Experimental manipulation of soil depth alters species richness and co-occurrence in restored tallgrass prairie


*Correspondence author. E-mail: dornbusm@uwgb.edu


1. In relatively fertile ecosystems, such as the tallgrass prairie, local species diversity is largely controlled by the competitive effects of community dominants. Despite the relative importance of soil resources in shaping competitive outcomes, we have a limited understanding of the ways by which plants partition below-ground space and resources while competing, and thus, how these interactions feedback to affect local diversity.

2. We experimentally tested whether potential rooting depth affected plant species diversity and composition by seeding 36 tallgrass prairie species into replicated, bare-ground plots in which soil depth was manipulated to produce shallow- (20 cm), medium- (42 cm) and deep-soil treatments all within one soil type. Because root architecture and foraging strategies differ among species, we hypothesized that soil depth alone could affect plant richness, diversity and community composition.

3. After 3 years, richness (S) significantly increased with soil depth (< 0.0001), but there was no significant change in species diversity (> 0.1) or composition (multi-response permutation procedure, > 0.2). The lack of a depth effect upon diversity resulted from the opposing effect of increasing soil depth enhancing S, but decreasing evenness.

4. Species presence among depth treatments was strongly nested, with species found in shallow soils reflecting a subset of the species found in the medium-depth treatment, and the species found within the medium-depth treatment reflecting a subset of those found in the deepest soils.

5. All depth treatments contained the same dominant grasses, thus differences in S resulted from the nested loss of forbs. Conversely, increasing soil depth added sets of new species, but the specific identity of the species present appeared interchangeable among replicates of a given depth.

6.Synthesis. Our results provide the first field-based experimental evidence that altering soil depth alters species occurrence and diversity in predictable ways in seeded tallgrass prairie. Our results have important theoretical implications for understanding the processes promoting plant co-occurrence in grasslands, and generate testable hypotheses concerning the conditions under which root-niche partitioning is probably important for maintaining local richness in grasslands. Future work is needed to elucidate the generality and mechanistic basis of our results.


In relatively fertile ecosystems, like the tallgrass prairie of North America, competitive exclusion of subordinate species by community dominants contributes to conditions of reduced local diversity (Tilman 1984; Huston 1994; Baer et al. 2004; Hautier, Niklaus & Hector 2009). In response to this understanding, researchers have begun to investigate the role of resource availability and partitioning in promoting high local diversity in remnant stands, and reciprocally their potential relevance for achieving diverse restorations (Baer et al. 2003, 2004, 2005; Blumenthal, Jordan & Russelle 2003; Perry, Galatowitsch & Rosen 2004). Despite the central role that soil resources play in altering competitive outcomes among co-occurring species, we have a rather limited understanding of the ways in which plants partition below-ground space and the resources held within that space while competing.

The role of below-ground niche separation in promoting species coexistence and its effect on ecosystem processes in grasslands have received previous attention. Weaver (1958) and Sun, Coffin & Lauenroth (1997) detailed differences in root distributions among North American grassland species and discussed the potential implications that these differences in root architecture have on the ability of these species to tolerate environmental stresses and coexist. Parrish & Bazzaz (1976) found greater below-ground niche partitioning in the root systems of North American perennial grasslands than in evolutionarily more nascent North American ruderal communities, commenting that these differences probably help to promote diverse grassland communities at the local scale. More recently, root distributions of grassland species have been directly linked to the acquisition of soil nitrogen (N) (Craine et al. 2002) and water (Craine et al. 2002; Nippert & Knapp 2007a,b). Nippert & Knapp (2007a,b) further documented that the vertical segregation of soil resources by plants varies temporally with resource availability. In general, these studies highlight the potential importance of species-specific differences in root architecture and foraging strategies for understanding community composition.

The physical construct of a plant’s below-ground environment also affects plant performance. McConnaughy & Bazzaz documented species-specific differences in growth and reproductive output of ruderal plants in response to changes in available soil volume (McConnaughay & Bazzaz 1991) and to changes in the fragmentation of soil space by small-diameter inert objects (McConnaughay & Bazzaz 1992). At the community level, Dimitrakopoulos & Schmid (2004) and von Felten & Schmid (2008) found that increasing soil volume enhances biodiversity effects on plant production. While there is accumulating evidence that below-ground niche partitioning is important for community and species-specific production, we retain a limited understanding of the reciprocal effect that differences in the physical below-ground environment have on grassland community diversity and composition.

Observational studies from the tallgrass prairie provide some insights into questions regarding the impacts of differences in below-ground environments on local community diversity. For example, plant-species richness varies predictably along topographic soil-depth gradients in the tallgrass prairie (Bliss & Cox 1964; Barnes & Harrison 1982; Barnes, Tieszen & Ode 1983, Abrams & Hulbert 1987; Gibson & Hulbert 1987). Specifically, deep-soil lowlands consistently support fewer species than adjacent, shallow-soil uplands (Abrams & Hulbert 1987; Gibson & Hulbert 1987; Hartnett & Fay 1998). The lower richness and diversity of lowland communities is generally attributed to higher soil fertility and the associated higher graminoid production, which is hypothesized to subsequently drive the competitive exclusion of subordinate forbs (Abrams & Hulbert 1987; Huston 1994; Turner & Knapp 1996). Thus, observational studies from the tallgrass prairie suggest that differences in soil fertility and rates of competitive exclusion of forbs by grasses explain observed differences in species diversity between shallow and deep soils. However, a number of environmental variables simultaneously change along natural soil depth-richness gradients, and we suggest that it remains unclear how soil depth, independent of other confounding factors, affects local community processes. For example, topographic gradients are associated with concomitant changes in rooting depth (Abrams & Hulbert 1987; Gibson & Hulbert 1987), soil texture and nutrient availability (Schimel, Stillwell & Woodmansee 1985; Schimel et al. 1991; Turner et al. 1997), soil moisture (Barnes & Harrison 1982; Barnes, Tieszen & Ode 1983), plant production (Schimel, Stillwell & Woodmansee 1985; Abrams, Knapp & Hulbert 1986; Belcher, Keddy & Twolan-Strutt 1995) and even seed predation (Reed, Kaufman & Kaufman 2004). The objective of this study was to explore the role that potential rooting depth may play in modulating community diversity for a given site fertility. For example, viewed from the perspective of niche complementarity, one could predict the exact opposite diversity trend along soil-depth gradients, where a greater potential rooting volume in deeper soils could support a more species-rich community via greater niche partitioning of root-foraging depths (Weaver 1958; Parrish & Bazzaz 1976; Berendse 1979; van Auken, Bush & Diamond 1994, Craine et al. 2002; Dimitrakopoulos & Schmid 2004; Fargione & Tilman 2005). From a community assembly perspective, one could likewise expect changes in community composition between shallow and deep soils to reflect differences in species’ root architectures. For example, within the tallgrass prairie, one could expect that shallower soils would be dominated by shallow-rooting species, such as sod-forming grasses, with a notable reduction in deeper-rooting subordinate forbs (Weaver 1958; Sun, Coffin & Lauenroth 1997). Support for a niche-based view of the effects of soil depth on plant communities appears partially supported by the high species richness of deep-soil tallgrass prairies found across the Midwest in the USA. For example, Dornbush (2004) reported 17.8 species m−2 and Martin, Moloney & Wilsey (2005) reported 19.0 species 0.4 m−2 from deep soil (> 1 m), high-productivity remnant prairies in Iowa. For these reasons, the direct effect of soil depth on local species diversity remains unclear.

In this experiment, we directly tested whether potential rooting depth affects plant species diversity and composition by physically manipulating soil depths to produce shallow- (20 cm), medium- (42 cm) and deep-soil (complete soil profile) treatments within one soil type. We suggest that variation in rooting depth alone could contribute to differences in grassland diversity (α-diversity) and composition at small spatial scales. Specifically, we hypothesized that plant diversity would increase with increasing soil depth, in agreement with niche complementarity theory. In addition, we predicted that shallower soils would support fewer forb species than deeper soils. This assertion comes from predictions by Weaver (1958), who hypothesized that forb species avoid direct competition with grasses by producing foraging roots below the active rooting zones of dominant grasses. Finally, we predicted a concomitant change in species composition along the soil-depth gradient, as the pool of species capable of surviving under the various rooting depths changed. By seeding an experimental tallgrass prairie community, we were able to control external propagule supply, species additions, and soil texture and nutrient concentrations among our depth treatments. Thus, we were able to directly test the effects of soil depth on tallgrass prairie plant communities without the confounding factors associated with natural soil-depth gradients. Because we focus on plant community diversity in this study, we held plot area constant, and thus acknowledge that there is an inseparable, confounding relationship between soil depth and soil volume in this study, but stress that this confounded structure is also most relevant for understanding soil-depth effects on plant community diversity and composition.

Materials and methods

This experiment was conducted from May 2003 to September 2006 at a private property in Story County, Iowa, USA (42°2.76′ N, 93°31.25′ W). Soil was a clarion loam, a fine-loamy, mixed mesic Typic Hapludoll that developed within glacial till under tallgrass prairie (DeWitt 1984). Prior to establishing experimental plots, the site had been maintained as a mowed, smooth brome (Bromus inermis Leysser) dominated field. The study site averaged approximately 890 mm of rain annually in a unimodal pattern with a mean annual temperature of 8.5 °C. Approximately 72% of annual precipitation falls from April through September (http://mcc.sws.uiuc.edu/). In close agreement with long-term site averages, annual precipitation during the years presented in this paper averaged 887 ± 27 mm (mean ± 1 SE) and growing season precipitation accounted for 70.8 ± 2.4% of annual precipitation.

We applied three discrete rooting-depth treatments to evaluate the effect of soil depth on plant community composition and diversity: a shallow-depth treatment with 20 cm rooting depth, a 42-cm-deep medium-depth treatment, and a deep treatment with unrestricted rooting depth. Soil depths were chosen to range from the shallow, rocky upland soils to the deep bottomland soils of Konza tallgrass prairie (Ransom et al. 1998) and the deep-soil mesic prairies of Iowa (DeWitt 1984). Depth treatments were constructed using large plastic tubs with a planting area of 0.17 m2. The bottoms of all three treatment tubs were removed to facilitate normal drainage, with the exception of two 3-cm-wide cross-supports and a lip around the bottom outer edge. Tub bottoms in the shallow- and medium-depth treatments were then resealed to restrict root growth by gluing 150-μm polyethylene mesh to the remaining cross-supports and lip with silicon caulk. The polyethylene mesh was then reinforced by attaching 1.27 cm galvanized wire mesh to the outside bottom of the tubs. The use of open bottoms in the deep-soil treatment may have altered resource availability by allowing the lateral movement of water and resources, or by allowing lateral root spread below 42 cm. Tubs were buried level with the soil surface, and refilled with the same soil that was removed to insert the tubs, with the exception of the former smooth brome sod, which was discarded (c. 4 cm thick). Soil was allowed to settle for several days, and then tubs were seeded with 36 native prairie species at a rate of 19 700 seeds m−2 (see Appendix S1 in Supporting Information). This seed rain approximates the natural seed rain for tallgrass prairie reported by Rabinowitz & Rapp (1980). Non-seeded species were removed by hand weeding throughout the experiment.

The entire experimental design was replicated in two consecutive years (i.e. blocks). The first block was established in May 2003 and the second block was established in May 2004. The 2003 and 2004 plantings were established in adjacent areas, approximately 2 m apart. The three depth treatments were replicated five times in the 2003 planting (= 15 plots total) and four times in the 2004 planting (= 12 plots total).

Plant species richness (S) and abundance were assessed by stem counts in June and September 2003, October 2004, and June and October 2005 in the 2003 planting, and in June and October 2004, June 2005, and June and September 2006 in the 2004 planting. After the third complete growing season, species richness and abundance were determined by clipping and removing plant biomass from all plots in early October 2005 and late September 2006 for the 2003 and 2004 plantings, respectively. Clipped samples were returned to the laboratory, litter was removed, and live plants were sorted by species, dried at 65 °C for 48 h and weighed. For each plot and sample date we recorded species richness, Simpson’s Reciprocal Diversity (inline image; where pi is species relative abundance) and evenness [E = (1/D)/S; where S is species richness). Canopy transmittance (%) of photosynthetically active radiation (PAR) was determined by dividing PAR measured above the soil litter by PAR measured above the canopy, following canopy closure in June 2006 using a LI-COR LI-190 Quantum Sensor (LI-COR, Lincoln, NE, USA).

Temporal changes in species richness (S) among soil-depth treatments, as determined from stem count surveys, were evaluated using Type IV SS from a general linear model repeated-measures anova analysed as a split plot to deal with differences in sample collection dates during the second growing season (i.e. the data set had two empty cells, one for each planting year during the second growing season). These data were also analysed without the second growing season, but this did not alter statistical conclusions so we discuss only results from the full data set. The model included planting year as a block (= 2), soil depth (= 3), time since planting (= 6) and the soil-depth-by-time interaction term. Non-significant interaction terms (> 0.05) were pooled into the residual error. Differences in richness among soil depths for a given sample date were subsequently analysed using one-way anova. Differences in clipped plant biomass, S, 1/D, and E among depth treatments after the third growing season were analysed using anova with planting year as a block (= 2) and soil depth (= 3) as the treatment. Biomass values were ln-transformed to meet normality and equal-variance assumptions. Canopy transmittance of PAR was square root-transformed to meet normality and equal-variance assumptions and analysed in an identical manner to biomass-based diversity measures. All analyses were conducted using sas version 9.1 (SAS Institute, Inc., Cary, NC, USA).

The degree of species nestedness among soil-depth treatments was assessed for each sample time by first creating a maximally packed species presence/absence matrix using the Nestdeness Temperature Calculator (Atmar & Patterson 1995). Each tub was then assigned a rank (1–27) equivalent to the packing rank assigned by the Nestdeness Temperature Calculator, which organizes all sites and species in a way that minimizes the occurrence of unexpected species presences and absences (Atmar & Patterson 1993). Following this procedure, increasing nestedness rank is generally associated with decreasing tub richness, and species assemblages that increasingly represent subsets of lower ranked tubs that have higher species richness. Differences in average nestedness rank among soil-depth treatments were then analysed using one-way anova for each sample time in sas version 9.1 (SAS Institute, Inc., Cary, NC, USA).

Patterns of community composition after the third growing season were analysed using relative abundance data in a multi-response permutation procedure (MRPP) with Bray–Curtis distance (PC-ORD version 4.25, MjM Software Design, Gleneden Beach, OR, USA). As with nestedness analyses, planting years were combined (= 27) for MRPP analysis.


Temporal trends in species richness and abundance

We observed 31 of the 36 seeded species in at least one of the tubs on at least one date. Initial seedling density (mean ± SE) in early summer of the first growing season did not differ significantly among depth treatments (= 2.8, = 0.08). The average number of seedlings in the first sampling date was 64.3 ± 7.1 in deep-soil tubs, 66.6 ± 4.8 in medium-depth tubs and 80.6 ± 7.2 in shallow-depth tubs. Thus, we found no evidence of a soil-depth effect upon seedling establishment at the start of our experiment.

Averaged across sample dates, species richness (S) was significantly affected by soil depth, time since planting, and the time × soil depth interaction (Table 1). Species richness did not differ (> 0.05) among soil depths on the first sample date of the first growing season, but differed among soil depths by the end of the first summer (Fig. 1). However, richness patterns among soil-depth treatments did not stabilize until the second growing season, and thereafter the general ranking of S among depth treatments remained consistent and in agreement with our niche-based hypothesis with richness highest in deep-soil tubs, intermediate in medium-depth tubs and lowest in shallow-depth tubs (Fig. 1). In general, as the treatments aged, richness in the deep-soil treatment remained comparable to maximum observed levels, while it declined through time in both the medium and shallow treatments (Fig. 1). Although not statistically significant (> 0.05), species loss trended higher in the shallow-soil depth than in the medium-soil depth treatment (Fig. 1).

Table 1.   Repeated-measures anova evaluating differences in species richness (S) between planting years, among soil depths and through time (= 6 dates) in seeded tallgrass prairie tubs located in central Iowa, USA. Measurements were taken over three growing seasons using stem counts
Source of variationd.f.FP
Planting year1, 232.5> 0.1
Soil depth2, 236.3< 0.01
Time5, 9312.8< 0.0001
Soil depth × Time10, 932.7< 0.01
Figure 1.

 Temporal changes in species richness (mean ± SE) among experimental soil-depth treatments seeded into tallgrass prairie.

Biomass-based comparisons after the third growing season

After three growing seasons, soil depth significantly affected above-ground plant biomass (= 14.7, < 0.0001; Fig. 2). Plant biomass was greatest in the deep-soil treatment, intermediate in the medium-depth treatment and least in the shallow-soil treatment (Fig. 2). Averaged across both planting years, forb biomass (mean ± SE) comprised 12% ± 6, 20% ± 8, and 31% ± 9 of total above-ground biomass in shallow, medium and deep treatments, respectively.

Figure 2.

 Above-ground biomass among experimental soil-depth treatments following the third growing season after seeding treatments tubs into tallgrass prairie. All values are mean ± SE; different letters represent statistically significant comparisons at  0.05.

In agreement with our initial hypotheses, both final species richness (S) (= 21.0, < 0.0001) and evenness (= 10.6, < 0.001) differed significantly among soil-depth treatments. Richness declined from deep (15.0 ± 0.6) to medium (11.1 ± 0.4) to shallow soil treatments (10.1 ± 0.7) (Fig. 3b). Evenness patterns among depth treatments were opposite to those of S, with E increasing as soils became more shallow (Fig. 3c). In contrast to our initial hypothesis, 1/D was unaffected by soil depth (= 2.0, = 0.2; Fig. 3a), and it appears that the non-significant effect of soil depth on 1/D probably resulted from the opposite effect that soil depth had on E and S (Fig. 3). These patterns appear related to changes among soil-depth treatments in above-ground plant biomass; S was positively correlated with ln plant biomass (Pearson product-moment correlation; = 0.39, < 0.05), E was negatively correlated with ln plant biomass (= −0.38, < 0.05), and there was no relationship between 1/D and ln plant biomass (r = −0.28, = 0.16).

Figure 3.

 Above-ground biomass based estimates of (a) Simpson’s reciprocal diversity, (b) species richness and (c) community evenness among experimental soil-depth treatments following the third growing season after seeding treatment tubs into tallgrass prairie. All values are mean ± SE; different letters represent statistically significant comparisons at  0.05.

Canopy PAR transmittance

Transmittance of PAR (%) was significantly affected by soil-depth treatments (= 7.1, < 0.002). As expected, PAR transmittance (mean ± SE) increased with decreasing soil depth from a low of 7.2% ± 1.9 in the deep-soil treatment, to 15.0% ± 2.0 in the medium-depth treatment, to 26.1% ± 5.1 in the shallow treatment. However, PAR transmittance only differed significantly between the deep- and shallow-soil treatments (Tukey–Kramer HSD: α = 0.05).

Nestedness and community composition after the third growing season

Nestedness rank did not differ (> 0.1) among depth treatments until early summer of the third growing season (= 4.4, < 0.03; Fig. 4). In agreement with predictions based on S, nestedness patterns had increased by the time of the late-summer sampling (= 18.6, < 0.0001), with differences changing from a significant difference in nestedness rank between shallow- and deep-soil treatments in early summer of the third growing season, to a significant difference in average nestedness rank between the shallow and medium and the deep treatment by late summer (Fig. 4). Nestedness trends among soil-depth treatments agreed with our hypothesis, with shallow-soil species representing a subset of species from the medium-depth treatment, and species in the medium-depth treatment consisting of a subset of species from the deep-soil treatment. In fact, the degree of nestedness among the depth treatments was so complete by the end of the third growing season that 17 of the 19 species found in the shallow-soil treatment were also found in the medium-depth treatment, and all of the species found in the medium-depth treatment were found in the deep-soil treatment (Fig. 5). There was only one species found in the shallow-soil treatment, Solidago rigida, that was not also present in deep-soil treatment and S. rigida was only found in a single, shallow-soil tub. The other species present in the shallow, but not medium treatment, was also rare (Anemone cylindrica); A. cylindrica was found in only two deep-treatment tubs, no medium-treatment tubs and one shallow-soil tub. Silphium laciniatum, Rudbeckia hirta, Viola pedata and Dalea candida were found only in the deep-soil treatment, whereas Desmanthus illinoensis, Lespedeza capitata, Solidago canadensis and Verbena stricta were found only in the medium- and deep-soil treatments. Thus, in agreement with our hypothesis, differences in soil depth generated significant differences in species occurances, despite the fact that all soil-depth treatments occurred within one soil type.

Figure 4.

 Temporal changes in the degree of nestedness (mean ± SE) among experimental soil-depth treatments seeded into tallgrass prairie. Increasing packing ranks correspond to treatments that increasingly support a subset of species found in other treatments.

Figure 5.

 Occurrence frequencies among soil-depth treatments for each of the 27 tallgrass species that remained within at least one treatment tub following the third growing season post-seeding.

Despite significant differences in both richness and nestedness among depth treatments, there was no significant difference in species composition among soil-depth treatments (MRPP, = 0.02, > 0.1). Stated in other terms, soil depth significantly affected species richness, and the species present in shallower soils were subsets of the species present in deeper soils, but the exact identity of the species present within each depth treatment was not consistent among replicates, with several species apparently interchangeable at a given depth. This is in contrast to our original hypothesis, which suggested that soil depth would be a strong predictor of community composition.


After three years of growth, we found strong empirical evidence that our depth treatments affected species richness, evenness and community assembly in seeded tallgrass prairie. Opposite responses by species richness and evenness to soil depth appear to have contributed to the non-significant effect of soil depth on plant diversity. Richness and above-ground biomass were positively correlated, and richness was not correlated with light at soil surface, suggesting that species losses among depth treatments were not driven by above-ground competition. Species losses with decreasing soil depth resulted from the loss of forbs, were non-random and were strongly nested among soil depths. Thus, all depth treatments contained the same dominant grasses and by increasing soil depth new sets of forbs appeared; however, the specific identity of the species within these sets appeared interchangeable among replicates of a given depth.

Because plant richness is often positively correlated with the number of individuals within an area (Oksanen 1996; Gotelli & Colwell 2001), differences in initial seedling establishment or the number of species that established among depth treatments could potentially produce lasting differences in species richness. However, this was not the case, as neither the number of initial seedlings or species that established differed significantly among depth treatments. Temporal divergences in richness, coupled with non-significant differences in initial seedling densities suggest that differences in species richness resulted from depth effects on species loss through time, rather than from initial differences in seed germination or species establishment among depth treatments.

In contrast to soil depth–species richness trends in remnant tallgrass prairies (Abrams & Hulbert 1987; Gibson & Hulbert 1987; Hartnett & Fay 1998), we found increasing species richness with increasing soil depth in our seeded prairies. Deeper soils have a greater total amount of moisture and nutrients (Schimel, Stillwell & Woodmansee 1985, Schimel et al. 1991; Turner et al. 1997), and are thus capable of supporting more plant biomass than drier, lower-nutrient, shallower soils (Schimel, Stillwell & Woodmansee 1985; Abrams, Knapp & Hulbert 1986, Belcher, Keddy & Twolan-Strutt 1995, Briggs & Knapp 1995). For this reason, field-based patterns of decreasing richness with increasing soil depth are generally explained in terms of increasing above-ground competition and competitive exclusion with increasing soil depth and fertility (Abrams & Hulbert 1987; Huston 1994; Turner & Knapp 1996). In agreement with this prediction, we found greater above-ground production and lower PAR transmittance in deeper soil treatments, and thus seemingly a greater potential for light-mediated competitive exclusion in our deeper-soil treatments. However, in our experiment species richness was positively correlated with above-ground plant biomass and not light availability, suggesting that light competition did not drive differences in species losses among depth treatments.

Baer et al. (2003) examined the effects of spatial heterogeneity in soil N and depth on diversity and productivity in seeded tallgrass prairie in Kansas, USA. In agreement with our results, their experiment failed to replicate previously reported tallgrass prairie diversity–soil depth patterns in remnants. However, in contrast to our experiment, they reported no soil-depth effects on richness, diversity or productivity. They suggested that their shallow-soil treatment (25-cm deep) was not sufficient enough to restrain above-ground production by dominant grasses and thus did not release forbs from competitive exclusion by the grasses. In our experiment, we found a positive relationship between richness and above-ground biomass, and production in our deep-soil treatment (908 g m−2) surpassed the 660 g m−2 of above-ground production reported by Baer et al. (2003), suggesting an even greater potential for light-mediated competition in our experiment. Our modest plot size (0.17-m2 circular plots) may have enhanced light availability in all depth treatments, irrespective of treatment specific alterations to PAR caused by differences in above-ground production. In addition, by holding soil type and landscape position constant among our soil-depth treatments, we probably moderated the normal, but confounding, changes in soil texture, soil nutrient content and soil moisture expected to further alter above-ground competition intensity along natural soil-depth gradients. These explanations may reconcile the positive relationship between above-ground biomass and richness that we report, with the strong negative relationship reported previously from natural soil-depth gradients (Abrams & Hulbert 1987; Gibson & Hulbert 1987; Collins et al. 1998; Hartnett & Fay 1998). Irrespective, combining the results of our experiment with previously published studies suggests the potential for a hierarchical filter controlling local species co-occurrence, whereby below-ground processes influence species coexistence only in the absence of strong light competition. The general pattern of increasing importance of below-ground interactions with decreasing above-ground biomass has long been suggested for grasslands along large precipitation gradients (Burke et al. 1998; Lane, Coffin & Lauenroth 2000).

We hypothesized that if altered soil depth affected species richness in a predictable manner, then species loss among soil-depth treatments should be nested and non-random. In agreement, the species present in shallower treatments were near perfect subsets of the species in deeper-soil treatments. Although we did not sample rooting depths in this study, the specific identity of the species lost with decreasing soil depth agreed with published descriptions of the species’ root architecture and hypothesized root-foraging strategies (Weaver 1958). For example, Weaver (1958) described Dalea candida, Dalea purpurea and Lespedeza capitata as taprooted plants with deep-growing branches that have minimal absorption in the top 30 cm of the soil. In agreement, D. candida was found only in the deep-soil treatment, L. capitata was found only in the medium and deep treatments, and D. purpurea was twice as common in deep, than shallow-treatment tubs. Silphium laciniatum and Echinacea pallida have deep, non-branching taproots hypothesized to forage below the maximum zone of grass-root uptake (Weaver 1958). Silphium laciniatum was found in 78% of deep-soil tubs, but was absent from all medium and shallow tubs. Echinacea pallida was found three times more frequently in deep and medium tubs, than in shallow-soil tubs. In contrast to clear differences in forb occurrence among depth treatments, seven of the eight grass species were present at nearly identical frequencies in all soil-depth treatments. Sun, Coffin & Lauenroth (1997) suggest that the subtle differences in root distributions among species within ecosystems explain local co-occurrence. Our results support this position, as changes in soil depth appear to have altered the occurrence of the variously rooted forbs that contributed most to species richness, but did not affect the occurrence of dominant grasses.

Briggs & Knapp (1995) found that forb production (NPP) in tallgrass prairie was unrelated to precipitation and evapotranspiration, but rather was strongly and negatively related to grass NPP, concluding that forbs are more susceptible to the biotic effects of competition with dominant grasses, than to abiotic factors such as variations in precipitation. In agreement, Nippert & Knapp (2007a,b) found evidence for the partitioning of soil moisture between grasses and forbs in tallgrass prairie. Specifically, they found that under conditions of limiting moisture, forbs were more dependent than grasses on switching their water-uptake strategies from shallow layers to deeper within the soil profile. von Felten et al. (2009) studied N uptake among diversity treatments in experimental grasslands and found that niche breadth decreased for subordinate species with increasing species richness, but did not change for the dominant species. In agreement with these results and the predictions of Weaver (1958), we found no change in the occurrence of the dominant grass species, but significant losses in forb richness with decreasing soil depth. Nippert & Knapp (2007a,b), however, also found, that during periods of high soil moisture, both grasses and forbs obtained the majority of their water from surface soil layers, suggesting that the segregation of soil resources among plant species may become less important as soil resources increase in abundance. Thus, the nested loss of forbs that we observed with decreasing soil depth appears to match the community-level consequences expected during periods of resource limitation, if soil-depth limitations prevent forbs from acquiring resources from deeper within the soil profile. However, we acknowledge that the mechanistic basis for the loss of forbs that we observed among depth treatments requires direct testing in future studies. For example, future work is needed to determine if the systematic loss of forbs observed with decreasing soil depth resulted from an increasing inability among forbs to tolerate the increasingly stressful conditions of shallower soils, or if, as we suspect, forb losses were driven by a decreasing ability of many species to compete with dominant grasses under increasingly restricted rooting depths and associated resource availability.

While richness decreased with decreasing soil depth, evenness increased, resulting in non-significant changes in plant diversity. Evenness increased with decreasing soil depth as a result of the loss of low relative abundance forbs, with no concomitant loss of, and little change to the biomass of, the dominant grasses. For example, the grasses Andropogon gerardii, Elymus canadensis, Panicum virgatum and Sorghastrum nutans were present at nearly identical frequencies in all treatments, and accounted for 51% and 64% of all above-ground biomass in deep and medium, and shallow treatments, respectively. In contrast, the eight forb species present in the deep, but absent from the shallow treatment, accounted for 31% of the species, but for only 5% of the above-ground biomass in the deep treatment. Evenness responds most quickly to changes in environmental factors (reviewed by Chapin et al. 2000; Hillebrand, Bennett & Cadotte 2008), suggesting that future work is needed to determine if the diversity patterns that we report among depth treatments will remain, or if they currently reflect the effect of the extinction of rare forb species on species evenness, and thereby on community diversity.

Hartnett & Fay (1998), summarizing the work of Glenn & Collins (1990) and Collins (1992), suggest that variation in grassland species richness and composition results primarily from differences in satellite forb species, not in the core grasses. Our data agree with this suggestion, supporting our prediction of changes in forb occurrences among depth treatments. However, in contrast to our initial expectation, we found no difference in community composition among depth treatments. Thus, soil depth appeared to select for a subset of forbs, but within each soil depth several forb species were apparently interchangeable and the occurrence of any single species from the subset was apparently stochastic.


We provide support from an experimental manipulation replicated across years that alterations to soil depth altered species richness, evenness and tallgrass prairie community assembly in restored grassland. Furthermore, the identity and known rooting strategies of the forbs lost with decreasing soil depth were non-random and agreed with previously published results. Although our experiment was not designed to isolate the mechanistic basis driving soil-depth effects on community structure, our results in combination with the results of previous authors lead us to hypothesize that the importance of available rooting depth in promoting species co-occurrence in tallgrass prairie may depend upon the relative importance of above-ground versus below-ground competition, and on the relative intensity of below-ground competition (high versus low soil resource availability). Future work needs to evaluate the mechanistic basis for our results and the validity of our hypotheses for different grasslands. Nevertheless, this study provides the first field-based, experimental evidence that altering soil depth alters species occurrence and diversity in predictable, non-random ways in seeded tallgrass prairie.


This work was funded in part by a grant from the Iowa DOT Living Roadway Trust Fund to B. Wilsey. A. Loan-Wilsey, K. Wahl, S. Dellemann, A. Knaack provided invaluable assistance with data collection. Additional support was provided by the Department of Natural & Applied Sciences at the University of Wisconsin-Green Bay and the Department of Ecology, Evolution and Organismal Biology at Iowa State University. Several anonymous referees provided valuable suggestions improving this manuscript.