Influence of Nitrogen Loading and Species
Composition on the Carbon Balance of
Grasslands
David A. Wedin* and David Tilman
In a 12-year experimental study of nitrogen (N) deposition on Minnesota grasslands, plots dominated by native warm-season grasses shifted to low-diversity mixtures dominated by cool-season grasses at all but the lowest N addition rates. This shift was associated with decreased biomass carbon (C):N ratios, increased N mineralization, increased soil nitrate, high N losses, and low C storage. In addition, plots originally dominated by nonnative cool-season grasses retained little added N and stored little C, even at low N input rates. Thus, grasslands with high N retention and C storage rates were the most vulnerable to species losses and major shifts in C and N cycling.
riumans have dramatically altered the cycling of nitrogen on Earth, doubling the natural rate of N fixation and causing atmospheric N deposition rates to increase more than tenfold over the last 40 years to current values of 0.5 to 2.5 g N m-2 year-1 in eastern North America and 0.5 to 6.0 g N m-2 year-1 in northern Europe (J). Because N is the primary nutrient limiting terrestrial plant production, N addition is causing shifts in plant species composition, decreases in species diversity, and changes in food-web structure in terrestrial ecosystems (2-5). This N-driven terrestrial eu-trophication parallels phosphorus-driven eutrophication in lakes. Increased N deposition may lead to greater C storage in soil organic matter and vegetation, thus providing a sink for C02 and potentially explaining the globally "missing C" (6). Despite this, almost no experimental data exist on
D. A. Wedin, Department of Botany, University of Toronto, Toronto, Ontario M5S 3B2, Canada D. Tilman, Ecology, Evolution and Behavior, University of Minnesota, St. Paul, MN 55103, USA.
"To whom correspondence should be addressed. E-mail; wecfinSbotarry.utoronto.ca
changes in ecosystem C in response to long-term N addition in nonagricultural ecosystems; rather, effects on C stores have been estimated from models, giving divergent predictions (6).
We present results ot 12 years of experimental N addition to 162 grassland plots in three N-limited Minnesota grasslands that varied in successional age, total soil C, and plant species composition (7, 8). The youngest field (Field A) was dominated by vegetation with the C3 photosynthetic pathway, primarily nonnative "cool-season" grasses and forbs, whereas the two older fields (Fields B and C) were dominated by native C4 "warm-season" prairie grasses. Because other potentially limiting nutrients were supplied and soil pH was controlled, our study addresses the eutrophication effects of N loading while controlling for acidification and related biogeochemical effects that might also affect natural ecosystems (9, 10).
Nitrogen loading dramatically changed plant species composition, decreased species diversity, and increased aboveground productivity in these plots (2, 7, 11). After 12
SCIENCE   •   VOL 274    •   6 DECEMBER 1996
years of N addition, species richness declined by more than 50% across the N addition gradient (Fig. 1A), with the greatest losses at 1 to 5 g N m-2 year-1—levels spanning current atmospheric deposition rates in eastern North America and northern Europe (1). This loss of diversity was accompanied by major shifts in composition, with C4 grasses (predominantly the native bunchgrass Schizachyrium scoparium) declining and the weedy Eurasian C3 grass Agropyron repens becoming dominant at high N addition rates (Fig. IB) (2, 7, 11).
As the vegetation shifted wich increasing N inputs from C4 species to C3 species, the C:N ratios of aboveground and below-ground plant tissues decreased (Fig. 1, C and D) (12). Two analyses indicate that interspecific differences in tissue chemistry together with the observed species shifts can account for most of this shift in biomass C:N ratios across the experimental N gradient. First, nitrogen-use efficiency (NUE), the ratio of plant production to N use [estimated following (13)], averaged 203 across the N addition gradient for S. scoparium (14). The high NUE of S. scopaiium and other perennial C4 grasses is well documented (/5, 16). In contrast, Poa pratensis and A. repens, the dominant C3 grasses, had mean NUE values of 107 and 78, respectively. Intraspecific plasticity for NUE—die shift in tissue chemistry within species across the N addition gradient—was small relative to the large interspecific differences among the three species (14). In addition, multiple regression showed that the best correlate, after the rate of N fertilization, for the C:N ratio of dead biomass in a plot was the S. scopaiium abundance in the plot (17).
Ac N addicion races of <5 g N m 2 year-1, soil N03- concencrarions were sig-
nificantly lower in the older fields dominated by S. scoparium (Fields B and C) than in the C3-dominated youngest field (Field A) (Fig. 2A) (18, 19). This parallels results from experimental monocultures of these praine and old-field grasses (20). SoilN03-did not respond significandy to N addition at rates <5gN m-2year_I (19), but N03-concentrations increased by a factor of ten at higher N addition rates (Fig. 2A). With the exception of two treatments in Field B, annual nee N mineralization races also showed relatively little change at low N addition rates, but increased linearly with increased N addition at races >5 g N m-2 year-1 (Fig. 2B).
At low N addition rates (1 co 2 g N m-2 year-1), che two C4-dominated fields retained approximately all of the N inputs after 12 years (Fig. 2C) (21). Nitrogen retention in these fields dropped as N addition increased, converging on an N retention of 35% of N inputs at the two highesc N addicion rates. Similar resulcs are reported for N-loading studies in European forests, where, on average, 43% of N inputs were retained ac N inputs ranging from 2.5 to 7.5 g N m-: year-1 (22) However, N retencion varied greacly from sice to sice in those studies, supporting che conclusion of Aber et al. (23) chac "N recention will vary non-linearly depending on the internal state of the system." In contrasc co the two older fields, the C3-dominated Field A re-cained essencially none of the added N at low input races (Fig. 2C) (24). Although the mechanisms of N loss in Field A are unresolved, our grassland result concrascs wich that of forest research, where early successional stands are hypothesized to have higher nutrienc retention (25, 26).
On a plot-by-plot basis, nee N losses (as g N m-2) (21) were highly correlated with
Fig. 1. Vegetation responses to 12 years of N addition. Points represent treatment means (6 replicates per N addition level, 12 for controls) for each of three fields. (A) Number of vascular plant species found in 0.3-m2 vegetation samples. (B) Biomass of grasses with the C4 photosynthetic pathway as a proportion of above-ground live biomass at mid-growing season. One species, S. scoparium, contributed >95% of the C4 biomass in the plots. Biomass C:N ratios for (C) aboveground                  N addjtion ( m.2 1} dead biomass (both recent and old) and (D) belowground biomass (live and dead).
the average growing-season concentration ofsoilNCV (Rg- 3A). SoilN03- is highly mobile, and high soil N03- concentrations frequently lead to large leaching losses of N, as presumably happened in this study (10, 27). We cannot partition N losses, however, because N leaching, ammonia volatilization, dissolved organic N losses, and deni-trification were, not measured (28).
Soil N03- concentrations were highly correlated with biomass C:N ratios (Fig. 3B) (29). A comparable relationship existed between soil N03- and the C:N racio of either belowground biomass or aboveground litter. Ac biomass C:N racios greacer than 30, soil N03- concentracions were low (<1 mg/kg). As C:N ratios dropped below 30, the immobilization sink for mineral N provided by dead organic matcer disappeared, rates of nee N mineralization increased, soil N03~ increased sharply, and overall N retencion rates decreased (Fig. 2). Thus, our results
100
10           20
N addition (g nv2 year1)
30
5       10        15       20       25 N addition (g nv2 year-1)
Fig. 2. Nitrogen dynamics after 12 years of N addition. (A) Seasonal average extractable soil NO3- concentrations, (B) annual in situ net N mineralization, and (C) net N retention after 12 years estimated as the change in total system N (relative to controls) divided by the sum of experimental N additions.
SCIENCE   •   VOL 274   •   6 DECEMBER 1996                                                                        1721
support the conclusion that microhial immobilization ot mineral N is a major factor regulating N retention (JO, 23, 25, 30).
Our arialy.se* indicate another potentially important factor regulating soil NO,-pools in these grasslands. Plant species diversity remained a significant negative cor-lelate of soil NO," in a multiple regression model that accounted tor the effects of littei C:N ratio and N addition rate (29, 31). This suggests that complementary spatial and temporal patterns of nutrient uptake associated with high plant-species diversity or functional group diversity also play a significant role in ecosystem N retention (32).
We conclude that the shift from N immobilization to mineralization, a threshold determined by microhial resource requirements and the C:N ratio of an ecosystem's detrital biomass, creates an inherent non-linearity in the response of these grasslands to chronic N loading. In our study, species shifts in the vegetation at low levels ot" N loading appear to he driving such a nonlinear response of the N cycle (4, 15). In addition to shifts in species composition,
r2 = 0.49
-400
0.01           0.1            1            10
Soil NO3-N (mg/kg)
100
B
I
I
1
Biomass C:N ratio
Fig. 3. (A) The relationship between net N losses or gains (the change in total system N minus the sum of experimental N additions) and seasonal average soil NO., concentrations in 162 experimental plots. The equation for the fitted curve (note log scale) is N,. , = (-75.56) - [94.89 :< log(NO,,j]. (B) The relationship between soil NO.,' and the C:N ratio of plant blornabb (ubovugrcund dead biomass plus belowground biomass). Vertical line represents a biomass C:N ratio of 32.
the loss of diversity, per se, during eutrophi-cation may contribute to decreased N retention in grassland ecosystems subjected to atmospheric N deposition (31).
Two patterns emerged for the net change in total ecosystem C stoies after 12 years (12, 21). First, although total C stores differed significantly among the three fields across the experimental N gradient, differences were greater at the low end of the gradient (33). At the high end of the N addition gradient, all fields were converging on total C stores of roughly 4000 to 5000 g C m_i. Second, total C stores increased significantly at low N addition rates in the C4-dominated fields (Fields B and C) hut not in the C,-dominated field. Averaging across the three lowest N addition levels (1, 2, and 3.4 g N m"2 year"1), total ecosystem C increased 21% (545 g C m"') in Field B, which had lower soil C initially, 10% (445 g C m"') in Field C, and only l',\'i (27 g C m"2) in Field A. In contrast, theoretical estimates of C storage for humid temperate grasslands in response to climate change, direct CO^ enrichment, or both range from 3% to -3% (34). Carbon storage resulting from anthropogenic N inputs, although highly dependent on grassland type, may be markedly greater than C storage in response to other components of global change.
Finally, we determined the net long-term change in total ecosystem C per unit of added N over our 12-year study. In regression analysis, there was significantly lower C storage (g C/g N) at N addition rates <5 g N m-2 year-1 for Field A than for Fields B and C, as well as a significant effect of N addition rate and a significant field-hy-N-addition interaction (Fig. 4) (35). Without field as a categorical variable, plot C, hioin.iss was the best single predictor of C storage (35).
At the lowest N addition rates (1 and 2 g N m"-' year" '), the Ľ storage rate averaged
30
■i    20
4-o
•S S  10
o
D)
E %    o
-10
•       Field  A(C3) —O—   Fields B and C (C4)
0                10                 20
N addition (g rrrz year1)
30
Fig. 4. Net C storage per unit experimentally added N after 12 years. Because C storage rates (g C/g N) did not differ significantly between Fields B and C (34), overall treatment means for the two C,-dominated fields are presented.
24.3 g C/g N (n = 24, SE = 7.6) in Fields B and C (Fig. 4). Although we know of no comparable values fiom other long-term experiments, our value of 24.3 g C/g N is low compared to most model estimates of net C storage in response to atmospheric N deposition, which range from 17 to > 100 g C/g N (6). This difference probably relates to ecosystem type. In our two Cj-dominated grasslands, 63% of the long-term C storage was in soils, which had a C:N ratio of roughly 11. Globally, woody vegetation with a higher C:N ratio becomes a more significant C sink.
Estimates of C storage in response to N loading are the product of two terms: net C storage per unit N retained and the N retention rate. In simulations with the CENTURY model of long-term C budgets for S. M.-uJHir/nm monocultures in our soils and climate, we found a long-term C storage rate of 22 g C/g N input from atmospheric deposition (36). Thus, our empirical and modeling estimates of C storage (g C/g N) were very similar for low N addition plots in Fields B and C, where N retention rares were -100%.
In contrast, the model (36) did a relatively poor job of predicting C storage rates for Fields B and C at medium to high N inputs and Field A across the N gradient. CENTURY simulations predicted a long-term C storage rate of 10 g C/g N for A. re/icns monocultures, the dominant C, grass in Field A and in high-N plots. However, no net C storage was observed for Field A at low N inputs, and at the high end of the gradient, net long-term C storage across all fields converged on roughly 4 g C/g N (Fig. 4). These results underscore the need lor a clearer understanding of why N retention rates differ among ecosystems it ecologists are to make reasonable estimates, whether on local or global sl.lies, of C sequestration in response to N loading.
The grassland types best able to retain added N and sequester C were also the types most vulnerable to N eutrophication through losses of diversity, changes in plant species composition, and the resultant changes in C and N cycling. Thus, N-caused shifts in species composition limit the ability of temperate giasslnnds to serve as significant long-term C stores. In our fields dominated by Q, prairie grasses, shifts in species composition at relatively low N addition rates led to decreased biomass C:N rarios and decreased N immobilization potential, and, consequently, increased soil NO, concentrations, high N loss rates, and low C sequestration rates (g C/g N). The nonlinear or threshold-dependent response that we observed in response to chronic N loading appears to have two causes: species shifts in response to N eu-
1722
SCIENCE    ■    VOL. 274    ■    h DECEMBER !')%
5
trophication and an N mineralization or immobilĽatíon threshold for the decomposition of litter and soil organic matter. Our results show that N loading is a major threat to grassland ecosystems, causing loss of diversity, increased abundance of normative species, and the disruption of ecosystem functioning, and that these responses are tightly linked.
REFERENCES AND NOTES
1. P. M. Vitousek, Ecology 75, 1861 (1994); E Matthews, Global Biogeochem Cycles 8, 411 (1994); J N. Galloway, W. H Schlesinger, H Levy II, A. Michaels, J. L. Schnoor, ibid 9, 235 (1995).
2 D. Tilman, Plant Strategies and the Dynamics and Structure of Plant Communities (Pnnceton Univ Press, Princeton, NJ, 1988).
3.  R Bobbink, J. Appl. Ecol. 28, 28 (1991); H. Olff and J P. Bakker, ibid., p. 1040; R. Aerts and G. W. Heil, Eds, Healhlands- Patterns and Processes In a Changing Environment (Kluwer Academic Press, Dordrecht, Netherlands, 1993).
4.  F Berendse, J. Ecol. 78, 413 (1990).
5.  D. L. DeAngelis, Dynamics of Nutrient Cycling and Food Webs (Chapman and Hall, New York, 1992).
6.  B. J. Peterson and J. M Mehllo, Tellus 37B, 117 (1985), D W. Schindler and S. E. Bayley, Global Biogeochem. Cycles 7,717 (1993), R. J. M. Hudson, S. A. Ghermi, R. A. Goldstein, ibid 8, 307 (1994); D. S. Schimel, Global Change Biol. 1, 77 (1995); A R. Townsend, B. H. Braswell, E. A. Holland, J. É Penner, Ecol Appl. 6, 806 (1996).
7.  D. Tilman, Ecol. Monogr. 57, 189 (1987)
8.  Methods are described in detail (7) The experimental fields are at the Cedar Creek Natural History Area m east-central Minnesota (mean annual temperature = 5.5°C, mean annual precipitation = 726 mm) In 1993, field A had been abandoned from agriculture ror 25 years, Field B for 36 years, and Field C for 59 years Each field contained 54 plots of 4 m by 4 m All plots were randomly assigned to one ot nine treatments: no nutrient addition, addition of macro- and micronutrients other than N, or macro- and micronu-trients plus one of seven N addition treatments ranging from 1 to 27 g Nm-2 year-1 applied as NHjN03 fertilizer in mid-May and mid-June each year. Crushed limestone was added to the plots as necessary to maintain constant pH. Because neither aboveground biomass nor species composition differed significantly between the two non-N treatments, we considered them as a single control treatment here
9.  G I. Agren and E. Bosatta, Environ Pollut. 54, 185 (1989); E.-D. Schulze, Science 244, 776 (1989)
10.  J. D. Aber, Trends Ecol. Evol. 7, 220 (1992)
11.  R. S. Inouye and D. Tilman, Ecology 69, 995 (1988); D. Tilman, ibid. 77, 350 (1996).
12.  Methods used to measure total C and N in each plot: for soils, three cores (2 cm diameter) were divided into 0- to 15-cm, 15 to 30-cm, and 30- to 50-cm sections; composited, and sieved (1 mm), for below-ground biomass, three cores (4.7 cm diameter) were divided into 0- to 15-cm and 15- to 30-cm sections, composited, handwashed over a 1 -mm screen, and dried at 40°C, for aboveground biomass, a 0.3 m2 area was clipped at the soil surface, sorted into litter and live biomass, the live biomass sorted by species, and dried (7). Belowground biomass contained both live and dead fractions. Portions of each sample (composites for rare grasses and forbs in each plot) were ground prior to tissue analysis All soil and plant samples {n = 1610) were analyzed for total C and N with a Carlo-Erba NA1500. Samples were collected between 1 July and 15 August 1993.
13.  P Vitousek, An War. 119,553(1982).
14.  We estimated NUE (g biomass/g N) as the inverse of the N concentration of newly senesced above-
ground tissue (73) for three dominant grasses (sampled in October 1994). In a general linear modeling (GLM) model predicting NUE, the species effect (F = 137.81, P < 0 0001) and the species-by-N-addition rate interaction (F = 6.07, P = 0.0038) were significant, whereas the N addition rate effect (F = 0.438, P = 0.511) was not significant (r2 = 0.95, n = 72). The significant interaction was caused byA repens, for which NUE decreased significantly at high N addition rates, whereas NUE for S. scoparium and P. pratensis did not
15 D. A. Wedm and D Tilman, Oecologia 84, 433 (1990)
16. R.H Brown,CropSci. 18,93(1978),T.R Seastedt, J M. Bnggs, D J. Gibson, Oecologia 87, 72 (1991).
17 In a multiple regression model predicting the C-N ratio of dead biomass (midseason aboveground dead biomass, both recent and old, plus below-ground biomass), the effect of N addition rate (F = 167.46, P < 0 0001), the proportional abundance of the C4 grass S scoparium (based on midseason live aboveground biomass, F = 62.6, P < 0.0001), and total soil N (g N m ~2, F = 18.32, P < 0.0001) were all significant (r2 = 0.791, n = 162). AH F values are partial Fs, that is, corrected for the correlated effects of other terms in the model
18. Monthly net N mineralization was measured with in situ incubations (polyvinyl chloride pipes 2 5 cm in diameter and 15 cm deep) from April to October 1993 Pre- and postincubation soil samples were extracted with 1 M KCl and analyzed for NHa * -N and N03"-N colonmetncally with an Alpkem autoana-lyzer (75). Soil N03~ concentrations presented are means for each plot of preincubation concentrations from May to August (n = 4). To estimate total annual N supply (the sum of N mineralization, N fertilizer inputs, and atmospheric N deposition), we assumed annual atmospheric N deposition equaled 0 6 g N m-2 year-1, the mean wet and dry deposition for 1985-94 measured at our site by the Minnesota Pollution Control Agency.
19   In a two-way analysis of variance testing the effects of field, the four low N addition treatments (0, 1, 2, and 3.4 g N m~2 year-'), and the field-by-N addition interaction on the seasonal average soil N03" concentration (In transformed), the field effect was significant (F = 11 93, P < 0.0001), whereas the N addition (F = 2.65, P = 0.051) and field-by N-addi tion interaction (F = 0.723, P = 0.632) were not. In a comparable analysis with the three medium N-addi-tion treatments (3.4, 5.4, and 9.5 g N m-2 year-1), the field effect was not signincant (F = 0.68, P = 0.51), whereas the N addition (F = 32.40, P < 0 0001) and field by-N-addition interaction (F = 2.69, P = 0.043) were.
20   D. Tilman and D. Wedm, Ecology 72, 685 (1991).
21.  The net change in C and N pools arter 12 years of N addition was calculated as the difference between total C or N stores (gm-2) of each plot and the mean for the 12 control (no N addition) plots in that field. Although insufficient data existed to calculate initial C and N stores in each plot, analyses of total C and N in archived soils (the largest C and N pool) found no significant pretreatment difference between controls and N addition treatments. Nitrogen retention (as a percent) is the ratio of net change in total system N and the sum of experimental N additions over 12 years. Net N lost or gamed (as g N m-2) is the former term minus the latter. Because control and treatment plots both received atmospheric N inputs, these were not included in estimates of net change. At the lowest N addition rates, where changes in total C and N average <3% of total C and N pools, considerable variance exists in our estimates of net change
22.  R. F. Wright era/, For. Ecol. Manage. 71,163 (1995), N. B. Dise and R F Wnght, ibid., p. 153.
23.  J. D. Aber eř a!, Ecol Appl. 3,156 (1993).
24.  In GLM analyses predicting percent N retention (Fig. 2C) (27), the effect of N addition rate was not significant (In transformed, F= 0 671, P = 0.414), whereas effects of field (categoncal variable, F = 18.97, P < 0 0001) and the field-by-N-addrtion interaction (F
SCIENCE   •   VOL 274    •   6 DECEMBER 1996
= 10.322, P < 0.0001) were significant (r2 = 0.261, n = 126). In a stepwise multiple regression model predicting N retention without field as a categorical effect, N addition rate was not significant (F = 0.585, P = 0 446), whereas the effects of Ca aboveground biomass (F= 20.85, P < 0 0001), biomass C:N ratio (Titter and roots, F = 11.25, P = 0 001), soil C N ratio (F = 8.93, P = 0.003), soil N03- (In transformed, F = 6.24, P = 0 014), and the C^-biomass-by-N-addi-tion interaction (F = 6 42, P = 0.013) were significant (n = 126, r2 = 0.300). All F values are partial F's.
25.  E Gorham, P. M Vitousek, W. A Raners, Annu. Rev. Ecol. Syst. 10, 53 (1979); L O Hedin, J. J. Amnesto, A H. Johnson, Ecology 76, 493 (1995).
26.  R G. Woodmansee, S/osc/ence 28, 448(1978). 27   P. M. Vitousek etal, Science 204, 469 (1979); J. C.
Ryden, P R. Ball, E. A Garwood, Nature 311, 50 (1984).
28.  In contrast to forests, where nitnfication (the proportion of net N mineralization ending up as N03-) often increases with increased N loading (70, 23), nitnfica tion exceeded 90?i in all treatments here Because of the high nitrification rates, sandy texture, and consistently aerobic status of our soils, our assumption that N losses are dominated by N03- teaching is reasonable.
29.  In a multiple regression model predicting the seasonal average soil N03- concentration (In transformed), biomass C.N ratio (aboveground litter and roots; slope = -0.0648, partial F = 59 7, P < 0 0001), N supply rate (N addition plus net N mineralization; slope = 0 0367, partial F = 33 78, P < 0.0001), and plant species diversity (Shannon-Wiener index calculated from aboveground proportional biomass, slope = -0.1701, partial F = 14.25, P = 0.0002) were significant (n - 162, r2 = 0.768) Root biomass, live biomass, root.shoot ratio, and soil C:N ratio were not significant correlates of soil N03-.
30 P. M. Vitousek and P. A. Matson, Science 225, 51 (19S4).
31.  D Tilman, D. Wedm, J. Knops, Nature 379, 718 (1996)
32.  K. H. Johnson, K A. Vogt, H. J. Clark, O. J Schmitz, D J. Vogt, Trends Ecol E\ol 11, 372 (1996).
33   Total ecosystem stores (g C m-2) in control plots differed significantly among fields (F = 53 18, P < 0.0001, means: Field A = 3639, Field B = 2537, Field C = 4619). Differences among fields were significant but smaller for the highest N addition treatment (F = 6.59, P = 0.0083; means- Field A = 4509, Field B = 3897, Field C = 5094).
34   W. J. Parton ef al, Global Change Biol. 1,13 (1995); D. S. Schimel et al.. Global Biogeochem. Cycles 8, 279(1994)
35   In GLM analyses predicting C storage per unit N input (Fig. 4), the efiects of field (categorical variable, F = 7.04, P = 0.0013), N addition rate (In transformed, F = 6.99, P = 0.009), and the field-by-N-addition interaction (F = 3 48, P = 0.034) were significant (r2 = 0.16, n = 126). In a GLM model predicting C storage per unit N without the field effect, the effects ot C4 aboveground biomass (F = 18.15, P < 0.0001), soil C:N ratio (F = 12 39, P = 0.0006), and root biomass (In transformed, F - 11.13, P = 0.0011) were signincant (r2 = 0.332, n = 126). All F values are partial Fs                                -^
36   CENTURY is a grassland simulation model of productivity and soil organic matter dynamics that has been used extensively and is descnbed m [W. J. Panon, D. S Schimel, C. V. Cole, D. S Ojima, Soil Sei. Soc. Am. J. 51,1173 (1987), see also (34)]. The long-term monoculture simulations for S. scoparium and A. repens used species-level data on productivity, allocation, and litter quality from (75, 20).
37   SupportedbyNSF(BSR-8811834andBSR-8807831) and the Natural Sciences and Engineering Research Council of Canada We thank S. Finley and E. Williamson for assistance and J Aber, W Cume, H Peat, and W. Schlesinger for comments
23 July 1996, accepted 27 September 1996
1723