A cross-taxonomic index for quantifying the health of farmland biodiversity

Authors


*Correspondence author. E-mail: s.j.butler@reading.ac.uk

Summary

1. The development of sustainable, multi-functional agricultural systems involves reconciling the needs of agricultural production with the objectives for environmental protection, including biodiversity conservation. However, the definition of sustainability remains ambiguous and it has proven difficult to identify suitable indicators for monitoring progress towards, and the successful achievement of, sustainability.

2. In this study, we show that a trait-based approach can be used to assess the detrimental impacts of agricultural change to a broad range of taxonomic groupings and derive a standardised index of farmland biodiversity health, built around an objective of achieving stable or increasing populations in all species associated with agricultural landscapes.

3. To demonstrate its application, we assess the health of UK farmland biodiversity relative to this goal. Our results suggest that the populations of two-thirds of 333 plant and animal species assessed are unsustainable under current UK agricultural practices.

4. We then explore the potential benefits of an agri-environment scheme, Entry Level Stewardship (ELS), to farmland biodiversity in the UK under differing levels of risk mitigation delivery. We show that ELS has the potential to make a significant contribution to progress towards sustainability targets but that this potential is severely restricted by current patterns of scheme deployment.

5.Synthesis and applications: We have developed a cross-taxonomic sustainability index which can be used to assess both the current health of farmland biodiversity and the impacts of future agricultural changes relative to quantitative biodiversity targets. Although biodiversity conservation is just one of a number of factors that must be considered when defining sustainability, we believe our cross-taxonomic index has the potential to be a valuable tool for guiding the development of sustainable agricultural systems.

Introduction

Agricultural intensification is one of the major drivers of ecosystem change and global biodiversity loss as a consequence of associated land conversion, habitat fragmentation and agrochemical application (Green et al. 2005; Tscharntke et al. 2005). Agricultural production is set to double by 2050 (Tilman et al. 2002) and factors such as climate change, agricultural policy reform and the introduction of genetically modified and bio-energy crops are all predicted to have substantial impacts on land-use patterns at local, national and global scales (Tilman et al. 2001; Rounsevell et al. 2003, 2005). Unless these changes are managed effectively, agriculture is likely to have further detrimental impacts on biodiversity and, consequently, the ecosystem services it is capable of providing (Chapin et al. 2000; Kremen, Williams & Thorp 2002).

Managing these predicted land-use and management changes effectively requires the development of sustainable, multi-functional agricultural systems whereby the needs of agricultural production are reconciled with objectives for environmental protection, including biodiversity conservation. To guide this process, appropriate biodiversity conservation targets must be identified. In addition, one must develop tools which can assess both the current health of farmland biodiversity and predict the impacts of agricultural changes, relative to these targets (Hails 2002). However, definitional and philosophical ambiguities surround the precise goals of sustainable development. Whilst most definitions incorporate a combination of development, environment and equity or economy, society and environment (Parris & Kates 2003), key questions remain over what should be sustained, what should be developed and the appropriate timeframe for such actions. These ambiguities mean that no definitive set of goals or indicators of progress towards those goals have been identified; the Compendium of Sustainable Development Indicator Initiatives currently lists 839 sustainability indicator efforts, including 178 covering the issue area of ‘Nature and Biodiversity’ (IISD 2008).

At the 2002 Johannesburg World Summit on Sustainable Development, representatives of 190 countries committed themselves to achieving a significant reduction in the rate of biodiversity loss by 2010 and adopted 18 indicators to monitor progress towards this target (UNEP 2002, 2004). However, reducing the rate of biodiversity loss does not equal achieving sustainability and the proposed indicators are likely to leave important gaps in our understanding of biodiversity loss (Balmford et al. 2005). We propose that a key biodiversity goal for sustainable agricultural systems must be for the populations of all associated species to be either stable or increasing. This is in line with the more stringent European Union target of halting biodiversity loss by 2010 (European Council 2001). To achieve this, it will be necessary to both offset the detrimental impacts of past agricultural changes that have driven biodiversity loss and to implement appropriate hazard prevention or risk mitigation strategies alongside current and future agricultural change (May et al. 2005; Pidgeon et al. 2007). As agriculture is currently a major driver of global biodiversity loss, achieving this goal will not only contribute to the delivery of sustainable, multi-functional agricultural systems but will also represent a fundamental step towards achieving wider biodiversity conservation targets.

Using farmland birds as a model system, Butler, Vickery & Norris (2007a) published a trait-based risk assessment framework capable of predicting the indirect impacts of agricultural change on biodiversity. In this study, we show that this trait-based approach can be used to assess the indirect risks of agricultural change to a broader range of taxonomic groupings comprising arable broadleaf plants, butterflies, bumblebees and mammals. We use these results, and those for farmland birds (Butler et al. 2007a), to derive a standardized cross-taxonomic index for quantifying biodiversity health in agricultural systems generally. This index is centred on a point of sustainability (POS), defined as the maximum level of risk in the agricultural landscape at which national populations are predicted to remain stable. This represents a biodiversity conservation target against which the condition of farmland biodiversity in the current landscape and the impact of future agricultural changes can be assessed. We demonstrate this by using the index to explore the potential contribution of agri-environment schemes (AES) to progress towards biodiversity sustainability targets under three scenarios of risk mitigation success. AES have been widely implemented by governments of developed countries to counteract the detrimental effects of modern agriculture on the environment (Kleijn et al. 2006). They are seen as key policy instruments for biodiversity conservation and the development of sustainable agricultural systems (EEA 2004). However, there has been considerable debate as to whether these schemes actually deliver the expected biodiversity benefits (Vickery et al. 2004; Kleijn et al. 2006; Potts et al. 2006).

Materials and methods

Butler et al. (2007a) defined the risk of agricultural change x to species y as the degree of coincidence between the environmental impacts of that change and the resource requirements of that species, adjusted for the species’ ecological resilience. Ecological resilience was defined by the breadth of a species’ resource requirements and its reliance on farmland for these resources. Using these definitions, we constructed risk assessment frameworks for arable broadleaf plants, butterflies, bumblebees and mammals. These groupings were selected to demonstrate the utility of this approach and their response to agricultural change was used as a proxy for the response of wider biodiversity; they encompass a broad range of taxonomic resolution and trophic levels, include key ecosystem service providers and vary to a certain extent in the quantity and quality of ecological data available. For each taxonomic group, resource requirements matrices were constructed, incorporating each species identified as having some association with UK farmland (Table 1). These included 190 arable broadleaf plant species, 23 butterfly species, 14 bumblebee species and 44 mammal species. Complete species lists for each taxonomic group are provided in Supporting Information Appendices S1–S4.

Table 1.   The resource requirements of each taxonomic group are classified according to simple, broad components
Taxonomic groupKey resource requirement categoriesCategories defined
  1. The potential impacts of agricultural change on diversity are then defined in terms of detrimental changes to the quantity or quality of these requirements.

BumblebeesActivity periodsEarly (April–May), mid (June–July) and late (August–September)
Foraging habitatWithin-field grass, within-field arable, margin and hedgerow
Forage plantsSee Table Bumblebee4 in Appendix S1 (Supporting information) for list of potential plant families used
Nesting habitatWithin-field grass, within-field arable, margin and hedgerow
Nesting requirementsVersatile, extensive grass, open grass, aerial nester, snail shells and bare ground
ButterfliesAdult activity periodsEarly (April–May), mid (June–July) and late (August–September)
Adult foraging habitatWithin-field grass, within-field arable, margin, hedgerow and riparian
Adult forage plantsSee Table Butterfly4 in Appendix S2 (Supporting information) for list of potential plant families used
Larval activity periodsEarly (April–May), mid (June–July) and late (August–September)
Larval foraging habitatsWithin-field grass, within-field arable, margin and hedgerow
Larval forage plantsSee Table Butterfly4 in Appendix S2 (Supporting information) for list of potential plant families used
MammalsDietBelow-ground invertebrates, above-ground invertebrates, plant material, seeds and vertebrates
Foraging habitatCropped area, margin, hedgerow and riparian
Nesting/roosting habitatCropped area, margin, hedgerow and riparian
Arable broadleaf plantsHabitatCropped area, margin and hedgerow
Germination periodSpring and autumn
Flowering periodApril, May, June and July
Nitrogen preferenceModified Ellenberg no.
Moisture preferenceModified Ellenberg no.
Susceptibility to herbicideHigh and low
BirdsDiet – summer and winterBelow-ground invertebrates, above-ground invertebrates, plant material, seeds and vertebrates
Foraging habitat – summer and winterCropped area, margin and hedgerow
Nesting habitatCropped area, margin and hedgerow

To validate these risk assessment frameworks, we assessed six key components of agricultural change in the UK – switch from spring to autumn sowing, increased agrochemical inputs, loss of non-cropped habitats (e.g. hedgerows and margins), land drainage, switch from hay to silage and increased stocking densities – for their impact on the abundance or availability of the key resource requirements of each taxonomic group (see Appendices S1–S4, Supporting information for details). These land-use and management changes are known to have had significant detrimental impacts on farmland biodiversity in the UK over the past five decades (Krebs et al. 1999). Using the resource requirements matrices, we then identified every species likely to have been adversely affected by any reduction in resource abundance or availability and calculated risk scores for each species. These risk scores reflect the proportion of a species’ resource requirements affected by that change and, when summed across all six changes, provide an overall risk score which represents the impact of past agricultural intensification on that species. Below we outline the risk assessment framework for bumblebees as an example. Full details of all risk assessment frameworks, with worked examples for each taxonomic group, are provided in Appendices S1–S4 (Supporting information).

Our risk scoring system assumes that each source of risk has equal weighting in terms of its relationship to population growth and that different risk sources have an additive effect (Butler et al. 2007a). To critically assess these assumptions, we constructed a series of more complex, alternative models that decomposed the total risk score into various component parts, allowing the weighting of different sources of risk to vary. We also created a set of models that assumed multiplicative rather than additive effects. Comparisons of these models using Akaike’s information criterion showed that our assumptions were reasonable – whilst additive models with risk score decomposed by resource requirements (arable broadleaf plants and bumblebees) and by life stage (butterflies) were supported, the basic model, with total score as the predictor variable, received substantial support across all taxonomic groups (see Appendix S6, Supporting information for details).

To assess the risk assessment framework for each taxonomic group, we related species’ risk scores with their national population trends over the period of recent agricultural intensification. Quantitative national population change data are available for arable broadleaf plants (Preston, Pearman & Dines 2002) and butterflies (Fox et al. 2006). General Linear Modelling was used to investigate the relationship between overall risk scores and population growth rates for these taxa. Only semi-quantitative national population trend data, where species are classed as having stable/increasing, possibly declining or declining populations, are available for bumblebees (Biesmeijer et al. 2006) and mammals (Battersby 2005). For these taxa, the relationships between overall risk scores and population trends were investigated using ordinal logistic regression. Full details of the validation process for all taxonomic groups are provided in Appendices S1–S4 (Supporting information).

Risk assessment framework for bumblebees

A list of 14 bumblebee species recorded as having an affinity with UK farmland was collated from the available literature (Prys-Jones & Corbet 1991; Edwards & Jenner 2005; Benton 2006). Habitat, life cycle and forage plant requirements for these species were extracted from a comprehensive data base of pollinator requirements (Biesmeijer et al. 2006) recording a simple binary response of use (Table 1). Species which are only found on farmland or only utilize one additional biotope were scored as having a major reliance on farmland, those that utilize two, three or four additional biotopes were scored as having a moderate reliance on farmland and those which utilize five or more additional biotopes were scored as having only a minor reliance on farmland. Other potential biotopes were: (i) woodlands; (ii) gardens; (iii) lowland heath; (iv) marshes and fenland; (v) montane and upland moorland; (vi) calcareous grassland; and (vii) acidic grassland and bracken.

The risk assessment framework assumes that the major mechanisms of impact on UK farmland bumblebees will be reduced forage plant abundance, reduced foraging opportunity because of loss of habitat during periods of activity and reduced nesting success. The maximum risk score potentially accrued by a species for any single agricultural change is 3. Agricultural change x will impact food abundance and foraging activity if it causes a change in foraging habitat availability and/or a change in forage plant abundance in the existing foraging habitat. It will impact nesting success if it causes a change in nesting habitat availability and/or a reduction in nest success in the existing nesting habitat. Thus:

image( eqn 1)

where Pt is the risk score associated with reduced foraging activity potential, Ft is the risk score associated with the reduction in forage plant availability, Nt is the risk score associated with reduced nest site availability and R is the species’ reliance on farmland.

image( eqn 2)

where Gt = number of generations of a species active in the activity periods affected by loss of habitat, = total number of life cycle components (i.e. sum of the number of generations in all activity periods) and = number of habitat components used by the species.

image( eqn 3)

where = number of points of coincidence between the impact on and species’ use of forage plant families and = points of coincidence between habitat use by a species and the location of its forage plants [see Table Bumblebee4 in Appendix S1 (Supporting information) for further explanation of F].

image( eqn 4)

where = number of points of coincidence between the impact on and species’ use of nest sites, = number of nest sites used and Hn = number of habitats used where nest sites are likely to occur.

Biodiversity health index

The maximum risk score potentially accrued by a species as a consequence of any one agricultural change varies between taxonomic groups according to the number of key resource requirements included in the risk assessment framework [range = 3 (bumblebees) to 6 (farmland birds), see Butler et al. (2007a) and Appendices S1–S4 (Supporting information) for details]. It is therefore not possible to use risk scores to directly compare the health of species across taxonomic groups or to assess the cross-taxonomic impact of agricultural change. To allow such comparisons, risk scores were first standardized by dividing each species’ risk score by the number of resource requirements included in their taxon’s risk assessment framework. Scores were also multiplied by −1 to reverse their polarity, so that increasingly negative scores represented increasing environmental risk. Regression models were then re-run, using these standardized risk scores as the dependent variable, to identify the POS for each taxonomic group. For taxa with quantitative population growth data, this was defined as the score at which annual population growth rates equalled zero. For taxa with semi-quantitative population growth data, this was defined as the point at which the probability of being classified as having a stable/increasing population was greater than the probability of having either a declining or possibly declining population. Given that the slope of the relationship between risk and population growth was not equal across taxonomic groups (see Results), we amalgamated species’ standardized risk scores into a cross-taxonomic index of biodiversity health by converting each species’ score using the equation:

image( eqn 5)

where POSt represents the taxon-specific POS for the given species. Species with a standardized risk score equal to their taxon’s POS were therefore given a sustainability score of zero, more negative scores were attributed to species further from the POS and positive scores were attributed to species with positive annual population growth rates. To estimate uncertainty, we re-calculated the POS for each taxonomic group based on the slope of the respective regression equation ± standard error (SE) and calculated alternative sustainability scores for each species accordingly.

Delivering sustainable agricultural systems

The potential contribution of AES to progress towards biodiversity sustainability targets will largely depend on scheme design, in terms of the management options available, and the quantity and quality of risk mitigation delivered through management agreements relative to the sources of risk in the agricultural landscape. To demonstrate our approach and explore the potential trajectory of biodiversity health recovery in response to the transition from a current agricultural landscape to a sustainable agricultural landscape, and the mechanisms guiding that trajectory, we investigated how varying levels of risk mitigation might influence progress towards biodiversity sustainability. Specifically, we separated each species’ risk score calculated during the assessment of the impacts of past agricultural change into two parts, risk derived from past changes in the cropped and the non-cropped (i.e. hedge and margin) areas of the agricultural landscape. We then re-calculated each species’ sustainability score under all possible combinations of risk offset, from 0% to 100%, in either the cropped and/or the non-cropped component.

A number of key issues, particularly relating to the spatial scaling of biodiversity response and the necessary quantity and quality of resource provision, need to be resolved before the contribution of AES to progress towards biodiversity sustainability can be examined explicitly (see Discussion). However, it is possible to use our biodiversity health index to explore the likely biodiversity response under plausible scenarios of risk mitigation success.

An example scheme, Entry Level Stewardship (ELS), was launched in England in 2005 and the UK Government has set a target of 70% of farms to be under ELS management by 2011. This is one of the main tools by which the UK Government will deliver its biodiversity objectives such as its Public Service Agreement target to reverse the long-term decline in farmland bird populations by 2020. Under this scheme, any farmer who achieves a minimum points target by implementing a range of relatively simple management options chosen from a menu receives a flat rate, per hectare payment (Grice et al. 2007). The range of options available for selection supports the targeted delivery of environmentally beneficial management to hedgerows, margins and the cropped areas of agricultural landscapes (Defra 2005). However, analysis of option uptake patterns has shown that the delivery of environmentally beneficial management is not evenly distributed between the three key components of the agricultural landscape; recent research suggests that the cropped area component of the agricultural landscape receives approximately half as much risk mitigation as the hedgerow and margin components under current ELS agreements (Butler, Vickery & Norris 2007b).

Using the response of the 333 species assessed here as a proxy for the response of wider biodiversity, we investigated the likely impact of ELS implementation on biodiversity under three scenarios: Firstly, we calculated species’ sustainability scores assuming that current patterns of scheme and option uptake lead to a 70% reduction in non-cropped area risk and a 35% reduction in cropped area risk. Secondly, we adopted a more optimistic projection of the benefits of ELS and assumed that all non-cropped area risk and 50% cropped area risk will be mitigated. Finally, we adopted a more pessimistic projection and assumed that only 50% of non-cropped area risk and 25% of cropped area risk will be mitigated.

Results

Validation of risk assessment frameworks

Risk scores derived from the assessments of the environmental effects of agricultural intensification in the UK were significantly related to the annual population growth rates of arable broadleaf plant and butterfly species over the same time period. Higher risk scores were associated with species with negative population growth rates and therefore experiencing population declines (arable broadleaf plants –F1, 184 = 16·8, < 0·001; Fig. 1a; butterflies –F1,18 = 16·2, = 0·001; Fig. 1b). A similar relationship between farmland bird population growth rates and their risk scores derived from the validation process has previously been reported (Butler et al. 2007a). For bumblebees and mammals, derived risk scores were significantly related to the probability of being classified as either having a stable/increasing, possibly declining or declining population (Fig 1c,d respectively). Species with higher risk scores had a greater probability of being classified as having declining populations (bumblebees – ordinal logistic regression: χ2 = 14·3, < 0·001; mammals – ordinal logistic regression: χ2 = 11·3, = 0·001). Risk scores for each species are provided in Tables Bumblebee6, Butterfly6, Mammal5 and Plant5 in Appendices S1–S4 (Supporting information) respectively.

Figure 1.

 The relationship between total risk score, derived from validation against past agricultural intensification, and species’ population change in each taxonomic group over the same time period. Annual population growth rates of (a) arable broadleaf plants and (b) butterflies decline with increasing risk score. Solid black line shows fitted model for all species (= 0·009–0·006 × x, R2 = 0·08; = 7·212−3·525 × x, R2 = 0·47 respectively), dashed lines show 95% confidence intervals. Mean ± 1 SE risk score for (c) bumblebees and (d) mammals increases as population trend category becomes more negative.

Biodiversity health in agricultural landscapes

Standardized risk scores, the taxon-specific risk score associated with stable populations (POS), and the sustainability scores for each species are detailed in Tables Bumblebee6, Butterfly6, Mammal5, Plant5 and Bird 1 in Appendices S1–S5 (Supporting information) respectively. The mean sustainability score (±SE) across the 333 species assessed here was −0·47 ± 0·05. Within each taxonomic group, the mean sustainability score was: −0·44 ± 0·05 (arable broadleaf plant), −0·02 ± 0·06 (butterfly), −0·35 ± 0·23 (bumblebee), −0·96 ± 0·18 (farmland bird) and −0·24 ± 0·08 (mammal). Of the 333 species included in our analyses, 223 (±SE = 121 and 280 species) had sustainability scores below zero, the threshold score associated with sustainable populations. These 223 species include 73% of mammals, 69% of farmland birds, 68% of arable broadleaf plants, 50% of bumblebees and 48% of butterflies (Fig. 2).

Figure 2.

 The distribution of sustainability scores for arable broadleaf plants, butterflies, bumblebees, farmland birds and mammals in current agricultural landscapes. A sustainability score of 0 is attributed to species at the point of sustainability (POS), more negative scores to species further from the POS and positive scores to species with positive population growth rates.

Contribution of agri-environment schemes to biodiversity sustainability

The delivery of biodiversity sustainability in UK agricultural systems requires risk mitigation in both the cropped and non-cropped components of the agricultural landscape (Fig. 3). However, it is evident from Fig. 3 that failure to deliver sufficient risk mitigation to the cropped area will be more restrictive to progress towards biodiversity sustainability than failure to deliver sufficient mitigation to the non-cropped area. For example, the threshold levels of risk mitigation to achieve a 25% reduction in the number of species with sustainability scores below zero are a 47% reduction in cropped area risk or a 27% reduction in the non-cropped area risk. To achieve a 50% reduction in the number of species with sustainability scores below zero requires at least a 65% reduction in either the cropped area risk or the non-cropped area risk. However, to achieve a 75% reduction in the number of species with sustainability scores below zero requires that at least 90% of cropped area risk and 10% of non-cropped area risk or 40% of cropped area risk and all non-cropped area risk are offset (Fig. 3).

Figure 3.

 The relationship between the degree of risk mitigation in the cropped and non-cropped components of the agricultural landscape and progress towards biodiversity sustainability. Contours identify levels of risk mitigation which are likely to deliver a 25%, 50% and 75% reduction in the number of species which currently have sustainability scores below zero.

Under our first scenario, in which we assume that current patterns of scheme and option uptake lead to a 70% reduction in non-cropped area risk and a 35% reduction in cropped area risk, populations of 80 (±SE = 47 and 95 species) of the 333 species in our analyses were predicted to continue declining. These 80 species include 39% of farmland birds, 36% of bumblebees, 25% of arable broadleaf plants, 5% of mammals and 4% of butterflies. Under the second scenario, with a more optimistic projection of the benefits of ELS, we would expect all butterfly, bumblebee and mammal species included in our analyses to have stable or increasing populations but 30% of farmland birds and 13% of arable broadleaf plants to still have sustainability scores below zero. However, under scenario 3, with a more pessimistic projection of the benefits of ELS, populations of 92 (±SE = 83 and 154 species) of the 333 species were predicted to continue to decline.

Discussion

We have shown that a trait-based framework can be used to assess and quantify the risk of agricultural change to over 300 species of plants and animals, including the providers of key ecosystem services. Whilst there are clearly a number of components of agricultural biodiversity not covered here, we have shown that this approach can be employed to assess the impact of agricultural change on biodiversity at different levels of taxonomic resolution and across a range of trophic levels, risk sources and resource requirements. We have also demonstrated that two key assumptions of this approach, that each source of risk has an equal weighting in terms of its relationship to population growth and that different sources of risk have an additive effect, are broadly justified when using this approach to undertake risk assessments. We purposefully defined broad categories of resource requirements within our risk assessment frameworks to maximize the range of species to which this approach could be applied and have shown that risk scores can be related to both quantitative and semi-quantitative population trend data. However, we recognize that data deficiencies, particularly relating to population trends, may currently preclude the potential application of this approach to a number of other taxonomic groups.

The development of clear biodiversity conservation objectives and appropriate indicators is crucial for assessing the contribution of conservation efforts to sustainable development and for guiding new strategies (Pereira & Cooper 2006). These indicators need to be rigorous, repeatable and easily understood and designed so that they can be used in models for future projections and they need to be policy-relevant, both in terms of sensitivity to changes at appropriate spatial and temporal scales and the targets used for comparisons (CBD 2003; Balmford et al. 2005; Scholes & Biggs 2005). By identifying the maximum level of risk at which stable or increasing populations are expected to persist, we show that a trait-based approach can be used to provide a cross-taxonomic assessment of biodiversity health. Contrary to most current biodiversity monitoring programmes (Balmford, Green & Jenkins 2003; Pereira & Cooper 2006), we have been able to combine disparate data sources from a range of taxonomic groups into a single index of biodiversity health which can be used to determine objective sustainability targets and to assess both the relative health of species across taxonomic groups in current agricultural landscapes and their response to land-use or management change. Our index of biodiversity health is based on the underlying objective of achieving stable populations in all species associated with farmland. However, there is clearly scope to modify this target where appropriate. For instance, it may be more appropriate to set a goal of achieving a sustainability score above zero, whereby the landscape can support increasing populations, for species which have undergone significant population declines or which provide valuable ecosystem services, such as pollination or natural pest control. The corollary is that species which are viewed as pests could be excluded from this objective. Furthermore, the POS we adopted for bumblebees and mammals, identified as the point at which the probability of being classified as having a stable/increasing population was greater than the probability of having either a declining or possibly declining population, is relatively conservative. If sustainability scores were calculated using a more stringent definition of sustainability, requiring that the likelihood of being classified as having a stable/increasing population is greater, more species would currently have sustainability scores below zero. When, for example, a probability of being classified as having a stable/increasing population of 0·8 was adopted as the POS for bumblebees and mammals, the number of species with sustainability scores below zero in the current agricultural landscape increased from 50% to 58% and from 73% to 95% respectively.

Within our framework, biodiversity is defined in terms of species richness and population abundance. Whilst maintaining other components of biodiversity, such as genetic variability or species’ interactions and processes, are equally important objectives underpinning conservation action, we believe that our index offers a valuable and informative metric for measuring biodiversity health. Indeed, Balmford et al. (2003) identified measures of populations and habitats as key global measures for the changing state of nature. Furthermore, many biodiversity objectives, such as those relating to the UK and European Farmland Bird Indices (Gregory et al. 2005), and conservation status classifications, such as the IUCN threat status (Hilton-Taylor 2000), are driven by population trends. As understanding of the link between biodiversity and ecosystem services and understanding of the value of these services increases, we believe our framework also has the potential to be used for risk assessments based on changes in ecosystem service provision and could therefore contribute greatly to the economic evaluation of proposed agricultural changes.

To demonstrate our approach, we use it to explore the health of farmland biodiversity in the UK. Our results suggest that two-thirds of the species assessed currently have sustainability scores below zero, suggesting that the level of risk in the current UK agricultural landscape is too high for it to support stable populations of these species. Furthermore, our assessment of risk mitigation following ELS implementation suggests that, even if we adopt an optimistic projection of the likely benefits, ELS is unlikely to provide sufficient environmental benefits to deliver biodiversity sustainability. In particular, our results emphasize the need to improve the delivery of environmentally beneficial management to the cropped area of the agricultural landscape. Whilst this demonstration highlights the value and applicability of our approach, we had to make a number of assumptions to undertake it which reveal key limitations and areas requiring further research which are discussed in detail below.

Firstly, our assessment of the impact of past agricultural change, which underlies the sustainability scores reported here, focused solely on the biodiversity risks of land-use and management changes and did not take into account any potential benefits to biodiversity that these changes may also have offered (Robinson, Wilson & Crick 2001). Secondly, we focused on nationally occurring land-use and management changes and relied on simple but crude assessments of the spatial congruence of species’ distribution and agricultural change to estimate national population trends. However, agricultural change can occur over regional, national and continental scales, driven by policy reform, technological advances, socio-economic and environmental factors (Mattison & Norris 2005). Further development, to take into account both the potential risks and benefits of land-use and management changes (Hails 2002; ACRE 2007) and to allow changes occurring over a range of spatial scales to be accommodated, is required to provide a more balanced impact assessment tool.

When exploring the biodiversity response to risk mitigation delivered by ELS, we assumed a linear relationship between the extent of uptake of management options and the resultant level of risk mitigation and that, where environmentally beneficial management is undertaken, the quantity of risk mitigation delivered is sufficient to ameliorate the impacts of past agricultural change. We also assumed that all species exposed to any given environmentally beneficial management will benefit to the same extent and that species’ responses to such management will be the same across all agricultural landscapes. Thirdly, we assumed that patterns of option uptake and risk mitigation reflect the patterns of risk accrued from past agricultural change so that, for instance, foraging habitat is provided where foraging habitat has been lost and nest sites are provided where nest sites have been lost. Given these assumptions, we believe our results are likely to be optimistic representations of the biodiversity response and that the actual contribution of ELS to progress towards biodiversity sustainability under each projection is likely to be even lower than that presented here. For example, a number of studies have shown that, in reality, species and taxonomic groups respond to varying extents to environmentally friendly management (Critchley et al. 2006; Marshall, West & Kleijn 2006; Walker et al. 2007) and that landscape context can have an important influence on how biodiversity responds (Rundlöf, Bengtsson & Smith 2008; Merckx et al. 2009). There is clearly a need for further research to explore the validity and consequences of these assumptions, particularly the relationship between levels of option uptake and risk mitigation, but we believe our approach and analyses provide valuable foundations for that research.

We demonstrate that a trait-based approach to risk assessment can be used to derive a quantitative measure of biodiversity health in current agricultural landscapes, to identify objective biodiversity targets and to predict the impacts of future agricultural change on progress towards those targets. Although biodiversity conservation is just one of a number of factors which must be considered when defining sustainability, we believe our cross-taxonomic index has the potential to be a valuable tool for guiding the development of sustainable agricultural systems.

Acknowledgements

This work was supported by Defra project AR0317. The authors wish to thank David Kleijn and three anonymous referees for their valuable comments on earlier versions of the manuscript.

Ancillary