Introduction
Humans are transporting species faster, farther, and more frequently than ever before, thus providing opportunities for species to establish populations far outside of their native range. These invasive species can have adverse effects on native species and recipient ecosystems (
Ricciardi 2007;
Ehrenfeld 2010). In response, significant effort is directed toward suppressing or eradicating invasive species using a variety of approaches such as pesticide application, mechanical removal, and biological control.
Reducing invasive species populations can mitigate the adverse effects on native species and ecosystems (
Simberloff 2009); however, control actions also have the potential for unintended and harmful effects on native species and ecosystems (
Zavaleta et al. 2001;
Bergstrom et al. 2009;
Rinella et al. 2009;
Lu et al. 2015). Given that both invasive species and invasive species control can have negative effects on native species and ecosystems, there is a clear need to compare their relative effects. Here, we directly compare the ecological effects of an invasive species relative to those of herbicide treatments often used for control. We ask: is the management “cure” for invasive species worse than the disease it is intended to treat? The ability to make this direct comparison is of great value to natural resource managers who often grapple with management trade-offs and are ultimately interested in minimizing negative impacts on native species and ecosystems.
This study examines the nonnative aquatic plant
Myriophyllum spicatum (Eurasian watermilfoil), which in many parts of North America is a notorious nuisance. For example, lakefront property values in the U.S. states of Wisconsin and Washington were 13%–19% lower on lakes invaded by
M. spicatum (
Provencher et al. 2012;
Olden and Tamayo 2014). Recreational impacts following
M. spicatum invasion have also been well documented (
Horsch and Lewis 2009;
Eiswerth et al. 2000). However, results contrast when it comes to
M. spicatum’s ecological effects.
Myriophyllum spicatum, like many invaders, is often assumed to have adverse ecological impacts—and this has been verified in a few studies that examine large or rapidly expanding populations (
Madsen et al. 1991;
Boylen et al. 1999). Other studies stop short of declaring adverse ecological effects, do not link
M. spicatum invasion to reductions in native species across the landscape, or reveal the abundance distributions are not remarkably different from those of native species (
Trebitz and Taylor 2007;
Hansen et al. 2013b;
Gräfe 2014;
Muthukrishnan et al. 2018).
Given the potential for undesirable consequences, herbicide treatments are often used as a management tool to control
M. spicatum populations. Lake-wide chemical treatments (frequently using one of several different formulations of 2,4-dichlorophenoxyacetic (2,4-D) acid alone or in combination with other herbicides) have been shown to produce short-term reductions in
M. spicatum populations (
Kovalenko et al. 2010;
Kujawa et al. 2017). Yet several studies have also found that large-scale herbicide treatments can cause significant declines in native aquatic plants, in addition to the target invasive species (
Wagner et al. 2007;
Nault et al. 2014,
2018).
While research suggests that both invasive M. spicatum and lake-wide herbicide treatment can have negative effects on native plant species, no documentation exists to compare the magnitude of these negative effects. Here, we used a large observational data set for lakes in the state of Wisconsin, USA, and statistically compared the effects of M. spicatum and lake-wide herbicide treatments on native aquatic plant communities. First, we evaluated the impacts of lake-wide herbicide treatment on native plant species using a pre–post comparison that assesses native species declines. Second, we used a multi-level modeling framework to statistically compare the effects of herbicide treatment and the effects of invasive M. spicatum on native aquatic plants. Taken together, we examine whether the negative ecological effect of lake-wide herbicide treatments used to control M. spicatum exceeds the negative ecological effect of M. spicatum. Our results underscore the need for developing a better understanding of the relative impacts of invasive species and the methods that are being used to control their populations.
Discussion
Our study used two complementary approaches to evaluate the ecological effects of the invasive aquatic plant
M. spicatum and the effects of lake-wide herbicide treatments. First, using an extensive set of data on aquatic plant communities in Wisconsin lakes, a pre–post comparison revealed that native aquatic plant species exhibited more significant declines following lake-wide herbicide treatment relative to untreated lakes (
Fig. 2). Pre–post comparisons are a direct and powerful approach for making inferences about ecological effects. Unfortunately, a similar pre–post comparison was not possible for
M. spicatum invasions, since invasions are unplanned, and pre-invasion data are exceedingly rare. Thus, to complement the pre–post analysis for herbicide treatment, we conducted a second analysis using comparative multi-level modeling to statistically compare the effects of herbicide treatment and
M. spicatum on native aquatic plant species and communities (
Jackson et al. 2012,
2014). We found that lake-wide herbicide treatment was negatively associated with native aquatic plant abundance overall with the majority (82%) of individual native aquatic plant species exhibiting a negative coefficient (i.e., negative responses;
Fig. 3). The highly divergent species-specific responses to herbicide treatment suggest that there is an association between lake-wide herbicide treatment and aquatic plant community composition.
Myriophyllum spicatum appears to have a relatively minor effect on native plant species abundance and community composition. In fact, for individual aquatic plant species, the association among
M. spicatum and native species abundance was usually positive: 67% of species-specific
M. spicatum coefficients in the multi-level model were positive. Our findings do not suggest that competitive displacement of native species by
M. spicatum is strong or ubiquitous, at least at a lake-wide, cross-system scale. In communities where competition is a major structuring force, covariance among population abundances is on average expected to be negative (
Houlahan et al. 2007). On the contrary, our findings suggest that factors other than interspecific competition, like facilitation or environmental filtering, may better explain broadscale aquatic plant community patterns. Native species and
M. spicatum may be responding in concert to environmental conditions, or
M. spicatum and other native plants may create conditions that are mutually supportive of aquatic plant establishment and expansion.
At first glance, previous work on the effects of
M. spicatum on native plant communities appears contradictory; evidence exists for negative, neutral, and positive effects. Upon closer examination, negative effects are often reported from local-scale studies on selected lakes or sites within lakes, whereas reports of neutral or positive relationships come from studies conducted on a larger spatial scale (
Boylen et al. 1999;
Trebitz and Taylor 2007;
Gräfe 2014;
Muthukrishnan et al. 2018). This latter explanation is consistent with our study, which failed to discern negative effects of
M. spicatum on native aquatic plants at the lake-wide scale across the landscape.
In a meta-analysis of 199 studies on invasive plant impacts,
Vilà et al. (2011) found that 86% of studies used comparative data to quantify impact, but most of those compared uninvaded sites with sites that were highly invaded. Such a comparison may not be realistic. Highly invaded sites are not necessarily representative, as studies have found that aquatic invasive species are most often present at relatively low densities (
Hansen et al. 2013b). By exploring the association among the abundance of
M. spicatum and native plant species at the lake-wide scale and in many lakes, we present a more realistic picture of the actual impact of the species on the landscape.
Quantifying invasive species effects is difficult for several reasons. Pre-invasion data are often lacking, thus making direct pre–post comparisons nearly impossible. Experimental manipulations of invasive species presence or abundance provide a solution, but these tightly controlled experiments are often expensive, impractical, and offer only a limited perspective on the full community dynamics of a lake. Comparative data sets involving multiple sites or waterbodies, such as those from governmental monitoring programs, are more readily available. Statistical approaches like the one used here provide a path toward rigorous evaluation of a comparative data set.
Quantification of recovery following invasive species control or eradication is another common approach to assessing invasive species effects, but that approach can be problematic as well: eradication is difficult to achieve, and the invaded community may never fully recover to pre-invasion or pretreatment conditions (
Hansen et al. 2013a). Different approaches to understanding both invasive species effects and the effects of their management can yield conflicting results. This underscores the importance of combining multiple lines of evidence, as we do here, when attempting to evaluate or quantify invasive species impacts.
The ecological response metrics used in our assessment relate to native aquatic plant species and communities. There are many other potential response variables that could be used for comparing ecological effects of invasive species and invasive species control. For example,
M. spicatum can change the structural geometry and composition of lake littoral habitat, alter light regimes, and influence lake biogeochemistry (
Madsen et al. 1991;
Barko et al. 1994). While there is little evidence that
M. spicatum directly affects fish abundance, there is support for a significant effect on macroinvertebrates (
Duffy and Baltz 1998;
Kovalenko and Dibble 2011). While we failed to find evidence for
M. spicatum effects on native plant communities, it is important to recognize other potential ecological effects of
M. spicatum, though more work is needed to clarify magnitude and mechanism.
In contrast to the patterns observed with
M. spicatum, lake-wide chemical treatments that are used to control this invasive aquatic plant are associated with significant negative effects on native aquatic plant species abundance and overall aquatic plant community composition. Previous research on the ecological effects of herbicide treatments is variable: some studies report minimal effects on native aquatic plants, whereas other studies observe species declines that can be long-lasting (
Kovalenko et al. 2010;
Wersal et al. 2010;
Nault et al. 2014). Contradictory findings may be explained by the spatial scale of treatment, water chemistry, and differences in herbicide products, rates, and exposure time (
Frater et al. 2016;
Nault et al. 2012,
2018). Our study uses data from many aquatic plant communities to reveal evidence that lake-wide herbicide treatments may be associated with ecological effects on nontarget species and aquatic plant communities. The paired analyses do suggest a treatment-related effect, but it is important to realize that there may be uncaptured factors that contribute to the patterns we observed. Accounting for environmental variation and matching treated lakes to untreated control lakes that had similar plant communities were two important steps that contribute to the strength of our inferences, but there may yet exist underlying causal factors common to treated lakes that may not be directly related to the lake-wide application of herbicide.
Our study associates lake-wide herbicide treatments with nontarget effects on native aquatic plants, but the timing and longevity of these effects is unknown. We should track species abundance and plant community change after lake-wide herbicide treatment for multiple years to identify whether observed ecological effects last. Unfortunately, treated lakes in this study were typically subjected to follow-up management actions after the initial treatment, which limited our ability to explore this question. We conjecture that if native species fail to recover from lake-wide herbicide treatments as quickly as
M. spicatum, the invasive species may continue to present a management problem despite ongoing investment in control, leading to synergistic negative effects on native species (
Rinella et al. 2009). In light of our findings, we recommend an adaptive, integrated, pest-management approach that utilizes diverse strategies to achieve management goals, especially given that some commonly utilized aquatic herbicides (i.e., 2,4-D, fluridone) have been associated with milfoil hybridization events and increased herbicide resistance (
Thum et al. 2012;
Larue et al. 2013;
Berger et al. 2015;
Gill and Goyal 2016).
In conclusion, whether the lake-specific effects of the invasive species are adverse and severe enough to justify the risk posed by herbicide treatment deserves much more careful consideration than has occurred in the past. Lake management decisions must consider diverse stakeholder values and ecological health, and our work provides insights that may be incorporated into aquatic plant management decision-making frameworks (
Kumschick et al. 2012). We conclude that unless there is strong evidence of high ecological, social, or economic impact for an invasive aquatic plant, aggressive chemical control at a lake-wide scale might do more harm than good.
Acknowledgements
We thank the many Wisconsin Department of Natural Resources staff, partners, and collaborators for collecting the data analyzed in the present study, including Meghan Porzky, Brenton Butterfield, Shauna Chase, Jesse Schwingle, Raffica La Rosa, Todd Hanke, Dan Cibulka and Nicholas Shefte. We thank Tim Asplund, Heidi Bunk, Murray Clayton, Paul Frater, Claudio Gratton, Mary Gansberg, Susan Graham, Kevin Gauthier, Ted Johnson, Brenda Nordin, Scott Provost, Tony Ives, Susan Knight, Carroll Schaal, Alex Smith and Pamela Toshner for their critical input and support for the project. This material is based upon work supported by the Wisconsin DNR under grant No. ACEI-060-09 as well as the National Science Foundation Graduate Research Fellowship, under grant No. DGE-1256259. Any opinion, findings, and conclusions or recommendations expressed in this material are those of the authors(s) and do not necessarily reflect the views of the National Science Foundation.