Research Article |
Corresponding author: Antoni Vivó-Pons ( avipo@aqua.dtu.dk ) Academic editor: Belinda Gallardo
© 2023 Antoni Vivó-Pons, Isa Wallin-Kihlberg, Jens Olsson, Peter Ljungberg, Jane Behrens, Martin Lindegren.
This is an open access article distributed under the terms of the Creative Commons Attribution License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Citation:
Vivó-Pons A, Wallin-Kihlberg I, Olsson J, Ljungberg P, Behrens J, Lindegren M (2023) The devil is in the details: exploring how functionally distinct round goby is among native fish in the Baltic Sea. NeoBiota 89: 161-186. https://doi.org/10.3897/neobiota.89.110203
|
Understanding the characteristics and conditions that make non-indigenous species (NIS) successful at establishing in recipient communities is a key in determining their potential impacts on native species, as well as to improve management actions such as prevention of future invasions. The round goby (Neogobius melanostomus) is one of the most widespread non-indigenous fish species in the Northern Hemisphere, including the coastal zones of the Baltic Sea. The impacts of round goby in the Baltic Sea are pronounced and multifaceted, yet our knowledge regarding the underlying assembly processes determining its establishment is limited. To overcome this knowledge gap, we applied a trait-based approach to assess the degree of niche overlap and functional (trait) similarity between round goby and native fish species in coastal areas from the Baltic Sea, based on the functional distinctiveness metric. Our results show that round goby is generally quite similar (or not dissimilar) to the native fish of the regional species pool, at least in terms of its overall trait composition. Conversely, round goby demonstrates pronounced differences compared to the native community in its display of parental care and territorial behaviour. Such differences in individual traits could play an important role in round goby’s invasion success in the Baltic Sea, including its interactions with native species (e.g. competition). Our results and their potential implications may be highly relevant for conservation and management if integrated within existing risk assessment tools for biological invasions in order to prioritise and enhance the effectiveness of preventative actions towards the expansion of round goby.
Baltic Sea, biological invasions, coastal fish, functional distinctiveness, NIS, round goby, species traits, trait-based approach
The introduction and spread of non-indigenous species (NIS) constitute a major threat to global biodiversity, ecosystems and their associated services (
Before having an impact on native communities, NIS need to be successfully established in the recipient area with self-sustaining populations (
The Baltic Sea is one of the largest brackish water bodies in the world, demonstrating a pronounced north-south salinity gradient from fully marine- to almost freshwater conditions in the northern parts (
The round goby (Neogobius melanostomus), originally from the Ponto-Caspian area, is one of the most widespread invasive fish species in the Northern Hemisphere (
Trait-based approaches provide a mechanistic way to address key aspects of biological invasions (
Monitoring data of coastal fish communities where round goby is present was obtained from the Swedish national and regional coastal fish monitoring programme as registered in the national coastal fish database - KUL (https://www.slu.se/kul). The data were extracted for 14 locations sampled between 2008–2021, covering from the south-western Baltic Sea (Stavstensudde) to the Bothnian Sea (Gävlebukten; Fig.
In order to represent the general ecology of the species, a total of 11 categorical traits were selected, with 37 different trait modalities: habitat switching, parental care, territorial behaviour, diet, temperature preference, development mode, pharyngeal bones, habitat, fin type, body type and length class (Table
List of included traits and modalities, the number and percentage of species displaying each modality and the explanation of each modality. Modalities in bold are displayed by round goby.
Traits | Nature of the trait | Categories (n = 37) | N species having that category | Frequency (% of species having that category) | Explanation |
---|---|---|---|---|---|
Diet | Multichoice nominal | Benthivorous | 18 | 46.22 | Feeding mainly on benthic invertebrates as adults |
Planktivorous | 7 | 16.59 | Feeding mainly on plankton as adults | ||
Generalist | 7 | 21.59 | Feeding on the other categories as well as on detritus, algae etc. as adults | ||
Piscivorous | 8 | 15.56 | Feeding mainly on fishes as adults | ||
Habitat | Categorical | Demersal | 16 | 59.26 | Living and feeding on or near the bottom as adults |
Benthopelagic | 7 | 25.93 | Living and feeding near the bottom as well as in mid-waters or near the surface as adults | ||
Pelagic | 4 | 14.81 | Living and feeding in the open water throughout ontogeny | ||
Fin type | Categorical | Emarginated | 5 | 18.52 | Caudal fin with a rather sharp and straight end with an indent in the middle |
Forked | 12 | 44.44 | Caudal fin with the indent deeper than in emarginated fins | ||
Absent | 1 | 3.70 | |||
Rounded | 8 | 29.63 | Caudal fin evenly rounded and convex | ||
Truncated | 1 | 3.70 | Caudal fin with a rather sharp edge that can be flat, square or straight | ||
Body type | Categorical | Deep | 6 | 22.22 | Body is compressed from the sides |
Elongated | 8 | 29.63 | Body is rather long and slender | ||
Flat | 2 | 7.41 | Body is flat (depressed) with eyes on the same side | ||
Normal | 11 | 40.74 | Body is proportional and neither compressed nor depressed | ||
Development mode | Categorical | Scattered | 4 | 14.81 | Eggs are scattered on the bottom |
Viviparous | 1 | 3.70 | Eggs receive nourishment from the female during development and hatch inside the body of the female | ||
Ovoviviparous | 1 | 3.70 | Eggs do not receive nourishment from the female during development | ||
Pelagic | 2 | 7.41 | Eggs float freely in the water column | ||
Adherent | 17 | 62.96 | Eggs adhere to a substrate in a layer | ||
Mass clump | 2 | 7.41 | Eggs adhere to each other, forming a clump | ||
Length class (maximum length according to FishBase) | Ordinal | 0–10 cm | 2 | 7.41 | |
10–20 cm | 11 | 40.74 | |||
21–30 cm | 7 | 25.93 | |||
31–40 cm | 2 | 7.41 | |||
41–50 cm | 5 | 18.52 | |||
Temperature preference | Categorical | Cold | 9 | 33.33 | |
Warm | 18 | 66.67 | |||
Territorial behaviour | Categorical | Yes | 8 | 29.63 | The species holds and defends a territory or has a very narrow home range, usually related to spawning, but not necessarily |
No | 19 | 70.37 | |||
Parental care | Categorical | Yes | 8 | 29.63 | The species exhibits some sort of parental care, for example, carries or guards the eggs/young |
No | 19 | 70.37 | |||
Habitat switching | Categorical | Yes | 21 | 77.78 | The species switches habitat due to spawning, feeding migration or winter migration |
No | 6 | 22.22 | |||
Pharyngeal bones | Categorical | Yes | 16 | 59.26 | The species has pharyngeal bones or branchial tooth plates |
No | 11 | 40.74 |
To assess the degree of (trait) niche overlap between round goby and native species, we used the functional distinctiveness index (D), weighted by species biomass. The functional distinctiveness index is defined as the mean functional distance of a single species to all other species present in a given community (
(Equation I)
where dij is the functional distance between species i and j, N accounts for the number of species in the community and Abj accounts for the relative importance (i.e. relative WPUE) of species j. A high D value indicates that a species is functionally distinct compared to the other species in the community (
We computed the functional distance between each pair of species (dij) given by Gower’s general coefficient of similarity (
In order to investigate if round goby was more or less distinct than the other species in the regional pool, we compared the value of functional distinctiveness of round goby relative to the values for all native species in the data set. Furthermore, to assess and compare the degree of niche overlap in trait space between round goby and the native species, we performed a Principal Coordinate Analysis (PCoA) on the overall pairwise dissimilarity matrix for the regional species pool (
To assess the effect and relative importance of each trait on functional distinctiveness, we calculated the difference between the distinctiveness values for each species based on all traits (Di, T) and the values when each individual trait was removed from the analysis (Di, T-t). We then divided the difference by regional distinctiveness, including all traits (Di, T) as follows:
(Equation II)
In order to reflect the key environmental conditions affecting the local distinctiveness of round goby at each sampling site, we compiled data of bottom salinity, temperature and depth, measured in situ as part of the fish monitoring programme. For some locations, bottom salinity and temperature data were incomplete, hence we complemented the monitoring programme data with data derived from the ice-ocean model NEMO-Nordic (based on NEMO-3.6, Nucleus for European Modelling of the Ocean; https://doi.org/10.48670/moi-00013) from Copernicus Marine Service (https://marine.copernicus.eu/). Before completing the available in-situ data with model-derived data, we compared values of available environmental variables derived from both sources. This sensitivity test showed a high correlation for both bottom temperature (r = 0.75) and salinity (r = 0.77) (Suppl. material
To determine how the local functional distinctiveness of round goby was affected by the selected environmental and biotic variables at each sampling event, we applied a multi-model approach using both Generalised Additive Mixed Models (GAMMs) and Random Forests (RFs). This allows for comparison of the derived response curves and variable importance to assess robustness and sensitivity of results to model choice (
Round goby Dl ,t = a + s (Bottom oxygenl,t) + s (Bottom salinityl,t) + s (Bottom temperaturel,t) + s (Depthl,t) + s (Chlorophylll,t) + s (Exposurel,t) + s (Richnessl,t) + s (Evennessl,t) + d (Location x Time step) + e (Gear) + ϵ
where the response variable D is the distinctiveness for the round goby at each sampling location l at a specific time t. The parameter a is the intercept, s is the thin plate smooth function for each of the covariates and ϵ the error term. To account for the potential effect of repeating measures within the same area, we also included a random effect d for each sampled location at a certain time (i.e. “Location × Time step” in the formula). The inclusion of this random effect in the model served to account for possible variation in distinctiveness between locations due to their different stages of invasion. Finally, e accounts for the random effects of the different gears used during the sampling. The degrees of freedom of the spline smoother function (s) were constrained to three knots (k = 3) to allow for non-linearities, but restricting its flexibility on the model fitting. Since D ranges between 0 and 1, the model was fitted with a beta-regression distribution (
Random forest (RF) is a machine-learning tool comprising ensembles of decision trees that rely on bootstrap aggregation (
To evaluate the predictive accuracy of the fitted models between methods, we also performed a formal cross-validation analysis by fitting the same model with a randomly sampled subset of the data (75% of the total observations) and predicting round goby distinctiveness with the remaining 25% of observations that were not used to fit the models. The cross-validation was repeated 100 times, selecting a new random subset of observations in each iteration for model training and testing. Subsequently, we assessed the range of uncertainty of the predictions (i.e. mean squared error) and the range of explained variance for both methods. All statistical analyses were conducted using the R software, version 4.1.0 (
Amongst all species, Eurasian perch (Perca fluviatilis) was the most abundant taxa in the monitoring data in terms of weight, representing 24% of the total biomass, followed by common roach (Rutilus rutilus; 14.8%) and Atlantic cod (Gadhus morhua; 8.3%), while round goby represented only 1.04% of the total biomass (Suppl. material
Position of round goby and its six most functionally similar species along the distribution of WPUE-weighted distinctiveness values from the regional species pool. The black vertical line indicates the median value of distinctiveness for the whole community. The highlighted species are ordered according to their values of distinctiveness. The numbers on top of them only indicate the corresponding names.
The first two axes of the PCoA of functional distances (trait space) explained 35.1% and 23.2% of the total variability between species, respectively. Round goby was located closer to the most functionally distinct species (defined by the 4th quartile) in the trait space (Fig.
Community trait space given by a PCOA of functional distances between all species (A). The green dot and drawing indicate the position of round goby. Red dots indicate the position of species classified as being the most distinct, while blue dots define species classified as most similar compared to the rest of the community. Names in bold indicate the position of the most functionally similar species to round goby B biplot of trait vectors and loadings showing which traits are influencing the position of each species in the PCOA C zoom of the central part of the biplot.
Amongst the set of traits considered, displaying parental care had the highest influence on round goby distinctiveness, with a relative increase of 7.33% in D values when including this trait. This positive effect of parental care was closely followed by having an elongated body type (7.26%), a rounded fin (6.61%) and territorial behaviour (6.40%; Fig.
Effect of traits on species functional distinctiveness, shown as the percentage change in overall distinctiveness if excluding each individual trait in the calculations. Results are shown when using either the whole species pool with the green dots representing the effect of each trait on round goby functional distinctiveness.
The influence of each trait on the whole fish community distinctiveness demonstrated that body shape and fin type had the highest median positive effect (5% and 4.73%), while diet showed the most negative effect (-6.86%; Fig.
Distinctiveness values for round goby at each location were highly variable, with the highest mean value found in Hanöbukten (0.64) and the lowest mean distinctiveness in Herrvik (0.44). In 7 out of 14 locations, the mean distinctiveness of round goby was higher than the overall value when compared to the regional fish community (Fig.
Distribution of round goby local distinctiveness at each sampling site (A) and over time (B).
From the selected set of potential abiotic and biotic drivers, species richness and evenness, depth, coastal exposure, bottom temperature and oxygen showed significant effects on round goby distinctiveness in the fitted GAMM (Table
Partial effect curves derived from the models fitted with both GAMM and Random Forest. Only the variables that had a significant effect in GAMMs are shown. The yellow line and ribbon represent the partial curve and the standard deviation derived from the GAMM. The blue and black lines represent the partial effect curve and the corresponding variability derived from the Random Forest.
Results of the GAMM models for round goby local functional distinctiveness. Edf refers to estimated degrees of freedom; significant effects are highlighted in black.
Variables | edf | Chi squared | p-value | R squared | Deviance explained | N | |
---|---|---|---|---|---|---|---|
Eveness | 1.803 | 9.768 | 0.015 | * | |||
Richness | 1.923 | 25.361 | < 0.001 | *** | |||
Depth | 1.001 | 7.621 | 0.006 | ** | |||
Bottom temperature | 1.906 | 18.323 | < 0.001 | *** | |||
Bottom salinity | 1.000 | 0.370 | 0.543 | ||||
Bottom oxygen | 1.000 | 4.994 | 0.025 | * | |||
Chlorophyll | 1.004 | 0.918 | 0.342 | ||||
Exposure | 1.927 | 22.167 | < 0.001 | *** | |||
Location x Time step (1st) | 1.331 | 2.314 | 0.205 | ||||
Location x Time step (2nd) | 1.812 | 15.147 | 0.017 | * | |||
Location x Time step (3rd) | 9.695 | 198.967 | < 0.001 | *** | |||
Gear | 0.889 | 93.769 | < 0.001 | *** | |||
0.498 | 51.7% | 762 |
Finally, both methods showed similar values of explained variance (i.e. 51.7% for GAMM and 53.8% for RF). The cross-validation analysis demonstrated a better overall performance for RF, illustrated by lower mean squared error of predicted round goby distinctiveness compared to observation not used for model training (Suppl. material
The degree to which NIS display similar or dissimilar traits compared to native species of recipient communities is debated largely due to contrasting results from available studies, primarily conducted in terrestrial ecosystems (
Although round goby is not generally different from the regional pool of native species in terms of its overall trait composition, we observed notable differences in terms of individual trait modalities, primarily by display of territorial behaviour and parental care. This indicates that native species generally display a reproductive strategy that does not involve defending a territory, nor protecting their offspring. More specifically, round goby males display several types of parental care, including egg inspection, ventilation and nest guarding (
While generally similar to native species at the regional scale, our study demonstrates pronounced spatio-temporal variation in terms of local distinctiveness of round goby between and within sampling locations over time. The wide range of values (i.e. from < 0.2 to > 0.6) indicates that round goby can locally co-exist with native species that are either functionally similar or different to itself, reflecting its broad environmental tolerance (
The ability to colonise a broad range of habitats and therefore co-exist with different pools of native species with different trait composition may help explain the derived relationships with the biotic variables included in our statistical analysis. For instance, the negative effect of species richness likely reflects the higher local distinctiveness of round goby when co-occurring with the fewer and functionally more dissimilar marine species from colder and deeper locations. Contrarily, when found together with the more species from the native community at more shallow and warmer locations, the likelihood of round goby co-occurring with more functionally similar species is higher, thus explaining its lower local level of distinctiveness at higher richness. This is likely facilitated also by the strong relationship between species and functional richness in the Baltic Sea region (
In summary, the application of this trait-based approach to the case of round goby in the Baltic Sea shows a partial (trait) niche overlap with native fish species that appears to increase locally when round goby occurs with communities from shallow, inshore and warmer areas. Despite this partial overlap with native species, we also demonstrated that round goby shows pronounced differences compared to the native community in its display of parental care and territorial behaviour. Such differences could play an important underlying role behind round goby’s invasion success in the Baltic, as well as in defining the type of interactions with native species. Based on our results, non-aggressive native species that partially share their niche with round goby might be harmed or displaced in the case of direct competition with this NIS for similar resources (e.g. feeding grounds, sheltered areas, nesting sites). Due to the context dependence (i.e. the species and traits selected) of this study, caution should be taken when expanding our conclusions to different scenarios of round goby invasion. For that reason, we encourage the use of similar trait-based approaches, based on functional distinctiveness to further address the invasion of round goby in other areas, with a different environment and species composition (e.g. the North-American Great Lakes or central European rivers). If similar patterns emerge, this would contribute to the understanding of why this species has managed to successfully establish in such different regions, as well as a better understanding if round goby shows similar interactions with native fishes in other areas. Additionally, investigating how round goby dominance could be affected when it co-exists with more similar or dissimilar native species in local communities could also be valuable to define the niche or conditions that this species needs to become invasive (
We wish to thank the Swedish Agency for Marine and Water Management and OKG Aktiebolag for funding the regional and national monitoring programmes together with all the people involved who sorted, identified and measured the included species in the database, ensuring the access and availability to high-quality time-series data. In addition, we would like to thank all the reviewers involved in the revision process of this work, as all the comments and feedback provided helped to improve the quality of this manuscript. A.V.P and M.L. acknowledge financial support from the European Union’s Horizon 2020 projects “Mission Atlantic” (ID: 862428) and “B-USEFUL” (ID: 101059823).
Supplementary information
Data type: docx
Explanation note: table S1. List of the different modalities for the diet trait. The colored cells indicate the specific modality combination that is displayed by a certain number of species within the regional pool. Numbers inside colored cells indicate the probability that a certain species displays such modality. table S2. Trait values for all fish species. The included traits and modalities are further described in Table