Science of the Total Environment 639 (2018) 58–66

Contents lists available at ScienceDirect

Science of the Total Environment

j ourna l homepage: www.e lsev ie r .com/ locate /sc i totenv

Temperature and hydrologic alteration predict the spread of invasiveLargemouth Bass (Micropterus salmoides)

Mi-Jung Bae a,⁎, Christina A. Murphy a,b, Emili García-Berthou a

a GRECO, Institute of Aquatic Ecology, University of Girona 17003, Girona, Catalonia, Spainb Department of Fisheries and Wildlife, Oregon State University, Corvallis, OR 97331, USA

H I G H L I G H T S G R A P H I C A L A B S T R A C T

• The distribution of Largemouth Bass, aninvasive species, was examined usingspecies distribution models.

• The most influential predictor of bassdistribution in the Iberian Peninsulawas temperature.

• Larger volumes of local and upstreamreservoirs also increased predictedpresence.

• Understanding the drivers promotingthe establishment of this global invaderwill be important in identifying areasat risk.

E-mail address: mjbae@nnibr.re.kr (M.-J. Bae).

https://doi.org/10.1016/j.scitotenv.2018.05.0010048-9697/© 2018 Elsevier B.V. All rights reserved.

a b s t r a c t

a r t i c l e i n f o

Article history:Received 30 January 2018Received in revised form 16 April 2018Accepted 1 May 2018Available online xxxx

Editor: Daniel Wunderlin

The successful establishment of an aquatic invasive alien species can bemediated by a suite of environmental fac-tors, including climate and anthropogenic disturbance. Dams and reservoirs are thought to promote freshwaterfish invasion success through hydrological alterations but the evidence for their role in the global invasion ofLargemouth Bass (Micropterus salmoides) on a landscape scale is limited. Here, we examine the distribution ofLargemouth Bass, one of the most widely introduced fish in the world, from the Iberian Peninsula using speciesdistributionmodels (SDMs), including an ensemble forecast. We used these models to test the role of twelve en-vironmental predictors expected to influence the distribution of Largemouth Bass, including the reservoir storagecapacity at local and upstream reaches. We found that the predictive accuracy, based on AUC criteria, of the en-semblemodelwas higher than any of the six individual SDMs for LargemouthBass. Themost influential predictorof bass distribution included in our model of the Iberian Peninsula was temperature, where warmer tempera-tureswere generally associatedwith bass presence, and cooler temperatureswith absence. In addition towarmertemperatures, increasing storage of local and upstream reservoirs increased predicted presence, suggesting animportant role of reservoirs in mediating the invasive success of this fish. Our results indicate that although nat-ural climatic factors may be crucial in the successful invasion of Largemouth Bass, hydrological alteration(e.g., regulated flow regimes and lentic habitats associated with dams and reservoirs) may be important. Under-standing the drivers promoting the establishment of this global invader will be important in identifying areas atrisk and in developing future efforts to control its spread, especially when those drivers are ongoing anthropo-genic disturbances such as the construction and operation of dams and reservoirs.

© 2018 Elsevier B.V. All rights reserved.

Keywords:DisturbanceInvasive alien speciesEnsemble forecastingNatural flow regimeReservoirsSpecies distribution modelling

tional Institute of Biological Resources, Gyeongsangbuk-do 37242, Republic of Korea.

⁎ Corresponding author at: Freshwater Biodiversity Research Division, Nakdonggang Na

59M.-J. Bae et al. / Science of the Total Environment 639 (2018) 58–66

1. Introduction

Invasive alien species drive freshwater biodiversity loss and haveenormous economic costs worldwide (Williamson, 1999; Mooney andHobbs, 2000; Simberloff et al., 2013). Anthropogenic disturbancessuch as land use change, river channelization, fragmentation, andwater abstraction produce changes in natural temperature and flow re-gimes of freshwater ecosystems (Poff et al., 2007; Bae et al., 2016).These and other anthropogenic alterations may promote the invasionof exotic species (Dudgeon et al., 2006; Vörösmarty et al., 2010). Damscreate novel lentic habitats and modify the timing and magnitude ofdownstream flows, in general greatly reducing seasonal and inter-annual variability (Poff and Hart, 2002; Poff et al., 2007). Dams mayalso alter temperature regimes, in general reducing seasonal variabilityand often increasing temperatures during winter periods (Bunn andArthington, 2002; Poff and Hart, 2002; Olden and Naiman, 2010).Dams and associated reservoirs may promote invasive alien speciesestablishment because many successful freshwater invaders are morelimnophilic (i.e. preferring lentic habitats) and thermophilic(i.e., thriving at relatively high temperatures) than the native speciesthey replace in hydrologically altered systems (Bunn and Arthington,2002; Vila-Gispert et al., 2005; Olden et al., 2006; Boix et al., 2010;Gido et al., 2013). There is an urgent need to improve our understandingof biological invasions, both to reduce future invasions and to predicttheir ecological effects (Shea and Chesson, 2002; Simberloff et al., 2013).

Species distributionmodels (SDMs) are increasingly used as a tool toexplain and predict the patterns and processes of biological invasions(Rodríguez et al., 2007; Smolik et al., 2010; Capinha and Anastácio,2011). SDMs relate species distribution data (occurrence or abundanceat known locations) with information on the environmental or spatialcharacteristics of those locations (Elith and Leathwick, 2009). In thefield of invasion biology, SDMs can be used to predict the potential dis-tribution of introduced species (Ficetola et al., 2007; Mika et al., 2008;Bradley, 2009), and to compare the invasive potential of different inva-sion stages (Václavík and Meentemeyer, 2012). SDMs also indicateimportant environmental drivers of distribution, which can inform in-vasive alien species management (Guisan et al., 2013).

Here, we analyze the factors mediating the invasion of LargemouthBass (Micropterus salmoides), a centrarchid species native to parts ofNorth America (Page and Burr, 1991). Largemouth Bass are among theten most frequently introduced aquatic species worldwide (García-Berthou et al., 2005) and are now found on all continents exceptAntarctica (Lever, 1996). Largemouth Bass are an apex predator inmost introduced streams and lakes (Carpenter and Kitchell, 1993;García-Berthou, 2002) and can cause trophic cascades that change com-munity structure (Carpenter and Kitchell, 1993; Ahrenstorff et al.,2009). Largemouth Bass can also alter the foraging behavior of nativefish (MacRace and Jackson, 2001), compete with native piscivores(Bacheler et al., 2004), and extirpate or decrease the abundance ofnative species (Maezono and Miyashita, 2003). Largemouth Bass areconsidered a warm-water species (Coutant, 1975; Brown et al., 2009;Cooke and Philipp, 2009), but species distribution models analyzingthe influence of temperature and other environmental factors onLargemouth Bass distribution are very limited (see Iguchi et al., 2004for the known example). Therefore, we assessed the influence of envi-ronmental factors on the invasion of Largemouth Bass outside of theirnative range, using the Iberian Peninsula as a case study.

Largemouth Bass were first introduced in Spain in 1955 (Elvira andAlmodóvar, 2001) for sport fishing activities. Many 20th century intro-ductions favored preferred fishing locations such as reservoirs(Godinho et al., 1998; Marta et al., 2001; Copp et al., 2005) andLargemouth Bass are now estimated to occur in about 50% of all reser-voirs in the Iberian Peninsula (Clavero et al., 2013). Although initialintroductions likely favored lentic habitats, Largemouth Bass are nowcommon in lotic areas as well (Hermoso et al., 2008).The IberianPeninsula offers a large region (ca. 582,000 km2) with strong spatial

heterogeneity in climate and anthropogenic disturbance (Ferreiraet al., 2007), providing an excellent opportunity to examine environ-mental factors influencing Largemouth Bass invasion and persistence.There are over 1200 large dams (storage capacity N 1 hm3) with atotal capacity of ca. 64,000 hm3 in the Iberian Peninsula (Berga-Casafont, 2003). The identification of environmental factors, especiallyfactors related to anthropogenic alterations, that promote or mediateits spread is critical to develop effective management strategies and toreduce the effects of Largemouth Bass on freshwater ecosystemsworldwide.

Our objectiveswere: 1) to identify the influential environmental fac-tors that regulate Largemouth Bass distribution in the Iberian Peninsula,2) to test whether hydrologic alteration influences Largemouth Bass oc-currence, and 3) to determine the potential for range expansion ofLargemouth Bass in the region. These objectives were designed to in-formmanagement efforts to identify areas at risk and to limit the furtherspread of this species. Based on the known habitat preferences forLargemouthBass,we hypothesized thatwarm temperatures and hydro-logic alteration would promote their invasion success at a landscapescale. Specifically, we predicted that bass would be more frequent:1) at temperatures known to maximize growth and performance with-out increasingmortality (i.e., around 15–25 °C) and 2) in reservoirs or inregulated rivers downstreamof reservoirswhich provide lentic habitatsand reaches with fewer extreme hydrological events.

2. Methods

2.1. Largemouth Bass data

Largemouth Bass occurrence data in the Iberian Peninsula weremainly obtained from the Spanish (Doadrio et al., 2011) and Portuguese(Ribeiro et al., 2007) national databases. This information wascomplemented with searches in the Global Biodiversity Information Fa-cility (GBIF, http://www.data.gbif.org/, last accessed in February 2014),published papers and our own unpublished records (Table S1). Werestricted data to occurrence records from 2000 to 2010. We used a10 × 10 km UTM (Universal Transverse Mercator) resolution, whichwas the finest resolution available for the majority of the LargemouthBass occurrence data. Overlapping or duplicate records within 10 kmUTM cells were removed to allow only one occurrence per UTM unit.A total of 590 occurrence records of Largemouth Bass were thus ob-tained for the Iberian Peninsula and included in the final database(out of 6138 grid cells, Fig. 1e).

2.2. Environmental data

We collected environmental data expected to determine the distri-bution of Largemouth Bass based on the literature. All the environmen-tal datawere obtained fromonlinedatabases (Table 1) and extracted forSDM with the Spatial Analyst toolbox in ArcGIS 10 (ESRI, 2009). Whenavailable environmental data were finer grain than the records forLargemouth Bass occurrences (10 km UTM), average values for each10 × 10 km grid cell were computed. Because collinearity can compro-mise the reliability of SDM (Heikkinen et al., 2006), only one variablefrom any pair of strongly correlated variables (i.e. Pearson's r ≥ |0.75|)was retained (Cord and Rödder, 2011; Dormann et al., 2013; Filipeet al., 2013). We based our decision on which to retain on literatureand our expert opinion. A total of 12 environmental predictors passedthe collinearity threshold (Table S2) and were used for SDM develop-ment (Table 1). Largemouth Bass occurrence can be influenced bymany environmental variables across various spatial scales (Maceinaand Bettoli, 1998; Suski et al., 2006; Taylor et al., 2014) (Table 1).Water temperature, water level and pool area have been reported ascritical factors for survival (Aggus and Elliott, 1975; Garvey et al.,2000), spawning (Kramer and Smith, 1960; Post et al., 1998), growth(Olson, 1996), and density (Sowa and Rabeni, 1995) of Largemouth

Fig. 1. The four environmental variables with highest variable importance in species distribution models (a) annual mean temperature, (b) upstream reservoir capacity, (c) flowaccumulation, and (d) local reservoir capacity and (e) observed and (f) predicted distribution (ensemble model) of Largemouth Bass in the Iberian Peninsula. The predicteddistribution of individual SDMs is provided in Fig. S1.

60 M.-J. Bae et al. / Science of the Total Environment 639 (2018) 58–66

Bass in its native range (Parkos and Wahl, 2002). However, many ofthese variables vary strongly at finer spatial and temporal scales thatcould not be obtained or extracted for the Iberian Peninsula as awhole. In this case, we used air temperature as representative ofwater temperature and precipitation as representative of water leveland discharge (Garvey et al., 2000). We also selected slope and flowaccumulation, measured as the number of cells flowing into eachdown-slope cell, as a proxy of drainage area (Domisch et al., 2011).Solar radiation, previously reported as influential for spawning andgrowth, was also included (Havens et al., 2005). Finally, as indicators

of anthropogenic disturbance including hydrologic alteration due todams and land use change, we compiled data on local reservoir capacity(the volume of water stored in each UTM cell), upstream reservoir ca-pacity (the accumulated volume of water stored in reservoirs upstreamof each UTM cell), human population density, and agricultural andurban land uses. Local reservoir capacity was included to measure thedirect influence of reservoirs onbass occurrence,whereas upstream res-ervoir capacity was included to approximate the degree of hydrologicalalteration affecting a site, whichmight promote bass invasion by reduc-ing extremes in flow (Batalla et al., 2004) and temperature (Prats et al.,

Table 1Environmental variables used (and their mean and standard deviation) in SDMs for Largemouth Bass in the Iberian Peninsula from all 10 km2 UTM grid cells. When environmental datawere originally of finer grain than our species data, data were downsampled using the mean of values within each grid cell. All values indicate the mean and standard deviation (in pa-renthesis) values of environmental variables in the Iberian Peninsula, Presence reports themean and standard deviation for cells with detected presence of Largemouth Bass and Absencerepresents those values for the pseudo-absences used in the SDMs. * indicates layers calculated from source data using ArcGIS geoprocessing tools.

Environmental variables Sourcea Abbreviation All Presence Absence

ClimateAnnual mean temperature (°C) 1 AnMeanTemp 13.6 (2.7) 15.2 (1.8) 13.4 (2.7)Annual temperature range (°C) 1* AnRangeTemp 11.9 (1.1) 12.2 (0.9) 11.9 (1.1)Annual precipitation (mm) 1 AnPrecip 713.9 (321.7) 636.1 (205.0) 729.4 (335.4)Solar radiation (mm/day) 1 SolarRad 2027.5 (28.8) 2032.3 (17.6) 2027.9 (28.0)

TopographySlope (°) 2* Slope 5.4 (4.5) 4.2 (3.2) 5.5 (4.6)Topographic index 2* TopoIndex 2.1 (0.6) 2.3 (0.6) 2.0 (0.5)Flow accumulation 2* FlowAcc 1321.3 (5489.0) 4893.4 (10,044.6) 1054.5 (5139.3)

Anthropogenic disturbanceUpstream reservoir capacity (km3) 3,4* MaxResVol 410.9 (1685.2) 1561.0 (3186.6) 316.5 (1453.0)Local reservoir capacity (km3) 3,4* ResLocVol 11.6 (110.0) 38.8 (208.9) 8.5 (107.2)Population density (people/km2) 5 Population 83.1 (322.1) 80.9 (242.2) 86.5 (393.5)Agricultural land use (%) 6 Agric% 48.7 (32.7) 53.5 (29.1) 47.3 (33.1)Urban land use (%) 6 Urban% 1.4 (5.8) 1.4 (5.1) 1.5 (6.4)

a Sources: 1=UniversitatAutònomadeBarcelona, Atlas climático digital de la Península Ibérica (http://www.opengis.uab.es/); 2=SpanishNational Center forGeographic Information(http://centrodedescargas.cnig.es/); 3=Melo andGomes (1992); 4= SpanishMinistry of Agriculture, Food and Environment (http://sig.magrama.es/); 5=DIVA-GIS data (http://www.diva-gis.org/datadown); 6 = National Center of Geographical Information (https://www.cnig.es).

61M.-J. Bae et al. / Science of the Total Environment 639 (2018) 58–66

2010). All the environmental predictors were transformed into z scoresto standardize themeasurement scales of the inputs and reduce their ef-fects on the SDM results (Vander Zanden et al., 2004).

2.3. Modelling approach

To develop the SDM of Largemouth Bass, we used the ‘biomod2’package (Thuiller, 2003; Thuiller et al., 2009) in R (http://cran.r-project.org) (R Core Team, 2013).We used six SDMs that have been fre-quently applied in a variety of taxa: Generalized linear models (GLM),generalized additive models (GAM), boosted regression trees (BRT), ar-tificial neural networks (ANN), multiple adaptive regression splines(MARS), and Random Forests (RF). GLM are an extension of linearmodels to allow for non-normal errors and heteroscedasticity(McCullagh and Nelder, 1989) and we used them with binomial distri-bution and a logit function to model presence-absence data (Thuiller,2003). GAMare a non-parametric extension of GLM that use a smootherto fit nonlinear functions (i.e., Spline function) (Hastie and Tibshirani,1990), they have been widely applied in biogeographic studies(e.g., Araújo et al., 2004; Thuiller et al., 2006). BRT, more recently intro-duced in ecology, combine the strengths of regression trees andboosting (Ridgeway, 1999; Elith et al., 2008) by proceeding through se-quential improvements using a numerical optimization algorithm tominimize a loss function (e.g., deviance) and add a new tree at eachstep (Elith et al., 2008). ANN are a powerful rule-based modelling tech-nique (Lek andGuégan, 1999). Compared to logistic regression or lineardiscriminant analysis, ANN have displayed higher predictive powerwhenmodelling nonlinear relationships (Olden and Jackson, 2002). Be-cause ANN can be applied to a variety of data types with nonlinear asso-ciations, ANN have been increasingly used in SDM (Thuiller, 2003;Heikkinen et al., 2006). MARS combine linear regression, mathematicalconstruction of splines and binary recursive partitioning to fit localmodels in which the relationships between the response and predictorscan be linear or nonlinear (Friedman, 1991). The purpose of MARS is totry to determine the appropriate intervals to run independent linear re-gressions, for each predictor, and identify interactions while avoidingoverfitting the data (Briand et al., 2004). RF (Breiman, 2001) aremodel-averaging approaches where bootstrap samples are drawn toconstruct multiple trees, grown with a randomized subset of predictors(Prasad et al., 2006). RF have shown better prediction accuracy withminimal overfitting than many other SDM techniques (Cutler et al.,2007; Marmion et al., 2009).

The performance and spatial predictions of SDMs depend on uncer-tainties from a number of factors such as measurement errors, samplesize, sample representativeness (Edwards et al., 2006; Marmion et al.,2009) and the statistical techniques used (Thuiller et al., 2004a,2004b). These uncertainties can beminimized by includingmultiple en-vironmental drivers, using appropriate ranges of data (Filipe et al.,2013), applying the suitable spatial resolution data reflecting theecological knowledge (e.g., dispersal ability) of the study taxon andconducting standardization or normalization of environmental vari-ables prior to SDM construction (Peterson et al., 2011). In addition, en-semble forecasting, combining the output of multiple individual SDMs(e.g., using means, medians or weighted averages, Araújo and New,2007), is used to overcome “prediction uncertainty” from differentmodelling techniques (Pearson et al., 2006; Carvalho et al., 2010) andgenerally increases prediction accuracy compared with any individualSDM (e.g., Marmion et al., 2009; Grenouillet et al., 2011).

Our data did not contain completely reliable absence locations ofLargemouth Bass because of imperfect capturability and inconsistentsampling effort, so we generated pseudo-absences among non-presence or background grid cells (n = 1850 of 6138 total cells)(Phillips et al., 2009; Barbet-Massin et al., 2012). Following Barbet-Massin et al. (2012), we used random selection of pseudo-absences asthis method yields themost reliable distributionmodels. Species occur-rence data were divided into a training set (70%) and a testing set (30%)(Araújo et al., 2005) and each model was replicated 10 times to avoidbias from the data split. Model performance was evaluated based onthe area under the curve (AUC) of the receiver-operating characteristic(ROC) (Swets, 1988), which ranges from 0 to 1. As a rule of thumb, AUCvalues above 0.9 indicate an excellent predictionmodel, whereas valuesbetween 0.7 and 0.9 indicate a fair model, and values below 0.7 indicatepoor model performance (Swets, 1988).

We estimated the importance of the environmental predictors fromall species distribution models with a permutation procedure availablein the “variables_importance” function of biomod2 (Thuiller et al.,2009). This procedure starts with the predictions from the trained (i.e.calibrated) model, randomizes each variable separately, and comparesthe new predictions using a randomized variable with the original pre-dictions based on the Pearson correlation coefficient (r). The variableimportance measure is obtained as 1 − r where higher values indicatemore influential variables (Thuiller et al., 2009). The probability ofLargemouth Bass occurrence across the range of variation for each envi-ronmental predictor was also examined using response curves for eachindividual SDM (Elith et al., 2005).

62 M.-J. Bae et al. / Science of the Total Environment 639 (2018) 58–66

We then applied ensemble modelling, which can increase the reli-ability of predictions as well as decrease model-based uncertainty(Thuiller, 2004; Grenouillet et al., 2011). For the ensemblemodel amin-imum AUC value of 0.7 was used to select adequate models (Araújoet al., 2005), the average of these SDM was then used to construct theensemble model and provide more robust forecasts (Marmion et al.,2009, Grenouillet et al., 2011). Finally, the AUC statistic was also usedto analyze the accuracy of the ensemble model and to compare its per-formance to the individual SDM.

3. Results

For the six modelling techniques applied, average AUC scores werefair (around0.8; Fig. 2). RF overall showed thehighest predictive perfor-mance based on AUC (0.838), followed by BRT (0.828), MARS (0.815),ANN (0.789), GLM (0.778), and GAM (0.774). As expected, the predic-tive accuracy (0.908) of the ensemble model based on AUC was higherthan any individual SDM.

Of the 12 environmental predictors, annual mean temperature (av-erage importance ± standard error: 0.320 ± 0.013), flow accumulation(0.120± 0.011), upstream reservoir capacity (0.143± 0.013), and localreservoir capacity (0.108± 0.005) had the greatest effects on predictedbass presence (Fig. 3). Other predictors, more related to the positionalong the river network, such as slope (0.035± 0.006) and topographicindex (0.054± 0.006) or to anthropogenic disturbance, such as popula-tion density: 0.013 ± 0.003, agricultural: 0.048 ± 0.006 and urban:0.015 ± 0.003 land uses were poor predictors of Largemouth Basspresence.

The shape of the response curves of RF and BRT, which had thehighest individual model performances based on AUC, were also themost similar among the models (Fig. 4). Most response curves of GLMand GAM, which had comparatively poor predictive performances,were also similar, though some predictors presented different shapes(e.g., flow accumulation, upstream reservoir capacity, local reservoir ca-pacity and urban land use) and contrasting predicted probabilities. Thebest performing model (i.e. ensemble model) predicted LargemouthBass tomore frequently occur in locationswith relatively highmean an-nual temperatures (from ca. 14.9 to 17.4 °C), intermediate levels of an-nual precipitation, and thermal ranges above 12.1 °C (Fig. 1).Largemouth Bass were less likely to occur at lower flow accumulations(i.e. headwaters), and where few large reservoirs were found locallyor upstream.

The ensemble model projection reproduced the reported occur-rences of Largemouth Bass in the Iberian Peninsula (Fig. 1). The geo-graphical areas with the highest environmental suitability weremostly in the south-western and eastern parts of the Iberian Peninsula.

Fig. 2.Model performance as determined by the mean area under the curve (AUC) of thereceiver-operating characteristic for species distribution models of Largemouth Bass:Generalized linear model (GLM), generalized additive model (GAM), boosted regressiontrees (BRT), artificial neural networks (ANN), multiple adaptive regression splines(MARS), and Random Forests (RF). Error bars are standard deviations.

Largemouth Bass were much less frequent in the northwest, corre-sponding to the Atlantic climate zone with much higher annual precip-itation (N1000 mm) and cooler temperatures, and in the southeast,corresponding to a semi-arid climate with some of the lowest rainfallin Europe (b300 mm per year). The mainstem of large Mediterraneanrivers (Douro, Tagus, Guadiana, Guadalquivir, and Ebro), with high up-stream reservoir capacity (i.e. hydrologic alteration) had the highestsuitability (Fig. 1).

4. Discussion

4.1. Largemouth Bass invasion

Temperature was the most important factor for predictingLargemouth Bass distribution. Largemouth Bass were less prevalent inthe cool, wet northeastern Iberian Peninsula, and in the hot, dry regionof southeastern Spain. This is not surprising, as many Largemouth Basslife history traits such as spawning, breeding, growth and activity areknown to depend on water temperature (Coutant, 1975; Brownet al., 2009; Cooke and Philipp, 2009). All our models predictedthat the occurrence of bass is unlikely below 10 °C mean air temper-ature and that it is most likely to occur in areas with temperaturesfrom14 to 18 °C. Temperature partly explains themore invasive charac-ter of this fish species in Mediterranean countries in contrast to manyparts of northern Europe. The Iberian Peninsula has amostlyMediterra-nean climate with mild winters and dry, warm summers; the tempera-tures required for Largemouth Bass growth are maintained for manymonths.

Water temperature would probably be a better predictor ofLargemouth Bass occurrence than air temperature, but was not gener-ally available. Nearly all SDMs for freshwater species have used air in-stead of water temperatures for this reason (e.g., Buisson et al., 2010;Capinha and Anastácio, 2011; Comte and Grenouillet, 2013). Althoughair and water temperatures are often well correlated, this is not alwaysthe case (Carmona-Catot et al., 2014; Bae et al., 2016) and we wouldrecommend the use of water temperature for SDM constructionwhere possible. Although temperature ranges, including minimumtemperatures, may be important for Largemouth Bass, we did not useminimum (air) temperature because it was strongly correlated tomean air temperature and its effects likely depend on acclimation tem-perature (Beitinger et al., 2000). In addition, more data of speciespresence-absence and temperatures at finer grains would be neededto separate the differential effects of minimum vs. mean temperatures(e.g. overwinter mortality vs. seasonality in growth; Post et al., 1998,Fullerton et al., 2000, Lookingbill and Urban, 2003, Cooke and Philipp,2009).

Following temperature, the upstream and local reservoir capacitieswere among the most influential variables predicting Largemouth Bassoccurrence. These anthropogenic indicators of hydrologic alterationhave been generally neglected in previous SDM studies even though ithas been continuously reported that these alterations may facilitatespecies invasions (Murphy et al., 2015). Local reservoir capacity indi-cated the presence and size of local reservoirs, whereas upstream reser-voir capacity, the cumulative volume of reservoirs upstream,was aimedat describing the degree of modification of the natural flow regime andother ecological features resulting from upstream impoundment.

In our study, both local and upstream reservoir capacities appear toaffect the prevalence of bass, although the latter appeared slightly moreimportant. The strong relationship between reservoir-related factorsand the occurrences of Largemouth Bass can be explained by altered en-vironmental conditions (i.e., the modification of natural flow and ther-mal regimes; Poff et al., 2007) and increased propagule pressure(i.e., the size and frequency of introductions) (Williamson and Fitter,1996; Simberloff, 2009; Woodford et al., 2013). The presence of localreservoirs likely modifies available habitat to the detriment of nativespecies which are more adapted to riverine conditions, and provides

Fig. 3.Mean variable importance and standard error from the six species distributionmodels for Largemouth Bass (GLM, GAM, BRT, ANN, MARS, and RF). Abbreviations of environmentalvariables are provided in Table 1.

63M.-J. Bae et al. / Science of the Total Environment 639 (2018) 58–66

lentic habitat for invasive alien species, frequently introduced for recre-ational opportunities including sport fishing (Havel et al., 2005; Raheland Olden, 2008). In the Iberian Peninsula, sport fishing activities arecommon in reservoirs (Marta et al., 2001) where Largemouth Basshave been stocked since the 1950s (Leunda, 2010). It is assumed thatLargemouth Bass and other gamefish continue to be illegally introducedinto reservoirs (Clavero and Hermoso, 2011). According to Clavero andHermoso (2011), invasive species, such as Largemouth Bass, dominateIberian reservoirs (71% of species richness) but not rivers and streams(33% of species richness). Most unimpounded Iberian freshwater eco-systems display Mediterranean extremes; during the summer dry sea-son stream flow is often low or ceases, whereas unpredictable floodsare common during spring and autumn. These extremes structureMed-iterranean freshwater communities (Gasith and Resh, 1999; Magalhãeset al., 2002) and exotic species are less likely to occur where these ex-tremes are the greatest (e.g., Marchetti and Moyle, 2001; Bernardoet al., 2003; Vila-Gispert et al., 2005; Olden et al., 2006). By contrast, ithas been proposed that Largemouth Bass thrive in Iberian regulatedstreams and reservoirs because of damped flow variation, availabilityof naive prey, and low predation pressure (Godinho et al., 2000;Almeida et al., 2012). Hydrologic alteration favors Largemouth Bassand similar limnophilic species by producing more suitable habitats

Fig. 4. Largemouth Bass response curves with four influential environmental predictors for theprovided in Fig. S2. Abbreviations of environmental variables are provided in Table 1.

and reducing their flood and drought induced mortality rates (Propstand Gido, 2004; Kiernan et al., 2012; Gido et al., 2013; Taylor et al.,2014).

4.2. Species distribution modelling

Modelling accuracy (based on AUC and ecological interpretation ofresults) of all individual SDM was weaker than for the ensemblemodel, indicating that the various modelling techniques did not pro-duce equivalent and equally plausible predictions (Roura-Pascualet al., 2009;Meller et al., 2013). The newer techniques ofmachine learn-ing, such as RF and BRT, consistently outperformed other modellingmethods, as in many other applications (e.g., Cutler et al., 2007;Marmion et al., 2009; Markovic et al., 2012). This is partly becausethese machine learning techniques combine several modelling algo-rithms, average the results of many models, and have fewer assump-tions than more classical techniques such as GLM (Prasad et al., 2006).An individual model's suitability (e.g., a model with the most accurateprediction) can depend on features of the study species (e.g., specialistvs. generalist, invasive vs. native species) or its spatial distribution(e.g., geographical range modelled) (e.g., Thuiller et al., 2004a; Melleret al., 2013). Therefore, ensemble modelling often provides more

six spatial distribution modelling techniques. The response curves of all the predictors are

64 M.-J. Bae et al. / Science of the Total Environment 639 (2018) 58–66

accurate results than the best individual modelling techniques (RF inour study) (Marmion et al., 2009).

In spite of the high predictive accuracy of our model, the perfor-mance of our model might be improved should additional data becomeavailable. First, the number and location of pseudo-absences can influ-ence model predictions (e.g., Graham et al., 2008; Lomba et al., 2010).Confirming “true” absences is very difficult in mobile species and re-quires higher sampling effort to assure their reliability compared withpresence data (Elith and Leathwick, 2009). In our study, we appliedlarger numbers of pseudo-absences (i.e. 1/3 of background data) thanpresences based on the recommendation of Barbet-Massin et al. (2012)and pseudo-absences were downweighted in order to emulate anequal number of presences and pseudo-absences (Ferrier et al., 2002;Lomba et al., 2010). Second, even though it is well known that climaticfactors have a more significant role at larger spatial scales (Guisan andThuiller, 2005), it is reasonable that biological factors (e.g., competitionand predation) as well as unmeasured environmental factors also con-tribute to Largemouth Bass distribution. Physical habitat (e.g., substratecomposition and riparian vegetation) and water quality (e.g., turbidity,pH, dissolved oxygen, and eutrophication) were not directly consideredin our study although they are known to affect the local distribution ofbass in individual lakes and reservoirs (Hanson and Butler, 1994;Shoup and Wahl, 2009; Gaeta et al., 2011). Third, dispersal limitation isgenerally not considered in SDM (Filipe et al., 2013) which may resultin an overestimation of the occurrence probability in some places. How-ever, our predictions seem statistically accurate and ecologically sound.

The importance of temperature and hydrologic alteration onLargemouth Bass invasion has many management implications. For in-stance: i) temperature is a key factor in habitat suitability for bass andthermal pollution might promote bass invasion (Bae et al., 2016); ii)building new reservoirs may increase the distribution of bass andother invasive limnophilic species, whereas removing dams might actas a controlling measure; iii) mimicking the natural flow regime andpreserving floods and droughts might provide an avenue for managingLargemouth Bass invasions in regulated rivers; iv) the spatial projec-tions based on the ensemble model results identify highly suitableareas for bass where it has not been recorded, supporting the need fortargeted surveys (Guisan et al., 2006) and providing basic informationfor managers of areas at risk (Farnsworth and Ogurcak, 2006). Our re-sults are also relevant for over 45 countries where Largemouth Bass ap-pear to have been successfully introduced (García-Berthou et al., 2005),by showing the environments where this species is likely to invade andhow reservoir construction is likely to promote its invasion. Althoughnatural abiotic factors such as temperature and habitat are important,hydrologic alteration through reservoir construction creates favorableconditions for this highly invasive alien species. Managing ongoing an-thropogenic disturbances such as dams and reservoirs may be criticalto future efforts to control the spread of Largemouth Bass.

Acknowledgements

We gratefully thank everybody who contributed with data onpresence of Largemouth Bass, particularly Ignacio Doadrio and FilipeRibeiro, and two anonymous reviewers for helpful comments on themanuscript. This research was financially supported by the SpanishMinistry of Economy and Competitiveness (projects: CGL2013-43822-R; CGL2015-69311-REDT; CGL2016-80820-R; and ODYSSEUS,BiodivERsA3-2015-26, PCIN-2016-168) and the Government of Catalo-nia (ref. 2014 SGR 484 and 2017 SGR 548). MJB benefited from apostdoctoral grant from the European Commission (Erasmus MundusPartnership “NESSIE”, 372353-1-2012-1-FR-ERA MUNDUS-EMA22).

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.scitotenv.2018.05.001.

References

Aggus, L.R., Elliott, G.V., 1975. Effects of cover and food on year-class strength oflargemouth bass. In: Stroud, R., Clepper, H. (Eds.), Black Bass Biology and Manage-ment. Sport Fishing Institute, Washington, D.C., pp. 317–322.

Ahrenstorff, T.D., Sass, G.G., Helmus, M.R., 2009. The influence of littoral zone coarsewoody habitat on home range size, spatial distribution, and feeding ecology oflargemouth bass (Micropterus salmoides). Hydrobiologia 623, 223–233.

Almeida, D., Almodóvar, A., Nicola, G.G., Elvira, B., Grossman, G.D., 2012. Trophic plasticityof invasive juvenile largemouth bass Micropterus salmoides in Iberian streams. Fish.Res. 113, 153–158.

Araújo, M.B., New, M., 2007. Ensemble forecasting of species distributions. Trends Ecol.Evol. 22, 42–47.

Araújo, M.B., Cabeza, M., Thuiller, W., Hannah, L., Williams, P.H., 2004. Would climatechange drive species out of reserves? An assessment of existing reserve-selectionmethods. Glob. Chang. Biol. 10, 1618–1626.

Araújo, M.B., Pearson, R.G., Thuiller, W., Erhard, M., 2005. Validation of species–climateimpact models under climate change. Glob. Chang. Biol. 11, 1504–1513.

Bacheler, N.M., Neal, J.W., Noble, R.L., 2004. Diet overlap between native bigmouthsleepers Gobiomorus dormitor and introduced predatory fishes in a Puerto Rico res-ervoir. Ecol. Freshw. Fish 13, 111–118.

Bae, M.-J., Merciai, R., Benejam, L., Sabater, S., García-Berthou, E., 2016. Small weirs, big ef-fects: disruption ofwater temperature regimeswith hydrological alteration in aMed-iterranean stream. River Res. Appl. 32, 309–319.

Barbet-Massin, M., Jiguet, F., Albert, C.H., Thuiller, W., 2012. Selecting pseudo-absences forspecies distribution models: how, where and how many? Methods Ecol. Evol. 3,327–338.

Batalla, R.J., Gómez, C.M., Kondolf, G.M., 2004. Reservoir-induced hydrological changes inthe Ebro River basin (NE Spain). J. Hydrol. 290, 117–136.

Beitinger, T.L., Bennett, W.A., McCauley, R.W., 2000. Temperature tolerances of NorthAmerican freshwater fishes exposed to dynamic changes in temperature. Environ.Biol. Fish 58, 237–275.

Berga-Casafont, L., 2003. Rivers and reservoirs in 20th century Spain. Rev. Obras Públicas3438, 37–40.

Bernardo, J.M., Ilhéu,M., Matono, P., Costa, A.M., 2003. Interannual variation of fish assem-blage structure in a Mediterranean river: implications of streamflow on the domi-nance of native or exotic species. River Res. Appl. 19, 521–532.

Boix, D., García-Berthou, E., Gascón, S., Benejam, L., Tornés, E., Sala, J., Benito, J., Munné, A.,Solà, C., Sabater, S., 2010. Response of community structure to sustained drought inMediterranean rivers. J. Hydrol. 383, 135–146.

Bradley, B.A., 2009. Regional analysis of the impacts of climate change on cheatgrass inva-sion shows potential risk and opportunity. Glob. Chang. Biol. 15, 196–208.

Breiman, L., 2001. Random forests. Mach. Learn. 45, 5–32.Briand, L.C., Freimut, B., Vollei, F., 2004. Using multiple adaptive regression splines to sup-

port decision making in code inspections. J. Syst. Softw. 73, 205–217.Brown, T.G., Runciman, B., Pollard, S., Grant, A.D.A., 2009. Biological synopsis of largemouth

bass (Micropterus salmoides). Can. Manuscr. Rep. Fish. Aquat. Sci. 2884, 1–26.Buisson, L., Thuiller, W., Casajus, N., Lek, S., Grenouillet, G., 2010. Uncertainty in ensemble

forecasting of species distribution. Glob. Chang. Biol. 16, 1145–1157.Bunn, S.E., Arthington, A.H., 2002. Basic principles and ecological consequences of altered

flow regimes for aquatic biodiversity. Environ. Manag. 30, 492–507.Capinha, C., Anastácio, P., 2011. Assessing the environmental requirements of invaders

using ensembles of distribution models. Divers. Distrib. 17, 13–24.Carmona-Catot, G., Santos, A.F.G.N., Tedesco, P., García-Berthou, E., 2014. Quantifying sea-

sonality along a latitudinal gradient: from stream temperature to growth of invasivemosquitofish. Ecosphere 5, 134.

Carpenter, S.R., Kitchell, J.F., 1993. The Trophic Cascade in Lakes. Cambridge UniversityPress, Cambridge, England.

Carvalho, S.B., Brito, J.C., Crespo, E.J., Possingham, H.P., 2010. From climate change predic-tions to actions–conserving vulnerable animal groups in hotspots at a regional scale.Glob. Chang. Biol. 16, 3257–3270.

Clavero, M., Hermoso, V., 2011. Reservoirs promote the taxonomic homogenization of fishcommunities within river basins. Biodivers. Conserv. 20, 41–57.

Clavero, M., Hermoso, V., Aparicio, E., Godinho, F.N., 2013. Biodiversity in heavily modifiedwaterbodies: native and introduced fish in Iberian reservoirs. Freshw. Biol. 58 (6),1190–1201.

Comte, L., Grenouillet, G., 2013. Species distribution modelling and imperfect detection:comparing occupancy versus consensus methods. Divers. Distrib. 19, 996–1007.

Cooke, S., Philipp, D.P., 2009. Centrarchid Fishes: Diversity, Biology, and Conservation.John Wiley and Sons, West Sussex, UK.

Copp, G.H., Bianco, P.G., Bogutskaya, N.G., Erős, T., Falka, I., Ferreira, M.T., et al., 2005. To be,or not to be, a non-native freshwater fish? J. Appl. Ichthyol. 21, 242–262.

Cord, A., Rödder, D., 2011. Inclusion of habitat availability in species distribution modelsthrough multi-temporal remote-sensing data? Ecol. Appl. 21, 3285–3298.

Coutant, C.C., 1975. Responses of bass to natural and artificial temperature regimes. In:Stroud, R.H., Clepper, H. (Eds.), Black Bass Biology andManagement. Sport Fishing In-stitute, Washington, D.C., pp. 272–285.

Cutler, D.R., Edwards, T.C., C, T., Beard, K.H., Cutler, A., Hess, K.T., Gibson, J., Lawler, J.J.,2007. Random forests for classification in ecology. Ecology 88, 2783–2792.

Doadrio, I., Perea, S., Garzόn-Heydt, P., González, J.L., 2011. Ictiofauna continentalespañola. Bases para su seguimiento. Ministerio de Medio Ambiente, Madrid.

Domisch, S., Jaehnig, S.C., Haase, P., 2011. Climate-change winners and losers: streammacroinvertebrates of a submontane region in Central Europe. Freshw. Biol. 56,2009–2020.

Dormann, C.F., Elith, J., Bacher, S., Buchmann, C., Carl, G., Carré, G., García Marquéz, J.R.,Gruber, B., Lafourcade, B., Leitão, P.J., Münkemüller, T., McClean, C., Osborne, P.E.,

65M.-J. Bae et al. / Science of the Total Environment 639 (2018) 58–66

Reineking, B., Schröder, B., Skidmore, A.K., Zurell, D., Lautenbach, S., 2013. Collinear-ity: a review of methods to deal with it and a simulation study evaluating their per-formance. Ecography 36, 27–46.

Dudgeon, D., Arthington, A.H., Gessner, M.O., Kawabata, Z.I., Knowler, D.J., Lévêque, C.,Naiman, R.J., Prieur-Richard, A.-H., Soto, D., Stiassny, M.L.J., Sullivan, C.A., 2006. Fresh-water biodiversity: importance, threats, status and conservation challenges. Biol. Rev.81, 163–182.

Edwards Jr., T.C., Cutler, D.R., Zimmermann, N.E., Geiser, L., Moisen, G.G., 2006. Effect ofsample survey design on the accuracy of classification treemodels in species distribu-tion models. Ecol. Model. 199, 132–141.

Elith, J., Leathwick, J.R., 2009. Species distributionmodels: ecological explanation and pre-diction across space and time. Annu. Rev. Ecol. Evol. Syst. 40, 677–697.

Elith, J., Ferrier, S., Huettmann, F., Leathwick, J., 2005. The evaluation strip: a new and ro-bust method for plotting predicted responses from species distribution models. Ecol.Model. 186, 280–289.

Elith, J., Leathwick, J.R., Hastie, T., 2008. A working guide to boosted regression trees.J. Anim. Ecol. 77, 802–813.

Elvira, B., Almodóvar, A., 2001. Freshwater fish introductions in Spain: facts and figures atthe beginning of the 21st century. J. Fish Biol. 59, 323–331.

ESRI, 2009. ArcGIS 10.0. Environmental Systems Resource Institute, Redlands, CA, USA.Farnsworth, E.J., Ogurcak, D.E., 2006. Biogeography and decline of rare plants in New

England: historical evidence and contemporary monitoring. Ecol. Appl. 16,1327–1337.

Ferreira, T., Caiola, N., Casals, F., Oliveira, J.M., de Sostoa, A., 2007. Assessing perturbationof river fish communities in the Iberian Ecoregion. Fish. Manag. Ecol. 14, 519–530.

Ferrier, S., Watson, G., Pearce, J., Drielsma, M., 2002. Extended statistical approaches tomodelling spatial pattern in biodiversity in northeast New South Wales. I. Species-level modelling. Biodivers. Conserv. 11, 2275–2307.

Ficetola, G.F., Thuiller, W., Miaud, C., 2007. Prediction and validation of the potentialglobal distribution of a problematic alien invasive species—the American bullfrog. Di-vers. Distrib. 13, 476–485.

Filipe, A.F., Markovic, D., Pletterbauer, F., Tisseuil, C., Wever, A., Schmutz, S., Bonada, N.,Freyhof, J., 2013. Forecasting fish distribution along stream networks: brown trout(Salmo trutta) in Europe. Divers. Distrib. 19, 1059–1071.

Friedman, J.H., 1991. Multivariate adaptive regression splines. Ann. Stat. 19, 1–141.Fullerton, A.H., Garvey, J.E., Wright, R.A., Stein, R.A., 2000. Overwinter growth and survival

of largemouth bass: interactions among size, food, origin, and winter severity. Trans.Am. Fish. Soc. 129, 1–12.

Gaeta, J.W., Guarascio, M.J., Sass, G.G., Carpenter, S.R., 2011. Lakeshore residential devel-opment and growth of largemouth bass (Micropterus salmoides): a cross-lakes com-parison. Ecol. Freshw. Fish 20, 92–101.

García-Berthou, E., 2002. Ontogenetic diet shifts and interrupted piscivory in introducedlargemouth bass (Micropterus salmoides). Int. Rev. Hydrobiol. 87, 353–363.

García-Berthou, E., Alcaraz, C., Pou-Rovira, Q., Zamora, L., Coenders, G., Feo, C., 2005. Intro-duction pathways and establishment rates of invasive aquatic species in Europe. Can.J. Fish. Aquat. Sci. 62, 453–463.

Garvey, J.E., Wright, R.A., Stein, R.A., Ferry, K.H., 2000. Evaluating how local-and regional-scale processes interact to regulate growth of age-0 largemouth bass. Trans. Am. Fish.Soc. 129, 1044–1059.

Gasith, A., Resh, V.H., 1999. Streams inMediterranean climate regions – abiotic influencesand biotic responses to predictable seasonal events. Annu. Rev. Ecol. Syst. 30, 51–81.

Gido, K.B., Propst, D.L., Olden, J.D., Bestgen, K.R., 2013. Multidecadal responses of nativeand introduced fishes to natural and altered flow regimes in the American South-west. Can. J. Fish. Aquat. Sci. 70, 554–564.

Godinho, F.N., Ferreira, M.T., Portugal e Castro, M.I., 1998. Fish assemblage composition inrelation to environmental gradients in Portuguese reservoirs. Aquat. Living Resour.11, 325–334.

Godinho, F.N., Ferreira, M.T., Santos, J.M., 2000. Variation in fish community compositionalong an Iberian river basin from low to high discharge: relative contributions of en-vironmental and temporal variables. Ecol. Freshw. Fish 9, 22–29.

Graham, C.H., Elith, J., Hijmans, R.J., Guisan, A., Townsend Peterson, A., Loiselle, B.A., 2008.The influence of spatial errors in species occurrence data used in distribution models.J. Appl. Ecol. 45, 239–247.

Grenouillet, G., Buisson, L., Casajus, N., Lek, S., 2011. Ensemble modelling of speciesdistribution: the effects of geographical and environmental ranges. Ecography 34,9–17.

Guisan, A., Thuiller, W., 2005. Predicting species distribution: offering more than simplehabitat models. Ecol. Lett. 8, 993–1009.

Guisan, A., Broennimann, O., Engler, R., Vust, M., Yoccoz, N.G., Lehmann, A., Zimmermann,N.E., 2006. Using niche-based models to improve the sampling of rare species.Conserv. Biol. 20, 501–511.

Guisan, A., Tingley, R., Baumgartner, J.B., Naujokaitis-Lewis, I., Sutcliffe, P.R., Tulloch, A.I.T.,Regan, T.J., Brotons, L., McDonald-Madden, E., Mantyka-Pringle, C., Martin, T.G.,Rhodes, J.R., Maggini, R., Setterfield, S.A., Elith, J., Schwartz, M.W., Wintle, B.A.,Broennimann, O., Austin, M., Ferrier, S., Kearney, M.R., Possingham, H.P., Buckley,Y.M., 2013. Predicting species distributions for conservation decisions. Ecol. Lett. 16,1424–1435.

Hanson,M.A., Butler, M.G., 1994. Responses to food webmanipulation in a shallowwater-fowl lake. Hydrobiologia 279, 457–466.

Hastie, T.J., Tibshirani, R.J., 1990. Generalized additive models. Monographs on Statisticand Applied Probability. Chapman and Hall, London, UK.

Havel, J.E., Lee, C.E., Vander Zanden, J.M., 2005. Do reservoirs facilitate invasions into land-scapes? Bioscience 55, 518–525.

Havens, K., Fox, D., Gornak, S., Hanlon, C., 2005. Aquatic vegetation and largemouth basspopulation responses to water-level variations in Lake Okeechobee, Florida (USA).Hydrobiologia 539, 225–237.

Heikkinen, R.K., Luoto, M., Araújo, M.B., Virkkala, R., Thuiller, W., Sykes, M.T., 2006.Methods and uncertainties in bioclimatic envelope modelling under climate change.Prog. Phys. Geogr. 30, 751–777.

Hermoso, V., Blanco-Garrido, F., Prenda, J., 2008. Spatial distribution of exotic fish speciesin the Guadiana river basin, with two new records. Limnetica 27, 189–194.

Iguchi, K.I., Matsuura, K., McNyset, K.M., Peterson, A.T., Scachetti-Pereira, R., Powers, K.A.,Vieglais, D.A., Wiley, E.O., Yodo, T., 2004. Predicting invasions of North Americanbasses in Japan using native range data and a genetic algorithm. Trans. Am. Fish.Soc. 133, 845–854.

Kiernan, J.D., Moyle, P.B., Crain, P.K., 2012. Restoring native fish assemblages to a regu-lated California stream using the natural flow regime concept. Ecol. Appl. 22,1472–1482.

Kramer, R.H., Smith Jr., L.L., 1960. First year growth of the largemouth bass, Microperussalmoides, and some related ecological factors. Trans. Am. Fish. Soc. 89, 222–233.

Lek, S., Guégan, J.F., 1999. Artificial neural networks as a tool in ecological modelling, anintroduction. Ecol. Model. 120, 65–73.

Leunda, P.M., 2010. Impacts of non-native fishes on Iberian freshwater ichthyofauna: cur-rent knowledge and gaps. Aquat. Invasions 5, 239–262.

Lever, C., 1996. Naturalised Fishes of the World. Academic Press, San Diego, CA.Lomba, A., Pellissier, L., Randin, C., Vicente, J., Moreira, F., Honrado, J., Guisan, A., 2010.

Overcoming the rare species modelling paradox: a novel hierarchical framework ap-plied to an Iberian endemic plant. Biol. Conserv. 143, 2647–2657.

Lookingbill, T.R., Urban, D.L., 2003. Spatial estimation of air temperature differences forlandscape-scale studies in montane environments. Agric. For. Meteorol. 114 (3),141–151.

Maceina, M.J., Bettoli, P.W., 1998. Variation in largemouth bass recruitment in fourmainstream impoundments of the Tennessee River. N. Am. J. Fish Manag. 18,998–1003.

MacRace, P.S.D., Jackson, D.A., 2001. The influence of smallmouth bass (Micropterusdolomieu) predation and habitat complexity on the structure of littoral zone fish as-semblages. Can. J. Fish. Aquat. Sci. 58, 342–351.

Maezono, Y., Miyashita, T., 2003. Community-level impacts induced by introducedlargemouth bass and bluegill in farm ponds in Japan. Biol. Conserv. 109, 111–121.

Magalhães, M.F., Batalha, D.C., Collares-Pereira, M.J., 2002. Gradients in stream fish assem-blages across a Mediterranean landscape: contributions on environmental factorsand spatial structure. Freshw. Biol. 47, 1015–1031.

Marchetti, M.P., Moyle, P.B., 2001. Effects of flow regime on fish assemblages in a regu-lated California stream. Ecol. Appl. 11, 530–539.

Markovic, D., Freyhof, J., Wolter, C., 2012. Where are all the fish: potential of biogeograph-ical maps to project current and future distribution patterns of freshwater species.PLoS One 7, e40530.

Marmion, M., Parviainen, M., Luoto, M., Heikkinen, R.K., Thuiller, W., 2009. Evaluation ofconsensus methods in predictive species distribution modelling. Divers. Distrib. 15,59–69.

Marta, P., Bochechas, J., Collares-Pereira, M.J., 2001. Importance of recreational fisheries inthe Guadiana River Basin in Portugal. Fish. Manag. Ecol. 8 (4–5), 345–354.

McCullagh, P., Nelder, J.A., 1989. Generalized linear models. Monographs on statIstics andApplied Probability vol. 37. Chapman Hall, London.

Meller, et al., 2013. Ensemble distribution models in conservation prioritization: fromconsensus predictions to consensus reserve networks. Divers. Distrib. 20, 309–321.

Melo, F.G., Gomes, A.S. (Eds.), 1992. Large Dams in Portugal. Portuguese National Commit-tee on Large Dams, Lisbon, Portugal Available from:. http://www.worldcat.org/title/large-dams-in-portugal/oclc/246906036.

Mika, A.M.,Weiss, R.M., Olfert, O., Hallett, R.H., Newman, J.A., 2008.Will climate change bebeneficial or detrimental to the invasive swede midge in North America? Contrastingpredictions using climate projections from different general circulation models. Glob.Chang. Biol. 14, 1721–1733.

Mooney, H.A., Hobbs, R.J., 2000. Invasive Species in a Changing World. Island Press,Washington, DC.

Murphy, C.A., Grenouillet, G., García-Berthou, E., 2015. Natural abiotic factors more thananthropogenic perturbation shape invasibility by Eastern Mosquitofish (Gambusiaholbrooki). Freshwat. Sci. 34, 965–974.

Olden, J.D., Jackson, D.A., 2002. A comparison of statistical approaches for modelling fishspecies distributions. Freshw. Biol. 47, 1976–1995.

Olden, J.D., Naiman, R.J., 2010. Incorporating thermal regimes into environmental flowsassessments: modifying dam operations to restore freshwater ecosystem integrity.Freshw. Biol. 55 (1), 86–107.

Olden, J.D., Poff, N.L., Bestgen, K.R., 2006. Life-history strategies predict fish invasions andextirpations in the Colorado River Basin. Ecol. Monogr. 76, 25–40.

Olson, M.H., 1996. Ontogenetic niche shifts in largemouth bass: variability and conse-quences for first-year growth. Ecology 77, 179–190.

Page, L.M., Burr, B.M., 1991. A Field Guide to Freshwater Fishes of North America North ofMexico. Houghton Mifflin Company, Boston, MA.

Parkos, J.J., Wahl, D.H., 2002. Towards an understanding of recruitment mechanisms inlargemouth bass. Am. Fish. Soc. Symp. 31, 25–45.

Pearson, R.G., Thuiller, W., Araújo, M.B., Martinez-Meyer, E., Brotons, L., McClean, C., et al.,2006. Model-based uncertainty in species range prediction. J. Biogeogr. 33,1704–1711.

Peterson, A.T., Soberón, J., Pearson, R.G., Anderson, R.P., Nakamura,M., Martinez-Meyer, E.,et al., 2011. Ecological Niches and Geographical Distributions. Princeton UniversityPress, New Jersey.

Phillips, S.J., Dudík, M., Elith, J., Graham, C.H., Lehmann, A., Leathwick, J., Ferrier, S., 2009.Sample selection bias and presence-only distribution models: implications for back-ground and pseudo-absence data. Ecol. Appl. 19 (1), 181–197.

Poff, N.L., Hart, D.D., 2002. How dams vary and why it matters for the emerging science ofdam removal. Bioscience 52 (8), 659–668.

66 M.-J. Bae et al. / Science of the Total Environment 639 (2018) 58–66

Poff, N.L., Olden, J.D., Merritt, D.M., Pepin, D.M., 2007. Homogenization of regional riverdynamics by dams and global biodiversity implications. Proc. Natl. Acad. Sci. 104,5732–5737.

Post, D.M., Kitchell, J.F., Hodgson, J.R., 1998. Interactions among adult demography,spawning data, growth rate, predation, overwinter mortality, and the recruitmentof largemouth bass in a northern lake. Can. J. Fish. Aquat. Sci. 55, 2588–2600.

Prasad, A.M., Iverson, L.R., Liaw, A., 2006. Newer classification and regression treetechniques: bagging and random forests for ecological prediction. Ecosystems 9,181–199.

Prats, J., Val, R., Armengol, J., Dolz, J., 2010. Temporal variability in the thermal regime ofthe lower Ebro River (Spain) and alteration due to anthropogenic factors. J. Hydrol.387, 105–118.

Propst, D.L., Gido, K.B., 2004. Responses of native and nonnative fishes to natural flow re-gime mimicry in the San Juan River. Trans. Am. Fish. Soc. 133, 922–931.

R Core Team, 2013. R: A Language and Environment for Statistical Computing. R Founda-tion for Statistical Computing, Vienna, Austria http://www.R-project.org/.

Rahel, F.J., Olden, J.D., 2008. Assessing the effects of climate change on aquatic invasivespecies. Conserv. Biol. 22 (3), 521–533.

Ribeiro, F., Beldade, R., Dix, M., Bochechas, J., 2007. Carta Piscícola Nacional. Direcção Geraldos Recursos Florestais-Fluviatilis Available at:. http://www.cartapiscicola.org.

Ridgeway, G., 1999. The state of boosting. Comput. Sci. Stat. 31, 172–181.Rodríguez, J.P., Brotons, L., Bustamante, J., Seoane, J., 2007. The application of predictive

modelling of species distribution to biodiversity conservation. Divers. Distrib. 13,243–251.

Roura-Pascual, N., Brotons, L., Peterson, A.T., Thuiller, W., 2009. Consensual predictions ofpotential distributional areas for invasive species: a case study of Argentine ants inthe Iberian Peninsula. Biol. Invasions 11, 1017–1031.

Shea, K., Chesson, P., 2002. Community ecology theory as a framework for biological inva-sions. Trends Ecol. Evol. 17, 170–176.

Shoup, D.E., Wahl, D.H., 2009. The effects of turbidity on prey selection by piscivorouslargemouth bass. Trans. Am. Fish. Soc. 138, 1018–1027.

Simberloff, D., 2009. The role of propagule pressure in biological invasions. Annu. Rev.Ecol. Evol. Syst. 40, 81–102.

Simberloff, D., Martin, J.-L., Genovesi, P., Maris, V., Wardle, D.A., Aronson, J., et al., 2013.Impacts of biological invasions - what's what and the way forward. Trends Ecol.Evol. 28, 58–66.

Smolik, M.G., Dullinger, S., Essl, F., Kleinbauer, I., Leitner, M., Peterseil, J., et al., 2010. Inte-grating species distribution models and interacting particle systems to predict thespread of an invasive alien plant. J. Biogeogr. 37, 411–422.

Sowa, S.P., Rabeni, C.F., 1995. Regional evaluation of the relation of habitat to distributionand abundance of smallmouth bass and largemouth bass in Missouri streams. Trans.Am. Fish. Soc. 124, 240–251.

Suski, C.D., Killen, S.S., Kieffer, J.D., Tufts, B.L., 2006. The influence of environmental tem-perature and oxygen concentration on the recovery of largemouth bass from exer-cise: implications for live–release angling tournaments. J. Fish Biol. 68, 120–136.

Swets, K.A., 1988. Measuring the accuracy of diagnostic systems. Science 240, 1285–1293.Taylor, J.M., Seilheimer, T.S., Fisher, W.L., 2014. Downstream fish assemblage response to

river impoundment varies with degree of hydrologic alteration. Hydrobiologia 728,23–39.

Thuiller,W., 2003. BIOMOD–optimizing predictions of species distributions and projectingpotential future shifts under global change. Glob. Chang. Biol. 9, 1353–1362.

Thuiller, W., 2004. Patterns and uncertainties of species' range shifts under climatechange. Glob. Chang. Biol. 10, 2020–2027.

Thuiller, W., Araujo, M.B., Pearson, R.G., Whittaker, R.J., 2004a. Uncertainty in predictionsof extinction risk. Nature 427, 145–148.

Thuiller, W., Brotons, L., Araujo, M.B., Lavorel, S., 2004b. Effects of restricting environmen-tal range of data to project current and future species distributions. Ecography 27,165–172.

Thuiller, W., Broennimann, O., Hughes, G., Alkemade, J.R.M., Midgley, G.F., Corsi, F., 2006.Vulnerability of African mammals to anthropogenic climate change under conserva-tive land transformation assumptions. Glob. Chang. Biol. 12, 424–440.

Thuiller, W., Lafourcade, B., Engler, R., Araújo, M.B., 2009. BIOMOD–a platform for ensem-ble forecasting of species distributions. Ecography 32, 369–373.

Václavík, T., Meentemeyer, R.K., 2012. Equilibrium or not? Modelling potential distribu-tion of invasive species in different stages of invasion. Divers. Distrib. 18, 73–83.

Vander Zanden, M.J., Olden, J.D., Thorne, J.H., Mandrak, N.E., 2004. Predicting occurrencesand impacts of smallmouth bass introductions in north temperate lakes. Ecol. Appl.14, 132–148.

Vila-Gispert, A., Alcaraz, C., García-Berthou, E., 2005. Life-history traits of invasive fish insmall Mediterranean streams. Biol. Invasions 7, 107–116.

Vörösmarty, C.J., McIntyre, P.B., Gessner, M.O., Dudgeon, D., Prusevich, A., Green, P.,Glidden, S., Bunn, S.E., Sullivan, C.A., Reidy Liermann, C., Davies, P.M., 2010. Globalthreats to human water security and river biodiversity. Nature 467, 555–561.

Williamson, M., 1999. Invasions. Ecography 22, 5–12.Williamson,M.H., Fitter, A., 1996. The varying success of invaders. Ecology 77, 1661–1666.Woodford, D.J., Hui, C., Richardson, D.M., Weyl, O.L., 2013. Propagule pressure rives estab-

lishment of introduced freshwater fish: quantitative evidence from an irrigation net-work. Ecol. Appl. 23 (8), 1926–1937.

Table S1. Sources for the occurrence data of Largemouth Bass (Micropterus salmoides) in the Iberian Peninsula.

Type Source

Online databases GBIF (Global Biodiversity Information Facility) https://www.gbif.org/

Personal

communications

Ignacio Doadrio (Spanish National Research Council)

Teresa Ferreira (Technical University of Lisbon)

Filipe Ribeiro (Technical University of Lisbon)

Emili García-Berthou (University of Girona)

Published

papers/Reports

Doadrio, I. 2002. Atlas y libro rojo de los peceps continentales de España. Ministerio

de Medio Ambiente, Madrid, Spain.

Doadrio, I., S. Perea, P. Garzón-Heydt, and J. L. González. 2011. Ictiofauna

continental española.

Bases para su seguimiento. Ministerio de Medio Ambiente, Madrid, Spain.

IBICAT database. Generalitat de Catalunya. (Details at: http://aca-

web.gencat.cat/aca/documents/ca/directiva_marc/annex_metodologia_ibicat.pdf)

Table S2. Pearson correlation among 12 environmental factors. Upper numbers: P value and lower numbers: correlation coefficients.

Environmental

variables

AnMean

Temp AnTRan AnPrecip SolarRad Slope TopoIndex FlowAcc MaxResVol ResLocVol Population AgPcnt UrbanPcnt

AnMeanTemp

0.436 0.000 0.000 0.000 0.000 0.000 0.000 0.017 0.000 0.000 0.000

AnTRan -0.010

0.000 0.000 0.000 0.000 0.000 0.000 0.129 0.000 0.000 0.000

AnPrecip -0.315 -0.512

0.000 0.000 0.000 0.000 0.000 0.984 0.099 0.000 0.006

SolarRad 0.311 0.133 -0.325

0.000 0.000 0.017 0.003 0.975 0.000 0.000 0.000

Slope -0.546 -0.252 0.476 -0.508

0.000 0.000 0.000 0.795 0.000 0.000 0.000

TopoIndex 0.176 0.138 -0.298 0.133 -0.338

0.000 0.000 0.108 0.000 0.000 0.000

FlowAcc 0.147 0.051 -0.082 0.030 -0.089 0.295

0.000 0.000 0.001 0.000 0.000

MaxResVol 0.197 0.056 -0.055 0.037 -0.107 0.197 0.726

0.000 0.003 0.000 0.006

ResLocVol 0.030 0.019 0.000 0.000 0.003 0.021 0.067 0.111

0.165 0.010 0.083

Population 0.126 -0.189 0.021 0.053 -0.080 0.084 0.044 0.038 -0.018

0.000 0.000

AgPcnt 0.339 0.285 -0.469 0.253 -0.634 0.308 0.066 0.071 -0.033 -0.050

0.000

UrbanPcnt 0.172 -0.211 -0.035 0.095 -0.118 0.100 0.045 0.035 -0.022 0.725 -0.079

Figure S1. Predicted distribution of Largemouth Bass in the Iberian Peninsula based on species distribution models.

Figure S2. Largemouth bass response curves with environmental predictors for the six spatial distribution modelling techniques. Abbreviations of

environmental variables are given in Table 1.

0.0

0.5

1.0

0 20 40 60 80

0.0

0.5

1.0

0 2 4 6 8

0.0

0.5

1.0

1630 1780 1930 2080

0.0

0.5

1.0

0 4000 8000 12000

0.0

0.5

1.0

0 30000 60000 90000 120000

0.0

0.5

1.0

5.8 8.8 11.8 14.8

0.0

0.5

1.0

200 700 1200 1700 2200

0.0

0.5

1.0

0 5 10 15 20

0.0

0.5

1.0

0 20 40 60 80 100

A nM eanTem p (℃) A nR angeTem p (℃) A nP recip (m m ) S olarR ad (m m /day)

S lope (º) TopoIndex Flow A cc M axR esV ol(km 3)

R esLocV ol(km 3) P opulation (people/km 2) A gP cnt (% ) U rbanP cnt (% )

Occurrence probability

0.0

0.5

1.0

0 20 40 60 80

G LM G A M B R T A N N M A R S R F

0.0

0.5

1.0

0 1000 2000 3000 4000

0.0

0.5

1.0

0 2000 4000 6000 8000

0.0

0.5

1.0

0 6 12 18 24