Thumbnail Image

Olfactory generalisation in dogs (Canis familiaris) trained to detect koi carp (Cyprinus rubrofuscus)

Koi carp (Cyprinus rubrofuscus) are an invasive fish species that threaten the biodiversity and ecological health of New Zealand’s freshwater systems. To prevent the spread of koi carp, populations must be identified early so management methods can be implemented. However, the methods currently used to detect koi carp populations are either not suited to particular waterbodies, are too expensive, or are not sensitive enough to detect koi carp populations at low biomasses. Domestic dogs (Canis familiaris) are an emerging technology for the detection of koi carp. Previous studies have shown that dogs can detect koi carp at a biomass of 9.3 kg/ha in aquaria water samples, well below the proposed 50 kg/ha threshold for ecological decline. The present study used an automated scent detection apparatus to present dogs with naturally sourced carp-present and carp-absent lakes. The use of lake samples presented a unique challenge to the dogs because of the large variety of distractor odours in a natural environment, potentially complicating the detection of koi carp. In Phase A, dogs evaluated two carp-absent lake samples spiked with koi carp scent at an ecologically relevant biomass of 310 kg/ha. All dogs met the criteria of a hit rate and correct rejection rate ≥ 80% on these lakes, demonstrating that dogs can detect koi carp amongst lake water distractor odours. Phase B presented dogs with 12 positive (carp-present) and 12 negative (carp-absent) probe samples interspersed among samples from the lakes used in Phase A. The dogs’ first responses to positive probes were examined to evaluate generalisation to koi carp scent in novel contexts, and negative probes tested the dogs’ ability to reject lake samples not containing koi carp. Accuracy was stable across Phase A (range = 76% - 84%) and Phase B first responses (range = 71% - 81%). This stability across phases was interpreted as evidence for generalisation. Overall, generalisation was observed, but these generalisation data are interpreted cautiously due to the complexity of lake samples. This complexity may have resulted in the dogs using cues irrelevant to the presence or absence of carp to make responses, such as the differing scent profiles of carp-present and carp-absent lakes. Despite these potential issues, this study acts as a proof of concept for the operational viability of dogs as a koi carp detection technology.
Type of thesis
The University of Waikato
All items in Research Commons are provided for private study and research purposes and are protected by copyright with all rights reserved unless otherwise indicated.