Impact assessment

We evaluated environmental and socioeconomic impacts using modified versions of the EICAT (Hawkins et al. 2015) and SEICAT (Bacher et al. 2017) protocols. These protocols focus on impacts on native nonhuman populations and human activities, respectively. Primary modifications were the inclusion of beneficial (rather than only detrimental) impacts and use of studies in species’ native and expanded ranges to estimate impacts (rather than only nonnative ranges). These modifications were intended to minimize the influence of study/publication bias, although we acknowledge that researchers historically focused on negative over positive interactions (Bertness and Callaway 1994) and were more likely to study/publish results of strong over weak interactions (Gurevitch and Hedges 1999).

Impacts were classified by mechanism. We identified the following mechanisms as responsible for negative impacts by shifting species on native (nonhuman) species: competition, predation, herbivory, disease transmission, interaction with other invaders, physical disturbance, poisoning/toxicity, and “other” negative impacts (including those with unknown mechanisms). We also found evidence of positive ecological impacts through the following mechanisms: food provisioning, habitat provisioning, and “other” positive impacts. Our SEICAT analysis revealed socioeconomic impacts relating to alterations in health; material and immaterial assets; and social, spiritual, or cultural associations.

We assigned levels of impacts based on categories described in the EICAT and SEICAT protocols (Hawkins et al. 2015; Bacher et al. 2017). Impacts ranged across a semi-quantitative gradient from 1 (lowest) to 5 (highest). For each published study, we scored impacts of shifting species based on the highest-level response from the categories shown in Table 1. Impact scores therefore reflected the maximum impact observed. Both EICAT and SEICAT protocols were modified to incorporate positive impacts, essentially switching the direction or sign of negative impacts (Table 1). Species for which no published papers on impacts were found were categorized as “data deficient”.

For both EICAT and SEICAT assessments, we collected additional information about the shifting species and the study, including taxonomic classifications, study location, and whether the study was conducted in the shifting species’ native (“native”) or nonnative (“expanded”) range. Ranges were defined as “native” or “expanded” based primarily on documentation within the source reporting the range shift (WebTable 1). “Expanded” ranges were designated as such conservatively, acknowledging potential lack of benchmark data for species’ ranges, with most range shifts documented after 1985 (Sorte et al. 2010). We evaluated the relationship between average EICAT impact levels in the native versus expanded range for the seven species that were studied in both range types. This analysis was performed with a linear mixed effect model in R (v4.0.2; R Core Team 2020) using the R package lmerTest (Kuznetsova et al. 2017) with range (native or expanded) as a fixed factor and species as a random effect. Visual inspection of Pearson residuals indicated no deviation from linearity or normality and no major outliers, and there was no deviation from homoscedasticity (Levene’s test, F = 0.0002, P > 0.05). We validated the fit of this model against a model without random effects using Akaike information criteria corrected for small sample sizes (R package nlme; Pinheiro et al. 2020).