Designing biosurveillance using target analysis for a cryptic invasive species, the brown treesnake (Boiga irregularis)

Biosurveillance is especially important for invasive species that are difficult to control or remove once established. Target analysis, or the strategic evaluation of factors that affect or improve detection of organisms, is one approach to improve biosurveillance over time. Due to their ecologically and economically devastating effects, brown treesnakes ( Boiga irregularis ), an invasive snake in the United States territory of Guam, have a dedicated biosecurity program to prevent their spread. Guam has been occupied by brown treesnakes since the 1940s, and they are distributed throughout the island. The neighboring Saipan Island, otherwise known as Saipan, participates in biosecurity programs designed to prevent their establishment. We evaluated biosurveillance protocols within rapid response and detection exercises for brown treesnakes. Our goal was to identify, account for, and reduce potential error when inferring absence from lack of detection using data collected in Guam and Saipan. We evaluated 1) whether brown treesnake detection probability was density dependent; 2) whether using terrestrial environmental DNA (eDNA) sampling detected snakes in occupied areas; 3) the likelihood brown treesnakes would have been detected in Saipan, given sampling effort during a biosurveillance exercise, and whether lizards could be used as a proxy on snake-free islands for estimating factors that affect detection; 4) the effect of transect design on pacing and subsequently pacing on lizard counts; and 5) the importance of observer training in surveying for cryptic reptiles and completing broad versus targeted bioinventories. We found that brown treesnake detection probability was density dependent when using visual surveys and declined with abundance in Guam, supporting that greater effort was required to have high confidence that lack of detection supported absence in areas with low snake densities. The terrestrial eDNA sampling protocol we developed was effective at detecting snakes in occupied locations in Guam and may be a complementary tool. However, further validation of eDNA protocols would still be required for early detection application. We also found that among trained observers, the ability to detect lizards was positively predictive of the number of snakes they detected at the standard snake densities in Guam. Thus, lizard detection appeared to be a reasonable proxy for evaluating visual search protocols for snakes on snake-free islands. Overall, creating easy pace-setting relationships between time and distance (i.e., 10 m per min), with flagging that clearly marked location on a transect, greatly reduced variability in search pace across survey nights. Observers that were continuously trained to maintain brown treesnake search profiles detected twice as many, on average, cryptic reptiles compared to individuals who had never received training, which was not true for the non-reptiles (birds and small mammals). Additionally, conducting broad bioinventories with many target species reduced detection of the target species, and this was increasingly true as years since the observer completed training for the target species increased. We thus found that designing biosurveillance exercises to evaluate search protocols can help improve or refine survey methods to reduce noise in the data, and accounting for that can improve the estimated confidence that failure to detect an invasive species infers absence.