Evaluating critical uncertainty thresholds in a spatial model of forest pest invasion risk

SAFARIS: a spatial analytic framework for pest forecast systems

Non-native pests and diseases pose a risk of economic and environmental damage to managed and natural U.S. forests and agriculture. The U.S. Department of Agriculture (USDA) Animal and Plant Health Inspection Service (APHIS) Plant Protection and Quarantine (PPQ) protects the health of U.S. agriculture and natural resources against invasive pests and diseases through efforts to prevent the entry, establishment, and spread of non-native pests and diseases. Because each pest or disease has its own idiosyncratic characteristics, analyzing risk is highly complex. To help PPQ better respond to pest and disease threats, we developed the Spatial Analytic Framework for Advanced Risk Information Systems (SAFARIS), an integrated system designed to provide a seamless environment for producing predictive models. SAFARIS integrates pest biology information, climate and non-climate data drivers, and predictive models to provide users with readily accessible and easily customizable tools to analyze pest and disease risks. The phenology prediction models, spread forecasting models, and other climate-based analytical tools in SAFARIS help users understand which areas are suitable for establishment, when surveys would be most fruitful, and aid in other analyses that inform decision-making, operational efforts, and rapid response. Here we introduce the components of SAFARIS and provide two use cases demonstrating how pest-specific models developed with SAFARIS tools support PPQ in its mission. Although SAFARIS is designed to address the needs of PPQ, the flexible, web-based framework is publicly available, allowing any user to leverage the available data and tools to model pest and disease risks.

Prevention is better than cure: Integrating habitat suitability and invasion threat to assess global biological invasion risk by insect pests under climate change

BACKGROUND

Invasive alien species cause substantial impacts on ecosystem, economy, and public health. Therefore, identifying areas at risk of invasion and establishment is essential for the development and implementation of preventive measures. In this study, we integrated information on species habitat suitability, location of airports and ports, and invasion threat maps to assess global invasion risk under climate change using the cucurbit beetle, Diabrotica speciosa (Germar, 1824), as a model organism.

RESULTS

Suitable and optimal habitats for D. speciosa were estimated in several regions beyond its native range and comprised all continents. A decrease in the extent of suitable and optimal habitats for D. speciosa was predicted in different climate change scenarios, resulting in a reduction in invasion risk in most regions. However, regions such as western Europe and isolated areas in southern Asia and Oceania were predicted to face an increase in invasion risk under climate change. Invasion pathways via airports and ports were identified in all continents.

CONCLUSION

Our findings can be used in the development of phytosanitary measures against D. speciosa in high-risk areas. Furthermore, the approach used in this study provides a framework for estimating the global risk of invasion by insect pests and other terrestrial organisms in different climate change scenarios. This information can be used by policy makers to develop preventive measures against species with potential to invade and spread in regions beyond their native range. © 2021 Society of Chemical Industry.

Modelling for risk and biosecurity related to forest health

Modelling the invasion and emergence of forest pests and pathogens (PnPs) is necessary to quantify the risk levels for forest health and provide key information for policy makers. Here, we make a short review of the models used to quantify the invasion risk of exotic species and the emergence risk of native species. Regarding the invasion process, models tackle each invasion phase, e.g. pathway models to describe the risk of entry, species distribution models to describe potential establishment, and dispersal models to describe (human-assisted) spread. Concerning the emergence process, models tackle each process: spread or outbreak. Only a few spread models describe jointly dispersal, growth, and establishment capabilities of native species while some mechanistic models describe the population temporal dynamics and inference models describe the probability of outbreak. We also discuss the ways to quantify uncertainty and the role of machine learning. Overall, promising directions are to increase the models’ genericity by parameterization based on meta-analysis techniques to combine the effect of species traits and various environmental drivers. Further perspectives consist in considering the models’ interconnection, including the assessment of the economic impact and risk mitigation options, as well as the possibility of having multi-risks and the reduction in uncertainty by collecting larger fit-for-purpose datasets.

Assessing the probability of freedom from pine wood nematode based on 19 years of surveys

Many quarantine pests, such as the pine wood nematode (PWN, Bursaphelenchus xylophilus), are surveyed annually in all EU countries. Although a lot of resources are spent in the surveys, the confidence in pest freedom achieved with them is not commonly analysed. We assessed the probability that Finland is free from PWN, based on the surveys done in 2000–2018. We used the methods employed in the risk-based estimate of system sensitivity tool (RiBESS), which has recently been recommended for quarantine pest applications. We considered two scenarios: 1) the surveys aimed to justify phytosanitary import requirements and to facilitate exports and 2) the surveys aimed to detect invasions early to enable eradication of outbreaks. These differed only in the pest prevalence that the surveys were expected to detect. The surveys appeared to support the assumption that PWN is not present in Finland, but they did not seem extensive enough to ensure early detection of invasions. The sensitivity of the import-export surveys was greater than 0.6 in 13 years, whereas that of the early detection surveys was always below 0.25. The probability of freedom achieved in 2018 following 19 years of surveys increased asymptotically with the mean time between invasions. For the import-export surveys, this probability was at least 0.95 unless the mean time between invasions was less than 13 years. For the early detection surveys, the probability of freedom was less than 0.73 unless the mean time between invasions was 63 years or more. The results were rather robust with respect to the parameters for which exact information was lacking. To improve the assessment, a quantitative estimate of the probability of PWN invasion to Finland and a thorough assessment of the maximum area of an eradicable infestation would be needed. To gain an understanding about the true impact of quarantine pest surveys on biosecurity, more assessments, like the one presented in this paper, are needed.

A safety rule approach to surveillance and eradication of biological invasions

Uncertainty about future spread of invasive organisms hinders planning of effective response measures. We present a two-stage scenario optimization model that accounts for uncertainty about the spread of an invader, and determines survey and eradication strategies that minimize the expected program cost subject to a safety rule for eradication success. The safety rule includes a risk standard for the desired probability of eradication in each invasion scenario. Because the risk standard may not be attainable in every scenario, the safety rule defines a minimum proportion of scenarios with successful eradication. We apply the model to the problem of allocating resources to survey and eradicate the Asian longhorned beetle (ALB, Anoplophora glabripennis) after its discovery in the Greater Toronto Area, Ontario, Canada. We use historical data on ALB spread to generate a set of plausible invasion scenarios that characterizes the uncertainty of the beetle's extent. We use these scenarios in the model to find survey and tree removal strategies that minimize the expected program cost while satisfying the safety rule. We also identify strategies that reduce the risk of very high program costs. Our results reveal two alternative strategies: (i) delimiting surveys and subsequent tree removal based on the surveys' outcomes, or (ii) preventive host tree removal without referring to delimiting surveys. The second strategy is more likely to meet the stated objectives when the capacity to detect an invader is low or the aspirations to eradicate it are high. Our results provide practical guidelines to identify the best management strategy given aspirational targets for eradication and spending.

A new hypervolume approach for assessing environmental risks

Assessing risks of uncertain but potentially damaging events, such as environmental disturbances, disease outbreaks and pest invasions, is a key analytical step that informs subsequent decisions about how to respond to these events. We present a continuous risk measure that can be used to assess and prioritize environmental risks from uncertain data in a geographical domain. The metric is influenced by both the expected magnitude of risk and its uncertainty. We demonstrate the approach by assessing risks of human-mediated spread of Asian longhorned beetle (ALB, Anoplophora glabripennis) in Greater Toronto (Ontario, Canada). Information about the human-mediated spread of ALB through this urban environment to individual geographical locations is uncertain, so each location was characterized by a set of probabilistic rates of spread, derived in this case using a network model. We represented the sets of spread rates for the locations by their cumulative distribution functions (CDFs) and then, using the first-order stochastic dominance rule, found ordered non-dominant subsets of these CDFs, which we then used to define different classes of risk across the geographical domain, from high to low. Because each non-dominant subset was estimated with respect to all elements of the distribution, the uncertainty in the underlying data was factored into the delineation of the risk classes; essentially, fewer non-dominant subsets can be defined in portions of the full set where information is sparse. We then depicted each non-dominant subset as a point cloud, where points represented the CDF values of each subset element at specific sampling intervals. For each subset, we then defined a hypervolume bounded by the outermost convex frontier of that point cloud. This resulted in a collection of hypervolumes for every non-dominant subset that together serve as a continuous measure of risk, which may be more practically useful than averaging metrics or ordinal rank measures. Overall, the approach offers a rigorous depiction of risk in a geographical domain when the underlying estimates of risk for individual locations are represented by sets or distributions of uncertain estimates. Our hypervolume-based approach can be used to compare assessments made with different datasets and assumptions.

Managing outbreaks of invasive species--a new method to prioritize preemptive quarantine efforts across large geographic regions

In pest risk assessment it is frequently necessary to make time-critical decisions regarding management of expanding pest populations. When an invasive pest outbreak is expanding rapidly, preemptive quarantine of areas that are under imminent threat of infestation is one of only a few available management tools that can be implemented quickly to help control the expansion. The preemptive quarantine of locations that surround an infested area also acts as a safeguard to counteract the risk of failed detections of the pest in field surveys. In this paper, we present a method that assesses the suitability of preemptive quarantine measures at the level of small geographical subdivisions (U.S. counties). The cost of a preemptive quarantine in a given county is weighed against the protective benefit of delaying the spread of an outbreak to other neighboring counties. We demonstrate the approach with a decision support model that estimates the suitability of preemptive quarantine across multiple counties that surround areas infested with the emerald ash borer (Agrilus planipennis Fairmaire (EAB), Coleoptera: Buprestidae), an emerging major threat to ash tree species (Fraxinus spp.) in North America. The model identifies the U.S. counties where the installation of preemptive quarantine would most effectively slow the spread of EAB populations and reduce risk to high-value areas.

A suite of models to support the quantitative assessment of spread in pest risk analysis

Pest Risk Analyses (PRAs) are conducted worldwide to decide whether and how exotic plant pests should be regulated to prevent invasion. There is an increasing demand for science-based risk mapping in PRA. Spread plays a key role in determining the potential distribution of pests, but there is no suitable spread modelling tool available for pest risk analysts. Existing models are species specific, biologically and technically complex, and data hungry. Here we present a set of four simple and generic spread models that can be parameterised with limited data. Simulations with these models generate maps of the potential expansion of an invasive species at continental scale. The models have one to three biological parameters. They differ in whether they treat spatial processes implicitly or explicitly, and in whether they consider pest density or pest presence/absence only. The four models represent four complementary perspectives on the process of invasion and, because they have different initial conditions, they can be considered as alternative scenarios. All models take into account habitat distribution and climate. We present an application of each of the four models to the western corn rootworm, Diabrotica virgifera virgifera, using historic data on its spread in Europe. Further tests as proof of concept were conducted with a broad range of taxa (insects, nematodes, plants, and plant pathogens). Pest risk analysts, the intended model users, found the model outputs to be generally credible and useful. The estimation of parameters from data requires insights into population dynamics theory, and this requires guidance. If used appropriately, these generic spread models provide a transparent and objective tool for evaluating the potential spread of pests in PRAs. Further work is needed to validate models, build familiarity in the user community and create a database of species parameters to help realize their potential in PRA practice.

Fooled by local robustness

One would have expected the considerable public debate created by Nassim Taleb's two best selling books on uncertainty, Fooled by Randomness and The Black Swan, to inspire greater caution to the fundamental difficulties posed by severe uncertainty. Yet, methodologies exhibiting an incautious approach to uncertainty have been proposed recently in a range of publications. So, the objective of this short note is to call attention to a prime example of an incautious approach to severe uncertainty that is manifested in the proposition to use the concept radius of stability as a measure of robustness against severe uncertainty. The central proposition of this approach, which is exemplified in info-gap decision theory, is this: use a simple radius of stability model to analyze and manage a severe uncertainty that is characterized by a vast uncertainty space, a poor point estimate, and a likelihood-free quantification of uncertainty. This short discussion serves then as a reminder that the generic radius of stability model is a model of local robustness. It is, therefore, utterly unsuitable for the treatment of severe uncertainty when the latter is characterized by a poor estimate of the parameter of interest, a vast uncertainty space, and a likelihood-free quantification of uncertainty.

The influence of satellite populations of emerald ash borer on projected economic costs in U.S. communities, 2010-2020

The invasion spread of the emerald ash borer (Agrilus planipennis Fairmaire) (Coleoptera: Buprestidae) is characterized by the formation of satellite populations that expand and coalesce with the continuously invading population front. As of January 2010, satellite infestations have been detected in 13 states and two Canadian provinces. Understanding how newly established satellite populations may affect economic costs can help program managers to justify and design prevention and control strategies. We estimate the economic costs caused by EAB for the 10-yr period from 2010 to 2020 for scenarios of fewer EAB satellite populations than those found from 2005 to 2010 and slower expansion of satellite populations found in 2009. We measure the projected discounted cost of treatment, removal, and replacement of ash trees (Fraxinus spp.) growing in managed landscapes in U.S. communities. Estimated costs for the base scenario with the full complement of satellites in 2005-2010 and no program to mitigate spread is $12.5 billion. Fewer EAB satellites from 2005 to 2010 delay economic costs of $1.0 to 7.4 billion. Slower expansion of 2009 satellite populations delays economic costs of $0.1 to 0.7 billion. Satellite populations that are both distant from the core EAB infestation and close to large urban areas caused more economic costs in our simulations than did other satellites. Our estimates of delayed economic costs suggest that spending on activities that prevent establishment of new satellite EAB populations or slow expansion of existing populations can be cost-effective and that continued research on the cost and effectiveness of prevention and control activities is warranted.

A harvest failure approach to assess the threat from an invasive species

We present the idea of using potential infringements on annual allowable harvest targets as an approach to estimate threats from invasive species to the forest products sector. The approach uses present-day harvest levels as a reference level to estimate when and where the impact of a nonnative forest pest could become economically damaging. We use a generic model that simulates spread and damage by nonnative invasive species, basic harvest and forest growth through time. The concept is illustrated with a case study of a new nonnative invasive pest, Sirex noctilio Fabricius on pine resources in eastern Canada. Impacts of invasion on wood supply, in particular, the point at which present-day harvest levels are not attainable, were identified for 77 non-overlapping geographical regions that delimit the primary wood supply areas around large mills and wood processing facilities in eastern Canada. The results identify the minimum area of a pest outbreak that could trigger harvest shortages (approximately 12.5-14 M ha of pine forests in Ontario and Quebec). Beyond this level, the amount of host resource available for harvesting in any given year declines rapidly. The failure to sustain broad-scale harvest targets may be an attractive and intuitive indicator for policy makers and regulators interested in developing control and "slow-the-spread" programs for non-native forest pests.

Detection capacity, information gaps and the design of surveillance programs for invasive forest pests

Integrated pest risk maps and their underlying assessments provide broad guidance for establishing surveillance programs for invasive species, but they rarely account for knowledge gaps regarding the pest of interest or how these can be reduced. In this study we demonstrate how the somewhat competing notions of robustness to uncertainty and potential knowledge gains could be used in prioritizing large-scale surveillance activities. We illustrate this approach with the example of an invasive pest recently detected in North America, Sirex noctilio Fabricius. First, we formulate existing knowledge about the pest into a stochastic model and use the model to estimate the expected utility of surveillance efforts across the landscape. The expected utility accounts for the distribution, abundance and susceptibility of the host resource as well as the value of timely S. noctilio detections. Next, we make use of the info-gap decision theory framework to explore two alternative pest surveillance strategies. The first strategy aims for timely, certain detections and attempts to maximize the robustness to uncertainty about S. noctilio behavior; the second strategy aims to maximize the potential knowledge gain about the pest via unanticipated (i.e., opportune) detections. The results include a set of spatial outputs for each strategy that can be used independently to prioritize surveillance efforts. However, we demonstrate an alternative approach in which these outputs are combined via the Pareto ranking technique into a single priority map that outlines the survey regions with the best trade-offs between both surveillance strategies.