Quantifying imperfect detection in an invasive pest fish and the implications for 3 conservation management 4 5

27 28 In managing non-native species, surveillance programmes aim to minimise the 29 opportunity for invasions to develop from initial introductions through early detection. 30 However, this is dependent on surveillance methods being able to detect species at 31 low levels of abundance to avoid false-negative recordings through imperfect 32 detection. We investigated through field experimentation the ability to detect 33 Pseudorasbora parva, a highly invasive pest fish in Europe, in relation to their known 34 density and sampling method. Secure pond mesocosms of area 100 m 2 contained P. 35 parva densities from 0.02 to 5.0 m-2 ; each density was in triplicate. These were 36 searched using point sampling electric fishing and deployment of fish traps (non-37 baited and baited). No fish were captured at densities < 0.5 m-2 using any method and 38 this was considered their detection threshold. Point sample electric fishing was the 39 least effective detection method, producing high proportions of false-negative data 40 even at high fish densities. Baited traps were the most effective detection method. 41 Probability of detection of P. parva was 1.0 for baited traps at all densities > 0.5 m-2 , 42 whereas for electric fishing it only exceeded 0.95 at 5.0 m-2 using high searching 43 effort. These data reveal that small pest fishes such as Pseudorasbora parva may be 44 prone to imperfect detection when at low densities and this is consistent with a 45 number of other invasive species. This indicates the importance of designing 46 surveillance programmes using methods of known statistical power to optimise 47 conservation resource expenditure and enhance management outcomes. 48 49