Date: Tue, 05 Oct 1999 17:39:16 -0400 From: "Keith R. Hopper" Organization: University of Delaware X-Mailer: Mozilla 4.5 [en] (Win95; I) X-Accept-Language: en MIME-Version: 1.0 To: bc-ntimpact Subject: Summary of discussion to date Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Sender: owner-bc-ntimpact@udel.edu Precedence: bulk Reply-To: bc-ntimpact@udel.edu I have attached below a summary of the postings to bc-ntimpact to date. I plan to present this summary at the upcoming IOBC Symposium on Evaluating Indirect Ecological Effects of Biological Control to be held in Montpellier, France, 17-20 October 1999. If any of you see important gaps or misstatements in this summary please let me know as soon as possible by contacting me directly at khopper@udel.edu. A modified version of this summary, as well as a summary of the workshop part of the upcoming symposium, will be published as a chapter in the proceedings of the symposium. At present, I plan to list all who posted messages to bc-ntimpact in the published summary. If you do not want to be listed, please let me know during the next few weeks. Yours, Keith Hopper -- ****************************************************************** Keith R. Hopper, Research Entomologist USDA, ARS, Beneficial Insect Introduction Research Unit University of Delaware, 501 S. Chapel St., Newark, Delaware 19713 Voice +1-302-731-7330 ext 38; Fax +1-302-737-6780 ****************************************************************** Summary of Internet Workshop on Research Needs Concerning Non-target Impacts of Biological Control Introductions Keith R. Hopper, 5 October 1999 Since October 1998, ARS has operated an Internet mail-list for discussion of research needs for assessing and reducing non-target impacts of biological control introductions. Over 200 participants from 28 countries subscribed to the mail-list and have posted thus far 150 messages. An archive of these messages is available at http://www.udel.edu//entomology/khopper. Here, I summarize the discussion to date. The emphasis has been on arthropods introduced to control weeds and arthropods. Although there have been a few postings on tangential issues (e.g. biology of particular natural enemies, transgenic plants, the definition of biological control), most of the postings can be grouped under four germane topics: (1) what sort of non-target impacts should concern us; (2) host range evaluation; (3) predicting impacts of biological control introductions versus other management options; and (4) retrospective studies to assess actual impacts. (1) Sorts of non-target impacts that should concern us. Agreement is not complete on the species that should be considered when evaluating non-target impacts. Some have argued that impacts on all species should be considered equally, others hold that some species should receive special consideration (e.g., endangered species, species that provide crucial ecosystem services, species that provide economic benefits to society), and others hold that only such species should be of concern and impacts on others should be ignored. Although most agree that we should be concerned about population-level impacts rather than attack on individuals, the magnitude and spatial/temporal scale of population-level impact that should concern us are not agreed upon. Depending on how impact is defined, its measurement may be extremely difficult. Field collection of hosts and exposure of sentinel hosts give impact on individuals which is hard to translate into impact on populations. Some argue that indirect impacts of arthropod introductions on parasitoids of non-target hosts or prey may be larger than impacts on the hosts or prey themselves. This is because parasitoids may be more liable to extinction than their hosts. However, such indirect impacts may be more difficult to measure than direct impacts on non-target hosts or prey. Not all impacts on non-target species need be considered detrimental. For example, introduced biocontrol agents might reduce the abundance of common native species that compete with or prey upon endangered native species so that the introduction would decrease the extinction probability of the endangered species. Ultimately, the definition of non-target impact and how such impacted is valued is a societal decision, but researchers attempting to evaluate non-target impacts need working definitions for design of surveys and experiments. Such definitions may be system specific, but it would be useful to agree upon guidelines for the sorts of impacts to consider. (2) Host range evaluation. Host range evaluation is usually the first, and often the only, step in predicting risks of non-target impacts. Some argue that the literature provides a good first screening for host range, especially for distinguishing candidates with broad versus narrow host ranges. Others argue that misidentifications in the literature, and especially the older literature, often cause problems of spurious host records. Some of these problems may be avoided by weighing the quality of literature data on host range, e.g., by taking into account numbers of individuals examined, percent attack, and spatial and temporal extent of studies. Laboratory or field experiments can be used to confirm data from field collections. However, physiological/behavioral host range measured in the laboratory and ecological host range measured in the field often differ. Host range evaluation in the area of origin can help with this problem by allowing comparison of field collections and laboratory measurements. An inherent assumption is that host range in the area of origin is a reasonably good predictor of host range in the area of introduction. Good design of experiments on host range usually involves system-specific factors. For example, the value of choice versus no-choice testing depends on metapopulation dynamics of system. If an biocontrol agent can disperse from pest patches to patches of non-targets where it cannot reproduce, choice testing may be crucial; if biocontrol agents will often find themselves where they have driven pest patches extinct, no-choice testing may be more relevant. Some argue that evolution of host range poses significant problems for prediction of host range of introduced natural enemies. Others argue that there is little evidence for evolutionary shifts in host range after introduction, especially for herbivores introduced to control weeds. High levels of feeding and development on non-target species by a few individuals in a population represents a greater risk of subsequent adaptation to the non-target than low levels of feeding or development by most individuals in the population. This means that host range testing should track feeding and development of individuals and families. This will at least allow estimation of the likelihood of shifts in frequencies of genes/genotypes after introduction, although it will not provide information on the likelihood of novel genes arising. Ultimately, a deeper understanding of mechanisms affecting host specificity would allow better prediction of host ranges. Problems with host range screening of entomophages (difficulty of obtaining and rearing non-target species, problems with anomalous responses in the laboratory) may make such understanding particularly important for introductions against insects. (3) Predicting impacts of biocontrol introductions versus other management options. Host range testing is the most tractable, and most used, approach to evaluating risks of impacts on non-target species. If one can show that a candidate for introduction will not attack any non-target species in the area of introduction, risk of non-target impacts is low. However, even in this case, indirect impacts may occur, e.g., if the target pest supports native species in some way. Furthermore, proving no feeding or development on non-target species is difficult because one cannot test all possible non-target species and because host range may evolve after introduction. What is perhaps more important is that although many candidates for introduction have narrow host ranges, few are monospecific. Inclusion of species in the host range of an introduction candidate is often not simply dichotomous, with a list of species that are suitable for feeding and development neatly separated from those that are not suitable. Thus, several non-target species may be exposed to some reduced risk of attack compared to the target, and evaluation of the risks of significant impacts on these less suitable non-targets requires translation of attack on individuals to effects on populations. But there are major problems with predicting such impacts for target pests, let alone for non-target species. Furthermore, most agree that non-target impacts should be weighed not only against the value of controlling the target pest, but also against the risks versus benefits of other management options. If introductions are not done, resources devoted to other strategies may lead to increased biocontrol by natives (conservation, switching of generalists) or other successful control of the target. On the other hand, the pest may continue unabated and cause enormous economic or environmental damage. Unfortunately, we don't yet have the theory and methods that would allow quantitative comparison of the risks versus benefits of all management options. Although many debate whether we can develop methods for predicting which natural enemies are likely to be most effective, most researchers agree that pre-release studies in the areas of origin and introduction are needed to determine the value of controlling the pest, whether biological control introductions are likely to prove useful, and whether non-target impacts are likely. Some argue that mathematical models of the population dynamics of introductions are needed to provide a common framework for comparing various systems and to handle the complexity of the potential interactions in most systems. The problem with modeling real systems is that so little is known that describing system structure, let alone estimating model parameters, may prove impossible at present and for the foreseeable future. (4) Retrospective studies to assess impacts. All recognize the need for retrospective studies of previous introductions and their impacts. Such studies can provide the testing ground for predictions about non-target impacts, whether these are based on estimates of host range from field collections or laboratory experiments or on mathematical models of system dynamics. However, it is crucial to assess impacts in light of host range testing and predictions made prior to introduction so that methods for assessing impacts prior to introduction can be evaluated. Releases have been done when some level of attack on non-target species was possible, but considered less risky than alternatives for management of the pest. Impacts on non-target species and evolution of host range have rarely been evaluated for these cases in the past, but these introductions provide excellent opportunities for study. Some examples are: (a) Chrysolina quadrigemina feeding on Hypericum calycenum, an exotic plant extensively used as a ground cover in northern California and on a native Hypericum species; (b) Rhinocyllus conicus introduced against Carduus nutans and feeding on nontarget Cirsium spp.; (c) biocontrol agents for introduced weedy Solanum spp. and feeding on eggplant Solanum melongena in South Africa; (d) Galerucella spp. released against introduced Lythrum salicaria and feeding on native Lythrum alatum and Decodon verticillatus; (e) an eriophyid mite feeding on native Hypericum species in Australia; (f) Tyria jacobaeae introduced against Senecio jacobaea feeding on a native Senecio spp.; (g) Diachasmimorpha tryoni imported to Hawaii for fruit fly control parasitizing non-target tephritid Eutreta xanthochaeta imported for biocontrol of lantana in Hawaii. Although several of these introductions are now being studied to assess non-target impacts (e.g., Gallerucella spp., R. conicus, D. tyroni), there are problems with the spatial and temporal scale of impacts versus our capacity to study introductions over large areas and long periods. All natural systems are dynamic, and introductions in particular may take a long time to reach some sort of equilibrium or at least relatively steady state. Such dynamic systems present a moving target so that one may need to wait until an introduced agent is established over a broad area for a long time before evaluating the impacts of introduction. Ten years is often used as the time horizon for evaluating impacts of introductions on targets, but this is just a rule of thumb and may not even be applicable for target impacts, let alone non-target impacts. Such long time horizons and large spatial scales often put evaluations of target, as well as non-target, impacts beyond the resources available to most researchers. Unless thorough sampling was done before introduction, studying introduced species that have established over a large area for a long time means that potential control sites without the introduced species will be difficult or impossible to find. On the other hand, the slow rate of spread and patchy distribution of many introduced agents means that control sites without the introduced agent are often available long after introduction. A major problem with extensive surveys for non-target impacts is that negative evidence is hard to quantify and publish. Showing in a convincing way that small but significant impacts have not occurred is much more difficult and time consuming than showing that large impacts have occurred. Although both conservationists and biocontrol practioners regard with dismay illegal introductions and temporal and geographical differences in societal views about acceptable procedures for biological control introductions, differences in procedures do provide an opportunity to evaluate the effect of procedures on non-target impacts. Novel approaches to retrospective analyses, as well as a renewed call to treat introductions as experiments to test how introductions should be conducted, are the most exciting output of this workshop so far.