Nitrification is linked to dominant leaf traits rather than functional diversity


  • Present address: Department of Biological Sciences, University of Waikato, Private Bag 3105, Hamilton 3240, New Zealand

Correspondence author. E-mails:,


1.  The internal cycling of nitrogen (N) is a critical process in terrestrial ecosystems. Nitrification occurs when soil microbes convert ammonium to nitrate. Nitrification is known to be regulated by abiotic soil properties, but less is known about how plant communities influence this important ecosystem function.

2.  Two contrasting hypotheses propose mechanisms for how communities influence ecosystem processes. The ‘mass ratio hypothesis’ proposes that dominant species control ecosystem processes, e.g. communities dominated by plants with high leaf N content may increase rates of N cycling. The ‘diversity hypothesis’ proposes that community diversity controls ecosystem processes, e.g. diverse communities may increase rates of N cycling by providing a more consistent supply of organic N as variable litter qualities break down at different rates. Each hypothesis was simultaneously evaluated using structural equation modelling in the context of a ponderosa pine forest.

3.  The first principle component extracted from a species–trait matrix captured interspecific covariation in leaf traits such as specific leaf area and leaf N content. This ‘leaf economics spectrum’was scaled to the community-level by calculating community-weighted mean leaf trait axis scores. Quadratic entropy was used as the index of understorey functional diversity to capture variation in functional traits within the community.

4.  Structural equation modelling results suggest that soil factors, such as pH, temperature, C:N mass ratio and total N were the dominant factors exerting direct control over nitrification potential. Understorey leaf traits were positively associated with nitrification potential, suggesting that high-quality litter inputs increase rates of nitrification. The negative indirect effects of pine abundance on nitrification mediated through soil properties were much stronger than the direct effect of understorey leaf traits.

5. Synthesis. The leaf economics spectrum is an important gradient of functional trait variation that has consequences for internal N cycling. Nitrification potential was more strongly linked to dominant leaf traits than to functional diversity, thereby lending more support to the mass ratio hypothesis. However, functional diversity still explained some of the observed variation in biomass production. Leaf traits can be used to understand how internal N cycling rates will be affected by changes in vegetation structure.


Ecosystem processes are regulated by multitudes of interacting factors. The principle abiotic factors that govern processes in terrestrial ecosystems have been well documented (Paul 2007), but research also indicates that biotic factors, such as community diversity and functional composition, are also important (Wedin & Tilman 1990; Hooper et al. 2005). Two contrasting hypotheses have been proposed to explain the mechanisms by which communities influence below-ground ecosystem processes (Mokany, Ash & Roxburgh 2008). The ‘diversity hypothesis’ proposes that the diversity of functional types of species affects ecosystem functioning through complementary use of resources (Tilman, Wedin & Knops 1996) or through richer arrays of chemical compounds involved in resource fluxes. The ‘mass ratio’ hypothesis proposes that species controls on ecosystem processes are in proportion to their relative input to primary production (Grime 1998). Here, each hypothesis is simultaneously evaluated in the context of nitrogen (N) cycling in a ponderosa pine forest.

The internal cycling of N is a critical function for terrestrial ecosystems since it accounts for about 88% of the global N demand of plants (Schlesinger 1997). This process includes the mineralization of organic N to ammonium, the oxidation of mineralized ammonium to nitrite and nitrate (nitrification), the microbial assimilation of these mineral forms of N by micro-organisms and plant uptake (Robertson & Groffman 2007).This study focuses on factors that influence nitrification because only a few studies have examined the relationship between plants and this step in the N cycle.

The major soil properties that limit nitrification are well documented, and these are included as potential factors that may regulate nitrification potential in the hypothesized model (Fig. 1). The amount of N and the ratio of carbon to nitrogen (C:N) in the soil will influence nitrification because if N is limiting relative to C, then N will be immobilized in the microbial biomass (Robertson 1982; Kaye & Hart 1997). The ratio of carbon to phosphorus (C:P) is also proposed to limit nitrification because nitrite-oxidizer growth rates appear to be P-limited (Purchase 1974; Haynes 1986). Soil pH < 4.5 appears to limit nitrifiers because of aluminium toxicity, and soil pH between 6 and 7 is considered optimal for nitrifiers (Haynes 1986). Nitrification rates are optimum at soil temperatures between 20 and 40 °C (Brady & Weil 1999) and at 50% water-filled pore spaces (Robertson 1982; Haynes 1986; Robertson & Groffman 2007). In ponderosa pine forests, these soil properties are influenced by gradients in two other factors: forest floor litter mass and soil texture (Fig. 1; Welch & Klemmedson 1975).

Figure 1.

 Initial hypothesized structural equation model used to evaluate the relative importance of soil versus plant community effects on nitrification potential. The six soil factors that are enclosed in the dashed box were screened using multiple regression prior to fitting the model to eliminate weak predictors and reduce model complexity.

In addition to these abiotic soil factors, the plant community itself may exert additional influences on nitrification rates. The ‘mass ratio hypothesis’ predicts that communities dominated by plants with high leaf N content will increase internal N cycling rates. Interspecific variation in leaf physiology, morphology and chemistry, known as the ‘leaf economics spectrum’ (Wright et al. 2004), spans a gradient from slow-growing species with low leaf N content to fast-growing species with high leaf N content. This fundamental spectrum of leaf litter quality can influence the structure and functioning of soil microbial communities (Bardgett et al. 1999; Griffiths et al. 1999). Litter decomposition rates of species in this study are highest in species with high specific leaf area (SLA) and high leaf N content (Laughlin et al. 2010a), which suggests that soils beneath a community dominated by such plants may have faster rates of decomposition, mineralization and nitrification (Pastor et al. 1984; Quested et al. 2007; Cornwell et al. 2008). Orwin et al. (2010) recently demonstrated that monocultures of grassland plants with high tissue N content were associated with higher soil nitrate production.

The diversity hypothesis predicts that functionally diverse plant communities can positively influence soil microbial communities by providing a richer array of plant compounds through leaf litterfall, root turnover and root exudates. More diverse communities could affect mineralization and nitrification rates by providing a consistent long-term supply of organic N as the different litter qualities break down at different rates over time. Plant diversity has been shown to be positively related to microbial biomass (Spehn et al. 2000; Stephan, Meyer & Schmid 2000; Zak et al. 2003), N mineralization (Zak et al. 2003; West, Hobbie & Reich 2006) and N retention (Tilman, Wedin & Knops 1996). Understorey species richness was positively associated with nitrification potential in a restored ponderosa pine forest, but the composition of the understorey was also an important factor (Laughlin et al. 2010b). Idiosyncratic or negative responses of ecosystem processes to plant diversity have also been detected (Wardle et al. 1999; Niklaus et al. 2001; Schmid, Joshi & Schlapfer 2002; Carney, Matson & Bohannan 2004).

The primary objective of this study is to simultaneously evaluate the mass ratio and diversity hypotheses by including community-weighted mean leaf traits and a functional diversity index in a multivariate structural equation model (SEM) of nitrification potential (Fig. 1).The mass ratio hypothesis was tested by calculating the mean score of a community along the leaf economics spectrum, weighted by the relative contribution of each species to the total cover of the community. The mass ratio hypothesis would find support if the community-weighted mean leaf trait axis is significantly related to nitrification potential after statistically controlling for abiotic soil factors. The diversity hypothesis will be supported if functional diversity is a significant predictor of nitrification potential after accounting for abiotic soil factors.

Materials and methods

Study system

The ponderosa pine (Pinus ponderosa C. Lawson var. scopulorum Engelm.) forest ecosystem covers approximately 3.5 million ha of land across uplands in the south-western United States. This study was conducted on a c. 12 000-ha landscape on the Coconino National Forest in northern Arizona between the elevations of 2000 and 2500 m a.s.l on relatively flat sites. Ponderosa pine is the dominant tree species and forms extensive pure stands, but sometimes occurs with Gambel oak (Quercu sgambelii Nutt.). The mean annual precipitation of Flagstaff, Arizona, is c. 560 mm and the mean annual temperature is 7.7 °C. Vegetation and soil data were collected in 2007, when annual precipitation was c. 420 mm, which is 25% below the long-term average.

Vegetation data

Vegetation data was collected on eighty-two 1-m2 quadrats located across the study area (see Table S1 in Supporting Information for the data set used in the analysis). Quadrats were located within a range of soil types developed in basalt (= 63), limestone (= 9) and limestone/sandstone (= 10) parent materials. These plots were originally established in the early 1900s as a study in range management and were recently used to understand how understorey plant strategies and functional diversity have changed over the past century (Laughlin, Moore & Fulé 2011). These plots are used here to understand the multivariate relationships between vegetation structure, soil properties and nitrification potential. Above- and below-ground vegetation data were collected at peak standing crop in late August and early September 2007. Visual estimates of foliar cover were made for every plant species that occurred in each quadrat. Only one shrub (Rosa woodsii) was detected in the plots and it only occurred in a few plots at low relative abundance; all other species were herbaceous plants. Above-ground herbaceous biomass was measured by clipping all living vegetation on four 0.25-m2 subplots located 2 m from the centre of the 1-m2 chart quadrats at each of the four cardinal directions. These subplots were used to avoid destructive sampling within the historically valuable quadrats. The relationship between log-transformed total foliar cover in the quadrat and above-ground biomass measured in the subplots was positive and linear (R2 = 0.74, < 0.0001). I also collected three soil cores (4 cm diameter by 15 cm deep) located just outside the perimeter of the quadrats to quantify below-ground herbaceous fine root biomass. I used a hydropneumatic elutriator (Scienceware, Pequannock, NJ, USA) to separate the fine roots (<2 mm) from the soil. Herbaceous roots were easily distinguished and separated from ponderosa pine roots based on colour and morphology (the dark and suberized pine roots exhibit dichotomous branching). Plant material was dried at 55 °C for 72 h prior to weighing. Total herbaceous biomass at each plot was calculated as the sum of above-ground herbaceous shoot mass and below-ground herbaceous fine root mass. While I recognize that a single observation of fine root mass is not a measure of annual below-ground production, I included fine roots in my estimate of total herbaceous biomass since fine root mass in this ecosystem increases monotonically to a maximum in the fall (Kaye & Hart 1998).

Overstorey trees were measured on 20 × 20 m plots centred on each quadrat. This plot size ensured that most trees that could influence the light regime and microclimate of the quadrat were included in the plot. Forests were composed primarily of a ponderosa pine overstorey; only a few plots contained Gambel oak. For each tree (live and dead), I recorded its species and d.b.h. (1.37 m). These measurements were used to determine live tree basal area (m2 ha−1), which was used as a measure of pine abundance.

Soil data

I collected three soil samples with round soil probes (15 cm deep by 5 cm wide) from three locations approximately 2 m from the centre of each 1-m2 chart quadrat. Soil samples were composited, air-dried, sieved through a 2-mm sieve and analysed for texture using the hydrometer method (Dane & Topp 2002).

I measured soil temperature and moisture content simultaneously at each plot at four times between early July and late August, corresponding to the warmest and wettest part of the growing season in northern Arizona. Measurements were made between 10:00 and 14:00 hours to avoid diurnal effects on soil microclimate. These four measurements were averaged to obtain mean growing season soil temperature and moisture content. I used thermometers (VWR International Inc., Batavia, IL, USA) with probes to measure soil temperature at 5 cm depths. Soil water content was estimated using the hand-held Moisture Meter HH2 with the Theta Probe ML2x (Delta-T Devices, Cambridge, UK) that extended 6 cm into mineral soil. This device senses the apparent dielectric constant of the soil to estimate volumetric water content (%), and estimates were integrated along the entire 0–6 cm soil depth. Coarse-textured soils had lower water content than fine-textured soils. Despite different water-release curves for soils of different particle size distributions, the differences in water content between coarse- and fine-textured soils were large enough for the water potential to still be more negative (i.e. less available) in the coarse-textured soils than in the fine-textured soils (Saxton & Rawls 2006).

Total soil N and C were determined using a Thermo-Finnigan Deltaplus Advantage gas isotope-ratio mass spectrometer (Thermo Scientific, West Palm Beach, FL, USA) interfaced with a Costech Analytical ECS4010 elemental analyser (Costech, Valencia, CA, USA) at the Colorado Plateau Stable Isotope Laboratory at Northern Arizona University in Flagstaff, Arizona, USA. I used total soil N and soil C:N mass ratio as predictors of nitrification potential because total N is more stable than instantaneous NH4+ pool sizes. While the availability of ammonium (NH4+) will directly limit nitrification rates (chemolithotrophic nitrifiers obtain energy by oxidizing NH4+ and NO2), the instantaneous available pool of NH4+ may not be a good predictor of nitrification potential because this pool is transient throughout a growing season. Indeed, contrary to expectations, the NH4+ pool size throughout the growing season was not correlated with nitrification potential in a nearby ponderosa pine forest (= 0.72; data from Kaye & Hart 1998). Therefore, I used total soil N and soil C:N mass ratio as predictors of nitrification potential in this study.

I determined P concentrations from mineral soil samples following the general methodology of Bowman & Moir (1993) described in Lajtha et al. (1999). First, 0.5 g of ground (<1 mm) dry soil were placed in 25 mL of extracting solution (0.25 mol L−1 NaOH in 0.05 mol L−1 Na2EDTA). These extracts were heated in a water bath for 2 h at 85 °C. After filtering the supernatant, one aliquot was reserved for determination of inorganic P. Total extractable P in 10-mL aliquots of the extract was determined after acidified (0.9 mol L−1 H2SO4) potassium persulfate digestion in an autoclave for 90 min at 121 °C. Concentration of PO4-P was determined on a Lachat Instruments QuickChem 8000 Flow Injection Autoanalyzer (Lachat Instruments, Loveland, CO, USA) using a molybdate-ascorbic acid method (Lachat Instruments, Inc. 1992, QuickChem method 10-115-01-1-A). I used total extractable P in the calculation of the soil C:P mass ratio.

I determined forest floor mass (i.e. O horizon; hereafter, ‘litter mass’) by collecting two ‘cores’ of litter on the east and west sides of each quadrat using a 15-cm diameter steel pipe. Litter was dried at 55 °C for 72 h prior to weighing. I corrected for mineral content of the litter via loss-on-ignition (Dane & Topp 2002). I ground subsamples of litter in a Wiley Mill (Thompson Scientific, Swedesboro, NJ, USA) to <1 mm, combusted the subsamples at 550 °C for 6 h, then multiplied the original dry weights by the proportion of organic matter in the combusted subsample.

Nitrification potential was assessed using the soil slurry method (Belser 1979; Hart et al. 1994) on soils sampled from the upper 15 cm of the mineral soil in August 2007, when nitrifier activity is expected to be at a maximum because of warm and moist soil conditions during this period (Kaye & Hart 1998). This method assesses the Vmax of nitrification for a soil sample under conditions of optimal water content, NH4+, aeration and P availability. Therefore, the factor that is assumed to limit nitrate production under these conditions is the size of the nitrifier community. This is why nitrification potential, using this assay, is an index of the size of the nitrifier community (Belser 1979; Davidson, Stark & Firestone 1990). Soil samples of 15 g were placed in Erlenmeyer flasks with 100 mL of a neutral pH solution of 1.5 mmol L−1 of NH4+ and 1.0 mmol L−1 of PO43−. Soil slurries were shaken on an orbital mixer, and aliquots of 10 mL were drawn at 2, 4, 22 and 24 h. These samples were analysed for NO3on a Lachat AEFlow-Injection Analyzer (Lachat Instruments, Inc., Milwaukee, WI, USA), using the cadmium reduction-diazotization method (Lachat Instruments, Inc. 1992). The slope of the best-fitting line between the four sampling periods yields an estimate of nitrate production per unit dry mass of soil per unit time (Hart et al. 1994). Although some nitrification has been shown to be caused by heterotrophic soil fungi (Laughlin et al. 2008), I assume that most nitrification in this system is autotrophic given the near-neutral pH of the soils.

Functional traits

I measured a core set of functional traits on all plant species detected on these plots, including SLA, leaf dry matter content (LDMC), leaf N content (hereafter, leaf [N]), fine root [N], seed mass, specific root length (SRL), canopy height and mean Julian flowering date. Trait measurement methodology is fully described in Appendix S1.

As species richness and functional group richness do not adequately account for the diversity of multiple functional traits in the community, various indices of functional diversity have recently been proposed. I used ‘quadratic entropy’ (FDQ), which is the sum of the distances between pairs of species in trait space, weighted by the product of their relative abundances (Botta-Dukát 2005). Weighting the pairwise distances causes the index to increase as the most dominant species in a community have increasingly different trait values. I used the eight functional traits listed in the previous paragraph in the calculation of FDQ using code written for R (version 2.6.0; R Development Core Team 2007) by Evan Weiher ( This index is also much less correlated with species richness than dendrogram-based indices, making it more suitable for studying the potential link between functional trait diversity and ecosystem processes. In this study, the correlation between species richness and FDQ was 0.56, whereas the correlation between species richness and Petchey & Gaston’s (2002) dendrogram-based functional diversity index (FD) was 0.94.

Traits of individual species were scaled to the community level by calculating ‘community-weighted mean’ leaf trait axis scores, which are community trait means weighted by the relative abundance of each species. Laughlin et al. (2010a) used principal components analysis (PCA) to extract the major gradients of functional variation from a species × trait matrix. The first principal component represented variation along the leaf economics spectrum (Wright et al. 2004) because leaf [N], SLA, LDMC and fine root [N] loaded strongly on this axis (Laughlin et al. 2010a). Fine root [N] is often positively correlated with leaf [N], which is why it is included here within the leaf economics spectrum (Kerkhoff et al. 2006). The leaf economics spectrum represents broad species-level variation in leaf physiological processes (e.g. photosynthesis and respiration), leaf morphology (e.g. SLA, LDMC) and leaf chemistry (e.g. leaf [N]). Therefore, rather than treating each trait separately, I used the first principal component (i.e. a linear transformation of the traits) to represent species-level variation along this fundamental plant strategy spectrum. To scale the species-level leaf trait axis scores to the community level, I calculated community-level means for each plot (Garnier et al. 2004) weighted by the relative abundance of each species: inline image, where pi is the relative foliar cover of species i, S is the total number of species, and axisscorei is the axis score of species i along the first PCA axis from Laughlin et al. (2010a), i.e. the leaf economics spectrum. This straightforward technique permits one to directly regress an ecosystem process on a multidimensional plant community attribute. For simplicity, I refer to this community-level quantification of the leaf economics spectrum as ‘leaf traits’ throughout the paper. This quantification operationalizes Grime’s (1998) mass ratio hypothesis because the functional traits of the species are weighted by the relative contribution of each species to the total cover.

Data analysis

I used observed-variable structural equation modelling (SEM) to gain a multivariate perspective about how soil and plant community factors interact to influence nitrification potential. SEM is an extension of regression and path analysis that can be used to model multivariate relations and to evaluate multivariate hypotheses (Grace 2006). The model illustrated in Fig. 1 represents what I hypothesized to be the most plausible structural relations based on a priori knowledge. I acknowledge that not all causal processes that act in this system are represented in Fig. 1. Rather, my objective was to determine whether the data were consistent with the expectations of the proposed model.

Continuous soil variables were used in lieu of categorical soil types (i.e. nominal categories of soils derived from different parent material) because discriminant analysis showed that continuous soil variables in the SEM predicted the correct soil parent material 93% of the time (see Appendix S2). As a result of this finding, I used continuous soil variables in lieu of nominal soil types to represent the relationship between soil properties and other variables in the model.

I used a maximum likelihood estimator that is robust to non-normality and the Satorra–Bentler chi-square to evaluate model adequacy with Mplus software version 3.12 (Muthén & Muthén 2005). Model fit statistics evaluate the discrepancy between the covariance structure of the observed data and the covariance structure implied by the model. Therefore, good-fitting models yield small chi-square values and large P-values (> 0.05), indicating no significant difference between model and data.

Good-fitting SEMs do not prove causal relationships (Bollen 1989). Inferences about the sign and strength of directional paths in SEM can only be made if sound theory guides both the model-building and the model-fitting processes (Grace 2006). Ultimately, my goal was to arrive at a model consistent with the data using the fewest modifications of the initial model as possible, thereby preserving the ability to draw inferences from model parameters. The final SEM predicts a covariance structure that is consistent with the covariance structure of the data set; therefore, theory guides the interpretation of the mechanistic nature of the directional paths.


Linear bivariate relationships between nitrification potential and the soil and plant community factors hypothesized to influence nitrification potential are shown in Fig. 2. Nitrification potential was positively associated with soil temperature, soil total N, soil pH, functional diversity, total herbaceous biomass and leaf traits (Fig. 2), where ‘leaf traits’ represents the community’s average location on the leaf economics spectrum. Nitrification potential was negatively associated with soil C:N mass ratio, litter mass, soil C:P mass ratio and pine density (Fig. 2). Nitrification potential did not exhibit significant linear bivariate correlations with sand content or soil moisture (Fig. 2) and did not differ between limestone, sandstone or basalt-derived soil (Appendix S2).

Figure 2.

 Linear bivariate relationships between nitrification potential and (a) soil temperature, (b) soil total nitrogen, (c) soil carbon to nitrogen mass ratio, (d) litter mass, (e) soil sand content, (f) soil carbon to phosphorus mass ratio, (g) soil moisture, (h) soil pH, (i) functional diversity (FDQ), (j) pine density, (k) community-weighted mean leaf traits (i.e. a gradient in specific leaf area, leaf dry matter content, and leaf and fine root [N]; see Materials and methods for complete explanation of the metric) and (l) total (above- and below-ground) herbaceous biomass. = 82 plots.

Prior to the evaluation of the initial SEM, I evaluated a multiple regression model to determine which of the six abiotic soil properties (i.e. the variables enclosed in the dashed box in Fig. 1) were most important to nitrification potential. Soil sand content and pine litter mass were not included in this model because these variables were hypothesized to be indirect drivers of other soil properties (Fig. 1). This preliminary analysis determined that soil C:P mass ratio and soil moisture were not significant predictors (> 0.05) of nitrification potential in a model that included four significant factors (< 0.05): soil temperature, soil total N, soil C:N mass ratio and soil pH. Therefore, to simplify the analysis, I dropped soil moisture and soil C:P mass ratio from the model.

The new, simplified SEM did not fit the data well (χ2 = 51.1, d.f. = 31, = 0.0129; low P-values indicate significant discrepancy between model and data). Several pathways shown in Fig. 1 were found to be non-significant and were dropped from the model. These dropped pathways included: total herbaceous biomass to nitrification potential, functional diversity to nitrification potential, sand to soil C:N mass ratio, litter to soil N and pine basal area to functional diversity. The residual covariance matrix indicated two other model–data discrepancies, so I also added two additional pathways: one from leaf traits to total herbaceous biomass and one from nitrification potential to total herbaceous biomass. These changes suggest that nitrification potential has a stronger influence on biomass production than vice versa. This new model, illustrated in Fig. 3, fit the data well (χ2 = 42.4, d.f. = 32, = 0.10) and all pathways in the model were significant (< 0.05). This model explained significant variation in nitrification potential (R2 = 0.71), litter mass (R2 = 0.57), soil temperature (R2 = 0.65), soil C:N mass ratio (R2 = 0.54), soil total N (R2 = 0.30), total herbaceous biomass (R2 = 0.41), soil pH (R2 = 0.48) and leaf traits (R2 = 0.17). Model results were stable and qualitatively similar if the community-weighted mean leaf trait axis was replaced with either community-weighted mean leaf [N] or SLA (see Appendix S3), suggesting that these individual traits could be used as predictors of nitrification potential.

Figure 3.

 Final structural equation model with standardized path coefficients (χ2 = 42.4, d.f. = 32, = 0.10). All path coefficients are significant (< 0.05) and the sizes of the arrows are proportional to the strength of the relationships. Coefficients can be interpreted as partial correlation coefficients that range from −1 to +1. Coefficients of determinisms (R2) are shown for every response variable in this set of eight linear equations.

Explained variance of leaf traits was low, but two pathways to leaf traits were significant. Soil sand content tended to increase community-level SLA, leaf and root [N], and decrease LDMC. Litter mass tended to decrease community-level SLA, leaf and root [N], and increase LDMC (Fig. 3). No other soil properties in the model were associated with variation in leaf traits.

Nitrification potential was more strongly associated with soil factors (e.g. soil pH, soil temperature, soil total N and soil C:N mass ratio) than plant community factors (Fig. 3). After accounting for the variation explained by the soil factors, community-weighted leaf traits were significantly associated with nitrification potential (Fig. 3). Despite significant positive correlations between nitrification potential and both functional diversity and herbaceous production (Fig. 2), neither were important predictors in the context of the multivariate model (Fig. 3).

Relationships with nitrification potential can be understood as either direct or indirect. Neither pine density nor litter mass had direct relationships with nitrification potential (Fig. 3). However, pine density exhibited a significant negative ‘total indirect effect’ on nitrification potential (standardized coefficient = −0.49) mediated through cascading effects on litter mass and other soil properties (Fig. 3). Litter mass exhibited a significant negative ‘total indirect effect’ on nitrification potential (standardized coefficient = −0.64) mediated through effects on soil C:N mass ratio, soil temperature, soil pH and leaf traits (Fig. 3). Interestingly, despite no significant bivariate relationship between soil sand content and nitrification potential (Fig. 2), sand content exhibited three indirect pathways to nitrification potential mediated through positive effects on soil temperature and soil pH, and through the offsetting negative effect on soil total N.


Nitrification potential was more strongly linked to dominant leaf traits than to functional diversity, thereby lending more support to the mass ratio hypothesis (Grime 1998). This suggests that herbaceous plant species’ controls on nitrification in ponderosa pine forests are in proportion to their relative input to total abundance, and that the diversity of leaf litter quality in a community may not be important for this critical link in the internal cycling of N. Plant communities dominated by species with high leaf [N] and root [N], high SLA and low LDMC (i.e. fast-growing species) are associated with soils that have high nitrification potential. Importantly, this result was obtained after statistically controlling for soil factors, which exerted strong direct control over nitrification potential. These results are in agreement with Orwin et al. (2010) who demonstrated a positive relationship between plant tissue [N] and both soil nitrate and net nitrification. The average location of a plant community along the leaf economics spectrum has consequences for below-ground ecosystem functioning.

Despite finding no support for a relationship between functional diversity and nitrification potential, functional diversity was positively associated with total herbaceous biomass. This suggests that there may be positive diversity benefits to community productivity, possibly resulting from complementary use of resources (Tilman, Wedin & Knops 1996), but it is difficult to rule out the possibility of a sampling effect (Huston 1997). It is possible that functional diversity may be more important for certain ecosystem processes (e.g. net primary productivity) than others (e.g. N cycling).

Interestingly, model results suggest that total herbaceous biomass was positively influenced by nitrification potential, suggesting that nitrate production (which can be influenced by dominant leaf traits) may, in turn, influence community biomass production. This plant–soil feedback could be important in any semi-arid ecosystem, such as ponderosa pine forest, where leaching of the mobile nitrate ion is limited compared with mesic ecosystems (Kaye et al. 1999). Such feedbacks are important to our understanding of how ecosystem processes will be affected by future vegetation change (Orwin et al. 2010).

Another important implication of the SEM is that ponderosa pine basal area is an important indirect driver of nitrification potential in these forest soils. It has been shown that nitrification rates are generally lower in conifer forests than in forests dominated by broad-leaved species that drop high quality litter (Pastor et al. 1984; Finzi, Van Breeman & Canham 1998). Pine dominance causes an increase in the mass of recalcitrant pine needles on the forest floor (Hunt et al. 1988). Increased litter mass reduces soil temperatures and increases the mass ratio of C:N in the mineral soil. Moreover, pine needles tend to acidify the mineral soil. Dense pine stands also have more pine roots in the surface soil (the correlation between pine basal area and log-transformed pine root mass per area is = 0.54, < 0.0001). Given that pine roots have a relatively high C:N mass ratio of 40 (Laughlin et al. 2010a), pine roots may also increase the mineral soil C:N mass ratio. Overall, pine density alters the abiotic soil environment in multiple ways, thereby indirectly controlling nitrification potential. Thus, the direct positive effects of understorey leaf traits on nitrification potential are weaker than the negative indirect effects of the overstorey. Ponderosa pine density is the most important indirect driver of nitrification potential in these dry montane forest soils, lending further support to the mass ratio hypothesis given that ponderosa pine can account for >80% of the total leaf area at the stand scale in these forests (Dore et al. 2010).

Fire suppression policies and other land uses in the western United States have led to widespread increases in ponderosa pine and Douglas-fir density in dry lower montane forests (Allen et al. 2002; DeLuca & Sala 2006). These changes in forest structure have led to significant shifts in the functional composition of the understorey plant community (Laughlin, Moore & Fulé 2011). Historically, open stands of large ponderosa pine contained productive, graminoid-dominated understoreys with higher herbaceous understorey canopy heights, higher SRLs, higher fine root and leaf [N] and later flowering dates (Laughlin, Moore & Fulé 2011). Thus, the increase in pine density has been associated with a shift toward stress- and shade-tolerant species. Along with this functional shift came deeper litter layers of recalcitrant pine needles. All of these changes have likely driven the observed declines in nitrification rates in these forest soils (Kaye & Hart 1998; DeLuca & Sala 2006), suggesting that long-term vegetation dynamics can have significant consequences for ecosystem functioning.

This study shows that the leaf economics spectrum is an important gradient of functional trait variation among plant species and communities in the ponderosa pine forest ecosystem. This trait spectrum has consequences for a key below-ground ecosystem process, such as N cycling. With respect to nitrification potential, the final SEM supports the mass ratio hypothesis, which proposes that plant species controls on ecosystem processes are in proportion to their relative input to primary production. Therefore, nitrification is more strongly linked to the dominant leaf traits in the community rather than the functional diversity of the community. Leaf traits of dominant species can be used to understand how internal N cycling rates will be affected by future changes in vegetation structure.


I thank M. Moore, P. Fulé, S. Hart, C. Sieg, M. Kearsley, P. Selmants, B. Sullivan, G. Newman and M. Luce for their helpful comments or contributions to the project. This research was supported by the Ecological Restoration Institute (ERI) at Northern Arizona University (NAU), a Joint Venture Agreement (#08-JV-11221633-233) with the U.S. Forest Service Rocky Mountain Research Station (RMRS) and McIntire-Stennis appropriations to the NAU School of Forestry.