Global warming confers performance advantages to a non-native predator

Anthropogenic impacts manifest in rising temperatures worldwide, with drastic consequences for communities and ecosystems. Ectotherms tend to reach a smaller size at maturity in warmer environments; however, whether these responses elicit performance advantages between native and invasive species remains poorly understood. Here, we compared foraging efficiency and growth of aquatic predators (invasive racer goby, Babka gymnotrachelus; native European bullhead, Cottus gobio) in current and predicted future water temperatures (15^ °C; 23 °^C) facing two types of prey: live and frozen gammarids (Gammarus jadzewskii). We show that the growth rate of both predators was negatively affected by a temperature increase, but was consistently higher for the invasive species compared to the native. This was driven by changes in predators’ foraging efficiency, increased in the invasive, and decreased in the native species. Fish foraging was also shaped by prey motility, with active prey consumed more often than inactive ones. The results indicate that a temperature increase may shift the competitive balance between invasive and native species, with the former being more robust to temperature increase under a limited food supply than native comparators. This could further modify the impact of invaders on the local communities, with implications for ecosystem stability.

Methods
Methods

Animal collecting and housing

We used the invasive racer goby and the native European bullhead as model organisms. Both species are shelter-associated bottom dwellers, with burst-and-hold swimming mode, feeding on a variety of macroinvertebrates (mostly insects, crustaceans) and co-existing in habitats with rocky and gravelly substrates (Kottelat and Freyhof 2007, Kakareko et al. 2016, Janáč et al. 2018, Teletchea and Beisel 2018, Płąchocki et al. 2020). Fish were collected in the summer (June) of 2024, from the Brda River (central Poland, 53° 08' 52.5" N 17° 58' 10.5" E) by diving, using an aquarium net. The fish were kept in the laboratory in single-species 120-L stock tanks, with a maximum density of 20 individuals per tank, and divided according to the experimental temperatures (i.e., two groups per species), ensuring a similar size range of individuals in both groups. Tanks were filled with conditioned tap water (mean ± SD: temperature 19.1 ± 0.4, pH 8.1 ± 0.2, electrical conductivity 621 ± 23 µS/cm, oxygen level 8.2 ± 0.5 mg/l; measured with a Multi 340i Meter, WTW, Weilheim, Germany). The fish were acclimated to temperature treatments by gradually increasing or decreasing water temperature by 1oC/day. All of the stock tanks were equipped with aquarium filters, aerators and ceramic and stony shelters, with no other bottom substrate. The photoperiod was set at a 12:12h light:dark cycle, with lights turned on at 07:00 and off at 19:00.

We chose Gammarus jazdzewskii (Rudolph, Coleman, Mamos & Grabowski, 2018) as a representative prey species, being commonly predated by the tested fish species in the wild (Mills and Mann 1983, Grabowska and Grabowski 2005). Gammarids were collected from the Zielona Struga, a forest stream in central Poland (52o99'72.1"N, 18o45'04.7"E), with a hand net, in June and July 2024. The prey individuals were kept in the laboratory in 60-l tanks equipped with aquarium filters, aerators, and woody debris from the collection site, and filled with conditioned tap water at 18.5 ± 0.5 oC. The gammarids were acclimated to the experimental temperatures in the same manner as the test fish.

All animals had 2 weeks to acclimate to experimental temperatures before use. The fish were fed daily ad libitum with frozen chironomids, and the gammarids were fed every two days with frozen chironomids and decomposing leaves. The water in the stock tanks was exchanged once a week (ca. 30% of water volume).

Experimental design

To investigate the effects of elevated temperature on fish foraging efficiency we selected two water temperatures: ambient (15oC) and elevated (23oC). The ambient temperature corresponds to the mean river water temperature between May and October in Polish rivers over the 1971-2015 period (Graf and Wrzesiński 2020). Despite the fact that we compare performance of a primarily cold-water tolerant native species with a higher temperature tolerant invader, both species coexist in the sampled system, experiencing the same temperature regime at the collection site for 10-20 years (Kostrzewa and Grabowski 2003, Radtke et al. 2015). Further, the temperature of 23oC, according to the IPCC report (2023) and our model for the fish collection site, is predicted to be the common mean summer water temperature in European freshwaters by the end of 2100. We collected the fish in the lower section of a tributary of the Vistula River. In the Vistula, in summer (July) the water temperature reaches a maximum of 23.6°C already (Marszelewski and Pius 2014). Moreover, it is likely that a similar maximum occurs in the shallow water within tributary sampled system. Thus, while acknowledging the large temperature increment in our study, comparing both species’ performance at the higher selected temperature will provide valuable information on how the species could respond to ongoing global warming and microhabitat contexts.

We used live (active) and frozen (inactive) prey individuals to assess fish foraging efficiency concerning the prey's ability to actively escape. To produce frozen prey, live gammarids were collected from holding tanks, allocated numbers of prey individuals were placed into 50 ml containers assigned to particular experimental tanks, and frozen at -80oC.

We used opaque experimental tanks (51 × 38 × 30 cm, length × width × height) filled with 40 l of conditioned (constantly aerated for 48 h) tap water as experimental units. The bottom of each tank was covered with a 1-cm deep layer of sand, and constant aeration and water filtering were provided. Two weeks before the start of the experiment, each fish individual was weighed (to the nearest 0.1 g using Radwag WLC 2.X2 scale, N = 56), assigned among the temperature treatments, and placed individually in the experimental tanks. Each fish individual was assigned to a particular experimental tank and stayed inside throughout the experiment. The ambient temperature in the experimental tanks was maintained using air conditioning within the laboratory. To maintain the elevated water temperature, we placed half of the experimental tanks inside larger 60-l tanks (58 × 39 × 35 cm, length × width × height) filled with tap water and containing an aquarium heater with a thermostat (i.e. the heater was placed outside the experimental tank).

Fish were acclimated to live gammarid prey in their respective arenas by feeding them ad libitum for two weeks. Next, they were starved for 48h before the first experimental trial. On the test day, live gammarids were collected from the holding tanks, and allocated numbers of them were placed into 50 ml containers assigned to particular experimental tanks scheduled to receive active prey on that day. At the same time, inactive gammarids were defrosted to be used in the experimental tanks scheduled to receive inactive prey on that day. Then, the aeration in the experimental tanks was turned off and filters were pulled out. Each trial started between 09:00 and 10:00. The gammarids were flushed into experimental tanks by spreading them evenly over the water column. The fish had 24h (12:12h light:dark cycle, with light turned off at 19:00 and on at 7:00) to forage, after which the fish were removed and gammarids remaining in each experimental tank were counted. After that, the fish were starved for 24h before the next trial. Note that each fish was exposed to all prey densities by the end of the experiment, and thus had a similar overall feeding regime. Each fish was thus tested 12 times, with six densities (5, 10, 15, 20, 25 and 40 individuals) of both types of prey (active/inactive) provided in a random order. All the experimental tanks (13-16 replicates per species per temperature) were set up at the same time. Tanks assigned to different fish species and water temperature treatments were randomly distributed in the laboratory. The water in the experimental tanks was exchanged once a week (ca. 30% of water volume).

Statistical analysis

Overall body sizes before the experiment between the racer goby and the European bullhead were compared with Wilcoxon rank sum exact tests. Feeding rates (expressed as a proportion of prey killed over the initial prey density per replicate) were analysed using generalised linear mixed models (GLMM). These models followed a binomial error family, and considered proportions of killed prey as a function of species, temperature, prey type, and fish initial body mass, alongside all of their possible interactions. Initial prey density was included as a further explanatory variable. Individual fish identities were included as a random effect, because individuals were reused across prey density and prey type treatments during the experiment. This approach allowed us to capture variations in body mass regarding raw feeding rates of fish, in interaction alongside species, temperature, and prey type. Residuals were checked for overdispersion through diagnostics with simulations (Hartig 2022). Post-hoc tests were computed using estimated marginal means for significant effects (Lenth 2023).

Functional responses were characterised using logistic regression of the proportion of prey eaten as a function of prey density for each fish species, temperature, and prey type separately. We applied the Type II Rogers’ random predator equation (Rogers 1972) to account for prey depletion during the experiments:

(1)

where Ne is the number of prey eaten, N0 the initial prey density, a the predator attack rate (classically interpreted as the search efficiency), h the predator handling time (defined as the time spent pursuing, subduing, and consuming each prey item plus the time spent preparing to search for the next prey item), and T the duration of the experiment. Functional responses were compared using non-parametric bootstrapped 95% confidence intervals, with 2000 iterations per functional response curve (Pritchard et al. 2017).

Fish growth rate was analysed using generalised linear models (GLM) with a Gaussian error family. We log-transformed final and initial fish body mass to linearize the relationship between them. and modelled final body mass as a function of species, temperature, and initial fish body mass with all of their possible interactions. The total consumption (expressed as the total number of prey individuals eaten in all experimental trials) was included as a covariate.

The Wilcoxon rank sum exact test was performed using the base “stats” package (R Core Team 2024). Both GLMM and GLM were fitted with “glmmTMB” R package (Brooks et al. 2017). Residuals were checked for overdispersion through diagnostics with residual simulations using the “DHARMa” package (Hartig 2022). Slopes and means estimates for significant interaction terms were compared using “emmeans” package (Lenth 2023). Functional response models were fitted with the “frair” R package (Pritchard et al. 2017). All statistical analyses were performed using R 4.4.0 (R Core Team 2024).

Reference list:

Brooks, M. E., Kristensen, K., van Benthem, K. J., Magnusson, A., Berg, C. W., Nielsen, A., Skaug, H. J., Mächler, M., & Bolker, B. M. (2017). glmmTMB Balances Speed and Flexibility Among Packages for Zero-inflated Generalized Linear Mixed Modeling. The R Journal, 9(2). https://doi.org/https://doi.org/10.3929/ethz-b-000240890

Grabowska, J., & Grabowski, M. (2005). Diel-feeding activity in early summer of racer goby Neogobius gymnotrachelus (Gobiidae): A new invader in the Baltic basin. Journal of Applied Ichthyology, 21(4), 282–286. https://doi.org/10.1111/j.1439-0426.2005.00676.x

Graf, R., & Wrzesiński, D. (2020). Detecting patterns of changes in riverwater temperature in Poland. Water, 12(5), 1327. https://doi.org/10.3390/W12051327

Hartig F. (2022). DHARMa: Residual Diagnostics for Hierarchical (Multi-Level / Mixed) Regression Models. R package version 0.4.6, https://CRAN.R-project.org/package=DHARMa.

Janáč, M., Roche, K., Šlapanský, L., Polačik, M., & Jurajda, P. (2018). Long-term monitoring of native bullhead and invasive gobiids in the Danubian rip-rap zone. Hydrobiologia, 807, 263–275. https://doi.org/10.1007/s10750-017-3398-6

Kakareko, T., Kobak, J., Poznańska, M., Jermacz, Ł., & Copp, G. H. (2016). Underwater evaluation of habitat partitioning in a European river between a non-native invader, the racer goby and a threatened native fish, the European bullhead. Ecology of Freshwater Fish, 25(1), 60–71. https://doi.org/10.1111/eff.12191

Kostrzewa, J., & Grabowski, M. (2003). Opportunistic feeding strategy as a factor promoting the expansion of racer goby (Neogobius gymnotrachelus Kessler, 1857) in the Vistula basin. Lauterbornia, 48, 91–100. https://www.researchgate.net/publication/259000131

Kottelat, M., & Freyhof, J. (2007). Handbook of European freshwater fishes. Cornol, Switzerland: Publications Kottelat.

Lenth R (2023). emmeans: Estimated Marginal Means, aka Least-Squares Means. R package version 1.8.7, https://CRAN.R-project.org/package=emmeans.

Marszelewski, W., & Pius, B. (2014). The impact of changes in air temperature upon the thermal regime of the river Vistula in Toruń (Poland) over the years 1961-2012. In P. Gâştescu, W. Marszelewski, & P. Bretcan (Eds.), Water resources and wetlands (pp. 231–236). http://www.limnology.ro/water2014/proceedings/30_Marszelewski.pdf

Mills, C. A., & Mann, R. H. K. (1983). The bullhead cottus gobio, a versatile and successful fish. In Fifty-first annual report for the year ended 31st March 1983 (pp. 76–88). Freshwater Biological Association. https://agris.fao.org/agris-search/search.do?recordID=AV2012055893

Płąchocki, D., Kobak, J., Poznańska-Kakareko, M., & Kakareko, T. (2020). Environmental factors associated with the occurrence of the Ponto–Caspian gobies in a lowland river belonging to the central European invasion corridor. River Research and Applications, 36(1), 25–35. https://doi.org/10.1002/rra.3543

Pritchard, D. W., Paterson, R. A., Bovy, H. C., & Barrios-O’Neill, D. (2017). frair: an R package for fitting and comparing consumer functional responses. Methods in Ecology and Evolution, 8(11), 1528–1534. https://doi.org/10.1111/2041-210X.12784

R Core Team (2024). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.

Radtke, G., Bernaś, R., Dębowski, P., Morzuch, J., & Skóra, M. (2015). Ichtiofauna systemu rzeki Brdy. Rocz.Nauk.PZW, 25, 31–47. https://doi.org/10.12823/sapaa.0860-648X.15003

Rogers, D. (1972). Random search and insect population models. The Journal of Animal Ecology, 41(2), 369. https://doi.org/10.2307/3474

Rudolph, K., Coleman, C. O., Mamos, T., & Grabowski, M. (2018). Description and post-glacial demography of Gammarus jazdzewskii sp. nov. (Crustacea: Amphipoda) from Central Europe. Systematics and Biodiversity, 16(6), 587–603. https://doi.org/10.1080/14772000.2018.1470118