The aim of this study was to design and test a novel bait hopper equipped with a PIT-tag reader and bait weighing device, to record bait uptake by individual grey squirrels for optimising the delivery of a contraceptive bait. The hopper was tested in a laboratory, in a trial using captive grey squirrels and then used to collect data on the feeding behaviour of free-living grey squirrels in two woods.

Captive trial to test hopper capacity to record feeding visits by squirrels

A trial was conducted using two outdoor pens (width = 2.7 metres, length = 9.7 metres, height = 2.4 metres), each containing one male and two female grey squirrels previously fitted with a PIT-tag (Identichip^®, York, UK) subcutaneously in the scruff of the neck. Throughout the trial, the squirrels were provided with a varied diet including maize, peanuts, bird seed and fruit, along with environmental enrichment including wool bedding, foliage, tubes, ropes, nest boxes, branches and small sticks. Two hoppers were placed in each pen and each hopper was positioned on top of a wooden stand (approximately 90 cm high) so that they were visible by closed-circuit television (CCTV) cameras. During the trial, one of the hoppers in pen 2 became obscured by a branch so was not included in the study. For four consecutive days per week, for two weeks, 10 g of fresh hazelnut paste was provided daily in each hopper. Feeding visits to the hoppers by individual squirrels were recorded using CCTV at peak times of feeding activity (4:00 to 8:00 and 16:00 to 20:00) in the second week of the trial. For the CCTV analysis, when a squirrel entered a hopper, the visit was assigned to one of the following categories: ‘feeding’ (when it entered the hopper and was subsequently observed masticating), ‘full’ (only the tail was visible and no evidence of feeding) or ‘partial’ (part of its hindquarters were still visible). Only visits that could be definitively assigned to one of these categories were used in the analysis. The accuracy with which the hoppers recorded feeding visits was determined by checking whether a PIT-tag was detected for each feeding, full or partial visit recorded on CCTV and whether the ID of the squirrel was established. A Fisher’s exact test was used to determine whether PIT-tag IDs were more likely to be recorded for feeding visits over non-feeding visits.

Field trial to test hopper capacity to record feeding visits by squirrels

A field trial was conducted in December 2018 in two 8 ha woods located in Yorkshire, UK. Single-catch squirrel cage traps (n=24) were deployed in each wood on one metre high wooden stands and pre-baited with peanuts and whole hazelnuts every two to three days for one week. Grey squirrels were trapped the following week over consecutive days. Traps were set early in the morning and checked in the afternoon. Trapped squirrels were anaesthetised using isoflurane via a mask, sexed and a PIT-tag (Identichip^®, York, UK) implanted subcutaneously in the scruff of the neck. Hair was clipped from the end of the tail for visual identification. Once recovered from anaesthesia, squirrels were released under a Natural England licence in the location at which they had been trapped.

Trapping was continued until it was estimated that the majority of the population had been PIT-tagged, based on the ratio of new individuals to recaptured (tail clipped) individuals per day. In both woods, within three weeks of PIT-tagging, each trap was replaced by a hopper, fixed to the wooden stand. To simulate the deployment of a contraceptive bait, each hopper was pre-baited with 40 g of hazelnut paste every two to three days for one week. The following week, each hopper was baited daily with 40 g of hazelnut paste for four consecutive trial days. The hopper trials for both woods were conducted consecutively, using the same hoppers.

All analyses were carried out in R (R Core Team, 2019) using the MASS package (Venables & Ripley, 2002). A linear model (assuming a Gaussian distribution) was used to test whether there was a significant relationship between the number of feeding visits (or PIT-tag records) per individual and the sex of individuals or the wood where they were trapped. Significance was tested using type II analysis of variance. A generalised linear model (GLM, assuming a negative binomial distribution) was used to test whether there was a significant relationship between the number of hoppers visited per individual and the sex of individuals or the wood where they were trapped. Significance was tested using type II analysis of variance with a likelihood ratio test (LRT). Diagnostic checks of residual plots for both models showed that the residuals were approximately normally distributed and model assumptions were met.

Laboratory trial and captive trial to test hopper capacity to measure bait uptake by individual squirrels

A trial was designed to test the accuracy of the built-in weighing scales in each hopper, using manually weighed baits. To measure the accuracy of bait weight taken per visit, the first part of the trial was conducted in a laboratory with nine hoppers. Approximately 70 g of 100% hazelnut paste was weighed in a bait tray and placed in each hopper. To simulate different field conditions, five hoppers were placed in a refrigerator and left for one hour to acclimatise to between 6^oc and 8^oc; the remaining four hoppers were left at room temperature, between 20^oc and 21^oc. A small amount (0.1 to 2.2 g) of paste was then removed from each hopper using a pre-weighed metal spoon, and the spoon and paste weighed (to the nearest 0.1 g). After at least 10 minutes, a larger amount of paste (4.9 to 18.5 g) was removed from each hopper and weighed (to the nearest 0.1 g). This was repeated until there were are at least 5 weights for small amounts of bait and at least 5 weights for larger amounts of bait for each hopper. To ensure the strain gauge was stable, the amount of bait taken from each hopper was calculated from the weight recorded by each hopper 2 minutes prior to the bait removal minus the weight taken 2 minutes after the removal. These were compared with weights obtained manually using a standard balance.

To test whether it was possible for the hoppers to weigh bait consumed by individual squirrels, a second trial was conducted using captive PIT-tagged grey squirrels. Two hoppers per pen were deployed in three outdoor pens containing 2-3 squirrels per pen, 7 squirrels in total. Hoppers were placed on the floor along each side of a pen, weighed down by bricks, to ensure the squirrels could not move or overturn them. In week 1, the hoppers were baited on Monday, Wednesday and Friday with approximately 40 g 100% hazelnut paste to encourage the squirrels to use the hoppers. On the Tuesday and Wednesday of week 2, 20 + 0.5 g of hazelnut paste was weighed into bait trays and installed in each hopper at 7:15 am, immediately before the squirrel peak feeding time. After 6 hours, the bait trays were removed and the remaining bait weighed (to the nearest 0.1 g). The manual weights taken for each trial period were compared to those recorded by the hoppers. The weight of bait taken for each event was calculated by taking the minimum weight from the most stable values recorded; those where there was less than 5 units difference between the first and fifth repeat reading. The weight decrease at each event was then matched with a PIT-tag record, if available, to calculate the amount of bait taken per visit by individual squirrels.