Consumptive and non-consumptive predator effects from native fish but not invasive crayfish on Ambystoma salamander larvae

Introduced species impose a variety of negative impacts on native biodiversity, including as consumers of native species. Invasive consumers can also impact native species indirectly through non-consumptive predator effects, where mere presence of the predator causes a reduction in prey fitness. Yet, when prey lack strong responses to non-native predators, non-consumptive effects may be imposed only by native predators. White River crayfish (Procambarus acutus) were recently introduced into Canada, including on Pelee Island, Ontario which is home to the country’s only population of small-mouthed salamander (Ambystoma texanum). Leveraging data from 105 surveys conducted over four years in 36 distinct waterbodies we test for evidence of consumptive and non-consumptive effects of P. acutus on larval Ambystoma salamanders. We concurrently evaluate, and control for, consumptive and non-consumptive effects from native fish species, density dependence, water temperature effects, and variation in habitat conditions. Additionally, we develop a habitat suitability map for P. acutus to identify potential areas of interaction with salamanders and to guide future monitoring. We found considerable overlap between areas of high habitat suitability for P. acutus and areas where Ambystoma salamanders occur. Between 2022–2025 there was a significant increase in the proportion of surveyed waterbodies where P. acutus was detected and a concomitant decrease in the proportion of surveyed waterbodies where Ambystoma larvae were detected. Nevertheless, we found no evidence that presence of P. acutus was associated with a reduction in catch-per-unit-effort (CPUE, larvae/sweep) or larval snout-vent length. In contrast, native fish presence was associated with an 8.4-fold reduction in CPUE and a 6.6% reduction in snout-vent length. Larval body size was also negatively impacted by high conspecific density and warmer water temperature. While we are relieved that P. acutus does not appear to affect salamanders in the ways we investigated, we remain concerned about other routes through which P. acutus may impact Canadian amphibians and other native biodiversity. Determining the true impacts of introduced species on native ecosystems hinges on objective evaluation of threats and remains essential to and safeguarding native biodiversity in human-affected landscapes.

The introduction of non-native species is a major threat to biodiversity globally (Bellard et al., 2016a; Doherty et al., 2016; Hulme et al., 2010). Introduced species can negatively affect ecosystems by outcompeting or consuming native species (e.g., Jensen et al., 2002; Pille et al., 2021), disrupting trophic systems (e.g., David et al., 2017; Hall and Mills, 2000), and developing large populations in areas beyond the range of their natural predators (Gherardi, 2007; Hall and Mills, 2000). These species may present novel and significant problems for species that are already at risk, leading to further declines or extirpation (Bellard et al., 2016b; Berglund et al., 2013; Dueñas et al., 2021). Yet, not all introduced species have demonstrated negative impacts on their new ecosystems (e.g., Bucciarelli et al., 2014; Nagy et al., 2011), and once an invasive species becomes established the costs and disturbances incurred from their removal may outweigh the benefits (Westman, 1990). It is therefore important to objectively evaluate the risks posed by newly detected non-native species.

Introduced species represent an important global threat to amphibian populations, and negatively impact populations though predation, competition, hybridization, habitat alteration, or through the spread of disease (Kats and Ferrer, 2003; Falaschi et al., 2020). Amphibian strategies for avoiding predators, such as reducing activity or increasing use of refuges, may not be effective against novel predators, and evolutionarily-naïve species or populations may not respond appropriately to introduced consumers (Cruz et al., 2006; Cruz and Rebelo, 2005; Davis et al., 2017; Kats, 1988). Even species that can respond to a predation threat from novel consumers may also experience non-consumptive predator effects, which are changes in prey traits related to the perceived risk of being consumed, which often have significant negative impacts on prey fitness (Preisser et al., 2005; Wirsing et al., 2020). In amphibians, such effects can take the form of predator-induced reductions in body size (e.g., Kenison et al., 2016; Werner and Anholt, 1996), which is particularly impactful on fitness given that larger body size is a good predictor of survival and fecundity in amphibians (Berven and Gill, 1983; Blackwell et al., 2003; Székely et al., 2020; Walls and Altig, 1986). In addition, larval and adult salamanders are gape-limited predators, and larger individuals have a competitive advantage because they are able to utilize a wider range of food resources (Wilbur and Collins, 1973; Zellmer, 2010). Many amphibians can also increase their developmental rate in the presence of predators and undergo metamorphosis at smaller body sizes (Melotto et al., 2020; Relyea, 2007), which could reduce survival of emerging metamorphs (Székely et al., 2020; Werner, 1986).

Non-native crayfish have demonstrated negative effects on aquatic ecosystems globally (Lynas et al., 2007; Twardochleb et al., 2013; Wilson et al., 2004). The red swamp crayfish (Procambarus clarkii), for example, has negatively affected amphibian breeding and larval survival in its introduced ranges (Cruz et al., 2006; Cruz and Rebelo, 2005; Gamradt and Kats, 1996; Pinto, 2022). There is renewed concern over the spread of invasive crayfish in Canada, with recent detections of several Procambarus species in Ontario and elsewhere in the country (Hamr and Schryer, 2023; Hossie and Hamr, 2022; Schryer and Hamr, 2024). The potential impacts of these non-native invaders on Canadian ecosystems, however, remains largely unknown, leaving conservation practitioners unprepared to mount appropriate responses to mitigate their effects. Previous work has found that amphibians, and particularly salamanders, may be highly vulnerable to the consumptive effects of introduced crayfish (e.g., Ficetola et al., 2011; Gherardi et al., 2001). When novel crayfish predators are present, salamander larvae might increase the amount of time engaged in vigilance or hiding behaviours instead of foraging, leading to reductions in growth (Cruz and Rebelo, 2005; Davis et al., 2017; Nunes et al., 2014; Relyea and Werner, 1999). Therefore, both consumptive and non-consumptive predator effects of introduced crayfish are potentially of concern for imperilled amphibian populations.

The small-mouthed salamander (Ambystoma texanum) is listed as endangered under Canada’s Species at Risk Act (COSEWIC, 2014). Pelee Island, Ontario is the northernmost limit of their distribution and their only population in Canada, where this species co-occurs with a federally endangered population of unisexual Ambystoma (Hamill, 2015; Hossie, 2018). Salamanders on Pelee Island were isolated from mainland populations roughly 4000–6000 years ago as melting glacial waters created what is now Lake Erie (Herdendorf, 2013). This population is primarily threatened by the loss, alteration, and fragmentation of their habitat, but also faces threats from road mortality and invasive species (Environment and Climate Change Canada, 2020; Kolozsvary and Swihart, 1999). White River crayfish (Procambarus acutus) were not historically present on Pelee Island and were first detected on Pelee Island in 2015, possibly having been introduced through bait bucket releases (Hossie and Hamr, 2022). These crayfish have now been detected in several waterbodies on Pelee Island where A. texanum breed (Hossie and Hamr, 2022). Although the salamanders on Pelee Island have coexisted with two native crayfish species (i.e., Creaserinus fodiens, Faxonius immunis; Hossie and Hamr, 2022), the severe impacts of introduced Procambarus crayfish on amphibian populations documented elsewhere in the world gives legitimate cause to be concerned about how P. acutus might impact this imperilled salamander complex.

Our study seeks to test the hypothesis that the White River Crayfish (Procambarus acutus) negatively affect Ambystoma salamanders through consumptive and non-consumptive effects. Specifically, we predict that the consumptive effects of P. acutus will result in a lower catch per unit effort (CPUE) of salamander larvae in ponds where P. acutus is present. Additionally, we predict that non-consumptive effects of P. acutus on salamander larvae will be manifested through a smaller larval snout-vent length (SVL) in ponds where P. acutus are present. To test these predictions, we analyse data collected during 105 dipnet surveys spanning 36 distinct waterbodies across Pelee Island between 2022-2025. Additionally, we generated a habitat suitability map for P. acutus on Pelee Island to identify potential areas for targeted monitoring and mitigation and to examine potential overlap with known salamander habitat. Importantly, this is one of the only studies of crayfish on Pelee Island (but see Crocker and Barr, 1968; Hossie and Hamr, 2022), and the first exploration of how introduced populations of P. acutus may be impacting Canada’s biodiversity.

Study location

Pelee Island, Ontario is a 42 km² island located in the western basin on Lake Erie. The island’s landscape was dramatically altered in the 1890s through extensive wetland drainage and subsequent tree removal. It is now approximately 62% agricultural land, with only 6% wetlands and 17% forests (Hossie, 2018; Hossie and Hamr, 2022). The island is home to Canada’s only population of small-mouthed salamanders (Ambystoma texanum). Other Ambystoma species that occur on the island include the blue-spotted salamander (A. laterale) and unisexual Ambystoma (Bare et al., 2023; Hossie, 2018). These three species are visually indistinguishable during the larval stage. In addition to P. acutus ¸ two native species of burrowing crayfish occur on the island: Digger crayfish (Creaserinus fodiens Cottle, 1863) and Calico crayfish (Faxonius immunis Hagen, 1870) (Hamr, 2024a, 2024; Ross, 2023). Northern Clearwater crayfish (Faxonius propinquus Girard, 1852) were recently discovered on the Lake Erie shorelines on Pelee Island through iNaturalist observations (Sinclair, 2025) but have not been observed in the island’s interior. The three burrowing crayfish (P. acutus, C. fodiens, F. immunis) construct burrows which can be distinguished visually (Schryer and Hamr, 2024). Salamanders on Pelee Island use crayfish burrows as terrestrial retreats, and possibly also as overwintering sites (Hossie and Hamr, 2022). Indeed, Ward and Hossie (2020) found a positive association between the presence of crayfish burrows and the presence of salamander larva in ponds on Pelee Island, possibly indicating that the burrows crayfish construct may facilitate or support occupancy by salamanders.

Field methods

We surveyed waterbodies across Pelee Island from 2022-2025. Each waterbody was sampled for Ambystoma salamander larvae with a systematic dip net approach of 80 sweeps, following methods described in Ward and Hossie (2020). Dipnet surveys for larvae were typically conducted in late May and early June, however logistical constraints in 2023 resulted in the larval surveys being delayed until July. We surveyed 31 waterbodies in 2022, 12 in 2023, 32 in 2024, and 30 in 2025 (see Supplementary Material). A total of 36 distinct waterbodies were surveyed, 31 of which were surveyed in multiple years. In each survey, the number of salamander larvae caught was recorded and converted to catch per unit effort (CPUE; total number of individuals caught/number of sweeps). Using digital callipers, we measured the snout-vent length (SVL) of up to 30 salamander larvae from each waterbody each year. In 2022, larval body size data was instead collected as total length. We therefore converted these measurements to SVL prior to analysis using a regression model constructed using measured total length and SVL data from 420 salamander larvae collected across 8 waterbodies on Pelee Island between June 3, 2021 – August 3, 2021 (see Supplementary Material).

At each pond, we measured several abiotic and biotic parameters, as described in Ward and Hossie (2020). These variables included water temperature (°C), fish presence (not detected, detected), max depth (cm), pH, conductivity (uS/cm), total dissolved solids (ppm), salinity (ppt), dissolved oxygen (mg/L), canopy cover around the pond margin (%), amount of leaf litter in the substrate (0 = none, 1 = thin layer, 2 = deep layer), and presence of submergent vegetation. Water parameters were measured within 1 m from the pond edge with both a Hoskin Scientific pH Meter and an ExStik II dissolved oxygen metre (ExTech Instruments ®) and used the average of the readings. As pond temperature differs throughout the day, and ponds were measured at different times, water temperature data was standardized to noon-corrected values prior to analysis, following methods outlined in Ward and Hossie (2020) (see Supplementary Material).

Crayfish and fish detected during our dipnet sweeps for larvae were identified to species. The fish we detected included fathead minnow (Pimephales promelas) and green sunfish (Lepomis cyanellus), both of which are known predators of amphibian eggs and larvae (e.g., Petranka, 1983; Drake et al., 2014). In addition, we employed a series of methods to investigate the presence of P. acutus. First, we set Promar collapsible minnow traps (TR-501) baited with canned cat food (Purina Fancy Feast™) at 11 ponds in 2023. Depending on the size of the waterbody, 2–5 traps were set per waterbody for a minimum of 2 consecutive days and were checked at least once every 24 hours. The combined effort of all the traps was 58 trap days. Two large ponds not surveyed for salamander larvae were actively searched at night in 2023 with dip nets for a combined total of 9 person-hours. Additionally, we recorded any crayfish caught when conducting dipnet surveys for salamander larvae. We also scanned pond margins for evidence of crayfish burrows. P. acutus construct burrows that are distinct from those made by the native species (Schryer and Hamr, 2024), and this was helpful for directing additional search efforts.

Animals were handled in accordance with Animal Care protocols approved by Trent University and Ontario Ministry of Natural Resources and Forestry (Protocols: 23906, 25301, 25344) and released back to their point of capture. Land access permits, Wildlife Scientific Collector’s Authorizations (1079527, 1082275, 1085623, 1088782), and a Notice of Activity under the Ontario Endangered Species Act (Confirmation ID: M-102-3802796883, M-102-9225853169, M-102-1254262277) were secured to conduct this work. A license to collect crayfish and authorization under the Invasive Species Act were also secured (License #: 1101982, Authorization Number: 2024-02).

Statistical methods

Using our empirical data collected from 2022-2025, we quantified temporal patterns in the proportion of surveyed ponds on Pelee Island where Ambystoma salamanders and Procambarus acutus crayfish were detected. Given our data set, we were unable to formally account for imperfect detection, however we estimated uncertainty in these binomial proportion point estimates using Wilson score confidence intervals. To assess whether the environmental conditions differed significantly among ponds without crayfish, with native crayfish only, or with P. acutus, we conducted a PERMANOVA using our pond environmental data (i.e., water temperature, fish presence, max depth, pH, conductivity, total dissolve solids, salinity, dissolved oxygen (% saturation), canopy cover around the pond margin (%), amount of leaf litter in the substrate (0 = none, 1 = thin layer, 2 = deep layer), and presence of submergent vegetation), and Crayfish Detected as a categorical response variable (i.e., None, Native Only, Invasive Present). Using the same environmental data (but excluding the fish presence variable), a second PERMANOVA model was run to evaluate whether ponds with versus without fish varied in their environmental conditions.

To test the hypothesis that invasive crayfish cause a reduction in salamander larvae abundance (i.e., have a consumptive effect), we examined CPUE data from 105 pond surveys conducted between 2022–2025 in 36 different ponds across Pelee Island, Ontario. Most of these ponds were surveyed in multiple years (n = 31). We used a linear mixed model, with CPUE as our response variable, and Crayfish Detected (i.e., None, Native Only, Invasive Present), Fish Detected (Not detected, Present), and Year as fixed effect predictor variables. Site identity was included as a random effect to account for possible lack of independence arising from sampling the same ponds in multiple years. Year was treated as a categorical fixed effect to test for annual variation in CPUE, which might reflect a decline in CPUE as P. acutus has spread and established. Year could not be included as a random effect in this model because there was insufficient among-year variation to reliably estimate a separate variance component for it, causing the model’s random effects structure to be singular. Including the Fish Detected variable as a predictor allowed us to evaluate the effect of fish on larval abundance, and control for any fish-effects on CPUE when testing for an effect of native or non-native crayfish. CPUE was transformed using a cube-root to meet model assumptions.

To test the hypothesis that the non-consumptive effects of invasive crayfish cause salamander larvae to grow more slowly and therefore achieve a smaller body size, we examined SVL data collected from n = 1447 larvae collected from 22 distinct ponds on Pelee Island between 2022-2025. Most of these ponds (n = 17) were surveyed in multiple years. We restricted this analysis to include data from surveys where we collected measurements from at least 29 individuals from a given pond, in a given year, to ensure that our size data reflected a representative sample. To analyse this data, we used a linear mixed model, with SVL as the response variable, and CPUE (larvae/sweep), Crayfish Detected (i.e., None, Native Only, Invasive Present), Fish Detected (Not detected, Present), Water Temperature, and Month (May, June, July) as predictor variables. We included Pond ID and Year as random effects to account for possible lack of independence arising from sampling individuals from the same pond or during the same year. Including CPUE allowed us to account for size variation arising as a result of density-dependent effects of conspecifics. We included water temperature because the growth rate of ectotherms is likely to be impacted by temperature (Finkler, 2006). Including Fish as a predictor in our model allowed us to avoid potentially confounding effects of fish on larval SVL. Month was included as categorical fixed effect to account for variation in size arising from sampling at different times of the year.

Statistical analyses were conducted in R, version 4.2.2 (R Core Team, 2022) and are shown in the Supplementary Material. Significance of predictors was evaluated using test statistics from a Type III sums of square Analysis of Variance table, with p-values computed using approximate denominator degrees of freedom estimated by Satterthwaite’s method. We used the ranova function from the lmerTest package (Kuznetsova et al., 2017) to conduct Likelihood Ratio Tests to confirm that the random effects in our model significantly improved model fit. The marginal and conditional R² values for our mixed models were calculated using the r.squaredGLMM function from the MuMIn package (Barton, 2023). Where appropriate, we computed the model-adjusted means using the emmeans function from the emmeans package (Length, 2023) to quantify the magnitude of the differences in our response variables between treatment levels for categorical predictors.

MaxEnt modelling

We used MaxEnt (version 3.4.4) to predict where P. acutus are most likely to occur on Pelee Island based on known occurrences and environmental variables (Phillips et al., 2024). We chose MaxEnt because it is suitable for presence-only data, can use both continuous and categorical predictor variables, and has demonstrated good predictive power (Biodiversity and Climate Change Virtual Library, 2021). We used all P. acutus records on Pelee Island from 2015–2025 as the species occurrences, derived from our field work as well as research-grade sightings on iNaturalist. Land cover, soil type, distance to the drainage network, and distance to the nearest wetland were used as the environmental variables (see Supplementary Material), since previous work has found them to be potentially important predictors of P. acutus occurrence (Hossie and Hamr, 2022). Land cover and soil type were treated as categorical variables. Land cover was based on a collapsed version of the Southern Ontario Land Resource Information System (SOLRIS) (SOLRIS Technical Team, 2015). The categories were wetland, forest, tallgrass, agricultural, transportation, and human disturbance. In the original SOLRIS classification, water is listed separately from wetlands; however, since we suspect crayfish would use these habitats similarly, we collapsed them into a single category (but excluding Lake Erie). Soil type was identified based on the Ministry of Natural Resources Land Information Ontario Data Description Soil Survey Complex (OMAFRA, 2019). Cells of all layers were approximately 15 m 2. We evaluated MaxEnt models based on the Area Under the Curve (AUC) of the Receiver Operator Characteristic, a comparison of the false-positive and false-negative rates (Hirzel et al., 2006). The proportion of false positives is plotted on the x-axis and the proportion of true positives (i.e., sensitivity) is plotted on the y-axis, providing a measure of how accurately a model can predict occurrences. In situations where false negatives are more harmful than false positives, models are recommended to prioritize sensitivity (Zweig and Campbell, 1993). We therefore chose to maximize AUC in our model selection, since this study aims to identify all potential P. acutus locations on Pelee Island. While AUC has been criticized as inappropriate for measuring absolute performance, it is acceptable to use AUC for comparing models predicting the same data, which is how we employ it here (Lobo et al., 2008). To evaluate the robustness of our AUC measure, we used k -fold cross-validation, which divides the data points into k equally-sized groups, training the model with k – 1 groups, then calculating the AUC with the k th group (Hirzel et al., 2006). This process is repeated k times so all groups are used as the left-out group, thus providing the mean and standard deviation of the AUC across runs. We used 7 groups of 118 points each, with the exception of 2 groups of 119 in each run, as our sample size of 138 is unable to be evenly divided, and 100,000 pseudo-absence training points. We examined variable contribution to each model through jackknife analysis of regularized training gain, test gain, and AUC, which tests the fit of a model with all variables but the one in question, then the fit of a model only including that model. We also examined the effects of each variable using response curves, where the y-axis displays the probability of the species in question being present at a given value of the variable (Phillips et al., 2017).

P. acutus was detected in 25 of the 36 distinct waterbodies that we surveyed for salamander larvae. Between 2022–2025 we saw an increase in the proportion of surveyed ponds where P. acutus was detected (2022: 48%, 95% CI = 32-65%, 15/31 ponds; 2023: 42%, 19-68%, 5/12; 2024: 69%, 51-82%, 22/32; 2025: 73%, 55-86%, 22/30; Figure 1). A two-proportion z-test indicates that the difference between 2022 and 2025 is significant (z = -1.99, P = 0.047). Over the same period, the proportion of surveyed ponds where salamander larvae were detected decreased (2022: 81%, 95% CI = 67-95%, 25/31 ponds; 2023: 75%, 47-91%, 9/12; 2024: 53%, 36-69%, 17/32; 2025: 57%, 39-73%, 17/30), with the difference between 2022 and 2025 also significant (z = 2.02, P = 0.043; Figure 1). Salamander CPUE ranged from 0-6.26 larvae/sweep, with a median of 0.09 larvae/sweep across all ponds, and 0.42 larvae/sweep across ponds occupied by salamanders.

We compiled confirmed detections P. acutus from 17 ponds and 14 iNaturalist observations in 2022, 7 ponds and 4 iNaturalist observations in 2023, 22 ponds and 24 iNaturalist observations in 2024, and 22 ponds and 23 iNaturalist observations in 2025. The rest of our point observations were solely from iNaturalist observations, with 1 in 2015, 2 in 2020, and 2 in 2021. Combined, this was a total of 138 presence points for use in the MaxEnt model (see Supplementary Material).

Environmental conditions

Our PERMANOVA test found that ponds where no crayfish were detected, only native crayfish were detected, and ponds where P. acutus were detected did not differ in their aquatic habitat conditions (F 2,102 = 0.80, P = 0.48). Ponds with fish, however, tended to have higher maximum depth, higher dissolved oxygen, higher pH, lower canopy cover, less leaf litter in the substrate, and warmer water compared to those without fish (F 1,103 = 3.31, P = 0.048).

Consumptive effects

We found no evidence of consumptive effects of native crayfish or P. acutus on salamander larvae (F 2,98.42 = 1.01, P = 0.37); however, CPUE was significantly reduced by the presence of fish (F 1,104.71 = 10.10, P = 0.002, Figure 2). We did not detect significant variation in CPUE among years (F 3,78.27 = 0.79, P = 0.50, Figure 2). Ponds where fish were detected showed an estimated 8.4-fold reduction in CPUE (Mean, 95% CI: No fish = 0.53, 0.40-0.67; Fish detected: 0.06, 0.00-0.38). The consumptive effects model had a marginal R² value of 0.13, and a conditional R² value of 0.41.

Non-consumptive effects

We did not detect a non-consumptive effect of native or non-native crayfish on larval salamander body size (F 2,1395.43 = 0.31, P = 0.74). Larval body size was, however, impacted by CPUE (F 1,1438.39 = 26.79, P < 0.001), fish presence (F 1,1438.24 = 4.72, P = 0.03), water temperature (F 1,1279.45 = 32.82, P < 0.001), and month of sampling (F 2, 2.13 = 26.85, P = 0.03). Specifically, body size declined with both higher CPUE and warmer water temperature, and body size increased from May-June (Figure 3). Interestingly, salamander larvae in ponds where fish were detected had an estimated 6.6% smaller body size (Mean, 95% CI: No fish = 24.2 mm, 21.8-26.7 mm; Fish detected: 22.6 mm, 19.7-25.4 mm). The non-consumptive effect model had a marginal R² value of 0.23, and a conditional R² value of 0.74. Data from two surveys resulted in very high CPUE estimates (i.e., 5.1 and 6.25 larvae/sweep). Running our analysis while excluding these two outliers did not qualitatively affect our conclusions (see Supplementary Material, Supplementary Figure 6).

MaxEnt

We ran models with all combinations of land cover, soil type, distance to the drainage network, and distance to the nearest wetland (see Supplementary Material). The full model had the highest sensitivity (mean training AUC 0.962, mean test AUC 0.9556). The most important variables in this model were the soil type (percent contribution = 46.7%, permutation importance = 10.8%) and the distance to wetlands (percent contribution = 39.1%, permutation importance = 80.3%), followed by distance to the drainage network (percent contribution = 8.1%, permutation importance = 7.8%) and landcover type (percent contribution = 6.2%, permutation importance = 1.1%). The most important soil type was Toledo clay, with a 68% probability of presence (Figure 4). The probability of presence is greater than 95% when the distance to wetlands is between 6–39 m, then declines to approximately 10% around 108 m, after which point it decreases gradually to nearly 0% (Figure 4). For the distance to the drainage network, probability of presence peaks at approximately 79 m, reaching 73% probability of presence (Figure 4). The probability of presence decreases linearly to approximately 63% at 175 m, then drops exponentially until reaching 0 at 922 m from the drainage network. The most important landcover type was tallgrass (85%), followed by agricultural (68%) and wetland (55%). Areas of high (≥ 0.80) P. acutus habitat suitability intersected with 64% (14/22) of 1 km × 1 km cells reported to contain A. texanum and/or their dependent unisexuals (Figure 5).

Using data from 105 dipnet surveys over four years spanning 36 distinct waterbodies across Pelee Island, Ontario, we tested for evidence of consumptive and non-consumptive effects from the recently introduced White River crayfish (P. acutus) on Ambystoma salamander larvae. We observed an increase in the proportion of waterbodies where P. acutus was detected between 2022–2025 alongside a concomitant reduction in the proportion of waterbodies where salamander larvae were detected. We did not, however, detect a long-term reduction in mean CPUE of salamander larvae over this period. In contrast to our predictions, we found no evidence indicating that the relative abundance of salamander larvae (CPUE, larvae/sweep) or salamander larva body size (SVL) were affected by the presence of P. acutus. Notably, we detected substantial consumptive and non-consumptive effects from native fish on salamander larvae, and found that salamander larvae were also smaller as the density of salamander larvae and water temperature increased. Based on our spatial modelling exercise, P. acutus presence on Pelee Island was associated with the Toledo Clay soil type, proximity to wetlands (<100 m), proximity to the island’s drainage network (< 500 m), and tallgrass, agricultural, or wetland land cover types. Areas of high predicted habitat suitability for P. acutus intersect with 64% of the 1 km² grid cells where federally endangered salamanders are known to occur. While we are relieved that P. acutus may not affect salamanders in the ways we investigated, we remain concerned about other routes through which these crayfish may impact amphibians and other native biodiversity in Canada.

The lack of a negative relationship between P. acutus presence and larval Ambystoma salamander density has several possible explanations. Pelee Island’s Ambystoma salamanders naturally coexist with the Digger crayfish (Creaserinus fodiens) and the Calico crayfish (Faxonius immunis) (Hossie and Hamr, 2022), and may therefore possess adaptions to crayfish that are general enough for them to tolerate P. acutus (Melotto et al., 2021; Ricciardi and Atkinson, 2004). Importantly, A. texanum co-exists with P. acutus elsewhere in its range (Loughman, 2007; Loughman and Simon, 2011), and the population on Pelee Island may still possess adaptations related to the detection and avoidance of P. acutus despite approximately 4000 years of separation (see also Blumstein, 2006). Cruz and Rebelo (2005) found that P. clarkii, a close relative of P. acutus, primarily impacted amphibians that are not usually exposed to crayfish and are thus less adapted to recognize them as threats. In fact, it is possible that P. acutus presence improves habitat quality for Ambystoma salamanders by constructing burrows that juvenile and adult salamanders use as terrestrial retreats, which offsets any consumptive effect. Indeed, Ambystoma salamanders are known to use crayfish burrows to maintain moisture and to overwinter, and P. acutus burrows may provide suitable habitat (Hossie and Hamr, 2022; Loughman, 2007; Owen and Juterbock, 2013; Trépanier and Dunham, 1999; Ward and Hossie, 2020). While it remains possible that P. acutus consume Ambystoma eggs and larvae, it appears that these effects are so far negligible when compared to the effects of naturally occurring carnivorous fish (i.e., Pimephales promelas, Lepomis cyanellus). Importantly, our study has limited capacity to evaluate whether P. acutus directly or indirectly impact salamander populations through impacts on later life stages, including late-stage larvae, metamorphs, and adults. Additional research exploring crayfish burrow use by Ambystoma salamanders and experiments that test for consumptive effects of native vs. invasive crayfish on amphibians would help us better understand the nature of these interactions and the associated risk to Canadian fauna.

Salamander larval body size was not significantly impacted by the presence of P. acutus, suggesting that this non-native crayfish did not induce a non-consumptive predator effect. One explanation is that the Ambystoma larvae studied here do not perceive the introduced P. acutus as a predation threat and therefore do not increase vigilance behaviour or reduce foraging in ways that impact their body size or growth rate (Polo-Cavia et al., 2010; Wirsing et al., 2020). Conversely, other studies have found that Ambystoma larvae increase refuge use with all potential threats equally (Holomuzki, 1989) or only with predators other than crayfish (Kats, 1988; Relyea and Werner, 1999). It remains possible that the current densities of P. acutus are too low to induce strong responses by Ambystoma larvae, or that non-consumptive effects are manifest in a way not examined here (e.g., though reductions in body condition and subsequent survival as metamorphs).

In contrast to the lack of detectable consumptive and non-consumptive effects of P. acutus, we found strong evidence that the presence of carnivorous fish dramatically reduced CPUE (8.4-fold reduction) and body size of salamander larvae (~6.6% decrease in SVL). We observed multiple cases of previously occupied ponds having no detectable larvae following colonization by fathead minnow (Pimephales promelas) or green sunfish (Lepomis cyanellus). Encouragingly, we also observed cases where ponds previously occupied by fish were recolonized by salamander larvae once fish were no longer detected (e.g., following pond drying), reinforcing the important role of ephemeral hydroperiod in salamander breeding habitat. Many fish readily eat Ambystoma salamander larvae, sometimes to the point of excluding the salamanders from habitats where fish are present (Petranka, 1983). Previous studies have demonstrated that fish affect Ambystoma survival (e.g., Semlitsch, 1987), though other studies found the effect of fish presence was exclusively on larval size (Figiel and Semlitsch, 1990). Effects on larval body size could also arise from size-selective predation, interspecific competition for shared food resources, or from intraspecific competition among salamander larvae competing for the narrower resource pool available within refugia from fish (Figiel and Semlitsch, 1990; Semlitsch, 1987). Importantly, these mechanisms are not mutually exclusive, and Ousterhout et al. (2015) found that while fish presence strongly predicted larval Ambystoma density, density-driven competition better predicted body size. We cannot fully exclude the possibility that fish-associated differences in salamander density and body size arise as a direct or indirect result of the significantly different physical and chemical properties of waterbodies containing fish. Indeed, Ward and Hossie (2020) found that pond conditions associated with larval occupancy and CPUE (i.e., cooler water temperatures, higher canopy cover and more leaf litter in the substrate) were opposite to what we found here to be associated with fish presence.

Larval body size decreased significantly with increases in the density of salamander larvae. This is consistent with previous work in pond-breeding Ambystoma, where density-dependent reductions in growth appear to be driven primarily through the mechanism of food limitation (Wilbur and Collins, 1973; Petranka and Sih, 1986; Scott, 1994; Wildy et al., 2001; Ousterhout et al., 2015). Reductions in larval body size can result in protracted time-to-metamorphosis (Scott, 1990; Wilbur, 1976) or metamorphosis at a smaller body size (Wilbur and Collins, 1973), conceivably leading to corresponding reductions in survival or future reproductive output (e.g., Altwegg and Reyer, 2003; Cabrera-Guzmán et al., 2013). Extended larval periods can also lead to higher mortality due to the combined risks of predation and pond drying (Petranka, 1984). Restoration efforts should aim to construct ponds large enough to minimize the effects of density-dependence on larval growth and support viable populations.

We acknowledge the difficulty in establishing true absence of P. acutus in the ponds where we did not detect them, and the known limitations of the dipnet and minnow trapping sampling methods we employed (e.g., Price and Welch, 2009). To overcome the potential weakness of any one methodology, we combined three distinct methods for detecting crayfish (dip-net surveys, baited minnow traps, visual scans of pond margins for burrows), and we contend that our search efforts were sufficient to reliably detect crayfish in our focal waterbodies when they were present. Herein, we did not attempt to quantify the abundance or density of P. acutus in each pond. It is possible that we failed to detect consumptive or non-consumptive predator effects of P. acutus due to low densities in the environment. We think this is unlikely based on our own anecdotal observations that some ponds with large numbers of P. acutus also supported high densities of salamander larvae.

Importantly, complementary experimental work would be helpful to more conclusively assess the impact of P. acutus on salamander body size and abundance. Future experimental work could also examine the risk-response curves of larval salamanders in a lab setting (e.g., see Petranka, 1989; Schoeppner and Relyea, 2008; Supekar and Gramapurohit, 2020), or examine whether crayfish density impacts their prey consumption rate to better resolve this possibility.

Our MaxEnt habitat suitability model revealed that P. acutus appear to use habitats similar to those used by the Ambystoma we studied, namely wetlands located in areas with clay soils. Indeed, a preference for clay soils has been previously reported for both A. texanum (Greenwald et al., 2016) and P. acutus (Albaugh, 1973; Bouchard and Robison, 1980; Schubert, 2022). Compared to sand or loam, clay is better at retaining the shape and moisture of burrows which is key to accessing ground water as surface water dries up and avoiding the freezing temperatures of winter (Taylor and Schuster, 2004). The importance of distance to wetlands could reflect the movement capabilities of P. acutus, with the high suitability of short distances, followed by the sharp decline in suitability until 100 m, indicating how they may travel short spans with the occasional long-distance dispersal (Hobbs, 1981; Taylor and Schuster, 2004). This may inform our understanding of future colonization patterns of P. acutus into Ambystoma habitats. Relatedly, we were surprised by the relative unimportance of the distance to the drainage network, as we suspected that crayfish may use these canals to disperse through the island (Burskey, 2007; Loughman and Simon, 2011; Taylor and Schuster, 2004). Given that drainage canals are designed to lower the water table, perhaps this also makes it more difficult for these crayfish burrows to reach water (Loughman, 2007, 2010). Critically, we did not survey within the drainage network due to landowner permission, and surveys directly within the drainage network are still needed to determine their usage by crayfish. More generally, P. acutus was only recently documented on the island, meaning that we still have few observations, despite our island-wide survey efforts. Doubtless, additional observations drawn from systematic surveys across the island will improve the accuracy of habitat models for this species.

We recognize the difficulty in drawing specific conclusions about the overlap in habitat suitability between P. acutus and the Ambystoma salamanders in our study given the difference in cell size (15 m 2 cells for P. acutus, 1 km 2 for Ambystoma salamanders). P. acutus clearly have overlapping habitat preferences with the salamanders in this system (i.e. areas with clay-based soils near wetlands; Greenwald et al., 2016; Smith, 2022; Hossie and Hamr, 2022). Areas within the 1 km 2 salamander squares identified as highly suitable P. acutus cells are often—though not always—suitable habitat for salamanders. Our pond-level sampling data also indicated that P. acutus co-occurred with Ambystoma larvae in 22 of the 36 (61%) distinct waterbodies that we surveyed at least once between 2022-2026, confirming substantial overlap in habitat use. More broadly, our findings demonstrate that many of Pelee Island’s remaining wetlands have been invaded by P. acutus, and further expansion is possible. Suitable habitat for Amybstoma salamanders on Pelee Island is limited, especially for A. texanum (Hossie, 2018; Smith, 2022), so the integrity of each remaining patch is especially important.

Our results should not be taken as an indication that these introduced crayfish are not a risk to Canadian ecosystems. Importantly, P. acutus may impact Ambystoma in ways not investigated here and their impacts may vary dramatically among species of amphibian. For example, while use of native crayfish burrows by Ambystoma salamanders as terrestrial retreats is well documented, it is not yet clear if the burrows constructed by P. acutus are equally suitable (Hossie and Hamr, 2022; Loughman, 2007; Norrocky, 1991; Owen and Juterbock, 2013; Trépanier and Dunham, 1999; Ward and Hossie, 2020). More generally, Procambarus crayfish may destabilize shorelines, which could lead to other unexpected impacts to ecosystems and infrastructure (Roessink et al., 2017; Sanders et al., 2021). Furthermore, the impacts of introduced species are often not immediately observable after initial introduction or even establishment, as it may take a new species several generations to grow large enough populations, create sufficient ecological disturbances, or expand their range to affect native species (Crooks, 2005). Regardless, objective evidence-based approaches remain critical for invasion science to effectively discern the true impacts of introduced species on amphibians and their habitats.