cores by measured depth rather than by soil horizon to provide more consistent sampling from year to year, given that multiple individuals sampled the plots over the course of this long-term study. Soils were sieved through an 8-mm screen and incubated at constant temperature (20°C) and moisture (60% of field capacity). Soils were extracted at the beginning and end of the incubations using 2 M KCl (10 g moist soil/100 ml of KCl). Soil solution was collected using low-tension soil lysimeters installed in the lower B horizon (average depth $40 cm), below the predominant rooting zone. A constant tension of 0.1 atm (10 kPa) was maintained on the lysimeters using a hanging column system as described by Johnson and Lindberg (1992). Our experience in these sites indicates that zero-tension lysimeters collect soil water unreliably, and that these constant, low-tension lysimeters provide a reasonable compromise between collection efficiency and obtaining a fraction of soil water that represents potential leachate. Samples were collected bi-weekly, stored at 4o C with chloroform as a preservative, and pooled into monthly samples for analysis. Leaching N fluxes were estimated from monthly concentrations and monthly water flow estimated from the Brook90 hydrologic model (Federer 1995), parameterized with data from the Cary Institute weather station approximately 800 m from the plots. Pools of C and N in the soil were measured using quantitative excavation of soil pits (20 9 20 cm, 4 pits per plot) in 2006.
A New Conceptual Model of Nitrogen Saturation 619
Tree growth was measured by repeated diameter measurements of all trees larger than 5 cm diameter on every plot. Wood N concentration was measured on tree core samples taken in 2004 (Wallace and others 2007). Percent C and N were analyzed on segments of the cores representing the pre-treatment period (1989–1995) and the treatment period (1996–2004).
Chemical Analysis
Soil solution nitrate was measured using a Dionex ICS-2000 ion chromatograph, soil solution NH4+ and soil extract NH4+ and NO3- were measured with a Lachat QuickChem 8000 FIA Ion Analyzer, NH4+ using the phenate method, and NO3- using the cadmium reduction method. Soil solution total N was measured as NO3- on the Lachat Ion Analyzer after persulfate oxidation. Carbon and N concentrations in foliage, wood, and soil were measured on dried and ground samples using a CE Elantech Flash EA1112 Elemental Analyzer.
RESULTS
The patterns of response of this forest to long-term N addition (Figure 2) differ in some important ways from those hypothesized in the existing conceptual model of N saturation (Figure 1). To facilitate comparison with the hypotheses, we plot the temporal trends in key N cycling variables in Figure 2 as ratios of treatment plots/control plots. The N concentration in the foliage increased shortly after the experiment began as predicted by Figure 1, but the increase was only about 10–20% above control plot values and was essentially flat after the second year of treatment, with a decline in 2005, the last year for which we have measurements (Figures 2A, 3). This stands in contrast to Figure 1, in which foliar N is hypothesized to continue rising and nearly double in concentration. The hypotheses diagrammed in Figure 1 indicate that the N mineralization rates should increase initially and then decrease, whereas in our study the potential N mineralization showed no trend through the study (Figure 2B). The mean N mineralization rate in the treated plots was about 50% higher than in the control plots, but that difference existed before the treatments began (summer 1996), and the treated and control plots were not significantly different from one another in any year. [We note that these are net N mineralization rates, and in some cases gross N mineralization can be stimulated in N saturated sites without concomitant increases in net rates (Tietema 1998). In
Figure 2. Temporal patterns of forest response to N addition in an oak forest in southeastern New York State. Values represent the ratio of the mean of treated plots to the mean of the untreated plots. A Foliar N concentration, B potential net N mineralization, C potential net nitrification, D soil solution NO3- concentration (B horizon).
other cases, gross rates are unaffected by N additions (for example, Venterea and others 2004; Christenson and others 2009).] In Figure 1, nitrification rate is hypothesized to increase in the late stages of N saturation, followed by an increase in NO3- leaching. In our study, potential nitrification rate in the N-treated plots was significantly elevated over the control plots by year 6 (2002) (Figures 2C, 4). However, NO3- leaching, expressed as concentration of NO3- in soil solution, increased well before nitrification did. Nitrate leaching in treated plots was significantly higher than controls in the second year of treatment and
620 G. M. Lovett and C. L. Goodale
Figure 3. Nitrogen concentration (% dry mass) measured in sunlit oak foliage on the study plots. NS not significant, **P < 0.01, ***P < 0.001. rose to almost 150-fold greater than the control by year 3 (1999) (Figure 2D). Estimated N (inorganic + organic) leaching fluxes reached almost 50 kg N ha-1 y-1 in 1999 (Figure 5), after which the N application rate was reduced from 100 to 50 kg N ha-1 y-1. Leaching fluxes remained at high levels (10–40 kg N ha-1 y-1), though they varied from year to year (Figure 5), primarily depending on water flow through the soil.
The control plots showed very little N leaching and retained 88% of the approximately 9 kg N ha-1 y-1 added to the site from ambient N deposition. By contrast, the treated plots retained only 65% of N inputs from deposition and the N amendments (Table 1). Ammonium and dissolved organic N (DON) were a significant fraction of the total N leaching in the control plots (25% and 38%, respectively) but were relatively much less important (2 and 6%, respectively) in the treated plots because of the elevated NO3- concentrations (Table 1). Mass balance calculations of N inputs, hydrologic losses, and tree N accumulation indicate that about 55% of the added N was retained in the soil in the treated plots during the 8 years of the study. (This budget assumes that N gas losses (NH3, NO, N2O, N2) are negligible in these acidic and wellaerated soils. We are in the process of testing this assumption.) In the untreated plots, however, the measured atmospheric deposition was insufficient to supply even the N accumulated in the wood over the study period, indicating that N was probably extracted from the soil N pool by the plants. Our best estimate is that this soil N ‘‘mining’’ activity totaled about 15 kg N ha-1 over the 8 years (Table 1).
Although these N budget calculations indicate incorporation of N into the soil in the treated plots and mining of N in the untreated plots, we observed no significant treatment effect on soil %N, C:N ratio or soil C or N pools (Table 2), probably because the soil pools are large and spatially variable, with expected changes in these pools
Figure 4. Mean rates of potential net N mineralization (Nmin) and nitrification measured in the study plots in the 0–3 cm depth samples. Error bars are standard errors. ND no data. For the 3–10 cm depth, the patterns were similar but the rates were lower (data not shown).
Figure 5. N leaching flux (inorganic + organic N) in treated and control plots based on measured soil solution concentrations and modeled water flows. Solid line shows rate of fertilizer addition to the treated plots. All plots also received an average of 8.6 kg N ha-1 y-1 from wet + dry atmospheric deposition.
A New Conceptual Model of Nitrogen Saturation 621
Table 1. Plot N Budgets for the Period 1997–2004
Inputs
Atmospheric deposition (wet + dry)
Fertilizer N added
Change in pools
N increment in wood (aboveground) Estimated N increment in soil1
0–3 cm
Outputs N Leaching
3–10 cm
NH4+ NO3-
DON
Total
Sum inputs
Sum outputs + pool increments
Difference (gaseous loss or change in subsoil pool)
Retention % (= 100*(inputs - outputs)/inputs)
Treated
69 550
64
26 83
4 200
12 216 619 389 230 65%
Untreated
69 0
76
0 0 2 3 3 8 69 84 -15 88%
Values are cumulative means of treated and untreated plots, in kg N ha-1 for the entire 8-year period. 1Soil N pools were not measured in 1996. These calculations assume that the soil N increment was 0 in the untreated plots, and that the N pool in the untreated plots represents
the pre-treatment (1996) pool in the treated plots. Thus the difference between the treated and untreated plots is the estimate of the N accumulation in the soil of the treated
plots.
Table 2. Concentrations and Pools of C and N in Wood and Soils
Treated (mean (SE))
Wood
%N (1997–2004) N pool 2004 (kg N ha-1) C pool 2004 (kg C ha-1) N increment (kg N ha-1 y-1) C increment (kg C ha-1 y-1)
Soil (0–3 cm)
%N
%C
C:N (mass) N pool (kg N ha-1) C pool (kg C ha-1)
Soil (3–10 cm)
%N
%C
C:N (mass) N pool (kg N ha-1) C pool (kg C ha-1)
Soil total (0–10 cm) N pool (kg N ha-1) C pool (kg C ha-1)
0.30 (0.03)1 409.8 (98.8) 98,390 (28,609)
8.0 (1.5) 1359 (285)
0.83 (0.09) 15.8 (1.8) 18.9 (0.3) 921 (48) 17,356 (820)
0.37 (0.04) 6.6 (0.6)
18.1 (0.4) 1297 (81) 23,442 (1309)
2219 (119) 40,798 (1876)
Untreated (mean (SE))
0.22 (0.01)1 543.6 (105.6) 123,961 (17,180)
9.5 (1.5) 1891 (251)
0.73 (0.03) 14.3 (1.0) 19.6 (0.9) 896 (67) 17,319 (921)
0.37 (0.03) 7.0 (0.4)
19.0 (0.6) 1214 (90) 22,827 (120)
2109 (152) 40,145 (1908)
Values are means of 6 plots, with standard error in parentheses. %N in wood is from wood accrued during the N treatment period. N and C increments in wood are based on
sampling in 1996 and 2004. Wood pools and increments include both live and standing dead trees on the plot, and include only aboveground wood. Soil depth intervals are measured from the surface; 0–3 cm approximates the O horizon, and 3–10 cm approximates the top 7 cm of the mineral horizon. 1%N in wood was significantly higher in treated plots than in untreated plots (paired t-test, P = 0.015); for the other variables in the table there was no significant treatment effect.
smaller than the precision with which they can be measured. We did observe a significant increase in wood N concentration in the treated plots (Table 2). However, because of the tree mortality in the treated plots (see Wallace and others 2007 and
discussion below), the increased N concentration did not result in a significantly elevated N pool in wood (live + standing dead) in the treated plots (Table 2). Similarly, C pools in wood did not differ significantly between treatments.
622 G. M. Lovett and C. L. Goodale
DISCUSSION
The data from this site do not entirely support the general conceptual model diagrammed in Figure 1, in which enhanced N deposition is taken up by N-limited plants and incorporated into their tissues, and then transferred to soil organic matter as plant litter, stimulating mineralization, nitrification, and then N leaching. We suggest that the amount of N that can flow through this plant–litter–soil pathway in mature forests is often quite limited because of both the restricted stoichiometric plasticity of most plants, and constraints on productivity imposed by stand age and the availability of light, water and other nutrients. The limited stoichiometric plasticity of these trees is illustrated by the foliar N concentration, which increased by a maximum of 20%, and then did not rise further, despite continued N additions. Other N addition studies in the U.S. and Europe have reported similar levels of foliar N enrichment (Gundersen and others 1998; Magill and others 2004), but some N-amended pine and spruce stands in the eastern U.S. have shown foliar N concentrations up to 100% greater than controls (Magill and others 2004, McNulty and others 2005).
The reduction in N retention efficiency in the Ntreated plots indicates that the N retention mechanism in the soil was overwhelmed or inhibited by the N additions. Nitrogen retention in forest soils may occur abiotically or microbially (Johnson and others 2000; Dail and others 2001; Fitzhugh and others 2003). Microbial immobilization of added N may be low in soils with a C:N ratio below about 20, assuming that microbes are about 40–50% efficient and have C:N ratios of around 8–10 (Cleveland and Liptzin 2007). Moreover, microbial immobilization of NO3- can be suppressed by NH4+ additions (Betlach and others 1981) and this has been suggested as the mechanism by which N leaching has responded quickly to N treatments in other field experiments (Moldan and others 2006; Emmett 2007).
After 2001, several of the treated plots showed elevated tree mortality, primarily of red oaks, and most likely associated with soil acidification due to the NO3- leaching (Wallace and others 2007). This onset of ecosystem decline is consistent with stage 3 of the Aber and others (1998) N saturation model, and has been observed in some N-amended conifer stands in the eastern U.S. (Magill and others 2004; McNulty and others 2005) but had not previously been observed in hardwood stands. We do not believe that the decline in tree vigor precipitated the increase in NO3- leaching, for two reasons.
First, elevated NO3- leaching began in year 2 of the experiment, well before the tree mortality was observed (Figure 5). Second, the capacity for wood N storage by healthy trees (estimated from the untreated plots) could account for only about 12% of the N inputs to the treated plots (Table 1), thus reduction in tree uptake would have made only a minor difference in N leaching.
Other lines of evidence also suggest inconsistencies between the conceptual model illustrated in Figure 1 and experimental results. For instance, field studies in which a 15N tracer is added to forest plots consistently show that the forest floor, rather than the vegetation, is the strongest initial sink for the added N (for example, Nadelhoffer and others 1999; Templer and others 2005). Experimental N addition and exclusion studies have shown that N leaching is often one of the first variables to respond to N manipulations, rather than the last (Emmett 2007), and cross-site comparisons have shown N leaching to be one of the most sensitive response variables across gradients of N deposition (for example, Aber and others 2003). Aber and others (1998) discussed the importance of soil N retention in controlling N leaching and evaluated several possible mechanisms by which N could be incorporated into SOM without first flowing through plant biomass, but these mechanisms are not included in the dynamic responses illustrated in Figure 1.
Rethinking the Conceptual Model
The conceptual model illustrated in Figure 1 focuses on the temporal dynamics of N saturation, particularly the sequential passage of N from plant to litter to soil. Although this conceptual view of the process may be sufficient in some cases, we do not believe it is general enough to explain the results of our study or other recent experimental studies. Our data show that the added N can reach all three measured fates (vegetation, soils, and leaching) simultaneously (we do not have data on gaseous losses). This makes it very difficult to predict temporal patterns of response of various parts of the system (as in Figure 1), because the patterns depend on the relative strength of the sinks, rather than the timing of transfer between one sink and the next. To move our predictive capacity forward, we need to consider the sink strengths within the context of the underlying mass balance of N in the forest ecosystem, as expressed in equation (1).
Our conceptual model of the N saturation process is diagrammed in Figure 6, which shows that N added to a forest ecosystem can flow simulta-
A New Conceptual Model of Nitrogen Saturation 623
Figure 6. Conceptual model of N saturation in an ecosystem, illustrating N flow to various fates (internal sinks and losses from the ecosystem), examples of controlling factors, and ecosystem responses associated with N flow to those fates. Detritus includes the forest floor and woody debris, and SOM represents soil organic matter.
neously to four possible fates: incorporation into vegetation, incorporation into to the detritus and soil organic matter, leaching loss, or gaseous loss. The rate of movement to any of these fates is variable among ecosystems and controlled by properties of the ecosystem, some of which are well understood based on decades of research on the N cycle, and others of which are poorly understood.
In contrast to Figure 1, this diagrammatic model emphasizes (1) the multiple possible fates of N in the system, (2) the controls on movement of N to these fates, and (3) the variable ecosystem responses depending on which sinks or loss processes receive the added N. The temporal pattern of response of the system to N saturation (of which Figure 1 represents one possible example) will depend on the extent to which N flows to these various fates, each of which is associated with specific responses (Figure 6). For example, nitrate leaching can cause soil acidification, root damage, and tree mortality, and can lead to species change in favor of species tolerant of acid conditions (McNulty and others 2005; Hogberg and others 2006; Wallace and others 2007). Nitrogen accumulation in the forest floor and soil can reduce the activity of some microbial enzymes used in decomposition and N mineralization (Carreiro and others 2000; DeForest and others 2004; Sinsabaugh and others 2005), decrease decomposition rates of late-stage or lignin-rich litter or older soil organic matter (Berg 2000; Neff and others 2002; Knorr
and others 2005), cause an increase in soil carbon stocks (Hyvonen and others 2008; Pregitzer and others 2008; Nave and others 2009; Liu and Greaver 2010), and either increase or decrease the C:N ratio of the forest floor, depending on the balance between C accumulation and N accumulation. Nitrogen accumulation in biomass may lead to increased productivity, increased herbivory, increased N cycling from plant to forest floor, altered root:shoot ratios, and species change in favor of nitrophilic species (for example, Stevens and others 2004; Throop and Lerdau 2004; Hyvonen and others 2008; Thomas and others 2010). Species changes are likely to be most evident in ecosystems that have plants with short generation times (for example grasslands, Stevens and others 2004), whereas in forests such species changes may initially be most evident in herbaceous or understory species rather than the more long-lived trees (Gilliam 2006). At any particular site, whether and when specific effects are manifested depends on the movement of N into the pathway that causes those effects. This conceptual model, though very simple, explicitly incorporates the known sinks and losses for N, allowing better explanation of recent experimental evidence and providing a focus for future research. Expressing N saturation effects in this way allows us to focus on when and why systems may differ in their response to N additions, rather than expecting a uniform pattern of temporal response as suggested by Figure 1.
624 G. M. Lovett and C. L. Goodale
Estimating Sink Strengths
Although the sink strengths shown in Figure 6 are simple to diagram, they can be difficult to measure. Leaching losses are typically measured either by sampling soil solution chemistry and estimating water flow below the rooting zone with a hydrologic model, or by using stream water flux in gauged small watersheds (Lajtha 2000). Gaseous losses are typically measured in small chambers placed on the soil, but the rates are highly uncertain because of high spatial and temporal variability in the nitrification and denitrification processes that produce the gases, and the difficulty in measuring N2 flux, which can be a major component of the total gaseous flux (Groffman and others 2006; Kulkarni and others 2008).
The vegetation and soil sinks represent the accumulation of N within the system, and each can be thought of as having two components: (1) a carbon accumulation sink, in which N is accumulated with C in a constant C:N ratio, and (2) a stoichiometric sink, in which N is accumulated without corresponding C accumulation by changing the C:N ratio of the organic matter. The two types of sinks can occur separately or together. In combination, the C accumulation sink and the stoichiometric sink form the total capacity for N accumulation within the vegetation or soil. The well-known hypothesis put forth by Vitousek and Reiners (1975) predicting that N export should be low in ecosystems with high net ecosystem production, is essentially a statement about the effectiveness of the C accumulation sink for N in ecosystems.
N Sink in Vegetation
The N sink associated with C accumulation in forest vegetation can be quite variable depending on species composition and growth rate of the stand. The untreated oak forest in our study accumulated N in woody biomass at a rate of 9.5 kg N ha-1 y-1 (Table 1). A similar oak stand at Harvard Forest in central Massachusetts accumulated woody biomass N at a rate of 7.1 kg N ha-1 y-1 over a 9-year period, whereas a nearby pine stand accumulated only 1.9 kg N ha-1 y-1 (Magill and others 2000). Forests that are growing rapidly can be expected to have a substantial C accumulation sink for N. For example, a young pine plantation in North Carolina accumulated N at a rate of 5–23 kg N ha-1 y-1, depending on the year, with a 4-year average of 15 kg N ha-1 y-1 (Finzi and others 2002). Reiners (1992) reported that in the first 10 years of regrowth of a northern hardwood forest at Hubbard Brook, NH, following experimental deforestation, the forest
accumulated N in biomass at a rate of 9.2 kg N ha-1 y-1. In contrast, in a nearby 90-year-old forest at Hubbard Brook, the C accumulation sink for N is essentially zero because there has been no measurable C accumulation in aboveground biomass for approximately the past 25 years (Fahey and others 2005).
If forest growth is limited by N, the sink for N can be expected to increase with N fertilization. For example, in the Harvard Forest oak stand amended with 150 kg N ha-1 y-1, the N accumulation increased to 13 kg N ha-1 y-1 (compared to 7.1 in the control stand) due to both increased C accumulation and increased N concentration in wood (Magill and others 2000). However, the N accumulation in the N-amended pine stand decreased to 1.1 kg N ha-1 y-1 (compared to 1.9 in the control pine stand) due to mortality of the trees (Magill and others 2000), similar to the decline seen in the present study (Table 1). Thomas and others (2010) report that N deposition enhances aboveground C accumulation in Northeastern U.S. forests by an average of 61 kg C ha-1 y-1 for each 1 kg ha-1 y-1 increase in N deposition rate. Thus an increase in N deposition of 8 kg N ha-1 y-1 (for example, from pre-industrial times to present in this area) would increase C sequestration by 488 kg C ha-1 y-1, and at a C:N ratio of 250 (typical for woody biomass) this would correspond to an enhanced N sink of about 2 kg N ha-1 y-1.
Vegetation can also produce a stoichiometric sink for N if plant tissue N concentrations change, for example, if they increase with N fertilization. However, the limited ability of plant tissues to use or absorb excess N may make this sink small and short-lived. The stoichiometric component of the sink strength (N accumulation rate) would decline to near zero when the concentration in tissues stabilizes, as occurred in the foliage of our study by year 3 of the experiment. Increases in wood and root N concentration can also provide a stoichiometric sink.
N Sink in Detritus and Soil
The forest floor and mineral soil can also have both C accumulation and stoichiometric sinks. A forest stand that is recovering from a disturbance such as harvesting or fire could have a significant N sink associated with the accumulation of C in the forest floor. For instance, in a century of forest regrowth after agricultural abandonment in Rhode Island, USA, Hooker and Compton (2003) reported a relatively rapid accumulation of N in the forest floor (11.6 kg N ha-1 y-1). However, this forest floor
Ó 2011 Springer Science+Business Media, LLC
A New Conceptual Model of Nitrogen Saturation Based on Experimental Nitrogen Addition to an Oak Forest
Gary M. Lovett1* and Christine L. Goodale2
1Cary Institute of Ecosystem Studies, Box AB, Millbrook, New York 12545, USA; 2Department of Ecology & Evolutionary Biology, Cornell University, E215 Corson Hall, Ithaca, New York 14853, USA
ABSTRACT
The dominant conceptual model of nitrogen (N) saturation in forests predicts the temporal patterns of key N cycling indicators as an initially N-limited forest is progressively enriched in N. We present the results from a long-term N addition experiment in an oak forest in southeastern New York State, USA, which do not conform to the predictions of the conceptual model in several ways. In contrast to the predictions of the conceptual model, the foliar N concentrations in the N-treated stands of our study increased to about 20% above the levels in the control stands and then remained essentially constant, and nitrogen leaching from the treated stands increased almost immediately after the start of the experiment, prior to the onset of elevated nitrification. Concentrations of N in soil solution of the Ntreated stands peaked at over 150-fold greater than the concentrations in the control stands. There were no significant changes in potential net N mineralization. Tree mortality increased in the treated stands, but the tree mortality did not appear to be the primary cause of the excess nitrate leaching. Based on these results and those of other recent studies, we present a new conceptual model of the N saturation process focused on the mass balance of N rather than the temporal dynamics of N cycling indicators. The
mass balance is characterized by inputs of N from atmospheric deposition and fertilization, internal sinks in the vegetation and soils, and outputs to leaching and gaseous losses. The key points of the conceptual model are (1) added N can flow simultaneously to all sinks and losses in the system, (2) the fate of the added N and the temporal patterns of flow of N depend on the strength of the sinks and the factors that control them, and (3) the movement of N to the various sinks determines how N saturation is manifested in the ecosystem. We distinguish capacity N saturation, in which the sinks in the vegetation and soil are zero or negative, from kinetic N saturation, in which the sinks are positive but lower than the N input rate. The sink strengths in the vegetation and soil have two components, one due to carbon (C) accumulation in the system and the other due to change in the stoichiometry (C:N ratio) of the pool. Further work quantifying the magnitudes and controlling factors for the N sinks will allow better prediction of the dynamics of N saturation in different types of forested ecosystems.
Key words: nitrogen; saturation; forest; leaching; oak; nitrification; nitrogen mineralization; foliar nitrogen; fertilization.
Received 29 April 2010; accepted 6 February 2011; published online 19 March 2011
Author Contributions: Lovett designed and executed the field study. Lovett and Goodale collaborated on the development of the conceptual model and the preparation of the paper.
*Corresponding author; e-mail: [email protected]
INTRODUCTION
Elevated levels of atmospheric nitrogen (N) deposition are a problem throughout the developed world, and are increasingly becoming a problem in
615
616 G. M. Lovett and C. L. Goodale
developing countries (Galloway and Cowling 2002). Atmospheric N compounds emanating from both fossil fuel combustion and agricultural activities are transported downwind and may be deposited on ecosystems far from the source of emission (Vitousek and others 1997). This excess N deposition may have a wide range of effects on ecosystems, including: fertilization effects, nutrient imbalances, accumulation of N in plants, soils and sediments, enhanced losses of gaseous nitrogen oxides, and acidification of soils and surface waters caused by leaching of nitrate (Fenn and others 1998; Driscoll and others 2003).
Nitrogen saturation is the term used to describe the changes that occur in a terrestrial ecosystem as it is continually exposed to elevated levels of N deposition (Agren and Bosatta 1988; Aber and others 1989). The dominant conceptual model of N saturation was proposed by Aber and colleagues (Aber and others 1989, 1998) as a hypothetical series of changes that would occur as an N-limited ecosystem accumulates deposited N (Figure 1). In this model, the ecosystem is viewed as progressing through a series of stages of N saturation, in which stage 0 is a system that is strongly N limited, stage 1 is characterized by changes indicating increasing N sufficiency, stage 2 represents the initial symptoms of N saturation such as elevated N leaching, and stage 3 represents N-induced decline in which the productivity of the forest decreases and nitrate leaching accelerates (Aber and others 1998). As this sequence progresses, enhanced N deposition makes N available to N-limited plants, which enriches plant tissues and litter, the litter N is transferred to soil organic matter, stimulating N mineralization and nitrification, and eventually resulting in elevated nitrate leaching from the ecosystem. The authors hypothesize a strong, continual increase in
Figure 1. Aber and others’ (1998) hypothesized patterns of response of forest ecosystem properties to continuing N additions. Used with permission from Aber and others (1998).
foliar N concentration through the progression of these stages, leading to an increase in net primary productivity (NPP) of the N-limited system, an initial increase and then a decrease in N mineralization, and decreasing ratios of calcium:aluminum and magnesium:nitrogen in the foliage. In stage 3, the system undergoes a decline in NPP and an increase in nitrate leaching. Though not explicitly diagrammed in Figure 1, Aber and others (1998) recognized that the major sink for added N in forest ecosystems is in the soil, and they described several potential mechanisms by which N could be incorporated in soil organic matter.
The fate of N added to an ecosystem is constrained by the N mass balance as expressed in the following equation
I À dNv=dt À dNs=dt ¼ L þ G
ð1Þ
where I = the N input rate to the system (deposition plus N amendments), dNv/dt is the vegetation sink rate (the net rate of incorporation of N into vegetation), dNs/dt is the soil sink rate (the net rate of incorporation of N into detritus and soil), L is the rate of leaching loss, and G is the rate of gaseous loss of N.
It is important to note that the mass balance in equation (1) is expressed in terms of rates, and that losses from the ecosystem can occur when the rate of N addition exceeds the rate at which N can be incorporated into the vegetation and soil sinks, even if the capacity of those sinks is not saturated. This suggests that we should distinguish kinetic N saturation, in which the rate of N input exceeds the vegetation and soil N net sink rates, from capacity N saturation, in which N has accumulated in the vegetation and soils to the point that further net N retention in these sinks does not occur. In kinetic N saturation, the vegetation and soil sinks are still active, though N losses will occur simultaneously with N retention. In capacity saturation, N retention in the vegetation and soils is negligible and system inputs should equal outputs. Both types of saturation are related to the accumulation of N in organic matter in the ecosystem. Capacity saturation occurs when the accumulation of N in the ecosystem is zero (or negative, in the case of a deteriorating system), and kinetic saturation occurs when the accumulation rate is positive but less than the N input rate. Capacity saturation is rarely observed in real forests unless disturbance reduces the vegetation sink rate or turns the vegetation into a net source of N within the system.
Ecosystem sinks for N are associated with a specific time frame, either implicitly or explicitly, and
A New Conceptual Model of Nitrogen Saturation 617
the time frame determines which processes are important. For instance, incorporation of N into microbial biomass which then dies and is remineralized over several days, may represent a sink or source of N on an hourly or daily time scale, but does not contribute to N sinks evaluated over annual time scales. For the purposes of this paper, we define the time scales of interest for the vegetation and soil sinks in equation (1) to be on the order of years to decades or longer, because this scale represents a policy-relevant time frame for ecosystem sequestration of pollution-derived N.
The responses of forests to elevated rates of N deposition can be evaluated by several approaches, each of which has strengths and weaknesses. The most straightforward approach is long-term observation of forests subjected to N deposition (for example, Peterjohn and others 1996). However, this requires decades of measurements and may be confounded by simultaneous changes in other factors affecting the forests, such as climate or disturbance. A second approach is comparative or gradient studies, in which ecosystem properties are measured in ecosystems receiving different rates of N deposition (for example, McNulty and others 1991; Lovett and Rueth 1999; Boggs and others 2005; Thomas and others 2010). This eliminates the difficulties of long-term studies but introduces uncertainty caused by potential differences in the ecosystems caused by factors other than the rate of N deposition. In gradient studies, inferring the temporal sequence requires accepting the assumptions of the space-for-time substitution, that is, that processes that occur first in a temporal sequence also occur first along a gradient from low to high N deposition. A third approach is experimental studies, in which the rates of N deposition are enhanced by adding N to experimental plots (for example, Magill and others 1996, 2004; Wallace and others 2007). Careful pairing of experimental and control plots can minimize the influence of factors other than the level of N addition. However, because the experimental N addition rate is higher than ambient rates, these experiments run the risk of altering the processes controlling the fate of N in the ecosystem. In particular, experiments enhance kinetic saturation by increasing N input rates without necessarily increasing the N sink strengths in the ecosystem. These approaches can also be used in combination, for example, long-term experiments and experiments done across gradients.
The Aber and others’ (1998) conceptual diagram (Figure 1) focuses on which ecosystem properties are expected to respond to elevated N deposition, and the temporal sequence in which they respond.
All three types of studies can be used to evaluate this conceptual model. Aber and others compared their conceptual model with results from several experimental studies in the eastern U.S. including studies in Massachusetts (Magill and others 2004), Vermont (McNulty and others 1996) and Maine (Magill and others 1996) as well as a gradient study in spruce-fir forests across northern New York and New England (McNulty and others 1991). In general the experimental responses agreed with the hypothesized trends, if one assumes that forest ecosystems start at different places along the N saturation continuum and may proceed at different rates, depending on the degree of initial N limitation induced mainly by past land use (Aber and others 1998).
However, not all studies have supported the hypotheses diagrammed in Figure 1. For example, in experimental studies in Europe at sites in which N was either experimentally added or removed, nitrogen leaching was often the first variable to respond to the treatment, rather than the last as suggested by Figure 1 (Emmett 2007). Recent studies across N deposition gradients in the eastern and midwestern U.S. have shown variable concordance with the hypotheses illustrated in Figure 1. Tree growth increases with N deposition for some tree species but not others (Thomas and others 2010). Foliar nitrogen sometimes increases along an N deposition gradient (Pardo and others 2006; McNeil and others 2007) and sometimes does not (Aber and others 2003). Likewise, net N mineralization and nitrification sometimes increase along a nitrogen deposition gradient (McNulty and others 1991; Aber and others 2003) and sometimes do not (Lovett and Rueth 1999). Across the Northeastern U.S., N leaching tends to be low in watersheds that receive less than about 8 kg N ha-1 y-1 of atmospheric N deposition (Aber and others 2003). For deposition rates above 8 kg N ha-1 y-1, some sites show N leaching and some do not (Aber and others 2003). Whether these watersheds leach N or not is controlled in part by tree species composition, soil C:N ratio, and history of disturbance (Goodale and others 2000; Lovett and others 2002). It is often difficult to determine if the absence of trends in these gradient studies results from a true lack of response in the system, or an inability to discern the signal within the noise of spatial variation inherent in gradient studies. Overall, however, the literature suggests substantial variation in the degree to which forest ecosystems conform to the hypotheses expressed in Figure 1.
In this study, we tested the hypothetical series of changes proposed by Aber and others (1998) with
618 G. M. Lovett and C. L. Goodale
an N addition experiment in a mixed-oak forest in southeastern New York State, USA. The research site had similar vegetation to the Harvard Forest (Massachusetts) oak site used as one of the test sites for the conceptual model by Aber and others (1998) but was in a ridgetop rather than a valley location. Our goal was to determine if the temporal sequence of changes in key N cycling variables matched those illustrated in Figure 1, and if not, to propose a different conceptual model in better agreement with the observations.
METHODS
Research site
The research site is located at the Cary Institute of Ecosystem Studies (41.797°N, 73.734°W) in the Hudson Valley of southeastern New York State, USA. The forest is an upland mixed-oak woods dominated by red oak (Quercus rubra L.), chestnut oak (Quercus prinus L.) and several species of hickory (Carya sp.) regenerated from selective cutting associated with use as a woodlot early in the twentieth century (Glitzenstein and others 1990). The annual average temperature at the site is 9.6°C and precipitation averages 111 cm y-1. The bedrock is mostly shale and slate and soils are predominantly thin and well-drained silt loams of the Nassau and Woodlawn series (Glitzenstein and others 1990) classified as lithic dystrudepts. Mean annual precipitation pH at the site ranged from 4.0 to 4.4 between 1984 and 2004 (Kelly and others 2002). Ambient inorganic N deposition (wet + dry) in the period of study (1996–2006) averaged approximately 9 kg N ha-1 y-1. Wet deposition was collected continuously with a wet-only precipitation collector, and dry deposition was estimated from continuous filter-pack measurements of air concentrations of HNO3 vapor and particulate NO3- and NH4+, applying a dry deposition model parameterized with meteorological data at the site (Kelly and others 2002).
We chose 6 pairs of plots, each 20 m in diameter. The two plots of each pair are within 40 m of each other. Pre-treatment measurements were made in July–August 1996, and the N addition began in November 1996. One plot of each pair was treated with granular NH4NO3 fertilizer at a rate equivalent of 100 kg N ha-1 y-1 from 1996 to 1999, and then the N addition rate was decreased to 50 kg N ha-1 y-1 from 2000 to 2006. This annual application was divided into 4 equal doses each year. Fertilizer was applied to the forest floor; although application to the canopy would have
been more realistic, it would be technically much more difficult, and prior data from these forests indicate that canopy uptake of ambient N deposition is negligible (Lovett and others 2000a).
Sample Collection and Processing
Foliar chemistry was monitored on full-sun leaves of the dominant oak trees in the plot collected with a shotgun annually in late July or early August. Potential net nitrogen mineralization and nitrification were measured annually or biennially in laboratory assays using a 14-day aerobic incubation following the methods detailed in Lovett and Rueth (1999). Soils for these assays were collected in July using 8 cores per plot (2.5-cm diameter, 10-cm deep) composited into 4 samples per plot, separated into 0–3 and 3–10 cm depth increments. The top 3 cm corresponds to the average depth of the forest floor at the sites; thus the 0–3 cm sample approximates the organic horizon and the 3–10 cm sample represents the top 7 cm of the mineral soil (typically a B horizon). We separated the cores by measured depth rather than by soil horizon to provide more consistent sampling from year to year, given that multiple individuals sampled the plots over the course of this long-term study. Soils were sieved through an 8-mm screen and incubated at constant temperature (20°C) and moisture (60% of field capacity). Soils were extracted at the beginning and end of the incubations using 2 M KCl (10 g moist soil/100 ml of KCl). Soil solution was collected using low-tension soil lysimeters installed in the lower B horizon (average depth $40 cm), below the predominant rooting zone. A constant tension of 0.1 atm (10 kPa) was maintained on the lysimeters using a hanging column system as described by Johnson and Lindberg (1992). Our experience in these sites indicates that zero-tension lysimeters collect soil water unreliably, and that these constant, low-tension lysimeters provide a reasonable compromise between collection efficiency and obtaining a fraction of soil water that represents potential leachate. Samples were collected bi-weekly, stored at 4o C with chloroform as a preservative, and pooled into monthly samples for analysis. Leaching N fluxes were estimated from monthly concentrations and monthly water flow estimated from the Brook90 hydrologic model (Federer 1995), parameterized with data from the Cary Institute weather station approximately 800 m from the plots. Pools of C and N in the soil were measured using quantitative excavation of soil pits (20 9 20 cm, 4 pits per plot) in 2006.
A New Conceptual Model of Nitrogen Saturation 619
Tree growth was measured by repeated diameter measurements of all trees larger than 5 cm diameter on every plot. Wood N concentration was measured on tree core samples taken in 2004 (Wallace and others 2007). Percent C and N were analyzed on segments of the cores representing the pre-treatment period (1989–1995) and the treatment period (1996–2004).
Chemical Analysis
Soil solution nitrate was measured using a Dionex ICS-2000 ion chromatograph, soil solution NH4+ and soil extract NH4+ and NO3- were measured with a Lachat QuickChem 8000 FIA Ion Analyzer, NH4+ using the phenate method, and NO3- using the cadmium reduction method. Soil solution total N was measured as NO3- on the Lachat Ion Analyzer after persulfate oxidation. Carbon and N concentrations in foliage, wood, and soil were measured on dried and ground samples using a CE Elantech Flash EA1112 Elemental Analyzer.
RESULTS
The patterns of response of this forest to long-term N addition (Figure 2) differ in some important ways from those hypothesized in the existing conceptual model of N saturation (Figure 1). To facilitate comparison with the hypotheses, we plot the temporal trends in key N cycling variables in Figure 2 as ratios of treatment plots/control plots. The N concentration in the foliage increased shortly after the experiment began as predicted by Figure 1, but the increase was only about 10–20% above control plot values and was essentially flat after the second year of treatment, with a decline in 2005, the last year for which we have measurements (Figures 2A, 3). This stands in contrast to Figure 1, in which foliar N is hypothesized to continue rising and nearly double in concentration. The hypotheses diagrammed in Figure 1 indicate that the N mineralization rates should increase initially and then decrease, whereas in our study the potential N mineralization showed no trend through the study (Figure 2B). The mean N mineralization rate in the treated plots was about 50% higher than in the control plots, but that difference existed before the treatments began (summer 1996), and the treated and control plots were not significantly different from one another in any year. [We note that these are net N mineralization rates, and in some cases gross N mineralization can be stimulated in N saturated sites without concomitant increases in net rates (Tietema 1998). In
Figure 2. Temporal patterns of forest response to N addition in an oak forest in southeastern New York State. Values represent the ratio of the mean of treated plots to the mean of the untreated plots. A Foliar N concentration, B potential net N mineralization, C potential net nitrification, D soil solution NO3- concentration (B horizon).
other cases, gross rates are unaffected by N additions (for example, Venterea and others 2004; Christenson and others 2009).] In Figure 1, nitrification rate is hypothesized to increase in the late stages of N saturation, followed by an increase in NO3- leaching. In our study, potential nitrification rate in the N-treated plots was significantly elevated over the control plots by year 6 (2002) (Figures 2C, 4). However, NO3- leaching, expressed as concentration of NO3- in soil solution, increased well before nitrification did. Nitrate leaching in treated plots was significantly higher than controls in the second year of treatment and
620 G. M. Lovett and C. L. Goodale
Figure 3. Nitrogen concentration (% dry mass) measured in sunlit oak foliage on the study plots. NS not significant, **P < 0.01, ***P < 0.001. rose to almost 150-fold greater than the control by year 3 (1999) (Figure 2D). Estimated N (inorganic + organic) leaching fluxes reached almost 50 kg N ha-1 y-1 in 1999 (Figure 5), after which the N application rate was reduced from 100 to 50 kg N ha-1 y-1. Leaching fluxes remained at high levels (10–40 kg N ha-1 y-1), though they varied from year to year (Figure 5), primarily depending on water flow through the soil.
The control plots showed very little N leaching and retained 88% of the approximately 9 kg N ha-1 y-1 added to the site from ambient N deposition. By contrast, the treated plots retained only 65% of N inputs from deposition and the N amendments (Table 1). Ammonium and dissolved organic N (DON) were a significant fraction of the total N leaching in the control plots (25% and 38%, respectively) but were relatively much less important (2 and 6%, respectively) in the treated plots because of the elevated NO3- concentrations (Table 1). Mass balance calculations of N inputs, hydrologic losses, and tree N accumulation indicate that about 55% of the added N was retained in the soil in the treated plots during the 8 years of the study. (This budget assumes that N gas losses (NH3, NO, N2O, N2) are negligible in these acidic and wellaerated soils. We are in the process of testing this assumption.) In the untreated plots, however, the measured atmospheric deposition was insufficient to supply even the N accumulated in the wood over the study period, indicating that N was probably extracted from the soil N pool by the plants. Our best estimate is that this soil N ‘‘mining’’ activity totaled about 15 kg N ha-1 over the 8 years (Table 1).
Although these N budget calculations indicate incorporation of N into the soil in the treated plots and mining of N in the untreated plots, we observed no significant treatment effect on soil %N, C:N ratio or soil C or N pools (Table 2), probably because the soil pools are large and spatially variable, with expected changes in these pools
Figure 4. Mean rates of potential net N mineralization (Nmin) and nitrification measured in the study plots in the 0–3 cm depth samples. Error bars are standard errors. ND no data. For the 3–10 cm depth, the patterns were similar but the rates were lower (data not shown).
Figure 5. N leaching flux (inorganic + organic N) in treated and control plots based on measured soil solution concentrations and modeled water flows. Solid line shows rate of fertilizer addition to the treated plots. All plots also received an average of 8.6 kg N ha-1 y-1 from wet + dry atmospheric deposition.
A New Conceptual Model of Nitrogen Saturation 621
Table 1. Plot N Budgets for the Period 1997–2004
Inputs
Atmospheric deposition (wet + dry)
Fertilizer N added
Change in pools
N increment in wood (aboveground) Estimated N increment in soil1
0–3 cm
Outputs N Leaching
3–10 cm
NH4+ NO3-
DON
Total
Sum inputs
Sum outputs + pool increments
Difference (gaseous loss or change in subsoil pool)
Retention % (= 100*(inputs - outputs)/inputs)
Treated
69 550
64
26 83
4 200
12 216 619 389 230 65%
Untreated
69 0
76
0 0 2 3 3 8 69 84 -15 88%
Values are cumulative means of treated and untreated plots, in kg N ha-1 for the entire 8-year period. 1Soil N pools were not measured in 1996. These calculations assume that the soil N increment was 0 in the untreated plots, and that the N pool in the untreated plots represents
the pre-treatment (1996) pool in the treated plots. Thus the difference between the treated and untreated plots is the estimate of the N accumulation in the soil of the treated
plots.
Table 2. Concentrations and Pools of C and N in Wood and Soils
Treated (mean (SE))
Wood
%N (1997–2004) N pool 2004 (kg N ha-1) C pool 2004 (kg C ha-1) N increment (kg N ha-1 y-1) C increment (kg C ha-1 y-1)
Soil (0–3 cm)
%N
%C
C:N (mass) N pool (kg N ha-1) C pool (kg C ha-1)
Soil (3–10 cm)
%N
%C
C:N (mass) N pool (kg N ha-1) C pool (kg C ha-1)
Soil total (0–10 cm) N pool (kg N ha-1) C pool (kg C ha-1)
0.30 (0.03)1 409.8 (98.8) 98,390 (28,609)
8.0 (1.5) 1359 (285)
0.83 (0.09) 15.8 (1.8) 18.9 (0.3) 921 (48) 17,356 (820)
0.37 (0.04) 6.6 (0.6)
18.1 (0.4) 1297 (81) 23,442 (1309)
2219 (119) 40,798 (1876)
Untreated (mean (SE))
0.22 (0.01)1 543.6 (105.6) 123,961 (17,180)
9.5 (1.5) 1891 (251)
0.73 (0.03) 14.3 (1.0) 19.6 (0.9) 896 (67) 17,319 (921)
0.37 (0.03) 7.0 (0.4)
19.0 (0.6) 1214 (90) 22,827 (120)
2109 (152) 40,145 (1908)
Values are means of 6 plots, with standard error in parentheses. %N in wood is from wood accrued during the N treatment period. N and C increments in wood are based on
sampling in 1996 and 2004. Wood pools and increments include both live and standing dead trees on the plot, and include only aboveground wood. Soil depth intervals are measured from the surface; 0–3 cm approximates the O horizon, and 3–10 cm approximates the top 7 cm of the mineral horizon. 1%N in wood was significantly higher in treated plots than in untreated plots (paired t-test, P = 0.015); for the other variables in the table there was no significant treatment effect.
smaller than the precision with which they can be measured. We did observe a significant increase in wood N concentration in the treated plots (Table 2). However, because of the tree mortality in the treated plots (see Wallace and others 2007 and
discussion below), the increased N concentration did not result in a significantly elevated N pool in wood (live + standing dead) in the treated plots (Table 2). Similarly, C pools in wood did not differ significantly between treatments.
622 G. M. Lovett and C. L. Goodale
DISCUSSION
The data from this site do not entirely support the general conceptual model diagrammed in Figure 1, in which enhanced N deposition is taken up by N-limited plants and incorporated into their tissues, and then transferred to soil organic matter as plant litter, stimulating mineralization, nitrification, and then N leaching. We suggest that the amount of N that can flow through this plant–litter–soil pathway in mature forests is often quite limited because of both the restricted stoichiometric plasticity of most plants, and constraints on productivity imposed by stand age and the availability of light, water and other nutrients. The limited stoichiometric plasticity of these trees is illustrated by the foliar N concentration, which increased by a maximum of 20%, and then did not rise further, despite continued N additions. Other N addition studies in the U.S. and Europe have reported similar levels of foliar N enrichment (Gundersen and others 1998; Magill and others 2004), but some N-amended pine and spruce stands in the eastern U.S. have shown foliar N concentrations up to 100% greater than controls (Magill and others 2004, McNulty and others 2005).
The reduction in N retention efficiency in the Ntreated plots indicates that the N retention mechanism in the soil was overwhelmed or inhibited by the N additions. Nitrogen retention in forest soils may occur abiotically or microbially (Johnson and others 2000; Dail and others 2001; Fitzhugh and others 2003). Microbial immobilization of added N may be low in soils with a C:N ratio below about 20, assuming that microbes are about 40–50% efficient and have C:N ratios of around 8–10 (Cleveland and Liptzin 2007). Moreover, microbial immobilization of NO3- can be suppressed by NH4+ additions (Betlach and others 1981) and this has been suggested as the mechanism by which N leaching has responded quickly to N treatments in other field experiments (Moldan and others 2006; Emmett 2007).
After 2001, several of the treated plots showed elevated tree mortality, primarily of red oaks, and most likely associated with soil acidification due to the NO3- leaching (Wallace and others 2007). This onset of ecosystem decline is consistent with stage 3 of the Aber and others (1998) N saturation model, and has been observed in some N-amended conifer stands in the eastern U.S. (Magill and others 2004; McNulty and others 2005) but had not previously been observed in hardwood stands. We do not believe that the decline in tree vigor precipitated the increase in NO3- leaching, for two reasons.
First, elevated NO3- leaching began in year 2 of the experiment, well before the tree mortality was observed (Figure 5). Second, the capacity for wood N storage by healthy trees (estimated from the untreated plots) could account for only about 12% of the N inputs to the treated plots (Table 1), thus reduction in tree uptake would have made only a minor difference in N leaching.
Other lines of evidence also suggest inconsistencies between the conceptual model illustrated in Figure 1 and experimental results. For instance, field studies in which a 15N tracer is added to forest plots consistently show that the forest floor, rather than the vegetation, is the strongest initial sink for the added N (for example, Nadelhoffer and others 1999; Templer and others 2005). Experimental N addition and exclusion studies have shown that N leaching is often one of the first variables to respond to N manipulations, rather than the last (Emmett 2007), and cross-site comparisons have shown N leaching to be one of the most sensitive response variables across gradients of N deposition (for example, Aber and others 2003). Aber and others (1998) discussed the importance of soil N retention in controlling N leaching and evaluated several possible mechanisms by which N could be incorporated into SOM without first flowing through plant biomass, but these mechanisms are not included in the dynamic responses illustrated in Figure 1.
Rethinking the Conceptual Model
The conceptual model illustrated in Figure 1 focuses on the temporal dynamics of N saturation, particularly the sequential passage of N from plant to litter to soil. Although this conceptual view of the process may be sufficient in some cases, we do not believe it is general enough to explain the results of our study or other recent experimental studies. Our data show that the added N can reach all three measured fates (vegetation, soils, and leaching) simultaneously (we do not have data on gaseous losses). This makes it very difficult to predict temporal patterns of response of various parts of the system (as in Figure 1), because the patterns depend on the relative strength of the sinks, rather than the timing of transfer between one sink and the next. To move our predictive capacity forward, we need to consider the sink strengths within the context of the underlying mass balance of N in the forest ecosystem, as expressed in equation (1).
Our conceptual model of the N saturation process is diagrammed in Figure 6, which shows that N added to a forest ecosystem can flow simulta-
A New Conceptual Model of Nitrogen Saturation 623
Figure 6. Conceptual model of N saturation in an ecosystem, illustrating N flow to various fates (internal sinks and losses from the ecosystem), examples of controlling factors, and ecosystem responses associated with N flow to those fates. Detritus includes the forest floor and woody debris, and SOM represents soil organic matter.
neously to four possible fates: incorporation into vegetation, incorporation into to the detritus and soil organic matter, leaching loss, or gaseous loss. The rate of movement to any of these fates is variable among ecosystems and controlled by properties of the ecosystem, some of which are well understood based on decades of research on the N cycle, and others of which are poorly understood.
In contrast to Figure 1, this diagrammatic model emphasizes (1) the multiple possible fates of N in the system, (2) the controls on movement of N to these fates, and (3) the variable ecosystem responses depending on which sinks or loss processes receive the added N. The temporal pattern of response of the system to N saturation (of which Figure 1 represents one possible example) will depend on the extent to which N flows to these various fates, each of which is associated with specific responses (Figure 6). For example, nitrate leaching can cause soil acidification, root damage, and tree mortality, and can lead to species change in favor of species tolerant of acid conditions (McNulty and others 2005; Hogberg and others 2006; Wallace and others 2007). Nitrogen accumulation in the forest floor and soil can reduce the activity of some microbial enzymes used in decomposition and N mineralization (Carreiro and others 2000; DeForest and others 2004; Sinsabaugh and others 2005), decrease decomposition rates of late-stage or lignin-rich litter or older soil organic matter (Berg 2000; Neff and others 2002; Knorr
and others 2005), cause an increase in soil carbon stocks (Hyvonen and others 2008; Pregitzer and others 2008; Nave and others 2009; Liu and Greaver 2010), and either increase or decrease the C:N ratio of the forest floor, depending on the balance between C accumulation and N accumulation. Nitrogen accumulation in biomass may lead to increased productivity, increased herbivory, increased N cycling from plant to forest floor, altered root:shoot ratios, and species change in favor of nitrophilic species (for example, Stevens and others 2004; Throop and Lerdau 2004; Hyvonen and others 2008; Thomas and others 2010). Species changes are likely to be most evident in ecosystems that have plants with short generation times (for example grasslands, Stevens and others 2004), whereas in forests such species changes may initially be most evident in herbaceous or understory species rather than the more long-lived trees (Gilliam 2006). At any particular site, whether and when specific effects are manifested depends on the movement of N into the pathway that causes those effects. This conceptual model, though very simple, explicitly incorporates the known sinks and losses for N, allowing better explanation of recent experimental evidence and providing a focus for future research. Expressing N saturation effects in this way allows us to focus on when and why systems may differ in their response to N additions, rather than expecting a uniform pattern of temporal response as suggested by Figure 1.
624 G. M. Lovett and C. L. Goodale
Estimating Sink Strengths
Although the sink strengths shown in Figure 6 are simple to diagram, they can be difficult to measure. Leaching losses are typically measured either by sampling soil solution chemistry and estimating water flow below the rooting zone with a hydrologic model, or by using stream water flux in gauged small watersheds (Lajtha 2000). Gaseous losses are typically measured in small chambers placed on the soil, but the rates are highly uncertain because of high spatial and temporal variability in the nitrification and denitrification processes that produce the gases, and the difficulty in measuring N2 flux, which can be a major component of the total gaseous flux (Groffman and others 2006; Kulkarni and others 2008).
The vegetation and soil sinks represent the accumulation of N within the system, and each can be thought of as having two components: (1) a carbon accumulation sink, in which N is accumulated with C in a constant C:N ratio, and (2) a stoichiometric sink, in which N is accumulated without corresponding C accumulation by changing the C:N ratio of the organic matter. The two types of sinks can occur separately or together. In combination, the C accumulation sink and the stoichiometric sink form the total capacity for N accumulation within the vegetation or soil. The well-known hypothesis put forth by Vitousek and Reiners (1975) predicting that N export should be low in ecosystems with high net ecosystem production, is essentially a statement about the effectiveness of the C accumulation sink for N in ecosystems.
N Sink in Vegetation
The N sink associated with C accumulation in forest vegetation can be quite variable depending on species composition and growth rate of the stand. The untreated oak forest in our study accumulated N in woody biomass at a rate of 9.5 kg N ha-1 y-1 (Table 1). A similar oak stand at Harvard Forest in central Massachusetts accumulated woody biomass N at a rate of 7.1 kg N ha-1 y-1 over a 9-year period, whereas a nearby pine stand accumulated only 1.9 kg N ha-1 y-1 (Magill and others 2000). Forests that are growing rapidly can be expected to have a substantial C accumulation sink for N. For example, a young pine plantation in North Carolina accumulated N at a rate of 5–23 kg N ha-1 y-1, depending on the year, with a 4-year average of 15 kg N ha-1 y-1 (Finzi and others 2002). Reiners (1992) reported that in the first 10 years of regrowth of a northern hardwood forest at Hubbard Brook, NH, following experimental deforestation, the forest
accumulated N in biomass at a rate of 9.2 kg N ha-1 y-1. In contrast, in a nearby 90-year-old forest at Hubbard Brook, the C accumulation sink for N is essentially zero because there has been no measurable C accumulation in aboveground biomass for approximately the past 25 years (Fahey and others 2005).
If forest growth is limited by N, the sink for N can be expected to increase with N fertilization. For example, in the Harvard Forest oak stand amended with 150 kg N ha-1 y-1, the N accumulation increased to 13 kg N ha-1 y-1 (compared to 7.1 in the control stand) due to both increased C accumulation and increased N concentration in wood (Magill and others 2000). However, the N accumulation in the N-amended pine stand decreased to 1.1 kg N ha-1 y-1 (compared to 1.9 in the control pine stand) due to mortality of the trees (Magill and others 2000), similar to the decline seen in the present study (Table 1). Thomas and others (2010) report that N deposition enhances aboveground C accumulation in Northeastern U.S. forests by an average of 61 kg C ha-1 y-1 for each 1 kg ha-1 y-1 increase in N deposition rate. Thus an increase in N deposition of 8 kg N ha-1 y-1 (for example, from pre-industrial times to present in this area) would increase C sequestration by 488 kg C ha-1 y-1, and at a C:N ratio of 250 (typical for woody biomass) this would correspond to an enhanced N sink of about 2 kg N ha-1 y-1.
Vegetation can also produce a stoichiometric sink for N if plant tissue N concentrations change, for example, if they increase with N fertilization. However, the limited ability of plant tissues to use or absorb excess N may make this sink small and short-lived. The stoichiometric component of the sink strength (N accumulation rate) would decline to near zero when the concentration in tissues stabilizes, as occurred in the foliage of our study by year 3 of the experiment. Increases in wood and root N concentration can also provide a stoichiometric sink.
N Sink in Detritus and Soil
The forest floor and mineral soil can also have both C accumulation and stoichiometric sinks. A forest stand that is recovering from a disturbance such as harvesting or fire could have a significant N sink associated with the accumulation of C in the forest floor. For instance, in a century of forest regrowth after agricultural abandonment in Rhode Island, USA, Hooker and Compton (2003) reported a relatively rapid accumulation of N in the forest floor (11.6 kg N ha-1 y-1). However, this forest floor