Effective wildlife population management requires an understanding of the abundance of the target species. In the UK, the increase in numbers and range of the non-native invasive grey squirrel Sciurus carolinensis poses a substantial threat to the existence of the native red squirrel S. vulgaris, to tree health, and to the forestry industry. Reducing the number of grey squirrels is crucial to mitigate their impacts. 

Camera traps are increasingly used to estimate animal abundance, and methods have been developed that do not require the identification of individual animals. Most of these methods have been focussed on medium to large mammal species with large range sizes and may be unsuitable for measuring local abundances of smaller mammals that have variable detection rates and hard-to-measure movement behaviour.

The aim of this study was to develop a practical and cost-effective method, based on a camera trap index, that could be used by practitioners to estimate target densities of grey squirrels in woodlands to provide guidance on the numbers of traps or contraceptive feeders required for local grey squirrel control.

Camera traps were deployed in ten independent woods of between 6 and 28 ha in size. An index, calculated from the number of grey squirrel photographs recorded per camera per day had a strong linear relationship (R² = 0.90) with the densities of squirrels removed in trap and dispatch operations. From different time filters tested, a 5 minute filter was applied, where photographs of squirrels recorded on the same camera within 5 minutes of a previous photograph were not counted. There were no significant differences between the number of squirrel photographs per camera recorded by three different models of camera, increasing the method's practical application.

This study demonstrated that a camera index could be used to inform the number of feeders or traps required for grey squirrel management through culling or contraception. Results could be obtained within six days without requiring expensive equipment or a high level of technical input. This method can easily be adapted to other rodent or small mammal species, making it widely applicable to other wildlife management interventions.

Study sites

The study was conducted in 10 mature woods at the same time of year, between mid-June and mid-July, from 2018 to 2021 (Table 1). Woods were located in two regions of the UK; eight in Yorkshire, England (54°N, 0°W) and two in Denbighshire, Wales (53°N, -3°W). Woods were between 6 ha and 28 ha in area and consisted of either broadleaf or a mix of broadleaf and conifer trees.  The area of each wood was measured from a satellite base map using a measure tool (Google My Maps 2018 to 2021). During the study, each wood was sampled once. To ensure independence, woods sampled within consecutive years were not directly connected to each other via wooded corridors or hedgerows and were located at least 600 metres apart. The first seven woods, sampled in 2018 and 2019, were discrete areas of woodland with little connectivity to other woodland areas. The last three woods sampled were highly connected to other woodland areas.

Camera deployment

At each wood, camera traps (Reconyx™ HC500 or HS2X) were deployed at a density of 1/ha. Camera placement in the field was guided by a 1-ha grid generated in ArcGIS (version 10.7.1) overlayed onto a satellite map using the ArcGIS Collector mobile phone application and was adjusted according to accessibility; for example, steep slopes or thick vegetation were avoided (Figure 1a). 

Cameras were fixed to trees at approximately 1 meter above the ground and with the lens angled between horizontal and 45° below horizontal (Figure 1 b). A laser pen or 1-meter wooden pole, placed parallel to the base of the camera, was used to position a pile of bait at the centre of the camera field of view, between 1 and 2 meters away from the camera lens. The bait pile consisted of approximately 1.5 kg of 50:50 whole maize and peanuts. The cameras were set to take one photograph per trigger and the passive infrared sensor to high sensitivity. Cameras were deployed for 3–6 days and the bait in front of each camera was checked every 1–3 days (guided by a prior assessment of potential bait uptake by non-target species) and replenished, if required.

At the end of each deployment, the cameras were removed and all the photographs containing squirrels were digitally tagged using the Reconyx MapView Professional™ software. For the first five woods, photographs were also tagged with the number of squirrels present in each photograph. The resulting data were quality checked by a second observer re-analysing a sub-sample of the photographs to ensure there was no observer bias in the records. The final photographs taken by each camera in each wood were checked for the amount of bait remaining, as this is likely to affect squirrel activity and the numb

Camera index design and selection

Four camera indices were considered as candidates for estimating grey squirrel densities. All indices were based on the number of squirrel photographs per number of working cameras per trial day and were designed to be practical, cost-effective and representative of squirrel activity. Trial days consisted of consecutive 24 hours. The differences between indices concerned the time the first trial day began and which trial days were used for the photograph counts. Index 1 used all squirrel photographs recorded during consecutive 24 hours from the time the last camera was deployed in each wood. Index 2 used all squirrel photographs recorded during consecutive 24 hours, from 24 hours after the last camera was deployed; this was to allow the squirrels time to find the bait piles before the assessment began. Index 3 used all squirrel photographs recorded within consecutive 24 hours, from 24:00 on the day the cameras were deployed. Index 4 used all squirrel photographs from the 24 hours that recorded the maximum number of squirrel photographs from each consecutive 24 hours starting from when the last camera was deployed; this was to provide the maximum level of activity. 

For all four indices, time filters of 0.5, 1, 2, 3, 4, 5, 10, 20 and 30 minutes were applied, where any photographs that were recorded within the specified interval after the previous photograph were excluded from the photograph counts. The application of a time filter was used to moderate inflated counts caused by individuals that remain in front of a camera for extended periods of time. This is especially applicable at bait piles, where some individuals may feed for longer than others. Linear regressions were used to test whether the values calculated for each index could be used to predict the density of squirrels trapped and removed in each wood. The coefficient of determination (R²) was calculated as a measure of fit and the statistical significance of the model with the greatest R² was assessed using an F-test. Data normality was confirmed using a Jarque-Bera test and through plots of the residuals. To make the data processing methods more widely accessible to practitioners, all data analysis was conducted using Microsoft Excel®.

Photograph data were not analysed, and the number of cameras adjusted accordingly for days when a camera ceased to work due to insufficient battery power or faults, when the bait had been completely removed, or when the camera was not focussed on any part of the bait pile, due to set up error or if it was subsequently knocked out of position by a person or an animal.

Microsoft Excel