Simulation to investigate site-based monitoring of pest insect species for trade

Abstract

Pest insect surveillance using lures is widely used to support market access requirements for traded articles that are hosts or carriers of quarantine pests. Modeling has been used extensively to guide the design of surveillance to support pest free area claims but is less commonly applied to provide confidence in pest freedom or low pest prevalence within sites registered for trade. Site-based surveillance typically needs to detect pests that are already present in the site or that may be entering the site from surrounding areas. We assessed the ability of site-based surveillance strategies to detect pests originating from within or outside the registered site using a probabilistic trapping network simulation model with random-walk insect movement and biologically realistic parameters. For a given release size, time-dependent detection probability was primarily determined by trap density and lure attractiveness, whereas mean step size (daily dispersal) had limited effect. Results were robust to site shape and size. For pests already within the site, detection was most sensitive using regularly spaced traps. Perimeter traps performed best for detecting pests moving into the site, although the importance of trap arrangement decreased with time from release, and random trap placement performed relatively well compared to regularly spaced traps. High detection probabilities were achievable within 7 days using realistic values for lure attractiveness and trap density. These findings, together with the modeling approach, can guide the development of internationally agreed principles for designing site-based surveillance of lure-attractant pests that is calibrated against the risk of non-detection.

Keywords: market access; quarantine pest; survey and detection; trapping.

Fig. 1.

Examples showing different site shapes and trap arrangements for a 10-ha block with 1 trap/ha. a) Shows a perimeter trap arrangement on a triangular site, b) shows random trap arrangement on a quadrilateral site, and c) shows regular trap arrangement on a pentagonal site. Not shown is the square site shape.

Fig. 2.

Variable importance plot of the 8 simulation parameters by mean squared error for the random forest models using Days 0, 7, 14, 21 and combined Days 0, 7, 14, and 21 simulation output. Parameters are ordered based on combined random forest model. The added “Day” parameter for the combined random forest model is not included in this plot.

Fig. 3.

The effect of pest release size and time since release (0 to 21 days) on mean probability of pest detection, grouped by outbreak origin (inside site left, outside site right) and trap arrangement (perimeter top row, random middle row, regular bottom row). Boxplots show the 25th, 50th, and 75th percentiles in the box and 1.5 times the interquartile range.

Fig. 4.

Line plots of the mean probability of detection over days, grouped by mean step size (line color), lure attractiveness (columns), and trap density per ha (rows). Note that uncertainty bounds are not included.

Fig. 5.

Boxplots of the mean probability of pest detection by trap arrangement on Day 0, grouped by trap density (see legend), outbreak origin (columns) and lure attractiveness (rows). Release size was 10 pests. Boxplots show the 25th, 50th, and 75th percentiles in the box and 1.5 times the interquartile range.

Fig. 6.

Boxplots of the mean probability of pest detection by trap arrangement on Day 7, grouped by trap density (see legend), outbreak origin (columns), and lure attractiveness (rows). Release size was 10 pests. Boxplots show the 25th, 50th, and 75th percentiles in the box and 1.5 times the interquartile range.