What's the best way to document the impacts of biocontrol agents on weeds?

It’s a smackdown contest between monitoring and population modelling. Spoiler alert: they’re both important.

written Aug 31, 2014 • by Jon Sullivan • Category: Wild Soapbox

tangle banner

I was at a workshop earlier this week on the biocontrol of New Zealand weeds. Weed biocontrol is where a specialist insect or pathogen from overseas is released, after lots of safety testing, in an attempt to control a weed. Sometimes biocontrol is spectacularly successful. Other times it’s less effective. However, taken as a whole, the endeavour is an extraordinarily cost effective way to deal with bad weeds.

You might be surprised that the detailed impacts of biocontrol agents on their target weeds, and the wider environment, are usually poorly known. That’s because government agencies and industry fund the identification, testing, and releasing of agents, but rarely fund studies of impacts. Most funders prefer to invest their money in the next biocontrol programme for the next most pressing weed. Agents that establish either cause big, obvious weed declines, or they don’t. Dice rolled. Move on.

The workshops were called by NZ regional councils to figure out how best to increase their knowledge of biocontrol impacts. Councils now need to produce quantified, costed impacts of biocontrol agents to meet the greater regulatory requirements from central government. They need better numbers on the bang for their biocontrol buck and better numbers on environmental safety.

To answer how best to do this, they ran a series of workshops throughout NZ to ask scientists. One thing they wanted feedback on was whether citizen science would be sufficient to monitor the impacts of biocontrol agents on weeds and the wider environment.

iNaturalist NZ

In the workshop I attended, the discussion became an interesting debate on the relative merits of monitoring and modelling. Monitoring is good for documenting environmental changes, often showing unexpected changes, but it doesn’t necessarily reveal the causes of these changes. You’ve also got to monitor over many years at many places to reveal trends and patterns. (The WildCounts project is all about getting more people monitoring their corner of the wild.)

In contrast, population modelling can be built using parameters, gathered from just a few years of detailed field and lab observations and experiments. These models can predict population dynamics in future times and undocumented places, with different densities of the biocontrol agent. Models don’t anticipate the unexpected, but they can show us all the implications of what we know. That can be a powerful thing.

Dave Kelly from the University of Canterbury convincingly made the case that mathematical population modelling was the best way to quantify the impacts of a biocontrol agent on its host weed populations. Dave and his colleagues have done this successfully modelling nodding thistle biocontrol. More science like this would document how other biocontrol agents impact their host weeds, at least in the subset of environmental conditions that can get quantified for the models.

A flower head of the pasture weed, nodding thistle, Carduus nutans. Population modelling has successfully quantified the impacts of biocontrol agents on these thistle populations in New Zealand.

Using monitoring to document biocontrol agent impacts in weeds takes much longer than population modeling. It is complicated by the many other environmental changes in land use, climate, and species invasions, that can affect weeds. Without a lot of replication, the exact impacts of the agent on the weed can be hard to unravel from the noise of everything else that’s changing.

Does this mean that there’s no need to monitor? No! Monitoring can reveal how the weed, the biocontrol agent, and the environments and species they interact with, are all changing in time and space. This level of complexity in nature is difficult (and expensive) to predict with mathematical population modelling, if that’s even possible. It’s better to monitor those changes and build statistical models that can summarise and extrapolate the trends and patterns into other places and times.

An excellent example of the value of monitoring weed control is an insightful study by Jane Barton and colleagues in Auckland. They studied the recovery of stream side vegetation in Auckland’s Waitakere Ranges after the successful biocontrol of the weed mistflower (Ageratina riparia). Mistflower is a fast growing perennial that had taken over stream banks and open areas.

Two biocontrol agents were released to control mistflower, and they did their job spectacularly well. At the time of their release, it was feared that a successful biocontrol programme would just lead to mistflower being replaced by other weeds. The monitoring by Jane and colleagues showed that this didn’t happen. Instead, native riparian vegetation reestablished after the demise of mistflower. The agents have been a complete success.

Ageratina riparia
A dense patch of mistflower in Auckland, prior to its successful biocontrol. Mistflower used to choke the sides of forest streams in Auckland. After the introduction of two biocontrol agents, it's now difficult to find dense patches of the weed like this, and native plants have returned to dominate many of the forest stream sides.

Would citizen science be sufficient to monitor the impacts of biocontrol agents on weeds and the wider environment? I think the general consensus of the room was that it would be quicker, and likely more cost effective, to fund postgraduate students to build population models of agent impacts. Wider monitoring of weeds is still important, and should be part of councils’ environmental monitoring. There are many advantages to involving citizens in this monitoring, from increasing the number and variety of locations monitored, to the benefits of engaging the public in biocontrol programmes.

It is also important that citizen science monitoring complements, rather than supplants, a core of consistent and regular professional monitoring. Both types of monitoring together are more powerful than either alone. Professional monitoring generates cleaner and more consistent data, ideal for analysing trends, but these data are by necessity from relatively few sites. Citizen science monitoring can add the bigger picture of how consistent these trends are across a much wider range of sites and environmental conditions.

To understand environmental changes caused by the introduction of biocontrol agents, it makes sense to deploy a mix of modelling with professional and citizen monitoring. It will be interesting to see whether the recent shift in councils’ regulatory responsibilities will see a larger proportion of biocontrol funding being directed to understanding what happens after agents are released.