Latest Cat Predation Survey in Full

Written for The Mammal Society by MICHAEL WOODS, ROBBIE A. MCDONALD and STEPHEN HARRIS

A questionnaire survey of the numbers of animals brought home by domestic cats Felis catus was conducted between 1st April and 31st August 1997. A total of 14370 prey items were brought home by 986 cats living in 618 households. Mammals made up 69% of the items, birds 24%, amphibians 4%, reptiles 1%, fish <1%, invertebrates 1% and unidentified items 1%. A minimum of 44 species of wild bird, 20 species of wild mammal, 3 species of reptile and 3 species of amphibian were recorded.
Of a sample of 696 individual cats, 634 (91%) brought home at least one item and the back-transformed mean number of items brought home by was 11.3 (95% CI 10.4-12.2). The back-transformed means and number of cats retrieving at least one item from each prey group were: 8.1 (7.4-8.9) mammals for 547 (79%) cats, 4.1 (3.8-4.5) birds for 506 (73%) cats, 2.6 (1.8-2.7) herpetofauna for 145 (21%) cats and 2.2 (1.8-2.7) other items for 98 (14%) cats.
The number of birds and herpetofauna brought home per cat was significantly lower in households that provided food for birds. The number of bird species brought home was greater in households providing bird food. The number of birds and herpetofauna brought home per cat was negatively related to the age and condition of the cat. The number of mammals brought home per cat was significantly lower when cats were equipped with bells and when they were kept indoors at night. The number of herpetofauna brought home was significantly greater when cats were kept in at night.
Based on the proportion of cats bringing home at least one prey item and the back-transformed means, a British population of approximately 9 million cats was estimated to have brought home in the order of 92 (85-100) million prey items in the period of this survey, including 57 (52-63) million mammals, 27 (25-29) million birds and 5 (4-6) million reptiles and amphibians.
An experimental approach should be taken to investigate the factors found by this descriptive survey to influence the numbers of prey brought home by cats. In particular, investigation of potential management practices that could reduce the numbers of wild animals killed and brought home by cats will be useful for wildlife conservation, particularly in suburban areas.

Introduction

Domestic cats Felis catus are the most abundant carnivores in Great Britain and their numbers appear to be growing. In 1981, the national population of cats was estimated to be 6 million (UFAW, 1981). In 1993, the Cats Protection League estimated that approximately 25% of British households owned at least one cat and that the national population was approximately 7.6 million. This was predicted to increase to 8 million by 2000 (Cats Protection League, 1993). A more recent estimate supported this predicted increase and estimated that there were 7.8 million cats in 1998 (Turner & Bateson 2000). In addition, a minimum of 813 000 cats are estimated to live in a feral or semi-wild state in rural areas and a further unknown number of cats have only loose associations with domestic households in urban areas (Harris et al., 1995). It therefore seems likely that the total cat population in Britain in 2003 is in the region of nine million. By comparison to native carnivores, this is nearly 20 times the estimated pre-breeding populations of stoats and weasels and 38 times the estimated pre-breeding population of foxes Vulpes vulpes (Harris et al., 1995).

Most domestic cats depend on food supplied by their owners. Therefore, their populations are not limited by the availability of wild prey. Cats frequently kill wild animals and the combined impact of predation by millions of cats may have a substantial effect on wildlife. This prospect has previously been highlighted by several authors (Churcher & Lawton, 1987; May, 1988; Barratt, 1997, 1998). The recent rapid declines seen in British populations of many farmland and garden birds (Mead 2000), and the increasing importance of gardens for a range of small birds (Mead 2000), has led to renewed concern over the potential impact of cat predation on bird populations. However, experimental evaluations of the impact of cat predation on wildlife remain scarce and assessments of factors that might mitigate any such impact often lead to public controversy, arising from concern about the welfare of cats (May, 1988; Proulx, 1988; Fitzgerald, 1990; Jarvis, 1990).

Churcher & Lawton (1987) calculated that in a single English village, cats were responsible for up to 30% of mortality in a house sparrow population and concluded that domestic cats were a major predator in a typical English village. They found that the average cat caught and brought home approximately 14 prey items over the 12 months of their survey. May (1988) extrapolated from this figure and suggested that about 100 million wild birds and small mammals could be killed by 6 million cats every year in Britain. Mead (1982) ascribed 31% of recoveries of ringed robins and dunnocks to cat predation, but believed that there was no evidence that cats affected the overall populations of these species. Sharing this view, Fitzgerald (1988) and Fitzgerald & Turner (2000) asserted that on continental landmasses, wildlife had co-evolved with cats for hundreds of generations and that any species that were susceptible to predation would be "long extinct". It has also been argued that in many situations cats may limit populations of other predators, such as rats Rattus spp. that could have a more pronounced effect on wildlife (Fitzgerald, 1990; Fitzgerald, Karl & Veitch, 1991). However, while cats may eat Rattus spp., larger individuals of one of the most damaging rat species Rattus norvegicus are often avoided since they are hard to handle (Childs, 1986). Most authors do agree, however, that wildlife on oceanic islands is likely to be particularly susceptible to the impact of predation by cats. A particularly well-known example is that of the Stephens Island wren Xenicus lyalli, the entire population of which was eliminated by a single cat belonging to the island's lighthouse keeper (Oliver, 1955). Likewise, the Socorro dove Zenaida graysoni has also been driven to extinction primarily by cats (Jehl & Parks, 1983). Similarly, mammalian species such as hutias Geocapromys spp. have been exterminated from several Caribbean islands as the result of predation by cats (Fitzgerald, 1988).

Previous investigators of predation of wildlife by cats have frequently taken advantage of their habit of bringing back their prey to their owners' houses. Owners have been asked to record the species caught and the date it was retrieved or have retained the prey animal and given it to investigators for identification (Borkenhagen, 1978; Howes, 1982; Churcher & Lawton, 1987; Barratt, 1997, 1998). Similarly, but working at a smaller scale, Carss (1995) recorded the captures of his own two domestic cats over two periods of two years. There are several qualifications for the use of this method to describe the killing behaviour of cats, as opposed to alternatives such as analysing faeces or gut contents (reviewed by Fitzgerald, 1988; Fitzgerald & Turner 2000). Qualifications include the following: The prey brought home by cats should be representative of the total and variety of prey they actually kill. Cats belonging to households participating in the survey should exhibit behaviour representative of the killing behaviour of cats in general. Cat owners should be competent and honest recorders of the items their cats bring home. Unfortunately, for logistical reasons, these qualifications have usually remained untested. Nevertheless, the participation of cat owners is the only logistically feasible method of investigating cat predation of a range of species at a large scale.

Since it seems unlikely that trends in cat ownership and numbers in Britain will reverse, non-invasive measures that could be taken to reduce the number of animals killed by cats, while avoiding public concern for cat welfare, could be useful from a conservation perspective. For this study, a large-scale descriptive survey was conducted of the animals brought home by domestic cats living in Britain. We investigated factors influencing predation rates and which might be the focus of experimental work to evaluate strategies for minimising the numbers of wild animals killed by cats. The scientific names for prey species mentioned in this paper are provided in Appendix 1 and follow Arnold & Burton (1978), Chinery (1989), Corbet & Harris (1991) and Cramp (1994).
Methods

The methods applied during this survey were based on the work of Churcher & Lawton (1987) and Barratt (1997, 1998). A survey was conducted by The Mammal Society between 1st April and 31st August 1997. Participants were recruited through Society membership and appeals through the national media, including radio, newspapers and magazines. Forms were sent out to approximately 1400 households. Cat owners were asked to record in as much detail as possible the prey items brought home by each cat in their ownership. Where several cats lived in the same house, it was not always possible to separate the items brought home by individual cats. In these cases, records were included in our basic description of prey items, but were omitted from statistical analysis, which was based on households that could assign all retrieved items to individual cats. In common with previous investigators, we have assumed that the numbers of prey brought home by the cats was closely related to the numbers and variety of prey they actually kill. We have also assumed that the cat owners participating in the survey are representative, competent and honest recorders of the items their cats brought home. The validity and consequences of these assumptions are discussed below.

The following information was recorded by each household: the total number of cats in the household, whether the householders provided food for birds, the county of residence and the residence type. Counties were classified into five regions of England (North, Midlands, South East, South West and East Anglia), Scotland and Wales. Because of small sample sizes, households from Ireland and the Channel Islands were excluded from statistical analysis. Residence types were; detached house (a multi-storey house surrounded by garden), semi-detached house (a house attached to one other and with garden on three sizes), terraced house (predominantly urban housing that is adjoined on both sides and with usually small garden to front and rear only), bungalow (a single storey detached house with surrounding garden) and flat. For each cat the owners recorded: age, sex, condition (on a qualitative scale of 1 = thin to 3 = fat), whether the cat wore a bell and whether it was allowed out at night or kept indoors.

Prey items were grouped by taxonomic Class and Order where possible. In some cases, householders were unable to identify the prey remains brought home by their cat at all and these items were omitted from statistical analysis and were included only in total counts of prey items brought home (Appendix 1). Statistical analyses were conducted first on the numbers of birds, mammals and herpetofauna (reptiles and amphibians) and then on the numbers of bird and mammal species brought home by each cat. Counts of prey and prey species were log10 (n +1) transformed and cat ages were square root (n + 0.375) transformed to improve their approximation to normality (Zar, 1996). Four analyses of covariance (ANCOVA) were conducted. The first investigated the factors associated with the household that influenced the numbers of animals brought home by cats to the house. Region, residence type and the provision of bird food were factors and the total number of cats in the household was a covariate. The second ANCOVA investigated the factors associated with individual cats that influenced the numbers of animals they brought home. This analysis considered only the cats that lived in households with no other cats. This was in order to avoid possible pseudoreplication (Hurlbert, 1984) arising from similar conditions being applied to several cats living in one household. Sex, allowing the cat out at night and equipping the cat with a bell were factors and the age and condition of the cat were covariates. Both of these analyses included cats that did not bring home any animals, i.e. zero counts. These analyses were then repeated, but with the number of bird and mammal species brought home as the dependent variables and the total number of birds and mammals brought home as an additional covariate, to control for the effect of predation rate on species range. In this case, since the identification of species clearly required at least one animal to have been brought home, these analyses included only cats that had brought home at least one bird or mammal. Few herpetofauna were identified to species and the range of species was small, hence analysis of the number of species was not undertaken. Interaction terms were initially included in both analyses but were removed if found not to be significant and the analyses were rerun. No statistical analysis was made of other prey groups. Residuals from the ANCOVAs were checked for normality. Adjusted group means, corrected for covariates, of significant factors from ANCOVAs were compared using the Bryant-Paulson-Tukey test with Kramer's modification for unequal sample sizes (Bryant & Paulson, 1976; Day & Quinn, 1989).

To estimate the order of magnitude of the total numbers of animals that may be brought home by cats in Britain, we first estimated the population of cats likely to bring home prey of each group. This was achieved by multiplying the estimated cat population, here taken to be 9 million, by the proportion of this sample that brought home at least one of each prey type (Table 3). Then, for all the cats that did bring home at least one of each prey category, we calculated the mean and 95% confidence intervals of the log-transformed data. The number of prey animals brought home by all cats in Britain was then estimated by multiplying the estimated population of cats that retrieved that prey type by the back-transformed means and the associated, asymmetric 95% confidence limits. In order to be sure of the confidence intervals for this estimate, it was derived from the records from households where all prey items were assigned to individual cats, rather than the returns from households where the kills of several cats could not be distinguished.