Predicting costs of alien species surveillance across varying transportation networks.
Abstract
Efforts to detect and eradicate invading populations before they establish are a critical component of national biosecurity programmes. An essential element for maximizing the efficiency of these efforts is the balancing of expenditures on surveillance (e.g. trapping) versus treatment (e.g. eradication). Identifying the optimal allocation of resources towards surveillance requires an underlying model of how costs and the probability of detection fluctuate with survey intensity across various landscapes. Here, we develop a model, widely applicable across biological systems, for predicating costs associated with varying surveillance intensities across diverse road networks. We assumed that surveillance is conducted across a set of point locations. Resources needed to conduct surveillance include the fixed costs associated with surveying a point (e.g. cost of materials or labour time spent at the survey point) and variable costs that correspond to the expense of the time and distance travelled between points. We estimated travel time and distance between points as functions of surveillance intensity and road network characteristics using data from simulated least cost driving routes connecting points located on real-world road networks. Time and distance estimates were then combined with cost data from an actual gypsy moth Lymantria dispar surveillance programme in the state of Washington to predict per trap costs of surveillance across varying road network densities and surveillance intensities. Per point driving time, driving distance and total costs all decline with increasing survey point density and increasing road density. Surveillance intensity (planned point spacing) explains ∼94% of the average time driven per point and 97% of the distance driven per point - thus representing the primary explanatory variable. Incorporating road density and dead end road density explains relatively little additional variance in the model, although they improve goodness of fit. Synthesis and applications. This work predicts costs associated with surveillance of invasive species populations. We find that the cost per survey point diminishes with increasing survey point density and also depends on road network characteristics. When combined with maps for the relative risk of alien species establishment across landscapes and measures of surveillance efficacy dependent on effort, these cost predictions can increase efficiency of surveillance and eradication efforts for the gypsy moth and other invasive species.