Abundance and productivity mediate invader effects on nitrogen dynamics in a California grassland

Soil nitrogen (N) transformations have been shown to be influenced by plant community composition. Identifying species traits that control nitrogen dynamics is more straightforward when species dramatically differ in N input via litter (e.g., N-fixing invaders in a non-fixing community) or in litter carbon:N or lignin:N ratios. Cases where invaders and residents are more similar for such traits are more challenging to evaluate. In these settings, a species’ relative abundance and its contribution to overall ecosystem productivity are likely to contribute significantly to the development of effects on N availability


INTRODUCTION
As shifts in plant species composition have come to be recognized as a significant ongoing component of environmental change (Vitousek et al. 1997, Mack et al. 2000), the links between species composition and ecosystem nitrogen (N) cycling have received a great deal of attention (e.g., Melillo and Aber 1982, Wedin and Tilman 1990, Hobbie 1992, Scott and Binkley 1997, Eviner et al. 2006, Laungani and Knops 2009).The most dramatic examples of the role that vegetation composition plays in soil N dynamics come from studies of N-fixing species (e.g., Boring and Swank 1984, Vitousek et al. 1987, Vitousek and Walker 1989, Stock et al. 1995, Hart et al. 1997, Maron and Jefferies 1999).N-fixers typically accelerate N cycling rates and increase ecosystem N pools as they elevate N inputs into the soil via N-rich plant litter (Vitousek et al. 1987, Hart et al. 1997, Maron and Jefferies 1999, Haubensak and Parker 2004).For non-N fixing woody species, plant effects on soil N mineralization are largely predictable based on leaf litter chemistry and several studies have demonstrated that rates of N cycling in ecosystems dominated by woody species are negatively correlated with litter C:N and lignin:N ratios (Reich et al. 1997, Scott andBinkley 1997).
Controls over herbaceous species' effects on N cycling have proven to be more complex than those of woody species (Wedin and Tilman 1990, Vinton and Burke 1995, Asner and Beatty 1996, Hooper and Vitousek 1998, Mack and D'Antonio 2003b, Eviner 2004, Sperry et al. 2006).Multiple factors, including morphological and physiological traits of the vegetation, productivity, relative abundances, soil chemistry, and microclimatic conditions, influence the impacts that herbaceous species have on N dynamics, sometimes simultaneously (e.g., Wedin and Tilman 1990, Steltzer and Bowman 1998, Ehrenfeld et al. 2001, Eviner and Chapin 2003, Mack and D'Antonio 2003a, Eviner et al. 2006, Rossiter-Rachor et al. 2009).In some cases, indirect effects of vegetation on environmental conditions such as microclimate, disturbance, or soil moisture may equal or exceed the importance of litter chemistry (Ehrenfeld et al. 2001, Mack and D'Antonio 2003b, Eviner et al. 2006).
We examined N dynamics in experimental California grassland communities dominated by three broad types of vegetation: native perennial bunchgrasses, non-native perennial grasses, and non-native annual grasses.All of our experimental vegetation communities were comprised of herbaceous species of relatively close phylogenetic lineage (e.g., all members of the Family Poaceae), though the species comprising the latter two communities arrived in California after European settlement in the 19th Century.The three groups of grasses differ from each other subtly in terms of phenology, life-span, productivity, and, potentially, litter decomposition rates.Previous research has shown differences in various aspects of soil N (Welker et al. 1991, Seabloom et al. 2003, Corbin and D'Antonio 2004a, Hawkes et al. 2005, reviewed in Eviner and Firestone 2007) and water (Hull and Muller 1977, Holmes and Rice 1996, Dyer and Rice 1999) between California grassland plots dominated by native perennial versus non-native annual grasses.Most of these studies focused on monotypic settings rather than species mixtures.Also, little is known about the impacts of the exotic perennial grasses although we have demonstrated that, once established, they are strong competitors against native perennial grasses ( Corbin and D'Antonio 2010).Their effects on soil resources have not been studied despite their presence in California for many decades and their seeming ability to dominate coastal grassland sites (Peart 1989, Corbin andD'Antonio 2010).
Our study focused on two main questions: First, how do soil N dynamics-specifically, the size of inorganic N pools, rates of net N cycling, and microbial biomass-N-differ between experimental communities dominated by the three grass vegetation groups and combinations of the groups?And, second, what species traits, including tissue chemistry, aboveground productivity, and litter inputs, are most useful in explaining differences in N dynamics that we observe?Our experimental communities included single-group treatments as well as mixtures of the native perennial grasses with each non-native group of grasses.The mixtures allowed us to examine the importance of invader abundance relative to native species along with species traits to the availability and cycling of N in soil.

Study area
We conducted our experiment at Tom's Point Preserve, a private nature preserve adjacent to Tomales Bay in Marin County, USA (38813 0 N, 122857 0 W).The portion of the preserve in which we established our plots was typical of a highly invaded coastal prairie grassland, particularly one that has been tilled (Heady et al. 1988, Stromberg et al. 2001).Dominant plants included the introduced grasses Avena barbata Link., Bromus diandrus Roth, and Vulpia myuros (L.) C. Gmelin, introduced perennial grasses such as Festuca arundinacea Schreber and Holcus lanatus L., and exotic annual and biennial forbs such as Carduus pycnocephala L. and Conium maculatum L. Nomenclature follows Hickman (1993).Few native grasses were present where we established our plots, although they were abundant in nearby areas that had never been tilled.The soil at the study site is a sandy loam (fine, mixed mesic Ultic Paleustalf ).
The climate in this region is Mediterranean, with most of the precipitation (mean annual precipitation ¼ 790 mm) falling between November and April as rain.Coastal fog, present in the summer months, moderates the summer drought (Ingraham andMatthews 1995, Corbin et al. 2005).Mean monthly temperature ranges from 13.98C in March to 18.18C in September.During the course of the experiment, annual precipitation and mean monthly temperatures were relatively similar to 30-year averages, though Year 1 (1998( -1999( ) and Year 3 (2000( -2001)), were slightly wetter and dryer than normal, respectively.

Experimental design
In September 1998 we cut and sprayed all standing vegetation with 5% glyphosate-based herbicide.All biomass and litter were subsequently removed.We established eight replicates of each of five different community types in a randomized block design (two blocks): native perennial bunchgrasses only (NP), exotic annual grasses only (EA), exotic perennial grasses only (EP), and native perennials with each exotic group (NP þ EA and NP þ EP).Each community type was a mixture of three (for NP, EA and EP communities) or six (for NP þ EA and NP þ EP communities) species.We did not work with monocultures of individual species, because such monocultures are not realistic for our system.Each of the 40 plots was 1.5 3 1.5 m with a 1-2 m buffer between plots.
Seeds of the native perennial grasses Agrostis oregonensis Vasey, Festuca rubra L., and Nassella pulchra (A.Hitchc) Barkworth and the exotic perennial grasses F. arundinacea, H. lanatus and Phalaris aquatica L. were collected at Tom's Point in spring 1998.Seeds were planted in individual Cone-tainers (Stuewe and Sons, Tangent, Oregon, USA) in September 1998 and allowed to germinate under greenhouse conditions.The three month-old seedlings were transplanted into appropriate plots in January 1999 at a density of 144 plants per plot.Dead individuals (,1% of all plants) were replaced for the first month only.
Seeds of the exotic annual grasses A. barbata, B diandrus, and V. myuros were collected on site and cast onto EA and NP þ EA plots beginning in December 1998.Seed densities-1500 A. barbata, 775 B. diandrus, and 4000 V. myuros per m 2 -were chosen to be similar to estimates for California annual grasslands (e.g., Biswell andGraham 1956, Heady 1958).Unlike the native and exotic perennial grasses, we reapplied exotic annual seeds each fall with seeds collected from outside the experimental plots.Seeds were added at similar levels in subsequent seasons, with the exception that the number of V. myuros seeded was reduced to 3500 seeds per m 2 after year one.All plots were weeded three times each year to maintain species composition and density.

Soil analyses
Soil was analyzed for extractable ammonium and nitrate and net rates of N mineralization and nitrification four times per year, corresponding to the climatic seasons (Eviner and Firestone 2007), with the exception of winter of 2000-01.At each sampling period three 10 cm deep 3 2 cm wide cores were taken from each plot, bulked, and sieved (,2 mm).One subsample of soil (10 g) was collected from each sample and immediately extracted with 50 mL of 2 M KCl.Another soil sample was incubated in capped polyethylene vials at field moisture for 14 days in the lab at 258C for the determination of potential net N mineralization and nitrification rates.The incubated samples were extracted with KCl as above.An additional subsample was dried overnight and weighed for determination of gravimetric water content (GWC).
We assessed microbial biomass-N in our soil samples three times per year beginning in October 1999 using the chloroform-fumigation method (Brookes et al. 1985).One 10 g subsample of soil from each plot was immediately extracted with 40 mL of 0.5 M K 2 SO 4 for determination of extractable N. The other subsample was fumigated with chloroform for 5 days then extracted with K 2 SO 4 as above.Organic N in the K 2 SO 4 extracts was converted to NO 3 À in a sulfuric-salicylic digestion (Howarth and Paul 1994).
In the third year of the study, we obtained an estimate of N availability to plants using ion exchange resins (Binkley and Matson 1983).We constructed small resin bags using nylon stocking and 3g (dry-weight) of mixed bed ionexchange resins (J.T. Baker, Phillipsburg, New Jersey).One bag was placed in each plot 5 cm below the soil surface in November 2001.Bags were retrieved in January 2002 and rinsed with 2 M solution of KCl overnight.The resulting extract was analyzed for NH 4 þ and NO 3 À .Ammonium and nitrate concentrations in all KCl and K 2 SO 4 extracts were measured using a Lachat flow-injection autoanalyzer at UC Berkeley, then converted to micrograms of NH 4 þ and NO 3 À /g soil using GWC-corrected soil weights.Net mineralization of N was calculated as extractable ammonium þ nitrate in the incubated sample minus extractable ammonium þ nitrate in the initial extracts.Net nitrification was calculated as extractable nitrate in the incubated sample minus extractable nitrate in the initial extract.Microbial biomass-N was calculated as the difference between N content of fumigated and unfumigated subsamples using the appropriate conversion factor (Brookes et al. 1985).

Vegetation biomass and productivity
Biomass of native and exotic perennial grasses was sampled nondestructively to minimize disturbance.We constructed allometric relationships between basal diameter, longest leaf length, and the number of flowering culms versus the aboveground biomass for each species by harvesting 46-87 individuals of each species representing a range of plant sizes.All plant material was collected from the outer 0.25 m of each plot so as to minimize disturbances to the area in which quantitative sampling took place.Separate allometric equations were used for the first sample date when perennial plants were very small (March 1999) and all subsequent sample dates.Regressions between plant size and biomass resulted in r 2 values above 0.85 for all species and sample dates.(Additional details, including allometric equations, can be found in Corbin andD'Antonio (2004a, b, 2010).)Native species were measured twice per season to estimate productivity over the course of the season.Productivity each season was calculated as the difference between the beginning and end of the growing season, or between March 1999 and the end of the growing season in the case of the first year.
Above-ground biomass of exotic annual grasses was destructively harvested in either late May or June of each season when all plants were dead.All aboveground exotic annual biomass in each EA and NP þ EA plot was clipped to ground level in three randomly selected 0.25 3 0.25 m subsamples and separated by species.This harvest accurately reflects total above-ground productivity because exotic annual grasses die back completely each spring.Following drying (608C) to constant mass, each sample was weighed and returned to appropriate plots.
Belowground biomass was sampled at the end of the growing season in 2000, 2001and 2002 (Years 2, 3 and 4, respectively) (Years 2, 3 and 4, respectively).We extracted three 2 cm wide 3 15 cm deep soil cores from each plot at each sample period, and bulked the cores by plot.Roots were removed from the cores through a combination of hand-picking using tweezers and gentle washing using a Bel-Art (Scienceware, Pequannock, New Jersey USA) root elutriator then dried (608C) and weighed.

Tissue chemistry and litter decomposition
In order to assess tissue chemistry and rates of mass and N loss from each species' litter, we decomposed aboveground litter under field conditions over a 20-month period.Recently senesced leaf and flowering stem material was collected from individuals outside our plots in July 2000.Following air drying, 1-2 g of plant material was weighed and placed into 10 3 10 cm bags made of 1 mm mesh fiberglass windowscreen.For six of the species-A.barbata, B. diandrus, F. arundinacea, H. lanatus, N. pulchra, and P. aquatica-leaf and stem material was decomposed separately.For the remaining three species-A.oregonensis, F. rubra and V. myuroswhose stem tissue is more similar to their leaf material and is difficult to separate, only leaves were decomposed.42 replicate bags of each species and tissue type were placed in the mixed grassland outside our plots in August 2000, and six bags of each species-tissue type were collected at seven intervals over a 20-month period: September and November 2000, January, March,  and May 2001, and January and April 2002.At each collection time, material in the bags was gently brushed to remove soil or debris, dried (408C), weighed, ground and analyzed for percent N using a Carlo Erba NA 1500 CHN autoanalyzer (Fisions Instruments, Beverly, Massachusetts, USA).Percent mass and N loss was calculated as the difference between original mass and total tissue N content in each bag and the collected mass and its total tissue N. Initial lignin content of litter was analyzed using the acetyl bromide method (Iiyama andWallis 1990, Hatfield et al. 1999).
We estimated the amount of C and N and the ratio of C:N input from decomposing litter in each plot in 2000-2002 in order to understand the influence of litter quantity and quality on net N cycling dynamics.Litter biomass for perennial species was estimated each year by subtracting our estimate of each plot's above-ground biomass in the fall from the above-ground biomass in the previous spring.For annual species, we assumed that all spring biomass became litter.
We calculated the amount of C, N, and the C:N ratio input from litter as a function of biomass, initial tissue chemistry, and decomposition rate of each species' litter in each plot (Appendix: Table A1).Each plot's C and N input was calculated using the following formulas: where sp ¼ each of the nine grass species, B sp ¼ litter biomass, C sp ¼ percent C of leaf or stem litter at start of decomposition, N sp ¼ percent N of leaf or stem litter at start of decomposition, and Decomp sp ¼ Rate of C or N loss over 9-month period, calculated from the litter decomposition experiment.We calculated separate inputs of C and N for leaf and stem tissue of two species, A. barbata and B. diandrus.Based on field surveys, we assumed that 36% of total A. barbata biomass was leaves and 64% was stems, and that 44% of total B. diandrus biomass was leaves and 56% was stems (Corbin and D'Antonio, unpublished data).

Statistical analyses
Soil nitrogen dynamics were summarized in three ways so that we could assess both withinyear and between-year variation.First, in order to describe N dynamics at different points during the growing season, we calculated mean seasonal values (e.g., fall, winter, spring, and summer) for inorganic N pools (ammonium and nitrate), net N mineralization, and net nitrification for each plot.Second, we calculated within-year net rates of N mineralization and nitrification by summing the seasonal N cycling values within a year, weighted by the number of days in each season.Sampling did not take place in winter 2000, so in order to maintain consistency across years, we omitted the winter N cycling period from calculation of within-year rates for all years.Seasons were applied as: fall ¼ 45 days (November 1-December 15); winter and spring ¼ 151 days (December 16-May 15); summer ¼ 169 days (May 16-October 31).Analysis of N cycling during winter did not reveal significant differences between treatments (see Results), so this is an accurate reflection of relative differences in within-year net rates of N cycling, though absolute values may be higher than if winter values were included.Finally, in order to assess variability in N dynamics over the course of the experiment, we calculated the coefficients of variation (CV's) (means divided by standard deviation) for inorganic N levels and net rates of N cycling in each plot during the 40 month sampling period.
We analyzed differences in soil N dynamics by treatment at three different time scales.First, we compared soil ammonium, nitrate and rates of net N mineralization and net nitrification over the course of the entire 40 month period of soil sampling, treating all 15 sample time points as unique measurements in repeated measures (RM) ANOVA.Second, we compared these same dependent variables by season (fall, winter, spring, and summer) using the mean seasonal rates described above.In this analysis, the estimates of seasonal N dynamics in each plot were treated as unique measurements varying by Season in RM ANOVA.Third, we compared within-year seasonal rates of net N mineralization and net nitrification in our three full seasons-1999-2000, 2000-2001 and 2001-2002-where summed seasonal rates in each plot in each year were treated as unique measurements varying by Year in RM ANOVA.The RM ANOVA models included Treatment, Date or Season or Year, and the interactions between Treatment and the appropriate time scale.The effects of time in each model and the interactions were analyzed by MANOVA using Roy's Greatest Root (Scheiner 1993).Separate ANOVA's for each Season and Year (but not Date) were also performed.Inorganic N levels and N cycling rates were log transformed to meet the assumptions of ANOVA.All analyses were performed using SAS version 9.1.
We used linear regression to test whether the log-transformed rates of within-year and spring net N mineralization in a plot were related to the proportion of above-ground productivity that derived from exotic species.We ran separate regressions for each year, and also for three different ways of calculating proportion of exotic productivity: proportion of all exotics, which compared all five plot types; proportion of exotic annual grasses, which only considered NP, EA, and NP þ EA plots; and proportion of exotic perennial grasses, which only considered NP, EP, and NP þ EP plots.The exotic proportion in NP plots was functionally 0, while exotic proportion in EA and EP plots was 1.
Differences in CV by treatment were compared using ANOVA (Sokal and Braumann 1980) on log-transformed CV's.Differences between treatments were compared using LSD.
We compared differences in above-and belowground productivity between treatments using RM ANOVA including Treatment, Year, and the Treatment 3 Year interaction as independent variables.
We used ANOVA to test for differences in ammonium and nitrate in ion-exchange resin bags and for differences in the rates of mass and N loss and the C:N ratio of decomposing litter in each species and tissue type.Because individual litter bags were not sampled repeatedly during the 20-month decomposition experiment, we did not use repeated measures analysis.
We compared the influence of vegetation productivity, litter chemistry and quantity, and soil chemistry on within-year rates of net N cycling and rates of N cycling in the spring.In order to eliminate co-linearity between independent variables, we used Principle Components Analysis (PCA) to create composite variables that were used in the regression analysis.Composite variables for vegetation productivity were created using each plot's productivity, above-ground biomass, below-ground biomass, and root:shoot biomass ratio for a given year; for litter dynamics, we used litter input from previous season, the amount of litter C and N input and ratio of C:N input from decomposing litter (see Appendix: Table A1 for details); for soil chemistry, we used soil C, soil N and soil C:N ratio.Each plot's position along the first two axes of each PCA were applied as explanatory variables versus summed seasonal net N mineralization and net nitrification, spring net N mineralization and net nitrification, and spring microbial biomass in 2000, 2001, and 2002 using separate backwards stepwise multiple regressions.Mean annual GWC and spring GWC for each year were also included as potential explanatory variables.

Soil nitrogen dynamics
Soil N dynamics varied over time and among treatments during the course of the three-plus years of our sampling (Table 1, Figs. 1 and 2).We found highly significant effects of season on KClextractable ammonium and nitrate, and net N mineralization and nitrification rates.(Table 1, Fig. 2).Ammonium levels were highest in the winter and summer, while nitrate levels were highest in the fall.Rates of net N mineralization and net nitrification were significantly higher in the fall, winter and spring as compared to the summer (Fig. 2).
Treatment effects on soil inorganic N depended on the season in which they were sampled (Table 1; Figs. 1 and 2).We found significant Treatment interactions with season for soil nitrate levels, and a trend (p , 0.11) toward the same interactions for soil ammonium levels.(Table 1: Season 3 Treatment).Specifically, inorganic N levels differed among treatments in the spring (pairwise comparisons using LSD p , 0.05), when Exotic Annual (EA) plots had significantly lower nitrate levels than Native Perennial (NP) or Exotic Perennial (EP) plots and lower ammonium levels than EP plots (Fig. 2).In no other season were there significant differences in inorganic N between treatments (p .0.11).
Plot composition consistently influenced rates of net N mineralization and nitrification.RM ANOVA analysis showed significant effects of Treatment when compared by Season (Table 1) and Year (net N mineralization: F 4,20 ¼ 5.1, p , 0.005; nitrification F 4,20 ¼ 5.1, p , 0.005).Both net N mineralization and net nitrification rates were significantly lower in EA plots compared to the other treatments (Figs. 2 and 3).Within-year rates of net N cycling were lower in EA plots than NP plots in all three years, and lower than EP plots in 2000-2001and 2001-2002 (Fig. 3 (Fig. 3).We found significant within-season differences between treatments in the spring only (N mineralization F 4,34 ¼ 3.6, p , 0.01; nitrification F 4,34 ¼ 3.0, p , 0.04), when rates in NP plots were significantly higher than rates in EA plots (Fig. 2).
N dynamics in mixed-group plots (NP þ EA and NP þ EP) were intermediate between their respective single-group plots (Figs.1-3).Withinyear rates of net N mineralization and nitrification and N transformation rates during spring were significantly faster in NP þ EA plots than in EA plots, but slower than in NP plots.Soil nitrate levels during the spring were also significantly higher in NP plots than EA plots, but did not differ from NP þ EA plots (Fig. 2).Spring rates of N cycling in NP þ EP plots were significantly lower than in NP plots, but not significantly different from EP plots (Fig. 2).Otherwise, there were no significant differences in N cycling between NP, EP, and NP þ EP plots.
Linear regression analysis demonstrates that the higher the proportion of a plot's productivity that derived from exotic plants, the lower the rates of net N mineralization, although the slope varied by year and the group of exotic species tested (e.g., EA versus EP, Table 2).In 1999-2000, there was a negative relationship between net N mineralization (within-year and spring only rates) and all three measures of exotic grass proportion.Exotic annual species contributed much more to the decline in net N mineralization than did exotic perennial grasses: the steepest slopes were observed for proportion of exotic annual grasses in the annual above-ground productivity (Table 2).In subsequent years, there was a significant (negative) relationship between proportion of exotic annual grass and net N mineralization (Table 2b), but not for proportion of exotic perennial grass or the proportion of exotic grass in all plots (Table 2a,c).
Comparing coefficients of variation (CV) in each treatment indicate that when averaged over time, N dynamics in EA plots were more variable than in other treatments.The CV for net nitrification during the 40 months of soil sampling in EA plots was higher than in NP or NP þ EA plots (F 4,35 ¼ 2.12, p , 0.099; pairwise comparison using LSD p , 0.05).Soil nitrate levels in EA plots also had a higher CV than EP or NP þ EA plots but not NP plots (F 4,35 ¼ 2.4, p , 0.07; pairwise comparisons using LSD p , 0.05).
Estimates of plant-available N, as measured by recovery of inorganic N in ion-exchange resin bags (only one time period measured: Nov. 2001-January 2002) were consistent with net N cycling rates in laboratory incubations.Significantly more inorganic N (ammonium þ nitrate) was found in NP plots as compared to EA or EP plots (Treatment F 4,33 ¼ 3.3, p , 0.03; Fig. 4), largely because of the differences in nitrate.Nitrate and total resin N amounts in combined plots were intermediate between NP plots and the other two single-group plot types.There was also significantly less ammonium recovered from EP plots than NP or combined plots; however, the amount Microbial biomass-N varied through the course of the experiment (RM ANOVA: Date F 8, 216 ¼ 46.8, p , 0.0001), with higher values during the growing season (winter/spring) than during the summer (Fig. 5).Microbial biomass-N also varied by treatment (RM ANOVA: Treatment F 4,27 ¼ 3.1, p , 0.04): it was generally higher in EP plots and lower in EA plots (Fig. 5).The v www.esajournals.orgdifference between EA and the rest of the treatments tended to be greatest during the spring.

Vegetation biomass and productivity
Plots that contained perennial species had significantly higher above-ground productivity (RM ANOVA: Treatment F 4,33 ¼ 46.3, p , 0.0001) and below-ground biomass (RM ANOVA: Treatment F 4,26 ¼ 9.8, p , 0.0001) than the EA plots (Fig. 6).In the first year, plots with exotic perennial species, whether alone (EP) or with native perennials (NP þ EP), were the most productive (above-ground).In years 2-4, Native Perennial (NP) plots were the most productive, although after year 2 there were no significant differences between all treatments containing perennials (Above-ground productivity RM ANOVA: Treatment 3 Year F 4,33 ¼ 16.1, p , 0.0001).We found a similar pattern for belowground biomass: EA plots had significantly lower root biomass as compared to the other plots (Fig. 6).EP and NP þ EP plots had the highest below-ground biomass each year, though they did not differ significantly from NP and NP þ EA plots.

Tissue chemistry and litter decomposition
At the initiation of the litter bag study, leaf and stem litter from exotic perennial species had the highest concentration of N and the lowest C:N and lignin:N ratios of the three species groups (Table 3).Exotic annual grass leaf and stem litter had the lowest concentration of N and the highest C:N and lignin:N ratios, largely because of the leaves of V. myuros, whose mixture of stems and leaves had the lowest concentration of N and highest C:N and lignin:N ratios.
The rate of loss of N and biomass from leaf and stem tissue was fastest in the first 9 months after the placement of experimental litter bags (Fig. 7).Little N was lost from any litter type during the remaining 11 months.Leaf litter derived from exotic annual species consistently lost a greater proportion of their original N and mass than leaf litter from either native or exotic perennial species (Fig. 7; N loss F 2, 349 ¼ 36.5, p , 0.0001; mass loss: F 2, 229 ¼ 74.7, p , 0.0001).There was no significant difference between the loss of N from stem litter from the three groups (F 2, 214 ¼ 0.1, p .0.5).

Multiple linear regression
Each Principal Component Analysis extracted two composite axes that collectively explained between 93% and 99% of the variation (Table 4).For Vegetation Productivity, the first axis (Prod1) was associated with above-ground biomass, above-ground productivity, and total biomass in each year's PCA analysis.The second axis (Prod2) was associated with below-ground biomass and, in 2001 and 2002, root:shoot ratio.For Litter Dynamics, the first axis (Litter1) was associated with litter biomass and C and N input from litter (Eqs. 1 and 2).The second axis (Litter2) was associated with the C:N ratio of the litter inputs.For Soil Chemistry, the first axis (Soil1) was positively associated with soil N and soil C levels and negatively associated with soil C:N ratio.The second axis was positively associated with soil C:N ratio and, in 2002, with soil C.
The six PCA axes (Prod1, Prod2, Litter1, Litter2, Soil1, Soil2) along with mean annual GWC and spring GWC in each year were used as composite variables in multiple linear regression models to determine factors best predicting within-year and spring net N cycling rates in 2000-2002.The resultant models had r 2 ranging from 0.10 to 0.48 (Table 5a,b).Within-year rates of net N mineralization and net nitrification were most strongly, and positively, influenced by PCA axis Prod1 (above-ground productivity, aboveground biomass, and total biomass) in all three years.Within-year net N mineralization was also

DISCUSSION
Plant community composition had a signifi-  v www.esajournals.orgcant influence on soil inorganic nitrogen, N cycling rates, and microbial biomass.The greatest differences were between plots dominated by exotic annual grasses (EA) and plots dominated by either native (NP) or exotic (EP) perennial grasses.Differences were most pronounced during the spring, the time of peak grass physiology and flowering and seed set for the annual species (Eviner and Firestone 2007).Extractable soil nitrate during this period was significantly lower in EA plots compared to either of the perennial groups, and rates of net N mineralization and nitrification were significantly lower in EA compared to either NP or NP þ EA plots.Thus EA plots appear to function differently than plots with perennial grasses while in general, origin of the perennial grasses (e.g., NP versus EP) made little difference.
Contrary to our findings, non-native species have generally been associated with increases rather than decreases in soil N availability or cycling compared to native species.In a metaanalysis of 88 studies, Liao et al. (2008) determined that soils associated with non-native species tend to have higher soil N pools and greater availability of inorganic N compared to native species.Nonetheless, our finding that N availability and cycling was lower where the exotic annual grasses grew alone compared to native dominated or co-dominated plots, is consistent with other, primarily grassland, studies that have reported decreases in N cycling rates following invasion (reviewed in Ehrenfeld 2003).For example, Christian and Wilson (1999) found less available N and total soil N in soils associated with the non-native C 3 grass Agropyron cristatum than in native grass-dominated successional prairie in southwest Saskatchewan.Evans et al. (2001) reported a decrease in N availability and net N mineralization rates following invasion of the exotic annual grass Bromus tectorum into a semi-arid native perennial grassland in Utah.
The impacts of non-native species on N dynamics are difficult to generalize even within California grassland ecosystems (for review see Eviner and Firestone 2007).For example, Eviner and Hawkes ( personal communication, cited in Eviner and Firestone 2007) reported higher net N cycling rates in Northern California experimental plots with the native perennial grass Nassella pulchra than those with the exotic annual grasses  chemical and texture differences between sites can influence species effects although this too has been poorly studied (Eviner and Chapin 2003).
EA plots exhibited the unusual combination of lower extractable N availability and slower rates of net N cycling as well as greater variability in extractable nitrate and net nitrification than NP plots.We are not aware of previous findings that N dynamics in habitats dominated by non-native species are more variable than those dominated by native species.Yet because EA plots consist only of annual species, the greater degree of variability over time is not surprising compared to perennial plots where plant activity is spread over a greater time of the year (Jackson et al. 1988).Differences in both mean levels of nutrient availability during the study and variability over time may help explain our observation (Corbin and D'Antonio 2004b) that the biomass of nonnative dicot seedlings such as Conium maculatum and Carduus pychnocephala collected during fall weeding was typically higher in plots dominated by exotic annual grasses than in plots dominated by native perennial grasses.One explanation is that higher light levels and available space for germination in EA plots compared to NP plots (Corbin and D'Antonio 2004b) facilitated their establishment.However, it is also possible that the high variability in N dynamics in EA plots allowed more windows of opportunity for invading species, as proposed by Davis et al. (2000).

Relating species traits to net N cycling rates
We found a negative relationship between the proportion of a plot's productivity that derived from exotic species and net N mineralization rates.Thus, the effects of the exotic species on N cycling were a function of the species' productivity relative to the rest of the community and not whether they were present or absent.The relationship was strongest and most consistent when comparing the proportion of exotic annual grasses in NP, EA, and NP þ EA plots.The fact that the exotic annual grasses were less productive than native perennial grasses, thereby decreasing nutrient uptake and litter inputs, was likely an explanatory factor.
Differences in litter decomposition rates, including N loss, have frequently been evoked to explain differences in N dynamics associated with various species (e.g., Hobbie 1992, Mack andD'Antonio 2003a).In our experiment, the rates of mass loss and N loss from decomposing litter were significantly faster for the exotic annual grasses than the native or exotic perennial grasses (Fig. 7), even though exotic annual grasses had the lowest percent N and the highest C:N and lignin:N ratios compared to the perennial grasses.Based solely on decomposition rates, we would then predict that N availability and rates of N cycling would be highest in EA plots.Instead, EA plots had significantly lower levels of extractable nitrate in the spring and lower rates of net N cycling in the spring and over the course of the whole growing season (Figs. 2 and 3).Thus, litter decomposition rates alone are a poor predictor of soil N dynamics in our system.
Our stepwise regression analysis revealed that, instead, vegetation productivity was the best predictor of N cycling rates.The composite variable that was made up of above-ground biomass, productivity, and total biomass was the most important variable in the regressions of within-year net N mineralization, within-year net nitrification, and spring net nitrification in all three years.Litter chemistry, including the amount of N input into soil from decomposing litter (Litter1), and soil chemistry also influenced N cycling dynamics in some time periods (Table 5a,b).We hypothesize that N cycling rates were driven by the size of the pool of N in vegetation and litter, and secondarily by soil chemistry and moisture, rather than by the ratio of leaf or litter C:N.For example, in 2000 and 2001, even though the loss of N from exotic annual grass litter was relatively fast, the lower productivity of the EA treatment (Fig. 6) meant that there was less litter and, therefore, a smaller pool of N to decompose than in the other treatments.The smaller pools of litter N, in turn, led to lower rates of net N mineralization and nitrification.
We can also evoke species traits to explain the lack of significant differences in N cycling dynamics between NP and EP treatments.These two groups of species were relatively similar in terms of productivity after the second year, allocation to roots versus shoots, tissue chemistry, and litter decomposition rates.Previous research has also shown that they are both capable of maintaining water uptake during the region's summer drought (Corbin et al. 2005).
Our data demonstrate that native and exotic perennial grasses did not differ enough in key species traits for detectable differences to develop.
Microbial biomass-N was also found to be positively related to vegetation productivity, along with soil C and N levels, in two out of the three years that we sampled.Lower C inputs from litter and root exudates in less-productive plots likely would be expected to support smaller microbial populations.Indeed, microbial biomass-N was lowest in EA plots.The importance of vegetation productivity for microbial biomass in our grassland system is consistent with Zak et al. (1994), who found a positive correlation between within-year net primary productivity and microbial biomass across a range of vegetation types.Other studies have also observed links between plant community composition and microbial biomass-N (e.g., Johnson et al. 2003, Carney andMatson 2005).
This study is part of a growing body of evidence that the development of species effects in herbaceous species is driven by vegetation biomass or productivity (e.g., Joffre 1990, Wedin and Tilman 1990, Aerts et al. 1992, Hobbie 1992, Grime 1998, Steltzer and Bowman 1998, Rossiter-Rachor et al. 2009).For example, Joffre (1990), studying similar species of annual grasses (Bromus hordeaceous and Vulpia spp.) and a perennial grass (Phalaris aquatica) in southwestern Spain found significantly lower net N mineralization rates in plots dominated by the less productive annual grasses.Aerts et al. (1992) correlated higher net N cycling rates with higher root productivity and root turnover by the grass Molinia caerula versus the grass Deschampsia flexuosa or the shrub Calluna vulgaris.Rossiter-Rachor et al. ( 2009) related higher levels of NH 4 þ availability in soils associated with the invasive grass Andropogon gayanus to greater biomass and above-ground N pools following invasion.Wedin and Tilman (1990) and Steltzer and Bowman (1998) found that below-ground biomass and fine-root productivity, along with tissue quality (e.g., leaf C:N or lignin:N ratios), in two different grass-dominated systems were positively correlated with net N mineralization and net nitrification.Eviner et al. (2006), using a series of herbaceous mesocosms, found that aboveground biomass was related to soil N cycling, through its effects on microclimate.These results contrast with comparisons in systems dominated by woody species (e.g., Hobbie 1992, Reich et al. 1997, Scott and Binkley 1997), where species effects on soil N dynamics are more simply correlated with litter chemistry including C:N and lignin:N ratios.
The abundance or biomass level above which a species invasion produces detectable differences in ecosystem processes is likely a function of the differences in species traits between the resident community and the new addition (Chapin et al. 1996, Parker et al. 1999, D'Antonio and Corbin 2003).Environmental factors such as soil physical or chemical conditions or microclimate that modulate decomposition and soil microbial processes also likely influence the development of species effects.For a given set of environmental conditions, the larger the difference in, for example, tissue chemistry, nutrient uptake, or effects on soil microclimate, the smaller a proportion of the community or the litter pool an invader might need to occupy before it alters soil N dynamics.For example, Peltzer et al. (2009) demonstrated that, on a river floodplain undergoing primary succession, relatively rare species (,3% of community biomass) with unique life histories compared to the rest of the community were capable of exerting disproportionate effects on belowground processes.In our study exotic perennial invaders were not different enough from native perennial grasses to cause any change in soil N dynamics even where they were completely dominant (e.g., EP plots).By contrast the exotic annual grasses exerted an effect both in mixture and monoculture although their effects were most clearly seen where they were completely dominant.Understanding the interactions between abundance, differences in species traits between native and invading species, and environmental conditions will help refine our ability to predict when and where invasive species will have ecosystem impacts.

Fig. 1 .
Fig. 1.KCl-extractable ammonium and nitrate and net rates of N mineralization and nitrification in each treatment from March 1999 through June 2002.Each symbol represents mean (6SE).Horizontal bars on x-axis indicate the rough period of the growing season (November-June) in California's Mediterranean climate.

Fig. 2 .
Fig. 2. Soil inorganic N and N cycling rates in each season.Each bar is the mean (6SE) value of 8 replicate plots per treatment after averaging across three years of soil sampling.See text for treatment descriptions.Letters indicate significant differences between treatments within seasons, as measured by LSD (p , 0.05).

Fig. 3 .
Fig. 3. Within-year rates of net N mineralization (top panel) and net nitrification (bottom panel) in each treatment during 1999-2000, 2000-2001 and 2001-2002 years.Amounts were calculated by summing N cycling rates during a growing season year multiplied by the number of days in each season.See Methods for details.Bars are means (6SE).Letters indicate significant differences between treatments within a year, as analyzed using LSD (p , 0.05).See text for further statistical analyses.

Fig. 4 .
Fig. 4. Ammonium, nitrate, and total inorganic N (NH 4 þ þ NO 3 À ) collected from each treatment using ionexchange resin bags as an estimate of plant-available N. Resin bags were installed November 2001 and extracted January 2002.Letters indicate significant differences between treatments, as analyzed using LSD (p , 0.05).

Fig. 5 .
Fig. 5. Mean (6SE) of chloroform-labile N in soil samples, an index of soil microbial biomass-N, in each treatment from October 1999 through June 2002.Horizontal bars on x-axis indicate the rough period of the growing season (November-June) in California's Mediterranean climate.Note that chloroform-labile N was not sampled during summer 2000 or fall 2001.

Fig. 6 .
Fig. 6.Above-ground productivity and below-ground biomass in each treatment.Letters indicate significant differences between treatments within years, as measured by LSD (p , 0.05).

Fig. 7 .
Fig. 7. Decomposition (percent loss of original N) over 20 months of leaf and stem tissue from the three groups of species: exotic annuals, native perennials, and exotic perennials.Litter bags containing 1-2 g tissue were initially placed in the field in October 2000.Amount of N was calculated by multiplying (biomass in each litter bag) 3 (tissue %N).Symbols represent means (6SE) of seven replicate bags per species pooled by species group.Asterisks indicate significant differences between species groups, as measured by LSD (p , 0.05).Horizontal bars on x-axis indicate the rough period of the growing season (November-June) in California's Mediterranean climate.

Table 1 .
Repeated measures ANOVA comparing the effect of Treatment on log-transformed inorganic N (NH 4) and net N cycling rates (N mineralization and nitrification) using mean values for each season (fall, winter, spring and summer) from October 1999 to June 2002.
v www.esajournals.org of ammonium was relatively low compared to the amount of nitrate.

Table 2 .
Regression analysis of relationship between proportion of plot productivity that was non-native (exotic annual and exotic perennial) and log-transformed rates of net N mineralization in a) all plots, b) NP, EA, and NP þ EA plots, and c) NP, EP, and NP þ EP plots.Analysis was done separately for annual rates and spring rates of net N mineralization.

Table 3 .
Initial tissue chemistry of leaf and stem litter at the time of placement of litter bags.Values are means 6 1 SE.

Table 4 .
Principle Components Analysis (PCA) of vegetation productivity, litter dynamics, and soil chemistry in2000-2002.For each PCA (Vegetation Productivity, Litter Dynamics, and Soil Chemistry), factor loadings for each axis are presented, as well as eigenvalues and proportion of the variation explained by the included variables.Variables with values greater than 0.45 or lower than À0.45 on either axis are highlighted and are assumed to be strong explanatory variables for that PCA analysis.

Table 5 .
Stepwise multiple regressions for net mineralization, net nitrification, and microbial biomass-N in2000- 2001.Results from PCA analysis (Table4) were used as composite variables for vegetation productivity (Prod1, Prod2), litter dynamics (Litter1, Litter2), and soil chemistry (Soil1, Soil2).Mean annual GWC and spring GWC were included in all regressions as untransformed variables.Only explanatory variables selected by backwards elimination (p , 0.10) are shown.See Methods for further details.Rows for N cycling variable show statistics, including r 2 for model.Rows for significant explanatory variables show F (for model) and t (for individual parameters) and p-value (where there is . 1 significant explanatory variable) and the direction of the relationship.
improved this study.Comments from R. David Evans and two anonymous reviewers improved the manuscript.Funding was provided by the National Science Foundation (DEB 9910008), the UC Berkeley Faculty Fund for Research in the Biological Sciences, and the Marin Community Foundation.The Eastern New York Chapter of The Nature Conservancy provided research space for JDC during analysis and manuscript prep-