Appendix 10
Objective Four: To develop a mechanistic model linking demographic mechanisms to resource provision
A mechanistic model for the persistence of granivorous farmland bird populations through winter with respect to food availability
Introduction
Increased annual mortality has been identified as the key demographic mechanism behind the declines of many farmland bird species (Siriwardena et al. 2000). In particular, population declines have been linked to a reduction in winter food availability due to the loss of seed-rich habitats such as over-wintered stubble from the agricultural landscape. This is the preferred winter foraging habitat for many granivorous species (Wilson et al. 1996; Moorcroft et al. 2002) but a number of factors, such as the switch from spring to autumn sowing, increased herbicide application and improved harvesting efficiency have led to a reduction in the quantity and quality of this key resource (Newton 2004).
Recent studies have shown that increasing the availability of winter food supplies can increase survival and local abundance (Peach et al. 2001) and, to some extent, can influence breeding population trends (Siriwardena et al. 2007). Along with options designed to provide nest sites and summer food, a number of potentially beneficial management techniques for delivering winter food resources have been included as options within agri-environment schemes such as Entry Level and Higher Level Stewardship (ELS and HLS respectively) (Defra RDS 2005a, b). These schemes are the main policy tool by which the UK Government will deliver its Public Service Agreement target to reverse the long-term decline in farmland bird populations by 2020. However, there has been considerable debate over how best to encapsulate the benefits of resource provision within effective and feasible management options as part of a national scheme. This is particularly true for delivering winter food resources, where much of the past success of this conservation approach has been achieved with the provision of ad libitum food resources (Siriwardena et al. 2007) or with species of restricted range (Peach et al. 2001). Indeed, key questions remain over the necessary levels of resource provision and the most appropriate management prescriptions to deliver those resources.
Gillings et al. (2005) investigated the relationship between stubble field availability and breeding population trends of farmland birds within 1km squares. They showed that, for a number of farmland specialists, the availability of winter cereal stubble positively influences breeding bird trends. Furthermore, they identified a threshold availability of 15-20ha cereal stubble associated with approximately stable breeding populations of skylark Alauda arvensis and yellowhammer Emberiza citrinella (Gillings et al. 2005). However, this study focused on the average area of cereal stubble available over a three-year period and did not account for temporal availability within winters. The temporal dynamics of resource availability is likely to be a key determinant of over-winter survival (Siriwardena et al. 2008); food availability needs to match energetic requirements throughout the winter if individuals are to survive.
Despite the importance of the link between resource availability, its exploitation and population trends in agricultural landscapes, surprisingly little is known about these ecological processes in wild farmland bird populations. Spatial depletion modelling, based on optimal foraging theory, provides a method for understanding the mechanistic links between land-use, food resources and bird distribution patterns (Stephens et al. 2003). In the approach’s simplest form, depletion of resources caused by the foraging population is assumed to drive changes in predator distribution (Sutherland & Anderson 1993) but both resource input (Sutherland & Allport 1994) and other causes of depletion (Percival et al. 1996) during the modelling period can also be incorporated. Importantly, this approach allows the integration of a spatially explicit mechanistic model with field data that describes the wider landscape within which farming activities occur. Bird populations function ecologically at relatively broad spatial scales and there is evidence that the wider landscape context may be very important in understanding spatial patterns of population decline (Gillings et al. 2005; Robinson et al. 2001).
Here we use a mechanistic seed depletion model to investigate the impacts of stubble field dynamics on the population persistence of seed-eaters within the agricultural landscape. Stubble fields are considered as ephemeral food patches whose availability is dictated by the timing of harvest and the resumption of cultivation, and whose quality (in terms of food resources) is dictated by the crop type grown and associated management. We investigate how management characteristics impact on resource provision within stubble fields and explore how much stubble needs to be introduced into an agricultural landscape before levels of population persistence over winter match those likely to deliver stable breeding population trends.
Methods
Model structure
The model arable landscape consists of 100 patches, each representing a field, in a 10x10 array (2D). Five types of stubble, linseed, oilseed rape, sugar beet, barley and wheat, are included in the model. These were the predominant stubble types in the Breckland area of Norfolk, on which the development of this model was based (data largely collected under Defra funded project BD1610). Management characteristics for each field, including harvest date, lifespan, the number of herbicides applied to the preceding crop and whether or not the crop was sprayed with pre-harvest desiccant, are included as input into the model. Stubble lifespan represents the number of days between harvest and the field being ploughed and reflects the length of time each field is available as foraging habitat. Outside this time period it is assumed that fields effectively have a resource density of zero and are not suitable foraging habitat for seed-eaters. The model runs for a non-breeding period of 224 days, corresponding with the time between the harvesting of the first arable fields in mid-July (day 1 = 19th July) and the start of spring (day 224 = 28th February).
Food resource dynamics
The dynamics of four seed types, cereal seeds, oil seed rape seeds, linseeds and weed seeds (represented by Knotgrass Polygonum aviculare), within fields in the landscape were modelled. All are important potential food resources for farmland birds (Wilson et al. 1999). Seed dynamics were only simulated for the period after a field was harvested and before it re-entered cultivation.
Initial crop and weed seed densities available to seed-eating birds in each field immediately after harvest were allocated by randomly sampling with replacement from the distribution of values in the appropriate crop type recorded from a sample of fields in Breckland (Robinson 2003). Subsequent resource dynamics are driven by additional resource input, via weed seed rain, and resource depletion, modelled as avian predation, germination and other losses (including non-avian predation, incorporation in to the soil and decomposition).
Seed rain: Analyses of weed seed rain data have shown that weed seed rain is largely independent of crop type and varies significantly in relation to chemical inputs and harvest date (Robinson 2003). These management components alter the temporal patterns of seed rain post-harvest; seed rain tends to peak 2-3 months after harvest, but this peak is substantially reduced or largely absent if fields have relatively high chemical inputs or are harvested relatively late in the season (Mattison 2007). Using parameter estimates from multiple regression models derived from these weed seed rain data (Table 10.1), resource inputs into each field were estimated from the number of chemicals applied, harvest date and whether or not the crop was sprayed before harvest. Separate models were generated for the seed rain data for each month post harvest with the exception of month five that lacked sufficient data. Parameter estimates from these multiple regression models were then used to estimate monthly seed rain for each field in our mechanistic model based on their management data. For month five, we used the multiple regression model for month four but devalued the resultant estimate by 15% since this reflected the difference in mean seed rain densities between months four and five (see below). Daily seed rain was simply estimated as the monthly estimate divided by the number of days per month.
Seed depletion: Seed depletion by birds is discussed below. Two further estimations of daily survival probability were made for each seed type. The first, relating to germination, was initially set at 0.94 for oilseeds, 0.999 for cereal seeds and 1 for weed seeds. These values reflect the limited dormancy of oilseeds relative to cereal seeds (Lutman 1993; Robinson 2003) and the assumption that weed seeds remain dormant for the period studied here. The second estimation of daily survival probability, relating to non-avian and non-germination losses, was initially set at 0.99 for cereal seeds and 0.95 for oilseed and weed seeds (Holmes 2002; Hulme 1994). Little is known about survival rates in relation to small mammal predation other than that survival rates are likely to be lower where seeds occur in high densities (Hulme 1994). Lower values were therefore assigned to weed seeds and oilseeds than cereal seeds as these were recorded in significantly higher densities within stubble fields (average seed density: weed seed = 1567 m-2, oil seed = 1061 m-2, cereal seed = 287 m-2, Robinson 2003). This gave composite daily survival probabilities (i.e. the probability of surviving both germination and other sources of seed losses) of 0.89, 0.95 and 0.989 for oilseeds, weed seeds and cereal seeds respectively. Since little is known of the processes driving seed dynamics in arable fields, these values were based on the limited literature available or expert opinion and we explored a range of probabilities about these values in a series of sensitivity analyses, increasing and decreasing these composite probabilities by 0.025.
Avian depletion: We assumed that the seed-eating bird community consisted of one ecotype, buntings, defined in the model on the basis of dietary preferences. Specifically, this ecotype was represented by yellowhammers, a species that has undergone a 50% population decline over the past few decades, probably as a consequence of reduced survival (Siriwardena et al. 2000), and is on the UK red list of high conservation concern (Gregory et al. 2004). We have focused on buntings, and yellowhammers in particular, rather than on skylarks (as originally proposed) for a number for reasons. First, colour ringing and radio-tracking undertaken as part of the fieldwork towards achieving Objectives 1 and 2 have focused primarily on yellowhammers. Results from this fieldwork have therefore been used to inform the correct spatial scale for our modelled landscape and outputs from a mechanistic model based on yellowhammers can be integrated with outputs from the wider project. Secondly, previous analyses have shown that the model is less capable of predicting the spatial and temporal distribution of skylarks based on stubble resources alone (Robinson 2003). Observed and predicted bird numbers were poorly correlated within crop types and model predictions showed consistent bias, suggesting that the model failed to incorporate an important ecological process that drives distribution patterns in the landscape.
In our model, yellowhammers were assumed to feed on the most profitable seed type in a field when more than one seed type were available, unless a threshold intake was reached at which it was optimal to forage unselectively (Norris & Johnstone 1998). Birds therefore had the choice between an exclusive diet of cereal seeds or a mixed diet of cereal and weed seeds on wheat and barley stubbles, an exclusive diet of oilseeds or a mixed diet of oilseeds and weed seeds on oilseed rape stubbles, an exclusive diet of linseeds or a mixed diet of linseeds and weed seeds on linseed stubbles and a diet of weed seeds only on sugar beet stubbles.
The daily energy requirements of a free-living yellowhammer were set at 103.25 kJ, calculated using allometric relationships (Walsberg 1983). It was assumed that birds attempt to maximise their rate of energy intake to survive. Energy intake rates for each field at each time step were estimated using Holling’s disc equation (Holling, 1959):
(1)
Where E/T = intake rate, e = energy content of one prey item (j), a = search efficiency (s m-2), d = prey density (m-2) and Th = handling time of each prey item (s). Seed mass (g) and energy content (J g-1) details were taken from the ECOFLORA database ( and Díaz (1990).
Handling times of each seed type included in our model were not directly available for yellowhammers. However, analysis of reported handling times based on type II functional responses for a range of passerines feeding on different seed types (Holmes 2002; Stephens et al. 2003) showed that measures of seed length, bird body mass and bill shape index (bill length/bill index) are good predictors of handling time (Mattison 2007). Handling times for rape seeds, linseeds, cereal seeds and weed seeds were calculated as 1.81s, 4.08s, 7.27s and 3.54s respectively. In our sensitivity analyses, we explored a range of handling times for all seed types by increasing each value by 10 and 20%.
Estimated search efficiencies of seed-eating passerines are highly variable, spanning four orders of magnitude and varying by an order of magnitude for single species (Holmes 2002; Stephens et al. 2003). We therefore investigated whether the substrate conditions during intake rate measurement (bare earth or vegetation), bird or seed traits (bill shape index, bird length and bird mass, seed length and seed mass) were related to estimates of search efficiency using linear regression analysis. Reported search efficiencies for passerines feeding on seeds (Holmes 2002; Stephens et al., 2003) were natural log transformed, as were the predictor variables (except for substrate condition which was treated as a factor). We found a statistically significant relationship between seed mass and search efficiency but predicted values based on the resulting equation were consistently overestimated. We therefore used the mean search efficiency of passerines feeding on seeds reported in Holmes (2002) and Stephens et al. (2003), 0.0702 m2s-1, initially to define yellowhammer search efficiency in our model and subsequently explored a range of values, reducing efficiency by one and two orders of magnitude, in our sensitivity analyses.
Potential energy intakes were calculated for each available field at each time step in the model and all birds were assumed to forage in the field where energy intake rate was maximised, or if more than one field met this criterion birds were evenly distributed between them. Seed depletion due to birds was calculated as the product of the rate of energy intake and the number of birds occupying the field. A field was considered unavailable if the maximum achievable energy intake rate was insufficient to meet the daily energy requirements of a single bird. The model simulates energy intake rates, bird distribution and seed depletion by birds on an hourly basis; seed rain and seed depletion due to other sources are simulated on a daily basis.
Two sources of data were used to validate the model and assess the impact of stubble availability and dynamics on yellowhammer population persistence over winter: To examine critically the model’s ability to predict yellowhammer distribution between stubble fields within an agricultural landscape, we used data on land-use and management and on arable weed and bird populations collected during the 2000/01 and 2001/02 winters from a 168 stubble fields on 22 farms located throughout and adjacent to Breckland in Norfolk, covering an approximate landscape area of 100,000ha (Robinson 2003). To explore the model’s ability to characterise landscape quality in terms of carrying capacity for yellowhammers, we used data from the Winter Farmland Bird Survey (WFBS) (Gillings 1999). WFBS involved three timed visits to 1090 1x1 km squares across lowland farmland areas of the UK to document winter abundance, distribution and habitat selection of farmland birds in three winters 1999/2000, 2000/2001 and 2002/2003. Our analyses are restricted to 601 of these squares which have also been surveyed as part of the Breeding Bird Survey (BBS) and for which both breeding and winter population data are therefore available. WFBS surveys took place between November and February and, for the purposes of the modelling described below, it was assumed that visit 1 (hereafter V1) took place on November 15th (day 120), visit 2 (hereafter V2) took place on 30th December (day 165) and visit 3 (hereafter V3) took place on 13th February (day 210).
Distribution validation
We constructed a model arable landscape representative of the actual Breckland landscape. Each field in the 10x10 array was assigned crop and management characteristics equivalent to those of surveyed fields, maintaining the relative proportions of stubble types observed. The average field size was 10.7ha and the total stubble area modelled was 1124.05 ha. The model was populated with the total number of yellowhammers counted in the real landscape on each two-weekly survey, i.e. the number of birds in the modelled landscape was re-set at two-weekly intervals to reflect the temporal variation in bird numbers recorded in the Breckland landscape (Mean count = 61 individuals, min count = 5 (early August), max count = 202 (early December); Robinson 2003). The model then tracked the distribution of these birds between fields and crop types by distributing the total number of birds known to be present in the landscape according to the rules described above.