The effect of agricultural diversity and crop choice on functional capacity change in grassland conversions

Authors

  • Brenda B. Lin,

    1. The Earth Institute, Columbia University, Hogan Hall B-19, 2910 Broadway, MC 3277, New York, NY 10025, USA
    2. Department of Ecology, Evolution, and Environmental Biology, Columbia University, 1200 Amsterdam Avenue, 10th Floor Schermerhorn Ext., New York, NY 10027, USA
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Dan F.B. Flynn,

    1. Department of Ecology, Evolution, and Environmental Biology, Columbia University, 1200 Amsterdam Avenue, 10th Floor Schermerhorn Ext., New York, NY 10027, USA
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Daniel E. Bunker,

    1. Department of Ecology, Evolution, and Environmental Biology, Columbia University, 1200 Amsterdam Avenue, 10th Floor Schermerhorn Ext., New York, NY 10027, USA
    Search for more papers by this author
    • Present address: Department of Biological Sciences, New Jersey Institute of Technology, 433 Colton Hall, University Heights, Newark, NJ 07102-1982, USA.

  • María Uriarte,

    1. Department of Ecology, Evolution, and Environmental Biology, Columbia University, 1200 Amsterdam Avenue, 10th Floor Schermerhorn Ext., New York, NY 10027, USA
    Search for more papers by this author
  • Shahid Naeem

    1. Department of Ecology, Evolution, and Environmental Biology, Columbia University, 1200 Amsterdam Avenue, 10th Floor Schermerhorn Ext., New York, NY 10027, USA
    Search for more papers by this author

Correspondence author. CSIRO, Marine and Atmospheric Research, 107–121 Station Street, Aspendale, VIC 3195, Australia. E-mail: bbclin@gmail.com, Brenda.Lin@csiro.au

Summary

1. Given that approximately one-quarter of grasslands worldwide have been converted to agriculture, understanding the consequences of agricultural conversion for ecosystem functioning can provide insight into potential changes in the world’s most intensively managed biomes. The Great Plains of the United States represents a major grassland region that has experienced substantial conversion of prairie grassland ecosystems to agriculture, leading to widespread changes in plant species composition and functional trait diversity. While the converted system dramatically improves food, fuel and fibre production, changes in plant trait diversity may alter the capacity of these ecosystems to provide sustaining and regulating services to society.

2. Using three key plant functional traits, we illustrate how the trait composition and trait diversity of the Great Plains plant communities has been dramatically altered when visualized in multidimensional trait space, showing strong displacement and a 10-fold reduction in trait space volumes, as systems have shifted from grassland to agricultural regimes. However, individual case studies demonstrate large variation in the direction and magnitude of trait diversity change during conversion, with some case studies exhibiting much larger reductions in trait diversity from the average on conversion, largely a consequence of the changes in species richness accompanying agricultural conversion.

3.Synthesis and applications. The conversion of grassland to agriculture does not necessarily lead to a loss of functional trait diversity, as crop choice and diversity can ameliorate the functional trait differences between native-dominated and agricultural communities. In this study, regardless of the sign and magnitude of change in trait space, a shift in trait range was observed, reflecting the dramatically different selection that agricultural plant functional traits experience compared with their native counterparts. Historical agricultural policy and modern land-use patterns have led to a landscape with decreasing plant diversity and functional capacity. However, initiatives to increase agricultural diversity on-farm and at the landscape level through the implementation of cover crops of the perennialization of cropping systems may allow ecosystems to recover and maintain greater functional capacity.

Introduction

The transformation of terrestrial ecosystems to agro-ecosystems involves reductions in biodiversity to improve provisioning ecosystem services of food and fuel, generally at the cost of other sustaining and regulating services (Altieri 1999). In the US, conversion of native ecosystems to agriculture has enormously increased the production of goods in the Great Plains, where corn and cotton production have increased 400% in the last 50 years, due to increases in cultivated land and production technologies (Parton, Gutmann & Ojima 2007). However, such conversions may result in long-term, large-scale losses in sustaining ecosystem services such as soil carbon storage, soil stabilization and soil production (Manies et al. 2001; Parton et al. 2005; Montgomery 2007). Given that approximately one-quarter of temperate and tropical grasslands worldwide have been converted to agriculture (Mock 2000), understanding the consequences of agricultural conversion in the Great Plains for ecosystem functioning can provide insight into mitigating potential negative changes in the world’s most heavily and increasingly intensively managed biomes.

Research into the relationship between biodiversity and ecosystem functioning has identified species traits as the biological mechanism for how plant diversity influences ecosystem functioning (Lavorel & Garnier 2002). Although changes in plant biodiversity are known to influence ecosystem functioning, the way in which changes in plant diversity associated with agricultural conversion have impacted ecosystem functioning remains understudied (Hooper et al. 2005; Balvanera et al. 2006; Flynn et al. 2009). Species traits represent the species’ potential contribution to ecosystem functioning; thus, the greater the range of expression of traits among species in a community, the greater the potential capacity for the community to provide ecosystem functioning across a range of environmental conditions (Hillebrand & Matthiessen 2009; Reiss et al. 2009). Greater trait diversity increases both the magnitude and stability of ecosystem functioning (Eviner & Chapin 2003; Hooper et al. 2005), and trait diversity research can provide insights into ecosystem functioning under different scenarios of biodiversity loss (Bunker et al. 2005; McIntyre et al. 2007). Although functional trait diversity research thus far has largely been theoretical in application, the use of functional trait diversity comparisons could be very useful in understanding the functional differences between ecological communities. The approach has not been applied widely because of the limited availability of trait information; however, here we apply this approach to some well-studied grasslands and agricultural systems to better understand how functional trait diversity has changed under conversion.

Agricultural plants have long been selected by breeding to function optimally in their cultivation environment (Altieri 1999). Thus, traits of crops may be similar in mean and range to those of native plants within a similar environment as crops must also be physiologically successful in that environment, as studies of managed landscapes in Europe suggest that land-use change may select for specific traits through environmental filtering because changes in the landscape will lead to changes in the environment (Lososova et al. 2006). Other research has shown that plant traits have the ability to respond to land-use change (Garnier et al. 2007), signifying that agricultural conversion would lead to a selection of different traits in agricultural species.

In this paper, we present four case studies of functional trait diversity change with conversion of native-dominated grasslands to agricultural plant communities in four counties within the US Great Plains. One way to measure trait diversity, and thus the functional capacity of ecosystems, is the multidimensional volume circumscribed by the trait values present in an ecosystem (Cornwell, Schwilk & Ackerly 2006). In this study, we measured the trait volume of native and agricultural communities as a surrogate for functional capacity using leaf mass per area, leaf nitrogen, and maximum photosynthetic capacity of each species within each community, and refer to this quantification of functional capacity as the functional trait volume (FTV) of the community. We use these case studies as examples in how measurement of traits and trait volumes can inform on the differences in trait diversity between and among specific communities.

There are four possible outcomes (illustrated in Fig. 1) for a change in FTV following conversion to agriculture, which we label contraction, expansion, shifting centroid and displacement. Contraction reflects reduced FTV in the agricultural community due to trait range contraction. Expansion reflects increased FTV in the agricultural community because of a necessity for a broader range of traits for crop production. Shifting centroid reflects no significant contraction in FTV, but a shift in FTV centroid between natural and agricultural ecosystems due to differing selection regimes, with overlap between the two. Finally, displacement reflects a shift in centroid with no overlap in FTV reflecting relatively independent ecological and selection processes shaping the two communities. This general framework for comparative studies of FTV can be broadly applied to comparative studies of functional trait diversity, not just grasslands versus croplands. Figure 1 serves primarily as a heuristic device, as more than three traits would lead to multidimensional space and not be easily illustrated.

Figure 1.

 Hypothetical outcomes for change in functional trait volume (FTV) and hull spacing following conversion to agriculture (native hull – blue; agricultural hull – red): (a) agricultural trait range contraction, (b) agricultural trait range expansion, (c) shifting centroid: no net loss of FTV with overlap, (d) displacement: no net loss of FTV without overlap. Note that this figure can be used for comparison between any two communities.

Within the analysis, we expect to see a large difference in the FTV of the native versus agricultural ecosystem, as the functional trait diversity of a community will, in general, decrease with reduced species richness. Although this may be true of the overall pattern of agricultural conversion, the change in functional diversity in specific smaller-scale landscapes may differ based on the extent of conversion experienced and the diversity of agricultural crops maintained in the converted system. Therefore, the contraction in space and the amount of displacement in space remains unknown and could vary across the various case studies.

Materials and methods

Site selection

In order to understand how the functional capacity has changed with the conversion of native grasslands to agriculture, we compared functional diversity between native and agricultural communities. Data on the composition and abundance of both native and agricultural communities were required for the analysis. We chose counties as the unit of analysis for the case studies because the USDA National Agricultural Statistical Survey collects crop acreage coverages at the county level. This information gave us both the abundance of each crop in the county as well as the overall acreage under cultivation within the county. Within the county selections, we chose locations where there would be ample information regarding species composition and trait data, and for this reason, we focused on counties where there are current long-term ecological research (LTER) sites already collecting these types of data.

We obtained data on native species composition and abundance from three LTER sites and a prairie reserve, all of which are characteristic of the original ecosystems that agricultural systems have replaced. The LTER sites are Konza Prairie (in Riley County, Kansas, 39°05′N, 96°35′W), Shortgrass Steppe (in Weld County, Colorado, 40°80′N, 104°80′W) and Cedar Creek (in Anoka County, Minnesota, 45°40′N, 93°20′W). Within these sites, species data of unmanipulated native communities without fertilization or irrigation were chosen to be most representative of the original habitat (Cleland et al. 2008). In addition to these LTER sites, plant data from Kalsow prairie in Pocahontas County, Iowa, (42°34′N, 94°34′W) were also used (Dornbush 2004). Within each of the four counties, county-level agricultural species composition and abundance data were collected from the state level reports of the 2002 National Census of Agriculture (USDA 2004). Although proportionally very little native grassland remains in many of these counties, these four sites are representative of the major ecosystems likely to have been in place and the major kinds of agricultural ecosystems that have replaced them in the Great Plains. It is important to note, however, that these grasslands have gone through many changes and the native communities should be considered ‘native-dominated’ versus completely composed of all native species, with certain exotic species (e.g. Poa pratensis and Echinochloa crusgalli) appearing in high abundance within the native grassland sites. These counties reflect a range of environmental conditions in the Great Plains (Table 1), as well as a range in the history of agricultural extensification.

Table 1.   Environmental and soil conditions in the study sites. Mean annual temperature (MAT) and mean annual precipitation (MAP) data from 1961–1990 and 1948–1995 averages, respectively, from the National Climatic Data Center; soil types from the National Cooperative Soil Characterization Data
SiteLocationMAP (mm year−1)MAT (°C)Soil type
Pocahontas County, Iowa42°34′N, 94°34′W786·17·7Clay loam to loam
Weld County, Colorado40°80′N, 104°80′W326·88·7Fine sandy loam to sandy clay loam
Riley County, Kansas39°05′N, 96°35′W837·912·8Silt clay to silt clay loam
Anoka County, Minnesota45°40′N, 93°20′W784·16·9Fine sandy loam to find sand

Data collection

Within the comparative analysis of native and agricultural communities, only species comprising the top 80% of the total abundance of plants were used in the analyses to focus on the traits of the dominant species only. This was also a practical measure as trait data for many species is still quite limited. For agricultural plants, acreage was used to represent abundance, while for native plants, abundance was represented by frequency of presence in study plots (IA), cover class (CO, KS) or biomass (MN). Within the native communities, the 80% cut-off resulted in 13 out of 46 total species at Shortgrass Steppe (CO) with the next species below the cut-off representing only 2·6% of the relative abundance; 14 out of 60 species in Konza Prairie (KS) with the next species representing less than 1% of relative abundance; 7 out of 35 species at Cedar Creek (MN) with the next species representing 2% of the relative abundance; and 32 out of 87 species in Kalsow Prairie (IA) with the next species representing less than 1% of the relative abundance. Within the agricultural community, this cut-off resulted in 5 out of 31 total species in Weld County, Colorado, with the next species only 4% of the relative abundance; 5 out of 27 species in Riley County, Kansas, with the next species representing 3·7% of the relative abundance; 5 out of 31 crops in Anoka County, MN with the sixth species representing 3·4% of the relative abundance; and 2 out of 7 total species in Pocahontas County, Iowa, with the third species representing less than 1% of relative abundance (Table 2).

Table 2.   Native and agricultural species lists (comprising 80% cumulative abundance) for the four county case studies. Relative abundance by percentage cover for native species and by percentage area under cultivation for agricultural species. For native species, availability of all three focal traits and thus inclusion in the trait volume analysis is indicated overall in parentheses and for each species; trait data were available for all agricultural species. Nomenclatures: IA: Gleason & Cronquist 1991, CO, KS: Great Plains Flora Association 1986, MN: Ownbey and Morley 1991
CountyNative speciesRelative abundanceComplete set of traitsAgricultural speciesRelative abundance
Pocahontas County, Iowa (16 of 30)Poa pratensis0·05xZea mays0·56
Andropogon gerardii0·04xGlycine max0·43
Zizia aurea (L.) Koch.0·04   
Helianthus grosseserratus Martens0·04x  
Solidago canadensis0·04x  
Aster ericoides L.0·04x  
Rosa arkansana0·04x  
Dichanthelium leibergii0·03   
Carex sp.0·04x  
Sporobolus heterolepis0·04   
Galium obtusum0·03   
Fragaria sp.0·03   
Fragaria virginiana0·03x  
Calystegia sepium (L.) R. Br.0·07   
Phlox pilosa0·03   
Aster lanceolatus Willd.0·02x  
Ratibida pinnata (Vent.) Barnh0·02   
Aster laevis L.0·02   
Vicia americana0·01   
Panicum virgatum0·02x  
Solidago rigida L.0·01x  
Desmodium canadense0·02x  
Equisetum laevigatum0·01   
Equisetum spp.0·01   
Helianthus rigidus (Cass.) Desf.0·01x  
Achillea millefolium L.0·01x  
Elymus canadensis0·02x  
Amorpha canescens0·01x  
Eleocharis compressa0·01   
Euthamia graminifolia (L.) Nutt. Ex Cass.0·01   
Weld County, Colorado (8 of 13)Buchloe dactyloides0·16xZea mays0·30
Bouteloua gracilis0·15xTriticum aestivum0·24
Stipa comata0·07xMedicago sativa0·22
Gutierrezia sarothrae0·06xPhaseolus spp.0·05
Aristida longiseta0·06   
Artemisia frigida0·05x  
Atriplex canescens0·04x  
Carex eleocharis0·05   
Chrysopsis villosa0·03   
Agropyron smithii0·04x  
Ratibida columnifera0·03   
Echinochloa crus-galli0·03x  
Opuntia polyacantha0·03   
Riley County, Kansas (8 of 14)Panicum virgatum0·11xTriticum aestivum0·27
Symphoricarpos orbiculatus0·11 Glycine max0·22
Solidago canadensis0·10xSorghum spp.0·18
Andropogon gerardii0·10xMedicago sativa0·07
Schizachyrium scoparium0·10xZea mays0·09
Ambrosia artemisiifolia0·09x  
Bouteloua curtipendula0·07x  
Carex bicknellii0·03   
Sorghastrum nutans0·03x  
Cirsium altissimum0·02   
Poa pratensis0·01x  
Solidago missouriensis0·01   
Ambrosia psilostachya0·01   
Cassia chamaecrista0·00   
Anoka County, Minnesota (7 of 7)Schizachyrium scoparium0·17xGlycine max0·30
Solidago rigida L.0·16xMedicago sativa0·22
Poa pratensis0·14xZea mays0·19
Euphorbia corollata0·11xRaphanus sativus0·06
Lathyrus venosus0·08xDaucus carota0·03
Aster azureus0·08x  
Carex sp.0·05x  

We selected three traits broadly reflective of the life-history strategies of plants and widely recognized to be key measures of plant physiology and ecology. These were leaf mass per area (LMA, g m−2), leaf nitrogen (N%) and maximum photosynthetic capacity (Amax, μmol m−2 s−1). These traits form the ‘leaf economic spectrum’, describing a range from species with fast tissue turnover, rapid resource acquisition, and little investment in defence against herbivores to the opposite combination of traits (Garnier et al. 2001; Reich et al. 2003; Dahlgren et al. 2006). Other key traits which have been identified in the literature as key indicators of plant life-history strategies include seed size, leaf size and height at maturity (Westoby et al. 2002), but fewer data were available across our focal species for these traits.

We compiled trait data from the scientific literature, including grey literature reported by state extension agencies. Our search extended from September 2007 to April 2008 and included direct contact with extension agents in each of the four counties for agricultural plant traits as well as an organized search of scientific literature of each trait for each species. Using full-text (Google Scholar) and key word (ISI Web of Science) searches for articles with potential data, we used the following specific search terms for each trait: leaf mass per area: LMA, SLA, leaf mass per area, and specific leaf area; leaf nitrogen: leaf nitrogen, leaf N, N content, and % N; maximum photosynthetic capacity: Amax, Pmax, photosyn* and CO2assimilation. Data were extracted from the articles and compiled into a database where trait data for each trait–species combination were averaged for the FTV analysis. Each trait–species combination was averaged in order to use a general representative number for the analysis although there was variation in the measurements. Of the 58 species across all of the native sites, data were completely unavailable for 14, while only one trait was available for eight, and only two traits were available for seven species. As such, these results represent a highly conservative assessment of the actual range of trait values in these communities.

Functional trait volume calculations

Convex hulls of the species’ traits distributions were calculated for native and agricultural communities to estimate the trait volume of each set of species. Convex hulls describe the minimum volume required to contain a set of points in multivariate (3D) space and represent the multivariate range of a set of data (see Fig. 2), which are essentially the FTVs of the communities. Applying this technique from computational geometry to species traits allows the calculation of the ‘volume of trait space occupied by species in a community’ (Cornwell, Schwilk & Ackerly 2006). In this technique, the number of species (vertices) must exceed the number of traits (axes) in order for a volume to be calculated; thus, with three traits, a minimum of four species is required for a volume calculation. In addition, displacement of the agricultural hull with respect to the native hull was calculated as the distance between the centroids of the two hulls, and reported as the percentage of the maximum range in trait space for those two hulls. Both volume change and displacement values are thus relative to the convex hull of the native plant species used in each comparison. Analyses were accomplished using the geometry package in the R statistical software package, based on the Quickhull algorithm. In ecology, this method has been applied to measures of plant trait diversity in oldfields (Schamp, Chau & Aarssen 2008) and in response to disturbance (Pausas & Verdu 2008), and has been found to be a powerful method for analyzing patterns of trait-based community assembly (Mouchet et al. 2010).

Figure 2.

 Convex hulls of the overall comparison between the native species (blue) and agricultural species (red) community functional trait volumes (FTV). There are 43 species included within the native FTV and eight species included in the agricultural FTV.

Results

The results of the FTV calculations, combining data from the four counties, show that the conversion of grasslands to agricultural lands in the Great Plains ecosystems, in general, reflect both a contraction and displacement of FTV three-dimensional space, representing a change in both the values and range of traits in those communities. Based on the three traits analysed, our results show that the current functional trait space of agricultural communities in the Great Plains has been dramatically reduced and has shifted from that of the native communities it replaced. The agricultural community now displays a hull volume more than 10 times smaller than the native community (native = 5·29, agriculture = 0·42), and the agricultural hull has shifted 66·2% across the total range of trait space (Fig. 2, Table 3). In the combined analysis, 43 species were included within the overall native community, and eight species were included in the overall agricultural community, explaining the difference in trait diversity only in part. The large displacement in the centroids of the hulls (66·2% of the trait space range) and lack of overlap in FTV space between native and agricultural communities indicates that the two communities are not confined to similar trait ranges through environmental filtering, but rather suggests that they have experienced independent ecological and trait selection processes.

Table 3.   Functional trait volume (FTV) calculations for native and agricultural communities combined and county hulls. Proportion of agriculture to native community FTV and percentage displacement of the centroids of the hull volumes for each county. See text for details on the calculations
CountiesNative FTVAgricultural FTVProportion Ag. to Native FTV (%)Percent displacement (%)
All counties5·290·427·966·2
Pocahontas County, Iowa1·050·000·070·1
Weld County, Colorado2·760·010·358·2
Riley County, Kansas0·200·1889·571·4
Anoka County, Minnesota0·240·1352·966·8

Although we see an extreme reduction of functional trait space as well as a greatly reduced range in all three traits in the combined analysis, the results for the individual county studies are more variable, indicating a range in potential FTV change that will occur under agricultural conversion. A large displacement and lack of overlap between the native and agricultural hulls was seen in each of the case studies, signifying difference in trait selection between the two communities. However, the individual county-level case studies show that differences in agricultural management and crop choice may allow for some agricultural conversions to maintain fairly large FTVs, equivalent to that of the grassland counterparts.

Dramatic reductions and shifts in the agricultural FTVs were seen in two counties, Pocahontas County, Iowa, (native = 0·745, agriculture = N/A) and Weld County, Colorado, (native = 2·756, agriculture = 0·007) (Fig. 3a,b). Pocahontas County, Iowa, presented an especially stark example of volume reduction because of the large difference in species diversity between the native (30 species) and agricultural (two species) communities (Table 2). Because only two species made up the 80% abundance for the agricultural community in Pocahontas County, the hull is represented simply as a line between these two points, and no volume could be calculated. Weld County, Colorado, had 13 species within the native community versus four species within the agricultural community, with the agricultural community accounting for only 0·3% of the FTV of the native community, despite only nine fewer species difference in richness.

Figure 3.

 Convex hull comparisons for the individual counties: Pocahontas County, Iowa, (a); Weld County, Colorado, (b); Riley County, Kansas, (c); Anoka County, Minnesota, (d). Note that the agricultural ‘hull’ in Pocahontas County, Iowa is a line with zero volume because two crop species make up over 80% of total crop abundance in the county. Axes as in Fig. 2.

Riley County, Kansas, (native = 0·20, agricultural = 0·18) and Anoka County, MN, (native = 0·24, agricultural = 0·13) (Fig. 3c,d) experienced smaller changes in FTV between native and agricultural communities, although there was still a displacement similar to the overall displacement (Table 3). In Riley County, 14 species were included in the native community and five species were included in the agricultural community, yet the FTV difference between the native and agricultural community is quite small. This result contrasts with the Weld County example which had a similar number of species in each community, yet dramatically reduced FTV in the agricultural community.

Discussion

These results show that the general overall pattern of agricultural conversion from grasslands leads to both a dramatic shift as well as a reduction in FTV space (Fig. 2), but this pattern varies in magnitude across the four case studies examined. Within individual counties, crop choice and plant diversity management within agricultural lands can lead to a wide range of FTV differences between native and agricultural communities. In communities where there is a large shift in species diversity, such as Pocahontas County (Fig. 3a), there is a large loss in FTV. In other counties such as Weld Riley, and Anoka (Fig. 3c) where species ratios are similar, the change in FTV can range widely based on original community composition and the agricultural species that have replaced them. These results show that agricultural FTVs do not necessarily have to be dramatically smaller than that of native FTVs and both the number of crops planted as well as the identity of the crops can have a large impact on the change in FTV with conversion to agriculture.

Neither the overall nor the individual county examples showed an overlap in the FTVs of the native and agricultural FTVs, indicating that there is a distinct difference in trait range and selection. Examining the individual traits used within the analyses provides insight into the trait differences between native and agricultural communities (Table 4). In general, the LMA of the agricultural community was less than 50% that of the native community, % N was 50–75% higher, and Amax was 10–25% higher in the agricultural community. We interpret the shift in hull volume and location as a result of agricultural management, where resource availability is highly subsidized and resource acquisition is highly monitored, leading to a high level of trait selection within the overall agricultural community. This demonstrates the strong differences between natural environmental filters and agronomic selection for species traits, whereby natural systems allow or even require a wide variety of plant form and function whereas agricultural species have been modified to grow in optimal agronomic conditions.

Table 4.   Summary of trait data of the native and agricultural communities of the four counties (LMA, N, Amax)
TraitCountyCommunityNo. speciesMeanSDRange
LMA (g m−2)Pocahontas County, IowaNative3077·8524·91 26·9–126·4
Agricultural243·0819·58  30·1–56·9
Weld County, ColoradoNative13137·07104·84 45·3–295·1
Agricultural432·736·88    28·6–43
Riley County, KansasNative1485·2622·47 57·1–132·1
Agricultural539·1711·38  29·2–56·9
Anoka County, MinnesotaNative794·1125·12   68·7–144
Agricultural541·4317·08    27·9–63
Leaf N (%)Pocahontas County, IowaNative302·410·76    1·3–4·2
Agricultural23·251·72    2·0–4·5
Weld County, ColoradoNative131·710·77    0·7–3·4
Agricultural42·600·78    1·8–3·4
Riley County, KansasNative141·890·73    0·9–2·9
Agricultural52·931·07    1·8–4·5
Anoka County, MinnesotaNative72·251·20   1·2–4·1
Agricultural53·421·03     2–4·5
Amax (μmol CO2 m−2 s−1)Pocahontas County, IowaNative3015·885·46   7·9–27·6
Agricultural221·412·15  18·4–23·4
Weld County, ColoradoNative1318·4011·18   1·4–35·6
Agricultural421·472·15  18·7–23·4
Riley County, KansasNative1420·506·00   7·7–27·6
Agricultural523·976·79  18·7–35·5
Anoka County, MinnesotaNative716·895·04  11·4–24·5
Agricultural521·065·74  12·2–27·5

The lower LMA in the agricultural community probably reflects the fact that many crops do not require high LMA and its associated high construction costs, since there is little need for long-lived, well-defended leaves (Paoletti & Pimentel 1996), especially with the use of agrochemicals. The higher Amax of crops shows that they have been engineered to maintain high photosynthetic capacity while concentrating carbohydrates toward fruit production to maximize yield. Percentage leaf nitrogen is higher in the agricultural communities probably due to nitrogen fertilization as well as the use of nitrogen-fixing crops such as Medicago sativa.

For all research using functional traits to reflect either community responses to land use change (Garnier et al. 2004, 2007; Flynn et al. 2009; Lalibertéet al. 2009) or potential for communities to effect ecosystem properties (D. Flynn, unpublished data; Griffin et al. 2009), the traits selected and the number of traits used critically influence the outcome of the analysis. The three traits chosen here, LMA, Amax and leaf %N, reflect an important range of life-history strategies. In the case of the native plants, these traits show plant adaptations to the Great Plains environment, while for agricultural plants, these traits reflect the signature of agricultural practices. If other trait data had been available, such as root : shoot ratio, seed size or height, the results would probably have demonstrated different magnitudes of contraction and displacement, and would have captured functional capacity in a broader way. Another consideration particular to the convex hull volume method used here is that when only two species are present, as for the agricultural community of Iowa, no volume can be calculated. Modifications to the convex hull volume method to include abundance information and intraspecific variation (D. Bunker, D. Flynn, S. Naeem, unpublished data) are also being developed in order to more broadly capture the functional representation of a community. However, given these caveats and understanding that future improvements will strengthen the analysis, this study serves to demonstrate that with three key traits, a striking contraction and displacement in functional diversity can be observed with the conversion to agriculture in the Great Plains, as well as the variation in functional change at the county level.

The variation in the degree of FTV change among counties may be attributed in-part to environmental conditions and historical county-specific conversion patterns. Historically, agricultural conversion in combination with the domestication of crop species, have significantly reduced native plant and crop genetic diversity in the landscape (Samson & Knopf 1994; Hyten et al. 2006), but the patterns differ depending on native grassland composition and the levels of agricultural conversion. For example, the favourable soil and climate conditions in Pocahontas County, Iowa, (Table 1) resulted in a long history of annual crop production. By 1850, this county was already under extensive wheat, corn and potato agriculture. This was followed by a dramatic shift toward large-scale agricultural production with currently 83% of the county covered by primarily two species, soybean Glycine max and corn Zea mays. Conversely, Anoka County in Minnesota represents a region that did not experience agricultural conversion until the early 1900s (Knops & Tilman 2000). Currently, only 7% of land in Anoka County, Minnesota, is under agriculture, as the sandy soils and the environmental conditions (Table 1) limit the large-scale development of agriculture, potentially also leading to a smaller difference in FTV change between native and agricultural communities. Thus, to some extent the degree of change in functional diversity with agricultural conversion may be predictable from environmental conditions. For example, areas only suitable for perennial pasture would be expected to demonstrate less contraction or displacement in FTV compared with areas suitable for annual row crops.

It should be noted that the current analysis calculates the FTVs assuming that all the species within the native and agricultural communities are present within the systems concurrently to form the full volume of the hull. However, in agricultural systems, there are often large temporal shifts in community composition due to crop rotations where one species may be present for part of the year and then replaced by another species in a different season. Therefore, two or three species may be rotated within a system throughout the year and not represented in the landscape at the same time. This lack of concurrent species representation within the agricultural systems may lower the true FTV for agricultural systems since the species are not present temporally at the same time, representing a less diverse species community than the ones considered in this analysis.

Regional and national agricultural policy, in addition to county-specific history, plays a large role in the observed shifts in FTVs between native and agricultural communities. Since the 1870s, farm policy has played a large role in the extensification and intensification of Great Plains agriculture (Samson & Knopf 1994). Data from the National Census of Agriculture (USDA 2004) show that agricultural development from the late 1800s until the 1960s was marked by a high number of crops planted, particularly following the Dust Bowl in the 1930s. But in all four focal counties, crop diversity has decreased steadily since the 1960s, due to governmental support for agricultural intensification (Samson & Knopf 1994), leading to increasing acreage of fewer crops, as well as the possibility of reduced agricultural diversity for functional trait diversity.

Such changes in functional capacity may have negative effects on the long-term sustainability of Great Plains agricultural systems. The combined and often synergistic effects of species loss can make a system more vulnerable to environmental change (Folke et al. 2004). Studies have linked greater hay production biodiversity to more resilient agroecosystems that are able to provide vital services, such as fodder production, when challenged by severe weather (Schläpfer, Tucker & Seidl 2002). When comparing low versus high diversity grassland plots, year-to-year biomass measures were less variable in species-rich plots, especially in regards to drought conditions (Tilman, Reich & Knops 2006), showing greater stability in production yield. In agricultural studies of hay production, high diversity grasslands communities were found to be more economically feasible than intensified, low diversity communities because of the ability to maintain and attain target yields (Schläpfer, Tucker & Seidl 2002). Such studies show that functional diversity influences production and a reduction in FTV may lead to a decline in the stability of production, potentially because the functional traits have been selected outside of the environmental filters in exchange for high agricultural management.

It is important to note, however, that there are few incentives toward increasing crop diversification, therefore maintaining a large difference in plant diversity and functional capacity between native grasslands and agricultural communities. Five main commodities (corn, soybean, wheat, cotton and rice) receive about 90% of the subsidized farm supports in the USA (Boody et al. 2005), incentivising farmers to specialize production to these one or few crops in order to obtain consistent and guaranteed prices per bushel. Furthermore, the introduction of agricultural inputs encourages the progression toward crop specialization because the inputs provide and replace some of the lost functionality within the functionally reduced agricultural systems, allowing them to maintain productivity and buffering them from lost ecosystem services (e.g. soil carbon and soil production). Irrigation additionally allows for the replacement of soil water holding capacity and water use efficiency by providing systems with water at will. In this sense, agricultural management has effectively allowed agricultural species to be decoupled from the environmental filters which would have selected for traits similar to native species (Robertson & Swinton 2005) to adapt to natural conditions.

Increasingly intensive land use will continue to shift the functional diversity of plant communities, potentially yielding a landscape with reduced functional capacity. However, the results of this paper suggest that the degree of change in functional capacity can vary, as crop choice and planting combinations can exhibit equivalent FTVs to native grasslands, although all exhibit dramatic shifts in the location in trait space. Maintaining greater species diversity within agricultural lands may be one way to reduce the amount of functional change between native and agricultural systems and preserve high levels of functional capacity within agricultural systems. The management of increased on-farm agricultural diversity, both spatially and temporally, and the use of ecological practices can develop greater functional capacity within agricultural systems (Altieri 1999; Robertson & Swinton 2005). Initiatives to increase on-farm species diversity, such as the increased implementation of cover crops or the perennialization of cropping systems, may increase both sustaining and regulating functions while maintaining provisioning functions. Such changes in land management would allow ecosystems to recover and maintain some of the functional capacity that may have been lost in the Great Plains and elsewhere in ecosystems that have undergone similar changes.

Acknowledgements

We thank the many extension agents and researchers who assisted with data compilation, especially J. Blair, M. Dornbush, R. Elmore, S. Gleeson, D.C. Hart, T. Maxwell, C. McAllister, M. McGinty, D. Mengel, J. Morgan, J. Nippert, P. Olmstead, M. Palmer, P. Pederson, D.H. Rogers, K. Roozeboom, A. Wiedman and B. Wilsey. We also thank R. Russell for access to the compiled Census of Agriculture data and the comments of two anonymous reviewers in improving the manuscript.

Ancillary