1Agri-environmental schemes (AES) are commonly adopted in Europe to reduce the loss of farmland biodiversity. These schemes have, however, been criticized as not fulfilling this goal, partly because their effectiveness is thought to differ depending on external factors such as landscape heterogeneity, the focal organism and scale of application.
2We used one AES, organic farming, as a landscape-scale experiment to test whether its effect on butterflies depends on the spatial scale at which it is applied. Our study system consisted of organically and conventionally managed fields within eight pairs of matched landscapes, differing in the proportion of land under organic management at the landscape scale. Butterflies and their nectar and host-plant resources were surveyed along the fields and adjacent field borders.
3Butterfly species richness and abundance were significantly increased by organic farming at the local scale. However, local butterfly species richness was also positively affected by a large proportion of organic farming in the surrounding landscape, independent of the local farming practice. Local and landscape farming practices interacted such that the farming practice within fields had a larger effect on butterfly abundance if surrounded by conventionally rather than organically managed fields. These results could only partly be explained by variation in local availability of nectar and host-plant resources.
4The total observed species richness (γ-diversity) was higher in organically managed landscapes, mainly because of higher within-field diversity (α-diversity), whereas the between-field diversity (β-diversity) tended to be similar in both landscape types.
5Synthesis and applications. Butterflies were positively affected by organic farming at a local scale, but the amount of organic farming in the surrounding landscape had either an additive (species richness) or interactive (abundance) effect. Therefore, the spatial distribution of AES must be taken into account to maximize their potential to increase farmland biodiversity. We have shown that organic farming affected butterfly species richness on nearby conventionally managed land. This suggests a landscape effect of organic farming that may indicate a wider benefit of AES for biodiversity conservation.
Differences in outcomes of AES might also be because the effect of these are scale-dependent, i.e. the effect differs depending on at what scale the scheme is applied. Bengtsson, Ahnström & Weibull (2005) suggested that different processes might be involved, depending on the scale at which organic farming is applied. If adopted on a small scale (plot or single field), the response of many organisms to farming practices may be mainly behavioural, while application on a larger scale results in a difference in population dynamics (Bengtsson, Ahnström & Weibull 2005). Assuming that areas with AES provide patches of higher quality (suitable habitat) than will be available in non-AES areas (unsuitable habitat), an increased area under AES may, according to the species–area relationship (Rosenzweig 1995), result in larger species richness within that area (i.e. increasing the species pool). In combination with the regional enrichment model (Ricklefs 1987), where local diversity is determined not only by local environmental factors but also by regional diversity, this might result in higher biodiversity in an AES area surrounded by other AES areas compared with isolated AES areas. This reasoning has been put forward by Whittingham (2007), who suggested that the value of an AES area might depend on the area of land under AES in the surrounding landscape. He proposed a ‘protected area approach’ to AES, such that if AES were adopted in large connected areas, such as whole farms or groups of farms, the biodiversity gain would be proportionally greater (Whittingham 2007). For example, small and/or isolated areas may not support populations large enough to avoid extinction (Hanski 1999), making an isolated area managed under an AES less likely to be occupied by a focal organism. On the other hand, a single field managed under an AES surrounded by non-AES fields might, by its higher habitat quality, relative to nearby habitat, attract mobile organisms from the surrounding landscape (cf. Matter & Roland 2002), thus the value of an area is relative to its surroundings. If the addition of high-quality patches results in a lower probability of local extinctions, isolated patches within AES may still be of high conservation value.
Agri-environmental schemes are large-scale manipulations of the way in which agricultural land is managed and may be an opportunity for landscape-scale experiments that are otherwise practically impossible to carry out (Herzog 2005). We used the Swedish agri-environment scheme for organic farming as an experiment to study whether its effect on butterflies differs depending on the scale to which it is applied. We hypothesized that organic farming on a local scale increases the species richness and abundance of butterflies but that the local effect depends on the proportion of organic farming in the surrounding landscape.
Materials and methods
Our study system consisted of organically and conventionally managed fields within eight pairs of matched non-overlapping landscapes (radii 1 km) differing in the proportion of arable land under organic management at the landscape level. Each pair included one landscape with a high proportion of organic arable land in the landscape [organic landscapes 57·4 ± 8·8% (mean ± SE) organic arable land within a radius of 1 km (minimum 31%, maximum 97%)] and one landscape with a low proportion of organic arable land in the landscape [conventional landscapes 9·8 ± 1·5% organic arable land (minimum 2%, maximum 16%)]. Landscapes within a pair were matched based on proximity, field size, crop composition and landscape features (e.g. stone walls, ditches and other uncultivated habitat islands) and were situated 2·5–7·5 km apart. The selected landscapes had a mean field size of 5·0 ± 0·6 ha (organic landscapes 4·9 ha, conventional landscapes 5·1 ha) and on average 5·1 ± 1·7% pasture (permanent grasslands that are grazed; organic landscapes 5·2%, conventional landscapes 5·0%), 18·3 ± 3·5% ley (semi-permanent fields for grass or grass/clover production; organic landscapes 18·3%, conventional landscapes 18·2%) and 68·4 ± 4·6% annually tilled arable land (mainly used for cereal production; organic landscapes 64·1%, conventional landscapes 72·7%). The mean field size (t = 0·23, d.f. = 7, P= 0·82) and the proportions of ley (t = 0·01, d.f. = 7, P= 0·99), pasture (t = 0·08, d.f. = 7, P= 0·94) and annually tilled arable land (t = 1·70, d.f. = 7, P= 0·13) did not differ between landscapes within a pair.
Within each landscape we selected two pairs of cereal fields (mainly spring-sown wheat and barley) and adjacent field borders as the sampling units. The fields were selected from within the 1-km landscape (as close to the centre as possible) based on the crops grown, the management in the adjacent field and the character and width of the adjoining field border. We selected one field pair matching the landscape farming practice, so that organic landscapes contained a field pair of two organically managed fields and the conventional landscapes contained a field pair of two conventionally managed fields (Fig. 1, lower field pairs). Independent of landscape type, all landscapes (except one landscape pair) also included one field pair with one field managed organically and the other managed conventionally (Fig. 1, upper field pairs). For one landscape pair, the field pair with one field of each local management type could not be obtained because the landscapes in this case were nearly totally organic (97% organic arable land) and conventional (2% organic arable land).
This design allowed us to make three different comparisons of the effects of organic farming on butterfly diversity and abundance. First, we tested the overall effect of farming practice by comparing organic fields situated in landscapes with a high proportion of organic farming with conventional fields situated in landscapes dominated by conventional farming, i.e. the same farming practice locally as in the landscape. Secondly, we tested the effects of local and landscape farming practice separately by comparing organic and conventional fields in landscapes of both types. Thirdly, we partitioned species richness (γ-diversity) into its diversity component (α- and β-diversity) and tested whether organic farming, compared with conventional farming, contributed to a higher γ-diversity because of a higher within-field (α-diversity) or between-field (β-diversity) diversity.
The study landscapes were located in a 50 × 50-km region in southern Skåne, the southernmost part of Sweden. Spatially explicit data on land use, field size and farming practice were available for this area from the Swedish Board of Agriculture. The spatial data were used in the GIS software ArcGIS (ESRI Inc., Redland, CA) to select and match landscapes and to calculate the landscape land-use statistics. The fields in our study have been managed according to the same farming practice, either organic or conventional, since at least 2001 (according to EEC regulation 2092/91; data from the Swedish Board of Agriculture).
Butterflies [Rhopalocera, including burnet moths (Zygaenidae)] were surveyed during the summers of 2004 and 2005 using a standardized transect count method (Pollard 1977; Pollard & Yates 1993). Each landscape included four 250-m transects, divided into 50-m segments, located between the selected field headlands (outer part of the cereal fields) and uncultivated field borders (permanent strip of grassy vegetation along field boundaries), with one transect per field (Fig. 1). This resulted in two pairs of transects located in the borders between fields within a field pair. The transect pairs were located 385 ± 34 m from the centre of the landscape (measured from the middle of the transects) and were on average 440 ± 50 m apart. The fields and borders within a field pair were separated by a small road, stone wall or ditch, with border type matched within landscape pairs.
All butterflies observed 5 m into field headlands and 2 m (the approximate width of the borders) into field borders were recorded as the observer walked along transects. Transects were visited five times in 2004 (between 3 June and 11 August) and six times in 2005 (between 24 May and 18 August) on sunny days with no strong winds and a temperature of at least 17 °C (Pollard & Yates 1993), with approximately the same time interval between visits. Transects within landscape pairs were visited on the same day but the order of visits to landscapes within a pair were altered between morning and afternoon to avoid bias. The landscapes were the same over both years, although the location of some transects changed between years because of crop rotation. In those cases, new fields and transects were selected based on the previously stated criteria.
nectar and host-plant surveys
We collected data on nectar-producing flowering plants and occurrence of potential larval host-plant species (excluding grasses) as measures of local habitat quality for the butterflies and used these measures as covariates in the statistical models including butterflies as the dependent variable. Nectar and host-plants (see Appendix S1 in the supplementary material) were surveyed in the same transects as the butterfly recordings. Nectar plants were recorded during two visits (between 21 June and 4 August in 2004 and between 15 June and 21 July in 2005) and host-plants during one visit (between 21 June and 7 July in 2004 and between 15 June and 27 June in 2005) to each transect in each year. The frequencies of flowering nectar plants were counted in two squares (0·5 × 0·5 m) per segment, with squares placed in the middle of the segments (i.e. 25 m from the starting point of the transect and thereafter every 50 m), one square being in the field headland and one in the uncultivated field border, i.e. 10 squares per transect and visit. Each square was divided into 25 equally sized sections, and the flowering nectar plant frequencies were noted on a scale from 0 to 25 based on the number of sections in which flowers occurred. The occurrences (presence/absence) of larval host-plant species in field borders were noted in 10 squares (0·3 × 0·3 m) in 2004 and five squares (0·5 × 0·5 m) in 2005 for each transect, with squares placed at equal distances along the transects in the uncultivated borders.
The location of squares were not marked permanently and, as some transects changed between years, the squares were not placed at the exact same location along the transects between visits and between years. The abundance of flowering nectar plants was summarized over visits at the transect level and the host-plant data were used to calculate the number of potential host-plant species per transect.
First, we tested the overall effect of farming practice by analysing species richness and abundance data from headlands and borders of field pairs with management matching the landscape farming practice, i.e. two adjacent organically managed fields in organically managed landscapes and two adjacent conventionally managed fields in conventionally managed landscapes (lower field pairs in Fig. 1). Secondly, we examined the separate effects of local and landscape farming practice by analysing data from headlands and borders of field pairs with fields differing in farming practice, i.e. adjacent organic and conventional fields in both types of landscapes (upper field pairs in Fig. 1).
We calculated the total observed species richness (γ-diversity) and used the additive partitioning approach (Lande 1996) to partition total species richness into its diversity components (α- and β-diversity) and related the three diversity measures to farming practice. In this analysis we used data from the three fields within each landscape that matched the landscape farming practice, i.e. three organically managed fields in organically managed landscapes and three conventionally managed fields in conventionally managed landscapes. Total observed species richness at the landscape level (γ-diversity) was calculated as the total number of species found within a landscape. The within-field diversity (α-diversity) was calculated as the mean number of species per transect, and the between-field diversity (β-diversity) as the difference between total species richness in a landscape and the mean number of species per transect (β=γ–α) (Gabriel et al. 2006).
We analysed the data using general linear mixed models (GLMM; SAS Proc Mixed) with normal error distribution. Species richness and abundance data were analysed at the transect level, with data pooled over visits, while the three diversity measures were analysed at the landscape level. The abundance data were log-transformed [log10 (x + 1)] prior to analyses to achieve normally distributed residuals. To account for the study design and differences between study sites, we included the following hierarchical structure of random factors in all statistical models: landscape pair and landscape type nested within pair and, in analyses at the transect level, we included transect nested within landscape type and pair.
When testing for the overall effect of farm practice, landscape type (organic or conventional, here the same as field type) was included as a fixed factor using marginal sums-of-squares. In some models we also included habitat quality measures (host-plant species richness in species richness models and nectar plant abundance) as covariates. Covariates were added to test whether farming practice affected butterflies after controlling for local habitat quality. In testing for both local and landscape effects of farming practice, we used the same fixed factor and covariates as for the first analysis but also included field type (organic or conventional) and the interaction between landscape type and field type as fixed factors. In models including covariates we determined the final model by sequentially removing non-significant factors and covariates until only significant (P < 0·05) factors/covariates remained (backward selection) and thereafter one by one adding the eliminated factors and covariates. For analyses at the landscape level we included landscape type as a fixed factor. Year was treated as a fixed factor in all statistical analyses.
Denominator degrees of freedom were estimated using the Satterthwaite method (Littell et al. 1996). We allowed variance components to be estimated occasionally as non-significantly negative to avoid inflated denominator degrees of freedom, but all results were qualitatively the same if they were constrained to zero. All statistical analyses were performed in SAS 9·1 for Windows (SAS Institute Inc., Cary, NC).
overall effect of farming practice
We found a strong effect of farming practice on both butterfly species richness (F1,7 = 98·55, P < 0·001; Fig. 2a) and abundance (F1,7 = 45·15, P < 0·001; Fig. 2b), with higher numbers of species and abundance in organic systems. The species richness was higher in 2005 than in 2004 (F1,31 = 5·51, P= 0·026), while the abundance did not differ between years (F1,31 = 0·03, P= 0·86) and year showed no interacting effects with the effect of farming practice (P > 0·5 in both cases).
The effects of farming practice on butterflies could partly, but not fully, be explained by the higher abundance of flowering plants in organically managed fields and borders. Farming practice and local abundance of flowers both explained significant portions of the variation in butterfly species richness (farming practice F1,29·8 = 52·81, P < 0·001; flower abundance F1,55·1 = 8·79, P = 0·0045) and butterfly abundance (farming practice F1,11·4 = 23·13, P < 0·001; flower abundance F1,48·5 = 5·67, P = 0·021), with higher butterfly species richness and abundance in headlands and borders of organically managed fields where flowering plants were more abundant. Host-plant species richness was eliminated from the species richness model because it did not significantly influence the number of butterfly species (F1,58·3 = 0·45, P= 0·50).
local and landscape effects of farming practice
We separated local and landscape effects of farming practice and found different effects of farming practice on butterfly species richness compared with butterfly abundance at the two scales. The species richness of butterflies was significantly higher in transects of organically managed fields (F1,12 = 12·69, P = 0·0039) and in landscapes with a high proportion of organic farming (F1,6 = 6·10, P = 0·048) compared with transects in conventionally managed fields and in conventionally managed landscapes (Fig. 3a). The local and landscape effects on butterfly species richness were additive because the interaction between local and landscape farming practice was non-significant (F1,12 = 1·16, P= 0·30; Fig. 3a). In contrast, we found a strong local effect of farming practice on butterfly abundance (F1,12 = 56·17, P < 0·001; Fig. 3b) but no significant main effect of landscape type (F1,6 = 0·36, P= 0·57; Fig. 3b). However, local farming practice interacted with the landscape type (F1,12 = 15·36, P= 0·0020; Fig. 3b) so that there was a significant effect of local farming practice only in landscapes dominated by conventional farming (F1,6 = 29·18, P= 0·0017) and not in landscapes with a high proportion of organic farming (F1,6 = 3·40, P= 0·11). Consequently, the farming practice within fields seemed to be of greater importance if the fields were surrounded by conventionally managed fields than if they were surrounded by organically managed fields. Neither species richness (F1,27 = 1·79, P = 0·19) nor abundance (F1,27 = 3·19, P = 0·085) differed significantly between years, and there were no interactions between year and the effect of local or landscape characteristics (P > 0·1 in all cases).
The abundance of flowering plants and the species richness of host-plants were significantly higher in borders along organically managed fields (flowering plants F1,12 = 36·85, P < 0·001; host-plants F1,12 = 10·96, P= 0·0062) and were not affected by the landscape farming practice (flowering plants F1,6 = 0·09, P= 0·77; host-plants F1,6 = 0·15, P= 0·71) or the interaction between the two factors (flowering plants F1,12 < 0·001, P= 0·99; host-plants F1,12 = 0·25, P= 0·63). The mean abundance of flowering plants and the mean number of host-plant species found within transects of the two farming practices at the local and landscape levels are given in Appendix S2 in the supplementary material.
When including abundance of nectar flowers and/or diversity of host-plants as covariates in the models already containing farming practice as fixed factors, we found that host-plant species richness explained additional variation in butterfly species richness but that local and landscape farming practice nevertheless affected butterfly species richness (Table 1). Neither species richness nor abundance of butterflies was significantly related to local abundance of flowers (Table 1). Instead the final model for butterfly abundance included only farming practice factors, showing the same results as previously, with an interaction between local and landscape farming practice (Table 1). On the other hand, when including only covariates as explanatory factors in the models, i.e. excluding the fixed factors for farming practice, high butterfly abundance was significantly related to high local abundance of flowers (F1,34 = 10·35, P= 0·0028) and high butterfly species richness to high species richness of host-plants (F1,39·2 = 12·95, P= 0·0010) but not to local abundance of flowers (F1,44·2 = 0·07, P = 0·79).
Table 1. Effects of farming practice at local and landscape scales on species richness and abundance of butterflies. Positive effects (+) indicate that the butterfly species richness/abundance was higher in borders adjoining organically managed fields than conventionally managed ones, or higher in borders with a high abundance of flowering nectar plants/species richness of host-plants
Local farm practice
Landscape farm practice
Local × landscape
within- and between-field diversity
The γ-diversity, the total observed species richness of butterflies, was significantly higher for the three organic fields in organic landscapes compared with the three conventional fields in landscapes dominated by conventional farming (F1,7 = 105·41, P < 0·001; Fig. 4). The mean total number of butterfly species found per year in organically managed landscapes was 12·38 ± 0·67 compared with 8·38 ± 0·71 in conventionally managed landscapes (for a species list see Appendix S3 in the supplementary material). Of all the butterfly species found in this study, seven species were found only in landscapes with a high proportion of organic farming, while one species was found exclusively in landscapes dominated by conventional farming (see Appendix S3 in the supplementary material). When the γ-diversity was partitioned into its diversity components, the α-diversity (within-field diversity) differed significantly between landscape types (F1,7 = 311·23, P < 0·001; Fig. 4), with higher diversity in organically managed landscapes, while the β-diversity (between-field diversity) showed only a tendency to be higher in organically managed landscapes (F1,7 = 4·81, P= 0·064; Fig. 4). Thus the largest contribution to differences in total diversity between farming practices was made by the within-field diversity, not the between-field diversity.
In this study we aimed to disentangle the effects of the spatial configuration of an AES (organic farming) on mobile organisms, using butterflies as a model species group. Our study design separated the local and landscape effects of organic farming, allowing us to draw conclusions regarding whether converting a single field to organic farming is sufficient to improve biodiversity or whether an aggregation of organically managed fields results in a greater effect. Organically managed fields or farms are often isolated units. This may be one reason why organic farming does not always promote biodiversity, because these isolated units may contain insufficient resources for population persistence of the organism under study (Fuller et al. 2005; Kleijn et al. 2006). To our knowledge, this is the first study to test empirically the effect of the spatial distribution of agri-environment schemes on biodiversity.
In our study system, farming practice had a significant effect on butterflies in field borders and headlands: organic farming yielded higher species richness and abundance compared with conventional farming, which is in line with other studies on butterflies (Feber et al. 1997; Rundlöf & Smith 2006; but see Weibull, Bengtsson & Nohlgren 2000). However, when we separated the local- and landscape-scale effects of organic farming, we found that butterfly species richness and abundance were differently affected by farming practice at the two scales. Butterfly species richness was positively affected by organic farming at both local and landscape scales, while the local effect of organic farming on butterfly abundance was dependent on the proportion of organic farming in the surrounding landscape.
Butterflies have at least two primary requirements for survival: larval host-plants and adult nectar resources (Boggs 2003). Farming practices that exclude the use of herbicides, such as organic farming, can lead to a higher diversity of plants in and around arable fields and a farmland habitat more attractive to butterflies (Feber et al. 1997; Hald 1999; Roschewitz et al. 2005; Rundlöf & Smith 2006). Although butterfly abundance was related to the local abundance of flowering nectar plants, the farming practice, both locally and in the surrounding landscape, was more closely linked to butterfly abundance. The surprisingly weak link between abundance of flowering plants and butterfly abundance might be because flowering plants were measured only twice during the inventory season and this may not reflect accurately the amount of available forage. Important flower aggregations along the transects may have been missed. However, the local measure of resource abundance might also be an inadequate assessment of forage resources in the surrounding landscape. The abundance of flowering plants and the species richness of host-plants were strongly related to local farming practice but not to landscape farming practice, while butterflies were influenced by both local and landscape farming practice. As a consequence, the local measure of resources, which was unrelated to resources in the surroundings, was insufficient to predict local butterfly species richness and abundance.
The local and landscape effects of farming practice seemed to be additive in predicting butterfly species richness, such that field borders situated in landscapes with a high proportion of organic farming had higher species richness irrespective of local management. This landscape effect of organic farming could be caused by at least three factors. First, the species richness of organic and conventional borders within a pair are, because of their proximity, influencing each other and creating a spill-over effect (Holt 1997). However, because of our design, with all possible combinations of local and landscape farming practices, this factor should not influence the result of a landscape effect. Secondly, organic fields in conventional landscapes might be managed more intensively than organic fields in organic landscapes, and conventional fields in organic landscapes more extensively than conventional fields in conventional landscapes. However, the organically and conventionally managed fields within a landscape are rarely managed by the same farmer (data from the Swedish Board of Agriculture), thus making this unlikely. Thirdly, organically managed fields isolated within a matrix of conventionally managed fields might not contain enough resources for viable butterfly populations, or the resources may exhibit high temporal variability, making them less likely to be occupied (Hanski 1999). Finally, landscapes with a high proportion of organic farming might contain a larger species pool than landscapes dominated by conventional farming because of a higher between-field ecological heterogeneity (cf. Benton, Vickery & Wilson 2003). Local species richness may, through dispersal, be influenced by such a larger regional species richness (Caley & Schluter 1997). However, in our study the larger species richness in landscapes dominated by organic farming was mainly caused by an increased diversity within each organic field (α-diversity), while the between-field diversity (β-diversity) only showed a tendency to be higher among organically managed fields. This result contrasts with that of Gabriel et al. (2006), who showed that organic farming made the largest contribution to total species richness of plants by an increased between-field and between-region diversity. This difference may be a consequence of the higher mobility of butterflies, as dispersal between areas increases the α-diversity at the expense of β-diversity (Cadotte 2006). The relative contribution of α- and β-diversity to γ-diversity has been shown to differ among groups of organisms (Clough et al. 2007) and the scale of the study and the environmental heterogeneity between sampled units could influence the balance between α- and β-diversity (Loreau 2000). We chose to study field headlands and borders, while Gabriel et al. (2006) conducted their study in field edges (corresponding to field headlands) and the centre of fields. Thus, if organic and conventional fields show a larger difference in environmental heterogeneity in field centres than borders, the measured difference in β-diversity between farming practices would be larger in a study conducted within the fields. Our decision to study field borders only along cereal fields may also have excluded any β-diversity on organic farms resulting from more complex crop rotations (although the land use was very similar between our organic and conventional landscapes; see the Materials and Methods). If so, between-field ecological heterogeneity could have contributed to the landscape effect in this study.
Butterfly abundance was, similarly to butterfly species richness, strongly related to local farming practice, but the effect of local management differed in magnitude depending on the farming practice in the surrounding landscape. The farming practice within a field appeared to be of larger importance if the field was surrounded by conventionally managed fields than if surrounded by organically managed fields, because the abundance of butterflies was highest in borders around isolated organically managed fields in landscapes dominated by conventional farming. Butterflies within landscapes dominated by low-quality habitat (conventional fields and borders) seemed to be able to actively find the field border containing more resources (borders adjacent to organically managed fields) and aggregate there (cf. Matter & Roland 2002). Butterflies have been shown to be able to use a non-random systematic search strategy to be able to find suitable habitat patches (Conradt et al. 2000) and can move over distances similar to our landscape scale (reviewed in Schneider 2003).
The difference in the landscape effect of organic farming on butterfly species richness and abundance might be because the most abundant butterfly species in our inventory are species that can reproduce and survive in intensively farmed agricultural landscapes (cf. Feber, Smith & Macdonald 1996; see Appendix S3 in the supplementary material). The majority of these species might therefore already be present within the landscape, unlike species depending on larger grasslands as population sources, for example (cf. Öckinger & Smith 2007). As the amount of habitat increases these species can easily respond with an increase in population size. This may also partly explain the low β-diversity in our study, as the regional species pool in the arable landscapes consists of the same farmland species. However, seven butterfly species were found only in landscapes with a high proportion of organic farming, while only one species was found exclusively in landscapes dominated by conventional farming, indicating that additional species are able to survive if a large proportion of the land is managed organically.
Butterflies are positively affected by organic farming at a local scale but the amount of organic farming in the surrounding landscape can have an additive (species richness) or an interactive effect (abundance). Our results have implications for the design of AES and highlight the importance of the effect of the spatial distribution of such schemes. Our results showed that some species were only found in landscapes with a high proportion of organic farming. We therefore suggest that the scale-dependent effect of organic farming may lead to the less common species benefiting from the spatial concentration of organic farming. However, because the farming practice at the landscape scale affected local butterfly diversity independently of local farming practice, with a positive effect of organic farming also operating outside organically managed areas, spatially dispersed organic farming would benefit local diversity. The most effective approach, spatial concentration or spatial dispersion, depends on the aim of the AES in question.
We cannot conclude from this study at what spatial scale organic farming should be concentrated or how high the proportion of organically managed land should be to increase biodiversity. These two factors are probably dependent on the organism group in focus. There are also important related questions, such as the biodiversity effects of AES outside the actual areas under schemes. Spatial concentrations of AES may reduce the negative impact on biodiversity from fragmentation of semi-natural habitats (Donald & Evans 2006). Particularly in cases where large parts of the landscape are affected by agriculture, farming practices that create a matrix facilitating interpatch movement could reduce the risk of species going regionally extinct (Vandermeer & Perfecto 2007). Organic farming adopted on a large scale has the potential to increase matrix quality and thereby improve biodiversity in agricultural landscapes.
We are grateful to the farmers who allowed us to work on their land; Emma Wendt, Maria Eng and Mathilda Edlund for help with the butterfly and plant surveys; and Erik Öckinger, James Russell and referees for valuable comments on the manuscript. The work was financed by a grant from the Swedish Research Council for Environment, Agricultural Sciences and Spatial Planning (FORMAS) (to J. Bengtsson and H. G. Smith).