There is currently little evidence that wider countryside agri-environment scheme (AES) management has led to population-level benefits for farmland birds and it is timely to consider why this might be the case, so that AES options can be improved if necessary. Two reasons why particular options might not deliver landscape-scale population benefits are that they do not take account of the full temporal scale of a resource gap for birds and that they do not provide resources at an appropriate spatial scale for the target population. This paper reviews the published evidence on both issues, focusing on the results of landscape-scale field experiments that used winter seed provision as a model AES option. There is strong evidence that most existing measures that aim to provide overwinter food resources for granivorous birds fail to do so during the late winter period, when the need for the resource is likely to be greatest. Consideration of within-season changes, such as vegetation growth and management of vegetation height, could also increase the value of options such as field margins for birds. Published research and new data together suggest that a separation of c. 1 km between winter food patches would provide all individuals in local Chaffinch Fringilla coelebs and Yellowhammer Emberiza citrinella populations with access to the resource. This separation tallies well with the recommendation in Environmental Stewardship in England that seed crop patches should be sown in patches no larger than 2 ha, with no more than 3 ha/km2. Spatial scale issues could also be critical for AES success in terms of the degree to which complementary resources, within or between seasons, are sufficiently close to one another for the same birds to access the resources that they need. Although they are based on sound evidence, modern AESs might still be improved by revision of their organization or management prescriptions. Revisions that take both the temporal and the spatial scale of resource provision into account will be necessary if farmland bird population recovery goals are to be met. Such revisions could include seed crops or mixes managed specifically to provide food in late winter and the coordination of AES agreements between neighbouring farms to ensure that sufficient resources are provided at the appropriate spatial scale.
Agri-environment schemes (AESs) are the most widely accepted management tool for promoting the conservation and population recovery of widespread but declining farmland birds. However, AESs compete with many other demands for government spending, so resources supporting them are limited. Add to this the importance of ensuring that public spending produces tangible results and it is clear that AESs need to be both successful and cost-effective.
AES options to fill resource gaps for birds (and other components of biodiversity) have tended to be based on sound logic or knowledge of natural history, but the evidence supporting that logic has sometimes been lacking, leading to failure of options to achieve their aims at the population level (Kleijn et al. 2001, Kleijn & Sutherland 2003). In Europe, many national AESs have now been in place long enough to expect effects on bird populations to be detectable. In addition, the wave of national agri-environment action that started with the reform of the Common Agricultural Policy in 2005 might now be expected to show conservation benefits. If AESs, as implemented, are not succeeding in delivering biodiversity benefits, it is important to understand why, so that timely revisions of key options can be made. In England, this is particularly relevant given the latest evidence that declines in the farmland bird population have not yet been reversed (Noble & Eaton in press) and that Entry-Level Environmental Stewardship (ELS) uptake has not yet been found to be associated with increasing bird abundance (Davey et al. 2010).
The evidence underlying agri-environment prescription design tends to come from observations of habitat use or experimental trials of areas under prescriptions demonstrating use by birds. For example, many studies have shown that granivorous farmland birds select overwintered cereal stubble fields and find much of their winter food resources there (Wilson et al. 1996, Robinson & Sutherland 1999, Hancock & Wilson 2003, Perkins et al. 2008). Studies have also shown that most granivorous birds preferentially forage in seed crops sown for game cover or specifically for wild birds, and have identified the preferred crop mixes (Hancock & Wilson 2003, Stoate et al. 2003, 2004, Henderson et al. 2004, Parish & Sotherton 2004, Perkins et al. 2008). Such studies can demonstrate the local value of the management concerned, but are rarely able to measure population-level responses, not least because of the logistical difficulties in doing so. An exception to this is the study by Gillings et al. (2005), which showed that Skylark Alauda arvensis and Yellowhammer Emberiza citrinella population trends were significantly related to the local area of cereal stubble, but even that study was unable to differentiate stubble fields in terms of quality of seed resource so as to recommend areas to be managed under specific regimes. For Skylark, it also remains uncertain whether the relationships reflected stubble area itself or the subsequent area of spring cropping, and for several granivorous species no associations with stubble area were found (Gillings et al. 2005), showing that increasing stubble alone is not a panacea for seed-eating birds.
Two factors that must be considered if local benefits of agri-environment measures are to produce population-level effects are the spatial scale necessary for the implementation of the measures, such that a sufficient proportion of the population of interest benefits, and the range of timing of the availability of the resources provided relative to the need for those resources (the ‘temporal scale’). If either of these is insufficient to fill the resource gaps at which AES options are aimed, the options will not deliver population increases. In this paper, these issues are illustrated by considering AES options that aim to provide winter seed food for granivorous farmland birds and thus to fill the resource gap that current evidence suggests is primarily responsible for limiting their abundance. Other aspects of agri-environment provision for birds where similar issues are likely to be important are also considered. First, the experimental system within which the British Trust for Ornithology has investigated these issues with the aim of providing winter food for farmland birds is described, specifically with respect to new analyses of field data. Secondly, a review of published information is combined with the results of these analyses to inform the understanding of spatial scale effects. Thirdly, the evidence relating to temporal scale issues is reviewed. Finally, the implications of the evidence for agri-environment management are discussed. The paper focuses on UK farmland because experimental and mechanistic agri-environmental research is better developed there than elsewhere and few, if any, studies addressing scale issues have been conducted in other countries. Spatial and temporal scales are potentially important, nevertheless, for the effective design of management solutions in all landscapes.
Winter food resources are likely to limit the populations of many granivorous farmland birds (e.g. Schluter & Repasky 1991, Peach et al. 1999, Siriwardena et al. 2000, 2007, Hinsley et al. 2010), and are therefore generally scarce in the farmland environment relative to demand. This means that birds should respond to the addition of winter food and that manipulating the supply of food is likely to provide an effective method of investigating patterns of resource dependence in time and space. Artificial food resource patches, consisting of regularly replenished, palatable and preferred seed types, have thus formed the basis for two landscape-scale experiments conducted in East Anglia, UK, between 2001 and 2007. The methods used in these experiments are described in detail elsewhere (Siriwardena et al. 2006, 2007, 2008) but, in summary, consisted of at least weekly monitoring of the use of the supplementary food resources by birds in winter, coupled with bi-monthly or monthly monitoring of the local populations available to use supplementary food, and radiotracking and colour-ring resighting to measure movements of individual birds directly. This intensive winter monitoring was followed by breeding season surveys, using standard protocols, to measure changes in abundance.
Optimal resource distribution across the landscape
In East Anglia, the responses of granivorous birds to the spatial distribution of supplementary food patches in winter were studied using fixed spatial arrays of replicate seed sources, each a set distance from its nearest neighbour. Analyses of patterns of bird use of these patches suggested that local groups of Yellowhammers, Chaffinches Fringilla coelebs, Goldfinches Carduelis carduelis and Reed Buntings Emberiza schoeniclus tended to share resources that were within 0.5–1 km of one another, i.e. seed patches needed to be separated by double that distance to serve discrete groups of birds (Siriwardena et al. 2006). This might, therefore, represent a sensible, cost-effective separation distance for seed-providing AES habitats. This result was supported by evidence from colour-ring resightings of Chaffinches and Yellowhammers and radiotracking of the latter (Siriwardena et al. 2006). However, the use of a fixed array of seed patches meant that the limited number of inter-patch distances could have biased the results and the reliability of regularly replenished food sources could have reduced birds’ propensity to range across the landscape. As a result, a different approach was taken in the second East Anglian experiment (2004–2007), in which radiotracking of Chaffinches and Yellowhammers was conducted in each of two fed and two control 2 × 2-km tetrads. ‘Fed’ tetrads featured a single, central food patch, which was replenished semi-weekly from November to mid-April, whereas ‘control’ tetrads were only provided with supplementary seed as targeted bait in order to concentrate existing flocks to facilitate trapping. The data collected have not been published before; the field and analytical approaches used to measure movement behaviour are described below.
Bird trapping was conducted during January and February 2005 and 2006. In fed areas, most trapping activity was focused on food patches themselves, which acted as long-term bait for bird flocks and facilitated trapping. In control areas, flocks of target species were located during field surveys and trapping was conducted at these locations. Some trapping away from food patches was also used in fed areas where other concentrations of birds had been found. Large numbers of Chaffinches and Yellowhammers were captured using mist-nets, walk-in traps and elastic-powered ‘whoosh’ nets. Samples of each species for radiotagging were selected from those trapped to include an even spread of adults, first-years, males and females. Twenty transmitters were fitted at each of two tetrads (one fed and one control) in 2005 (Bawburgh, Ordnance Survey central grid reference TG149078, and Attleborough, TM025931, respectively) and 2006 (Balsham, TL577501, and Wyken, TL954713). The radiotags consisted of Pip3 transmitters and Ag379 batteries, weighed 0.8 g, were tail-mounted and had a nominal life of 26 days at a frequency of 43 pulses/min (Biotrack Ltd, Wareham, UK). Tags of this size should not affect behaviour or survival (Naef-Daenzer et al. 2001). The initial aim was to split the sample equally between Yellowhammers and Chaffinches, but the latter proved harder to catch, so the sample was biased towards Yellowhammers. Sample sizes were also complicated by the early failure of some tags and some tags being lost by birds and re-attached to other individuals.
Birds were tracked on alternate days from fixed points located at the intersections of a 500 × 500-m grid covering focal tetrads (day 1) or from further fixed points at 500-m intervals around a square 500 m outside the focal tetrad (day 2). Occasional scans of other apparently suitable habitat (stubbles and cover strips) up to 2–3 km outside focal squares were also made to check for more extensive movements. These wider checks only rarely identified bird locations. Trials of tag detectability with respect to distance revealed that tags 2–4 m above the ground could be detected at distances of >1 km in flat terrain, but tags on the ground could only be detected up to 300 m, and those in shallow ditches at half that range. In practice, undulating terrain and obstructions such as woody vegetation will have reduced detection ranges further. Given the short tag detection ranges when birds were on the ground or in ditches, where they forage, it is likely that failures to detect individuals within or near to focal squares on many occasions probably reflected microhabitat use rather than movements beyond the area sampled. Tracking proceeded for up to 50 days after tagging, but most tags failed (or birds were predated or lost their tags) less than 30 days (median 28) after capture. Preliminary analyses indicated that data from around 10 tracking days were required to reach an asymptote in terms of range size or maximum distances moved, so individuals for which fewer days of data were available were omitted from the analyses.
Bird locations were mapped using a combination of visual records in the field of likely habitat within tag range given the bearing and strength of signals obtained and triangulation from multiple sampling points within a given day. Each different location on each day identified in this way was then digitized using ArcMap 9.2 GIS software (ESRI 2006), within which home-range areas for each individual were calculated using minimum convex polygons. As a guide to what these areas imply about linear movements, radius estimates were calculated on the assumption that the range areas were circular. Simple statistics for mean, minimum and maximum movement distances in fed and control contexts were calculated using all locations of each bird, as well as daily average (centroid) locations. Differences between birds in fed and control contexts were tested using mixed models in sas 9.2 (SAS Institute, Inc. 2008), treating bird identity and tetrad as random effects.
To identify how food resource distribution affected the observed patterns of movement in the study tetrads, preferred habitat types were identified objectively and their distributions in each tetrad were quantified. Habitat patches (as defined by status in February 2005 or 2006, as appropriate) were mapped at the field scale and added to the GIS and were differentiated into 15 categories: grass, human (rural or suburban), winter cereal, winter broadleaved crops (mostly rape or beans), stubble, pigs, permanent set-aside, spring crops (mostly bare till when the study was conducted), water, woodland, scrub, wild bird cover, roads/tracks/railways, roses (commercial rose growing covered a small, but non-trivial, area in one tetrad) and miscellaneous arable (mostly arable land outside focal tetrad boundaries that was not mapped in detail). The areas of each habitat type within each tetrad were used to define ‘background’ habitat availability (proportions of the average tetrad). Areas within buffers of a radius of 100 m from each location identified for each radiotracked bird were then used to characterize the habitats the birds selected within each tetrad. This radius was selected as being likely to encompass the habitat actually being used for foraging if birds were detected while perched near to foraging sites (because of signal detectability, most radio-locations were probably not of birds on the ground). It also reflected the likely precision of field-estimated locations. Means and 95% confidence intervals for the proportion of the area around a location for each species covered by each habitat type were calculated using a logistic model in sas 9.2 (SAS Institute, Inc. 2008), treating all records of each bird as repeated measures. These estimates were then compared with the proportion of each habitat in the ‘background’ data to identify bird-selected habitat types: if the ‘background’ proportion of a habitat lay below the lower confidence limit of the mean for the ‘selected’ proportion, the habitat was considered to be selected, whereas, if it lay above the upper limit of the selected proportion, it was considered to be avoided. Compositional analysis (Aebischer et al. 1993) was not used because most buffers around bird locations did not include all habitats considered, which would have meant the insertion of many artificial, small values in lieu of zeroes.
Spatial distributions of habitat patches in each tetrad identified as selected by birds in this way were quantified using mean inter-patch distances and the mean ratio of the square root of patch area to patch perimeter length (an index of fragmentation of the selected foraging habitats present). The experimental seed patch in each fed area was included in these calculations as an extra, bird-selected habitat patch. Mean inter-location distances for each species in each tetrad were then plotted against these indices to search for evidence of dependence of movements on resource distribution.
The movements undertaken by individual birds of both species varied considerably, from <500 m between any two average daily locations to almost 3 km between consecutive detections (Table 1). Average individuals tended to move up to c. 1.5 km between daily centroid locations and up to c. 250 m between consecutive detection points (Table 1). Despite this variation, there was no indication that movement patterns varied between fed and control areas (for Chaffinch and Yellowhammer, respectively: daily centroids: F1,251 = 0.29, P =0.592 and F1,532 = 0.02, P =0.902; all consecutive locations: F1,358 = 0.32, P =0.573 and F1,740 = 0.65, P =0.420). Estimated home-range sizes showed correspondingly little difference between fed and control areas, but the radii of these home-ranges indicated typical movement distances similar to those revealed by the earlier experiment involving fixed arrays of feeding sites (Table 2; Siriwardena et al. 2006).
Table 1. Maximum distances (metres) moved between daily centroids and consecutive locations, averaged across individuals.
Table 2. Home-range areas (estimates as minimum convex polygons around all locations of each individual) and radii (estimated assuming that the range areas were circular).
LCL and UCL are lower and upper 95% confidence limits.
Range area (ha)
Range radius (m)
Comparison of ‘background’ and ‘selected’ habitats showed that most habitats were neither selected nor avoided by either species: the habitat composition of the tetrads as a whole fell within the confidence intervals of those of the locations selected by both species, with the exception of stubble and wild bird cover crops for both species, and woodland and pig fields for Yellowhammer (Fig. 1). Stubbles and bird covers were more associated with bird locations than would be expected by chance, as were pig fields for Yellowhammers, but the latter pattern depended on the results from one study area, where the only pig field present was adjacent to a well-used stubble, so is probably an artefact. Yellowhammers also showed an avoidance of woodland (Fig. 1b). The selection of stubbles and bird covers is perhaps not surprising, but it confirms that the distribution of these habitat types is likely to influence patterns of movement in the study areas. Accordingly, plotting movement distances against the indices of bird-selected habitat distribution suggested that more dispersed/fragmented habitats tended to be associated with longer movements in both species (Fig. 2). This result cannot be strong statistically because only four tetrads were available to provide data on patterns of habitat distribution, each contributing just one data point, such that sample sizes were too small to generate reliable test statistics, and because the variation in those patterns could not be controlled. It is, however, reassuring that the patterns shown in Figure 2 are consistent with those that the rest of the study would have predicted.
Review: Other Spatial and Temporal Scale Effects
Evidence for the importance of spatial scale
AESs are funded by the taxpayer, so it is in the public interest that they achieve their aims as cost-effectively as possible. In addition, habitat patches that together supply all the resources that birds require need to be available to a substantial proportion of a population if that population is to respond to management. The spatial scale of resource provision, i.e. the location of AES option plots relative to other plots of the same or different types, as well to existing features of the environment, is critical to both of these questions. Combinations of habitats that are required at different stages of birds’ life cycles must be sufficiently close for birds to move between habitat patches providing different resources when they need to do so. In some cases, maximum distances are fairly clear, as in the case of Northern Lapwings Vanellus vanellus, which require arable habitats for nesting and grassland or damp areas for chick-rearing that must be within walking distance for very young chicks, because chick mortality increases as the distance to be covered increases (Blomqvist & Johansson 1995, Sheldon et al. 2004). If birds are able to fly between patches, however, it is much more difficult to assess what distances they regularly cover, especially with respect, for example, to movements between wintering and breeding locations, which occur only twice each year. Measuring the costs of travelling further is still more difficult, but winter colour-ringing and breeding season resighting provide one practical approach to quantify winter-to-breeding movements of granivorous birds in order to inform AES management by indicating the optimum separation of resource patches (G.M. Siriwardena, G.Q.A. Anderson & J.R. Calladine unpubl. data).
Timing of resource supply vs. demand: seed food during winter
Providing a consistent supply of supplementary food throughout the winter means that the level of exploitation of that food by birds can be taken as an index of their demand for food over and above what is present elsewhere in the environment, if it can be assumed that the additional food does not represent a ‘honeypot’ resource that draws birds away from other resources. Field experience supported this assumption: birds used the supplementary food as an additional stop on daily foraging circuits, and interference competition and perceived predation risk were likely to have reduced the attractiveness of food patches, counteracting the benefit of the locally high seed density. Both crop and weed plants set seed in summer and autumn, creating an ambient seed food resource that is only depleted after mid-winter (or earlier). Effective depletion may also not be smooth, as major perturbations such as ploughing make significant proportions of the seed resource suddenly no longer available. However, demand for seed by birds is likely to vary over time for several reasons other than the level of food supply. For example, population reduction due to mortality, population augmentation due to immigration (either of wintering or of spring and autumn passage birds) and variation in weather conditions affecting birds’ physiological energy balances can all play a part. Seasonal patterns in the use of supplementary food will vary according to which of these dominates for a given species.
Using data from the East Anglian winter-feeding experiments conducted in 2001–2004 and 2004–2007, the use of supplementary food by 11 species over winter was modelled as a curvilinear (quadratic) function of date, with 95% confidence intervals for the modelled functions being estimated by bootstrapping by feeding site (distributions of functions were created by fitting the models to each bootstrap replicate). Separate analyses were conducted on the data from the two experiments, and data on weight-of-use (species-specific bird-minutes spent on food patches) as well as on maximum counts at food patches were available for the second experiment, so a further analysis was conducted on these data. These analyses are described in full in Siriwardena et al. (2008). The results revealed a range of function forms across species, with Blue Tit Cyanistes caeruleus and Great Tit Parus major displaying early winter peaks, Common Blackbird Turdus merula, European Goldfinch, House Sparrow Passer domesticus and European Robin Erithacus rubecula mid-winter peaks, and Chaffinch, Dunnock Prunella modularis, Reed Bunting Cyanistes and Yellowhammer late winter peaks (after mid-February; Fig. 3). The species demonstrating late winter peaks are granivorous in winter (Wilson et al. 1999), are predominantly associated with farmland (e.g. Gregory & Baillie 1998), have resident breeding populations (Wernham et al. 2002) and showed population trends that were positively affected by supplementary feeding during 2001–2004 (Siriwardena et al. 2007). The same is true for House Sparrow and European Goldfinch, which had mid-winter peak counts, with the exception that House Sparrow populations are predominantly associated with human habitation (e.g. Robinson et al. 2005) and breeding Goldfinches in the UK are partially migratory (Wernham et al. 2002). It should also be noted that the data were relatively sparse for the latter two species (only one or two of the three tests used could be conducted for each of them; Fig. 3). Nevertheless, it is noteworthy that the demand for seed food of the four species whose UK breeding populations are most dependent upon winter seed resources in UK farmland peaked in late winter, between late February and April.
This result is particularly important with respect to AESs because almost all of the options designed to provide overwinter food resources for farmland birds, even experimental prescriptions designed to improve on standard ones (e.g. Hinsley et al. 2010), cease to do so after February. This occurs because of both seed depletion and vegetation dieback (especially in the case of sacrificial wild bird seed crops) and because most option rules specifically allow stubbles and seed crop plots to be ploughed in before or during the periods of peak seed demand from vulnerable species shown by Figure 3 (Table 3). This means that birds may be supported through early and mid-winter only to experience a resource bottleneck in late winter. Such a bottleneck is likely to be a natural part of granivore population dynamics, but has been made more severe by contemporary agriculture. Existing AES seed options only seem likely to provide the seed resources needed for farmland bird recovery if the additional food earlier in the winter prevents the depletion of seed in semi-natural habitats, allowing the latter to support populations in late winter. This seed resource could be important but it has never been quantified and is subject to heavy depredation from many other animals through the winter (Holmes & Froud-Williams 2005), and so seems unlikely to support high densities of granivorous birds. Among AES options, only unsprayed fodder crops in Scotland (a preferred, seed-rich habitat there; Hancock & Wilson 2003) includes a stipulation against ploughing until well into the breeding season (Table 3), although a new option for 2010 in ELS seeks to address this in England (see Discussion). The importance of AES seed resources can only have increased after the demise of set-aside, which was the principal means by which seed-rich stubbles were retained through the late winter period (S. Gillings, I.G. Henderson, A.J Morris & J.A. Vickery in review). Moreover, the evidence base underlying the choice of crops and crop mixtures for bird seed crops in AESs relies upon data on the relative value of different crop types during October to February (Stoate et al. 2003, 2004, Henderson et al. 2004, Parish & Sotherton 2004, 2008), a pattern that persists even in the ongoing monitoring of Higher-Level Stewardship (HLS) in England (R.H. Field, A.J. Morris, P.V. Grice & A. Cooke in review). Seed mixtures have rarely, if ever, been examined in late winter or early spring, so features of crop varieties, mixtures or management that determine food provision late in winter, such as seed retention capabilities, have not been considered in making recommendations for AES design. Even where studies have considered this period, it has been included in averages over the whole winter rather than being investigated in its own right (Hancock & Wilson 2003, Perkins et al. 2008). There is, therefore, reason to believe that significant modifications to the management rules for overwintered stubble and wild bird seed crop options in AESs, as well as the crop composition of the latter, will be necessary before the limiting factors preventing the population recovery of many farmland birds are addressed. Moreover, there is currently little evidence for the crop varieties or management that would ensure retention of seed over the late winter period and avoid earlier depletion by non-target species, suggesting that further research should be a priority.
Latest date to which seed-providing habitats must be retained
Overwinter stubbles followed by spring crops
Low-input arable crops preceding stubbles
Seed crops (wild bird covers)
Unharvested cereal crop conservation headlands
Fodder crops in pastoral areas
Rural Development Programme
Countryside Management Scheme
Tir Gofal/Tir Cynnal
Environmental Stewardship (Entry- and Higher-level)
The issue of food provision throughout the winter period illustrates the importance of AES options providing resources through the entirety of the period that they are intended to cover. In this instance, the root of the problem may be that ‘winter’ and ‘non-breeding’ have been used synonymously when, in fact, the former usually describes a much shorter period and, consequently, encompasses only a fraction of the period for which non-breeding resources are required.
Timing of resource supply vs. demand: breeding season habitat change
During the breeding season (at least until harvest), the principal cause of habitat change in farmland is crop growth, which is typically faster and more pronounced than that of naturally occurring plants because of selective breeding and fertilization. This means that crop areas that appear to be suitable habitat early in the season may cease to be so later in the year. For species for which multiple breeding attempts during a single season are critical, such as Skylark and Yellow Wagtail Motacilla flava, this can mean that winter crops only provide suitable nesting or feeding habitat early in the season, making it unlikely that a landscape dominated by such crops can support a stable population (Wilson et al. 1997, Gilroy et al. 2009, 2010). This has led to recognition of the value of spring cropping for such ground-nesting species. However, spring-sown fields in April, for example, are often just bare soil, so they need to be combined with autumn-sown cereals to produce a matrix that provides suitable levels of nesting cover throughout the breeding season. An AES approach to the same problem that has no implications for crop rotations but has demonstrable benefits for Skylark breeding success is to introduce undrilled (or subsequently sprayed-off) patches within crops (Morris et al. 2004). In either case, sufficient heterogeneity is required to provide habitats suitable for nesting during the full temporal extent of the potential breeding period. Similar vegetation heterogeneity is also likely to benefit Corn Buntings Emberiza calandra, another threatened species that breeds on or near the ground and for which late breeding attempts in cereal crops, as opposed to grassy, non-crop vegetation, are vulnerable to destruction at harvest (Gillings & Watts 1997, Brickle & Harper 2002).
Plants in uncropped areas also grow during the breeding season, of course, and the problem of accessibility to food or other resources in late summer vegetation could affect the success of AES management outside the cropped area. Boundary and field margin management form the most popular agri-environment options among farmers in ELS (Boatman et al. 2007) and could provide significant food and nesting resources for birds as a consequence of the combination of semi-natural vegetation and low agrochemical inputs. However, although richer, denser vegetation in field margins is associated with greater invertebrate abundance and diversity, this pattern does not necessarily lead to greater use by foraging birds because that very density restricts access to the food source (e.g. Vickery et al. 2001, 2002, Wilson et al. 2005, Douglas et al. 2009). Again, the solution probably lies in heterogeneity, in this case at the territory scale, such that individual territories contain a mosaic of short and tall vegetation, providing both food-abundant and food-accessible microhabitats. Douglas et al. (2009) found that the use of field margin vegetation by Yellowhammers decreased through the summer despite the invertebrate density being greater than that in crops, the alternative foraging location. However, experimentally cut patches within tall margin swards were used significantly more in late summer, reflecting the extra access to food resources provided. In this case, the key to gaining maximum value from land taken out of production for AES management as field margins is to revise the option prescription to take account of habitat characteristics throughout the summer. Late summer breeding attempts could be important for many species, often showing lower failure rates than earlier ones (e.g. Kyrkos 1997, G.M. Siriwardena & H.Q.P. Crick unpubl. data), but studies investigating habitat selection, including associations with patches of AES management, typically focus on the early part of the summer, when singing activity is greater. As with the winter, it is clear that management over the full temporal scale of the breeding season needs to be considered if AES option effectiveness is to be maximized.
Many current AESs are now maturing to a stage at which it might be expected that any biodiversity benefits would become apparent, but so far there is little evidence of broad-scale benefits (e.g. Davey et al. 2010). It is timely, therefore, to consider why such benefits might not have occurred while there is still time to address any problems with the design or implementation of the schemes. If schemes are failing, shortfalls in either the quality or the quantity of the new resources intended to be provided by AES options could be responsible. The quantity of AES options required to support any given species at a particular time of year is difficult to judge and there has been rather little research scaling up encouraging local responses (e.g. that found by Hinsley et al. 2010) or relating quantities of landscape-scale habitat management to population changes (but see Gillings et al. 2005). Overall uptake of ELS in England, for example, has been encouraging, with almost 70% of land now under some agri-environment agreement. However, the uptake of many in-field options has been disappointing, so the area actually under AES management is considerably lower and there has probably been little real change in cropping practices (Boatman et al. 2007).
In terms of delivering AES management of an appropriate quality, evidence from pilot studies and the research that identified the problems initially should allow the design of suitable prescriptions, while scheme management guidelines should ensure that these prescriptions are followed. However, most of the evidence base for option design comes from plot- or farm-scale experiments that tend not to have considered effects across whole landscapes; this matches the scale of current AES implementation but not necessarily that required to generate detectable biodiversity responses. There are also important instances where option delivery throughout the entire duration of resource requirement, such as seed provision through all of the non-breeding period, has not been considered in option design. This suggests that AES efficacy would now benefit from the re-design of certain options. In fact, it is an example of how management and policy would best be served by a flexible approach in which AES protocols can be changed as significant additions to the evidence base are made, rather than being rigid throughout the period of an individual AES agreement.
The results described here lead to several specific management recommendations, in the form of revisions to existing AES options. Greater support for stubble retention through a whole cropping year (i.e. overwintered stubble followed by an unploughed summer fallow) would be extremely valuable. Such an option is available in HLS, and has been added to ELS for 2010 (Natural England 2010a,b), but it is unknown how popular it will be with farmers and it has not been included in any other UK scheme to date or in the Campaign for the Farmed Environment, the industry-led programme designed to mitigate the negative effects of set-aside loss in England (http://www.cfeonline.org.uk), which includes no consideration of late winter food. Simply delaying the earliest ploughing date for wild bird seed crops is unlikely to be effective in the same way because of vegetation die-back and seed depletion, although sparsely vegetated and bare ground can produce emergent annual weeds and thus provide a limited additional seed resource. Instead, 2-year plots including annual and biennial crops to provide seed in years 1 and 2 would be ideal. Biennial plots are already allowed in some AESs (Natural England 2010a, Scottish Government 2008) but are not promoted above annual ones and do not incorporate specific late winter food provision. Annual crops alone could provide a resource in year 1, but the requirement for late winter seed provision means preserving the plot through to April. This is likely to prevent the establishment of another annual crop in such a plot in time for year 2, but incorporating a biennial crop, such as kale, into the initial mixture would mean that a plot has value in the second winter as well. Finding crop varieties that offer sufficient seed retention is likely to require research, possibly including crop-breeding trials, but certainly involving field trials to identify optimal mixes for a range of soil conditions and to ensure efficacy in real landscapes. Successful crop mixes (or sets of mixes) would need to cater for the dietary requirements of all species of interest (although highly palatable seed types, such as cereals, are edible to a wide range of species) and also, critically, to ensure seed supply throughout the winter. This means late winter seed retention, but also, ideally, ensuring that seed is not taken preferentially early in the winter or by non-target species. This could be a particular challenge with respect to Woodpigeon Columba palumbus depredation, and research is required as to how location, habitat context and crop structure might be used to reduce this problem.
Recommendations can also be made for improving AES options that aim to provide breeding season resources. The value of grass field margins to birds throughout the breeding season might be improved by employing partial cutting in mid-summer as an additional management tool for grass field margins, but protocols would need to be designed carefully to ensure that the benefits of uncut margins for other taxa are not compromised. In addition, the requirement for additional work from farmers might make this unpopular and unlikely to be taken up widely. Other approaches to the generation of sward heterogeneity might be more effective in practice, such as partial margin cultivation when the rest of the field is ploughed, perhaps in patches amounting to a fixed proportion of the area covered by the margin. In this case, revised management would have to avoid compromising the value of a margin for resource protection. However it is achieved, more heterogeneous vegetation in grass margins will increase their value to birds (Vickery et al. 2009) and other biodiversity (e.g. Carvell et al. 2007, Potts et al. 2009). Undrilled patches (or ‘Skylark plots’) in winter cereals have a strong evidence base at the field scale as a means of promoting in-field heterogeneity in late summer (Morris et al. 2004) and are easy to deliver, but have proved unpopular with farmers (Boatman et al. 2007). This would seem to be a clear example of an option whose value at the national scale could be increased dramatically by improved advice to farmers (e.g. J. Phillips, R. Winspear, S. Fisher & D.G. Noble in review). Other possibilities for increasing heterogeneity include, at the field scale, promoting crop diversity, especially of spring cropping in winter-crop-dominated areas and, within fields, controlling stubble height and areas of bare ground to meet the habitat requirements of different species as they occur during the winter (Whittingham et al. 2006).
In terms of spatial distributions of AES patches, the evidence for winter seed-providing habitats is that habitat patches would best be spread fairly regularly through the environment to be certain of catering for all birds in local populations; specifically, they should be spaced 1–2 km from one another. In practice, this would mean some habitat management on every farm, which fits well with the ‘broad and shallow’ ethos of ELS and Tir Cynnal, for example. Interestingly, it also supports the contention that a principal benefit of set-aside was that it provided habitat on every farm (S. Gillings, I.G. Henderson, A.J Morris & J.A. Vickery in review). It is important to note that these conclusions are derived from the average distances that birds tended to move in the studies reported here: far greater distances are not only possible but frequently recorded. Moreover, there are numerous reports of very large flocks of wintering granivorous passerines, some of which must have travelled large distances to find a food source in, for example, particularly rich seed crops. Birds clearly can, therefore, cover much larger distances to find food, so it is possible that larger separations of AES food sources than those suggested by the results here would be equally effective. Correspondingly, research on breeding habitat selection by Ortolan Buntings Emberiza hortulana has shown rapid movements within a season that are very large compared with daily home-ranges (Dale et al. 2006). Decisions made by individuals as to where to settle and directions in which to move, and therefore the distances travelled, can also be influenced by social factors (the presence of conspecifics: Betts et al. 2008) and behavioural barriers preventing the crossing of certain habitat features (Harris & Reed 2002). More generally, movement propensities are likely to vary with habitat/landscape structure, such that the distances animals tend to move will be landscape-specific, not just species-specific (Fahrig 2007). In addition, the costs birds incur by travelling further, whether energetic or due to increased predation risk in transit or in an unfamiliar area, are unknown and modelling suggests that search time, in particular, can have great effects on the selectivity of birds choosing habitats in which to settle (Stamps et al. 2005, Fahrig 2007). If spatial distributions are unfavourable, such costs could detract from any positive effects on survival derived from supplementing food resources.
Spatial considerations are also critical in the combination of different AES measures to provide resources in different seasons: addressing a winter population bottleneck will not lead to population recovery unless there is also sufficient breeding habitat, for example, although declines due to winter resource shortages seem likely, intuitively, to leave some ‘spare’ breeding season resources. However, complementarity of seasonal resources will not necessarily be provided most efficiently simply by ensuring that an AES agreement on a given farm includes options that aim to deliver all that a species needs. The quantitative balance of habitat requirements is likely to be complex and bird movements between seasons also need to be taken into account. Regional AES targeting (J. Phillips, R. Winspear, S. Fisher & D.G. Noble in review) provides part of the mechanism whereby ensuring resource complementarity at larger scales than that of a single farm could become possible, but the organization of agreements at the landscape level could be problematic under systems where agreements are holding-specific.
Nationally, the uptake of broad and shallow AESs, such as ELS, is very encouraging in terms of the area under agreement (Boatman et al. 2007), suggesting that there is real potential to benefit wider biodiversity. However, the constituent options need to be effective in terms of both design and spatial organization to maximize scheme value, as well as being present in sufficient quantity to meet the needs of target populations. Now that many AESs are sufficiently mature that their success in terms of habitat creation and biodiversity responses can begin to be assessed, it would be timely to re-evaluate all of the component options with respect to spatial organization (asking whether this is sufficient to meet all requirements of target species and efficient in terms of meeting needs for minimum cost) and temporal scale (whether the complete period of resource shortage at which the option is aimed is actually covered by the design employed). This would help to identify where schemes would benefit from re-design or re-organization as suggested for certain options here.
The work described here was funded by Defra under projects BD1616 and BD1628: many thanks to Richard Brand-Hardy for his support. Thanks also to Guy Anderson and to the many BTO staff who helped with fieldwork, project design and organization, notably Juliet Vickery, Neil Calbrade, Greg Conway, Diana de Palacio, Trevor Girling, Chas Holt, Markus Handschuh, Jeff Stenning and Aaron Boughtflower. Valuable discussions with John Bingham and Lydia Smith have helped to guide this work; Simon Gillings, Juliet Vickery, Andrew Hoodless, Jeremy Wilson, Paul Donald and an anonymous referee provided helpful comments on earlier drafts.