Discussion:
Understanding what factors promote and what factors inhibit the establishment of candidate biological control agents is critical if biological control is to transition from a reactionary science to a predictive one (Kimberling, 2004). There has, justifiably, been a major focus on predicting the potential for non-target impacts (Barratt, 2011; Barratt et al., 2010), leading to extensive modeling of the potential extent of a candidate’s introduced range based upon its native host’s range (Barton, 2004; Kaser & Heimpel, 2015; Raghu et al., 2007). Increasingly, studies are conducting pre-release ecological niche modeling to identify geographic regions from which candidate biological control agents might be selected (for some examples see Banerjee et al., 2019; Manrique et al., 2014; Mukherjee et al., 2011; Zalucki & van Klinken, 2006). These types of pre-release comparisons can even help prioritize the suitability of different strains of biological control agents (Manrique et al., 2014), thus reducing both the environmental and political risks of introducing ineffective agents (McClay & Balciunas, 2005). Of course, these types of models are only estimates and make numerous assumptions to quantify and reduce complex biological and ecological processes. For example, models that integrate biological knowledge in addition to climatic variables produce more accurate models than those based on climate data alone (Low et al., 2020). In addition, the choice of climate variables can have important implications on results (see Booth, 2021). Furthermore, climate variables have been found to be more closely associated with some species compared to others, even when the species have overlapping distributions (Shabani et al., 2016). Most importantly, these approaches fail to account for the evolutionary potential of species (e.g., Bean et al., 2012, Diamond, 2018). Yet, despite these limitations, ecological niche models can prove useful as part of a larger discussion of factors that might influence species distributions (Warren, 2012). Here we find that of the three strains of the biological control agent Aphalara itadoricurrently being considered or currently being released for the biological control of invasive knotweed species (Reynoutriaspp.), only the Hokkaido strain is predicted to be suitable in both Europe and North America based on climate comparisons between the current distributions of knotweed species in these two regions and the source localities. The suitability of the other two strains differs by location and target species and are discussed in more detail below.
In our analyses and the analyses published in Andersen and Elkinton (2022), the Kyushu strain is found to have no-to-low climate suitability for any of the target knotweed species in either Europe or North America (Figure 4 and 5; Andersen & Elkinton 2022). The lack of climate suitability of the Kysuhu strain mirrors field observations in North America, where to date, efforts to establish the Kyushu strain have been hindered by both biotic and abiotic factors (Andersen & Elkinton, 2022; Grevstad et al., 2022; Jones et al., 2021; Jones et al., 2020). In contrast, our analyses of the Murakami strain shows that this agent has greater potential, particularly in North America. Analyses, based on the distributions of all three species of knotweed in North America, suggest that the Murakami strain has medium-to-high climate suitability in this region (Figures 4 and 5). In addition, laboratory testing has shown that this strain is capable of laying eggs on all three target species in a choice experiment, and that feeding results in significant reductions for all three species in plant height (8% total reduction), and rhizome biomass for R. × bohemica , and R. sachalinensis(35% and 50%, respectively) (Camargo et al., 2022). Unfortunately, our climate analyses suggest that this strain has no-to-low suitability based on the distributions of all three target species in Europe (Figures 1-3). Analyses based on records of the Hokkaido strain suggest that this agent has at least some climate suitability based on knotweed records from both Europe and North America. We should note that this strain has been shown in the laboratory to have reduced fitness on species of knotweed other than R. sachalinensis (Grevstad et al., 2013); however, it is possible that additional populations feeding on the other target knotweed species might be present on Hokkaido as well, and we encourage further investigations in this region.
In an effort to create datasets with large enough geographic distributions to be statistically meaningful, these continental-wide analyses might mask more localized locations where climate suitability of the different strains might be achieved. For example, despite our previous findings that the Kyushu strain has no-to-low climate suitability for most of North America against R. japonica(Andersen & Elkinton, 2022), in the spring of 2022 we did note the presence of 15 overwintering adults at one field release site in western Massachusetts (Andersen & Elkinton, unpublished data). Similar reports of small numbers of overwintering adults of the Kyushu strain were also reported in coastal Rhode Island (Dr. Lisa Tewksbury, personal communication) and in North Carolina (Dr. Fritzi Grevstad, personal communication), and individuals of the Murakami strain have been observed persisting and dispersing in the field in the Netherlands (Dr. Suzanne Lommen, personal communication), despite our results suggesting they have no or low climate suitability. While records from several more years will be necessary before we can consider these localized populations “established”, they do suggest that even in areas where suitability might be predicted to be low based on our analyses, that persistence and eventually establishment might be possible.
Lastly, we would like to point out an interesting result from our analyses. Our climate models, based on the invasive distributions of each knotweed species, tend to predict low-climate suitability to areas across much of the Japanese archipelago. On one hand, readers should interpret this result as an indication that our ecological niche models are capturing only a portion of the factors that shape the potential and realized niches of an organism (as reviewed above). On the other hand, we believe this result highlights the fact that local adaptation has occurred in this system. Invasive knotweeds have been present in North America and Europe for nearly 200 years (Conolly 1977), and that during that time they have successfully adapted to the local environments in their introduced ranges – this evolutionary potential is likely one of the reasons that they are listed amongst worlds 100 most invasive species (IUCN 2021). Given that coevolutionary forces form the basis of sustainable biological control services (Holt & Hochberg 1997), this potential mismatch between the newly evolved realized niche of an invasive species and the existing potential niche of the natural enemy in its native range, such as A. itadori , could have profound implications on the “success” of a biological control program if the target and the natural enemy no longer share the same climatic constraints.