The β-diversity of arable weed communities on organic and conventional cereal farms in two contrasting regions

Authors


corresponding author, larmengot@ub.edu

Abstract

Questions

Do diversity components (α, β and γ) differ across contrasting regions? What is the relative contribution of organic and conventional farming practices to the structuring of arable weed species diversity and different functional groups (legumes, grasses and broad-leaves)? To what extent do arable weed communities differ between regions and farming systems (organic vs conventional)?

Location

Twenty-six farms in total in northeast Spain (Catalonia) and north Germany (Lower Saxony).

Methods

We examined the weed flora in paired organic and conventional farms at each locality and assessed diversity components by additive partitioning of species richness (α, β and γ).

Results

The weed species composition differed greatly between the two regions. Only 18 of 135 arable weed species were shared. The α-, β- and γ-diversity of all functional groups was significantly higher under organic farming practices for both regions, indicating an increasing homogenization of local communities by agricultural intensification within each region. β-diversity contributed most to the total observed species richness in both regions (69.8% and 35.8% on organic and conventional farms, respectively, in Catalonia; and 62.4% and 53.0% on organic and conventional farms, respectively, in Lower Saxony). These results demonstrate the great importance of environmental heterogeneity and of farm-specific differences in agricultural practices for the richness of arable weed species.

Conclusions

Regardless of the substantial differences in arable weed community composition across regions, our study emphasizes the great importance of organic farming for arable weed species richness. Understanding the patterns and causes of the dissimilarity of local communities appears to be a key factor for species conservation and the development of effective European-wide agri-environmental schemes at landscape and regional scales. This approach is in contrast to current nature conservation practices that are restricted primarily to local (field- and farm-scale) implementation.

Nomenclature
Bolòs et al.

(2005)

Introduction

Intensive agricultural practices have caused declines in arable weed diversity. These declines are of great concern because arable weed diversity plays a key role in supporting biological diversity within agroecosystems (Stoate et al. 2001; Marshall et al. 2003; Franke et al. 2009). Agricultural land includes a large proportion of terrestrial biodiversity (Pimentel et al. 1992) and occupies most of the land surface of Europe (Donald et al. 2006). Therefore, conservation efforts in this habitat type are a crucial issue. Measures such as organic farming have been promoted to reverse the current decrease of arable vegetation, the most endangered vegetation type in Europe (van Elsen 2000). However, the adoption of organic management is still infrequent. Organic farming represents only 4.3% of the total area of agricultural land across Europe (Bengtsson et al. 2005; Hole et al. 2005; European Communities 2009).

The weed (i.e. non-crop plant) vegetation of arable land, consisting primarily of annual plants, is a widespread and highly dynamic vegetation type (Lososová & Cimalová 2009) that responds rapidly to changing conditions. The species richness and composition of a weed community are influenced by the regional species pool and by the local environment (Ricklefs 1987), including the microclimate, soil properties (Andreasen et al. 1991; Lososová et al. 2004; Pyšek et al. 2005) and farming practices (e.g. crop rotations, use of synthetic fertilizers and herbicides). Agricultural management therefore influences weed community structure (Pyšek & Lepš 1991; Andersson & Milberg 1998; Hyvönen & Salonen 2002; Hyvönen et al. 2005) due to differences in the species-specific sensitivity of broad-leaves, legumes and grasses (Moreby et al. 1994; Rydberg & Milberg 2000; Romero et al. 2008).

Traditional analyses of local (α-) diversity do not allow the identification of the relative role of factors operating at both local and regional spatial scales (Ricklefs 1987; Collins et al. 2002; Weiher & Howe 2003), whereas knowledge of the spatial components appears to be essential for understanding the determinants of the diversity and composition of weed communities (Clough et al. 2007) and for identifying the spatial scales that contribute most to the total diversity of a region (Gabriel et al. 2006).

Additive partitioning of diversity is widely used for different taxa (e.g. plants and insects) in studies addressing effects on a spatial scale (Roschewitz et al. 2005; Gabriel et al. 2006; Clough et al. 2007). The total species richness for a given study site (γ-diversity) can be partitioned into α-diversity (the average number of species in a sampling unit) and β-diversity (the difference between γ-diversity and α-diversity), which is a measure of the heterogeneity in community composition. This method allows direct comparisons between diversity components because they are expressed in the same units. The method can be applied at any spatial scale (Veech et al. 2002).

In this study, we assessed the diversity components of arable weeds on organic and conventionally managed farms in two contrasting regions (Catalonia and Lower Saxony). Although information is available about the components of weed diversity in central Europe (Roschewitz et al. 2005; Gabriel et al. 2006; Clough et al. 2007), information from the Mediterranean region is scarce. We expected that (1) weed communities would differ between regions due to the geographical distance and differences in climate and soil conditions; (2) the diversity under low-intensity farming (organic farming) would be higher; and (3) the heterogeneity in community composition (β-diversity) would play an important role in the total species richness in both regions.

Methods

Study site and sampling design

The study was conducted in Catalonia (northeast Spain), a Mediterranean region, and in Lower Saxony (north Germany), an oceanic, temperate region. We analysed 11 localities in Catalonia (upper and lower limits: 41°24′–42°03 N, 1°05′–2°05′E) and 15 localities in Lower Saxony (51°23′–51°52′N, 9°32′–10°10′E; Fig. 1). Both of the regions studied covered similar areas (70 km × 40 km in Catalonia and 50 km × 50 km in Lower Saxony), ranging from structurally simple to complex agricultural landscapes (mean percentage of arable land ± SE and minimum and maximum within 1 km of the centre of each locality: 54.5 ± 7.4%, 24.1% and 95.9%, respectively, in Catalonia and 56.6 ± 3.7%, 21.45, and 88.41%, respectively, in Lower Saxony). In central Catalonia, the mean annual precipitation is 515 mm, the mean annual temperature is 12.5 °C, and the elevation ranges from 420 to 715 m a.s.l. In southern Lower Saxony, the mean annual precipitation is 645 mm, the mean annual temperature is 8.7 °C, and the elevation ranges from 121 to 428 m a.s.l. The soils are predominantly loamy and clayish in both regions.

Figure 1.

Locations of the study areas (a) and study sites in Lower Saxony (b) and Catalonia (c).

We selected one organic and one conventional farm from each locality. The organic farms were managed according to European Union regulation 2092/91/EEC, which prohibits the use of synthetic fertilizers and pesticides. We surveyed from one to five cereal (wheat or barley) fields per farm, depending on the available number of cereal fields on each farm. All sampled fields from different farms were at least 1 km apart. The distance between paired farms (organic vs conventional) did not significantly differ between regions (Catalonia: 2.0 ± 0.50 km, mean ± SE; Lower Saxony: 1.45 ± 0.12 km, Student's t-test = 1.71, P = 0.12). We conducted a floristic survey before the crop harvest to evaluate the weed species richness within each farm. The surveys were performed on four 2 m × 2 m plots per farm in Catalonia and on 15 2 m × 2 m plots per farm in Lower Saxony. The plots were evenly distributed among the available cereal fields. This plot size was previously shown to be sufficient based on use in both regions (Geiger et al. 2010). The plots were randomly placed in the centre of the fields and at least 10 m from the field edge. In total, 88 plots were surveyed in Catalonia (44 on organic and 44 on conventional farms) and 450 plots in Lower Saxony (225 on organic and 225 on conventional farms).

Arable weed community composition

The species composition was analysed using a multivariate method based on presence/absence data. This analysis was conducted at the farm level, i.e. by considering all weed species surveyed within each farm. The species present in only one locality were omitted. We performed a permutational multivariate analysis of variance using distance matrices (the Jaccard dissimilarity index was applied) to analyse the effect of the region and the farming system on weed composition. This analysis permits the partitioning of a distance matrix among sources of variation and the fitting of a linear model to the matrix. The partial R-squared (r2) obtained indicates the percentage of variance that is explained by the factor analysed. The significance of each explanatory variable was obtained from F-tests based on sequential sums of squares from permutations of the raw data. The permutations within each locality were restricted to incorporate hierarchical sampling. The analyses were performed in R 2.7.1 (R Development Core Team, Vienna, Austria, 2008) using the ‘vegan’ package (R package version 1.15-2, R Core Development Team, http://vegan.r-forge.r-project.org). For each farming system, we also calculated the frequency of each species in relation to the total number of organic and conventional farms surveyed in each region. The expected number of species in each region was computed using EstimateS (Version 8.2, http://purl.oclc.org/estimates).

Additive partitioning of species richness

We used the additive partitioning method (Allan 1975) to partition the arable weed species richness of each region and farming system into its diversity components. The experimental design included the following scales: (1) plot; (2) field within farm; (3) farm within farming system; and (4) farming system within region. Therefore, the total species richness of each region (γregion) was partitioned into αplot (the mean number of species found per plot in a farm), βplot/farm (the difference between γfarm, the total species richness of each farm, and αplot) and βfarm/region (the difference between γregion and αfarm, the mean number of species found per farm). Thus, the total observed species richness of each study region (γregion) was the sum of αplot, βplot/farm and βfarm/region. We repeated the complete procedure for the comparison of different functional groups: broad-leaves (without legumes), grasses and legumes. In addition, for each farming system, we calculated the relative contribution of αplot, βplot/farm, βfarm/region and γregion to the total observed weed species richness within each region (considering both organic and conventional farms). To minimize any difference due to variation in sampling effort and to allow direct comparison between regions, we performed a permutation analysis. The data from Lower Saxony, with 15 plots per farm, were randomly resampled without replacement in groups of four surveys. The average of 1000 permutations was used to compute all diversity components.

The effect of the farming system and region on arable weed diversity components at the farm level (αplot, βplot/farm and γfarm) was analysed for all weed species and for the different functional groups using mixed models. Mixed models account for random effects and nested sampling designs. Therefore, we considered each locality (with one organic and one conventional farm) as a block, for which the departure from the average model can be considered a random effect. Orthogonal contrasts were fixed a priori to compare the different levels of the farming system (conventional vs organic) and the region (Catalonia vs Lower Saxony). The adequacy of each model was assessed through the normality and unbiasedness of the residuals and through the predictive power of the model. For Lower Saxony, we used the diversity components computed with the permuted data to correct for differences in sampling effort between the two study regions. The variables were log- or square-root transformed if necessary to meet the normality requirements for the residuals. The analyses were carried out using the ‘lme4′ package (R package version 0.999375-27, R Core Development Team, http://lme4.r-forge.r-project.org) for mixed models and ‘languageR’ (R package version 0.953, R Core Development Team, http://cran.r-project.org/web/packages/languageR/index.html) to evaluate the P-values.

Species accumulation curves

Species accumulation curves for each farming system within both regions were constructed by plotting the average number of species against the total number of samples. This plot provides a measure of the rate of accumulation of different species as the sampled area increases (Scheiner 2003). This analysis incorporates the number of species and their identity, and the slope of each curve represents a measure of the increases in diversity across scales. The order in which samples are added to species-accumulation curves affects the shape of the curve produced (Colwell & Coddington 1994). To avoid this heterogeneity, we randomized all of the samples and used the average of 9999 permutations. Because the sample area on all organic and conventional farms was the same within each region, any difference in the curve parameters is due to differences in the frequency of species among farming systems and is not a result of variation in sampling effort. The analyses were performed using the ‘vegan’ package (R Core Development Team, http://vegan.r-forge.r-project.org).

Results

Overview

Overall, we mapped 135 arable weed species, of which 76 occurred in Catalonia and 77 in Lower Saxony (Table 1). In Catalonia and Lower Saxony, 42 and 33 species, respectively, were found exclusively on organic farms, and six and five species, respectively, were found exclusively on conventional farms (for a species list, see Appendix S1).

Table 1. Total species richness in Catalonia and Lower Saxony for all weeds and for the different functional groups under organic and conventional farming. Data with four replicates per farm in Lower Saxony were generated by resampling without replacement in groups of four the 15 replicates per farm.The mean of 1000 permutations per farm is shown
DiversityCatalonia (n = 4)Lower Saxony (n = 15)Lower Saxony (n = 4)Catalonia (n = 4)Lower Saxony (n = 15)Lower Saxony (n = 4)
OrganicConventionalOrganicConventionalOrganicConventional
  1. n = number of replicates per farm.

  2. a

    Includes Equisetum arvense (Monilophyta).

  3. b

    Legumes are excluded.

All Weeds7677a63.717134704857.3938.86
Broad-Leavesb605144.155530493241.1727.49
Legumes9108.0191937.513.04
Grasses71511.857311128.978.62

Arable weed communities differed markedly between Catalonia and Lower Saxony (results of permutational analysis: SS = 3.889; r2 = 0.205, P < 0.001). For all species, only 18 weed species (ca. 13%) were shared between the two regions. This result revealed substantial differences in community composition. These differences were also apparent for the different functional groups. Only two grasses, two legumes and 14 broad-leaves were shared between regions (Appendix S1).

Weed communities differed significantly between the farming systems (SS = 1.160; r2 = 0.061, P < 0.001). Twenty-eight species (ca. 37%) were shared by both farming systems in Catalonia (grasses: three, legumes: one, broad-leaves: 24). Forty species (ca. 52%) were shared in Lower Saxony (grasses: six, legumes: two, broad-leaves: 32; Appendix S1).

Additive partitioning of diversity components

Inspection of all weed species and of different functional groups showed that αplot, βplot/farm and γfarm were significantly higher under organic management in Catalonia and Lower Saxony (Table 2). This result indicated an increasing homogenization of local communities through agricultural intensification in both regions. In general, the diversity components responded similarly to agricultural intensification. However, αplot for broad-leaves had higher values in Catalonia, and αplot for grasses had higher values on the organic farms in Lower Saxony (interaction: farming system × region). Consistently for both regions and farming systems, the comparison of the relative contribution of each diversity component of all weed species and of different functional groups showed that βfarm/region made the greatest contribution to total species richness, with higher values on organic farms compared with conventional farms. In Catalonia, the relative contribution of βfarm/region in organic and conventional systems was 69.8% and 35.8%, respectively, and 62.4% and 53.0%, respectively, in Lower Saxony (Fig. 2, Table 3).

Table 2. Coefficients and their standard errors for the linear mixed models and levels of significance for the tests of the effect of the farming system (conventional vs organic) and region (Catalonia vs Lower Saxony) on weed species richness at the farm scale
 αplotβplot/farmγfarm
  1. Permuted data were used for Lower Saxony (see text for details).

  2. a

    Legumes are excluded.

  3. +P < 0.09; *P < 0.05; ***P < 0.001.

All weeds
Intercept5.97 ± 0.315.66 ± 0.3511.63 ± 0.63
A: Farming system−3.47 ± 0.28***−2.38 ± 0.28***–5.83 ± 0.50***
B: Region0.48 ± 0.31–0.24 ± 0.350.23 ± 0.63
A × B0.12 ± 0.280.30 ± 0.280.42 ± 0.51
Broad-leavesa
Intercept4.45 ± 0.284.18 ± 0.308.63 ± 0.53
A: Farming system–2.60 ± 0.22***–1.79 ± 0.22***–4.39 ± 0.38***
B: Region0.54 ± 0.28*–0.13 ± 0.300.41 ± 0.53
A × B–0.04 ± 0.220.21 ± 0.220.16 ± 0.38
Legumes
Intercept0.45 ± 0.05–2.18 ± 0.0250.93 ± 0.11
A: Farming system–0.34 ± 0.04***–1.45 ± 0.23***–0.76 ± 0.10***
B: Region–0.04 ± 0.05–0.18 ± 0.25–0.04 ± 0.11
A × B0.02 ± 0.04–0.00 ± 0.230.05 ± 0.10
Grasses
Intercept1.05 ± 0.070.61 ± 0.051.02 ± 0.05
A: Farming system–0.45 ± 0.07***–0.07 ± 0.04+–0.20 ± 0.04***
B: Region–0.07 ± 0.07–0.08 ± 0.05+–0.06 ± 0.05
A × B–0.18 ± 0.07*0.05 ± 0.040.09 ± 0.04+
Figure 2.

Mean (±SE) of the α- and β-diversity components of all weed species and all functional groups under organic (org) and conventional (con) management in Catalonia and Lower Saxony. The summed heights of the three bars represent the total number of weed species (γregion) for each farming system. The data for Lower Saxony were generated from 1000 random subsamples for each farm by selecting four of the 15 replicates (see text for details).

Table 3. Relative contribution (%) of diversity components to total observed species richness (considering organic and conventional farms together) in Catalonia and Lower Saxony. The data for Lower Saxony were generated from 1000 random subsamples for each farm by selecting four of the 15 replicates. Catalonia: 76 weed species, 60 broad-leaves, nine legumes and seven grasses; Lower Saxony: 63.71 weed species, 44.15 broad-leaves, 8.01 legumes and 11.85 grasses
DiversityCataloniaLower Saxony
OrganicConventionalOrganicConventional
  1. a

    Legumes are excluded.

All weeds
αplot12.74.014.33.0
plot/farm9.74.413.45.1
farm/region69.835.862.453.0
region92.244.290.161.0
Broad-leavesa
αplot12.63.814.73.1
plot/farm9.24.114.35.3
farm/region68.441.364.354.0
region90.249.293.362.3
Legumes
αplot10.10.510.80.7
plot/farm11.11.511.51.4
farm/region78.89.171.535.9
region100.011.193.838.0
Grasses
αplot17.910.114.84.1
plot/farm12.010.711.16.9
farm/region70.122.149.961.8
region100.042.975.772.7

The expected number of species was very similar in both regions: 77 ± 4.0 (±SD), 69.2–84.8 (95% CI lower–upper bounds) in Catalonia and 76 ± 2.7, 70.7–81.3 in Lower Saxony. Species accumulation curves suggested that the weed diversity at large spatial scales was greater under organic than under conventional management in both regions because species accumulated more rapidly with increasing sample size (i.e. within the range of samples explored the slope of the curve was steeper for the organic farms; Fig. 3).

Figure 3.

Species accumulation curves (mean ± SE) for organic and conventional farms in (a) Catalonia and (b) Lower Saxony.

Discussion

The comparison of arable weed communities in Mediterranean (i.e. in Catalonia) and temperate (i.e. in Lower Saxony) agroecosystems revealed very similar patterns in additive diversity components as affected by farming practices, despite substantial differences in species composition across these contrasting regions.

α-, β- and γ-diversity in contrasting regions

Mediterranean climatic conditions, with irregular, stormy rainfall and dry, hot summers, have selected a particular arable weed flora. This flora differs from those of humid and temperate climates (Guillerm & Maillet 1982; Nekola & White 1999; Fried et al. 2008). Nevertheless, the α-, β- and γ-diversity were two to three times higher in organic farms compared to conventional ones despite the marked differences in the species composition of the arable weed communities in the two study regions. The inputs of herbicides and mineral fertilizers and the less-diverse rotational schemes in conventionally managed arable fields have often been shown to reduce weed species richness (Benton et al. 2003), whereas chemical-free management and more complex crop rotations in organic fields can favour species-rich weed communities (Hole et al. 2005; Romero et al. 2008), as shown in our study for broad-leaves, legumes and grasses.

A pool of broad-leaves occurred at high frequencies in both farming systems. These species included ruderal and generalist, widespread arable weeds, such as Fumaria officinalis L. (common fumitory), Galium aparine L. (goosegrass), Myosotis arvensis (L.) Hill (field forget-me-not), Papaver rhoeas L. (corn poppy), Stellaria media (L.) Vill. (common chickweed), Veronica hederifolia L. (ivy-leaved speedwell) and Viola tricolor L. (European wild pansy); perennials, such as Cirsium arvense (L.) Scop. (creeping thistle); or ruderal species with long-persistent seedbanks, such as Polygonum aviculare L. (prostrate knotweed). These generally tolerant species are less strongly affected by intensive management practices and are well known to become increasingly dominant, thereby causing agricultural problems (Tiley 2010; Torra et al. 2010). In contrast, most of the species that only occurred on organic farms were characterized by a low percentage of cover and low frequencies of occurrence, i.e. they were relatively rare. For example, Bifora testiculata (L.) Roth (European bishop), Coronilla scorpioides (L.) Koch (yellow crown vetch), Ranunculus arvensis L. (corn buttercup) and Kickxia spuria (L.) Dumort. (roundleaf fluellen) could only be found on organic farms in Catalonia, whereas Ranunculus repens L. (creeping buttercup), Mentha arvensis L. (wild mint) and Anagallis arvensis L. (scarlet pimpernel) could only be found on organic farms in Lower Saxony. Apparently, these species are sensitive to intensive agricultural practices. Diverse arable weed communities may therefore experience a marked homogenization under highly intensive farming practices (McKinney & Lockwood 1999; Albrecht 2003). The higher α-diversity of broad-leaves encountered in Catalonia could also be related to the relatively low intensity of historical agricultural production compared with the situation in Lower Saxony (Suárez et al. 2003).

The low contribution of legumes, especially on conventionally managed farms, was a common pattern in both regions. In our survey, the number and frequency of legumes substantially increased on organic farms [e.g. Medicago lupulina L. (black medick), Medicago polymorpha L. (bur clover), Vicia cracca L. (bird vetch) and Vicia hirsuta (L.) Gray (hairy vetch)]. However, this pattern was dominated primarily by the presence of cultivated species, such as Medicago sativa L. (alfalfa), Vicia ervilia (L.) Willd. (blister vetch), Vicia faba L. (broadbean), and Trifolium sp. (clover). According to van Elsen (2000), the cessation of mineral nitrogen fertilization may restore the presence of legumes in organic farming. However, the occurrence of legumes in organic fields may not only result from the abandonment of mineral fertilizers but could also be influenced by other farming practices, such as more complex rotations compared to conventional fields or the re-use of crop seeds, which allow the maintenance of weed diversity (Armengot et al. 2011).

The number of grass species was relatively low compared to total species richness in both regions, but grasses were among the most frequently recorded species. This result is consistent with the findings of Daehler (1998) that species of the plant family Poaceae were over-represented among both serious and widespread weeds. Moreover, in cereal crops, the long-term application of herbicides can favour genotypes that are tolerant and resistant to herbicide applications (Heap 1997), such as Apera spica-venti (L.) P. Beauv. (loose silkybent) and Lolium rigidum Gaudin (wimmera ryegrass) (Krzakowa & Adamczewski 2007; Preston et al. 2009), which often cause crop yield losses (Melander et al. 2008; Cirujeda & Taberner 2009). The higher diversity of grasses encountered in Lower Saxony was due to cultivated species [Avena sativa L. (common oat)] or species frequently used in pastures [e.g. Lolium multiflorum Lam. (Italian ryegrass), Lolium perenne L. (perennial ryegrass) and Alopecurus pratensis L. (meadow foxtail)]. This result could be related to the higher percentage of pastures in Lower Saxony compared to Catalonia (14.54 ± 1.33% vs 0.42 ± 0.00%, respectively, of pastures relative to total land in landscape sectors of 1 km around each farm). Therefore, the larger number of grass species may result from farming practices that introduce fodders in their rotational schemes and/or from the spillover of seeds from adjacent farms, but not from the abandonment of herbicides and mineral fertilizers.

Relative importance of β-diversity

The comparison of diversity components at both farm and regional scales (for all weeds and for the different functional groups) showed that βfarm/region contributed most to overall species richness. This finding indicated that substantial differences in weed species composition occurred among farms. Apparently, the local weed plant assemblies and weed species composition change markedly with local site characteristics, such as differences in soil properties and type, microclimate and/or water regime. The effects of environmental heterogeneity were higher at large spatial scales, i.e. higher between than within localities. In addition, the use of farm machinery may have served to homogenize weed diversity among fields within farms. Organic farming exhibited the greatest contribution to overall observed weed species richness in both regions, despite substantial dissimilarities in community composition across regions. The largest contribution of this farming system may be related to the less intensive agricultural practices and higher heterogeneity in management among farms compared with conventional farming (Clough et al. 2007; Armengot et al. 2011).

Implications for farmland management and arable weed species diversity

Heterogeneity in community composition among organic farms (β-diversity) accounts for the majority of arable weed species diversity in agricultural landscapes, despite substantial differences in community composition across regions. Therefore, future European-wide agri-environmental schemes should recognize the dissimilarity of local communities and focus on the conversion to organic farming at landscape scale to optimize limited budgets and thereby maximize arable weed species diversity across regions.

Acknowledgements

We thank M. Ohlemuller and three anonymous referees for valuable comments on the manuscript, the farmers for their collaboration, and A. Romero and J. M. Blanco-Moreno for field and office assistance. L.A. thanks her colleagues at the Agroecology Department for their kind hospitality during the research period in Göttingen. This research was funded by the Spanish Ministry of Education and Science with a fellowship to the first author and the project CGL2009-13497-C02-01. C.F., A.F., T.T. and C.T. were financially supported by the German Ministry of Research and Education (BMBF) and the German Science Foundation (DFG).

Ancillary