| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
What's this? |
Article |
Département de biologie, Université Laval, Québec, Québec, Canada G1K 7P4
Received for publication April 14, 2006. Accepted for publication November 8, 2006.
ABSTRACT
The specific shape of the relationship between plant diversity and productivity and the causal mechanism(s) behind the observed pattern(s) are still highly debated. Recent advances suggest that the relationship depends on several environmental variables and may change with the observational scale. In this study, a multivariate, multiscale approach was used to identify the variables that determine the relationship between species richness and annual production along a forest/old field edge in southern Québec (Canada). Various relationships between richness and production were found at different distances to the edge. In the forest, most relationships were positive and linear, while in the old field the relationship shifted from positive linear to non-significant with increasing distance from the edge. In the forest or in the old field, the shape of the relationship (all distances from the edge combined) was unimodal. Path analyses showed that species richness was determined mostly by production, which was influenced by different limiting resources, depending on the community (forest or old field). An increasing range in production created by pooling across community types can confound the resources and/or conditions determining the diversity–productivity relationship.
Key Words: diversity–productivity relationship • edge effect • light gradient • multivariate approach • path analysis • production • spatial scale • species richness • unimodal curve
Understanding the relation between species richness and productivity has been of interest to ecologists for almost half a century (Connell and Orias, 1964
; Leigh, 1965
). Yet, the specific shape of the relation and the causal mechanism(s) behind the observed pattern(s) are still highly debated (Abrams, 1995
; Oksanen, 1996
; Grace, 1999
; Waide et al., 1999
; Mittelbach et al., 2001
; Rajaniemi, 2003
). Grime (1973)
and Al-Mufti et al. (1977)
were among the first to report maximal richness at intermediate levels of productivity. Grime (1973
, 1977
) discussed this pattern in terms of the ability of species to perform at different productivity (i.e., resource availability) levels: at low productivity, few species can tolerate the stress characteristic of low-resource habitats. As productivity increases, more resources become available, thereby relaxing the abiotic stress and allowing the establishment and growth of more species. At still higher levels of productivity, a few dominant species tend to competitively exclude most other species, thereby decreasing overall species richness. Rosenzweig (1992
, p. 728) considered this unimodal relation "the true productivity pattern," whereas Huston and deAngelis (1994, p. 972) referred to it as the "ubiquitous unimodal diversity response along productivity gradients." However, in a recent literature review, Waide et al. (1999)
showed that a majority of the studies conducted in natural environments do not support a unimodal relation between diversity and productivity. The multitude of relations found (positive linear, negative linear, u-shaped, unimodal, or even no relation) seems to come from the difficulty in applying the theoretical models to the patterns observed in nature because such patterns are scale dependent.
Indeed, Moore and Keddy (1989)
in a study of a wide variety of wetlands did not find any significant diversity–productivity relation (DPR) within vegetation types. Yet, they observed a clear unimodal pattern when observations were pooled across vegetation types. These results suggested to the authors that Grime's model described variations in species richness among but not within vegetation types; they concluded that "... the higher level processes which structure species richness patterns among vegetation types are not the same processes which determine patterns within a vegetation type" (Moore and Keddy, 1989
, p. 99). However, Grace (1999)
reviewed empirical evidence on DPRs in herbaceous plant communities and suggested that the same process could operate at different scales and still generate different patterns. The often-reported absence of a significant relation within community types might be due solely to the small range of both richness and productivity. Pooling across community types has the effect of increasing the range of both the dependent and the independent variables, thus uncovering the whole unimodal DPR.
Indeed, Guo and Berry (1998)
used an approach similar to that of Moore and Keddy (1989)
and found a variety of significant relations (positive or negative linear, unimodal) or no significant relation within community types. By pooling data across community types, the different patterns previously found merged to create the unimodal relation as the productivity gradient was extended. Guo and Berry (1998, p. 2558) suggested that the patterns observed within community types represented in fact ‘sections' of the general unimodal relation and concluded that "the greater the range in environmental conditions, the more complete ... [is] ... the development of the ‘hump-shape' relationship." Scheiner et al. (2000)
coined the phrase pattern accumulation hypothesis to invoke this between-community pattern as a simple accumulation of local effects and patterns.
However, pooling data from contrasting community types has confounding effects. First, when combining data from spatially distant communities, one is ignoring the fact that the species involved might belong to distinct pools. Second, and more importantly, key environmental factors (e.g., climate, disturbance dynamics, habitat characteristics) are likely to differ significantly among distinct community types, and environmental factors have been shown to affect diversity and productivity (Gough et al., 1994
; Grace and Pugesek, 1997
; Weiher, 2003
). Therefore, processes responsible for the relation that is observed when pooling data across community types may remain obscured by unaccounted differences in species pool and environmental factors.
The objectives of the present study were to (1) examine changes in the DPR along a direct gradient corresponding to a forest/old field edge, (2) explain the processes underlying these relations in terms of key environmental factors using a multivariate approach, and (3) demonstrate how these individual relations can be integrated at an increasing range in productivity.
A forest edge corresponds to a complex environment in which species richness and community composition are influenced by a suite of interacting factors such as light, soil and air moisture, and temperature (Ranney et al., 1981
; Matlack, 1993
; Fraver, 1994
; Matlack, 1994
; Murcia, 1995
; Fox et al., 1997
; Gehlhausen et al., 2000
). Thus, we may expect different microhabitats at different distances from the edge, in the forest as well as in the old field. In the present study, we describe a DPR for each distance from the edge (i.e., each microhabitat) for the forest as well as for the old field. We then combine the microhabitat DPRs into two habitat DPRs (i.e., one for the forest and one for the old field). We also determine a DPR for the entire forest/old field edge by combining the two habitat DPRs. Because we measure specific environmental variables for each microhabitat, we can describe the causal mechanics of the DPR using a specific case of structural equation modelling, namely path analysis.
We expect ground layer production to increase from the forest to the old field, following the light availability gradient. However, we predict a peak in plant species richness (a measure of diversity) at the edge under intermediate light conditions (Gehlhausen et al., 2000
) and a decrease toward both the forest (characterized by a few shade-tolerant specialists) and the old field (characterized by a few, highly productive species). Thus, positive linear DPRs are expected in the forest, where both plant richness and production should increase with light availability; the slopes of the linear relations should gradually decrease from the forest interior to the edge. Negative linear DPRs should be found in the old field, where competition for light should cause a decrease in richness with an increase in production; the slopes of the relations should become more negative from the edge to the old field. When combined over the entire forest/old field edge, these relations should generate the classic unimodal DPR (Fig. 1).
|
Study site
The study site is located in southern Québec (Bécancour, Québec, Canada), in the Léon-Provancher Ecological Reserve (46°17' N, 72°30' W). The area falls within the Great Lakes–St. Lawrence forest region of Rowe (1972)
, subsection Mid St-Lawrence (L-3) characterized by thick marine deposits from the Champlain Sea. Mean annual rainfall and daily temperature at the nearby Nicolet weather station are, respectively, 931.1 mm and 5.1°C. The reserve comprises a forest tract of ca. 200 ha, which includes a sugar maple stand that was exploited for sap until the early 1970s. An old field (ca. 2 ha), which was used as a pasture and has been abandoned for at least 40 years, borders the southeastern edge of the sugar maple stand and creates an uninterrupted, 200 m long edge. The study site is relatively flat and has not been subjected to recent human disturbances. In the absence of such disturbances, trees from the forest are encroaching on the old field, forming a progressively shorter canopy across the original edge (embedded edge sensu Matlack, 1994
).
Tree species in the forest include Acer saccharum Marsh., Tsuga canadensis (L.) Carr., Fagus grandifolia Ehrh., Tilia americana L., and Carpinus caroliniana Walt., all relatively shade tolerant. The first woody species to colonize the old field/forest edge are the less shade-tolerant Betula populifolia Marsh., Populus tremuloides Michx., and Salix spp.
Ground layer vegetation in the forest community is diverse (ca. 50 species) and relatively dense in the spring mostly due to ephemeral species (Erythronium americanum Ker-Gawl., Claytonia caroliniana Michx.), but becomes more open in the summer with ferns [Onoclea sensibilis L., Athyrium filix-femina (L.) Roth], tree seedlings and juveniles (e.g., Acer saccharum, Fraxinus spp., Fagus grandifolia), Carex spp., and several other perennials [e.g., Symplocarpus foetidus (L.) Salisb., Equisetum sylvaticum L., and Toxicodendron radicans (L.) Kuntze]. Herbaceous vegetation in the old field is somewhat more diverse (ca. 60 species), quite dense during the entire growing season, and characterized by several perennial species such as Solidago canadensis L., S. rugosa Mill., Rubus idaeus L., R. alleghaniensis Porter, Onoclea sensibilis, Toxicodendron radicans, Poa pratensis L., and Phalaris arundinacea L. Nomenclature follows that of Marie-Victorin (1995)
.
Fifteen 90 m long transects, 4.25 m apart were established perpendicular to the edge such that each transect had 45 m in the forest and 45 m in the old field. Most studies on edge effects in temperate forests have used transects of similar length: for instance, Matlack (1993)
used 50 m transects, Matlack (1994)
used 40 m transects, and Gehlhausen et al. (2000)
used 175 m transects but found that edge effects disappeared after 40 m on south-oriented edges. One 1.25 x 1.25 m quadrat was positioned along each transect at 2.5, 5, 10, 20, and 45 m in the forest and in the old field (for a total of 150 quadrats). Quadrats were subdivided into four, 0.25 m2 subquadrats with a 0.25 m buffer strip in between. Because our study focused on ground layer species, whenever a tree (height > 2.5 m) was present inside the quadrat, we moved the quadrat laterally from the transect. Edge position was determined from the presence of rock hummocks and wire fencing embedded in trees placed when the old field was used for pasture.
Three harvests were accomplished during the 2003 growing season (spring: 26 May–2 June, summer: 3–8 July, and fall: 26–28 August). At each harvest in each of the 150 quadrats, one subquadrat was randomly selected without replacement, and all herbaceous plants were cut at the ground level and sorted by species. Annual net primary production in a quadrat (g/0.25 m2) was calculated from the sum of the maximum aboveground biomass of each species over the growing season. Annual aboveground biomass of shrubs and tree seedlings or juveniles (
1 m high) was evaluated using foliage and twigs produced during the 2003 growing season (therefore omitting cambial woody biomass increment). In the laboratory, all samples (aboveground biomass of each species in each subquadrat) were dried at 75°C for 48 h and weighed.
Soil characteristics were measured from samples gathered on 3 August 2003 after a period of at least three days without rain. Soil samples (4 cm diameter, 10 cm deep) were collected from the center of each quadrat in the buffer area (H horizon and top of A horizon). In the laboratory, samples were passed through a 2-mm mesh sieve to sort out rocks, twigs, and roots. Soil moisture was determined gravimetrically. Soil pH was measured in a volume ratio 1 : 1 soil : water solution. Soil organic matter was determined by mass loss on ignition (450°C for 5 h).
Irradiance was measured on 29 July 2003, under uniform conditions between 1000 and 1400 h, using a portable radiometer (LI-250, LI-COR, Lincoln, Nebraska, USA). An average reading was taken from a 10-s scan at 1 m from the ground over each quadrat. Irradiance measured above ground level is an estimate of incident light available to even the tallest individuals included in the production and richness measures (e.g., tree juveniles and shrubs 1 m high and individuals of Solidago species), while accounting for the shade created by the canopy layer. Canopy openness was also determined from photographs taken at 1.5 m over each quadrat with a 50-mm lens. Percentage canopy openness was evaluated with a grid comprising 100 points superposed onto the photographs (data not shown).
Data analysis
Prior to statistical analysis, some data were transformed as necessary to achieve linearity with richness and normality of the residuals (Table 1).
|
). For each variable generating a significant Wilks 
, profile contrasts were made. This procedure is suited for finding significant differences (breaking points) between adjacent distances along a continuum.
Diversity–productivity relationships
Raw (untransformed) data from each of the 10 distances to the edge were analyzed to test for a relationship between richness and production. Additionally, data from each side of the edge (75 forest quadrats and 75 old field quadrats) and from the entire site (for a total of 150 quadrats) were analyzed. In all cases, we tested both linear and quadratic regressions. To interpret a quadratic regression as a unimodal relationship (that is, a relation with a significant peak and not just a curvature), three conditions had to be met. First, the R2 had to be significant. Second, the quadratic term had to be negative and significant. Third, the relationship had to conform to the Mitchell-Olds and Shaw's test (MOS test; Mitchell-Olds and Shaw, 1987
). As applied here, the MOS test verifies that a peak in richness is inside the measured range of production. All tests were performed using Statistica release 5 (StatSoft, Tulsa, Oklahoma, USA).
Path models
Structural equation modeling (SEM) methods, of which path analysis is a specific case, can be used to determine whether data sets possess commonality with regards to a set of structural (dependence) relationships (Grace and Pugesek, 1998
). The first step is to propose a structural model based on a priori theoretical knowledge. In this study, path analyses were carried out to explain richness in terms of annual production and environmental factors (Fig. 2). In our structural model, all environmental factors (irradiance, soil pH, moisture, and organic matter) are directly associated with distance to the edge (Matlack, 1993
). In turn, the two variables of interest, annual production and species richness, are influenced by all environmental factors (Gough et al., 1994
; Grace and Pugesek, 1997
; Weiher, 2003
). In addition, soil organic matter controls soil pH and allows the retention of soil moisture. Soil moisture also directly affects soil pH. The next step in our SEM was to test the conformity of our data to the hypothesized structural model; this was done using EQS 6.1 for Windows (Bentler, 2004
). The hypothesized path model is translated into structural equations containing free parameters (path coefficients, variances, or covariances) that are further estimated from each data set. From these equations, predicted covariance matrices are formulated. The free parameters in these predicted covariance matrices are estimated to make the predicted matrices as close as possible to the observed covariance matrices using maximum likelihood (ML). To test model fit, the null hypothesis is that there is no difference between the observed and the predicted covariance matrices, except for what would be expected given random sampling variation. With this hypothesis, a maximum likelihood chi-squared (ML
2) statistic asymptotically follows a central chi-squared distribution under the assumption of multivariate normality (Shipley, 2000
). A corrected ML statistic, the Satorra-Bentler 2 was used to account for deviations from multivariate normality (Iriondo et al., 2003
). In SEM, when the probability associated with the statistic is small (<0.05), the hypothetical model has to be rejected; on the other hand, a probability >0.05 means that empirical data are consistent with the causal processes suggested by the structural model (Shipley, 2000
). This does not prove causation but instead it shows that the initial assumptions made by specifying the causal structure are not contradicted and might be valid (Bollen, 1989
). Statistically, a model has an exact fit to the data when 2 has a P > 0.05, a close fit when the root mean square error of approximation (RMSEA) has a P > 0.05 and an acceptable fit when its goodness of fit index (GFI) is above 0.9 (Iriondo et al., 2003
).
|
). In this study, standardized coefficients are presented, thus allowing comparisons between all path coefficients. Indirect effects are the effects of a given variable on another one transmitted through mediator variables, and these effects can be estimated by multiplying the coefficients along the entire path. The coefficient of determination of a variable is equivalent to the linear or multiple regression R2 (depending on the number of arrows leading to the focus variable) with its associated significance level.
There are several rules to determine the minimum sample size necessary to test a model (Bentler and Chou, 1987
; Petraitis et al., 1996
; Stevens, 1996
). In the present study, we use 75 cases, which is marginally above the minimum (70 cases) following the rules described by Petraitis et al. (1996)
. Shipley (2000)
suggests that if the sample size is too small to confidently assume that the sampling distribution of the ML2 statistic is close to the theoretical 2 distribution, one can use the d-separation (d-sep) method. We tried the d-sep method with our data and obtained the exact same results. We are thus confident that our path models are robust with a sample size of 75.
RESULTS
Gradient analysis
Significant changes were found from the forest interior to the old field for each one of the variables (MANOVA; P values ranging from 0.0001 to 0.0177), with profile contrasts indicating at least one major breaking point along the continuum (Fig. 3). As expected, irradiance (Fig. 3A) was an important variable associated with the forest/old field continuum: although there were four significant breaking points along the distance gradient, and a nine-fold increase in irradiance was found between 5–10 m to the edge in the old field. This abrupt rise in irradiance coincided with a sharp increase in annual production (Fig. 3E). Another simultaneous increase in these two variables was found between 20–45 m to the edge in the old field. Soil moisture was relatively constant along the continuum, apart from a sharp decrease between 45–20 m from the edge in the forest (Fig. 3B). Figure 3C shows that soil pH increased from ca. 4.2 at 20 m in the forest interior to ca. 5.1 at 45 m to the edge in the old field; however, values at 45 m in the forest were significantly higher than any of the other forest values. From the forest interior to the old field, soil organic matter displayed a significant two-fold decrease (Fig. 3D). The relatively high values in soil moisture (Fig. 3B), pH (Fig. 3C), and organic matter (Fig. 3D) at 45 m to the edge in the forest were associated with a shallow depression (de Lafontaine, 2004
).
|
Diversity–productivity relationships
Table 2 and Fig. 4 show that significant linear regressions were found at 10 m and 2.5 m to the edge in the forest (r2 = 0.558, P = 0.0014; r2 = 0.534, P = 0.0020, respectively). Quadratic regressions were found at 10 m, 5 m, and 2.5 m to the edge in the forest, however, the relationship was unimodal only at 2.5 m (R2 = 0.846, P < 0.0001). No significant relationships were found at 45 m and 20 m to the edge in the forest. In the old field habitat, only one linear regression was significant, that at 5 m (r2 = 0.268, P = 0.0479); no quadratic regressions were significant there. When combining data from the different distances, significant quadratic regressions gave a better fit over linear regressions for the forest (R2 = 0.376, P < 0.0001), the old field (R2 = 0.190, P = 0.0005), and the entire site (R2 = 0.325, P < 0.0001), all with the quadratic term significant and the MOS test indicating a unimodal relationship.
|
|
2 = 4.2872, df = 5, P = 0.5089; old field: GFI = 0.967, RMSEA = 0.103, Satorra-Bentler 2 = 6.7284, df = 5, P = 0.2416). However, care must be taken when interpreting the diagrams shown in Fig. 5. First, we arbitrarily assigned negative values to distances in the forest, the edge being 0 m and distances in the old field being positive. Second, the variables organic matter and soil moisture were submitted to an inverse transformation, thus changing the sign of the path coefficients associated with them. Finally, the linear relationships between annual production and species richness are the results of transformations that follow unimodal quadratic regressions with positive and negative components.
|
DISCUSSION
Forest/old field continuum
Our irradiance data indicated that the original edge (0 m) was not the functional edge: because the edge was embedded, the functional edge was displaced somewhere between 5- and 10 m in the old field. This had an impact on the production gradient, which was also different from that expected (Fig. 1). Indeed, production followed this "new" edge definition, with a slow increase from 0 to 5 m in the old field and then a sharp increase at 10 m (Fig. 3E). High values of production were found at 45 m to the edge in the old field. In contrast, production from the edge to the forest interior was rather constant at a low level.
We expected a peak in richness at the original edge and a decrease toward both the forest and the old field, albeit for different reasons. Instead, mean richness was high at 45 m in the forest (Fig. 3F): this was associated with a shallow depression at this distance (Fig. 3B) and the presence of several hydrophilic species, such as Carex gynandra Schwein., C. projecta Mack., C. tuckermanii Dewey, Galium palustre L., Lycopus americanus Mühl., and Rubus pubescens Raf. These rather small specialist species contributed to increase richness more so than production (see next section). Richness was low at the original edge, but maximal at the functional edge (10 m into the old field); however, it did not decrease further in the old field. This is not consistent with our original model (Fig.1).
Diversity–productivity relationships
In the forest, two DPRs were linear (and positive) with the slope of the regressions higher than in the old field. A negative (although not significant) slope was found at 20 m to the original edge in the forest, but this was due to an outlier (Fig. 4). Indeed, one subquadrat was entirely filled by a single individual of skunk cabbage [Symplocarpus foetidus (L.) Salisb.], thus increasing production to a value that was not found elsewhere in the forest. Removing this outlier gave the steepest regression slope of the entire site (ß1 = 0.466, P = 0.0623, and r2 = 0.260). Therefore, two significant linear regressions (10 m and 2.5 m) and two marginally significant relations (20 m, outlier removed: P = 0.0623; 5 m: P = 0.0653), suggested positive linear relationships for the forest. Furthermore, when removing the outlier data point at 20 m, the MOS test for the entire forest habitat no longer found a unimodal relationship; however, the quadratic regression remained significant (ß1 = 0.606, ß2 = –0.012, P < 0.0001, R2 = 0.406). The quadratic relationship suggested a rapid increase in richness with production at low levels of production and a decrease in the rate of species accumulation at higher values of production. Therefore, we may conclude that in the forest habitat, DPRs were mostly consistent with our model.
In the old field, a significant positive and linear DPR was found at 5 m to the edge and a marginally significant one (P = 0.0567) at 2.5 m to the edge. These two regressions had shallower slopes than those found in the forest. DPRs at 10, 20, and 45 m were not significant. Considering these results in the light of the "new" edge definition, there was a shift from positive DPRs between the original and the functional edge, where production was low, to non significant DPRs farther in the old field, where production was high (Fig. 4). However, combining the data of the entire old field habitat, a significant unimodal DPR was found. Therefore, the new edge definition changed the proposed DPR patterns (Fig. 1) and displaced the shift from positive to null slopes in the old field between 5–10 m, which corresponded to the functional edge. The negative DPRs we expected at the different distances from the edge in the old field were not found. This could be a statistical artefact of the low number of data points or a demonstration that clear negative DPRs are found only at high values of production. Indeed, the DPR for the entire old field habitat showed a clear decrease in richness only at high values of production (>400 g/0.75 m2). However, only a few high production points were present at 10 m (1 point) and 20 m (3 points) to the edge in the old field (none at 2.5 m or 5 m). Yet, when combined over the entire old field habitat, these data points combined with those recorded at 45 m had enough statistical weight to significantly develop the decreasing section of the unimodal relationship (Table 2, MOS test significant). This negative relationship between richness and production at high values of production has often been discussed both in theory and in fertilization experiments. The negative relationship between richness and production has been attributed to the increased shading and competition for light by tall plants associated with increased biomass production (Tilman and Pacala, 1993
; Huston and deAngelis, 1994
; Abrams, 1995
; Collins et al., 1998
; Baer et al., 2003
) and accumulation of litter (Carson and Peterson, 1990
; Foster and Gross, 1998
; Gough et al., 2000
).
As expected (Fig. 1), combining the several within-distance mostly linear relationships over the entire site gave the classic unimodal DPR (Fig. 4; this still remained true even when the outlier [20 m, in the forest] was removed: R2 = 0.330, P < 0.0001, MOS test significant). Our results are thus consistent with those of Moore and Keddy (1989)
and Guo and Berry (1998)
.
Multivariate causes of diversity (path models)
Grace and Pugesek (1997)
used a multivariate approach to identify what environmental variables determined most species richness in a series of coastal wetlands. Their findings suggested that richness was most strongly affected by light at the soil surface and by abiotic stress (related to soil quality) and biomass. The negative effect of biomass on richness had a direct component and an indirect one associated with reduced light. Weiher (2003)
tested Grace and Pugesek's model along several oak savanna/prairie edges: he found that fire disturbance and biomass, not light availability, were the major determinants of richness. Our multivariate model included all those factors associated with richness in the two previously-mentioned studies, except for disturbance (there was no apparent recent disturbance at our study site). Yet, our path models gave results different from those of Grace and Pugesek (1997)
and of Weiher (2003)
.
Overall, our models (Fig. 5) suggested that richness was entirely determined by production via mostly linear positive DPRs in the forest. In the old field, richness was also mostly determined by production, but via DPRs of various shapes, depending on the production range considered. In this habitat, soil moisture also had a direct effect on richness; however, this was independent of the spatial gradient studied because soil moisture was not determined by distance to the edge (Fig. 5B). Therefore, along our forest-old field gradient, production stood out as the major determinant of species richness. In turn, production in the old field was mostly determined by light availability, while in the forest, under constant low light availability, it was mostly determined by edaphic features (soil pH and organic matter and, indirectly, soil moisture). This stresses the importance of being careful when pooling DPRs across community types: indeed, even if production is a major determinant of species richness, the determinants of production (the resources and/or conditions limiting most plant growth) might vary with community types, making difficult the identification of the processes behind the pooled DPRs.
Conclusion
We have shown here that the shape of the DPR changed with the production range considered, as suggested by Grace (1999)
: indeed, distinct DPRs were found at different production ranges along our forest/old field gradient. When the production range was wide enough, a clear unimodal relationship was obtained. We used a multivariate approach to test an a priori causal model of DPRs along our forest/old field gradient. We found that species richness was determined mostly by production, which was driven by varying limiting resources, depending on the community type studied.
The present study highlights the fact that pooling data across community types may obscure the processes behind the pooled DPRs; thus these data must be interpreted with caution. Although production alone is sufficient to explain a significant part of the variance in species richness, one must keep in mind that by increasing the production range to reveal a global unimodal relationship, one may confound the environmental variables that determine both production and richness.
FOOTNOTES
1 The authors thank V. Bolduc-Tremblay, A. Lenière, P. Marchand, and F. Sahim for assistance in the field and the laboratory and J. Grace, L. Lapointe, S. Payette, and two anonymous reviewers for their comments on an earlier version of this manuscript. Financial support was provided by the Natural Sciences and Engineering Research Council of Canada to G. Houle. 
2 Author for correspondence (e-mail: gilles.houle{at}bio.ulaval.ca
)
LITERATURE CITED
Abrams P. A.. 1995. Monotonic or unimodal diversity—-productivity gradients: what does competition theory predict?. Ecology 76: 2019-2027.[CrossRef]
Al-Mufti M. M. Sydes C. L. Furness S. B. Grime J. P. Band S. R.. 1977. A quantitative analysis of shoot phenology and dominance in herbaceous vegetation. Journal of Ecology 65: 759-791.[CrossRef]
Baer S. G. Blair J. M. Collins S. L. Knapp A. K.. 2003. Soil resources regulate productivity and diversity in newly established tallgrass prairie. Ecology 84: 724-735.[CrossRef][Web of Science]
Bentler P. M.. 2004. EQS structural equations program manual Multivariate Software, Inc., Encino, California, USA..
Bentler P. M. Chou C. P.. 1987. Practical issues in structural modeling. Sociological Methods and Research 16: 78-117.[CrossRef]
Bollen K. A.. 1989. Structural equations with latent variables Wiley, New York, New York, USA..
Carson W. P. Peterson C. J.. 1990. The role of litter in an old-field community: impact of litter quantity in different seasons on plant species richness and abundance. Oecologia 85: 8-13.[CrossRef][Web of Science]
Collins S. L. Knapp A. K. Briggs J. M. Blair J. M. Steinauer E. M.. 1998. Modulation of diversity by grazing and mowing in native tallgrass prairie. Science 280: 745-747.
Connell J. H. Orias E.. 1964. The ecological regulation of species diversity. American Naturalist 98: 399-414.[CrossRef][Web of Science]
Crowder M. J. Hand D. J.. 1990. Analysis of repeated measures Chapman and Hall, New York, New York, USA..
de Lafontaine G.. 2004. Richesse spécifique le long d'un gradient de production: utilisation d'une approche multivariée M.Sc. thesis. Université Laval, Québec, QC, Canada..
Foster B. L. Gross K. L.. 1998. Species richness in a successional grassland: effects of nitrogen enrichment and plant litter. Ecology 79: 2593-2602.[CrossRef][Web of Science]
Fox B. J. Taylor J. E. Fox M. D. Williams C.. 1997. Vegetation changes across edges of rainforest remnants. Biological Conservation 82: 1-13.[CrossRef][Web of Science]
Fraver S.. 1994. Vegetation responses along edge-to-interior gradients in the mixed hardwood forests of the Roanoke River Basin, North Carolina. Conservation Biology 3: 822-832.
Gehlhausen S. M. Schwatrz M. W. Augspurger C. K.. 2000. Vegetation and microclimatic edge effects in two mixed-mesophytic forest fragments. Plant Ecology 147: 21-35.[CrossRef][Web of Science]
Gough L. Grace J. B. Taylor K. L.. 1994. The relationship between species richness and community biomass: the importance of environmental variables. Oikos 70: 271-279.[CrossRef][Web of Science]
Gough L. Osenberg C. W. Gross K. L. Collins S. L.. 2000. Fertilization effects on species density and primary productivity in herbaceous plant communities. Oikos 89: 428-439.[CrossRef][Web of Science]
Grace J. B.. 1999. The factors controlling species density in herbaceous plant communities: an assessment. Perspectives in Plant Ecology, Evolution, and Systematics 2: 1-28.[CrossRef]
Grace J. B. Pugesek B. H.. 1997. A structural equation model of plant species richness and its application to a coastal wetland. American Naturalist 149: 436-460.[CrossRef][Web of Science]
Grace J. B. Pugesek B. H.. 1998. On the use of path analysis and related procedures for the investigation of ecological problems. American Naturalist 152: 151-159.[CrossRef][Web of Science]
Grime J. P.. 1973. Competitive exclusion in herbaceous vegetation. Nature 242: 344-347.[CrossRef]
Grime J. P.. 1977. Evidence for the existence of three primary strategies in plants and its relevance to ecological and evolutionary theory. American Naturalist 111: 1169-1194.[CrossRef][Web of Science]
Guo Q. Berry W. L.. 1998. Species richness and biomass: dissection of the hump-shaped relationships. Ecology 79: 2555-2559.[Web of Science]
Huston M. A. deAngelis D. L.. 1994. Competition and coexistence: the effects of resource transport and supply rates. American Naturalist 144: 954-977.[CrossRef][Web of Science]
Iriondo J. M. Albert M. J. Escudero A.. 2003. Structural equation modelling: an alternative for assessing causal relationships in threatened plant populations. Biological Conservation 113: 367-377.[CrossRef][Web of Science]
Leigh E. G. J.. 1965. On the relation between the productivity, biomass, diversity, and stability of a community. Proceedings of the National Academy of Sciences, USA 53: 777-783.
Marie-Victorin F.. 1995. Flore Laurentienne, 3rd ed Les Presses de l'Université de Montréal, Montréal, Québec, Canada..
Matlack G. R.. 1993. Microenvironment variation within and among forest edge sites in the eastern United States. Biological Conservation 66: 185-194.[CrossRef][Web of Science]
Matlack G. R.. 1994. Vegetation dynamics of the forest edge: trends in space and successional time. Journal of Ecology 82: 113-123.[CrossRef]
Mitchell-Olds T. Shaw R. G.. 1987. Regression analysis of natural selection: statistical inference and biological interpretation. Evolution 41: 1149-1161.[CrossRef][Web of Science]
Mittelbach G. G. Steiner C. F. Scheiner S. M. Gross K. L. Reynolds H. L. Waide R. B. Willig M. R. Dodson S. I. Gough L.. 2001. What is the observed relationship between species richness and productivity?. Ecology 82: 2381-2396.[CrossRef][Web of Science]
Moore D. R. J. Keddy P. A.. 1989. The relationship between species richness and standing crop in wetlands: the importance of scale. Vegetatio 79: 99-106.[Web of Science]
Murcia C.. 1995. Edge effects in fragmented forests: implications for conservation. Trends in Ecology and Evolution 10: 58-62.[CrossRef]
Oksanen J.. 1996. Is the humped relationship between species richness and biomass an artefact due to plot size?. Journal of Ecology 84: 293-295.[CrossRef][Web of Science]
Petraitis P. S. Dunham A. E. Niewiarowski P. H.. 1996. Inferring multiple causality: the limitations of path analysis. Functional Ecology 10: 421-431.[CrossRef][Web of Science]
Rajaniemi T. K.. 2003. Explaining productivity–diversity relationships in plants. Oikos 101: 449-457.[CrossRef][Web of Science]
Ranney J. W. Bruner M. C. Levenson J. B.. 1981. The importance of edge in the structure and dynamics of forest stands. In R. L. Burgess, D. M. Sharpe [eds.] Forest island dynamics in man-dominated landscapes, 67-95 Springer Verlag, New York, New York, USA..
Rosenzweig M. L.. 1992. Species diversity gradients: we know more and less than we thought. Journal of Mammalogy 73: 715-730.[CrossRef]
Rowe J. S.. 1972. Les régions forestières du Canada Ministère de l'Environnement, Service Canadien des Forêts, Ottawa, Ontario, Canada..
Scheiner S. M. Cox S. B. Willig M. R. Mittelbach G. G. Osenberg C. W. Kaspari M.. 2000. Species richness, species-area curves, and Simpson's paradox. Evolutionary Ecology Research 2: 791-802.[Web of Science]
Shipley B.. 2000. Cause and correlation in biology: a user's guide to path analysis, structural equations and causal inference Cambridge University Press, Cambridge, UK..
Stevens J.. 1996. Applied multivariate statistics for the social sciences, 3rd ed Lawrence Erlbaum, Mahwah, New Jersey, USA..
Tilman D. Pacala S.. 1993. The maintenance of species richness in plant community. In R. E. Ricklefs, D. Schluter [eds.] Species diversity in ecological communities: historical and geographical perspectives, 13-65 University of Chicago Press, Chicago, Illinois, USA..
Waide R. B. Willig M. R. Steiner C. F. Mittelbach G. G. Gough L. Dodson S. I. Juday G. P. Parmenter R.. 1999. The relationship between productivity and species richness. Annual Review of Ecology and Systematics 30: 257-300.[CrossRef][Web of Science]
Weiher E.. 2003. Species richness along multiple gradients: testing a general multivariate model in oak savannas. Oikos 101: 311-316.[CrossRef][Web of Science]
CiteULike Complore Connotea Del.icio.us Digg Facebook Reddit Technorati Twitter What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
