Economic impacts of non-native forest insects in the continental United States

Abstract

Reliable estimates of the impacts and costs of biological invasions are critical to developing credible management, trade and regulatory policies. Worldwide, forests and urban trees provide important ecosystem services as well as economic and social benefits, but are threatened by non-native insects. More than 450 non-native forest insects are established in the United States but estimates of broad-scale economic impacts associated with these species are largely unavailable. We developed a novel modeling approach that maximizes the use of available data, accounts for multiple sources of uncertainty, and provides cost estimates for three major feeding guilds of non-native forest insects. For each guild, we calculated the economic damages for five cost categories and we estimated the probability of future introductions of damaging pests. We found that costs are largely borne by homeowners and municipal governments. Wood- and phloem-boring insects are anticipated to cause the largest economic impacts by annually inducing nearly $1.7 billion in local government expenditures and approximately $830 million in lost residential property values. Given observations of new species, there is a 32% chance that another highly destructive borer species will invade the U.S. in the next 10 years. Our damage estimates provide a crucial but previously missing component of cost-benefit analyses to evaluate policies and management options intended to reduce species introductions. The modeling approach we developed is highly flexible and could be similarly employed to estimate damages in other countries or natural resource sectors.

Figure 1. Framework for estimating economic costs of invasive species.

A) The hypothetical cost curve is the frequency distribution of annual economic cost caused by invasive species belonging to a feeding guild. The counts of low and intermediate impact species, as well as the low impact threshold (LT) and the level of damages caused by the poster pest (HT) are known; however, the exact shape of the curve is unknown. B) Alternative cost curves. The data are fit using different parameter values, for four alternative models: gamma (illustrated), log-normal, power and Weibull distributions. C) Illustrative Bayesian posterior probability distribution of cost curve parameters. The posterior probability distribution is the relative probability for each cost curve (defined by parameter values). Some cost curves are more likely than others, given the observed data. The posterior probability allows us to consider and incorporate the relative evidence for each cost curve, thereby accounting for parameter uncertainty. This process is repeated for all four models (Weibull, log-normal, gamma (shown), and power function), and then integrated using Bayesian model averaging, which accounts for model uncertainty. The relative probabilities are shown as a heat graph. D) Probability distribution of total annual cost across species in the guild. We converted the cost curves from the Bayesian analysis into a more meaningful metric - total costs from invaders (Appendix S1). Each cost curve and its corresponding total cost has a relative probability of being true given the observed data. The entire process is repeated for each guild and cost category.

Figure 2. Results for borer feeding guild and local government cost category showing the posterior probability distributions of a) estimated total annual cost of all known borer species, b) probability that a newly introduced borer will cost local governments as much or more than the poster pest (emerald ash borer).

Panels c) and d) show the low sensitivity of the posterior predictions to alternative specifications of the low impact threshold (LT) which was based on expert opinion (Appendix S1). Shown are alternative specifications for total annual guild costs (c) and probability of a high impact borer (d) across two orders of magnitude, where low, medium and high costs are defined as 150, 1,500 and 15,000 annual USD, respectively (Table S6).