How Could Ecological Research be Used to Improve Safety of Weed Biocontrol?
We need to avoid rejecting safe, effective agents like the St John’s wort beetle, where host testing results overestimate likely non- target damage in the field.
Many ecologists have a negative opinion of biocontrol while practitioners argue that it offers a cost-effective solution to many invasive weed problems. Meanwhile practitioners are under pressure to implement effective weed biocontrol more quickly, cheaply and safely.
Recently Simon Fowler and colleagues were asked to prepare a paper for the Journal of Applied Ecology on how advances in ecological research could help to deliver on these aims while minimising any potential negative outcomes. Below is a summary of that paper which is divided into two parts: avoiding direct and indirect non-target effects.
Part One: Avoiding Direct Non-Target Effects
Host range testing is undertaken to determine which plant species will be damaged by a biocontrol agent. The most conservative test is a ‘no-choice’ test where an agent can either attack a plant or not. Such tests may involve feeding by an insect, egg-laying by a female insect or infection by a pathogen (in all cases, the stark alternative is starvation, death and failure to reproduce). Other tests provide potential biocontrol agents, particularly insects, with a selection of test plants to see which ones they eat (a ‘choice’ test).
There are problems with using ‘no choice’ and ‘choice’ tests. In ‘no-choice’ tests, an agent is rejected if it attacks plants of value (native species or plants of economic importance). However, there can be a high rate of ‘false-positive’ results meaning that potential biocontrol candidates are rejected or overlooked. “There is a real danger of eliminating potential agents unnecessarily as often the non-target impact of agents may not even occur or be minimal once they are released into the field,” explained Simon.
The potential for missed opportunities is highlighted by a current study led by Ronny Groenteman that retrospectively tested the host range of two beetle species introduced into New Zealand in 1943 and 1965 to control St. John’s wort (Hypericum perforatum). These tests showed that, under current protocols, the two beetles would not meet the requirements for release in New Zealand. However, the beetles have undoubtedly been successful in controlling this serious pasture weed with minimal impacts on native Hypericum species.
On the other hand, ‘choice’ tests are not always good indicators of the agent’s behaviour when the host plant’s availability is reduced, creating a ‘no choice’ situation in the field. For example, the gorse pod moth (Cydia succedana), which was released in New Zealand to target gorse (Ulex europaeus), was later found in the seed pods of other plants, especially during times when the gorse pods were not abundant. Similarly, there is uncertainty surrounding whether ‘choice’ tests adequately predict whether an agent will damage plants of value when they disperse away from their normal hosts, and again find themselves in a no-choice environment.
Ideally, to get the most realistic results potential agents would be tested on plants of value in the field or in the environment that they are going to be released into. But the practicalities of keeping potential agents safely contained in the field prevent this. Another option is to do the host range testing in the native range of the biocontrol agent using plants from New Zealand, but the logistical issues involved are often insurmountable.
It is apparent that unless we improve our risk assessment procedures, we will either reject potentially successful and safe agents based on overly conservative host range testing or conversely, underestimate the damage that agents can do when faced with limited host plant availability. How could better ecological studies help? “What is required,” said Quentin Paynter, “is a framework for better assessing whether non target impacts are significant without having to rely on field based testing.”
Another area that would benefit from further study is the risk of host-range expansion over time. This is currently believed to be a minor risk but we need to improve our understanding of which agents might be at greater risk of evolving to do this, and under what conditions. For example, if we go down the track of targeting multiple weeds in one hit by using less highly specific agents, are we setting ourselves up for trouble? Or is there a higher risk where highly specific agents are known to have slightly different host-ranges in different geographical areas?
Part Two: Avoiding Indirect Non-Target Effects
Large scale changes in food webs resulting from the introduction of biocontrol agents are rare but something we need to continue to be careful to avoid. As well as food web effects, other “indirect effects” arising from the introduction of biocontrol agents include changes to pollination systems or mutualistic relationships between species.
Some potential food web effects are obvious and are already considered in weed biocontrol programmes. For example, some broom biocontrol agents could also attack tree lucerne (Cytisus proliferus) which is an important spring food for native pigeons. These “conflicts of interest” are carefully addressed early on during biocontrol programmes, and under the Hazardous Substances and New Organisms (HSNO) Act (1996) any species that have the potential to “cause any significant displacement of any native species within its natural habitat” are not approved for release. Retrospective studies have allowed us to get better at identifying, and therefore avoiding, agents that have a high potential for indirect non-target effects, e.g. those that become common but have no impact on the target plant, or become popular targets themselves (through disease, predation or parasitism –for recent breakthroughs on the latter see “Is it Possible to Predict Parasitism?” in Issue 49).
“Although many of the implications of releasing weed biocontrol agents for food webs are predictable, there may be more subtle effects that have previously fl own beneath the radar,” explained Simon. For example, a recent high profile debate in the USA centred on the implications of releasing gall flies (Urophora spp.) to control knapweed (Centaurea spp.). As the gall flies increased in number, the food supply and abundance of native deer mice (Peromyscus maniculatus) also increased. Elevated numbers of deer mice could have a range of effects in food webs, even potentially impacting human health because the mice are the main reservoirs of hantavirus (a rare but serious disease in humans).
It is clear that we need to avoid major impacts on “keystone species” and avoid triggering “trophic cascades” where one disturbance creates a ripple effect on other species.
However, while our understanding of how food webs function is good enough to predict major unwanted consequences from relatively simple interactions, our ability to predict the impact of more subtle or less obvious indirect effects is not. More research is needed, especially to allow us to forecast the magnitude of effects, and determine whether they are significant in the context of reduced weed abundance should an agent be successful.
“While some indirect non-target effects are inevitable when a new biocontrol agent is established, the key issue is the size and importance of these effects both geographically and over time,” confirmed Simon. If effects are highly localised around the target weeds, or confined to a small period of the year, or only happen while the agent is common and bringing the weed under control (although this might take 5-10 years in typical programmes), then they may not be important in the overall picture.
Biocontrol practitioners are often constrained from delving into studying the ecological systems associated with the release of biocontrol agents by the realities of delivering applied research. However, biocontrol systems offer a wealth of experimental opportunities for ecologists ranging from food-webs, to advanced predator-prey interactions, interspecific plant competition and modelling parasitic relationships. Clearly there is a wealth of fascinating research opportunities that could better inform biocontrol science, and any interest by universities in becoming involved would be welcomed.
This project was funded by the Ministry of Science and Innovation through the Beating Weeds Programme.
Fowler SV, Paynter Q, Dodd S, Groenteman R 2012. How can ecologists help practitioners minimize non-target effects in weed biocontrol? Journal of Applied Ecology 49 (2): 307−310. http://onlinelibrary.wiley.com/doi/10.1111/j.1365-2664.2011.02106.x/full