General enquiries on this form should be made to:Defra, Procurements and Contracts Division (Science R&D Team)Telephone No. 0207 238 5734E-mail: research.competitions@defra.gsi.gov.uk

SID 5 Research Project Final Report

SID 5 (Rev. 05/09) Page 1 of 22

NoteIn line with the Freedom of Information Act 2000, Defra aims to place the results of its completed research projects in the public domain wherever possible. The SID 5 (Research Project Final Report) is designed to capture the information on the results and outputs of Defra-funded research in a format that is easily publishable through the Defra website. A SID 5 must be completed for all projects.

This form is in Word format and the boxes may be expanded or reduced, as appropriate.

ACCESS TO INFORMATIONThe information collected on this form will be stored electronically and may be sent to any part of Defra, or to individual researchers or organisations outside Defra for the purposes of reviewing the project. Defra may also disclose the information to any outside organisation acting as an agent authorised by Defra to process final research reports on its behalf. Defra intends to publish this form on its website, unless there are strong reasons not to, which fully comply with exemptions under the Environmental Information Regulations or the Freedom of Information Act 2000.Defra may be required to release information, including personal data and commercial information, on request under the Environmental Information Regulations or the Freedom of Information Act 2000. However, Defra will not permit any unwarranted breach of confidentiality or act in contravention of its obligations under the Data Protection Act 1998. Defra or its appointed agents may use the name, address or other details on your form to contact you in connection with occasional customer research aimed at improving the processes through which Defra works with its contractors.

Project identification

1. Defra Project code BD1641

2. Project title

Assessing the impact of the loss of set-aside stubbles on the winter ecology, behaviour and distribution of farmland birds and their breeding population trends

3. Contractororganisation(s)

British Trust for Ornithology                         

54. Total Defra project costs £ 242,158(agreed fixed price)

5. Project: start date................ 03 November 2008

end date................. 31 March 2010

SID 5 (Rev. 05/09) Page 2 of 22

6. It is Defra’s intention to publish this form. Please confirm your agreement to do so...................................................................................YES NO (a) When preparing SID 5s contractors should bear in mind that Defra intends that they be made public. They

should be written in a clear and concise manner and represent a full account of the research project which someone not closely associated with the project can follow.Defra recognises that in a small minority of cases there may be information, such as intellectual property or commercially confidential data, used in or generated by the research project, which should not be disclosed. In these cases, such information should be detailed in a separate annex (not to be published) so that the SID 5 can be placed in the public domain. Where it is impossible to complete the Final Report without including references to any sensitive or confidential data, the information should be included and section (b) completed. NB: only in exceptional circumstances will Defra expect contractors to give a "No" answer.In all cases, reasons for withholding information must be fully in line with exemptions under the Environmental Information Regulations or the Freedom of Information Act 2000.

(b) If you have answered NO, please explain why the Final report should not be released into public domain

Executive Summary7. The executive summary must not exceed 2 sides in total of A4 and should be understandable to the

intelligent non-scientist. It should cover the main objectives, methods and findings of the research, together with any other significant events and options for new work.1. Previous research (including BD1639 and BD1640) identified the historical value of set-aside land as

habitat for farmland birds and demonstrated that the establishment of the zero-rate for set-aside in 2007 had been followed by significant reductions in areas of winter crop stubble across Britain. Bird survey work failed to find declines in bird numbers in the first winter after set-aside disappeared, however, and suggested that birds might be making increased use of field boundary habitats. The present project was then established to investigate whether the patterns of land-use and bird population sizes remained the same during the second winter post-set-aside and whether breeding bird abundance nationally had been affected after the two winters. In addition, novel approaches were used to investigate bird behaviour in and what influences bird use of “hotspot” areas away from

2. Areas of stubble in winter were compared to the set-aside era via re-surveys of 596 1-km squares nationally previously covered under the JNCC-funded Winter Farmland Bird Survey (WFBS) in 1999-2002, focusing on February each winter, because set-aside was probably most valuable later in the winter. A second analysis of stubble area was conducted using resurveys in winter 2008-09 of 19 2×2km tetrads in East Anglia previously covered in 2004-07 under BD1628. These data were more limited geographically, but included data from later in the winter (March) and were more closely matched temporally to the post-set-aside context.

3. Stubble areas in all but the very end of winter 2008-09 (from the East Anglian data) had risen again, on average, back to the levels they had been at during the set-aside era, although there were considerable survey-square-specific changes and some polarization of stubble areas towards “a lot” or “very little”. A wet autumn restricting ploughing and a fall in wheat prices from the very high historic peak in 2007-08 influenced these patterns.

4. Possible effects of set-aside loss on national population trends were assessed using five analyses of BTO/JNCC/RSPB Breeding Bird Survey (BBS) data, supplemented by additional survey work in 2009: (i) a repeat of the analyses of Gillings et al. (2005) to investigate population trends from 2007 to 2009 in BBS squares with zero, <10 and >10ha of stubble during the WFBS period; (ii) a repeat of analyses from BD1639 to investigate population changes between the WFBS period and 2009 with respect to the change in the area of stubble over the same period; (iii) a novel analysis to look for evidence of changes in BBS population trajectories after 2007, the year that set-aside disappeared, comparing the fits of models allowing abundance to vary as one linear trend before a “change point” and one after it, with change points each year from 1995 to 2008; (iv) a repeat of (iii) for arable and pastoral trends, separately, for skylark and yellowhammer; (v) analyses of changes in skylark and yellowhammer BBS counts from 2006 to 2008 relative to changes in the areas of uncropped land around each lowland farmland BBS square, as recorded by Defra statistics. (i), (ii) and (v) considered 22 species with different a priori likelihoods of being affected by set-aside loss: 11 “stubble species” (sensu Gillings et

SID 5 (Rev. 05/09) Page 3 of 22

al. 2005) likely to be affected by set-aside loss, and 11 “control” species. Skylark and yellowhammer were the species most likely to be affected by set-aside loss for which sample sizes were large.

5. There was little evidence of a significant effect of stubble area, as measured in 1999-2003, on population change from 2007 to 2008 or from 2007 to 2009: results were either non-significant, involved non-stubble species that are unlikely to have been affected by set-aside in reality, or showed a biologically unlikely pattern.

6. Results for effects of changes in stubble area from 1999-2003 to 2008 on bird changes from 2007 to 2008 were inconclusive, but a larger sample for changes up to 2009 identified positive associations between bird population change and stubble area change for six species, of which House Sparrow, Yellowhammer and Stock Dove could reflect a real effect of stubble.

7. The change-point analyses revealed 2007 as a poor single break year for Skylark and Yellowhammer and there was no general pattern for “stubble” species to decline further after 2007: downturns for Greenfinch and, perhaps, chaffinch, probably reflect disease effects. Considering arable and pastoral regions separately made little difference for Skylark but 2007 was well-supported as a change point for Yellowhammer, providing some evidence consistent with an effect of set-aside loss.

8. There were no significant associations between change in uncropped land area from Defra statistics and population change for either Skylark or Yellowhammer.

9. The 19 BD1628 tetrads in East Anglia were re-surveyed for birds in winter 2008-09 following the approach taken in BD1639 to reveal total population sizes and the extent of use of seed-rich versus other habitat types. The results were similar to those of BD1639: counts were higher than in 2004-07, probably because of differences in field methods, but there was no evidence for declines in abundance post-set-aside.

10. Surveys of the same areas in the breeding season, using the same methods and observers as BD1628, revealed a range of population trends across 42 species, but these included nine showing small but significant changes consistent with negative effects of set-aside, including chaffinch, greenfinch, yellowhammer, house sparrow, skylark and turtle dove. There was, however, no strong evidence that areas of stubble affected trends of species that use stubble fields in winter, so these patterns could have other causes. Nevertheless, they represent the first landscape-scale evidence consistent the large-scale negative effects of set-aside loss on biodiversity.

11. The possible importance of foraging “hotspots” away from accepted seed-rich habitats was investigated in the East Anglian tetrads using a novel field observation approach, aiming (i) to quantify the extent to which boundary habitats are used and to assess how this differs between seed-rich and other fields, (ii) to assess what habitat features characterize hotspots of bird-use in field boundaries, (iii) to assess what birds are using hotspot areas for and (iv) to measure the stability of such hotspots over time. Hotspots were identified during survey visits in December-January, February-March and early April 2008-09, with follow-up visits to record bird presence and behaviour after periods of one day to four months.

12. Large proportions of granivorous bird populations were found in non-“seed-rich” hotspots, but stubbles and wild bird covers were still selected actively. Bird counts were more biased towards records in boundary vegetation in non-seed-rich field types than in stubbles and covers for chaffinch, yellowhammer and grey partridge, suggesting that the resources the birds find in cropped fields are more concentrated in the boundary vegetation than in the cropped area.

13. There was clear pattern for birds to be found in a combination of small and large groups. There were at least two significant effects of habitat on hotspot definition for each more common species, but there was no evidence for consistent patterns across species.

14. Chaffinch, yellowhammer and linnet all spent a greater proportion of their time feeding in hotspot areas that were not adjacent to recognized seed-rich habitat than they did in locations next to such habitat. This probably shows that birds need to forage for longer on semi-natural food sources than they do in cropped habitats, probably because seed densities are higher in the latter, but it demonstrates that birds use hotspots for foraging. Habitat within hotspots affected foraging behaviour, but again with few generalities across species. The most general and useful results showed a positive influence of the percentage bare ground and a negative one of herbaceous vegetation height on time spent foraging.

15. Comparisons of survey visits revealed very little evidence of hotspot persistence. For example, only 6.9% of chaffinch hotspots identified in December-January were also used in February-March. Significant habitat influences on persistence were again highly variable between species.

16. This project has not revealed conclusive evidence for national effects of set-aside loss on habitat quality or availability for farmland birds, but three separate new lines of evidence have revealed patterns of population trend consistent with such effects on species like yellowhammer. Larger, more conspicuous effects might not have occurred for several reasons, including time lags to detectability thresholds caused by ecological/demographic factors and methodological limitations of surveys, weather effects combined with volatility in agricultural economics affecting cropping patterns and confounding effects of agri-environment management. In addition, the arguments made previously as to the value of set-aside for farmland birds remain valid.

17. Patterns of hotspot use reveal that they could be represent important foraging habitats, but are

SID 5 (Rev. 05/09) Page 4 of 22

unpredictable in location and ephemeral. Birds seem to use a wide range of habitats rather opportunistically, with weed, crop and livestock-associated seed sources being exploited where they are found, and in flock sizes broadly appropriate for the density of the food resource. This makes a specific prescription for a form of agri-environment option to promote “hotspot habitats” difficult to define spatially or in terms of specific management rules. However, the principal recommendation for management must be to manage habitats peripheral to farmland for seed availability: flexible and opportunistic birds may use seed resource in many potential contexts. Approaches might best focus on the promotion of non-injurious weeds and their seeds and sacrificial seed crops in areas that are peripheral to the main economic activities of the farm, with management of crop or semi-natural vegetation to promote access to bare ground for foraging. This could fit well into guidelines for habitat management under the Campaign for the Farmed Environment or Environmental Stewardship.

Project Report to Defra8. As a guide this report should be no longer than 20 sides of A4. This report is to provide Defra with

details of the outputs of the research project for internal purposes; to meet the terms of the contract; and to allow Defra to publish details of the outputs to meet Environmental Information Regulation or Freedom of Information obligations. This short report to Defra does not preclude contractors from also seeking to publish a full, formal scientific report/paper in an appropriate scientific or other journal/publication. Indeed, Defra actively encourages such publications as part of the contract terms. The report to Defra should include: the scientific objectives as set out in the contract; the extent to which the objectives set out in the contract have been met; details of methods used and the results obtained, including statistical analysis (if appropriate); a discussion of the results and their reliability; the main implications of the findings; possible future work; and any action resulting from the research (e.g. IP, Knowledge Transfer).

Introduction

Although it was not designed to meet conservation aims, set-aside provided considerable benefits for biodiversity, notably farmland birds (reviewed in BD1640: Vickery et al. 2008b). These benefits varied between species, including breeding season nesting habitat and food resources and, especially, winter foraging habitat for the open field bird species that have suffered most from agricultural intensification. The disappearance of set-aside has, therefore, widely been predicted to have negative consequences for the species concerned, leading to policy formulation to mitigate these consequences, which culminated in the launch of the Campaign for the Farmed Environment as a set of measures designed to recapture the environmental benefits of set-aside without the legislative burden of a formal scheme.

To date, however, little evidence has accrued as to the real environmental costs of set-aside loss in practice. Previous research by the BTO (BD1639: Vickery et al. 2008a) identified that areas of crop stubble (the key foraging habitat for many granivorous farmland birds) fell considerably in the first winter post-set-aside (2007-08), both nationally (compared to 1999-2002) and in East Anglia (compared to 2004-2007). The difference was particularly large during the late winter period that is probably when habitat is most scarce and the resources provided by set-aside most valuable (Siriwardena et al. 2008). This was the pattern expected given the lack of a requirement to keep land uncropped as set-aside, combined with record high wheat prices (caused by global market forces) giving farmers a strong incentive to crop as much land as possible. It also supported the expectation that removing compulsory set-aside would have strong negative effects on the farmland environment.

An initial investigation of changes in wintering bird populations, however, showed no evidence for declines caused by the change in habitat during 2007-08 (BD1639: Vickery et al. 2008a). Constraints on study design imposed by the reactive nature of this research (it made use of reference data collected for different purposes, under BD1628: Siriwardena et al. 2007) meant that these results were not definitive. In addition, as well as it being possible that set-aside loss actually had no effect on the birds investigated, responses could be subject to time lags that mean that they are difficult to detect in (less than) a single year. It could also be that effects of this habitat change are only significant biologically in years when conditions are harsh, and 2007-08 was a mild winter. Another key result of BD1639 was, however, that farmland granivores were frequently found in hedgerows and other habitat patches that were not adjacent to stubble fields, wild bird seed/game crops or other accepted

SID 5 (Rev. 05/09) Page 5 of 22

“seed-rich habitats”. Although no earlier data were available against which to compare this pattern, it suggests that these species could be finding significant food resources in habitats not previously considered to provide them and that this use of alternative habitats could have increased after set-aside disappeared. This is important both because it could indicate that other habitats are, indeed, available that make set-aside unnecessary (although note that use of such habitats does not necessarily mean that they provide an adequate quantity or quality of food resources or that they are equivalent in terms of predation risk) and because the promotion and/or management of key features of the alternative habitats could provide an effective replacement for the benefits of set-aside that would be acceptable to farmers.

This project was commissioned by Defra to improve the evidence base on effects of set-aside loss by building on the previous field research on the issue conducted by the BTO to consider (a) patterns of winter habitat in the second winter post-set-aside, (b) patterns of bird abundance in that winter, (c) whether bird breeding populations have been affected by set-aside loss and (d) the nature and importance for bird foraging in habitats not identifiable as seed-rich a priori. The results then inform recommendations as to policy for the post-set-aside context. Fieldwork and analyses were structured around the points at which population limitation due to the effects of set-aside loss were likely to occur (Figure 1) and some analyses were conducted twice, once to provide “early” results using data up to 2008 and once using full data sets including 2009. Specifically, the project addressed the following Objectives, around which the rest of this report is structured:

1) to determine whether there is evidence for a breeding-population-level response for farmland bird species, at the national scale;2) to determine whether there is evidence for breeding- and winter-population-level responses for a range of farmland bird species, at the local (tetrad) scale in East Anglia;3) to assess the nature of secondary foraging habitats, particularly hedgerow and other field boundary vegetation, and extent of use by birds;4) to strengthen the evidence base relating to the need for action to mitigate the loss of set-aside and provide recommendations about the possible habitat management measures that may be of value in this context.

Figure 1. Schematic diagram showing the timing of various surveys in relation to seasons, the possible loss of set-aside habitat and periods of potential population limitation for resident farmland birds.

Objective 1: national-scale breeding population patterns

The first part of this objective (“Approach One” in the SID3) was to repeat and to extend the analyses under BD1639 using a further and larger resurvey of late winter habitat in long-term BTO study areas in February 2009 to give greater analytical power in assessing whether winter stubble availability has been affected by the loss of set-aside. The second part (“Approach Two”) was to measure bird population trends and to relate them to the timing and location of changes in uncropped land (i.e. winter stubble) availability by repeating an analysis by Gillings et al. (2005) that investigated winter habitat effects on breeding populations of key species. Set-aside was present on every farm and rotational set-aside, especially, provided high quality habitat in the breeding season and/or winter for a range of species. Its loss could therefore have affected birds throughout the wider countryside and at large spatial scales. Set-aside was far from being the only influence on bird abundance, however, and changes in bird numbers could well not be apparent either because other environmental effects have over-ridden them or because bird responses are only detectable after a certain time-lag. The latter could occur either for biological reasons, such as effects on breeding success taking a few generations to work through into the abundance of adults, or as a consequence of survey methods: changes might occur at a low annual percentage rate, but need to reach a certain, higher, threshold to be detectable statistically. Given data for more years after

SID 5 (Rev. 05/09) Page 6 of 22

2007 2008 2009

WinterSpring

Stubble / fallow

Periods of potential

population limitation

Breeding Bird Surveys

Winter Bird Surveys

Winter Habitat Surveys

No. potential population limiting periods 1 32-3*0-1*

the setting of the zero-rate for set-aside, it is more likely, therefore, that any effects it has on national bird populations will be detectable, but it is still possible that effects might be undetectable. The analyses here concentrated initially on the trends from 1994 to 2008 for rapid reporting (see 2009 SID4 for this project), but we were subsequently able to analyse data up to 2009 to consider responses two winters after the loss of set-aside.

Full details of the Methods and Results under this objective are provided in Appendix 1.

Methods

All analyses used 1-km squares surveyed under the BTO/RSPB/JNCC Breeding Bird Survey (BBS), and in particular, the 601 1-km squares that were surveyed by both BBS and the BTO/JNCC Winter Farmland Bird Survey (WFBS; Gillings et al. 2008).

Winter habitat surveysSurveys conducted during the winters 1999/2000, 2000/01 and 2002/03 of all 601 squares under WFBS were repeated between 15th January and 5th March 2009 (initial plans were to match the WFBS methods precisely, but unusually heavy snowfall in early February delayed the work – and farmers’ field operations, so serious bias in terms of habitat sampling is unlikely). In each square, fieldworkers aimed to survey the same fields as were surveyed by volunteers during the early 2000s, using a simplified habitat recording scheme to estimate key habitat areas.

Upon completion, the area of cereal stubble and the total surveyed area in each square in 2009 were determined, using a GIS. These data were compared to WFBS data from (the mean across) February visits in the early 2000s. For 39 of 601 squares, data had to be taken from January WFBS visits because no February data were available. Habitat areas from the early 2000s and 2009 were scaled to standardize to the area of farmland actually present in each square to prevent geographic biases and to give estimates of total habitat availability in 1-km squares. Areas were compared using Kruskal-Wallis and Sign Tests performed in SAS (version 9.2).

Breeding Bird Survey population trendsTo maximise the power to measure population changes in relation to winter habitat availability we required population trend data for as many as possible of the 601 BBS squares. The aim of BBS is to retain the same volunteer covering the same square but inevitably there is some turnover or loss of volunteers. We therefore supplemented the volunteer dataset with as many of the currently uncovered 1km squares in the sample of 601 as funding allowed, focusing on those that provided the best temporal coverage for assessing effects of set-aside loss. These data were used in two analyses:

Analysis 1. 1999-2003 stubble area and change in bird abundance 2007 to 2008, and 2007 to 2009. This investigated possible effects of set-aside loss on population trends by comparing recent changes among the squares identified by Gillings et al. (2005) as having relatively low, intermediate and high areas of stubble between 1999 and 2003. The squares that were also surveyed in 2007, 2008 and 2009 showed changes in breeding-season counts of different species concurrent with the loss of set-aside, which were analysed with respect to the relative amounts of stubble in the WFBS period. For those squares that were surveyed in both 2007 and 2008, models were conducted to compare counts between 2007 and 2008, in squares that had no stubble (none), less than 10 ha of stubble (little) and more than 10 ha of stubble (lots) between 1999 and 2003. These models were based on ratios of counts between 2008 and 2007 for the species in question, using the formula: ratio change = (2008 count + 1) / ( 2007 count + 1). Adding one to each count allows increases from zero to one or more recorded birds to be considered, providing a measure of population growth where no birds were seen in 2007 but where birds were subsequently recorded in 2008. This was repeated for the comparison of 2007 and 2009 counts. All models were fitted in the GENMOD procedure of SAS 9.2 (SAS Institute, Inc., 2007) using an identity link function and normal error distribution, and treating the “stubble amount” variable as a three-level factor.

Analysis 2. Changes in stubble area and change in bird abundance 2007 to 2008, and 2007 to 2009. These population analyses considered change in bird count (as above) in relation to change in stubble area between 1999-2003 and 2008 (a subset of 129 of the squares considered by Gillings et al. 2005 and reported in Defra project BD1639), and between 1999-2003 and 2009 (results herein). Results from 2008 were compared with the previous results from the same squares between 1999 and 2003 closest to the date of the re-survey and showed whether, considering the stubble categorization used by Gillings et al. (2005) of no stubble, little stubble (<10ha) or lots of stubble (>10ha) in the 1km square, winter stubble area had increased, decreased or showed no change. These changes were converted to an interval scale consisting of values from -2 for squares which had decreased from lots to no stubble, -1 for squares which had decreased from lots to little or little to none, 0 for squares showing no change in stubble amount, 1 for squares increasing from none to little or little to lots of stubble and 2 for squares increasing from none to lots of stubble. Ratio changes in bird abundance between 2007 and 2008 were calculated as above and analysed with respect to the stubble change variable, using a similar model

SID 5 (Rev. 05/09) Page 7 of 22

structure. The results indicated whether population growth rate varied between squares that differed in the change in the amount of stubble between 1999-2003 and 2008. The same approach was adopted for comparisons between 1999-2003 and 2009 using as many as possible of the 601 squares that had both habitat data and BBS visits for 2009.

The national BBS sample was used to investigate possible relationships between national-scale bird population changes and (a) the timing of the loss of set-aside (analyses (3) and (4) below) and (b) the area of uncropped land in around each BBS square, as recorded by Defra statistics (analysis (5)). Initial analyses concentrated on the period 1994 to 2008, with subsequent analyses extending to 2009 when those data became available. Analyses considered 22 species with different a priori likelihoods of being affected by set-aside loss: 11 designated as “stubble species” by Gillings et al. (2005) (most likely to be affected by set-aside loss), and 11 species more associated with hedges and other habitats and thus not expected to respond directly to set-aside loss.

Analysis 3. Break year analysis of trends. We developed a novel modelling approach to analyse temporal changes in national BBS abundance. Species’ population trends were modelled using log-linear negative binomial models (as are standard for over-dispersed count data), as simple linear functions of year modified to allow the population trend slope to change at a single point during the time series. This approach is known as “piece-wise” or “broken stick” regression (Toms & Lesperance 2003). BBS data are highly heterogeneous and are potentially influenced by many factors in addition to the disappearance of set-aside, and there was only one year post-set-aside from which to characterize population trends in the new environmental context. This means that the statistical power of the analyses to detect break points around 2007 was low. We therefore took a novel approach of identifying which of the possible break points in the time series (1995 - 2007) was best supported by the data, and whether any of these was better supported than a model with no break points at all. This process revealed whether there was any evidence supporting the existence of a discontinuity in the population trend and when a discontinuity is most likely to have occurred. If 2007 were identified as the “best” break point year for the species likely to be affected by the habitats provided by set-aside, this would suggest that the loss of set-aside could have been the largest influence on these species’ population trends between 1994 and 2007. Note, however, that such a result would be consistent with an effect of set-aside, but would not prove that it had occurred. Further, an effect of set-aside could have occurred even if 2007 were not identified as the best supported potential break year, because another habitat or environmental change at a different time could have had a larger effect, so providing a better single break point across the entire time series. These analyses were repeated for the “best” break year and breaks in 2007 and 2008 when 2009 BBS data were available. Model fits were compared using the quasi-likelihood information criterion (QIC; Pan 2001, Hardin and Hilbe 2003).

Analysis 4. Arable and pastoral trends for Skylark and YellowhammerAreas of habitats like over-wintered stubble can have different influences on abundance in arable and pastoral landscape contexts (Robinson et al. 2001), so separate arable and pastoral analyses were also run for the most widespread species most likely to be affected by areas of stubble: Yellowhammer and Skylark.

Analysis 5. National changes in uncropped land and birds, 2006-08Data collected by Defra to inform the Single Payment Scheme (SPS) and the Integrated Administration and Control System (IACS) provide an alternative source of spatially referenced information on areas of uncropped land. All farms that claim a payment under the SPS submit data that are used to compile this dataset, so it provides a comprehensive summary of the uncropped land present in the landscape. In turn, this means that areas of uncropped land can be summarized at a range of spatial scales, rather than just the 1km square scale. The results of Defra project BD1628 suggest that Yellowhammer and Chaffinch often move more than 1km between winter and breeding locations, so a 3×3km square centred on a focal BBS square provides an appropriate spatial scale at which to consider the influence of winter habitat on breeding bird numbers in the square. Summaries of the SPS/IACS uncropped land data for 3×3km squares centred on each lowland farmland BBS square in Britain for 2006, 2007 and 2008 were supplied by Defra (S.D. Langton, pers. comm.). Although set-aside disappeared after 2007, this final year featured an increased proportion of set-aside under industrial (non-food) crops, thus presumably reducing environmental benefits, so 2006 data could be a valuable addition. BBS data from 2006-08 for Skylark and Yellowhammer were analysed using a “change-change” approach, relating proportional changes in bird numbers in each 1km square to proportional changes in the area of uncropped land, using the GENMOD procedure of SAS. Changes in uncropped land were considered from each of 2006 and 2007 to 2008, considering all uncropped land and rotational uncropped land only and both across all BBS squares and in squares more than 50% covered by farmland.

Results and Discussion

Coverage and sample sizes

SID 5 (Rev. 05/09) Page 8 of 22

Of the original 601 squares used in the analyses by Gillings et al. (2005), 129 and 596 were surveyed for habitat in February 2008 and 2009 respectively. For the national analyses, 3208 squares were used in the final analyses of BBS trends up to 2008 and 2690 for trends up to 2009. Change analyses were constrained by squares having to have been surveyed in two years: 372 squares for the comparison of bird numbers in 2007 and 2008 with respect to stubble availability and 89 with respect to stubble change. For 2007-2009 the stubble change sample was 367 squares.

Stubble availability in 2009 and change since 1999-2003Slightly more than half of all surveyed squares had zero cereal stubble in 2009 and we estimate that 27% of squares contained more than 10ha (Table 1). Of the 596 squares, 178 (30%) had zero stubble in both the WFBS period and in 2009. In 204 squares (34%), the area of stubble decreased between WFBS and 2009, and in 214 squares (36%) the area increased. Stubble availability differed markedly among farm landscape types and countries in both periods, but percentages of squares showing increases and decreases in stubble were similar. Across all squares, there was no overall change in the median area of stubble per square (combining all regions: Kruskal-Wallis 21 = 0.06, P > 0.8); there were changes at the square level, but both numbers showing increases and decreases (Table 2) and the magnitudes of the changes in each direction (Figure 2) tended to cancel one another out.

Table 1. Percentage of surveyed squares falling in each category of cereal stubble availability. Frequencies differed significantly among WFBS, 2008 and 2009 groups (24 = 67.6, P < 0.01). The column 2009* gives figures from 2009 surveys for the subset of squares surveyed in 2008.

Cereal stubble availability

Percentage of squaresWFBS (n = 601)

2008 (n = 129)

2009 (n = 596)

2009* (n = 129)

0ha 41% 72% 53% 49%10ha or less 35% 17% 20% 28%More than 10ha 24% 11% 27% 23%

Table 2. Frequencies (and percentage in brackets) of squares falling in each category of cereal stubble change in different farming regions and countries. The sign test determines significant deviation from an equal number of increases and decreases.

Stubble change Farming region Country TotalArable Mixed Pastoral England Scotland Wales

Decrease 71 (40%) 108 (38%) 25 (19%) 177 (35%) 21 (46%) 6 (14%) 204 (34%)Increase 74 (42%) 107 (38%) 33 (25%) 189 (37%) 17 (37%) 8 (19%) 214 (36%)Zero 32 (18%) 70 (25%) 76 (57%) 142 (28%) 8 (17%) 28 (67%) 178 (30%)Sign Test M = 1.5,

P = 0.87M = -1.0. P = 0.95

M = 4.0, P = 0.36

M = 5.5, P = 0.60

M = -2.0, P = 0.63

M = 1.0, P = 0.79

M = 4.5, P = 0.70

These patterns of change in non-cropped land were less consistent than those prior to February 2008 and show an increase back to pre-2008 levels. The Defra Farm Business Survey predicted a 10% increase in non-cropped land area from 2008 to 2009 but acknowledged that there was a wide variation in response from farmers, which is consistent with our finding of localised increases and decreases. Why has set-aside loss not led to a clear reduction in stubble? Defra June Statistics indicate that the set-aside area was almost completely replaced by cereals in 2008, so the following winter could have seen an increase in stubble following these crops, depending on rotations. Further, the increase in cereal areas in 2008 was driven by high wheat prices, which had fallen considerably by the following year, so farmers had less incentive to crop all of their previously uncropped land. It is also likely that the snow cover in late winter 2009 delayed or prevented some farming activities, producing a land-use pattern unrepresentative of what the long-term post-set-aside pattern will be.

Figure 2. Frequency distribution of decreases and increases in stubble area (ha) between WFBS and the 2009 survey. The zero bar is divided into the four squares where stubble was present and constant (shaded part) and the 178 squares where stubble was absent in both periods (white part).

SID 5 (Rev. 05/09) Page 9 of 22

Table 3. Tests of change in bird abundance (2007-2009) in relation to change in the area of stubble (2007-2009). The table presents parameter estimates and Type1 p-values for models with three different dependent variables. Significant P values and species where one ore more tests were significant are highlighted in bold.

Species NameParameter Estimates Type1 P-values

Ratio Count+1 Ratio

Count Difference Ratio Count+1

RatioCount Difference

Blackbird 0.136 0.111 0.417 0.015 0.006 0.187Blue Tit 0.154 0.120 0.441 0.020 0.011 0.110Chaffinch 0.030 0.017 0.377 0.366 0.611 0.272Corn Bunting 0.090 0.331 0.115 0.328 0.333 0.803Dunnock 0.113 0.089 0.025 0.101 0.148 0.877Goldfinch -0.049 0.091 -0.072 0.619 0.454 0.774Great Tit 0.050 0.093 0.356 0.434 0.108 0.108Greenfinch 0.116 0.052 -0.119 0.110 0.384 0.650Grey Partridge -0.007 -0.112 -0.075 0.942 0.388 0.746House Sparrow 0.173 0.289 -0.123 0.315 0.043 0.847Linnet -0.084 0.155 -0.065 0.541 0.405 0.873Mistle Thrush -0.111 -0.100 -0.109 0.136 0.153 0.424Pied Wagtail -0.017 0.003 0.040 0.781 0.950 0.707Reed Bunting 0.060 0.120 -0.141 0.713 0.373 0.632Robin 0.169 0.118 0.438 0.008 0.034 0.041Skylark 0.038 0.088 0.099 0.528 0.121 0.705Song Thrush -0.065 -0.046 -0.240 0.267 0.355 0.079Starling 0.129 0.207 0.815 0.669 0.283 0.546Stock Dove 0.382 0.078 0.606 <0.001 0.545 0.010Tree Sparrow 0.081 0.138 0.266 0.749 0.521 0.634Wren 0.020 0.012 0.039 0.763 0.809 0.861Yellowhammer 0.022 0.160 0.309 0.716 0.030 0.135

Breeding bird population changes1. 1999-2003 stubble area and change in bird abundance 2007 to 2008, and 2007 to 2009. There was little evidence of a significant effect of stubble area on population change in 2007-2008 or 2007-2009: results were either non-significant, involved non-stubble species that are unlikely to have been affected by set-aside in reality, or showed a biologically unlikely pattern (Reed Bunting). The exception was Skylark for 2007 to 2008, in which trends were progressively more positive as stubble areas increased, but this result should be treated with caution because it was only apparent for a ratio model which excluded squares with zero counts in 2007, providing a smaller sample size and probably biasing the results

2. Changes in stubble area and change in bird abundance 2007 to 2008, and 2007 to 2009. Only two species showed a significant association between bird population change and stubble area change with the limited sample from 2007 to 2008: both Great Tit and Tree Sparrow apparently increased in abundance where stubble

SID 5 (Rev. 05/09) Page 10 of 22

decreased in area, but direct effects of stubble on Great Tit are unlikely and the sample size for Tree Sparrow was too small (<30) to give reliable results. For 2007-2009, however, the larger sample identified positive associations between bird population change and stubble area change for six species, of which House Sparrow, Yellowhammer and Stock Dove could reflect a true stubble effect (Table 3).

3. Break year analysis of trends. For all species tested, one or more models including a discontinuity in the temporal trend were better supported by the data for 1994-2008 than a simple linear trend (Table 4). Models with 2007 as the break year and downturns at this time were supported best for Blue Tit and Chaffinch, potentially indicating an effect of set-aside for Chaffinch (Blue Tit is unlikely to have been affected by in-field habitat change; Goldfinch and Corn Bunting showed upturns in 2007). In addition, a 2007 break-year model was among the best five models for ten other species, five of which could plausibly have been affected by set-aside loss: Dunnock, Greenfinch, Linnet, Reed Bunting and Tree Sparrow (Table 4). Note that 2007 was revealed as a poor single break year for Skylark and Yellowhammer, the two species most likely to have been affected by set-aside loss and for which such an effect is most likely to be detectable. Overall, there was no real pattern for “stubble” species (sensu Gillings et al. 2005) to decline further after 2007: the downturns around this time for Greenfinch probably has a different explanation, which could also affect Chaffinch (disease: Robinson et al. in press), while the Goldfinch, Linnet and Corn Bunting patterns all featured upturns after 2007.

The repeat analysis performed using data for the period 1994-2009 found that the best break-year was the same as that for the 1994-2008 analysis for 16 species and linear models were now best for Chaffinch and Starling (Table 4). 2007 was now the best break year for Greenfinch and Great Tit), while 2008 was best for Blackbird and Wren, all of which turned down. The downward trend for Greenfinch is almost certainly a continuation of the decline due to disease, while the lack of reinforcement of the Chaffinch downturn with an extra year of data argues against an effect of set-aside. The other species’ patterns are unlikely to be consequences of set-aside loss.

Table 4. Summary of model results identifying years when species’ population trends are most likely to have featured a discontinuity. For the trend 1994-2009 analysis species are highlighted where the best model is different to that identified from the 1994-2008 models.

Trend 1994-2008 Trend 1994-2009Species “Best”

break yearRank of 2007 break model (out of 12, 1 = highest)

Simple linear model better?

Best model

Blackbird 1997 5 No 2008Blue Tit 2007 1 No 2007Chaffinch 2007 1 No LinearCorn Bunting 2007 1 No 2007Dunnock 2006 2 No 2006Goldfinch 2007 1 No 2007Great Tit 2002 3 No 2007Greenfinch 1997 3 No 2007Grey Partridge 1996 4 No 1996House Sparrow 1997 12 No 1997Linnet 1995 5 No 1995Mistle Thrush 2002 11 No 2002Pied Wagtail 2003 12 No 2003Reed Bunting 1998 5 No 1998Robin 1997 3 No 1997Skylark 1999 12 No 1999Song Thrush 2003 8 No 2003Starling 2004 9 No LinearStock Dove 2003 10 No 2003Tree Sparrow 1996 4 No 1996Wren 1996 5 No 2008Yellowhammer 1998 11 No 19984. Arable and pastoral trends for Skylark and Yellowhammer. Considering arable and pastoral regions separately made little difference to the 1994-2008 patterns revealed for Skylark but, for Yellowhammer, the best break point for both habitat types was towards the end of the time series and 2007 was, respectively, the second and third best-supported break-point. This provides some evidence consistent with an effect of set-aside loss, although it is odd that patterns were detectable in arable and pastoral areas individually, but not when they were combined.

SID 5 (Rev. 05/09) Page 11 of 22

5. National changes in uncropped land and birds, 2006-08. There were no significant associations between change in uncropped land area and population change for either Skylark or Yellowhammer. Smaller P-values (closer to statistical significance) were associated with the results for 2006-2008 and considering rotational uncropped land only, i.e. with the largest changes in the area of the most valuable uncropped habitat for these species. These results suggest a positive relationship between uncropped land and Yellowhammer abundance (i.e. loss of uncropped land leading to declines), but a negative one for Skylark, although it must be stressed that the results do not support such patterns at all strongly because they fell well short of statistical significance.

Objective 2: breeding and winter population changes in East Anglia

The rationale behind this part of the project was to use breeding season and winter field survey data collected for 20 22km tetrads in East Anglia in 2004-2007, under BD1628, as a baseline against which to compare population levels post-set-aside. For reference and interpretation, this also meant recording areas of uncropped land in these areas. In winter, therefore, the work towards this objective constituted a repeat of BD1639, although here we conducted an additional round of “winter” fieldwork in early April in order to assess habitats and birds during the period when set-aside is likely to have been most valuable. The work towards this objective is described in detail in Appendix 2.

Winter habitat and birds

MethodsOne of the 20 BD1628 tetrads could not be covered this time due to access restrictions imposed by the principal landowner, but the tetrad concerned was atypical, being dominated by a farm in arable reversion, so the overall sample was not particularly adversely affected by its omission. Survey methods followed those for BD1639, i.e. each tetrad was visited once during December-January and once during February-March by a professional fieldworker. On each visit (typically spread over 3-4 days), every field in the tetrad was visited, the field contents noted and all birds seen and heard recorded. Birds were surveyed by walking around all field boundaries and by conducting whole-area searches of stubble fields and other fields likely to hold large numbers of birds that would not be visible from field edges (e.g. weedy “set-aside” fields and wild bird/game cover strips). “Hotspots” featuring groups of finches, buntings or sparrows were noted on maps and re-visited to assess foraging behaviour (see Objective 3). During late March and the first two weeks of April, the tetrads that had been surveyed earliest in the February-March period were visited once again for a further complete habitat and bird survey (time did not permit such “late” visits to all 19 tetrads). This meant that all tetrads were visited once after 15th February, thus considering the critical late winter period (Siriwardena et al. 2008).

Field areas were calculated using a GIS and areas under different land-uses, including stubble types, wild bird cover and crops, were compared with those from BD1628, matching data by month. Bird abundance was estimated as the sums of all records from all fields in each tetrad on each visit (December-January, February-March and late March/early April), i.e. as three separate counts. These counts were compared with data, matched by month, from BD1628, which focused only on seed-rich habitats, so are likely to under-record species that also use other field types or non-agricultural habitats, the boundaries of which with cropped fields were surveyed in 2008-09.

Changes in bird abundance were analysed using generalized linear mixed models in SAS, modelling counts as a function of project (data collected under the present project in 2008-09 versus either data from 2004-07 or 2007-08) to inform about temporal changes in abundance and random effects for tetrad and observer. It was important to account for observer differences because fieldworkers changed between projects; although fieldworkers were trained, some differences between observers in surveying approaches and ability are inevitable.

Results and Discussion(i) Areas of uncropped landThere was remarkably little difference in the average areas of stubble seen in the presence and absence of set-aside, both during December-January and during February-March (Table 5), between 2004-07 and 2008-09. This contrasts with the results of BD1639 for 2007-08, which indicated average areas of stubble per tetrad for December-January and February-March, respectively, of 21.09ha and 10.49ha, i.e. reductions in stubble areas of around one third post-set-aside (BD1639 – Vickery et al. 2008a). The general pattern therefore seems to be for stubble areas to have been higher in 2008-09 than in 2007-08, throughout the winter, and to be back to the levels that were seen when set-aside was in place. This matches the pattern found in the national analysis and is likely to have the same causes (Objective 1). However, in late March 2009, stubble covered 8.67ha, which was higher than the corresponding figure for 2008 (5.91ha: BD1639 – Vickery et al. 2008a) but also lower than that for 2005-07: 12.47ha. Note also that differences in visit dates mean that the 2009 estimate may be biased relatively high. This suggests that the increase in over-winter stubble during December 2008 to February 2009, relative to the

SID 5 (Rev. 05/09) Page 12 of 22

previous winter, was largely not reflected in areas of the late winter stubble that was a critical feature of the benefits of set-aside to farmland birds (BD1640 – Vickery et al. 2008b). An additional loss of uncropped land involved areas recorded as “fallow”, which fell from averages of 2.9 and 1.9ha/tetrad in December-January and March-April, respectively, during 2007-08 to 1.1 and 0.4 the following winter, suggesting that more of the potentially usable arable land was being cropped in 2008-09. Cropping records from spring 2009 supported the loss of season-long uncropped land: the results of the poor weather were clearly visible, with many “winter” cereal and rape crops clearly having been sown later and thus maturing later, but there was little evidence of fields being left as unplanned fallows.

Table 5. Comparison of areas of cereal stubbles in December-January, February-March and March-April 2004-05 – 2006-07 and 2008-09 (with and without set aside respectively). Data for 2004-07 were taken from BD1628.

SITE (2×2-km tetrad)

Cereal stubble area (ha/tetrad)

December-January February-March March-AprilMean 2004-07 2008-09 Mean 2004-07 2008-09 2008-09

Attleborough 40.04 55.00 11.85 53.40 53.40Balsham 10.52 6.23 10.03 6.23 0.00Bawburgh 12.73 23.45 11.48 5.88 5.88Belchamp 24.89 0.00 11.52 0.00 0.00Bovingdon 25.53 38.96 22.09 13.99 13.99Bowbeck 31.22 0.00 12.57 0.00 0.00Brooke 9.08 12.51 4.63 3.64 0.00Clare 28.80 18.77 17.77 1.68 1.68Dickleburgh 29.41 60.71 11.20 5.56 5.56Glemham 24.96 20.51 23.14 17.23 17.23Great Barton 61.16 55.58 37.32 55.58 0.00Great Tey 0.15 8.74 0.00 8.74 8.74Hildersham 66.35 54.98 40.64 25.23 25.23Mattishall 35.79 42.16 5.44 0.00 0.00Morley 21.43 38.86 4.50 11.60 11.60Raveningham 69.77 54.65 12.33 21.52 21.52Shelfanger 5.68 2.49 8.51 0.00 0.00Thrandeston 48.79 62.54 26.02 8.44 0.00Wyken 83.06 105.91 52.43 80.50 0.00MEAN 33.12 34.84 17.03 16.80 8.67% of farmed area 9.74 10.25 5.01 4.94 2.55

(ii) Winter bird abundanceThere was no evidence for negative effects of set-aside loss on winter bird counts: counts either did not differ significantly or showed increases (Table 6). For December-January, the latter involved chaffinch and goldfinch and, for February-March, these species as well as linnet, grey partridge and skylark (Table 6). Similarly, there were apparent increases for corn bunting, chaffinch, goldfinch, greenfinch, linnet and skylark comparing late winter counts (March surveys from 2005-07 versus February to April ones from 2009; Table 6), although these comparisons are less reliable because of the variation in survey dates. There were no significant differences between counts in 2007-08 and 2008-09 for visit 1 or visit 2 (Table 6).

These results are similar to those of BD1639 and, as then, apparent increases probably largely reflect differences in field methods from those used in BD1628. The difference in habitat focus means that birds not using obviously seed-rich habitats will only have been recorded from 2007-08 onwards and the recording of data from “hotspots” away from seed-rich habitats reported under Objective 3 shows that significant numbers of many species in such locations could have contributed to apparent increases. Nevertheless, it is clear that the results provide no evidence for reductions in the winter abundance of the species considered after the loss of set-aside, although breeding season surveys and long-term patterns will provide a more reliable assessment.

Table 6. Comparison of winter counts of key bird species in December-January, February-March and March-April 2004-05 – 2006-07 and 2008-09 (with and without set aside respectively). Data for 2004-07 were taken from BD1628. Figures shown in bold for the overall mean from BD1628 and the mean from 2008-09 were significantly different from one another at at least the 5% level. There were no significant differences between counts in 2007-08 and 2008-09.

SID 5 (Rev. 05/09) Page 13 of 22

Species Max sites

December-January February-March MarchBD1628 2004-07

BD16392007-08

BD16412008-09

BD1628 2004-07

BD16392007-08

BD16412008-09

BD1628 2004-07

BD16412008-09

Overall mean (SE)

Mean (Range)

Mean (Range)

Overall mean (SE)

Mean (Range)

Mean (Range)

Overall mean (SE)

Mean (Range)

Corn Bunting 3 0.6

(0.5)1.6 (0-300)

5.3 (0-74)

1.5 (1.1)

0.4 (0.0-7.0)

1.7 (0-21)

1.5 (1.1)

4.1 (0-56)

Chaffinch 19 71.5 (12.7)

114.4 (6-326)

137.4 (40-444)

36.7 (6.6)

71.6 (8-218)

130.4 (33-365)

31.3 (5.0)

115.4 (36-365)

Goldfinch 19 7.3 (1.4)

19.4 (1-56)

33.4 (3-85)

4.8 (1.5)

13.5 (0-38)

27.8 (9-53)

5.0 (1.5)

25.1 (4-53)

Greenfinch 19 22.7 (4.4)

17.4 (0-76)

29.7 (1-145)

14.8 (3.0)

20.4 (0-58)

24.3 (3-57)

10.3 (1.6)

25.9 (3-57)

Linnet 19 28.9 (9.2)

20.8 (0-71)

53.1 (0-411)

11.1 (3.9)

23.0 (0-77)

50.0 (1-327)

8.2 (2.3)

45.8 (5-327)

Grey Partridge 17 3.0

(1.2)1.8 (0-14)

3.0 (0-13)

1.1 (0.3)

3.0 (0-12)

2.9 (0-12)

1.2 (0.4)

2 (0-12)

Reed Bunting 19 5.2

(3.0)8.0 (0-53)

2.6 (0-11)

3.9 (1.2)

5.9 (0-31)

2.8 (0-12)

12.4 (3.3)

3.3 (0-17)

Skylark 19 65.7 (15.0)

107.0 (2-349)

88.1 (11-262)

22.7 (7.9)

58.3 (1-239)

65.7 (1-247)

12.9 (3.2)

58.5 (1-145)

Yellow-hammer 19 48.3

(9.7)56.1 (1-175)

61.5 (5-251)

44.6 (13.6)

40.4 (0-135)

62.6 (9-163)

35.3 (7.6)

50.7 (9-155)

Breeding bird populations

MethodsBreeding bird surveys were conducted in April-June 2009 in the same 19 study tetrads in East Anglia that were covered in winter. These surveys potentially provided a more sensitive assessment of the effects of set-aside loss on breeding farmland bird populations than the national analyses under Objective 1 because we were able to repeat surveys conducted annually, under BD1628, from 2005 to 2007 precisely, matching observers, transect routes and survey dates, and because the sample of squares was exactly the same in all years considered. In addition, the data on winter habitat areas (specifically those of stubbles and other fallows) allowed us to investigate the effects of areas remaining under stubble locally. Field methods, adapted from the standard BBS approach, followed those for BD1628 precisely. Survey timing was standardized from year to year during BD1628 and the re-surveys.

Data were summarized as square- and species-specific totals per survey visit, and the maximum across the two visits was then used as the measure of annual species abundance in each square. All species observed were considered for analysis and results are presented for all that were recorded sufficiently frequently to produce potentially useful results (i.e. reasonably tight confidence intervals). All analyses used standard models for count data in SAS, correcting for over-dispersion and using a repeated measures structure to allow for correlations between adjacent squares. Counts were analysed to reveal temporal changes in abundance, while a second analysis replaced used a two-level “project effect”, i.e. whether data came from BD1628 (with set-aside) or this project (post-set-aside) and a direct test of whether counts were higher or lower in 2009 than during 2005-2007.

A further analysis tested for direct effects of winter foraging habitat area on the change in abundance between 2005-2007 and 2009 by using areas of stubble in each winter during BD1628 and the current project as a predictor of the abundance in the following breeding season of species likely to be affected by winter seed availability. Significant, positive stubble area effects would indicate that changes in abundance had been consistent with their having been driven by local fluctuations in winter stubble areas. Stubble areas were considered separately for December-January and February-March.

Results

There was a wide variety of temporal trends among the 42 species considered. If the loss of set-aside had affected a population negatively, the trends would be expected to show a downward step between 2007 and 2009: such a pattern was apparent to the eye for chaffinch, greenfinch, mallard, starling and yellowhammer, as well, less clearly, as for blue tit, goldcrest, house sparrow, rook, skylark and turtle dove (Figure 3; see Appendix 2 for all species’ trends). Among these species, the formal “project effect” tests indicated that all of these changes were significant, with the exception of rook and goldcrest. There was, however, no strong evidence that areas of

SID 5 (Rev. 05/09) Page 14 of 22

stubble affected trends of species that use stubble fields in winter. There positive associations between stubble and trend for grey partridge (December/January stubble), chaffinch and (weakly) linnet (February/March stubble), but negative ones for skylark (December/January) and goldfinch (February/March). It is quite possible that at least some of the significant associations identified simply reflect other, correlated environmental effects or chance patterns.

Figure 3. Population trends in East Anglian tetrads for four “stubble species” showing population changes consistent with negative effects of set-aside loss.

The downward step changes in abundance found here (e.g. those in Figure 3) represent the first landscape-scale evidence consistent the large-scale negative effects of set-aside loss on biodiversity, although the changes recorded were not large. A caveat to this conclusion is, of course, that the results are only correlative and other changes in the environment could have been responsible. This is underlined by the occurrence of similar patterns in species that cannot, plausibly, have been influenced by set-aside (e.g. blue tit, mallard and starling). In addition, an effect of disease on greenfinches and, perhaps also chaffinches, is highly likely to have caused a decline at this time (Robinson et al. in press), so factors independent of set-aside could have affected “stubble species” too. This would be clarified by direct analyses of population change with respect to set-aside area, but there was no clear evidence for such an effect within or between species using stubble areas. This does not necessarily indicate a lack of an effect of set-aside because stubble was probably a poor indicator of set-aside availability and quality and because winter habitat areas probably affect birds at a larger spatial scale than was considered here (BD1628). Overall, therefore, this study provides no conclusive proof that set-aside loss has begun to cause declines in farmland bird abundance, but the East Anglian results do provide the first evidence

SID 5 (Rev. 05/09) Page 15 of 22

(a) Chaffinch

0

0.2

0.4

0.6

0.8

1

1.2

1.4

2005 2006 2007 2008 2009

Year

Inde

x

(b) Greenfinch

0

0.2

0.4

0.6

0.8

1

1.2

1.4

2005 2006 2007 2008 2009

Year

Inde

x

(c) Skylark

0

0.2

0.4

0.6

0.8

1

1.2

1.4

2005 2006 2007 2008 2009

Year

Inde

x

(d) Yellowhammer

0

0.2

0.4

0.6

0.8

1

1.2

1.4

2005 2006 2007 2008 2009

Year

Inde

x

consistent with such an effect having occurred at a large scale, using surveys likely to be more sensitive than the national BBS.

Objective 3: the nature and use of secondary foraging habitats

The results of BD1639 suggested that granivorous birds might make greater use than had previously been suspected of peripheral farmland habitats such as field boundary vegetation and raised the question of whether the resources they found there actually replaced those previously provided by set-aside (or could be managed so to do). The work towards this objective aimed (i) to quantify the extent to which boundary habitats are used and to assess how this differs between seed-rich and other fields, (ii) to assess what habitat features characterize ‘hot spots’ of bird-use in field boundaries, (iii) to measure time budgets of birds in hotspots and (iv) to measure the stability of such ‘hotspots’ over time. This work is described in detail in Appendix 3.

(i) Use of boundary versus in-field habitats with respect to field type

MethodsUsing the 2008-09 winter bird survey data from Objective 2, proportions of birds found in in-field versus boundary vegetation were compared between (a priori) seed-rich fields (crop stubbles and wild bird covers) and other crops. Birds present in the boundaries of seed-rich fields are most likely to be resting or loafing between foraging bouts, whereas there is no obvious reason why these species should use the boundaries of non-seed-rich fields unless they are finding food there. Percentages of the local populations of each species in the boundaries and field centres of seed-rich and other field types were compared using 2 tests.

ResultsLarge proportions of the populations of all these species were found in field habitats not usually considered “seed-rich”, i.e. fields under active cropping and bare plough/till, and these proportions rose between successive survey visits through the winter as the available areas of seed-rich habitat (chiefly stubble and wild bird cover) fell. Note, however, that stubbles and wild bird covers were still selected actively. This pattern mirrors the results of BD1639 and suggests that granivorous birds are finding food resources outside stubbles, covers and fallow fields, as well as within these accepted seed-rich habitats. Bird counts were more biased towards records in boundary vegetation in non-seed-rich field types than in stubbles and covers for chaffinch, yellowhammer and grey partridge, suggesting that the resources the birds find in cropped fields are more concentrated in the boundary vegetation than in the cropped area. However, it should be noted that the opposite pattern was found for linnet and (less clearly) skylark (Table 6): both species tend to avoid field boundary habitats in winter and this result seems primarily to reflect a reduced avoidance of boundary habitats in seed-rich fields (Robinson & Sutherland 2000). It is unknown whether the use of boundary habitats has increased since set-aside disappeared and the data collected under BD1628 do not allow this to be tested.

(ii) Assessment of habitat features defining bird-use “hotspots”

MethodsData on the structure and vegetation content of all field boundaries in the 20 study tetrads in East Anglia were collected under BD1628, allowing the features of areas of boundary being used by farmland birds (“hotspots”) during winter 2008-09 to be assessed with respect to the range of habitats available. Boundary recording was conducted on the basis of discontinuities in structure delimiting sections in which habitat was constant, considering all boundaries of agricultural habitats. The habitat in each section was then recorded using variables intended to capture all of the structural variation observed. Section lengths were extracted from a GIS. The locations of all chaffinches, yellowhammers, greenfinches, goldfinches, reed buntings, linnets, tree sparrows and corn buntings seen or heard within tetrads during the winter field surveys described under Objective 2 were recorded on field maps. All records of groups of more than two individuals of each of these species that were not in open field areas were used to define the associated boundary habitat section as a “hotspot” for the species concerned. Boundary sections adjacent to known seed-rich habitats (stubbles or wild bird covers) were identified as such. Separate sets of hotspots were defined for each survey visit (December-January, February-March and April, the latter involving just seven tetrads).

Selection of a boundary section as a “hotspot” by a species, on a given survey visit, was tested using logistic regression models of the binary variable for selection as a function of boundary habitat features, controlling for boundary section length. Data were pooled across study tetrads. All boundary variables were tested for influences on boundary selection by each species; hedge and tree dimension variables were tested (using SAS) among the hedgerow and remnant hedgerow habitats only. These tests were then repeated using a second definition of hotspot habitat selection, considering only flocks of more than ten birds, in order to reveal whether particular features were associated with larger feeding flocks. Note that these data included fewer “selected” locations, increasing the potential for apparent patterns of selection merely to reflect chance associations.

SID 5 (Rev. 05/09) Page 16 of 22

Results Tests on hotspot selection with respect to habitat were conducted for all the target species except reed bunting, tree sparrow and corn bunting, for which none were possible because no groups, or only one or two groups, respectively, of more than two individuals were found away from seed-rich habitats. There was clear pattern for birds to be found in a combination of small and large groups (Figure 4). Larger groups in hotspots became rarer as the winter progressed (Figure 4), so tests could not be conducted for tree sparrow, corn bunting and linnet for one or both of the later two visits. More individual greenfinches and linnets were found in larger flocks. The proportion of the birds recorded found in hotspots (i.e. groups of three or more away from seed-rich habitat fell rapidly after December-January for all species (e.g. 39% for chaffinch falling to 5% in February-March and 3% in April, respective figures for yellowhammer 33%, 17%, 2%). Large flocks were presumably associated with particularly valuable food resources while the fall in their occurrence over the winter probably reflected a combination of increased territorial behaviour and decreased food densities that only supported a few individuals.

Figure 4. Frequency of occurrence of different flock sizes at hotspot locations for chaffinch and yellowhammer. Fewer data were available for other species (see Appendix 3). Data are separated by visit. Visit 1 was December-January, visit 2 February-March and visit 3 April. Note that only seven tetrads were visited in April.

There were at least two significant results for each species (except for corn bunting and tree sparrow, for which samples were small) across the range of tests conducted for general habitat features of selected versus non-selected boundary sections, but there was no evidence for consistent patterns across species. Chaffinches in December-January tended to select boundary sections without field margins and showed some, relative avoidance of isolated bushes and “narrow” boundaries, while never being found in treelines or banks. In February-March, goldfinches showed preferences for boundaries without ditches and with adjacent tracks or, to a lesser extent, no human thoroughfare. Greenfinches showed preferences for boundaries with margins and adjacent to gardens in December-January. Linnets selected boundaries with ditches in December-January and without margins in February-March. Yellowhammers showed significant habitat selection only in December-January, when they tended to select boundaries with ditches but to avoid those along roads and with adjacent gardens. Among solid and gappy hedges, and isolated bushes, there were rather few significant associations with structural variables and no clear, consistent patterns across or within species. Chaffinches showed a weakly negative association with margin width during April visits, while greenfinches showed a positive association in February-March. Greenfinches and yellowhammers (April) showed positive associations with boundary width, but greenfinches also showed a weakly negative association in December-January. It is important to note that hotspots were defined as locations of groups of more than two birds, because March/April locations were always likely to include pairs and territorial singles as opposed, strictly, to birds foraging for “winter” food alone. However,

SID 5 (Rev. 05/09) Page 17 of 22

(a) Chaffinch

0

5

10

15

20

3 4 5 6 7 8 9 10 11 12 13 14 15

16-2

0

21-3

0

31-5

0

51-1

00

>100

Hotspot bird count

Freq

uenc

y

Visit 1

Visit 2

Visit 3

(b) Yellowhammer

0

2

4

6

8

10

3 4 5 6 7 8 9 10 11 12 13 14 15

16-2

0

21-3

0

31-5

0

51-1

00

>100

Hotspot bird count

Freq

uenc

y

Visit 1

Visit 2

Visit 3

a weakness of this approach was that scattered, sparse food resources might not support larger groups, so foraging birds might also only be in ones or twos. Such a pattern might be more likely to occur later in the winter, once more concentrated food sources have been depleted. This would limit the ability of the analyses here to identify influences on foraging later in the winter, when individuals divide their time between prospective breeding locations and seed-rich food sources. Nevertheless, the results clearly indicate that birds are not becoming more concentrated in hotspot areas away from acknowledged seed-rich habitats as the winter progresses. This could be because such habitats do not support larger groups of foraging birds because seed densities are too low. The reduction in the numbers of hotspots through the winter (Figure 4) would support this, suggesting that seed food in field boundary habitats could have been depleted significantly during the winter. This pattern would also suggest that such habitats are actively selected in early and mid-winter, despite the relative commonness of stubble fields and wild bird covers at that time

Some of these issues should have been addressed by considering “selected” locations only to be those with flocks of more than ten birds. However, fewer tests were then possible, especially later in the winter, because large flocks were rarer or absent entirely (Figure 4). Some of the results identified with all group sizes were still apparent, but some disappeared and a few new associations appeared, but there were still no clear general patterns across or within species (Appendix 3).

(iii) Time budgets of birds in hotspots

Critical to the assessment of the role that habitat “hotspots” play in providing an important food resource for granivorous birds in winter is to make some measurement of the resources that they provide. This would be extremely difficult to do directly. First, seed abundance on plants is at best only weakly related to seed availability to birds that forage on the ground (as is shown by weak relationships between seed indices and bird numbers, e.g. Hinsley et al. 2010). Second, measuring seed on the ground is both difficult to achieve at a sufficiently shallow depth to reflect bird foraging abilities and fails to account for competition between species, involving both bird species and a range of other taxa (Holmes & Froud-Williams 2005), which affects the amount actually supplied to a given species. Third, seed densities are naturally very heterogeneous (S. Queenborough & R. Freckleton, unpublished), so it would be critical to identify the exact location in which birds were foraging, perhaps to a resolution as small as a square metre, to measure the resource that they are using accurately.

An alternative approach is to consider bird behaviour as an indicator of the value that habitats have for the species that are found there. Birds found in known seed-rich habitats such as stubble fields and sacrificial seed crops can be assumed to be feeding there, while the use of hedgerows adjacent to such habitats is likely often to reflect its role as cover, i.e. a refuge between foraging sorties. Birds found elsewhere could be feeding or loafing or, later in the winter, prospecting for breeding territories. Thus, one way to assess what value birds find in the locations in which they are found in winter is to compare behaviour between locations where birds are known to feed with “hotspots” away from accepted seed-rich habitats.

MethodsSix or more locations in each tetrad were selected for detailed observation after the December-January survey visit from the hotspots described above and boundary areas next to known seed-rich habitats. These locations were then revisited on a subsequent day as soon as possible after the original survey was carried out and again around one month later. In six tetrads, the same hotspots were revisited again, monthly, until the end of the project while, in the other 13 tetrads, the hotspot list was “reset” after the second (February-March) and third (April, if it occurred) survey visits. This protocol was designed to record whether hotspots persisted and for how long, considering periods of days, weeks and months. On each visit, formal observations were conducted to record the numbers of feeding birds present and, critically, the proportion of the time observed that they were doing so. The output was an index of the use that birds made of a given hotspot, derived from the balance between time spent feeding and time spent perching. This is a coarse measure, but was the best possible given the difficulties inherent in observing behaviour in small farmland passerines foraging on the ground or in dense vegetation. Data on habitat structure at hotspot locations were also collected, including gross structure and the density/nature of field layer (herbaceous) vegetation. Full details of this protocol are given in Appendix 3.

The proportion of time spent feeding was analysed to reveal how feeding versus perching profiles for each species across the range of hotspots identified varied between known seed-rich habitats and other hotspots and with respect to key habitat structural variables in the hotspots. Models were run on both a species-specific basis and combining species’ data.

ResultsChaffinch, yellowhammer and linnet all spent a greater proportion of their time apparently feeding in hotspot areas that were not adjacent to recognized seed-rich habitat than they did in locations next to such habitat, and this pattern was reflected in a strongly significant difference when all species were considered together. This could indicate that birds find greater feeding value in hotspots away from stubbles and covers, but it is more likely that

SID 5 (Rev. 05/09) Page 18 of 22

they actually need to forage for longer on semi-natural food sources than they do in cropped habitats, probably because seed densities are higher in the latter. As a measure to minimize predation risk, wintering granivorous passerines often feed rapidly if they can, collect a crop-full of seed and then retreat to safe cover to digest the seed (pers. obs.). Digestive constraints can also drive similarly alternating foraging and resting behaviour in other species, for example waders (e.g. Kersten & Visser 1996). Nevertheless, the results suggest that birds are using hotspot locations for foraging: if they were not, significant differences in the opposite direction would have been expected.

There was no evidence for variation in patterns of behaviour with respect to date during the winter, so it appears that these boundary locations are used similarly for foraging throughout the winter and there is no evidence that the behaviour of birds in hotspots changes as other seed food resources disappear. Note, however, that the data and analyses may be too coarse to allow the detection of subtle changes in behaviour that may result from shifts in foraging effort between habitats. However, there were three significant patterns of behaviour with respect to habitat that were shared across species, suggesting that they show real, biological effects. First, chaffinch, reed bunting and greenfinch all spent more time feeding at hotspots with a higher proportion of bare ground amongst the boundary vegetation. Second, yellowhammer and chaffinch both fed more where the herbaceous vegetation in the boundary was lower. Access to bare ground is critical for granivorous birds because that is where they are most able to forage for seed (refs), while shorter vegetation both promotes access to food (e.g. Douglas et al. 2009) and facilitates foraging by reducing the demand for anti-predator vigilance (Butler & Gillings 2004 BOU Ibis). A third pattern saw goldfinch and linnet feeding more where the trees at the hotspot location were taller, probably reflecting the influence of observations at a few hotspots in one tetrad where these species were observed feeding on fallen elm tree buds. There was little further evidence for any effect of woody vegetation structure on feeding behaviour.

(iv) Hotspot stability

As well as the habitat content of previously unrecognized feeding habitats for farmland birds, it is important to identify the extent to which the value of such habitats persists through the winter. If “good” habitats remain so for extended periods, then it would suggest that they might be promoted and protected as specific “bird areas” within agri-environment schemes. Alternatively, if these hotspots tend to be ephemeral, i.e. only to be valuable for short periods, it would be harder to promote them via area-based measures, and the focus might need more to be on dispersed management across the farm to increase food supplies for birds.

MethodsThe degree of stability of hotspot areas over time was addressed using the field data in two ways. First, maps of granivorous bird locations were compared between the two or three survey visits to reveal whether the same areas were used by each species on each visit. Second, hotspot locations selected for ease of observation were revisited at intervals from one day to four months after the survey visit on which they were identified in order to record details of bird behaviour there (see (iii) above), thus providing a more sophisticated assessment of hotspot persistence.

For the first approach, maps of granivore locations in each tetrad were simply compared between the two or three survey visits to reveal whether groups of birds of a given species had been found in the same places. The effective width or diameter of “a location” for this purpose was around 100m. For the second approach, formal statistical analyses were conducted to assess the extent to which hotspot use was related (i) to whether that hotspot had initially identified as a location for the focal species and (ii) to the number of days that had elapsed since the hotspot was identified (considering only hotspots identified for the focal species).

ResultsComparisons of maps between survey visits revealed very little evidence of hotspot persistence. Only 6.9% of chaffinch hotspots identified on the December-January visit and 6.6% of yellowhammer ones were also used in February-March, while only 1.5% and 3.3%, respectively, were also used during April visits, and no hotspots were used on all three visits. In addition, of the hotspots identified in February-March, only one, where a group of chaffinches was found, was also used in April. Fewer hotspots were found for other species, but the patterns were the same. The general pattern was, therefore, for hotspot areas to be ephemeral, which is consistent with the hotspot selection and behaviour results.

The use of hotspots by each species was, unsurprisingly, strongly related to whether the species had been present when the hotspot was initially identified, but the highest average probability of presence on subsequent visits was only 0.46 (Table 7), showing again that hotspot locations were not particularly stable. In addition, probabilities of finding a species at a hotspot at which it had not been present when the hotspot was identified were not zero (Table 7), showing either that the resources there were used by different species at different times or that there was considerable stochasticity in bird locations. There was also little evidence that patterns of hotspot use varied with the time elapsed between hotspot identification and observation, despite the periods considered ranging from days to months.

SID 5 (Rev. 05/09) Page 19 of 22

Hotspots that tended to persist for longer had different associated habitat features for different species. For chaffinch, persistence was greater where hotspots features more bare ground both in the boundary and in the margin, where the ground vegetation was dominated by broadleaved weeds. Persistence for goldfinches was also associated with higher hedges and taller trees, while that for greenfinch was also associated with more bare ground in the boundary and taller trees. There was a converse association with bare ground for reed bunting, but sample sizes were low for this species. There were no significant results for corn bunting and only weak associations for linnet and yellowhammer. These varied patterns could reflect differences in feeding ecology or behaviour, but all the species considered mostly feed on seeds on the ground in winter and are sufficiently commonly found in mixed flocks to suggest that they often use broadly the same resource base. Results for the commoner species might be more reliable because sample sizes were larger, so the results for chaffinch could be particularly important in showing what affects a species that forages on the ground on weed seeds, i.e. greater hotspot persistence in locations where there was easier access to bare ground and more broadleaved weeds.

Objective 4: update of the evidence base and potential for habitat management to mitigate effects of set-aside loss

(i) Population responses to set-aside loss

National BBS monitoring has yet to show any obvious declines in response to the loss of set-aside and our analyses specifically looking for such changes support this pattern. We did not detect effects even for Skylark and Yellowhammer, the two species for which there is the strongest expectation of a response, and for which power to detect changes is greatest due to the high proportion of BBS squares that are occupied. So, why have we not detected a response of farmland birds to the loss of set-aside? The main reason for monitoring population trends over two seasons following the loss of set-aside was that the policy would lead to an irreversible loss of uncropped habitat and that allowing more time to elapse would increase our capability to detect a decline following a lag. However, stubble occurrence in 2008-09 resembled that in 2004-07 more than that in the first winter post-set-aside. Two factors probably explain this pattern. First, wheat prices, although still high during autumn 2008, had fallen considerably from their historic peak a year earlier, so farmers may have been more predisposed to leave land fallow or to grow other crops, some of which will have been spring-sown and so often preceded by stubbles. Second, wet weather in autumn and winter 2008-09 delayed ploughing and sowing in many fields, so many winter crops were delayed and farming logistics meant that some fields may never have been sown at all. In any case, the partial return of non-cropped habitats in winter 2008-09 may have overridden the loss of set-aside, undermining our ability to detect population responses. If this were true, in the long term it could, clearly, mean that the effects of set-aside can be compensated for by other cropping changes, making further mitigation unnecessary. However, the cropping pattern in the previous winter demonstrates that areas of uncropped land in such a scenario are volatile and subject to rapid change due to market forces. Only patterns of cropping over an extended run of years will show whether market and weather conditions will lead to a consistent replacement of the uncropped land lost with set-aside. In practice, this seems unlikely to occur without management or policy intervention.

Given that clear, national bird declines have not occurred, we must consider the possibility that set-aside did not actually have the importance for farmland birds that it is widely believed to have had. This would certainly fit the observation that populations have not responded and that the resources it provided could have been replaced by a simple increase in stubble caused by changes in wheat prices and poor weather. However, the national BBS may not be sensitive enough to detect subtle changes in abundance in the short term (and the confounding patterns described above may mean that the changes are subtle): there is substantial turnover of squares and observers between years, while varying annual weather conditions and stochastic variation in counts in individual squares will affect the power to detect effects of habitat change. The loss of set-aside also did not occur in a vacuum: other environmental variation has occurred in parallel, including the wet, snowy winter of 2008-09, which could have had complex effects directly via the climate and indirectly via effects on cropping as described above, and also including the broad uptake of potentially ameliorative management under agri-environment schemes, notably ELS. It is possible, although very difficult to prove because both have the potential to have affected very broad areas of farmland, that the effects of set-aside loss and ELS have cancelled each other out for some species, preventing the detection of effects of either (see Davey et al. 2010). Further, effects of the disappearance of habitats should be detectable quickly, but the winter results from East Anglia suggest that birds are finding alternative habitats in the absence of set-aside (and may, indeed, always have used such habitats). Thus, the effects of set-aside loss will depend on the difference in quality between set-aside and the habitats that have replaced it. If the differences are small, negative effects will be small and may only be detectable in extreme conditions or after several years’ cumulative effects. For example, it is known that Skylarks will graze young cereal leaves when weed seeds are absent but intake rates achieved in this way are inferior (Green 1978): increased use of cereals might cause a decrease in individual fitness, but the birds still survive and effects on

SID 5 (Rev. 05/09) Page 20 of 22

population numbers might not be seen for some years, as was found in correlations between agricultural change and farmland bird decline in the 1970s (Chamberlain et al. 2000). Another way in which set-aside loss effects could occur subtly is via the timing of food resource requirements: increases in cereal stubbles could replace set-aside adequately in mid-winter, but are unlikely to do so in late winter or during the breeding season, because they will be destroyed in time for spring cropping. Thus, only where breeding season or late winter habitat is scarce will the effect of set-aside loss be seen (but such effects can nevertheless be significant: Siriwardena et al. 2008). In this context, it could be critical that the March and April survey visits in East Anglia indicated that even if there had been a long-term return to set-aside-era stubble areas, this was only true in terms of mid-winter stubble rather than the late-winter habitat that was critical in determining much of the value of set-aside for farmland birds. No data were available to test for late winter habitat change nationally.

Overall, therefore, the evidence is not yet conclusive that the loss of set-aside has had important effects nationally on habitat quality or availability for farmland birds, but there are good reasons why the effects might be subtle at present and that lags might occur before they are detectable. In addition, the arguments made previously as to the value of set-aside for farmland birds (Vickery et al. 2008b: BD1640) remain valid. Thus, it may be noteworthy that we have found three forms of evidence consistent with negative effects of set-aside on key species. First, there were positive associations between change in stubble area in 1-km squares and population change between 2007 and 2009 for three such species including yellowhammer. Second, once national yellowhammer population trends were split into arable and pastoral regions, there was some support for both having turned down around the time when set-aside disappeared. Third, we found small, but significant declines consistent with negative effects of set-aside loss in our breeding season surveys in East Anglia: survey methods and observers were more consistent than in the national BBS and the arable east is the region in which the most set-aside is likely to have been made up of seed-rich, rotational fallows. Thus, effects in this dataset might have been both the largest and the most detectable of those investigated to date and this provides the best evidence so far of landscape-scale effects of set-aside on bird populations.

(ii) Possible habitat management to mitigate set-aside loss effects

This project has identified both that field boundary habitats are where a large proportion of wintering finch and bunting populations are found throughout the winter and that they are used as foraging habitat for all of that time as well. We were unable to measure the potential of these habitats to replace the winter food resources that set-aside provided and it is obvious that they cannot replace other benefits of set-aside, such as the provision of in-field nesting habitat. Nevertheless, it would be possible to manage boundaries to provide the highest quality winter habitat in order to provide some of the resources found previously in set-aside if the habitats to manage could be identified. Unfortunately, however, the results in terms of the selection of “hotspot” areas, their persistence and bird behaviour within them all suggested that hotspot locations were ephemeral, variable between species (despite broad dietary overlap and flexibility) and unpredictable in terms of structural habitat variables. This makes guidelines as to habitat features to protect or to promote difficult to draw up. It seems most likely that birds use a wide range of habitats rather opportunistically, with weed, crop and livestock-associated seed sources being exploited where they are found in flock sizes broadly appropriate for the density of the food resource, birds often cueing in on food availability in ways that could be too subtle to detect easily by gross habitat recording. The pattern for three species to respond positively to areas on bare ground in the analyses of foraging behaviour in hotspots and for it positively to influence hotspot persistence supports the latter idea and fits with the management that would have been expected to benefit bird foraging a priori. Anecdotal observations of large flocks during the fieldwork for the project (summarized in Appendix 3) also support the idea of a wide, unpredictable and ephemeral range of selected habitat types: for example, seeding ryegrass attracted up to 250 yellowhammers for over a month in one tetrad, while up to 400 linnets and chaffinches were recorded using cereal fields which had previously held borage crops, probably eating spilt seed from the previous crop.

All of the above makes a specific prescription for a form of agri-environment option to promote “hotspot habitats” difficult to define spatially or in terms of specific management rules. However, some clear pointers for management policy do come out of this work, such as a lack of clear value for grass field margins as winter foraging habitat and no evidence for farmland species to show increased preferences for boundaries with gardens. Margins are a very popular form of agri-environment management, but this result is consistent with the lack of positive population changes associated with margins under Entry-Level Stewardship (ELS) for farmland birds that are likely to be limited by over-winter food supplies (Davey et al. 2010), while feeding in rural gardens represents one possible management route to aid farmland birds via private individuals without affecting farm economics. Conversely, there was some positive evidence for a preference for ditches by yellowhammers, supporting earlier anecdotal results from BD1628. Generally, however, the principal recommendation for management must be to manage habitats peripheral to farmland for seed availability; if the birds are flexible and opportunistic, it could mean that a wide range of potential contexts into which such seed resources could be placed would be equally valuable. Such flexibility could suit both ELS, where boundary management is the most popular type of option and promoting seed availability in boundaries could be a farmer-friendly option, and the Campaign for the Farmed Environment, which has a central premise the idea that currently uncropped odd corners of farmland can be managed more sympathetically to benefit the environment. Sowing small plots of seed

SID 5 (Rev. 05/09) Page 21 of 22

crops or keep soils broken to promote arable weeds in such areas could achieve this aim for granivorous birds. In both approaches, any means of providing seed resources that works logistically within the constraints of scheme organization might be valuable. Approaches might therefore best focus on the promotion of non-injurious weeds and their seeds and sacrificial seed crops in areas that are peripheral to the main economic activities of the farm, with management of crop or semi-natural vegetation to promote access to bare ground for foraging. Particular focus should also be given to seed provision in the later winter period when most recognized seed-rich habitats have disappeared (Siriwardena et al. 2008).

References to published material9. This section should be used to record links (hypertext links where possible) or references to other

published material generated by, or relating to this project. BD1639 Final report: Quantifying the magnitude of the loss of set-aside stubbles and its impact on the

winter ecology and distribution of farmland birds (Vickery, J.A., Siriwardena, G.M. & Gillings, S. 2008a British Trust for Ornithology).

BD1640 Final report: Zero rate of set-aside: evaluating the potential impact on farmland birds and the implications for requirements for ELS uptake and related agri-environment measures. (Vickery, J.A., Henderson, I.G., Aebischer, N.J., Chamberlain, D. 2008b, British Trust for Ornithology).

Butler, S.J. & Gillings, S. 2004. Quantifying the effects of habitat structure on prey detectability and accessibility in farmland birds. Ibis 146 (Suppl. 2):123-130.

Chamberlain, D.E., Fuller, R.J., Bunce, R.G.H., Duckworth, J.C. & Shrubb, M. 2000. Changes in the abundance of farmland birds in relation to the timing of agricultural intensification in England and Wales. J. Appl. Ecol. 37: 771–788.

Davey, C.M., Vickery, J.A., Boatman, N.D., Chamberlain, D.E. Parry, H.R. & Siriwardena, G.M. 2010. Assessing the impact of Entry Level Stewardship on lowland farmland birds in England. Ibis, Online Early.

Douglas, J.T., Vickery, J.A. & Benton, T.G. 2009. Improving the value of field margins as foraging habitat for farmland birds. J. Appl. Ecol. 46: 353–362.

Gillings, S., Newson, S.E., Noble, D.G. and Vickery, J.A. 2005. Winter availability of cereal stubbles attracts declining farmland birds and positively influences breeding population trends. Proceedings of the Royal Society B 272 (1564), 733-739.

Gillings, S., Wilson, A.M., Conway, G.J., Vickery, J.A. & Fuller, R.J. 2008 Distribution and abundance of birds and their habitats within lowland farmland of Britain in winter. Bird Study in press

Green, R.E. 1978 Factors affecting the diet of farmland skylarks Alauda arvensis. Journal of Applied Ecology 47, 913-928

Hardin, J.W. and Hilbe, J.M. 2003. Generalised Estimating Equations. Chapman & Hall, New York. Hinsley, S.A., Redhead, J.W., Bellamy, P.E., Broughton, R.K., Hill, R.A., Heard, M.S. & Pywell, R.F.

2010. Testing agri-environment delivery for farmland birds at the farm scale: the Hillesden experiment. Ibis, in press.

Holmes, R.J. & Froud-Williams, R.J. 2005. Post-dispersal weed seed predation by avian and non-avian predators. Agric. Ecosyst. Environ. 105: 23–27.

Kersten, M. & Visser, W. (1996) The rate of food processing in Oystercatchers: food intake and energy expenditure constrained by a digestive bottleneck. Functional Ecology, 10, 440-448.

Pan, W. 2001. Akaike’s Information Criterion in Generalised Estimating Equations. Biometrics 57, 120-125.

Robinson, R.A. & Sutherland, W.J. 2000. The winter distribution of seed-eating birds: habitat structure, seed density and seasonal depletion. Ecography 22: 447–454.

Robinson, R.A., Wilson, J.D., & Crick, H.Q.P. 2001. The importance of arable habitat for farmland birds in grassland landscapes. Journal of Applied Ecology 38, 1059-1069.

Robinson, R.A., Lawson, B., Toms, M.P., Peck, K.M., Kirkwood, J.K., Chantrey, J., Clatworthy, I., S.A., Evans, A.D., Hughes, L.A., Hutchinson, O.C., John, S.K., Pennycott, T.W., Perkins, M.W., Rowley, P.R., Simpson, V.R., Tyler, K.M. & Cunningham A.A. Emerging Infectious Disease leads to Rapid Population Declines of Common British Birds. In press, PLoS One.

Siriwardena, G.M., Calbrade, N.A. & Vickery, J.A. (2008) Farmland birds and late winter food: does seed supply fail to meet demand? Ibis 150: 585–595.

Toms, J.D. and Lesperance, M.L. 2003. Piecewise Regression – a tool for identifying ecological thresholds. Ecology 84 (8) 2034-2041.

SID 5 (Rev. 05/09) Page 22 of 22