Classical Weed Biological Control Outcomes: A Catalogue-based Analysis of Success Rates and Their Correlates Submitted by: Mike Barbetta Supervisors: Dr. Jay Malcolm and Dr. Sandy Smith Final Capstone Submission for FOR3008H April 12th, 2018 Abstract Classical weed biological control (hereafter CWBC) is an important and effective management tool that may see an expanded role as more problematic weeds emerge worldwide, more countries begin to adopt this approach, and broader applications of the practice are explored. However, success is never certain in CWBC, and predicting success remains an elusive goal. Here, a series of catalogue-based analyses of CWBC outcomes are conducted in order to quantify success rates and to explore potential factors that may be associated with success. Statistical analyses utilizing chi-squared tests were utilized to search for correlations between CWBC efficacy and various host and agent characteristics. Multiple such relationships were identified, including previously identified correlations between CWBC outcomes, host habitat types, and agent orders. Significant differences in CWBC outcomes are also correlated to agent feeding guilds, which represents a novel association revealed by this study that warrants further investigation. For example, defoliating agents were associated with lower establishment rates, and root boring agents were associated with higher control efficacy. Success rates derived from the most recent catalogue were also delineated by region and time period, and management implications are discussed. INTRODUCTION The deliberate introduction of exotic agents to control problematic invasive species; that is, classical biological control, is an important tool in the management of weeds worldwide. In many scenarios, use of classical biological control for weed control (hereafter CWBC) is touted as the only and best option because conventional control techniques are often ineffective against problematic weed species, and because CWBC offers the potential for significant, cost-effective management impacts in the form of self perpetuating, permanent, landscape-wide control mechanisms (Van Driesche et al. 2010). To date, over 450 insect agents have been released against over 150 plant species in over 90 countries and regions worldwide (Winston et al. 2014), with many notable successes (McFadyen 2000). Although the practice does entail the risk of non-target impacts occurring (Myers and Bazely 2003), the ongoing and potentially expanded use of this practice is encouraged by a number of factors: significant agent impacts on non-target hosts are rare (Suckling and Sforza 2014), the number of problematic invasive weed species is expected to increase worldwide (Seastedt 2015), the goals and applications of CWBC (including relatively unexplored forestry applications) continue to evolve and expand (Van Driesche et al. 2010; MacQuarrie et al. 2016), and previously reluctant European jurisdictions appear willing to make use of biological control (Shaw et al. 2016; Varia et al. 2016). However, the efficacy of CWBC as a weed management tool is plagued by several pervasive issues. Success is uncertain, widely variable, and is often inadequately monitored so as to make accurate evaluation difficult (Suckling 2013; Seastedt 2015). Success prediction and improvement in control agent selection and release is a complex and challenging task that, despite recent attention and efforts, remains an elusive goal (Seastedt 2015). This is problematic given the high costs of initiating and carrying out new biological control programs (Myers and Bazely 2003), and the significant cost increases (Paynter et al. 2015) and inherent risks (Delfosse 2005) that are incurred with each additional agent that is required for any given program. Determining success rates in CWBC can be difficult. Depending on the metric of success being used (e.g., if the success of each individual agent is being evaluated, or if the net success of a set of agents deployed against a specific host plant is being evaluated), reported success rates can vary significantly (McFadyen 1998; Fowler et al. 2008). Furthermore, there is tendency in CWBC publications (even those published very recently) to cite dated articles for this information despite much more up to date information being available (Fowler et al. 2008; DiTomaso et al. 2017). Attempting to establish generalized ‘rules-of-thumb’ to guide CWBC releases is also a contested issue, with context-specific factors often cited as more important determinants of success in CWBC programs (McFadyen 1998; Center et al. 2014; Seastedt 2015). Nevertheless, several authors have explored or suggested basic host and agent characteristics as potential indicators of success in CWBC (Boughton and Pemberton 2008; Clewey et al. 2012; Paynter et al. 2012) and the potential merits of post-hoc analysis can only increase with time as more data is accumulated. An exhaustive and regularly updated catalogue of worldwide CWBC releases and their results is available (Winston et al. 2014) and there may be value in analyzing the information within this catalogue to identify and explore trends in the successes and failures of CWBC worldwide. Several reviews of the results of CWBC releases in relation to potential explanatory variables have been conducted. Paynter et al. (2012) discovered that CWBC programs deployed against plant hosts that resided in aquatic habitats, reproduced vegetatively, and/or were not major weeds within their native ranges were much more effective than programs deployed against plant hosts that lacked any of these characteristics. Using 61 published CWBC studies, Clewey et al. (2012) discovered that Coleopteran agents from Chrysomelidae and Curculionidae were more effective CWBC agents than other insect types. While both of these studies utilized an analysis of precise quantitative data, it is still possible to take a similar approach using the more generalized information available in the Winston et al. (2014) catalogue. Doing so involves an important trade-off: the information within the catalogue consists of broad, generalized categories, but it also much more comprehensive, readily available, and therefore easily applied to evaluations of CWBC efficacy. While Boughton and Pemeberton (2008) utilized a catalogue-based approach to analyze establishment success rates of previous CWBC agents across different insect orders and lepidopteran families, further catalogue analyses of CWBC outcomes has been limited. The purpose of the current paper is to build on this approach through a series of post-hoc catalogue-based analyses. My first objective was to quantify success rates of CWBC releases in general, by major practicing jurisdiction, and in respect to how these rates may be changing over time. Crawley (1989) indicated that success in CWBC was decreasing over time, whereas more recently improved rates of success have been reported in certain regions (Hayes 2000; Fowler et al. 2008). There is a need to update the information concerning CWBC success rates and to present in a way that is useful for informing management decision making, avoids bias, and clearly differentiates the different evaluation criteria that is used to assess success rates. A second objective was to assess CWBC release outcomes in relation to pertinent host and agent characteristics. Catalogue analysis with up to date information may allow the confirmation of previously identified correlations (e.g., improved outcomes when utilizing coleopteran agents (Crawley 1989; Boughton and Pemberton 2008; Clewley et al. 2012) or against aquatic hosts (Crawley 1989; Paynter et al. 2012)), as well the exploration of other possible trends. Myers and Bezely (2003) suggest that certain agent feeding guilds (e.g., seed predation) may act as less effective control mechanisms than others. Therefore, the evaluation of CWBC release outcomes in relation to agent feeding guilds will also be conducted to in order provide an initial cursory evaluation of these previously unexplored relationships. The results of this meta-analysis may be significant in a number of ways. If the results obtained from this catalogue meta-analysis are found to emulate results from non-catalogue based analysis (i.e., better CWBC against aquatic hosts, and when utilizing coleopteran agents), this will help support the legitimacy of the catalogue-based approach. As such, the results obtained from the novel relationships being explored (i.e. agent efficacy by feeding guilds) may be more confidently used to help inform agent prioritization, or identify areas of interest for more robust investigation. Furthermore, the investigation of the overall efficacy of CWBC in general, by jurisdiction, and over time will evaluate how often and to what extent this weed management strategy has been successful in the past, where it has been most successful, and whether or not the efficacy of the practice has stagnated, improved, or deteriorated over time. This will assist in informing management expectations and strategies when deciding upon options for the management of problematic weeds. METHODS Data compilation of catalogue and supplementary information Catalogue information, which contains the date, location, basic host (i.e., family and species name) and agent characteristics (i.e., species name and classification), and outcome of every CWBC release worldwide up until 2012 (Winston et al. 2014), was utilized as the primary source for analysis. Information from the catalogue website