Estimating potential global sources and secondary spread of freshwater invasions under historical and future climates

We employed a climate‐matching method to evaluate potential source regions of freshwater invasive species to an introduced region and their potential secondary spread under historical and future climates.


| 987 1 | INTRODUC TI ON
Invasive species are a major cause of global biodiversity loss (Bellard et al., 2016), can transform biogeography (Capinha et al., 2015) and lead to significant economic disruption within the regions they invade (Diagne et al., 2021). Recent invasions have been dominated by newly recorded alien species, due to the growth of economic trade among source regions of new exotic species and receiving regions (Seebens et al., 2018). Climate change is predicted to create more favourable conditions for the survival of some freshwater species in receiving regions that were previously unfavourable (Britton et al., 2010;Rahel & Olden, 2008). Due to the difficulty of implementing successful control or eradication measures to manage invasions after introductions, prevention remains the most cost-effective method of invasive species management (Leung et al., 2002;Lodge et al., 2006Lodge et al., , 2016. Prevention of non-native species introductions requires management frameworks that consider potential species and/or their source regions, introduction vectors  and the implications of climate change on regional species survival, establishment and impact (Kriticos, 2012).
Climate change has begun altering ecological processes including species invasions, and it is likely that as climate change intensifies, so too will the alterations to these ecological processes (Walther et al., 2002). Climate change is predicted to alter regional air temperatures, and variability in precipitation regimes that, in turn, will lead to increases in global and regional water temperatures and altered hydrological cycles in many regions.
These climatic changes are predicted to lead to shifts and contractions in freshwater fish distributions (Bond et al., 2011;Buisson et al., 2010;Chu et al., 2005;Lyons et al., 2010). Regional climatic conditions also act as a strong filter that may affect the establishment of species with human-mediated distributions (Magnuson et al., 1997;Rahel & Olden, 2008). This filter may be diminished or strengthened with changing climatic conditions and, in some cases, lead to opportunities for some introduced species to persist where they were previously not capable of surviving (Rahel & Olden, 2008). As a result, there is an emerging need to understand how future climate change will impact the stages of the invasion process .
Researchers and managers often assess the probability of nonnative species survival or establishment in an area of invasion by quantifying the similarity in climate between the native and exotic ranges of potentially invasive species, which is known as climate matching (Bomford et al., 2010;Britton et al., 2010;Howeth et al., 2016). Determining from where invasive species may originate can incorporate a geographic lens by examining the climate match of a receiving region to potential global source regions (Kriticos, 2012), which is useful for predicting individual invasive species and potential source pools of invasive species, such as regions with trade pathways (e.g. shipping ports; Keller et al., 2011;Seebens et al., 2013). Freshwater ecoregions are biogeographic units representative of the ecosystems and evolutionary processes that represent the distinct freshwater fish communities of the world (Abell et al., 2008) and are, thus, a suitable unit of spatial delineation for climate-based analyses. Incorporation of climate projections into climate-matching assessments has been used to predict the survival of potential aquatic invasive species to Britain (Britton et al., 2010) and global source regions of terrestrial weed species to New Zealand (Kriticos, 2012). Incorporating climate projections with climate matching can be broadened to include analyses of potential secondary spread of invaders beyond an area of initial introduction. Such analyses would provide valuable information to biosecurity managers to minimize or avoid the impacts of future invasive species.
The Laurentian Great Lakes ecoregion is one of the most highly invaded aquatic ecosystems in the world (GLANSIS, 2022;Pagnucco et al., 2015;Ricciardi, 2006). There have been 69 recorded nonnative freshwater fish species introductions, 35 of which have established reproducing populations (GLANSIS, 2022;Mandrak & Cudmore, 2010). These species have had significant negative impacts to Great Lakes ecosystems, including the extirpation of native species (Mandrak & Cudmore, 2010). Freshwater non-native species have been introduced to the Great Lakes through a variety of pathways including canals, legal and illegal stocking, commercial shipping (ballast water) and intentional release from live trades such as aquarium, pond, food and bait markets (Davidson et al., 2021;Mandrak & Cudmore, 2010). Many of the existing freshwater invasive species in the Great Lakes originated from temperate regions of Eurasia (Kolar & Lodge, 2002;Mills et al., 1993). The Great Lakes has been a beachhead of invasive species, as it has been a location of first introduction of non-native species, some of which have subsequently spread to other North American ecosystems (Rothlisberger & Lodge, 2013).
Secondary pathways, such as shipping and recreational boating, have led to 21 of 65 beachhead species, representing a variety of taxa, to spread beyond the Great Lakes, leaving 44 species poised for introduction or spread throughout North America should they follow the beachhead process (Rothlisberger & Lodge, 2013). The Great Lakes has a temperate climate and, as climates warm, species that previously were unable to survive due to the cold may be able to survive if introduced. Climate change will also affect regions into which invasive species may spread, creating potential opportunities for existing and future non-native species to spread beyond the Great Lakes. The high level of freshwater species invasions and subsequent spread make the Great Lakes ecoregion a model study system for the development of methods and tools for assessing global sources, and primary and secondary spread, of non-native species that incorporate climate change.
Here, we examine all global freshwater ecoregions as possible sources of freshwater invasive species to a recipient ecoregion and the vulnerabilities of all North American ecoregions that may experience the spread of invasive species beyond the recipient region through secondary pathways. Further, we assess the implications of incorporating climate-change projections in the predictions. We apply a climate-match model to the Laurentian Great Lakes freshwater ecoregion as the primary recipient region with global ecoregions as potential sources. We also include the Great Lakes ecoregion as a possible source to all other North American ecoregions as recipients, representing the regions where spread would not be hindered by climate. Our objective was to provide a method and guidance for the assessment of regional vulnerabilities of introduced non-native species under historical and future climatic conditions, which will aid managers, policy-makers and researchers in identifying invasion risks. Further, we aimed to provide an assessment of our model system, the Laurentian Great Lakes, that can be used for determining potential sources of freshwater nonnative species under historical and future climatic conditions in risk assessments of invasive species and their pathways.

| Climate data
Climate data (WorldClim v2.1 and six global climate models) were accessed via the WorldClim platform (www.world clim.org).

| Ecoregions
To conduct our climate-match assessment of an area of primary introduction to all possible global sources and where secondary spread would not be hindered by climate, we used the biogeographic units of Freshwater Ecoregions of the World (mean surface area = 312,315 km 2 ; n = 426). Freshwater ecoregions encompass distinct biodiversity features at large scales that are widely regarded as useful for conservation planning (Abell et al., 2008). In this analysis, the ecoregion of primary introduction was the Laurentian Great Lakes, and the remaining ecoregions of North America were assessed as regions of potential spread should freshwater species follow the beachhead process.

| Climate matching
To quantify the similarity in climates between possible source and recipient regions, we used the Euclidean distance-based measuring algorithm, CLIMATE (Crombie et al., 2008;Pheloung, 1996), which computes a standardized distance between a source region and a recipient region and has been empirically validated for predicting the establishment of introduced non-native freshwater fishes (Bomford et al., 2010;Howeth et al., 2016). The algorithm calculates a score from 1 (very poor match) to 10 (perfect match) for each climate station or grid cell within a recipient region, which denotes the closest match of any grid cell in the source ecoregion. Following the methods of Bomford et al. (2010) andHoweth et al. (2016), the percentage of grid cells within the recipient region with a match of six or higher to the source region was the final retained climate-match value. Howeth et al. (2016) found that a climate match of ≥71.7% between the native range of introduced non-native freshwater fish species and the recipient Great Lakes predicted the establishment of species populations in the region with 85% accuracy (i.e. predicting successful vs. failed establishment of introduced freshwater fishes) using classification tree analysis. Therefore, we used the climate match of ≥71.7% to indicate ecoregions whose freshwater fishes, if introduced to the Great Lakes, would not be hindered by climate. For the secondaryspread analyses, we relaxed the threshold value to ≥50% because the climate-match threshold is unlikely to be the same for all North American ecoregions due to differences in spatial size and heterogeneity in climate. Hereafter, for the analyses with the Laurentian Great Lakes as the potential recipient ecoregion, ecoregions that entered the climate-matching algorithm as source ecoregions were termed possible source ecoregions and the subset of possible source ecoregions with a climate match above the thresholds were termed high-match source ecoregions. For the spread analyses, ecoregions of North America other than the Great Lakes were termed possible recipient ecoregions with the subset of ecoregions with a climate match above the threshold as high-match recipient ecoregions.
We conducted the climate-matching analysis using climate data of historical and future climatic scenarios representing possible ranges and climatic niches of potentially invasive species, similar to Kriticos (2012). as potential recipients of freshwater invasive species. A sensitivity analysis was conducted to examine the impact of the thresholds used in our analysis on the predictions of the ecoregions that hold or will hold species whose survival would not be hindered by climate in the recipient region (i.e. high-match sources). For this analysis, we counted the number of ecoregions that were equal to or greater than climate-match thresholds at 5% increments from 5% to 100%. This sensitivity assessment was repeated for each of the four climate scenarios described above and for the Laurentian Great Lakes as a recipient ecoregion to global ecoregions as possible sources and for the Great Lakes as a source to North American ecoregions as recipients. The climate-match threshold sensitivity analysis revealed that the number of high-match sources was often affected by the threshold used ( Figure S1). With Scenarios 1 and 4, at each 5% increment increase in the threshold there was a reduction in the number of high-match source ecoregions ( Figure S1). For Scenarios 2 and 3, there was no clear pattern in the numbers of possible source ecoregions predicted to rise above the threshold and become high-match sources under climate change ( Figure S1). The threshold with the highest numbers of high-match sources was 75% with 14 high-match sources for Scenario 2, and the 80% threshold had the greatest number (9) of high-match sources for scenario 3. The threshold sensitivity analysis for spread under scenario 1 had 34 high-match recipient ecoregions up to a threshold of 50%, followed by declining number of high-match recipients above the 50% threshold ( Figure S2a). For Scenario 2, climate-match thresholds of 65%-80% had the greatest number of high-match recipients with 4 ( Figure S2b). Under Scenario 3, the thresholds with the greatest number of high-match recipient ecoregions were 75%, 80% and 95%, with six high-match recipients ( Figure S2c). Under Scenario 4, there was a trend of declining numbers of high-match recipient ecoregions with increasing climatematch thresholds ( Figure S2d).

| DISCUSS ION
Adaptation to climate-change impacts is a leading strategy for reducing ecological exposure and vulnerabilities to climate change (IPCC, 2022). Adaptation planning involves the assessment of vulnerabilities, including those due to biological invasions resulting from climate change. In this study, we describe methods for predicting origins of freshwater invasive species and their potential spread beyond areas of original introduction using climate matching. Using the Laurentian Great Lakes ecoregion as a model system, we cartographically represented climate matches between all global ecoregions as possible sources of freshwater non-native species and a receiving ecoregion under historical and future climate-change scenarios that have immediate applications for species and pathway risk assessments. We demonstrate the use of a climate-matching tool using a biogeographic perspective that incorporates climate-change projections to provide an assessment of potential geographic origins of freshwater biological invasions. Further, we demonstrate a method of climate matching that incorporates secondary spread of non-native species from an initial point of establishment in the Laurentian Great Lakes, which is a beachhead for the spread of aquatic invaders in North America (Rothlisberger & Lodge, 2013).
The ecoregions and the species that occupy them, with high  (Love et al., 2015) and have been identified as high-risk invasive species to the Great Lakes (Cudmore & Mandrak, 2005). Species native to those ecoregions may also become invasive if they move beyond their native range (Mandrak, 1989). Under the climate-change scenario 2090 SSP5-8.5, a new continent, South America, was predicted to have three ecoregions become potential sources of freshwater invasions to the Great Lakes where no ecoregions had a climate match ≥71.7% under historical climates. These ecoregions in South America were Lower Paraná, Lower Uruguay and Bonaerensean ecoregions (Abell et al., 2008) and they have 331, 230 and 42 known freshwater fish species, respectively (Albert & Reis, 2011), many of which are in the aquarium trade (Mandrak & Cudmore, 2015). As only some of these species may have an impact if introduced to the Great Lakes under climate change, assessment of invasion risks of these species to the Great Lakes is warranted, particularly those species likely to occur in invasion pathways that have greater probability of propagule release, such as the live trades (Blackburn et al., 2015;Lockwood et al., 2005). Further, some ecoregions with historical climate matches above the threshold increased in climate match under the projected climate scenarios, creating more favourable climatic conditions for the species inhabiting these potential source regions leading to reduced limitations of introduced propagule survival. Climate change will likely impact the spread of invasive species beyond regions of primary introduction that occurs via secondary pathways such as natural dispersal and recreational boating and other vectors Lodge et al., 2016). Climate change provides new opportunities for native and non-native species of the Great Lakes to survive in other areas of North America. Species currently in the Great Lakes will have greater climate matches in the ecoregions located in northern Canada (Western Hudson Bay, Central Arctic Coastal), which supports current predictions of climate change as a major driver of biological invasions in Arctic regions (Essl et al., 2020). Climate change is also expected to increase shipping lanes and trade connections at higher latitudes, facilitating new connections to source pools of non-native species and increased species survival during transport due to shorter travel times (Melia et al., 2016). In addition to climatic suitability, the pathways and vectors that result in species introductions are important considerations in biological invasion risk assessment. Historically, shipping ballast water has been the most prominent invasion vector to the Great Lakes, although regulations have reduced species introductions resulting from this vector (Ricciardi & MacIsaac, 2022). Still, arrival and establishment of species has not been completely prevented (Drake et al., 2020), so it will be important to understand the survival of potentially introduced species that could result from the primary, secondary and tertiary shipping ports located in the South American ecoregions (e.g. Lower Paraná; Keller et al., 2011) whose source pools of non-native species may be invasive in the Great Lakes should they be introduced in the future. Further, live trade has been identified as a prominent pathway for freshwater species introductions currently and is expected to continue due to expansion or growth in globalization, economic trade, urbanization and a lack of effective tools to mitigate introductions (Chan et al., 2019;Chapman et al., 2017;Howeth et al., 2016;Lockwood et al., 2019). Specifically, e-commerce for the aquarium trade has been identified as a vector likely leading to increased freshwater non-native species introductions in the future (Reid et al., 2019). White Cloud Mountain Minnow (Tanichthys albonubes), native to ecoregions in South America predicted to have a high match to the Great Lakes under climate change, is a highly traded species in the aquarium trade, meaning it will have greater numbers of propagules, resulting in a higher probability of introduction. This species is predicted to pose a risk of invasion in the Great Lakes under historical climatic conditions due to its low thermal tolerance of 5°C (Rixon et al., 2005) and had an estimated propagule pressure of 117 individuals per year released in the St.
Lawrence Seaway, a waterway connected to the Great Lakes (Gertzen et al., 2008). The freshwater ecoregions of the world represent distinct units of biodiversity, reflecting major ecological and evolutionary patterns of each region, with hydrologic drainages that act as natural barriers to species movement (Abell et al., 2008). In response to climate change and other global changes, such as urbanization, ecoregions and their delineations may shift, homogenize or be remade anew as species distributions change (Chu et al., 2005;Comte et al., 2013) and may require updating for future conservation assessments (Abell et al., 2008). Nevertheless, the attributes of ecoregions offer utility for biological invasion risk assessments and research under historical and future climate scenarios. For individual species, climate matching using species occurrence within terrestrial ecoregions has been found to be more predictive of non-native species establishment for herpetofauna than point-locality data available in We recognize that our study is limited to the consideration of climate and climate change on the invasion process using climate matching. Despite climate matching providing useful predictions of freshwater invasion outcomes (Bomford et al., 2010;Bradie et al., 2015;Howeth et al., 2016;Lodge et al., 2016) In assessments of ecological impacts resulting from climate change, uncertainties arise from the climate data used in the assessment. Uncertainty in socio-economic development and future greenhouse gas emission policies complicates the application of GCMs to predict ecological impacts resulting from climate change.
One approach is to assume all emissions scenarios as being equally likely due to a lack of sufficient information to determine which scenario is most likely to occur (Wigley & Raper, 2001), although some research has suggested that the use of the highest emissions scenarios is the most precautionary for use in impact studies (Rahmstorf et al., 2007). In this study, we adopt a risk-averse approach and, thus, used the highest emissions SSP. However, because the SSPs represent broad narratives in climate and societal change that also affects mitigation and adaptation (O'Neill et al., 2017), there is uncertainty in predicting the scenarios that will bring the most severe impacts to non-native species spread and impact. Further uncertainty stems from the GCMs used in the assessment. The advantage of a climate ensemble is the reduction in bias and uncertainty associated with individual GCMs (Fenech et al., 2007); however, this can make it difficult to interpret the full range of possible impacts. An operationalized strategy could incorporate a wider array of stand-alone GCMs (i.e. without averaging) and multiple SSPs to provide a more complete insight to the range of possible source regions and potential spread of non-native species in a changing climate.
The purpose of this study was to develop and apply an ecoregionscale climate-matching procedure for the assessment of possible sources of primary and secondary species introductions, under baseline climates and climate-change scenarios. These procedures are an important step in developing a framework for evaluating climate-change impacts on invasion risk from a biogeographic perspective. Understanding possible sources of freshwater invasive species for an ecoregion under future climate scenarios is valuable for risk assessors to help anticipate future invasions and inform invasive species management. These findings underscore the importance of considering the impacts of climate change on the survival of freshwater non-native species to inform invasive species management decisions and regulatory actions such as limiting transport or sale of species (Mandrak & Cudmore, 2015). Our analysis provides guidance for climate-match assessments with applications for rapid and screening-level risk assessments, horizon scanning and other methods used to evaluate the risk associated with non-native species and their pathways.

ACK N O WLE D G E M ENTS
This project was funded by NSERC Strategic Partnership Grants to Nicholas E. Mandrak. We thank two anonymous reviewers whose comments improved the clarity of our manuscript.

CO N FLI C T O F I NTER E S T S TATEM ENT
The authors declare no conflict of interest.

PEER R E V I E W
The peer review history for this article is available at https:// www.webof scien ce.com/api/gatew ay/wos/peer-revie w/10.1111/ ddi.13695.

DATA AVA I L A B I L I T Y S TAT E M E N T
The datasets that support the findings of this study are available in