Recommendations for depletion modelling of granivorous birds
|
|  Yüklə 6.33 Mb.
|
6.3.4 SeedsThere is an extensive literature on seed banks and seed rain, however, very little in relation to food availability for seed-eating farmland birds. Seeds on plants, seeds on the soil surface and seeds buried within the soil can be measured, depending on the foraging behaviour of the species being considered. Seeds on plants can be estimated from numbers of flowers, or by relationships to plant weight (Wilson et al., 1988). Diaz & Telleria (1994) measured seed availability by cutting all seed bearing plants within a 20 cm x 20 cm quadrat, and counting all seeds and fruit in the laboratory. Sample points were at 18 m intervals along a transect. Two contiguous soil samples 3.8 cm x 3.8 cm, 1 cm deep, were also taken 15 cm away from the quadrat in the transect direction, and seeds extracted and counted. Seeds were identified to species level and average seed composition and energy value determined from the literature. Average energy abundance per unit area was then calculated and expressed as kJ 10 ha-1. Pascual et al. (1999a) and McKay et al. (1999) estimated the density of exposed seed available to woodpigeons Columba palumbus on winter cereal fields by counting the number of visible seeds in 0.25m2 quadrats. Twenty sampling points were chosen at random in a diagonal line across the main field and 10 points along each of the two headlands walked in a zigzag pattern. In the RSPB Skylark Alauda arvensis Project (Donald et al., 2001) ten soil samples per field were taken and analysed to provide a measure of seed density within the top 5 mm of the soil surface. Similarly, in a project to measure the indirect effects of pesticides on farmland birds, Hart et al. (2002) estimated availability of seed to skylarks and yellowhammers by taking 5 mm deep, 20 cm x 20 cm soil scrapes. Five soil scrapes were taken at each sampling point and the number of seeds counted. Two sampling points in the centre and two at the edge of each field were used to make an estimate of seed density. Robinson & Sutherland (1997) quantified seed density and related it to density of skylarks. In each field, 10 samples were taken at randomly selected positions. For each sample, eight 5 cm soil cores were taken to a depth of 3 mm at random positions within a 50 cm x 50 cm quadrat. Seeds were then separated and identified to species level with the aid of a binocular microscope. Draycott et al., (1998) measured weed seed and waste cereal availability to pheasants, in arable fields on 16 farms in southern and eastern England. Sampling involved randomly throwing a 0.25m2 quadrat within 20 m of the field boundary and scraping the top 1 cm soil and seed bearing vegetation into a plastic bag (after Klute et al., 1997). Wakeham-Dawson & Aebischer (1998) measured availability of weed seeds to foraging skylarks by placing two 0.25 m2 quadrats 50 m from the field margin and 25 m apart in each field, and sucking up the seeds in two passes of a Halfords 12 V car vacuum cleaner, pressing the head down to the ground to allow all seeds to be collected. Robinson & Sutherland (1999) measured seed density and related this to habitat preferences of seed-eating birds. Seed density in the upper layer of soil was estimated by taking samples at ten random positions in each field (in the last two weeks of November and the last two weeks of March). Each sample consisted of eight individual cores (each 5 cm in diameter) 6 mm in depth bulked to give a total area of 0.016 m2, seeds extracted by washing through a stack of sieves, and hand-counted. Marshall & Vickery (2000) used a ‘cyclone sampler’ to take surface soil samples to measure seed availability on stubble fields. Seeds were subsequently extracted from soil samples by eye. Table 6.3 gives the frequency by which the different methods have been used in the studies on farmland birds reviewed above. In summary, a wide variety of techniques have been used to estimate the abundance of farmland bird foods. The choice of method has mainly been based on knowledge of the diet and foraging behaviour of the species concerned, but also on time available, practicality and cost. For each method that has been used, the sampling regime also varies widely between studies. Few studies justify the sampling regime used, and few provide data (e.g. the distribution and number of sampling points per field, depth of soil core taken, etc.). Some studies do not describe it adequately. Table 6.3 The frequency with which methods for estimating the abundance of bird food resources have been used in the studies reviewed in this section.
Most studies comparing sampling methods or devices have tended to concentrate on their relative efficiencies (e.g. Stewart & Wright, 1995) rather than on their repeatability. Also, the repeatability of destructive methods, such as the various vacuum sampling methods for seeds, will be difficult to separate from spatial variability without undertaking carefully designed experiments. Perhaps for this reason, few studies on terrestrial arthropods have attempted to investigate repeatability of the methods used to measure food availability. The seed bank is highly variable in space and time, in addition to any variability due to the method used. Nevertheless, a few studies give an indication of the repeatability of the methods which have been used. Marshall & Vickery (2000) investigated the efficiencies of four different methods of seed sampling: visual counts, brush sweeping (using a stiff plastic brush and dustpan), cyclone-sampling (using a prototype petrol-driven vortex vacuum sampler) and Aquavac sampling (using a car battery operated vacuum sampler). Seeds were sampled on wet soils in December and in a designed experiment in which known densities of rape and grass seeds were sown on dry soil in March. Overall, the cyclone sampler gave the best seed recoveries, while sweep sampling consistently underestimated seeds. Cyclone samples varied from 15 to 100 seeds, equivalent to 167 to 1111 seeds m-2. This extent of variation suggests that, at first sight, the repeatability of the cyclone technique does not appear to be very satisfactory. In addition, the process is time-consuming, as seeds must be separated from soil in the samples back in the laboratory. The efficiency, and hence repeatability, by which seeds visible on the soil surface are counted is likely to be high and dependent on the care and time taken by the observer. However, a good estimation of the mean density of exposed seeds on fields will depend on the spatial variability and the number and size of quadrats per field. For example, Pascual et al. (1999a), using 20 sampling points and 0.25 m2 quadrats, estimated means ranging from 2.9 to 10.58 wheat seeds per quadrat and standard errors around 2. That is, 95% of samples were within about 4 seeds of the mean. Holland et al. (2002) measured total invertebrate food availability for farmland birds. Using the D-Vac method, they estimated Chick Food Index (CFI) for grey partridge and chick food availability (CF) for yellowhammers at study sites in Hampshire and Lincolnshire. They estimated variability between and within fields and found significant differences between crops and between the edge and centre of fields. For a mean total number of insects of 1.72 for the edge and 1.63 for the centre of fields, the pooled standard error was 0.02, and for a mean CFI ranging from 0.07 for spring barley to 0.14 for winter wheat, the pooled standard error was 0.05. These data suggest that 95% of D-vac samples will fall within about 0.1 of means ranging from about 0.1 to 2. That is, repeatability of the D-vac method appears satisfactory.
Only items known to be present in the diet of the bird species being studied should be measured when estimating food availability. Diet composition can be investigated by observation of foraging birds (e.g. corn buntings: Gillings & Watts, 1997), from crop or gizzard contents of some species (e.g. woodpigeons: McKay et al. 1999), faecal analysis (e.g. yellowhammers: Moreby & Stoate 2001), and bolus analysis of neck-ringed chicks (e.g. corn buntings: Gliemann, 1973). For invertebrates, methods appropriate to the species in the diet should be selected, e.g. sweep sampling for mobile species or the D-Vac method for surface active species. For example, Holland et al. (2002) used the D-Vac technique to estimate general invertebrate food availability for farmland birds and availability specific to grey partridge and yellowhammers from knowledge of their diet. If invertebrate availability, as opposed to abundance, is to be measured, samples should be taken at the time of year, time of day, weather conditions and in the habitat and microhabitats in which the birds are known to forage. It should be appreciated that methods for assessing diet composition also have associated problems. Larger items in the diet are more easily seen and identified than smaller items in observation studies, some items may not be retained in the crop in crop content studies, and some may be completely digested and not be detectable in faecal analyses. All three methods, therefore, are usually qualitative rather than quantitative, without very careful interpretation of the data. Choice of sampling site is also crucial if availability rather than overall abundance is to be measured. For example, Brickle et al. (2000) investigated habitat use by corn buntings collecting food for nestlings by comparing the proportion of foraging visits to each habitat with the proportional availability of habitats around the nest. However, they then sampled chick food in every discrete habitat block within the study area using sweep netting. In contrast, Green (1984) sampled the abundance of grey and red-legged partridge chick foods during the main hatching periods, within the foraging areas of the birds (which were radio-tracked) and Hill (1985) sampled pheasant chick insect food within the home range of radio-tracked broods. It is not clear how seed is available to seed-feeding birds. Seed shed onto the surface is available to ground feeders, but if seed is covered by soil, is it unavailable? The seeds of many weed species are very small and possibly not actively sought. Availability of different seed sources is likely to be bird species-specific (Marshall & Vickery 2000). The timing of seed sampling is also crucial if availability rather than abundance is to be measured. Seed density must be measured at the time of year, and in the particular habitat, that birds forage. In general, seed not visible on the soil surface is assumed to be unavailable to most seed-feeding birds. Marshall & Vickery (2000) made this assumption and so only estimated seeds on plants and on the soil surface. Similarly, McKay et al. (1999) and Pascual et al. (1999a) estimated the density of winter cereal seed visible on the surface of newly drilled fields. Pascual et al. (1999a and 1999b) found that agricultural methods which resulted in more seed being buried, such as sowing seed deeper, and rolling and harrowing, reduced the availability of seed to woodpigeons in pen trials. If it is true that most birds do not dig for buried seed, then the methods which rely on removing the top layer of soil (which includes vacuum sampling techniques under some conditions) may be overestimating the availability of seeds to birds.
Farmland birds can be separated into guilds based on diet and foraging behaviour. Both diet and foraging behaviour are reviewed in Buxton, Crocker & Pascual (1998), the CSL ‘bird bible’. This information was used to separate the species considered in this report into six categories (Table 6.4). Species feeding only on the wing and only by probing the soil, are not considered in this project and are excluded from the table. For each category, a method for estimating availability was selected, based as far as possible on information on repeatability and efficiency. Because there are few studies comparing different methods, and assessing their efficiencies and repeatability, this should be regarded as an informed suggestion based on current knowledge. A considerable amount of further work is required before a definitive list of standard methods can be put forward.
The standard techniques for measuring the abundance of plants and arthropods are listed, and summaries given of their known efficiencies and limitations. The techniques which have been used in studies of farmland bird food resources are reviewed with a view towards standardisation. A wide variety of methods are found to have been used, governed mainly by the species of bird being considered and its foraging behaviour and habitat. Other considerations include cost, staff resources available and time. The methods vary in terms of their efficiency and probably repeatability, though few data are available on the latter. Also, different methods are affected in different ways by external factors such as vegetation structure and climatic conditions. It is therefore very difficult to compare the results of different studies, and to combine their data when constructing models. Generally, methods measure relative rather than absolute availability (Southwood, 1978) and this becomes important when attempting to compare the amount of food available to the amount of food needed to survive or reproduce (see Objective 1). More work is needed, therefore, on the efficiencies of the different methods in a range of different habitats, and hence on the calibration coefficients needed to convert relative to absolute abundance for each device used. Farmland bird species can be divided into guilds based on their diet and foraging behaviour. When this was done for the species being considered in this report, we were able to suggest standard methods for assessing food availability. The method selected for each guild was based on the frequency by which the various methods have been used in previous studies, as well as efficiency and any data on repeatability. Cost and ease of use have also been taken into consideration, although these are regarded as being of secondary importance.
Further work is needed on the repeatability and efficiency of most methods reviewed in this Section, in particular the D-Vac method and sweep netting. The D-Vac device, despite being the most popular method for sampling surface arthropods, is expensive, heavy, cumbersome to use and has safety implications (the operator walks about with a heavy, noisy petrol-driven motor strapped to his/her back). Further work is therefore needed to improve upon its design, while still maintaining a comparable efficiency of extraction. Table 6.4 Categories of farmland birds based on diet and foraging behaviour, used to select standard methods for the measurement of food availability. Species considered in the present study are placed into these categories based on their winter and spring/summer diets. w= winter, s=spring/summer
|