Quantifying economic damages caused by invasive species is crucial for cost-benefit analyses of control measures. Most studies focus on short-term damage estimates, but evaluating exclusion or prevention measures requires estimates of total anticipated damages from the time of establishment onward. The magnitude of such damages critically depends on the timing of damages relative to a species’ arrival because costs are discounted back to the time of establishment. Using theoretical simulations, we illustrate how (ceteris paribus) total long-term damages, and hence the benefits of prevention efforts, are greater for species that a) have short lags between introduction and spread or between arrival at a location and initiation of damages, b) cause larger, short-lived damages (as opposed to smaller, persistent damages), and c) spread faster or earlier. We empirically estimate total long-term discounted impacts for three forest pests currently invading North America - gypsy moth (Lymantria dispar), hemlock woolly adelgid (Adelges tsugae), and emerald ash borer (Agrilus planipennis) - and discuss how damage persistence, lags between introduction and spread, and spread rates affect damages. Many temporal characteristics can be predicted for new invaders and should be considered in species risk analyses and economic evaluations of quarantine and eradication programs.
As the table below shows the total present value of residential damages from the gypsy moth ranges from $32 million with a discount rate of 5% to $6.415 billion at a discount rate of 1%, and $0.7 million to $20 million for the hemlock wooily adegid and from $9.2 billion to 26.4 billion for the emerald ash borer invasion
Table 1. Estimated total present value (millions 2011 USD) of residential damages from gypsy moth and hemlock woolly adelgid and of ash tree removal and replacement costs from emerald ash borer invasion. Present values are calculated from the time of both introduction and initiation of damages (1869 and 1880 for gypsy moth, 1911 and 1971 for hemlock woolly adelgid, and 1990 and 2002 for emerald ash borer).
Fig. 1. Panels show the (a) radial extent, (b) invaded area, (c,e) nondiscounted annual damages, and (d,f) present value of annual damages for the first 100 years of invasion for spread models A, B, and C (lines) with the baseline parameterizations. The second row (c,d) shows damages when persistence P is 1 year, and the third row (e,f) shows damages when persistence P is 100 years.
Quantification of gypsy moth spread and damages.
[The authors] used historical invasion records to characterize past invasion spread from the time of its initial introduction in Medford, Massachusetts, in 1868 through 2012. Past spread of gypsy moth from 1868 to 1912 is summarized in Liebhold et al. and Liebhold and Tobin from a variety of historical records. Spread from 1912 to the present is recorded in county-level quarantine records published by the US Department of Agriculture (US Code of Federal Regulations, Title 7, Chapter III, Section 301.45). Future spread of the gypsy moth was projected based on an assumption of constant radial spread, using a rate of 5 km/yr (the average spread rate estimated for 1999–2012 along the expanding population front in the United States; ... Year of establishment was predicted for each county in 39 eastern states . [They] excluded spread into counties whose climates are unsuitable for gypsy moth development.
Following Aukema et al., [the authors] assumed that residential damages from gypsy moth invasion depend on the number of one- and two-unit houses that experience defoliation from gypsy moth outbreaks in each year and estimated damages as the household’s willingness to pay to avoid gypsy moth damage, including the loss of value from nuisance, defoliation, and tree mortality. Aukema et al. assumed a willingness to pay of $503 per urban household and $433 per rural. [The authors] used the lower value estimate of 433 USD 2011 per house per year of defoliation as [their] baseline estimate.
Statewide records of the area annually defoliated by the gypsy moth are available for 1924–2010. From this [they] calculated the percentage of susceptible invaded habitat that was defoliated in each state and year. Susceptible land area is assumed to consist of forested lands with basal area comprising >20% tree species preferred by the gypsy moth; these estimates were derived from national forest inventory data. [They] also calculated the average percentage of susceptible invaded area defoliated across all years (1924–2010) and states, and we used this value as an estimate of the percentage of invaded susceptible area defoliated annually in years and states with no data (i.e., pre-1924 and post-2010). In this way [they] assumed that future outbreak dynamics would be comparable to past dynamics when projecting future defoliation. Analyses of historical dynamics indicate some temporal variation but overall a relatively constant pattern of gypsy moth outbreak periodicity over the past 85 years. The fungal pathogen, Entomophaga maimaiga, appeared in North American populations in 1999 and previously had not been known to be present. Though this pathogen causes considerable mortality of gypsy moth, there is no conclusive evidence that this disease has altered gypsy moth outbreak dynamics, since recurrent large regional outbreaks have continued subsequent to its emergence.
[They] estimated the number of potentially affected households in each infested county as the number of one- or two-unit housing structures in the county times the fraction of susceptible area in the county. [They] calculated the number of affected households in each county in each year as the number of potentially affected households times the average percentage susceptible area defoliated. We multiplied this by the willingness to pay to avoid gypsy moth damage to estimate the expected annual damages to residential properties from gypsy moth invasion.
Estimates of the numbers of one- and two-unit housing structures are not available across the full timeframe of historical and future gypsy moth invasion. Thus [they] estimate these numbers using available housing structure and human population data. US Census estimates of the number of one-unit housing structures are available by state for each decadal year from 1940 to 2000 (https://www.census.gov/hhes/www/housing/census/historic/units.html). In addition, the number of one- and two-unit structures is available at the county level for 2011... [They] estimated state-level numbers of housing structures for years without data based on state population data. Specifically, [they] modeled the ratio of one-unit housing structures to human population numbers as a function of year and its standardized quadratic using GLM estimation in Stata. [They] used a logit link and state fixed-effects to jointly estimate state-specific coefficients and then used the results to predict the house: population ratio for years without housing data. From this, [they] backed out the number of one-unit housing structures per state based on state population estimates, which were available annually for 1880–2012.... Only country-level population predictions are available for 2013–2060..., so [they] estimate state-level populations using these data and each state’s proportion of the country’s population in 2012. Furthermore, [they] estimated county-level housing structure numbers from state estimates by assuming that housing structures were allocated across counties in proportion to their observed 2011 distributions, and that the ratio of one-unit to two-unit housing structures also was constant over time at 2011 values. Given the lack of population data and considerable uncertainty about the ratio of housing structures to population in the future, [they] assumed constant housing numbers beginning in 2060. Although housing numbers are likely to continue to increase, this assumption also recognizes that the number of houses with susceptible vegetation may decline over time because of loss of habitat.
by Rebecca Epanchin-Niell, Andrew M. Liebhold
Resources For the Future (RFF) www.RFF.org
Discussion Paper 15-18; May, 2015