Aspects of Applied Biology 118, 2013

Aspects of Applied Biology 118, 2013

Environmental Management on Farmland

Multi-taxa benefits of a targeted single-species

agri-environment option

By J C DUNN1, V HARTWELL1,2 , P GRICE3 and A J MORRIS1

1Centre for Conservation Science, RSPB, The Lodge, Potton Road, Sandy, Bedfordshire SG19 2DL, UK

234 Wisbech Road, Long Sutton, Spalding, Lincolnshire PE12 9AG, UK

3Natural England, Suite D, Unex House, Bourges Boulevard, Peterborough PE1 1NG, UK

Corresponding Author Email:

Abstract

Agri-environment schemes such as the Environmental Stewardship in England provide a range of options, including some targeted at single species, which may benefit a wide range of other species or taxa. To ascertain whether trials of an agri-environment measure designed for a single species had wider biodiversity benefits, we describe monitoring of pollinators (bumblebees and butterflies) on plots of a novel seed mix aimed at providing a source of seed food throughout the breeding season for the rapidly declining European Turtle Dove Streptopelia turtur. We compare pollinator abundance on these turtle dove plots to that on six control habitat types: fallow, grass margins, meadow, nectar flower mixture, wild bird cover and floristically enhanced margins. The abundance of foraging bumblebees and butterflies was higher on turtle dove plots than in fallow and wild bird cover habitats. Surprisingly, foraging bumblebee abundance was higher, and foraging butterfly abundance marginally higher on turtle dove plots than on nectar flower mixture. Whilst the specific mix described here is to be altered to increase access to foraging turtle doves, we suggest that the modified mix, to be rolled out through Higher Level Stewardship, is also likely to provide valuable resources for pollinators, adding to the growing body of literature demonstrating multi-taxa benefits of agri-environment options designed to benefit one species.

Key words: Wider biodiversity benefits, pollinator, environmental stewardship, turtle dove

Introduction

Agricultural intensification has been closely linked with declines in farmland bird populations in the UK and across Europe (e.g. Chamberlain et al., 2000; Donald et al., 2001; Robinson & Sutherland, 2002). Agri-environment schemes (AES) aimed at reversing this decline in farmland biodiversity have been in place in the UK since 1987, and the current scheme in England, launched in 2005, is Environmental Stewardship (ES), which has multiple elements. ‘Broad and shallow’ Entry-Level Stewardship (ELS), and associated Organic and Uplands ELS, rewards farmers for delivering simple environmental management above that required by cross-compliance, whereas the Higher-Level Stewardship (HLS) provides more targeted, longer duration agreements (usually underpinned by ELS) aimed at delivering significant environmental benefits in high-priority situations and areas, including for certain key species, which may require individually tailored options aimed at providing a specific resource. For example, fallow plots for stone curlew (Evans & Green, 2007) and low-input spring barley or permanent pasture without inputs for cirl buntings (Peach et al., 2001) in HLS and, in Scotland, early and late cover and delayed mowing for Corncrake Crex crex (O’Brien et al., 2006). Options designed for all three of these bird species have successfully reversed declines of the target species (Wilson et al., 2009) and demonstrated benefits for wider biodiversity (MacDonald et al., 2012a,b; Wilkinson et al., 2012).

The European Turtle Dove Streptopelia turtur is one of Europe’s fastest declining farmland birds, with a UK population trend of -93% since 1970 (Eaton et al., 2012), paralleled by a 73% decline across Europe since 1980 (PECBMS, 2012). As such, it is one of six bird species that HLS specifically targets to provide habitats and food. The species is unique in the UK, as a trans-Saharan migrant solely reliant upon seeds all year round. Turtle dove reproductive output in England probably halved during the period of decline (Browne & Aebischer, 2004), with a concurrent switch in their diet from the seeds of arable weeds to cereal and oil seed rape seeds, most commonly found in anthropogenic situations such as farmyards and food for outdoor livestock, game or wild birds (Browne & Aebischer, 2003). Turtle dove nesting habitat is not currently thought to be limiting, although Browne et al. (2004) suggested that a reduction in suitable nesting habitat may have played a role in the long-term decline and a recent analysis suggested that habitat selection may be partly determined by the availability of suitable foraging habitat (Dunn & Morris, 2012). Together with the recorded dietary switch (Browne & Aebischer, 2003), this suggests that food availability may be limiting breeding productivity and, elsewhere, we describe a trial of a new HLS management option aimed at providing a source of seed food for turtle doves throughout the breeding season (J C Dunn & A J Morris, in preparation). This seed mix is clover-based, but incorporates other arable plants that are known components of turtle dove diet, and is designed to provide a sparse sward to allow turtle doves to access the seeds. As the mix is designed to start flowering and seeding when turtle doves return from migration in early May, and to continue flowering and seeding throughout the breeding season, we expected this mix to have wider biodiversity benefits beyond the target species, comparable to existing agri-environment options aimed at other taxa. Here, we investigate this question by assessing pollinator (bumblebee and butterfly) abundance on turtle dove plots and on six alternative habitats provided by ES.

Materials and Methods

A total of 23 turtle dove plots were established on five farms during autumn 2010, and, due to unsuitable ground conditions in autumn 2010, six further turtle dove plots were established on a sixth farm in March 2011. Plots ranged in size from 0.063 to

Table 1. Trial plot seed mixtures. Mix 1 is that described in this study; mix 2 is that now available under HLS as a modified nectar flower mix

% weight
Species / Mix 1 / Mix 2
Common Fumitory Fumaria officianalis / 2.88 / 5.0
Corvus Red Clover Trifolium pratense / 14.3 / 10.0
Avoca White Clover Trifolia repens / 14.3 / 20.0
Virgo Black Medick Medicago lupolina / 14.3 / 20.0
Early English Common Vetch Vicia sativa / 54.1 / 25.0
Common Mouse-Ear Cerastium fontanum / 0.12 / 0
Bird’s Foot Trefoil Lotus corniculatus / 0 / 20.0

1.178 ha (mean ± se: 0.301 ± 0.046 ha), were of variable shape to fit in with existing crops and ES options and covered a total of 2 ha on each farm (except one farm where plots covered 1 ha). Farms were located across East Anglia, with two in Essex, two in Cambridgeshire, one in Norfolk and one in Suffolk, on farms known to have at least two territorial turtle doves during 2010. The trial plot seed mix consisted of plants known to be important in turtle dove diet, to seed early, and to be largely non-pernicious to agricultural operations (detailed as Mix 1 in Table 1). Plots were sown at the rate recommended by the seed supplier (Kings of Holbeach) at 20 kg ha-1, intended to form a fairly sparse ground cover and ensure accessibility of seed to foraging turtle doves.

During 2011, between five and six (mean ± se: 5.83 ± 0.17) control plots were selected on each farm. Control plots were of comparable size to trial plots and consisted of discrete areas of habitat considered to form alternative turtle dove foraging habitats and to be potentially beneficial to pollinating arthropods that were currently available on farms either through agri-environment schemes or voluntary measures. These fell into the following categories (followed by sample size during 2011 and 2012 in parentheses): fallow, including lapwing fallow plots and other sparsely vegetated areas (6, 7), meadow (5, 2), floristically enhanced margins (4, 5), grass margins including grass paths (7, 0), nectar flower margins (3, 12), wild bird cover (10, 0).

Pollinator surveys were carried out only on still, dry days when temperatures were greater than 12°C (Redpath et al., 2010), between 10:00 and 15:00 h. If cloud cover was greater than 60%, then temperatures were greater than 17°C before surveys were carried out (Carvell et al., 2007). Controls and turtle dove plots were surveyed on the same day, alternately where possible, to preclude confounding temporal effects. A diagonal transect was walked across each trial and control plot between one and three (mean ± se: 1.97 ± 0.05) times during each of 2011 and 2012. All butterflies and bumblebees within 2.5 m of the observer were identified to species (where possible) or genus using standard texts (Anon., n.d.; Natural History Museum, 2012). Whether an individual was actively foraging and, if so, on what host plant, was recorded for each observation. Temperature (measured using a thermometer ± 1°C) and cloud cover (by eye, ± 10%) were recorded at the start and end of each transect, and the length of time spent carrying out each transect was recorded using a stopwatch (± 5 s).

Fig. 1. Trial plot flower abundance as proportion of total flower abundance during each survey round. Bars show mean proportion ± 1 se.

Flower abundance is tightly linked with pollinator abundance (Potts et al., 2009), so we recorded flower abundance at four evenly spaced points within each transect. A 0.5 m × 0.5 m quadrat divided into 25 equally-sized squares was dropped at each of these four points, and the number of squares within which open flowers were present was noted. In order to establish the success of establishment of the mix, to confirm that the trial plot species were those influencing pollinator abundance, we also counted the number of flowers of each species within each quadrat.

Data analysis

Generalised Linear Mixed-Effect Models (GLMMs) were constructed using the lmer function in the lme4 library in R v. 2.14.2 for Mac. Two models were constructed to examine effects of habitat on the total abundance of each of foraging bumblebees and butterflies per survey separately, pooled across species. Poisson error distributions were fitted to the models, and individual level random effects were included in models to control for overdispersion by translating to a lognormal-Poisson model. To control for geographical variation in pollinator abundance, Farm ID was included a random term within each model. Whilst data collection was undertaken by different fieldworkers within and between years, data collection on matched trial and control plots was undertaken on the same farm by a single fieldworker. Thus, fieldworker ID was not controlled within the statistical model.

Each minimal model contained the response variable and random terms only. Terms included to examine their impact upon pollinator abundance were Plot habitat, round (1, 2 or 3), Year, the hour in which the survey started, temperature (the average of start and finish temperatures), cloud cover (the average of start and finish temperatures), the number of squares containing flowers (log transformed) and the time spent walking the transect (in s). Each term was tested in turn by adding it into the model and using model comparisons to determine its importance in influencing the model fit. The most significant terms were then added sequentially in a stepwise fashion until no terms influencing the fit of the model at P<0.1 remained.

Table 2. Factors influencing the abundance of a) foraging bumblebees and b) foraging butterflies. Estimates are provided for terms included in the final model (for factors: significant levels in bold; non-significant levels in italics). Statistics given for non-significant terms are from comparisons of the models with and without the term. Estimates for factors are compared to Round 1 (Round), trial plots (Plot habitat), and 2011 (Year). All models contain Farm ID as a random term

a / Estimate / se / c2 / df / P
Intercept / -7.476 / 0.433
Round (2) / 3.121 / 0.215 / 216.62 / 2 / <0.001
Round (3) / 2.317 / 0.260
Plot habitat (fallow) / -2.079 / 0.407 / 77.200 / 6 / <0.001
Plot habitat (grass) / -0.293 / 0.427
Plot habitat (meadow) / -0.308 / 0.358
Plot habitat (pollen and nectar) / -1.033 / 0.248
Plot habitat (wild bird cover) / -2.173 / 0.505
Plot habitat (floristically enhanced margin) / 0.647 / 0.263
Year (2012) / 0.973 / 0.157 / 35.890 / 1 / <0.001
Transect length (s) / 0.672 / 0.050 / 47.181 / 1 / <0.001
Squares with flowers / 0.514 / 0.076 / 143.02 / 1 / <0.001
Transect start hour / 0.458 / 1 / 0.498
Mean temperature / 0.001 / 1 / 0.992
Mean cloud cover / 0.013 / 1 / 0.910
b / Estimate / se / c2 / df / P
Intercept / -2.218 / 0.320
Round (2) / 0.524 / 0.176 / 8.954 / 2 / 0.011
Round (3) / 0.454 / 0.193
Plot habitat (fallow) / -1.473 / 0.365 / 53.311 / 6 / <0.001
Plot habitat (grass) / 0.569 / 0.230
Plot habitat (meadow) / -0.206 / 0.282
Plot habitat (pollen and nectar) / -0.568 / 0.236
Plot habitat (wild bird cover) / -1.511 / 0.317
Plot habitat (floristically enhanced margin) / 0.408 / 0.062
Year (2012) / -0.728 / 0.140 / 26.110 / 1 / <0.001
Transect length (s) / 0.146 / 0.041 / 12.026 / 1 / <0.001
Squares with flowers / 0.408 / 0.062 / 34.788 / 1 / <0.001
Transect start hour / 1.345 / 1 / 0.246
Mean temperature / 0.631 / 1 / 0.427
Mean cloud cover / 1.862 / 1 / 0.172

Results

One hundred and sixteen transects were walked on 29 trial and 35 control plots during 2011, and 118 transects were walked on 29 trial and 26 control plots during 2012. Within the turtle dove plots, sown species constituted between 65% and 85% of total flower availability across all survey rounds (Fig. 1). Plot habitat influenced pollinator abundance in both models, with turtle dove plots supporting a higher abundance of foraging bumblebees and butterflies than some control habitats (Tables 2a & 2b). Turtle dove plots supported more foraging bumblebees and butterflies than fallow plots, nectar flower mixtures and wild bird cover (Tables 2a & 2b), although fewer bumblebees than floristically enhanced margins (Table 2a) and fewer butterflies than grass margins (Table 2b.