Developing predictive methods to forecast the impacts of existing and emerging invasive species is of critical importance to biodiversity conservation. However, invader impacts are context-dependent, making reliable and robust predictions challenging. In particular, it is unclear how temporal variabilities in relation to temperature regime shifts influence invader ecological impacts. In the present study, we quantify the functional responses of three coexisting freshwater fishes: the native freshwater River Goby Glossogobius callidus, and the non-native Mozambique Tilapia Oreochromis mossambicus and Western Mosquitofish Gambusia affinis, under two temperature treatments using chironomid larvae as prey. This was used along with fish abundance data to determine temporal differences in ecological impacts of each fish species between seasons (i.e. at two corresponding temperatures). All three fish species exhibited potentially population-destabilizing Type II functional responses. Their maximum feeding rates were consistently higher in the warm temperature treatment, whereas attack rates tended to be reduced. Non-native Mozambique Tilapia had the highest maximum feeding rate under both temperature treatments (18 °C and 25 °C), followed by the non-native Western Mosquitofish and lastly the native River Goby, suggesting greater per capita impacts on native prey by non-native fishes. The predatory fish abundances differed significantly according to season, with native River Goby and non-native Mozambique Tilapia generally more abundant than non-native Western Mosquitofish. By multiplying functional response maximum feeding rates with abundances of each fish species across the seasonal gradient, the relative impact potential of non-native Mozambique Tilapia was consistently higher compared to that of native gobies. Western Mosquitofish impacts were less apparent, owing to their low abundances. We demonstrate how seasonal temperature fluctuations affect the relative impact capacities of introduced species and the utility of consumer functional response and the relative impact potential metric in impact forecasting.

Context-dependence, impact assessment, introduced species, relative impact potential, seasonal abundance, thermal regime

Biological invasions are a central driver of global biodiversity loss (Sala et al. 2000; Turak et al. 2016; Bertelsmeier and Keller 2018; Shuai et al. 2018). This loss has not only socio-economic impacts, but also threatens ecosystem functions and services (Ricciardi et al. 2017). Biological invasions can occur through numerous pathways, such as human-mediated introduction, climate change and connectivity of systems, thus allowing extra-limital movement of species (Latombe et al. 2017). Upon arrival in a new environment, non-native species can cause ecological impact on native species assemblages through a range of biotic interactions (e.g. predation, competition and parasitism) (Vitousek et al. 1996; Thomsen et al. 2011; Havel et al. 2015; Seebens et al. 2018). Competition and predation play particularly important roles in the structuring of ecological communities (Paine 1980; Gurevitch et al. 1992). Although the impact of invaders through predation and competition is well documented, the context-dependency of these processes is often overlooked. In particular, direct biotic interactions (i.e. predation) can drive trophic cascades through alterations of prey abundance and native predator fitness (Gallardo et al. 2016; Penk et al. 2017).

Despite the considerable work conducted on invasive species, predicting ecological impacts of biological invasions has remained elusive (Simberloff et al. 2013; Dick et al. 2014). Ricciardi et al. (2013) highlighted context-dependency as the largest confound for impact predictions in invasion biology. Therefore, robust predictive methods that include environmental contexts as factors are needed in invasion studies to improve impact forecasting. In particular, temperature regime is a key abiotic context that is pervasive across all ecosystem types, and particularly in aquatic ecosystems (Lang et al. 2017). Specifically, temperature is a central determinant of the strength of predator-prey interactions and mediates food web stability (Rall et al. 2010, 2012; Englund et al. 2011). Fish have physiological mechanisms (i.e. metabolism and reproductive success) that are directly and/or indirectly dependent on temperature (Roessig et al. 2004). These mechanisms may differ between native and non-native species given differences in geographical origins and their physiological tolerances (Sorte et al. 2013). If high temperatures are more physiologically optimal for invaders, ecological impacts may be intensified (Iacarella et al. 2015), and seasonal changes, coupled with ongoing climatic warming, are key drivers of such temperature change in aquatic ecosystems. Indeed, interaction strengths are known to vary even with slight changes to seasonal temperatures (Sanford 1999, 2002). Therefore, failure to incorporate these factors in predictive approaches limits our ability to forecast invasive species impacts under changing environmental conditions across different spatiotemporal scales (Dick et al. 2013, 2014, 2017).

Methodological developments, which incorporate native/non-native species resource utilization across context-dependencies, have recently provided robust predictive tools for invasion science (Laverty et al. 2015; Dick et al. 2017; Dickey et al. 2018; Cuthbert et al. 2019). In particular, the functional response quantifies resource consumption as a function of resource density, and, in a predator-prey context, can quantify per capita ecological impacts of predators towards lower trophic groups (Holling 1959; Adams 1980; Dick et al. 2013, 2014, 2017; Alexander et al. 2014; Cuthbert et al. 2018). The types and magnitude of FRs quantify whether consumers will likely stabilize or destabilize resource populations (Murdoch and Oaten 1975; Rip and McCann 2011; Uszko et al. 2017). The relationships between consumer resource uptake and resource densities results in three broad functional response 'Types', and each Type has a different effect to resource population stability: an increasing linear relationship with no handling time constraint (Type I, mechanistically exclusive to filter feeders; Jeschke et al. 2004); an inversely density-dependent response characterized by high resource consumption at low resource density (Type II, resulting in rapid resource depletion at low densities); and a sigmoidal positively density dependent relationship (Type III, where resources have a low-density refuge) (Holling 1959). Despite the two functional response components, i.e. attack rate and handling time, being strongly associated with variations in temperature (Englund et al. 2011; Rall et al. 2012; Sentis et al. 2012; South et al. 2017; Cuthbert et al. 2018), there is, however, very limited information available on how temperature mediates species interactions at the population-level (Viherluoto and Viitasalo 2001; Fussmann et al. 2014; O’Gorman et al. 2017). Temperature and/or season effects may differ depending on how species functional traits directly influence responses, and these traits may too change along environmental gradients (Chapin et al. 2000).

Classically, the functional response has been combined with the ‘numerical response’ to determine the ‘total response’ of consumers (Solomon 1949; Holling 1959). The numerical response describes the consumer population-level response to changes in resource densities, while ‘total response’ can be defined as the multiplication of species’ numerical response with functional response (Solomon 1949; Holling 1959). Given that the numerical response, in comparison to the functional response, is difficult to ascertain, consumer abundance has recently been proposed as a proxy for numerical response in the development of the ‘impact potential’ and ‘relative impact potential’ metrics (Dick et al. 2017; Dickey et al. 2018). The ‘impact potential’ is the product of functional responses and abundance of consumers, while the relative impact potential compares the impact of the invader to that of a native (Dick et al. 2017). The strength of the relative impact potential metric lies in its ability to incorporate both species abundance and functional response under different environmental conditions (e.g. temperature change) and thus predict the influence of context-dependencies on invader impact (Laverty et al. 2017). This metric provides a novel approach for assessing existing and potential ecologically damaging species through the use of actual field abundance data under different environmental conditions.

The current study focuses on one native and two non-native fish species that co-occur in irrigation ponds within the Sundays River Valley, Eastern Cape, South Africa. These are the native River Goby Glossogobius callidus (Smith, 1937), and two non-native species, Mozambique Tilapia Oreochromis mossambicus (Peters, 1852) and Western Mosquitofish Gambusia affinis (Baird and Girard, 1853). The native River Goby is naturally found in estuarine and freshwater habitats (Engelbrecht and Mulder 1999; James et al. 2007). The River Goby is an invertivorous species (Wasserman 2012; Mofu et al. 2019). The non-native Mozambique Tilapia is native to eastward flowing rivers of central and southern Africa but its natural distribution does not extend to the Sundays River (Skelton 2001). The non-native Mozambique Tilapia is an omnivorous species, with clear ontogenetic shifts in diet, where juveniles feed predominantly on zooplankton and insects, while the diet of adults comprises of vegetative detritus (Zengeya et al. 2011). The non-native Western Mosquitofish is native to the lowland ponds, lakes and drainages of North America from Mexico to Alabama (Skelton 2001; Pyke 2008). It is an opportunistic omnivore feeding on algae, crustaceans, insects and amphibian larvae (Pyke 2008). Both non-native Mozambique Tilapia and Western Mosquitofish have been listed in the top 100 worst global invasive species database (IUCN 2006). Given that these three fish co-occur and are the most abundant within the Sundays River Valley irrigation ponds, this study sought to comparatively assess the potential relative ecological impacts of non-native Mozambique Tilapia and non-native Western Mosquitofish relative to the native River Goby towards native benthic prey across a seasonal temperature gradient.

Generalized linear models (GLMs) assuming a Poisson error distribution and log link were used to analyze overall prey consumption with respect to species, temperature and prey supply. Likewise, GLMs were used to compare fish abundances with respect to species, season and pond. Non-significant terms and interactions were removed stepwise to obtain minimum adequate models (Crawley 2007). Tukey’s comparisons were used to undertake post hoc tests of significant effects in each resulting model (Hothorn et al. 2008).

To distinguish between Type II and III functional responses, logistic regression of the proportion of prey consumed as a function of initial prey density was performed (but see also Rosenbaum and Rall 2018). Selection between Type II and Type III models was further confirmed via comparison of Akaike’s information criterion. A significantly negative first-order term indicates a Type II functional response, whereas a significantly positive first-order term followed by a significantly negative second-order term indicates a Type III response (Juliano 2001). Rogers’ random predator equation was used to model functional responses as prey were not replaced as they were consumed (Rogers 1972):

(1)

where N_e is the number of prey eaten, N₀ is the initial density of prey, a is the attack rate, h is the handling time and T is the experiment duration (fixed at 1). To enable model fitting, the Lambert W function was used (needed as N_e appears on both sides of the equation; (Bolker 2008)). Differences in attack and handling parameters were assessed pairwise between fishes at each temperature using indicator variables (Juliano 2001; Pritchard et al. 2017). Bonferroni corrections were used on raw p-values to account for multiple comparisons. Multiple estimates of the handling time parameter h were generated using non-parametric bootstrapping (n = 100), with maximum feeding rates then calculated via 1/h.

We calculated relative impact potential (RIP) of native (i.e. River Goby) and non-native (i.e. Mozambique Tilapia, Western Mosquitofish) species using the mean bootstrapped functional response maximum feeding rate (FR) and abundance (AB) for the three species at each season and pond (Dick et al. 2017):

(2)

when RIP < 1, the predicted impact of the non-native fish is predicted to be less than the native; when RIP = 1, there is no difference in impact between the fish species; whereas when RIP > 1, the non-native has a greater impact than the native. To integrate uncertainty into the RIP score, a probability density function (pdf) was applied using the standard deviation (SD) of the FR and AB estimates and this generated 80% confidence intervals (CIs) (see Dick et al. 2017). Biplots were then generated to illustrate the RIP for both for non-native Mozambique Tilapia and Western Mosquitofish relative to the native River Goby at each season between ponds (Laverty et al. 2017). All analyses were carried out in R v. 3.4.2 (R Development Core Team 2017).

Using the relative impact potential metric proposed by Dick et al. (2017), this study provides insights into how the ecological impacts of non-native species are mediated by temporal variabilities associated with seasons through the multiplying of functional responses and population abundances. Irrespective of seasonal variations, our results corroborate with studies that identified Mozambique Tilapia as a particularly impactful non-native species (Canonico et al. 2005; Maddern et al. 2007), whilst Western Mosquitofish impacts were less pronounced. We first show that all three fish species display a Type II functional response across the seasonal gradient, conducive to high resource utilisation at low densities. Whilst Type II functional responses are common in comparative laboratory-based studies (e.g. Dick et al. 2013), if included experimentally, additional context-dependencies such as habitat structure may have driven a significant impact on functional response form (Vucic-Pestic et al. 2010a; Vucic‐Pestic et al. 2010b; Kalinkat et al. 2013; Barrios-O’Neill et al. 2016). Moreover, greater incremental low-density prey resolution, different feeding durations and larger experimental aquaria volumes may further alter functional response forms (e.g. to Type III) (Sarnelle and Wilson 2008; Uiterwaal and DeLong 2018). Nevertheless, in the present comparative study, interspecific variation in functional responses between the species showed that both the non-native species exert higher per capita impacts than the native species on native prey and that predatory impacts are more profound during the summer season. These findings concur with a considerable number of studies comparing impact between invasive and native species (Alexander et al. 2014; Dick et al. 2014; Cuthbert et al. 2019).

Temperature differences had a significant effect on the functional response parameters, wherein attack rates were high in spring (i.e. 18 °C) and were reduced in summer (i.e. 25 °C). This result concurs with Grigaltchik et al. (2012), where an increase in temperature resulted in reduced attack rates, but contrasts with other studies (e.g. Wasserman et al. 2016) wherein attack rates exhibit a non-monotonic temperature response. Furthermore, we showed that during the summer season, handling times were reduced and hence these species exhibited higher maximum feeding rates. The findings from Englund et al. (2011) corroborate with ours, and this effect is mostly related to predators’ metabolic rate changes. For instance, for a predator’s metabolic activity to reach its maximum efficiency (i.e. high per capita effects), temperatures need to be optimal; yet if temperatures are too high this will result in reduced metabolic rates through catabolism (Clarke and Johnson 1999).

Secondly, we show that there was significant variation in fish abundances among species according to season, and also between ponds. Such variation in fish abundances seems to be a common theme, especially in fish communities that co-occur in environments and this is driven by spatial and temporal variation in life-history traits (Amezcua and Amezcua-Linares 2014). All three fish species were generally less abundant in spring and more abundant in summer. Mozambique Tilapia were the most abundant species overall, followed by River Goby and, lastly, Western Mosquitofish. By combining the fish maximum feeding rates and abundances (as per Dick et al. 2017) to give the relative impact potential score, we showed that the non-native Mozambique Tilapia consistently had the highest impact across seasons whereas, in the majority of cases, impacts of non-native Western Mosquitofish were less apparent relative to the native River Goby given currently low abundances.

Changes in relative impact potential scores with seasonal temperature fluctuations and fish abundances from different localities demonstrate how such context-dependencies can have a critical effect on the relative field impact capacities of introduced species through time (Dick et al. 2017). The effects of temperature regime shifts on interaction strengths are profound across habitat types and trophic groups (Englund et al. 2011; Rall et al. 2012), and increasing temperatures may exacerbate invader ecological impacts as species approach thermal optima (Iacarella et al. 2015). This is supported by the heightened functional responses observed for the non-native Mozambique Tilapia and non-native Western Mosquitofish as experimental temperature was increased to near their thermal preferendum (̴ 28°C; Jobling 1981). Therefore, the explicit inclusion of temperature change will be critical in future studies which seek to predict invader impacts across regime shifts associated with climatic warming and seasonal variability. Since the relative impact potential metric was 100% predictive of ecological impact across taxonomic and trophic groups (Dick et al. 2017), the current results, whereby relative impact potential is high for the non-native Mozambique Tilapia, gives confidence that this species can be forecast to cause major ecological impacts.

The present study further demonstrates the usefulness of numerical response proxies such as abundances in rapid assessments of potential impacts of introduced species. Indeed, in many cases, impact predictions are inherently limited if based on per capita impacts alone, given the importance of abundances in discernments of overall offtake rates by consumer populations (Dick et al. 2017). Importantly, our results suggest that ecological impacts of non-native species are likely to change across seasonal gradients associated with both changing functional responses and abundances, with summer impacts generally more profound than those in spring. We thus propose that further studies should incorporate such seasonal variability. Our study demonstrates that species-specific shifts in abundances may alter interaction strengths within ecosystems towards native populations. Therefore, quantitative assessments of species abundances can ultimately bridge the gap in decision-making and can be used to forecast future invader impacts under different climatic conditions when combined with per capita effects. Nevertheless, our study additionally demonstrates that individual systems (e.g. ponds) can differ substantially in predator community composition over time, and this system-specific population variability should be also considered in future studies.

Overall, this study provides further evidence of the strength of the relative impact potential metric in predicting ecological impacts of species and provides an extension to the framework by integrating an environmental gradient, which reflects seasonal temperature fluctuations. The identification of temporal shifts in impact across seasons and habitats in our study presents novel insights into invader impact. In many ecosystems, data on species abundances are still lacking, but since the relative impact potential metric enables impact predictions for species without invasion histories, we recommend more surveys to estimate abundance of potential invaders and/or for practitioners to incorporate other proxies (such as fecundity) into the metric (see Dickey et al. 2018). Crucially, the ability of both Mozambique Tilapia and Western Mosquitofish to thrive in novel habitats highlights their ecological plasticity, and with an increase in environmental temperatures, their impacts may be intensified through changes to functional responses and fish abundances. The relative impact potential metric thus allows for rapid assessment of current and future invasive species under shifting environmental contexts and can identify priority species for management.

This study forms part of a PhD research project supported by the Professional Development Programme Doctoral Scholarship with the South African Institute for Aquatic Biodiversity (NRF, Grant No. 101039), the CIB/DST Centre of Excellence for Invasion Biology (CIB) and the National Research Foundation – South African Research Chairs Initiative of the Department of Science and Technology (Inland Fisheries and Freshwater Ecology, Grant No. 110507) and NRF incentive funding (Grant Nos. 109015 to O.L.F.W., 103581 to D.J.W and 88746 to R.J.W). We acknowledge use of infrastructure and equipment provided by the NRF-SAIAB Research Platform and the funding channelled through the NRF-SAIAB Institutional Support system. Eastern Cape Department of Economic Development and Environmental Affairs (DEDEA) is thanked for issuing research permits. Any opinion, finding and conclusion or recommendation expressed in this material is that of the authors. Consequently, the National Research Foundation of South Africa and CapeNature do not accept any liability in this regard.