Detection dogs are increasingly used to locate cryptic wildlife species, but their use for amphibians is still rather underexplored. In the present paper we focus on the great crested newts (Triturus cristatus), a European species which is experiencing high conservation concerns across its range, and assess the ability of a trained detection dog to locate individuals during their terrestrial phase. More specifically, we used a series of randomised, double-blinded experiments to document whether a range of distances between target newts and the detection dog affects the ability of localisation, and to assess the ability and efficiency of target newt detection in simulated subterranean refugia through 20 cm of two common soil types (clay and sandy soil, both with and without air vents to mimic mammal burrows, a common refuge used by T. cristatus). The detection dog accurately located all individual T. cristatus across the entire range of tested distances (25 cm - 2 m). The substrate trials revealed that the detection dog could locate individuals also through soil. As expected, the detection time during the soil discrimination trials was significantly reduced for treatments with vents. Contrary to existing studies with detection dogs in human forensic contexts, however, detection was generally faster for T. cristatus under clay soil compared to sandy soil. Our study provides a general baseline for the use of detection dogs in locating T. cristatus and similar amphibian species during their terrestrial phase.   

Soil interference trials

Thirty T. cristatus (16 males and 14 females) were retrieved from a quarry in south Gloucestershire between July and September 2021.  All T. cristatus used for the experiments were individually weighed to the nearest 0.1 g and measured from the tip of the snout to the end of the cloaca to the nearest mm. Two soil types were used: sandy soil collected from Calne Without, Bowden Hill, Wiltshire, Southwest England (OS grid reference ST954684), and clay soil collected from Broad Lane, Westerleigh Road, Westerleigh, South Gloucestershire, Southwest England (OS grid reference: ST695795). For soil characterisation, three reference samples were obtained from each site using a 20 cm long core, and merged into one sealed plastic bag for analysis at the University of Salford. In order to measure the pH, 20 ml of fresh soil was placed into a clean 50ml beaker. Distilled water was added to the 40 ml mark, and the mixture was left on an automatic stirrer for 15 minutes. The probe of a calibrated pH meter was then placed into the sample and left for 30 seconds for the reading to settle. This was repeated three times for each soil type with the probe cleaned with distilled water between each reading.  To gauge moisture content for each soil type, the soil was weighed, oven dried overnight at 105°C and re-weighed. The moisture content was obtained by subtracting the dry weight from the wet weight, divided by the dry weight and multiplied by 100. Soil texture was analysed using an LA-960 Slurry Sampler, which uses laser diffraction to measure light scattered from the soil particles as it passes through the measurement cell. The percentage of silt, clay and sand was determined using a soil texture chart.

The soil interference trials were carried out in the same field as the channelled distance perception trials. Four plots measuring 10 m x 4 m were erected 2 m apart and demarcated using barrier tape. Each plot had eight holes dug to 250 mm depth and spaced evenly 2 m apart in a 2 x 4 setup (Figure 2a). Compartments to house individual T. cristatus were made from a guttering pipe socket (50 mm high, 61 mm in diameter), which was considered to supply enough air to the target newts during the trial. A plastic bung was placed at the base of the compartment to keep the newt in place and to reduce the chance of residual scent being left in the hole following the trial. A fibreglass 2 mm mesh was placed at the top of the compartment, a 200 mm plastic guttering pipe with 68 mm diameter (Figure 2b) was then slotted over the top of the pipe socket, holding the netting in place. The pipe was weighed prior to the newt entering the compartment, and assigned to one of four treatments: Sandy Soil Full, Sandy Soil Vented, Clay Soil Full and Clay Soil Vented. The vented treatments comprised a plastic-coated wire mesh with 4 x 4 mm gaps. The mesh was cut to 200 mm length to sit within the 200 mm long pipe. The mesh was rolled up and secured with cable ties with diameter of approximately 25 mm to represent small mammal burrows (a common subterraneous refuge of T. cristatus ([11], see also [39] for the closely related Triturus carnifex). The soil was placed into the pipes using a trowel with soil placed around the wire mesh vents to ensure they stayed open. The pipes containing soil and newts were weighed to provide the weight of the soil by subtracting the pipe weight and the individual weight of the T. cristatus. Locations of each pipe within the plots was allocated using a random number generated in Microsoft Excel. Pipes with all treatment types and blank compartments served as controls. Two pipes of every treatment type was present in each plot with one soil type (e.g clay) located to the left of the plot and the other soil type (e.g. sandy) was positioned to the right. The location of each soil type changed between each run. Vented and full treatments were positioned alternatively within each plot.  Three out of the four plots contained a single T. cristatus with one plot serving as a control (Figure 2c). All pipes were placed out 30 minutes prior to the start of the trial to allow the scent to penetrate through the tubes. This was considered enough time to allow the scent to flow through 20cm of soil whilst also maintaining the welfare of the target newts.

The experiments involved one or two assessors and the dog/handler team. Unlike the channelled distance perception trials, the experiments were undertaken blind as opposed to double blind due to limited numbers of assessors available. The assessor therefore performed the role of the observer and the tester and was present during the test runs. The dog/handler team was not aware of the location of the T. cristatus in each plot. During each trial, the assessor measured weather conditions (using same apparatus and methodology as described for the channelled distance perception trials), air temperature, air humidity, wind speed, wind direction and cloud coverage. Soil temperature and moisture content was measured by placing the probe 5 cm into the soil before trial runs took place using a Soil Condition Meter DSMM600, with moisture levels displayed as <5% (Dry+), 5 to 10% (Dry), 10 to 20% (Normal), 20 to 30% (Wet) and >30% (Wet+). Soil condition measurements were taken immediately before each 4-plot trial but not for single plot trial runs as a 4-plot trial would only take a maximum of 10 minutes, therefore conditions were unlikely to change within this timeframe. As measures of performance by the dog/handler team, numbers and locations of false negative, false positive, true negative and true positive indications were noted.

The handler positioned the dog at the entrance of the plot. The assistant held the detection dog in place as the handler performed a ‘dummy drop’, where the handler moves between the tube locations and pretends to drop the odour, moving in numerical order to focus the dog.  The handler then returned to the dog, unclipped the lead and ask them to search. The handler  monitored the detection dog’s behaviour during the search, and when the dog indicated, the handler confirmed the location with the assessor and if correct the dog was rewarded, followed by a two-minute play session. If the dog/handler team was incorrect, then the search in each plot continued, noting down the number and locations of false indications until the correct indication was given. The trial was terminated if the dog/handler team could not locate the correct pipe within 180 seconds and the handler could not determine if the plot was blank. This length of time was considered sufficient given the size of the plot and the time taken to locate correct pipes during the practice sessions. The handler would call blank before the 180 seconds if she believed the plot did not have a T. cristatus individual present due to lack of interest by the detection dog.  A 45-minute break was given between each 4-plot trial. New pipes were used during each trial. Fresh gloves were worn by the assessor when placing newts into the compartments to avoid transferring newt scent onto the outside of pipes. All pipes were moved to new positions between each trial. Each 4-plot trial was repeated four times within one day, and trials were carried out over two consecutive days in four successive weeks (19^th August 2021 and 17^th September 2021), resulting in a total of 128 single-plot trial runs. To minimise stress, twelve individual T. cristatus were selected at random to be used for the trial day, placed in a labelled container to prevent re-use. Twelve additional individuals were used for the consecutive day, taken from the tanks at random. The dog/handler team trained for three weeks in advance of the soil trial assessment, receiving consultation from Conservation K9 Consultancy.

Data can be accessed using Microsoft Excel.