Ecological Complexity 22 (2015) 59–64

Original Research Article

Modeling the spread of invasive nutrias (Myocastor coypus) over Iran

Azita Farashi a,*, Mitra Shariati Najafabadi b

a Department of Environment, Faculty of Natural Resource and Environment, Ferdowsi University of Mashhad, Iranb Faculty of Geo-Information Science and Earth Observation (ITC), University of Twente, Enschede, The Netherlands

A R T I C L E I N F O

Article history:

Received 8 December 2014

Received in revised form 7 February 2015

Accepted 15 February 2015

Available online

Keywords:

Biological invasion

Species distribution modeling

Environmental variables

A B S T R A C T

Nutria (Myocastor coypus) is a native aquatic rodent to South America, and was introduced to Europe,

Asia, Africa and North America for fur farming. The South American nutria or coypu is now considered a

pest in the area of introduction, because of its negative impact on biological diversity and ecological

relationships. Having information on the invasion range of exotic species is crucial for understanding the

ecology of invasive spread and for making good conservation and management planning to address this

problem. At the beginning of the 20th century, nutria was introduced into Asia. Nutria was recorded for

the first time in Iran in 1995. In the present study we proposed a multiple spatial scale approach to

predict the invasion trends of the nutria in Iran, and to define up the ‘‘suitable scale’’ for predicting the

invasion trends of this species. Our results highlighted the importance of environmental variables

including vegetation density (for food and nesting) and water resource (streams, rivers, and lakes) in

distribution of the nutria. Potential areas for the presence of the nutria are located near the Caspian Sea,

west and central Iran which receive more precipitation than other parts of the country. Therefore, these

parts of Iran may face a much greater risk of invasion risk in the future. Moreover, these results can show

the possible risk of nutria invasion to the northern and western neighbors of Iran.

� 2015 Elsevier B.V. All rights reserved.

Contents lists available at ScienceDirect

Ecological Complexity

jo ur n al ho mep ag e: www .e lsev ier . c om / lo cate /ec o co m

1. Introduction

Species distribution modeling (SDM) is one of the mostcurrently tools available as a mean to assess species-environmentrelationship and defining the potentially suitable areas for aspecies (Elith and Leathwick, 2009). Researchers have in manycases relied on models for predicting and assessing patterns ofspecies distribution from environmental data (Guisan and Thuiller,2005; Yang et al., 2013). Although species distribution data areoften sparse, the species distribution models (SDMs) provide one ofthe best ways to overcome this problem, and have been widelyapplied to determine the relationships between species and theirenvironments (Guisan and Thuiller, 2005; Robertson et al., 2004).The representation of ecological processes is usually affected byscale and resolution. For instance, the choosing of wrong scale and/or resolution in the modeling process may cause an incorrectprediction of habitat pattern to reproduce ecological processes(Tamis and Van’t Zelfde, 1998; Naugle et al., 1999; Meyer, 2007;Storch et al., 2007; Convertino et al., 2009; McGill, 2010). It ischallenging to deal with biodiversity patterns at different scales

* Corresponding author. Tel.: +985138805460.

E-mail address: farashi@um.ac.ir (A. Farashi).

http://dx.doi.org/10.1016/j.ecocom.2015.02.003

1476-945X/� 2015 Elsevier B.V. All rights reserved.

and resolutions especially when it is combined with detecting anycritical bio-ecological scales (Levin, 1992; Rahbek, 2005; Storchet al., 2007; McGill, 2010) or a set of commanding processes(Graves and Rahbek, 2005). Guisan et al. (2007) found out that thehabitat range predicted by SDMs is unaffected by resolution, thusthey confirmed that species distribution can be correctly predictedat any resolution. However, other studies found that there is anupper threshold of resolution which above it the performance ofSDM decreases with respect to the precision of prediction(Hurlbert and Jetz, 2007; Seo et al., 2009; Yamakita and Nakaoka,2009; Pineda and Lobo, 2012).

Nutria (Myocastor coypus Molina, 1782) is a native aquaticrodent to South America, and was introduced to Europe, Asia,Africa and North America for fur farming (Carter and Leonard,2002; Bertolino and Genovesi, 2007). However, the rodentsrepeatedly escaped or were released from the farms to the wildand have since become established throughout the river banks andin wetlands. The South American nutria, or coypu is nowconsidered a pest in the area of introduction, because of itsnegative impact on biological diversity, ecological relationships,crop and irrigation systems (Linscombe et al., 1981; Shaffer et al.,1992; Llewellyn and Shaffer, 1993; Kaplan et al., 1998; Carter et al.,1999; Cabral et al., 2004; Randall and Foote, 2005). Because ofthese reasons, coypu is listed among the 100 World’s Worst

Table 1Habitat variables considered for the distribution models.

Environmental variables Code

Topography variables

Altitude T-A

Slope T-S

Climatic variables

Annual mean temperature (8C) C-AMT

Mean temperature of coldest quarter (8C) C-MTCQ

Annual precipitation (mm) C-AP

Land use/land cover variables

Distance of settlements in urban area L-DSU

Distance of settlements in rural area L-DSR

Human population density in urban area L-HDU

Human population density in rural area L-HDR

Distance of road L-DR

Distance of stream L-DS

Distance of river L-DRI

Distance of lake L-DL

Distance of dry farm L-DDF

Distance of irrigated farm L-DIF

Distance of forest L-DF

Distance of woodland L-DW

Distance of scrubland L-DSL

Distance of range L-DRAN

Distance of bare L-DB

Distance of rocky area L-DRA

Distance of protected area L-DPA

NDVI NDVI

A. Farashi, M.S. Najafabadi / Ecological Complexity 22 (2015) 59–6460

Invasive Alien Species (Bertolino, 2009). The presence of invasivespecies cause many ecological disruptions to natural ecosystemsand sometimes economic consequences, with establishment ofexotic food chains (Amori and Battisti, 2008). Having informa-tion on the invasion range of exotic species is crucial forunderstanding the ecology of invasive spread and for makinggood conservation and management planning to address thisproblem. At the beginning of the 20th century, nutria wasintroduced to Middle East (Carter and Leonard, 2002), and for thefirst time in Iran was recorded near the Iran-Azerbaijan border in1995. The Iran Neighboring countries such as Russia, Armenia,Azerbaijan and Turkmenistan are now the host of this species.However, the situation of nutria remains unknown in many partsof Asian, especially in Middle Eastern countries. This speciesexperienced a range expansion during the last 100 years, whichwas mostly due to human introduction. It was assumed that thenutria entered Iran by way of this border. Considering the centrallocation of Iran in Eurasia and western Asia, the nutriapopulation in Iran may be considered as kind of sourcepopulation to the other countries. This highlights the importanceof the study about distribution, invasion trend, habitat selection,and effects of nutria in Iran. However, the current knowledgeabout this species still remains extremely poor, although it hasbeen a long time since the nutria has recorded in Iran. Thereforein the present study we proposed a multiple spatial scaleapproach to predict the invasion trends of the nutria in Iran, andto define up the ‘‘suitable scale’’ for predicting the invasiontrends of this species. The aims of our study were: (1) todetermine the relationship between landscape composition withthe species ecological requirements, (2) to develop a robuststatistical framework for prediction of nutria distribution in Iran,and (3) to compare the model performances at different spatialscales.

2. Method

We used Genetic Algorithm for Rule-set Production (GARP) andMaximum Entropy (Maxent) to predict range expansion ofinvasive nutrias in Iran and determine species ecological require-ments at three extents (30, 100 and 1000 m).

2.1. Environmental parameters

Environmental parameters that we used included land cover/land use characteristics, climatic variables, topography variables(Table 1) which were extracted on three different spatial extents/resolutions (30/100/1000 m). Land cover/land cover data wereobtained from the Iranian Forests, Range and Watershed Manage-ment Organization (IFRWO) and Iran Department of Environment.The data was derived from 30 m Landsat Enhanced ThematicMapper Plus (ETM+) imagery for the conterminous Iran in the year2010 (7% forest, 4.7% woodland, 6.3% Irrigated farm, 9.1% Dry farm,42.3% Range, 5% Scrubland, 4.2% Rocky land, 18.9% Bare land, 2.5%Lake) (Fig. 1s). Human densities were interpolate from the dataderived from the Statistical Center of Iran that this data wascollected in the year 2011. Topography variables were obtainedfrom a Digital Elevation Model (DEM) generated by the NationalCartographic Center of Iran (NCC) at 1:25,000 scale. NormalizedDifference Vegetation Index (NDVI) was extracted from 30 mLandsat TM imagery for the conterminous Iran in the year2011. Climatic variables were derived from the temperature andprecipitation datasets of the Iran Meteorological Organizationfrom 1970 to 2011. Here too, the highly correlated variables wereeliminated using a cluster analysis and the initial set of 19 climaticvariables was reduced to 3 (Fig. 2s). ArcGIS 9.3 spatial analyst toolwas used to resample the data layers to different spatial extents

(30/100/1000 m). The multicollinearity test was conducted byusing Pearson correlation coefficient (r) to examine the cross-correlation and the variables with cross-correlation coefficientvalue of >�0.8 were excluded (see Table 1).

2.2. Species sampling

Presence points in Iran were collected from regional inventoriescovering the period 1995–2013. All the locations where the nutriahad been reported were examined in the field by searching for theanimal or signs of its presence, trapping, and taking pictures usingcamera traps.

2.3. Analytical/statistical procedure

GARP (Desktop GARP v1.1.6) is a non-parametric, machine-learning modeling method, which develops rule sets and predictsspecies potential distribution based on a genetic algorithm(Stockwell and Noble, 1991; Stockwell and Peters, 1999). GARPmodels are built by an iterative process of rule selection,evaluation, testing, and incorporation or rejection of the rulesproduced (Peterson and Cohoon, 1999). In the first phase, GARPselects a random population, based on a combination of initialprediction rules, which might represent suitable solutions for theproblem. The fitness to the characteristics of the population is thenevaluated for each pixel in the search space. If the performance ofthe rule is adequate as determined by the rule’s significancemeasure, the rule is retained for further runs of the algorithm, untilan end condition, consisting of a convergence limit and maximumnumber iterations, is satisfied (Stockwell and Peters, 1999).Maxent (Maxent v3.3.3e) is a general-purpose machine-learningmethod based on maximum entropy theory for species distribu-tion modeling. Maxent estimates niches by finding the distributionof probabilities closest to uniform (maximum entropy), con-strained to the fact that feature values match their empiricalaverage (Phillips et al., 2006). Maxent can evaluate the importanceof environmental variables by using Jackknife tests (Elith et al.,2011). In order to assess the average behavior of the models,

Table 2Habitat predictions’ evaluation for nutria in Iran.

Extent Model TSS Sensitivity (%) Specificity (%)

30 m GARP 0.60 64.1 61.8

Maxent 0.76 77.3 78.0

Combine 0.78 79.6 78.9

100 m GARP 0.62 64.3 61.9

Maxent 0.79 78.7 79.0

Combine 0.87 82.4 82.3

1000 m GARP 0.63 62.4 61.6

Maxent 0.78 79.5 77.0

Combine 0.79 80.1 79.3

The highest values of TSS, sensitivity and specificity were highlighted in bold.

A. Farashi, M.S. Najafabadi / Ecological Complexity 22 (2015) 59–64 61

10 random partitions were used in the three models (Phillips et al.,2006; Tittensor et al., 2009). Each partition was generated byrandomly choosing 75% species occurrence (logged from the divevideos) as calibration data, and the remaining 25% speciesoccurrence as evaluation data. Model evaluation was based ondifferent criteria: (1) the true skill statistic (TSS) which corre-sponds to the sum of sensitivity and specificity minus 1, and isindependent to prevalence (Lobo et al., 2008), (2) the sensitivity(‘true positives’) and (3) specificity (‘true negatives’) (Thuiller et al.,2009; Barbet-Massin et al., 2012). Finally, all two modelingtechniques were combined in an ensemble-forecasting frameworkas recommended by Araujo and New (2007). The ensemble wasbuilt out of all modeling techniques, giving higher importance tomodels with a better performance according to the TSS criterion(Thuiller et al., 2009).

3. Results

3.1. Model evaluation

After doing correlation analyses, we recognized twenty-threeenvironmental parameters as independent variables which werechosen for modeling (see Table 1). In this study, all the invasionpredictions of nutria by GARP and Maxent performed significantlybetter than the random performance (Table 2). The modelingevaluation results based on the TSS, sensitivity and specificityvalues showed that the combination of models at three extentsperformed averagely better than each individual model (Table 2).Moreover, the modeling evaluation indicated that Maxentperformed better than GARP at three extents, and 100 m extentwork better than other extents in all of the examined models.

3.2. Importance of environmental variables

The relative importance of different environmental variableswas examined based on the results of jackknife tests in Maxent(Fig. 1). These results showed that the importance of environmen-tal variables changed in different extents. The relative importance

Fig. 1. Importance of environmental variables for Maxent models ((a) extent 30 m, (b)

models containing each variable in isolation: gray bars, models with all variables: red ba

referred to the web version of this article.)

of distance of river in extent 30 and 1000 m and the relativeimportance of distance of lake in extent 100 m were the mostimportant predictors of nutria distribution in Iran. However, ingeneral, NDVI and distance of stream, river, and lake have the first,and mean temperature of coldest quarter, annual precipitation,human population density in urban area, distance of road, distanceof settlements and urban area, bare and rocky area have the secondrelative importance for nutria distribution.

3.3. Prediction of the potential invasion range of nutria

A large part of Iran was predicted as suitable habitat for nutriausing all models at three extents. In other words, all three modelspredicted almost same area as suitable habitat for nutria.According to our obtained results, the combination of modelsgenerally performed better in predicting the potential distributionof the invasion and indicated higher levels of predictive power andaccuracy in all three extents based on the modeling evaluationresults (Table 2). The results showed that north, north-west, andwest Iran had a higher probability of invasion range at all extents.The 80% of predicted area of occupancy of nutria within threedifferent scales and three different models had overlap (Fig. 2). Theonly difference in selection of suitable habitats at multiple scales

extent 100 m, (c) extent 1000 m) (models with each variable omitted: black bars,

rs). (For interpretation of the references to color in this figure legend, the reader is

Fig. 2. Estimated probability of occupancy of nutria in Iran. (a) GARP (extent 30 m), (b) GARP (extent 100 m), (c) GARP (extent 1000 m), (d) Maxent (extent 30 m), (e) Maxent

(extent 100 m), (f) Maxent (extent 1000 m), (g) combine model (extent 30 m), (h) combine model (extent 100 m), (i) combine model (extent 1000 m).

A. Farashi, M.S. Najafabadi / Ecological Complexity 22 (2015) 59–6462

was related to the areas along the Persian Gulf which were selectedat 30 and 100 m, but not at 1000 m. Furthermore, the predictionmap showed that 36.8%, 35.7% and 31.9% of Iran can be consideredas a suitable habitat for the nutria at 1000 m, 100 m, and 30 mextent, respectively (Fig. 2).

4. Discussion

The effects of scale on species distribution modeling got moreattention in the recent years (Pineda and Lobo, 2012). Ferrier and

Watson (1997) found that the model performance fitted onpresence-only data was not significantly affected by coarsening theresolution from 0.2 km to 5 km grid cells. However, Seo et al.(2009) found that the accuracy of model and spatial outputagreement declined with increasing the grid size. They alsoshowed that the correct selection of extent size would improve theaccuracy of the prediction. In the other study, by Song et al. (2013),the accuracy of the prediction model improved by selection ofproper extent size. There are no general rules about theappropriate scale for the species distribution modeling, because

A. Farashi, M.S. Najafabadi / Ecological Complexity 22 (2015) 59–64 63

it depends on both the ecology of the species under study and theobjectives of the study (Boyce et al., 2003). Due to the managementgoals that we followed, and the lack of knowledge about theecology of nutria in Iran, we decided to carry out our study in threeextents. Based on Johnson (1980), four different order selectionsexist: first-order selection that includes the overall geographicrange of species, second-order selection that only consider specificareas within a study sight, third order-selection that uses differentareas within the individual’s home range, and fourth-orderselection that uses vegetative or terrain characteristics withinthe areas of use. However, the focus of vast majority of animalhabitat selection studies is on second, third and fourth-orderselection (e.g., Moore et al., 2002; Gosselink et al., 2003; Lyonset al., 2003; McCorquodale, 2003; Franco et al., 2004; Nikula et al.,2004; Nolfo-Clements, 2012). In a case that an animal has atemporally variable home range or habitat use is dependent onsmall-scale phenomena such as nesting or feeding sites, habitatselection has a higher accuracy (Nolfo-Clements, 2012) first-orderselection, and in the present study, we used our results indicatedthat the models created at 100 m scale had higher accuracy whencompared with models created at 30 and 1000 m scales. Since first-order selection was used in the present study, therefore we cannotconclude that lower scale has a higher accuracy. One reason tochoose 100 m as the most accurate scale for the habitat selection ofnutria in this study, may be related to appropriate distribution ofvital environmental variables for this species (e.g. river, lake andvegetation along them) in this scale.

Our results highlighted the importance of environmentalvariables including vegetation density (for food and nesting)and water resource (streams, rivers, and lakes) in distribution ofnutria which is consistent with the results of previous studies(Laurie, 1946; Baroch and Hafner, 2002; Hong et al., 2014).Although it was shown that nutria locally distributed to urbanareas (Meyer et al., 2005), the results of our study as well as thefindings of other researchers indicated that human disturbance isone of the main factors affecting the nutria distribution within itsnatural range (Guichon and Cassini, 2005; Bertolino and Ingegno,2009).

Potential areas for the presence of nutria are located near theCaspian Sea, west and central Iran which receives more precipita-tion than other parts of the country. Therefore, these parts of Iranmay face a much greater risk of invasion risk in the future.Moreover, these results can show the possible risk of nutriainvasion to the western neighbors of Iran (Fig. 2). Our resultsshowed a continuous distribution of nutria in the north Iran thatmight be due to the density of rivers and streams with suitablevegetation. This might be because of nutria feeding which is mostlyon crops along rivers, weaken river banks, and dikes. However, thevegetation cover nearby rivers in south Iran are so weak because ofthe hot and dry climate of this region. It also was found by otherresearchers that climatic factors may limit the nutria dispersal(Wilson et al., 1966; Aliev, 1973; Gosling et al., 1980; Doncasterand Micol, 1990; Reggiani et al., 1995; Carter and Leonard, 2002;Hong et al., 2014).

The keeping of nutria as pets by local people in north Iran cancause concerns with regard to dispersion speed of this species.Therefore, some control methods may be required to prevent thedissemination of this species. Based on current knowledge, the bestmethods of control can be supporting farmers and NGOs for legalhunting of nutria, doing more research about population controlmethod, increasing the knowledge of local communities aboutinvasion risk and monitoring the population trends of nutria inIran.

This study is a first step toward a higher understanding of theinteractions between nutria and environmental variables in Iran.The results of our study provide the opportunity to develop

ecological management tools for biodiversity conservation andfight against introduced species. Such tries need to includemanagers, conservationists, and local communities on a largespatial scale to properly respond the widespread invasion process.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.ecocom.2015.02.003.

References

Aliev, F., 1973. Cases of mass mortality of nutria in the wetlands of Azerbaidzhan inwinter 1971–1972. Mammalia 36, 539–540.

Amori, G., Battisti, C., 2008. An invaded wet ecosystem in central Italy: an arrange-ment and evidence for an alien food chain. Rend. Lincei 19, 161–171.

Araujo, M.B., New, M., 2007. Ensemble forecasting of species distributions. TrendsEcol. Evol. 22, 42–47.

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

Baroch, J., Hafner, M., 2002. Biology and Natural History of the Nutria, with SpecialReference to Nutria in Louisiana. Nutria (Myocastor coypus) in Louisiana.Genesis Laboratories Inc., Wellington.

Bertolino, S., 2009. Species account of the 100 of the most invasive alien species inEurope: Myocastor coypus (Molina), coypu, nutria (Myocastoridae, Mammalia).DAISIE Handbook of Alien Species in Europe. Invading Nature-Springer Series inInvasion Ecology, vol. 3. Springer, Dordrecht, The Netherlands.

Bertolino, S., Genovesi, P., 2007. Semiaquatic mammals introduced into Italy: casestudies in biological invasion. In: Gherardi, F. (Ed.), Biological Invaders in InlandWater. Profiles, Distribution and Threats. Springer, Netherlands, pp. 175–191.

Bertolino, S., Ingegno, B., 2009. Modelling the distribution of an introduced species:the coypu Myocastor coypus (Mammalia, Rodentia) in Piedmont region, NW,Italy, Ital. J. Zool. 76 (3), 340–346.

Boyce, M.S., Mao, J.S., Merrill, E.H., Fortin, D., Turner, M.G., Fryxell, J., Turchin, P.,2003. Scale and heterogeneity in habitat selection by elk in YellowstoneNational Park. Ecoscience 10, 321–332.

Cabral, J.A., Miero, C.L., Marques, J.C., 2004. Environmental and biological factorinfluence the relationship between a predator fish, Gambusia holbrooki, and itsmain prey in rice fields of the Lower Mondego River Valley (Portugal). Hydro-biologia 382, 41–51.

Carter, J., Foote, A.L., Johnson-Randall, L.A., 1999. Modelling the effects of coypu(Myocastor coypus) on wetland loss. Wetlands 19, 209–219.

Carter, J., Leonard, B.P., 2002. A review of the literature on the worldwide distribu-tion, spread of, and efforts to eradicate the coypu (Myocastor coypus). Wildl. Soc.Bull. 30 (1), 162–175.

Convertino, M., Muneepeerakul, R., Azaele, A., Bertuzzo, E., Rinaldo, A., Rodriguez-Iturbe, I., 2009. On neutral metacommunity patterns of river basins at differentscales of aggregation. Water Resour. Res. 45, W08424.

Doncaster, C.P., Micol, T., 1990. Response by coypus to catastrophic events of coldand flooding. Ecography 13 (2), 98–104.

Elith, J., Leathwick, J.R., 2009. Species distribution models: ecological explanationand prediction across space and time. Annu. Rev. Ecol. Evol. Syst. 40, 677–697.

Elith, J., Phillips, S.J., Hastie, T., Dudık, M., Chee, Y.E., Yates, C.J., 2011. A statisticalexplanation of MaxEnt for ecologists. Divers. Distrib. 17, 43–57.

Ferrier, S., Watson, G., 1997. An Evaluation of the Effectiveness of EnvironmentalSurrogates and Modelling Techniques in Predicting the Distribution of Biologi-cal Diversity. Environment Australia, Canberra, Australia.

Franco, A.M.A., Catry, I., Sutherland, W.J., Palmeirim, J.M., 2004. Do different habitatsurvey methods produce the same conservation recommendations for LesserKestrels? Anim. Conserv. 7, 291–300.

Gosling, L., Guyon, G., Wright, K.M.H., 1980. Diurnal activity of feral coypus(Myocastor coypus) during the cold winter of 1978–9. J. Zool. 192 (2), 143–146.

Gosselink, T.E., Van Deelen, T.R., Warner, R.E., Joselyn, M.G., 2003. Temporal habitatpartitioning and spatial use of Coyotes and Red Foxes in east-central Illinois.Wildl. Manage. 67, 90–103.

Graves, G.R., Rahbek, C., 2005. Source pool geometry and the assembly of continen-tal avifaunas. Proc. Natl. Acad. Sci. 102 (May), 7871–7876.

Guichon, M.L., Cassini, M.H., 2005. Population parameters of indigenous popula-tions of Myocastor coypus: the effect of hunting pressure. Acta Theriol. 50,125–132.

Guisan, A., Graham, C.H., Elith, J., Huettmann, F., NCEAS Working Group, 2007.Sensitivity of predictive species distribution models to change in grain size.Divers. Distrib. 13, 332–340.

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

Hong, S., Do, Y., Kim, J.Y., Kim, D.K., Joo, G.J., 2014. Distribution, spread and habitatpreferences of nutria (Myocastor coypus) invading the lower Nakdong River,South Korea. Biol. Invasions 1–12.

A. Farashi, M.S. Najafabadi / Ecological Complexity 22 (2015) 59–6464

Hurlbert, A.H., Jetz, W., 2007. Species richness, hotspots, and the scale dependenceof range maps in ecology and conservation. PNAS 104, 13384–13389.

Johnson, D.A., 1980. The comparison of usage and availability measurements forevaluating resource preference. Ecology 61, 65–71.

Kaplan, D., Oron, T., Gutman, M., 1998. Development of macrophytic vegetation inthe Agmon Wetland of Israel by spontaneous colonization and reintroduction.Wetl. Ecol. Manage. 6, 143–150.

Laurie, E., 1946. The coypu (Myocastor coypus) in Great Britain. J. Anim. Ecol. 15 (1),22–34.

Levin, S.A., 1992. The problem of pattern and scale in ecology: the Robert H.MacArthur award lecture. Ecology 73 (December), 1943–1967.

Linscombe, G., Kinledr, N., Wright, U., 1981. Nutria population density andvegetative changes in brackish marsh in coastal Louisiana. In: Chapman, J.A.,Pursley, D. (Eds.), Proceeding of the Worldwide Furbearer Conference, vol. 1. pp.129–141.

Llewellyn, D.W., Shaffer, G.P., 1993. Marsh restoration in the presence ofintense herbivory. The role of Justicia lanceolata (Chapm.) small. Wetlands13, 176–184.

Lobo, J.M., Jimenez-valverde, A., Real, R., 2008. AUC: erratum: predicting speciesdistribution: offering more than simple habitat models. Glob. Ecol. Biogeogr. 17,145–151.

Lyons, A.L., Gaines, W.L., Servheen, C., 2003. Black bear resource selection in thenortheast Cascades, Washington. Biol. Conserv. 113, 55–62.

McCorquodale, S.M., 2003. Sex-specific movements and habitat use by Elk in theCascade Range of Washington. Wildl. Manage. 67, 729–741.

McGill, B.J., 2010. Matters of scale. Science 328, 575–576.Meyer, C.B., 2007. Does scale matter in predicting species distributions? Case study

with the marbled murrelet. Ecol. Appl. 17 (5), 1474–1483.Meyer, J., Klemann, N., Halle, S., 2005. Diurnal activity patterns of coypu in an urban

habitat. Acta Theriol. 50, 207–211.Moore, B.D., Coulson, G., Way, S., 2002. Habitat selection by adult female Eastern

Grey Kangaroos. Wildl. Res. 29, 439–445.Naugle, D.E., Higgins, K.F., Nusser, S.M., Johnson, W.C., 1999. Scale-dependent

habitat use in three species of prairie wetland birds. Landsc. Ecol. 14, 267–276.Nikula, A., Heikkinen, S., Helle, E., 2004. Habitat selection of adult Moose, Alces alces,

at two spatial scales in central Finland. Wildl. Biol. 10, 121–135.Nolfo-Clements, E.L., 2012. Habitat selection by nutria in a freshwater Louisiana

marsh. Southeast. Nat. 11 (2), 183–204.Peterson, A.T., Cohoon, K.P., 1999. Sensitivity of distributional prediction algorithms

to geographic data completeness. Ecol. Model. 117, 159–164.Phillips, S.J., Anderson, R.P., Schapire, R.E., 2006. Maximum entropy modeling of

species geographic distributions. Ecol. Model. 190, 231–259.

Pineda, E., Lobo, J.M., 2012. The performance of range maps and species distributionmodels representing the geographic variation of species richness at differentresolutions. Glob. Ecol. Biogeogr. 21, 935–944.

Rahbek, C., 2005. The role of spatial scale and the perception of large-scale speciesrichness patterns. Ecol. Lett. 8, 224–239.

Randall, L.A.J., Foote, A.L., 2005. Effects of managed impoundments and herbivoryon wetland plant production and stand structure. Wetlands 25, 38–50.

Reggiani, G., Boitani, L., Stefano, R., 1995. Population dynamics and regulation in thecoypu Myocastor coypus in central Italy. Ecography 18 (2), 138–146.

Robertson, M.P., Villet, M.H., Palmer, A.R., 2004. A fuzzy classification technique forpredicting species’ distributions: applications using invasive alien plants andindigenous insects. Divers. Distrib. 10, 461–474.

Seo, C., Thorne, J.H., Hannah, L., Thuiller, W., 2009. Scale effects in species distribu-tion models: implications for conservation planning under climate change. Biol.Lett. 5 (1), 39–43.

Shaffer, G.P., Sasser, C.E., Gosselink, J.G., Rejmanek, M., 1992. Vegetation dynamicsin the emerging Atchafalaya Delta, Louisiana, USA. J. Ecol. 80, 677–687.

Song, W., Kim, E., Lee, D., Lee, M., Jeon, S.W., 2013. The sensitivity of speciesdistribution modeling to scale differences. Ecol. Model. 248, 113–118.

Stockwell, D., Peters, D., 1999. The GARP modelling system: problems and solutionsto automated spatial prediction. Int. J. Geogr. Inf. Sci. 13, 143–158.

Stockwell, D.R.B., Noble, I.R., 1991. Induction of sets of rules from animal distribu-tion data: a robust and informative method of analysis. Math. Comput. Simul.33, 385–390.

Storch, D., Marquet, P., Brown, J., 2007. Scaling biodiversity. In: Storch, D., Marquet,P., Brown, J. (Eds.), Series: Ecological Reviews. Cambridge University Press,Cambridge.

Tamis, W.L.M., Van’t Zelfde, M., 1998. An expert habitat suitability model for thedisaggregation of bird survey data: bird counts in the Netherlands downscaledfrom atlas block to kilometre cell. Landsc. Urban Plan. 40 (4), 269–282.

Thuiller, W., Lafourcade, B., Araujo, M., 2009. Mod Operating Manual for BIOMOD.Universite Joseph Fourie, LaboratoireD’Ecologie Alpine.

Tittensor, D.P., Baco, A.R., Brewin, P.E., Clark, M.R., Consalvey, M., Hall-Spencer, J.,Rowden, A.A., Schlacher, T., Stocks, K.I., Rogers, A.D., 2009. Predicting globalhabitat suitability for stony corals on seamounts. Biogeography 36, 1111–1128.

Wilson, E., Newson, R., Aliev, F., 1966. Enemies and competitors of the nutria inUSSR. J. Mammal. 47 (2), 353–355.

Yang, X.-Q., Kushwaha, S.P., Saran, S., Xu, J., Roy, P.S., 2013. MaxEnt modeling forpredicting the potential distribution of medicinal plant, Justicia adhatoda L. inLesser Himalayan foothills. Ecol. Eng. 51, 83–87.

Yamakita, T., Nakaoka, M., 2009. Scale dependency in seagrass dynamics: how doesthe neighboring effect vary with grain of observation. Popul. Ecol. 51, 33–40.