Context . Failure to acknowledge potential bias from imperfect detection of cryptic organisms such as frogs may compromise survey and monitoring programmes targeting these species.
Aims . The aims of the present study were to identify proximate factors influencing detection probabilities of a range of frog species in monsoonal northern Australia, and to estimate the number of repeat censuses required at a site to have confidence that non-detected species are absent.
Methods . Data on detection or non-detection of frog species based on calling individuals were recorded during 10 wet-season censuses of 29 survey sites in the Darwin region. Factors influencing detection probabilities were identified using occupancy models; model selection was based on the Akaike information criterion. Sampling effort for individual species was calculated using model predictions at different stages of the wet season.
Key results . The covariate water temperature featured in the best-supported models for 7 of the 14 frog species. Six of these species were more likely to be detected when water temperatures were below 30°C. Detection probabilities were also correlated with the number of days since the commencement of the wet season, time since last significant rainfall, air temperature and time after sunset. Required sampling effort for individual species varied throughout the wet season. For example, a minimum of two repeat censuses was required for detection of Litoria caerulea in the early wet season, but this number increased to 13 in the middle stage of the wet season.
Conclusions . Variability in environmental conditions throughout the wet season leads to variability in detection probabilities of frog species in northern Australia. Lower water temperatures, mediated by rainfall immediately before or during surveys, enhances detectability of a range of species. For most species, three repeat surveys under conditions resulting in a high detection probability are sufficient to determine presence at a site.
Implications . Survey and monitoring programmes for frogs in tropical northern Australia will benefit from the results of the present study by allowing targeting of conditions of high detection probability for individual species, and by incorporating sufficient repeat censuses to provide accurate assessment of the status of individual species at a site.