Biodiversity at the farm scale: A novel credit point system

Agriculture, Ecosystems and Environment 197 (2014) 195–203

Contents lists available at ScienceDirect

Agriculture, Ecosystems and Environment journal homepage: www.elsevier.com/locate/agee

Biodiversity at the farm scale: A novel credit point system Simon Birrer a, *, Judith Zellweger-Fischer a , Sibylle Stoeckli b , Fränzi Korner-Nievergelt a , Oliver Balmer b,1, Markus Jenny a , Lukas Pfiffner b a b

Swiss Ornithological Institute, Seerose 1, CH–6204 Sempach, Switzerland Research Institute of Organic Agriculture (FiBL), Ackerstrasse 113, P.O. 219, CH–5070 Frick, Switzerland

A R T I C L E I N F O

A B S T R A C T

Article history: Received 29 April 2014 Received in revised form 30 July 2014 Accepted 5 August 2014 Available online xxx

Farmland biodiversity has often been assessed, but seldom at the farm scale, although it is ultimately the farm level at which decisions are taken. Therefore, a credit point system (CPS) was developed based on 32 options known to enhance farmland biodiversity. It was verified whether the resulting CPS score and farm-scale biodiversity are correlated considering four indicator groups (plants, grasshoppers, butterflies and birds) on 133 farms in the Swiss lowland. We further compared the suitability of the CPS score in reflecting farm-scale biodiversity to three alternative habitat measures, i.e. the amount of ecological compensation areas (ECAs, i.e. agri-environment scheme options), ECAs with a high ecological quality and valuable semi-natural elements (SNEs). Species richness and density of plants, grasshoppers, butterflies and birds were analysed, for ‘all species’, stenotopic farmland species and ‘red-listed’ species within each group, resulting in 19 biodiversity measures (dependent variables). Basic models were built, first without, then by including a range of environmental variables and compared to models expanded by the CPS score or one of the three habitat measures (ECAs, high-quality ECAs or SNEs). For each of the 19 biodiversity measures, the CPS score and the three habitat measures were ranked by how much their inclusion improved the basic model, to determine which measure best captured biodiversity at the farm scale. We demonstrate that the CPS score reflects farm-scale biodiversity. For 13 out of 19 biodiversity measures, models including the CPS score performed better than those without. The CPS score was found to be the most suitable predictor for a fast and efficient assessment of farm-scale biodiversity, which makes it suitable for use in large scale agri-environment schemes. ã 2014 Elsevier B.V. All rights reserved.

Keywords: Agri-environment scheme Biodiversity indicator Birds Butterflies Ecological compensation areas Farmland Grasshoppers Plants

1. Introduction Farmland biodiversity has undergone strong declines over the past decades (Donald et al., 2002; EEA, 2013), a trend which has often been linked to agricultural intensification (Donald et al., 2002). To reverse this negative trend, agri-environment schemes (AES) have been set up in a number of EU countries and in Switzerland (Kleijn et al., 2004; Aviron et al., 2009). A decade of evaluation, however, showed that impacts of AES on biodiversity are mixed (Batáry et al., 2011), and no general increase in farmland biodiversity has been observed (EEA, 2006; Lachat et al., 2010).

* Corresponding author. Tel.: +41 41 462 97 38. E-mail address: [email protected] (S. Birrer). 1 Current address: Swiss Tropical and Public Health Institute, Socinstrasse 57, CH–4051 Basel, Switzerland. http://dx.doi.org/10.1016/j.agee.2014.08.008 0167-8809/ ã 2014 Elsevier B.V. All rights reserved.

Positive effects on biodiversity were mostly achieved in ‘narrow-and-deep’ schemes targeted at local scales or rangerestricted populations (Perkins et al., 2011) rather than in ‘broadand-shallow’ programmes (Baker et al., 2012). One reason for partial success at smaller scales (plots, farms) but failures at regional or national levels might lie in the fact that, despite participating in AESs, farmers might base their management decisions on farming optimisation processes, economic aspects or subsidy payments rather than on what is most effective for biodiversity (Jahrl et al., 2012). This has led to poor ecological quality of many implemented conservation options (Jeanneret et al., 2010) or to conservation areas being insufficient in size and connectivity (Aviron et al., 2011; Rösch et al., 2013). The principle unit of decision making is the farm (Dallimer et al., 2009; Schneider et al., 2014), and decisions about participating in AES schemes are also taken at that level. Onfarm experience shows that many farmers are in fact interested in biodiversity, but a general lack of information about ecology,

196

S. Birrer et al. / Agriculture, Ecosystems and Environment 197 (2014) 195–203

biodiversity and agri-environmental issues (Jahrl et al., 2012) seems to hinder them from managing their land in a more sustainable and wildlife-friendly manner (Home et al., 2014). Even farmers who are attentive to biodiversity matters are usually uncertain about the contribution they (could) make to enhancing biodiversity on their own farms. To fill this gap, a tool was designed to help farmers with the assessment of biodiversity-favouring measures on their land, the Credit Point System (CPS; Jenny et al., 2009). The CPS combines quantity as well as ecological quality and connectivity (spatial distribution) of 32 options known to enhance farmland biodiversity (Table 1). The CPS yields a total score for each farm. In contrast to simple proportions of ECAs per farm, the CPS weights the measures according to their presumed or measured impact on biodiversity and also includes additional biodiversity-favouring measures, such as in-field grassland and in-field arable options. The aim of this study is to investigate whether the farm-based CPS score is indeed correlated with various measures of biodiversity (derived from plants, grasshoppers, butterflies and birds), i.e. whether a farm with a higher CPS score harbours a higher biodiversity than a farm with a lower CPS score. To our knowledge, this is one of very view studies where several biodiversity indicators are assessed and where this occurs on the entire farm area (but see Schneider et al., 2014). We compared whether models including CPS score fitted the data better than models with the mere quantity of ECAs or valuable semi-natural elements (SNEs; Graf et al., 2011). ECAs would be even simpler to record for farmers while SNEs are usually used as a measure of habitat quality and diversity by ecologists. They comprise all natural elements on a farm, also those which are not/cannot be managed by the farmer. We also tested whether models including the CPS score were still valid when various environmental variables likely to affect biodiversity were added to the models

and whether these models fitted the data better than models containing proportions of ECAs or SNEs. 2. Methods 2.1. The credit point system Since it is nearly impossible for farmers to quantify biodiversity on their farms we developed a tool allowing them to assess the measures they take to enhance biodiversity, the Credit Point System (Jenny et al., 2009). The CPS was designed to compose a wide range of options with which farmers can positively influence biodiversity on their farms. The CPS consists of a catalogue of 32 such options. Farmers can “score points” by applying these measures on their farms (Jenny et al., 2009). The majority of them are options from the Swiss agri-environment scheme, so called Ecological Compensation Areas (ECAs, i.e. extensively managed meadows, hedges, wildflower and rotational fallows etc.). Additionally, ecological quality and size of individual ECAs are also recorded, according to the ‘quality’ and ‘connectivity scheme’ (Ordinance for Ecological Quality (ÖQV); Schweizerischer Bundesrat, 2001). Further, application of arable and grassland options (e.g. no herbicide application, staggered mowing etc.) as well as for the conservation of genetic diversity (heritage breeds/heirloom crops) yield points. The point assignment accounts for farm size, i.e. points are assigned for the proportion of a given measure. An overview of the options in the CPS and their assignment to credit points is given in Table 1 (a demo version of the CPS can be filled in on http://www.ipsuisse.ch/secret/frmMain.aspx?SID=248). The scores are weighted according to their known (expertbased) benefit for biodiversity, i.e. larger-sized meadows will yield more points than smaller ones and meadows with a high ecological quality (according to the ‘quality scheme’) more than those

Table 1 Contents and assignment of points in the credit point system (CPS).

A

B

C

a b

Definition/content Assessed data/ options yielding credit points Average field/parcel A plot cultivated with one crop or grassland/pastures. Average parcel size size = (UAAa ECAb)/number of fields.

Credit point assignment and range of scores

1–3 points, with smaller parcels yielding more points (only inverse relationship between a measurement and credit point assignment) Number of land-use Arable crops, mown grass, pastures, litter meadows (similar to rush pastures, 1–3 points types but cut rather than grazed and originally used as litter for cattle), horticultures, vineyards, vegetables, other special/permanent crops. ECAs – registered To receive any subsidy payments (direct payments), farmers must manage at ECAs are summed and calculated as percentage of UAA. 1 to least 7% of their UAA as ecological compensation areas (cross-compliance). 6 points if ECAs account for more than 7% of UAA. There is a defined set of ECA types which can be registered and for which payments can be received. ECAs – high quality Farmers can apply for extra payments for ECAs with a high ecological quality A certain threshold of high-quality ECAs will yield 2 to 6 (monitored and verified periodically by experts). additional points. ECAs – structurally ECAs can be structurally enriched by stone walls, ponds and pools or by 2–6 points enriched retaining at least 5% of rough grass. High-quality ECAs larger than 0.25 ha. These ECAs are divided into 0.25-ha- All units are summed for the point score. 2 to 6 points ECAs – size units. An ECA of 1 ha thus equals four 0.25-ha-units. Number of ECA which are larger than 0.1 ha on arable and grassland, Number of ECAs per 20 ha arable and grassland, respectively is ECA – spatial distribution respectively. calculated. 2 to 6 points Several, homogenously distributed ECAs of a certain minimum size (0.1 ha) will improve connectivity of habitats on a farm. Arable options Skylark plots (undrilled patches), wider sown rows, spring crops, catch crops, 0.5–2 points per option based on the proportion of arable and under-sown crops, wildflower area management, no pesticide, no growth grassland regulators, no herbicides, no mechanical weeding after mid-April. Grassland options In extensively managed ECA grassland: use of bar mowers, staggered mowing, 0.5–2 points per option based on the proportion of arable and no-input meadows in fruit orchards, double fences. grassland Intensively managed grassland: no silage, use of bar mowers. Further options Structured forest edges, genetic diversity: Heritage breeds/heirloom crops, 0.5–2 points based on the proportion of arable and grassland specific measures for defined target species (monitored by experts).

UAA = utilised agricultural area. ECAs = Ecological Compensation Areas (options of the Swiss agri-environment scheme).

S. Birrer et al. / Agriculture, Ecosystems and Environment 197 (2014) 195–203

without. The weighting is based on results of studies addressing (i) the monitoring of AES (Kleijn et al., 2004; Knop et al., 2006; Pywell et al., 2006; Birrer et al., 2007; Meichtry-Stier et al., 2014), (ii) farming intensity (Mäder et al., 2002; Pfiffner and Luka, 2003; Hole et al., 2005; Billeter et al., 2008; Kleijn et al., 2009), (iii) landscape and habitat heterogeneity (Benton et al., 2003), (iv) nature conservation measures for (endangered) target species (Morris et al., 2004; Evans and Green, 2007), (v) own data sets, and is supplemented by expert knowledge (Smallshire et al., 2004; Schmitzberger et al., 2005). At the end, the CPS returns one single total point score for each farm. 2.2. Study area and farms To assess the correlation between biodiversity measures and the CPS score, 133 farms were selected which were located in the lowland and hill production zones of the Swiss Central Plateau below 800 m a.s.l. (Appendix A). The farms had an average size of 24.6 ha (SD 4.3) (Table 2) which corresponds to the national lowland average (BFS, 2012). They all covered arable and grassland farming (mixed production farms), with an average proportion of arable crops of 39.1% (SD 17.1). 42 farms were certified organic, 80 were integrated farms (integrated production), and 11 holdings were conventional farms (meeting only cross-compliance regulations; Schweizerischer Bundesrat, 1992). To minimise edge effects, the farms had to be as spatially consolidated as possible. For each farm, detailed information on field area, grown crop types and ECAs was collected during interviews with the farmers in the winter before field work. At the same time, the CPS was filled in together with the farmers (totally ca. 2 h). Valuable semi-natural elements (SNEs), fulfilling a minimum of ecological criteria (Graf et al., 2011) were mapped by trained field workers during the vegetation period (May–August). One to two days per farm were necessary to collect the SNE data in the field. Within the basic Swiss agri-environment scheme, farmers are obliged to manage at least 7% of their utilised agricultural area as ecological compensation areas (ECAs) to get subsidy payments (cross-compliance). Farmers can voluntarily apply for the ‘quality scheme’ to receive additional payments if an ECA has a certain ecological quality (Schweizerischer Bundesrat, 2001). High-quality

197

ECAs are therefore a subset of all general ECAs (Appendix A). It is possible that plots are registered and managed as ECAs, but that they are not visible in the field due to lacking ecological quality (especially meadows). We mapped all SNEs visible in the field as a measure of all habitats of a certain ecological value (according to Graf et al., 2011). Nearly all high-quality ECAs were mapped as SNEs, but only a few regular ECAs (without ecological quality) were SNEs. Per definition, hedges were mapped as SNEs, but they are often not registered ECAs because farmers want to avoid prescriptions linked with ECA-hedges. Some SNEs cannot be officially registered as ECAs, e.g. ditches, brooks and banks because they are defined as being ‘outside of the utilised agricultural area’. Such semi-natural habitats are missing on ‘official lists’ despite their value for biodiversity. 2.3. Biodiversity measures Four taxa, plants, grasshoppers, butterflies and birds, were used to compose a total of 19 biodiversity measures. Birds were surveyed on the entire farm area, while plants and insects were recorded on transects. Transects were deployed so that all ecological compensation areas (ECAs) and crop/grassland types were represented. The average number of transects per farm was 19.9 (range: 10–38 transects per farm), transect length ranged from 10 m to 447 m and was on average 126 m (SD 67). The lengths of all transects was chosen to sum up to a predefined total length of 2500 m per farm (realised mean was 2513 m, SD 5.6). This predefined limit was always sufficient to cover all ECAs and at least one transect for each crop type on a given farm. All individuals of butterflies found within 2.5 m from the transect line (=5 m wide strip) were determined to the species level in the field, and the data stored on digital handhelds (Palmã Tungsten E2). They were recorded on six visits due to narrow search windows (and compliance with sunny and calm weather conditions). For plants and grasshoppers, abundance classes for each transect were estimated on two visits. Plants were grouped into three abundance classes (‘sporadic’, ‘reoccurring but not ground-covering’ and ‘ground-covering’) and grasshoppers into four abundance classes (1, 2–10, 11–100 and >100 individuals per 100 m of transect). Abundance classes were determined per 100 m

Table 2 Explanatory variables used in this analysis. Non-normally distributed variables were logarithmised. Farm size was scaled (mean = 0, SD = 1). Mean  SD

Group

Variable name

Explanation

Environmental variables

Year

Year of survey. Each farm was visited once between 2009 and 2011. Included as random Factor: 2009 (n = 48), 2010 (n = 48), 2011 factor. (n = 37) Farms were grouped into 4 regions, accord. to the nearest city: Berne, Solothurn, Lucerne, Zurich. Included as random factor. Farms were either organic, integrated production or conventional. Factor: Organic (n = 42), integrated (n = 80), conventional (n = 11) Categorical variable based on temperature level maps from phenotypic data (Schreiber 10.9 (SD 0.7) et al., 1977) Area per farm (ha). Scaled and squared and cubed before built into statistical models. 24.6 (SD 4.3)

Region Farm type

CPS score ECAs High-quality ECAs SNEs

Temperature Level Area, Area2, Area3 Consolidation Degree of Consolidation: Farm area by circumference (logarithm). The bigger the value, the more consolidated the farm. Proportion (%) of arable land of farm area (value obtained from CPS) Arable Ley grass Proportion (%) of farmed area seeded with ley grass (value obtained from CPS) Livestock unit Livestock unit/ha Woodland Proportion (%) of forest/woodland edge of the circumference (m) of the farm (logarithm) Settlement Proportion (%) of settlements of the circumference of the farm (logarithm) Trees Number of trees not registered as ecological compensation area Point score per farm from the CPS Proportion (%) of ECAs on farm area Proportion of ECAs with a high ecological quality (registered and verified) on farm area. Proportion of SNEs of entire farm area.

53.3 (SD 17.6) 39.1 (SD 17.1) 24.2 (SD 13.7) 1.3 (SD 0.8) 18.5 (SD 16.5) 13.8 (SD 11.3) 5.6 (SD 12.2) 14.6 (SD 7.4) 12.5 (SD 7.0) 3.4 (SD 4.9) 2.6 (SD 3.6)

198

S. Birrer et al. / Agriculture, Ecosystems and Environment 197 (2014) 195–203

12

Number of farms

10 8 6 4 2 0 0

5

10

15 20 CPS score

25

30

35

Fig. 1. Histogram of the distribution of CPS scores on the 133 investigated farms.

of transect. When sections with differing abundance classes were observed on the same transect, a mean was taken. Since transects were relatively short and structurally homogenous, averaging was rarely needed. Birds were monitored three times on the entire farm area. All visits took place during suitable weather conditions (minimal wind, no rain) between dawn and 11 a.m. All the bird species heard or seen were recorded, resulting in presence/absence data. For bird species depending on farmland, i.e. species listed by the government as “species of the Federal Environmental Objectives of the Agriculture Sector” (EOAS; BAFU and BLW, 2008), each visual or acoustic contact or breeding behaviour was recorded on a map. The number of territories per EOAS bird species was estimated (Bibby et al., 2000; Schmid et al., 2004). Two types of species diversity measures – species richness and density – for each of the four indicator groups (plants, grasshoppers, butterflies and birds) were calculated. Species richness was defined as the total number of species observed on a given farm, and was thus expected to depend on the area of the farm. To obtain density of plants, grasshoppers and butterflies collected on transects, the data were pooled at the farm level according to proportions of crop types and ECAs; therefore density measures were expected not to depend Table 3 Number of species in total and per farm for (a) all species, (b) species defined in the Swiss “Environmental Objectives of the Agriculture Sector” (EOAS species) and (c) red-listed species. The number of farms (out of 133) where indicator groups occurred, the total number of species (i.e. species richness) and the average number of species per farm (SD) are given. Indicator group

(a) All species Plants Grasshoppers Butterflies Birds

(b) EOAS species EOAS plants EOAS grasshoppers EOAS butterflies EOAS birds

Number of farms with occurrence

Species total

Mean species per farm (SD)

133 133 133 133

773 33 69 103

152.4 (41.8) 9.0 (3.0) 18.4 (5.3) 23.5 (5.2)

133 104

257 14

34.9 (16.8) 1.7 (1.7)

131 133

36 30

5.1 (2.8) 5.4 (2.1)

94 9

1.7 (2.2) 0.5 (0.9)

19

0.8 (1.2)

16

1.3 (1.3)

(c) Red-listed species Red-listed plants 81 Red-listed 43 grasshoppers Red-listed 63 butterflies 92 Red-listed birds

on farm area. First, transect density was calculated by dividing the observed abundance on each transect by the transect area (transect length  5 m transect width). In case of butterfly density, all observed individuals of a given species on a transect over six surveys were summed. For grasshoppers and plants, the maximum abundance class observed on the first or the second visit was used. For parcels without transects (24.1 % of the parcels, n = 3139), an average transect density per crop type was used which was derived from the transect data of the corresponding crop type on that farm. When transects on a certain land-use/crop type were entirely missing on a farm (e.g. by mistake or by an unexpected change in cultivation), the average transect density per crop type had to be derived from average densities of other farms (true for 2.3% of the parcels). Individuals per parcel were then calculated by multiplying (averaged) transect densities with parcel size. To obtain the density of a certain species on a given farm, all individuals of all parcels were summed and subsequently divided by the farm area. For each of the four taxa, species richness and density were examined for the sets ‘all species’ as well as EOAS species. The latter group entails native, stenotopic species which occur mainly on farmland or depend on agricultural use as listed by the government (731 listed plant species; 48 grasshopper species, 140 butterfly species and 47 bird species; BAFU and BLW, 2008). Further, species of the Red List (Gonseth, 1994; Moser et al., 2002; Monnerat et al., 2007; Keller et al., 2010) were examined. Due to their scarcity, occurrence (presence/absence per farm) was calculated instead of species richness and density (Section 2.4). Thus, in total, 19 combinations of indicator groups, organism sets and species diversity measures were obtained and referred to as ‘biodiversity measures’ (Table 4). 2.4. Data analysis Generalised linear mixed models (Bates et al., 2011; R Development Core Team, 2011) were built, starting with basic models containing only the random factors region, year and farm. To these models the CPS score was added as a fixed factor (expanded models). The fit of each pair of basic and expanded models was compared by calculating the difference of their Akaike Information criterion (DAIC) (Akaike, 1974; Burnham and Anderson, 2002). A DAIC of at least 2 indicated that the CPS models fitted the data better than the basic models (Table 4, first column). For the models with species richness as dependent variable, Poisson error distribution was assumed and the logarithm link function was used. The random factor farm-id contained one separate level for each observation and therefore allowed for overdispersion (Gelman and Hill, 2007). For the models with density as dependent variable, normal error distribution and an identity link function were assumed. For the models with occurrence of red-listed species, a binomial error distribution was assumed and the logit link function used. In the normal and binomial (which were de facto Bernoulli) models, year and region were included as random factors. For all models, we graphically assessed whether the model assumptions were met. Particularly, the following plots were examined: residuals vs. fitted values, residuals vs. each of the explanatory variables, qq-plots of the residuals and all random effects vs. the quantiles of the normal distribution. In a second step, the CPS score was replaced with either the proportion of ECAs, high-quality ECAs or SNEs (Table 4). Again, a DAIC of at least 2 indicated that ECA or SNE models fitted the data better than the basic models. When comparing CPS score, ECAs and SNEs (i.e. between the four columns), the models with lower values were better fitting models. Further, we tested whether the correlation between the 19 biodiversity measures and CPS score was still valid when

S. Birrer et al. / Agriculture, Ecosystems and Environment 197 (2014) 195–203

100

Plants (EOAS species)

200 100

40

0 5

10

15 20 CPS score

25

5

30

Butterflies (EOAS species)

Density (territories/10 ha)

Density (individuals/ha)

60

20

0

150

Grasshoppers (EOAS species)

80

300

Density (index)

Density (index)

400

199

100

50

12

10

15 20 CPS score

25

30

25

30

Birds (EOAS species)

10 8 6 4 2 0

0 5

10

15 20 CPS score

25

5

30

10

15 20 CPS Score

Fig. 2. Relationship between the CPS score and density of EOAS plants, grasshoppers, butterflies and birds. Shown are regression lines including 95% credibility intervals (dotted lines) of the models including environmental variables. The raw data are plotted as dots.

various possibly confounding environmental variables were included: temperature level, area of the farm, degree of consolidation, proportion of arable land and ley grass, number of livestock unit per ha, trees, proportion of adjacent woodland and settlement (definitions at Table. 1). These covariates could be included because multi-collinearity among them was not strong (Appendix A). Confounding variables were retained so as not to falsely overestimate the effect of the CPS score. Although normal distribution was assumed, the skewness of each environmental variable was assessed by plotting its histogram beforehand. Heavily skewed variables were log-transformed (consolidation, woodland, settlement and trees). For numerical reasons, the area of the farm (ha) was scaled so that it's mean was zero and its standard deviation one. To account for possible nonlinear relationships between species numbers and area, the quadratic and cubic effect of area was included in addition to the linear effect in the models with species richness as dependent variables. In the models with density as dependent variable, only the linear effect of area was included as predictor. Interactions were not tested because the number of possible interactions (91 two-way interactions) would have inflated the type I error, and there were no a priori reasons to include specific interactions. Again, basic models (environmental variables and random factors) were compared with expanded models (environmental variables, random factors and CPS score, ECAs or SNEs, Table 5). To assess the significance of environmental variables and regression slopes of CPS score, ECAs and SNEs, Bayesian methods were used as recommended for generalised linear mixed models (Bolker et al., 2008). The function sim from the R-package arm (Gelman and Hill, 2007) was used to draw random simulations from the joint posterior distribution of the model parameters.

Based on the quantiles of these simulated samples from the posterior distributions, the 95%, 99% and 99.9% credible intervals (CrI) were obtained for each model parameter. If zero was not included in the 95% CrI, a significant effect was denoted. If zero was not included in the 99% or 99.9% CrI, a highly and a very highly significant effect were assumed, respectively. 3. Results The CPS scores of the 133 analysed farms ranged from 3.8 to 32.6 with an average of 14.6 (SD 7.4) (Fig. 1). The proportion of ECAs per farm area was on average 12.5% (SD 7.0) while high-quality ECAs made up a fourth of that (3.4%, SD 4.9). The proportion of SNEs was on average 2.6% (SD 3.6; Table 2). A total of 773 plant species, 33 grasshopper species, 69 butterfly species and 103 bird species were found during our survey (Table 3). 337 of the encountered species were EOAS species (34.5%; Table 3), and 138 were red-listed species (14.1%; including ‘near threatened’). Overall, 37,247 butterfly individuals were discovered, 5720 of them were EOAS species (15.4%), and 507 were red-listed species (1.4%). EOAS bird species occupied a total of 1252 territories (red-listed bird territories could not be assessed, as some of the red-listed bird species were not EOAS species). On average 152.4 (SD 41.8) plant species, 9.0 (SD 3.0) grasshopper, 18.4 (SD 5.3) butterfly and 23.5 (SD 5.2) bird species were found per farm. In all models (with or without environmental variables), the relationships between the various biodiversity measures and the CPS score, ECAs, high quality ECAs and SNEs were positive, the majority of them significantly (Appendix A), in particular in the case of CPS score and EOAS biodiversity measures (see Fig. 2).

200

S. Birrer et al. / Agriculture, Ecosystems and Environment 197 (2014) 195–203

Table 4 Comparison of models without environmental variables. The basic model (containing only random factors) was compared with the expanded models including either CPS score, ECAs, high-quality ECAs (ECA.Qs) or valuable semi-natural elements (SNEs). Shown are differences of AIC (DAIC) between the basic and expanded models for (a) all species, (b) species defined in the Swiss “Environmental Objectives of the Agriculture Sector” (EOAS species) and (c) occurrence of red-listed species. Negative values of at least 2 indicate that expanded models fitted the data better than the basic model. Comparing among the expanded models (between the columns), the best fitting model is the one with the lowest DAIC. Biodiversity measure

CPS score

ECAs

ECA.Qs

SNEs

(a) All species Plant richness Grasshopper richness Butterfly richness Bird richness Plant density Grasshopper density Butterfly density better than basic model (max 7)

31.2 4.4 19.5 14.6 23.0 7.1 23.1 7

22.9 5.1 15.0 20.2 20.1 11.6 26.7 7

15.4 2.9 7.0 5.5 9.3 5.6 0.6 6

31.4 6.7 16.0 8.3 12.9 3.2 8.8 7

(b) EOAS species Plant richness Grasshopper richness Butterfly richness Bird richness Plant density Grasshopper density Butterfly density Bird density (territories) better than basic model (max 8)

47.7 7.7 12.5 10.2 38.5 0.2 11.4 7.2 7

29.2 7.0 11.0 6.3 34.4 12.1 11.6 1.2 7

22.6 10.7 5.4 4.2 6.7 +1.0 +2.9 4.4 6

50.6 8.5 9.7 12.3 32.0 +0.2 1.7 14.3 6

(c) Red-listed species Plant occurrence Grasshopper occurrence Butterfly occurrence Bird occurrence better than basic model (max 4) Total better than basic model (max 19)

15.5 0.9 0.4 +1.5 1 15

8.2 +1.2 +2.0 0.0 1 15

7.2 7.4 4.7 +1.2 3 15

1.7 5.4 2.7 3.4 3 16

In the analysis without environmental variables, the models including CPS score were better (DAIC > 2) than the basic models (containing only random factors) for the seven measures of ‘all species’ and for seven out of eight EOAS species measures (except EOAS grasshopper density, Table 4). The CPS score improved the model fit for red-listed plant occurrence, but not for the occurrence of other red-listed taxa (Table 4). The models including ECAs were better (DAIC > 2) than the basic models for the seven measures of ‘all species’ as well as for seven out of eight measures of EOAS species (except EOAS bird density). Like the CPS score, ECAs improved the model fit for red-listed plant occurrence (Table 4). In direct comparison, the DAIC indicated better model fit for CPS score models than ECA models for ten biodiversity measures while ECAs fitted the data better than CPS score in six cases (Table 4). Judging from their AICs, high-quality ECAs improved the basic model for 15 biodiversity measures. Compared to the CPS score however, high-quality ECAs were better only for the prediction of EOAS grasshopper richness, red-listed grasshopper and red-listed butterfly occurrence whereas the CPS score was better in 14 cases (Table 4). SNE models were better than basic models for 16 out of 19 biodiversity measures (except densities of EOAS grasshopper, EOAS butterfly species and richness of red-listed plant species, Table 4). Specifically, SNEs improved models of red-listed species occurrence. Compared to the CPS score, SNEs were better predictors of biodiversity in nine cases and less suitable in nine cases (Table 4). When including environmental factors into the models, the CPS score improved the model fit (lower DAIC) for 13 biodiversity measures compared to the basic models (containing environmental and random factors). The models including CPS score had a

better fit for six measures of ‘all species’ and for six EOAS measures (Table 5, Fig. 2). The CPS score did not improve models of grasshopper richness of ‘all species’, nor densities of EOAS grasshopper and EOAS bird species. For the occurrence of red-listed species, models including CPS scores were not better than the basic models with the exception of plant occurrence. In models with the CPS score, all of the added environmental factors were significant in at least one model with the exception of squared and cubed area and livestock unit. As expected, the linear term of area was positively related to richness of EOAS plant species, all plant and all butterfly species. Density measures were not correlated with area except density of EOAS bird species which was significantly negatively related. Consolidation was significantly negatively related to nine biodiversity measures, and a negative relationship was also found in six models for the proportion of arable land (Appendix A). Models including environmental variables and ECAs were better than the basic models for ten out of 19 biodiversity measures. ECAs improved the models of three EOAS species groups, while the CPS score did so in six cases (Table 5). As for ‘all species’ (six expanded models better than basic models) and red-listed species (one expanded model better than basic models), the CPS score and ECAs performed identically. In direct comparison, the CPS score models were better than the ECA models in nine cases and less predictive in four cases. The CPS score models were better for six EOAS measures, while ECAs were never better. Directly compared, the ratio of CPS score being better than ECAs (lower DAICs) was 2:4 for all species, 6:0 for EOAS species and 1:0 for red-listed species occurrence. Models including environmental variables and high-quality ECAs instead of the CPS score improved the models for 12 out of 19 biodiversity measures, among them five models of ‘all species’,

S. Birrer et al. / Agriculture, Ecosystems and Environment 197 (2014) 195–203

201

Table 5 Comparison of models including environmental variables. The basic model included the environmental variables listed in Table 2 and was compared with the expanded models by adding either CPS score, ECAs, high-quality ECAs (ECA.Qs) or valuable semi-natural elements (SNEs). Shown are differences of AIC (DAIC) between the basic and expanded models for (a) all species, (b) species defined in the Swiss “Environmental Objectives of the Agriculture Sector” (EOAS species) and (c) occurrence of red-listed species. Negative values of at least 2 indicate that expanded models fitted the data better than the basic model. Comparing among the expanded models (between the columns), the best fitting model was the one with the lowest DAIC. Biodiversity measure

CPS score

ECAs

ECA.Qs

SNEs

a) All species Plant richness Grasshopper richness Butterfly richness Bird richness Plant density Grasshopper density Butterfly density better than basic model (max 7)

20.4 1.3 11.1 8.1 9.7 6.9 12.4 6

9.7 +1.4 5.5 10.2 10.9 8.8 15.6 6

14.3 1.8 3.9 4.8 6.1 5.4 +1.0 5

19.1 3.9 9.8 4.6 10.8 6.8 5.7 7

(b) EOAS species Plant richness Grasshopper richness Butterfly richness Bird richness Plant density Grasshopper density Butterfly density Bird density (territories) better than basic model (max 8)

32.8 4.1 6.1 3.1 18.0 +3.9 6.7 +0.2 6

13.0 0.0 0.9 1.2 17.3 1.5 4.7 0.2 3

22.4 11.6 3.3 1.2 2.9 +3.0 +3.6 1.6 4

36.4 5.6 6.4 5.0 19.0 +2.0 2.8 9.4 7

(c) Red-listed species Plant occurrence Grasshopper occurrence Butterfly occurrence Bird occurrence better than basic model (max 4) Total better than basic model (max 19)

10.4 +0.8 +1.8 +1.5 1 13

4.4 +1.9 +0.1 1.5 1 10

5.0 9.2 2.6 +1.8 3 12

0.1 2.2 0.3 2.6 2 16

and four models of EOAS species. High-quality ECAs improved the model fit for the occurrence of red-listed pants, grasshoppers and butterflies (Table 5). In direct comparison, the CPS score models were better than the high-quality ECA models in twelve cases and less good in three cases. The CPS score models were better for five EOAS measures, while high-quality ECAs were better only for EOAS grasshopper richness (all species = 6:0; EOAS species = 5:1; red-listed species = 1:2) Including SNEs improved models for 16 out 19 biodiversity measures, amongst them in all seven models of all species‘. Models with SNEs were better than the basic models in seven models of EOAS species (all except EOAS g density), and in two red-listed species occurrence models (Table 5). In direct comparison, the CPS score models were better than SNE models in seven cases and less good in ten cases. The CPS score models were better for one EOAS measure (butterfly density), while SNEs were better for six EOAS species measures (all species = 5:2; EOAS species = 1:6; red-listed species = 1:2). 4. Discussion The CPS was developed as a novel tool to measure biodiversity at the farm-scale. Analysing several indicator groups, our model comparison revealed that models including the CPS score fitted the data better than basic models in 15 cases without and 13 cases with environmental variables. The correlations between CPS score and biodiversity measures were all positive (Appendix A, Fig. 2) and the inclusion of the CPS score improved model fit (Tables 4 and 5). We did not perform an AIC model selection in the usual sense, but deliberately retained a range of possibly confounding variables in the models so as not to falsely overestimate the effect of the CPS score. But despite this conservative approach, the CPS score

remained significant in most of the models including environmental variables. Furthermore, the CPS score turned up significant more often than ECAs and high-quality ECAs. This means that measures controllable by the farmer (all the various options from the CPS) contribute substantially to predicting biodiversity. Our analysis showed that existing habitat measures, i.e. the proportion of (high-quality) ECAs and SNEs can also be used to predict farm-scale biodiversity. Creating a new predictor might seem useless at first. However, regarding our main target, the EOAS species, the CPS score was a more suitable predictor than ECAs or high-quality ECAs (CPS in six out of eight models with environmental variables, ECAs and high-quality ECAs in three out of eight models). Overall, SNEs were as suitable as the CPS score. In contrast to (high-quality) ECAs, however, proportions of SNEs are not readily available but have to be mapped, which is time-consuming and has to be done by experts. It has been repeatedly shown that the amount of semi-natural habitat is a good predictor of species richness and density, especially in differently intensive used farmland (Billeter et al., 2008; Meichtry-Stier et al., 2014), and this was confirmed in our study. SNEs and high-quality ECAs both fulfil specific criteria of habitat quality (Schweizerischer Bundesrat, 2001) which are distinctly higher than those of general ECAs. Highquality habitats are often characterised by a large structural richness and heterogeneous vegetation, e.g. herbaceous strips, low hedges with thorny shrubs, piles of stones or branches. Such a high habitat quality seems indispensible for red-listed species, but remains rare on intensively managed farmland. ECAs and high-quality ECAs were taken up in the CPS, as they were thought to be crucial determinants of farm-scale biodiversity. Our model comparisons now confirmed that the proportion of these ECAs had a positive effect on farm-scale biodiversity, but that there are obviously further important factors promoting

202

S. Birrer et al. / Agriculture, Ecosystems and Environment 197 (2014) 195–203

biodiversity. Some of those factors have apparently been included in the CPS and helped to better predict biodiversity at the farm scale than mere proportions of ECAs. A large variety of habitat types and options, specifically grassland management (e.g. mowing regime with uncut refuge areas etc.) and arable options (low-input cropping, no herbicides etc., Table 1) were adopted in the CPS, as these elements are known to benefit, among others, butterfly (Blake et al., 2011) and grasshopper species diversity (Pasinelli et al., 2013; Buri et al., 2013). It is therefore important to maintain such a broad range of options in the CPS. If necessary, the point scoring can be adjusted based on such findings; e.g. options receiving too little weight in the scoring could be elevated while apparently less important options could be downgraded. The CPS is adapted to the existing Swiss agri-environment scheme, but can be modified for other countries in relation to their AES nomenclature. The CPS can also be applied to larger farms typically found across Europe, as points are assigned for the proportion of a given measure (e.g. Neumann and Dierking, 2014). When considering red-listed species occurrence, SNEs were a suitable predictor of farm-scale biodiversity in this study. For in-depth ecological studies of such rare or endangered species, a detailed assessment of SNEs is most probably inevitable. For broadscaled conservationprojects, however, mapping and classifying SNEs is very time consuming and has to be done by experts. Our trained staff needed about one day per farm (25 ha) to map all SNEs, not including the time for subsequent data preparation. In contrast, the CPS is a simple and fast tool to perform, and ca. one hour per farm was needed to fill in the CPS in cooperation with the farmer. Another advantage of the CPS is that farmers can directly see how their decisions, for example improving the ecological quality of an ECA, affect their CPS score and they can easily distinguish the effects of different options. Thus, the CPS can also be applied as a self-evaluation tool with which farmers can assess their current biodiversity score and also run various scenarios on how to further promote biodiversity on their land. This in turn increases their motivation and self-initiative, a prerequisite for sustainable conservation of farmland biodiversity (de Snoo et al., 2013; Home et al., 2014). The CPS was originally developed within the framework of the long-term project “scoring with biodiversity - farmers enrich nature” in which the effects of farm-tailored advisory have also been investigated (Jenny et al., 2013). It was demonstrated that such a farm-tailored advisory service in combination with the CPS led farmers to significantly increase the area of suitable, highquality and site-adapted AES options compared to non-advised farms (Chevillat et al., 2012). In the meantime, the CPS has been successfully implemented in the agricultural practice. Today, 9700 Swiss farmers following rules of integrated production (IP-Suisse; Jenny et al., 2013), must apply the CPS and attain a minimal point score to receive the production label of the IP-Suisse farmers’ association. Farmers have begun to enhance biodiversity on their farms with the aid of the CPS. Thereby, registered high-quality ECAs have increased from 5.5 % of UAA in 2010 to 8.3 % in 2014 and the average CPS score has gone up by 3.8 points from 2009 to 2013 (Wunderlin and Birrer, 2013). By a large-scale use of the CPS, an increase of biodiversity is anticipated, not only at the field but also at the farm level. Acknowledgements This project was financially supported by Swiss Federal Office for Agriculture and Swiss Federal Office for the Environment, Ernst Göhner Foundation, AVINA STIFTUNG, MAVA Foundation, Sophie and Karl Binding Foundation, Dreiklang Foundation, Vontobel Foundation and Strafin Foundation. We would like to thank all participating farmers, field assistants and the entire project team

for their support. Two anonymous reviewers gave valuable comments on the manuscript. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.agee.2014.08.008. References Akaike, H., 1974. A new look at the statistical model identification. IEEE Trans. 19, 716–723. Aviron, S., Herzog, F., Klaus, I., Schüpbach, B., Jeanneret, P., 2011. Effects of wildflower strip quality, quantity, and connectivity on butterfly diversity in a Swiss arable landscape. Restor. Ecol. 19, 500–508. Aviron, S., Nitsch, H., Jeanneret, P., Buholzer, S., Luka, H., Pfiffner, L., Pozzi, S., Schüpbach, B., Walter, T., Herzog, F., 2009. Ecological cross compliance promotes farmland biodiversity in Switzerland. Front. Ecol. Environ. 5, 247–252. BAFU, BLW, 2008. Umweltziele Landwirtschaft. Hergeleitet aus bestehenden rechtlichen Grundlagen. Umwelt-Wissen 0820. Bundesamt für Umwelt (BAFU) und Bundesamt für Landwirtschaft (BLW), Bern. Baker, D.J., Freeman, S.N., Grice, P.V., Siriwardena, G.M., 2012. Landscape-scale responses of birds to agri-environment management: a test of the English Environmental Stewardship scheme. J. Appl. Ecol. 49, 871–882. Batáry, P., Báldi, A., Kleijn, D., Tscharntke, T., 2011. Landscape-moderated biodiversity effects of agri-environmental management: a meta-analysis. Proc. Roy. Soc.: Biol. Sci. 278, 1897–1902. Bates, D., Maechler, M., Bolker, B.,2011. lme4: Linear mixed-effects Models using S4Classes. R package version 0. 999, 375–42. http://CRAN.R-project.org/ package=lme4 Benton, T.G., Vickery, J.A., Wilson, J.D., 2003. Farmland biodiversity: is habitat heterogeneity the key? Trends Ecol. Evol. 18, 182–188. Bibby, C.J., Burgess, N.D., Hill, D.A., Mustoe, S.H., 2000. Bird Census Techniques. Academic Press, London. Billeter, R., Liira, J., Bailey, D., Bugter, R., Arens, P., Augenstein, I., Aviron, S., Baudry, J., Bukacek, R., Burel, F., Cerny, M., De Blust, G., et al., 2008. Indicators for biodiversity in agricultural landscapes: a pan-European study. J. Appl. Ecol. 45, 141–150. Birrer, S., Spiess, M., Herzog, F., Kohli, L., Lugrin, B., 2007. The Swiss agrienvironment scheme promotes farmland birds: but only moderately. J. Ornithol. 148 (Suppl. 2), 295–303. Blake, R.J., Woodcock, B.A., Westbury, D.B., Sutton, P., Potts, S.G., 2011. New tools to boost butterfly habitat quality in existing grass buffer strips. J. Insect Conserv. 15 (Suppl. 2), 221–232. Bolker, B.M., Brooks, M.E., Clark, C.J., Geange, S.W., Poulsen, J.R., Stevens, H.H., White, J.-S., 2008. Generalized linear mixed models: a practical guide for ecology and evolution. Trends Ecol. Evol. 24, 127–135. Buri, P., Arlettaz, R., Humbert, J.-Y., 2013. Delaying mowing and leaving uncut refuges boosts orthopterans inextensively managed meadows: evidence drawn from field-scale experimentation. Agric. Ecosyst. Environ. 181, 22–30. Burnham, K.P., Anderson, D.R., 2002. Model selection and multimodel inference, A Practical Information – Theoretic Approach. 2nd ed. Springer, New York. Chevillat, V., Balmer, O., Birrer, S., Doppler, V., Graf, R., Jenny, M., Pfiffner, L., Rudmann, C., Zellweger-Fischer, J., 2012. Gesamtbetriebliche Beratung steigert Qualität und Quantität von Ökoausgleichsflächen. Agrarforsch. Schweiz 3, 104–111. Dallimer, M., Acs, S., Hanley, N., Wilson, P., Gaston, K.J., Armsworth, P.R., 2009. What explains property-level variation in avian diversity? An inter-disciplinary approach J. Appl. Ecol. 46, 647–656. de Snoo, G.R., Herzon, I., Staats, H., Burton, R.J.F., Schindler, S., van Dijk, J., Lokhorst, A. M., Bullock, J.M., Lobley, M., Wrbka, T., Schwarz, G., Musters, C.J.M., 2013. Toward effective nature conservation on farmland: making farmers matter. Conserv. Lett. 6, 66–72. Donald, P.F., Pisano, G., Rayment, M.D., Pain, D.J., 2002. The common agricultural policy, EU enlargement and the conservation of Europe's farmland birds. Agric. Ecosyst. Environ. 89, 167–182. EEA, 2006. Progress towards halting the loss of biodiversity by 2010. Report 5/2006. European Environment Agency, Copenhagen, the Netherlands. EEA, 2013. The European grassland butterfly indicator 1990–2011. EEA Technical Report Nr. 11/2013. European Environment Agency, Luxembourg. Evans, A.D., Green, R.E., 2007. An example of tow-tired agri-environment scheme designed to deliver effectively the ecological requirements of both localised and widespread birds in England. J. Ornithol. 148 (suppl. 2), S279–S286. Gelman, A., Hill, J., 2007. Data Analysis Using Regression and Multilevel/Hierarchical Models. Cambridge University Press, Cambridge. Gonseth, Y., 1994. Rote liste der Gefährdeten Tagfalter der Schweiz. In: Duelli, P. (Ed.), Rote Listen der gefährdeten Tierarten in der Schweiz. Bundesamt für Umwelt, Wald und Landschaft, Bern, pp. 48–51. Graf, R., von Sury, R., Bühler, C., 2011. Monitoring naturnaher Lebensräume auf dem Landwirtschaftsbetrieb – Methode zum Projekt mit Vielfalt punkten. Schweizerische Vogelwarte Sempach, Sempach. Hole, D.G., Perkins, A.J., Wilson, J.D., Alexander, I.H., Grice, P.V., Evans, A.D., 2005. Does organic farming benefit biodiversity? Biol. Conserv. 122, 113–130.

S. Birrer et al. / Agriculture, Ecosystems and Environment 197 (2014) 195–203 Home, R., Balmer, O., Jahrl, I., Stolze, M., Pfiffner, L., 2014. Motivations for implementation of ecological compensation areas on Swiss lowland farms. J. Rural Stud. 34, 26–36. Jahrl, I., Rudmann, C., Pfiffner, L., Balmer, O., 2012. Motivationen für die Umsetzung von Ökoausgleichsmassnahmen. Agrarforsch. Schweiz 3, 208–215. Jeanneret, P., Walter, T., Birrer, S., Herzog, F., 2010. Evaluating the performance of the Swiss agri-environmental measures for biodiversity: methods, results and questions. Asp. Appl. Biol. 100, 35–42. Jenny, M., Zellweger-Fischer, J., Balmer, O., Birrer, S., Pfiffner, L., 2013. The credit point system: an innovative approach to enhance biodiversity on farmland. Asp. Appl. Biol. 118, 23–30. Jenny, M., Fischer, J., Pfiffner, L., Birrer, S., Graf, R., 2009. Leitfaden für die Anwendung des Punktesystems Biodiversität IP-SUISSE, Version 2009. IP-SUISSE, Zollikofen und Schweizerische Vogelwarte, Sempach. Keller, V., Gerber, A., Schmid, H., Volet, B., Zbinden, N., 2010. Rote Liste Brutvögel. Gefährdete Arten der Schweiz, Stand. Bundesamt für Umwelt BAFU, Schweizerische Vogelwarte, Bern, Sempach, pp. 1019. Kleijn, D., Berendse, F., Smit, R., Gilissen, N., Smit, J., Brak, B., Groeneveld, R., 2004. Ecological effectiveness of agri-environment schemes in different agricultural landscapes in the Netherlands. Conserv. Biol. 18, 775–786. Knop, E., Kleijn, D., Herzog, F., Schmid, B., 2006. Effectiveness of the Swiss agrienvironment scheme in promoting biodiversity. J. Appl. Ecol. 43, 120–127. Lachat, T., Pauli, D., Gonseth, Y., Klaus, G., Scheidegger, C., Vittoz, P., Walter, T., 2010. Der Wandel der Biodiversität in der Schweiz seit 1900. Ist die Talsohle erreicht? Haupt, Bern. Mäder, P., Fliessbach, A., Dubois, D., Gunst, L., Fried, P., Niggli, U., 2002. Soil fertility and biodiversity in organic farming. Science 296, 1694–1697. Meichtry-Stier, K.S., Jenny, M., Zellweger-Fischer, J., Birrer, S., 2014. Impact of landscape improvement by agri-environment scheme options on densities of characteristic farmland bird species and brown hare (Lepus europaeus). Agric. Ecosyst. Environ. 189, 101–109. Monnerat, C., Thorens, P., Walter, T., Gonseth, Y., 2007. Rote Liste Heuschrecken. Ausgabe 2007. Umwelt-Vollzug 0719. Bundesamt für Umwelt BAFU. Schweizer Zentrum für die Kartographie der Fauna, Bern. Morris, A.J., Holland, J.M., Smith, B., Jones, N.E., 2004. Sustainable arable farming for an improved environment (SAFFIE): managing winter wheat sward structure for Skylark Alauda arvensis. Ibis 146 (suppl. 2), 155–162. Moser, D.M., Gygax, A., Bäumler, B., Wyler, N., Palese, R., 2002. Rote Liste der gefährdeten Arten der Schweiz: Farn- und Blütenpflanzen. Vollzug Umwelt 9006. Bundesamt für Umwelt, Wald und Landschaft BUWAL, Zentrum des Datenverbundnetzes der Schweizer Flora. Bundesamt für Umwelt, Wald und Landschaft BUWAL, Zentrum des Datenverbundnetzes der Schweizer Flora, Conservatoire et Jardin botaniques de la Ville de Genève, Bern, Chambésy.

203

Neumann, H., Dierking, U., 2014. Ermittlung des "Biodiversitätswerts" landwirtschaftlicher Betriebe in Schleswig-Holstein. Natursch. Landschaftspl. 46, 145–152. Pasinelli, G., Meichtry-Stier, K., Birrer, S., Baur, B., Duss, M., 2013. Habitat quality and geometry affect patch occupancy of two Orthopteran species. PLoS ONE 8, e65850. Perkins, A.J., Maggs, H.E., Watson, A., Wilson, J.D., 2011. Adaptive management and targeting of agri-environment schemes does benefit biodiversity: a case study of the corn bunting Emberiza calandra. J. Appl. Ecol. 48, 514–522. Pfiffner, L., Luka, H., 2003. Effects of low-input farming system on carabids and epigeal spiders – a paired farm approach. Basic Appl. Ecol. 4, 117–127. Pywell, R.F., Warman, E.A., Hulmes, L., Hulmes, S., Nuttall, P., Sparks, T.H., Critchley, C.N.R., Sherwood, A., 2006. Effectiveness of new agri-environment schemes in providing foraging resources for bumblebees in intensively farmed landscapes. Biol. Conserv. 129, 192–206. R Development Core Team, 2011. A Language and Environment for Statistical Computing. Foundation for Statistical Computing, Vienna, Austria. Rösch, V., Tscharntke, T., Scherber, C., Batáry, P., 2013. Landscape composition, connectivity and fragment size drive effects of grassland fragmentation on insect communities. J. Appl. Ecol. 50, 387–394. Schmid, H., Zbinden, N., Keller, V., 2004. Überwachung der Bestandsentwicklung häufiger Brutvögel in der Schweiz. Schweizerische Vogelwarte, Sempach. Schmitzberger, I., Wrbka, T., Steurer, B., Aschenbrenner, G., Peterseil, J., Zechmeister, H.G., 2005. How farming styles influence biodiversity maintenance in Austrian agricultural landscapes. Agric. Ecosyst. Environ. 108, 274–290. Schneider, M.K., Lüscher, G., Jeanneret, P., Arndorfer, M., Ammari, Y., Bailey, D., Balázs, K., Báldi, A., Choisis, J.-P., Dennis, P., Eiter, S., Fjellstad, W., et al., 2014. Gains to species diversity in organically farmed fields are not propagated at the farm level. Nat. Commun. 5, 4151. Schreiber, K.F., Kuhn, N., Hug, C., Häberli, R., Schreiber, C., 1977. Wärmegliederung der Schweiz auf Grund von phänologischen Geländeaufnahmen in den Jahren 1969 bis 1973. Eidg. Justiz- und Polizeidepartement, Bern. Schweizerischer Bundesrat, 1992. Verordnung über die Direktzahlungen an die Landwirtschaft. Bundesrat, Bern, Switzerland. Schweizerischer Bundesrat, 2001. Verordnung vom 4. April 2001 über die Regionale Förderung der Qualität und der Vernetzung von ökologischen Ausgleichsflächen in der Landwirtschaft. Ökoqualitätsverordnung, ÖQV, Bern, Switzerland. Smallshire, D., Robertson, P., Thompson, P., 2004. Policy into practice: the development and delivery of agri-enironment schemes and supporting advice in England. Ibis 146 (suppl. 2), 250–258. Wunderlin, J., Birrer, S., 2013. Punktesystem Biodiversität der IP-Suisse – Stand und Entwicklung der Labelbetriebe. Schweizerische Vogelwarte, Sempach.