Predicting potential impacts of climate change on ...

Our reference: ECOINF 530

P-authorquery-v11

AUTHOR QUERY FORM Journal: ECOINF

Please e-mail or fax your responses and any corrections to: Richard Louis, Allwyn E-mail: [email protected] Fax: +1 619 699 6721

Article Number: 530

Dear Author, Please check your proof carefully and mark all corrections at the appropriate place in the proof (e.g., by using on-screen annotation in the PDF file) or compile them in a separate list. Note: if you opt to annotate the file with software other than Adobe Reader then please also highlight the appropriate place in the PDF file. To ensure fast publication of your paper please return your corrections within 48 hours. For correction or revision of any artwork, please consult http://www.elsevier.com/artworkinstructions. We were unable to process your file(s) fully electronically and have proceeded by Scanning (parts of) your article

Rekeying (parts of) your article

Scanning the artwork

Any queries or remarks that have arisen during the processing of your manuscript are listed below and highlighted by flags in the proof. Click on the ‘Q’ link to go to the location in the proof.

Location in article

Query / Remark: click on the Q link to go Please insert your reply or correction at the corresponding line in the proof

Q1

Please specify the significance of footnote symbols “*” and “**” cited in Table 3, as corresponding footnote text have not been provided.

Q2

Please confirm that given names and surnames have been identified correctly.

Q3

The country name “Republic of Korea” has been inserted for the authors' affiliations. Please check, and correct if necessary.

Q4

Citation “Li et al., 2014” has not been found in the reference list. Please supply full details for this reference.

Q5

Citation “MA, 2005” has not been found in the reference list. Please supply full details for this reference.

Q6

Citation “Brook et al., 2008, 2010” has not been found in the reference list. Please supply full details for this reference.

Q7

The citation “Montgomery and Peck, 1982” has been changed to match the author name/date in the reference list. Please check here and in subsequent occurrences, and correct if necessary.

Q8

Citation “StatSoft, 2004” has not been found in the reference list. Please supply full details for this reference.

Q9

Citation “Li et al. (2014)” has not been found in the reference list. Please supply full details for this reference.

Q10

This sentence has been slightly modified for clarity. Please check that the meaning is still correct, and amend if necessary.

Page 1 of 2

Our reference: ECOINF 530

P-authorquery-v11

Q11

Uncited references: This section comprises references that occur in the reference list but not in the body of the text. Please position each reference in the text or, alternatively, delete it. Thank you.

Q12

Please provide the corresponding grant number for the grant sponsor “Ministry of Environment and the National Institute of Environmental Research (Korea)”. Please check this box if you have no corrections to make to the PDF file.

□

Thank you for your assistance.

Page 2 of 2

ECOINF-00530; No of Pages 1 Ecological Informatics xxx (2014) xxx

Contents lists available at ScienceDirect

Ecological Informatics journal homepage: www.elsevier.com/locate/ecolinf

2

Predicting potential impacts of climate change on freshwater fish in Korea

3

Yong-Su Kwon a, Mi-Jung Bae a, Soon-Jin Hwang b, Sang-Hun Kim c, Young-Seuk Park a,⁎

b

7

c

8 9 10 11 12

• • • •

Department of Biology and Department of Life and Nanopharmaceutical Science, Kyung Hee University, Seoul 130-701, Republic of Korea Department of Environmental Health Science, Konkuk University, Seoul 143-701, Republic of Korea Watershed Ecology Research Team, Water Environment Research Department, National Institute of Environmental Research, Incheon 404-170, Republic of Korea

R O

a

6

We projected the impacts of climate change on the distributions of fish species. The random forest model showed the best prediction power for the distributions of species. The model predicted decrease of endemic species richness and occurrence probabilities in the 2080s. The model predicted the extinctions of five species from their habitats in the 2080s.

U

N

C

O

R

R

E

C

T

E

D

5

P

4

Ecological Informatics xxx (2014) xxx – xxx

F

Highlights

O

1

http://dx.doi.org/10.1016/j.ecoinf.2014.10.002 1574-9541/© 2014 Published by Elsevier B.V.

Please cite this article as: Kwon, Y.-S., et al., Predicting potential impacts of climate change on freshwater fish in Korea, Ecological Informatics (2014), http://dx.doi.org/10.1016/j.ecoinf.2014.10.002

ECOINF-00530; No of Pages 10 Ecological Informatics xxx (2014) xxx–xxx

Contents lists available at ScienceDirect

Ecological Informatics journal homepage: www.elsevier.com/locate/ecolinf

Predicting potential impacts of climate change on freshwater fish in Korea

2Q2

Yong-Su Kwon a, Mi-Jung Bae a, Soon-Jin Hwang b, Sang-Hun Kim c, Young-Seuk Park a,⁎

F

1

3 4Q3 5

a

6

a r t i c l e

7 8 9 10 11

Article history: Received 1 May 2014 Received in revised form 30 July 2014 Accepted 11 October 2014 Available online xxxx

12 13 14 15 16 17 18

Keywords: Species distribution models Endemic fish Random forest Climate change Geographical distribution Stream

O

R O

i n f o

a b s t r a c t

D

P

Climate change is expected to have profound effects on the distribution and phenology of species and the productivity of aquatic ecosystem. In this study, we projected the impacts of climate change on the distributions of 22 endemic fish species in Korea with climatic and geographical variables by using species distribution models (SDMs). Six different SDMs – linear discriminant analysis, generalized linear model, classification and regression trees, random forest, support vector machine, and multivariate adaptive regression splines – were implemented for the prediction, and compared for their prediction capacity. The results showed that the random forest displayed the highest predictive power for the prediction of current species distributions. Therefore, the random forest was used to assess the potential impacts of climate change on the distributions of 22 endemic fish species. The results revealed that five species (Acheilognathus yamatsutae, Sarcocheilichthys variegatus wakiyae, Squalidus japonicus coreanus, Microphysogobio longidorsalis, and Liobagrus andersoni) have a high probability of becoming extinct in their respective habitable sub-watersheds by the 2080s due to climate change. The sensitivity analysis of the model showed that geo-hydrological variables such as stream order and altitude and temperature-related variables such as mean temperature in January and difference between the minimum and maximum temperatures exhibited relatively higher importance in their contributions for the prediction of species occurrence than that other variables. The decline of endemic fish species richness, and their occurrence probability due to climate change, would lead to poleward and upward shifts, as well as extinctions of species. Finally, we believe that our projections are useful for understanding how climate change affects the distribution range of endemic species in Korea, while also providing the necessary information to develop preservation and conservation strategies for maintaining endemic fish. © 2014 Published by Elsevier B.V.

E

C

T

c

Department of Biology and Department of Life and Nanopharmaceutical Science, Kyung Hee University, Seoul 130-701, Republic of Korea Department of Environmental Health Science, Konkuk University, Seoul 143-701, Republic of Korea Watershed Ecology Research Team, Water Environment Research Department, National Institute of Environmental Research, Incheon 404-170, Republic of Korea

E

b

R

38

47 48 49 50 51 52 Q4 53 54 55 56 57 58 59

The evaluation of climate change effects on the structure and function of the ecosystem is one of the most researched issues in modern ecology (Daufresne and Boët, 2007), and various studies have shown that the recent climate change has already affected species' geographical distributions and persistence of their populations (Heikkinen et al., 2006; Moore, 2003; Parmesan, 1996; Parmesan and Yohe, 2003; Walther et al., 2002). Furthermore, the projected climate change is likely to have an even greater impact on biota (Berry et al., 2002; Hill et al., 2003; Thomas et al., 2004; Thuiller et al., 2005; Li et al., 2013; Li et al., 2014). The most immediate effects of climate change are shifts in species' geographical range, prompted by shifts in the normal patterns of temperatures delimiting species boundaries (Thuiller, 2007). In particular, climate change can have a greater effect on freshwater biodiversity than either terrestrial or marine biodiversity, by increasing water temperatures and altering stream flow patterns (Jenkins, 2003; Poff et al., 2002; UNESCO, 2003). Revenga et al. (2005) showed that

N C O

45 46

1. Introduction

U

44

R

42 40 39 41 43

19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37

⁎ Corresponding author. Tel.: +82 2 961 0946; fax: +82 2 961 0244. E-mail address: [email protected] (Y.-S. Park).

the projected rate of species loss in freshwater ecosystems is estimated to be five times greater than that of terrestrial fauna in North America, and rates are likely to increase in the future. These changes are expected to have profound effects on the distributions and phenology of species and productivity of aquatic ecosystems (Parmesan, 2006). Changes of environmental factors can be stresses on ecosystems and biota. Environmental stresses have caused extensive transformation and deterioration of various environments (MA, 2005; Brook et al., 2008, 2010; Staudt et al., 2012). Climate change has serious impacts on ecosystems mentioned above and climate change effects are projected to be an increasingly important source of stresses in the future (Staudt et al., 2012). Therefore, ecological risk assessment of climate change is needed. Ecological risk assessment is the assessment of environmental effects of certain stresses and their immediate and long-term damage or harm to an ecosystem and it has been widely recommended for environmental decision making (Chen et al., 2011, 2013). Fish are an important component of aquatic ecosystems through functions, including their consumption of organisms at lower trophic levels and their regulatory effects on a variety of ecosystem-level properties (Carpenter et al., 1985; Power et al., 1985; Wootton and Power, 1993). Fish can also reflect habitat changes, environmental degradation,

http://dx.doi.org/10.1016/j.ecoinf.2014.10.002 1574-9541/© 2014 Published by Elsevier B.V.


60 61 62 63 64 65 66 67 Q5 Q6 68 69 70 71 72 73 74 75 76 77 78 79 80

126

131

The dataset consisting of presence/absence of fish species was obtained from the database of the National Aquatic Ecological Monitoring Program, which is operated by the Ministry of Environment and the National Institute of Environmental Research, Korea. According to the standardized sampling protocol of the program (MOE/NIER, 2008), the samples were collected at 960 sampling sites in five major rivers (Han,

t1:1 t1:2

Table 1 Characteristics of fish assemblage at different watersheds.

106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122

127 128 129 130

t1:3

C

104 105

E

102 103

R

100 101

Watershed

t1:4 t1:5 t1:6 t1:7 t1:8 t1:9 t1:10

Han River Nakdong River Geum River Youngsan River Seomjin River Total

Number of species

132 133 134 135 136 137 138 139 140 141 142 143 144

145

To develop the species distribution models and evaluate the potential effects of climate change on the distribution of fish species, we obtained climate data regarding a climate change scenario from the Korea Meteorological Administration (KMA, http://www.climate.go. kr). The representative concentration pathway (RCP) scenario was based on the 5th report of the Intergovernmental Panel on Climate Change (IPCC) and reflects the recent trends in the changing greenhouse gas concentrations. The RCP scenario is comprised of four different scenarios of greenhouse gas concentrations (RCP 2.6, 4.5, 6.0, and 8.5). Among these scenarios, we considered the RCP 8.5 scenario (i.e., 936 ppm atmospheric CO2 in 2100), in which greenhouse gases are emitted at the current level, to predict changes in the species distributions due to climate change. From the RCP 8.5 scenario data, we extracted 12 climatic variables related to temperature and precipitation, because temperature (Begon et al., 2006; Somero, 1997) and precipitation (Milly et al., 2005) are the key climatic factors influencing fish species distribution (Buisson et al., 2008b; Gillanders et al., 2011; Matthews, 1998; Pont et al., 2005; Wehrly et al., 2003). Among the 12 climatic variables, 8 variables, including average annual temperatures and precipitations for January, April, July, and October, were used to account for seasonality of three different decades: the 2000s, 2040s, and 2080s (Table 2). In addition, the differences between the maximum (July) and minimum (January) temperature and precipitation values were also used to account for annual variation. The 2000s were selected as the current climate situation, and the other two decades (2040s and 2080s) were selected to represent climate change. To reflect the physical characteristics of fish habitats and the limitations of fish distributions due to geo-hydrological characteristics, we also considered three geo-hydrological variables: altitude, slope, and stream order of each sampling site. Altitude was measured from a Digital Elevation Map (DEM), and slope and stream order were obtained from the Water Management Information System (WAMIS, http:// www.wamis.go.kr) of the Ministry of Land, Transport and Maritime Affairs, Korea. Both RCP scenario and geo-hydrological data were extracted from the ESRI-GRID format using ArcGIS (Version 9.3, ESRI, 2008). Finally, 15 environmental variables (12 climatic variables and three geo-hydrological variables) were used in the study. Environmental

R

98 99

2.2. Climatic and geo-hydrological data

O

96 97

C

94 95

N

92 93

U

90 91

F

2.1. Fish data

88 89

O

125

87

R O

2. Materials and methods

85 86

P

124

83 84

Nakdong, Geum, Yeongsan, and Seomjin rivers) and their tributaries and small streams across the entire stream system of the country from 2008 to 2012 (Table 1). In the dataset, a total of 145 fish species belonging to 29 families were recorded. Based on the occurrence frequency in the sampling sites, Zacco platypus was the most frequently observed species in Korea. Pseudogobio esocinus, Odontobutis platycephala, and Carassius auratus were the most abundant species in the Han River watershed, the Nakdong River, and Seomjin River watersheds, and the Geum River and Youngsan River watersheds, respectively. Endemic fish species that occurred in greater than 10% of the 953 sampling sites (excluding the 7 sites on Jeju Island) were retained to reduce errors by the rare species, resulting in a dataset with 22 endemic fish species. The presence–absence of species was used in this study.

T

123

and overall ecosystem health (Karr and Freemark, 1985; Kouamélan et al., 2003; McCauley, 1990). Distribution and abundance of fish are affected by various environmental factors, including geological, hydromorphological, physicochemical, climatic, and biological factors (Kwon et al., 2012). Among them, climatic ones are fundamental to fish distributions because most fish do not have the physiological ability to regulate their body temperatures (Wood and McDonald, 1997). Freshwater fish also cannot disperse across terrestrial areas because their dispersal abilities are constrained by the network structure of their drainage basins, and are consequently limited to the river basin in which they currently live, making them vulnerable to broad-scale environmental changes, such as climate change (Buisson et al., 2008). Thus, climate change is of considerable importance for determining future fish distributions. Species distribution models (SDMs), also known as habitat modeling and niche modeling, are useful tools for predicting the potential spatial distribution of a phenomenon because they relate sites of known occurrences to predictor variables (Hijmans and Elith, 2013). The common applications of this method assess the potential impacts of climate change and predict species distribution ranges based on climate data by using various methods that include the following: generalized additive models (GAM; Leathwick et al., 1996; Araújo et al., 2004; Luoto et al., 2005), generalized linear models (GLM; Araújo et al., 2004; Thuiller et al., 2005), multivariate adaptive regression splines (MARS; Prasad and Iverson, 2000), and classification and regression trees (CART; Iverson and Prasad, 1998; 2002). Each method has been shown to produce high levels of predictability in each of the aforementioned studies. However, they have different properties and have been applied under different conditions. Therefore, it is valuable to compare the performance of each of the different modeling methods with the same datasets. Meanwhile, Fukuda et al. (2013) demonstrated the applicability of several modeling methods to model the species–habitat relationship of spawning European graylings (Thymallus thymallus), and Buisson et al. (2010) have focused on applying a range of SDMs to a set of fish species occurring in French streams. However, they did not consider climatic factors in the models. In this study, we predicted occurrences of endemic fish species with climatic and geographical variables using six different modeling methods, and compared their prediction capacities. In addition, we intended to estimate the potential distribution range of each species under a climate change scenario by using the best-suited modeling method, and to evaluate the relative importance of environmental variables for the prediction of species distribution.

D

81 82

Y.-S. Kwon et al. / Ecological Informatics xxx (2014) xxx–xxx

E

2

Dominant species

Total

Exotic

Endemic

1st

2nd

110 103 92 76 75 145

6 4 5 3 3 6

35 37 32 22 26 54

Zacco platypus Zacco platypus Zacco platypus Zacco platypus Zacco platypus Zacco platypus

Pseudogobio esocinus Odontobutis platycephala Carassius auratus Carassius auratus Odontobutis platycephala Pseudogobio esocinus


146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182

Y.-S. Kwon et al. / Ecological Informatics xxx (2014) xxx–xxx Table 2 Description of variables used in the models. Range 1–7 0–89 10–823 10.1–19.7 −3.0–8.5 8.9–17.9 22.9–31.4 13.1–23.4 20.5–29.2 102–261 27–80 75–312 244–1069 26–107 213–1030

F

Variable and description Stream order Slope Altitude Average annual temperature during 10 years Mean temperature in January Mean temperature in April Mean temperature in July Mean temperature in October Difference between maximum temperature (TempJUL) and minimum temperature (TempJAN) Average annual precipitation during 10 years Mean precipitation in January Mean precipitation in April Mean precipitation in July Mean precipitation in October Difference between maximum temperature (PrcpJUL) and minimum temperature (PrcpJAN)

data were standardized using a z-transformation to remove collinearity (Montgomery et al., 2001; Rodríguez, 1995).

185

2.3. Species distribution models (SDMs)

186

To build the SDMs for the endemic fish species based on climatic and geo-hydrological variables, we used the following six different methods: linear discriminant analysis (LDA), generalized linear model (GLM), classification and regression tree (CART), random forest (RF), support vector machine (SVM), and multivariate adaptive regression splines (MARS). These SDMs have frequently been used to evaluate the potential impacts of climate change and predict species distribution ranges based on climate. LDA is based on the extraction of linear discriminant functions of the independent variables by means of a qualitative dependent variable and several quantitative independent variables (McLachlan, 2004; Pop and Sârbu, 2013). GLM uses mathematical extensions of linear models able to handle nonlinear relationships by fitting parametric terms (McCullagh and Nelder, 1989) and represent the most commonly used technique for species distribution modeling (Austin and Meyers, 1996; Brito et al., 1999). GLM is technically closely related to traditional practices used in linear modeling and analysis of variance (ANOVA) (Guisan et al., 2002). CART uses recursive partitioning to split the data into increasingly homogenous subsets, in terms of the dependent variable, to yield a binary decision tree, and the decision rules at the nodes use one or more of the independent variables (Breiman et al., 1984). In CART analysis, trees subdivide the space spanned by the predictor variables into regions for which the values of the response variable are approximately equal, and then estimate the response variable by a constant in each of these regions (Moisen and Frescino, 2002). The tree is called a classification tree if the response variable is qualitative and a regression tree if the response variable is quantitative (Moisen et al., 2006). RF is a non-parametric method for predicting and assessing the relationship between a large number of potential predictor variables and response variables (Breiman, 2001). It is a model-averaging approach that generates hundreds or thousands of random trees built from a set of randomly selected predictors and observations (Breiman, 2001; Buisson et al., 2010). After the trees have been built, data are entered into them and each grid square is classified by all trees. At the end of the run, the classification given by each tree is considered as a “vote”, and the classification of a grid square corresponds to the majority vote among all trees (Breiman, 2001). SVM is a learning system that uses a hypothesis space of linear functions in a high-dimensional space and is trained using a learning algorithm from optimization theory that implements learning bias derived from statistical learning theory (Cristianini and Shawe-Taylor, 2000). In its classical implementation, it uses two classes (e.g. presence/absence) of training samples within a multidimensional feature space to fit an optimal separating hyperplane (in each dimension, the vector component is image gray-level). In this way, SVM

198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228

C

197

E

195 196

R

193 194

R

191 192

N C O

189 190

U

187 188

T

184 Q7

O

183

Climate

Order Slope (%) Altitude (m) TempAVE (°C) TempJAN (°C) TempAPR (°C) TempJUL (°C) TempOCT (°C) TempDIF (°C) PrcpAVE (mm) PrcpJAN (mm) PrcpAPR (mm) PrcpJUL (mm) PrcpOCT (mm) PrcpDIF (mm)

tries to maximize the margin that is the distance between the closest training samples, or support vectors, and the hyperplane itself (Pouteau et al., 2012). MARS models use a nonparametric modeling approach that does not require assumptions about the forms of the relationships between the predictor and dependent variables (Friedman, 1991). Consequently, MARS models have the ability to characterize relationships between explanatory and response variables that are difficult, if not impossible, for other regression methods (e.g., linear models) to reveal (Balshi et al., 2009). All SDMs were run in the R statistical computing environment (http://cran.r-project.org) with R package: MASS (Ripley et al., 2014) for LDA, glm2 (Marschner, 2013) for GLM, rpart (Therneau et al., 2014) for CART, CORElearn (Robnik-Sikonja and Savicky, 2013) for RF, e1071 (Meyer et al., 2014) for SVM, and earth (Milborrow and Tibshirani, 2014) for MARS. The continuous values of model outputs were transformed to show the presence or absence of each species at each sampling site (1 representing presence, 0 for absence) based on a threshold (=0.5) that maximized the sum of two measures: sensitivity (i.e., the percentage of correctly predicted presence) and specificity (i.e., the percentage of correctly predicted absence) (Fielding and Bell, 1997) in the preliminary study. Three measures of model performance, accuracy, area under the curve (AUC), and Cohen's kappa, were used to evaluate the predictive ability of each SDM. The accuracy (i.e., the correct prediction rate) is based on binary predictions and measures the percentages of both correctly predicted presence and absence, allowing one to quantify the match between the predicted and observed distributions (Buisson et al., 2010). The AUC of a receiver operating characteristic plot (Fielding and Bell, 1997; Pearce and Ferrier, 2000) is used as a measure of a model's overall performance (Buisson et al., 2010). AUCs usually range from 0.5 (random) to 1.0 (perfect discrimination or accuracy) but can also be below this range, which is indicative of a model that is worse than random (Engler et al., 2004). The following approximate guide for classifying the accuracies of AUCs was proposed by Swets (1988): 0.90–1.00, excellent; 0.80–0.90, good; 0.70–0.80, fair; 0.60–0.70, poor; and 0.50–0.60, fail. Cohen's kappa (Cohen, 1960) ranges from 0 (completely random prediction) to 1 (perfect prediction). Landis and Koch (1977) proposed the following scale to describe the degree of concordance based on Cohen's kappa: 0.81–1.00, nearly perfect; 0.61–0.80, substantial; 0.41–0.60, moderate; 00.21–0.40, fair; and 0.00–0.20, fail. These three methods are widely used in ecology (Garzón et al., 2006), and were applied to select the best model from the six different SDMs employed in this study. The overall modeling procedures are summarized in Fig. 1. After the selection of the best suited modeling method to predict the current distributions of endemic species, the importance of environmental variables used in the prediction model was evaluated using the Minimum Description Length (MDL), which measures the quality of attributes as their ability to compress the data (Robnik-Sikonja,

R O

Geo-hydrology

P

Category acronym

t2:4 t2:5 t2:6 t2:7 t2:8 t2:9 t2:10 t2:11 t2:12 t2:13 t2:14 t2:15 t2:16 t2:17 t2:18

D

t2:3

E

t2:1 t2:2

3


229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276


richness displayed negative correlation with altitude (r = − 0.16, P b 0.01), slope (r = − 0.21, P b 0.01), PrcpAVE (r = − 0.12, P b 0.01), PrcpAPR (r = − 0.14, P b 0.01), and PrcpOCT (r = − 0.23, P b 0.01), whereas it was positively correlated with stream order (r = 0.54, P b 0.01), TempJUL (r = 0.18, P b 0.01), and TempDIF (r = 0.19, P b 0.01) (Table 3). The geo-hydrological variables were significantly correlated with all climatic variables with the exception of PrcpJAN, TempOCT, and TempDIF (P b 0.05). Altitude and slope were negatively correlated with most of the temperature-related variables and positively correlated with most of the precipitation-related variables (P b 0.05).

294 295

3.2. Species distribution models

305

F

4

P

R O

O

The six different SDMs in predicting current species distributions showed significantly different performances (K–W test, P b 0.05) (Fig. 2). Across the 22 species, RF displayed the best performance (median accuracy: 0.90 ± 0.06, median AUC: 0.97 ± 0.01, and median Cohen's kappa: 0.72 ± 0.15), whereas LDA exhibited the worst performance (median accuracy: 0.79 ± 0.07, median AUC: 0.81 ± 0.05, and median Cohen's kappa: 0.35 ± 0.18) according to Dunn's test (P b 0.05). Four other SDMs (CART, MARS, SVM and GLM) showed similar performance (median accuracy: 0.81–0.87 and median AUC: 0.83–0.88), though Cohen's kappa averaged from 0.59 to 0.42 (Dunn's test, P b 0.05). In the RF models, the performance measures were different depending on species, ranging from 0.65 to 0.94 in accuracy and from 0.27 to 0.83 in Cohen's kappa (Table 4). Among the 22 endemic species, O. platycephala exhibited a relatively low predictive power (accuracy: 0.65 and kappa: 0.27), whereas Acheilognathus koreensis and Microphysogobio longidorsalis displayed the greatest predictive performance in the accuracy (0.94) and Cohen's kappa (0.83), respectively. Zacco koreanus, which was the most frequently observed endemic species in Korea, displayed a relatively high prediction power (accuracy: 0.90 and kappa: 0.80). The sensitivity analyses were conducted to evaluate the contribution of variables for predicting the occurrence of each species using the MDL in the RF model, which proved to be the best performance model in this study. Geo-hydrological variables, such as stream order and altitude, and temperature-related variables, such as TempDIF and TempJAN, exhibited relatively higher importance in their contributions for the prediction of species occurrence than other variables, although each species showed different patterns in the importance of environmental variables (Table 4, Fig. 3). For instance, geo-hydrological variables, such as stream order, were the most important variables for the prediction of endemic fish species, such as Microphysogobio yaluensis and Squalidus gracilis majimae. Temperature-related variables, such as

Fig. 1. A flowchart of the modeling procedures.

277

290

3. Results

291

3.1. Relations between variables

292 293

To evaluate relationships between fish species richness and environmental variables, we applied Spearman's rank correlation. Total species

t3:1 t3:2 Q1

Table 3 Spearman's rank correlation coefficients between 15 environmental variables and species richness.

t3:3 t3:4 t3:5 t3:6 t3:7 t3:8 t3:9 t3:10 t3:11 t3:12 t3:13 t3:14 t3:15 t3:16 t3:17 t3:18 t3:19 t3:20

E

T

C

E

R

R O

286 287

C

284 285

Category

Variable

SR

N

282 283

U

280 281

D

288 289 Q8

2004). The values of MDL were rescaled to range from 0 to 100 to compare the relative importance of each environmental factor. The potential impacts of climate change on the distribution range of endemic species were assessed at each watershed, as well as subwatershed levels, by considering the number of sub-watersheds in which species are predicted to be present by the model. Spearman's rank correlation coefficients were calculated between environmental variables and species richness to evaluate their relationships. The Kruskal– Wallis test (K–W test) and Dunn's multiple comparisons tests were conducted to compare the average predictive performances (i.e. accuracy, AUC, and Cohen's kappa) of 22 endemic fish species in the six different SDMs. Statistical analyses were carried out with the statistical software STATISTICA (StatSoft, 2004).

278 279

GeoAltitude hydrology Slope Order Climate TempAVE TempJAN TempAPR TempJUL TempOCT TempDIF PrcpAVE PrcpJAN PrcpAPR PrcpJUL PrcpOCT PrcpDIF

−0.16⁎⁎ −0.21⁎⁎ 0.54⁎⁎ −0.01 −0.06 0.03 0.18⁎⁎

Geo-hydrology

Altitude

0.52⁎⁎ −0.26⁎⁎ −0.73⁎⁎ −0.60⁎⁎ −0.67⁎⁎ −0.74⁎⁎ −0.05 −0.76⁎⁎ 0.19⁎⁎ 0.29⁎⁎ −0.12⁎⁎ 0.30⁎⁎ 0.00 0.00 −0.14⁎⁎ 0.27⁎⁎ −0.01 0.27⁎⁎ −0.23⁎⁎ 0.13⁎⁎ 0.00 0.27⁎⁎

Slope

−0.25⁎⁎ −0.28⁎⁎ −0.17⁎⁎ −0.30⁎⁎ −0.41⁎⁎ −0.28⁎⁎

Climate Order

Temp AVE

Temp JAN

Temp APR

Temp JUL

Temp OCT

Temp DIF

Prcp AVE

Prcp JAN

Prcp APR

Prcp JUL

Prcp OCT

−0.68⁎⁎ −0.10⁎⁎ 0.22⁎⁎ 0.01 −0.34⁎⁎ 0.03 −0.35⁎⁎

−0.11⁎⁎ −0.60⁎⁎ −0.47⁎⁎ 0.48⁎⁎ −0.20⁎⁎ 0.52⁎⁎

0.27⁎⁎ 0.63⁎⁎ 0.46⁎⁎ 0.68⁎⁎ −0.25⁎⁎ 0.12⁎⁎ 0.69⁎⁎ 0.29⁎⁎ 0.44⁎⁎ 0.38⁎⁎ 0.63⁎⁎ −0.33⁎⁎ 0.08 0.99⁎⁎ 0.34⁎⁎

0.10⁎⁎ 0.07⁎ 0.16⁎⁎ 0.24⁎⁎

0.94⁎⁎ 0.91⁎⁎ 0.80⁎⁎ 0.78⁎⁎ 0.60⁎⁎ 0.82⁎⁎ 0.04 0.97⁎⁎ 0.92⁎⁎ 0.85⁎⁎ 0.73⁎⁎ −0.04 0.04 −0.68⁎⁎ −0.87⁎⁎ −0.50⁎⁎ −0.20⁎⁎ 0.22⁎⁎ −0.25⁎⁎ −0.12⁎⁎ −0.04 −0.18⁎⁎ −0.30⁎⁎ 0.06 −0.05 0.21⁎⁎ 0.42⁎⁎ 0.01 −0.08⁎ 0.26⁎⁎ −0.13⁎⁎ 0.07⁎ 0.24⁎⁎ −0.06 −0.26⁎⁎ 0.09⁎⁎ −0.19⁎⁎ −0.37⁎⁎ −0.48⁎⁎ −0.30⁎⁎ −0.23⁎⁎ 0.18⁎⁎ −0.29⁎⁎ −0.05 0.03 −0.20⁎⁎ −0.37⁎⁎ 0.08⁎ −0.17⁎⁎ −0.38⁎⁎ −0.50⁎⁎ −0.29⁎⁎ −0.22⁎⁎


296 297 298 299 300 301 302 303 304

306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337


AUC 1.2

0.0

Kappa

0.6

1.2

a

RF

0.0

0.6

1.2

a

a

CART

ab

bc

ab

MARS

bc

b

bc

bc

SVM

bc

GLM

bc

bc

LDA

c

c

c

c

The distribution ranges of each species were also evaluated in subwatersheds (Table 5). In Table 5, presence in the 2000s indicates the current total number of sub-watersheds where corresponding species inhabit, and addition and disappearance in the future represent the number of sub-watersheds in which these species are predicted to be present or absent under the climate change conditions, respectively. The number of habitable sub-watersheds for most endemic species is expected to decrease under the climate change conditions (Table 5). In particular, one species (Acheilognathus yamatsutae) observed in 51 sub-watersheds in the 2000s, and four species (Sarcocheilichthys variegatus wakiyae, Squalidus japonicus coreanus, M. longidorsalis, and Liobagrus andersoni) observed in fewer than 35 sub-watersheds, were predicted to be extinct in the future because of climate change. Meanwhile, only two species, Z. koreanus (77 sub-watersheds in 2000s and 91 sub-watersheds in 2080s) and S. gracilis majimae (88 subwatersheds in 2000s and 107 sub-watersheds in 2080s) were predicted to be present in an increased number of habitable sub-watersheds. Occurrence probabilities of endemic species predicted in the RF models were different along the longitudinal gradient (Figs. 5, 6). The mean occurrence probabilities of endemic species were decreased slightly in midstream areas of the river (stream order levels from 3 to 6), whereas those in upstream (stream order levels 1 and 2) and downstream areas (above stream order level 7) increased after 2040s (Fig. 5). On the other hand, the mean occurrence probability of endemic species increased at lower than 50 m of altitude.

349 350

4. Discussion

374

F

0.6

c

O

Accuracy 0.0

5

D

P

R O

Fig. 2. Comparison of average predictive performances of six different models based on predictions of current distributions of the 22 endemic fish species. (a) Accuracy (i.e., percentage of correctly predicted presence and absence), (b) area under the ROC curve (AUC), and (c) Cohen's kappa. RF: random forest, CART: classification and regression trees, GLM: generalized linear models, SVM: support vector machine, LDA: linear discriminant analysis, and MARS: multivariate adaptive regression splines. Different letters indicate statistically significant differences between six different models based on Dunn's multiple comparison test after a Kruskal–Wallis test (P b 0.05). Error bars indicate standard deviations.

338 339

TempDIF and TempJAN, were relatively important for predicting the distribution of Iksookimia koreensis and A. koreensis.

340

3.3. Effects on species distribution range

341 342

348

To assess the potential effects of climate change on the distribution range of endemic species, the RF models were developed with the climate change scenario. The distributions of most endemic species were predicted to change under the climate change conditions (Table 5). For instance, Coreoleuciscus splendidus and Hemiculter eigenmanni, the major indicator species for environmental pollution, were predicted to be affected by climate change and would nearly disappear from their current distributional areas in the future (Fig. 4).

t4:1 t4:2

Table 4 Predictions of 22 endemic species using the random forest models and evaluations of the importance of the variables in each model.

t4:3 t4:4 t4:5 t4:6 t4:7 t4:8 t4:9 t4:10 t4:11 t4:12 t4:13 t4:14 t4:15 t4:16 t4:17 t4:18 t4:19 t4:20 t4:21 t4:22 t4:23 t4:24 t4:25 t4:26

E T

C

E

R

R

347

Species name

N C O

345 346

Acanthorhodeus gracilis Rhodeus uyekii Squalidus gracilis majimae Coreoperca herzi Pseudobagrus koreanus Microphysogobio yaluensis Odontobutis platycephala Odontobutis interrupta Liobagrus mediadiposalis Acheilognathus yamatsutae Sarcocheilichthys nigripinnis Squalidus chankaensis Sarcocheilichthys variegatus wakiyae Hemiculter eigenmanni Acheilognathus koreensis Squalidus japonicus coreanus Coreoleuciscus splendidus Zacco koreanus Iksookimia koreensis Microphysogobio longidorsalis Koreocobitis rotundicaudata Liobagrus andersoni

U

343 344

Predicting changes in the distribution of species as a result of climate change has now become a major goal in ecology and conservation biology (Brooker et al., 2007), and it can contribute to the development of adaptation and mitigation strategies for biodiversity conservation (EEA, 2004; SCBD, 2003; Thomas et al., 2004). Generally, the determinants of species distribution may differ along environmental gradients in addition to climate change (Comte and Grenouillet, 2013). However, many studies that have been performed have considered only a limited view of ecological responses to climate change (Hampe and Petit, 2005; Parmesan, 2006), although using several descriptors is relevant, and necessary, to find out whether species are shifting across their entire distribution range. In this study, therefore, we investigated the relationships between the occurrence and environmental variables, including

Acronym

ACGR RHUY SQGR COHE PSKO MIYA ODPL ODIN LIME ACYA SANI SQCH SAVA HEEI ACKO SQJA COSP ZAKO IKKO MILO KORO LIAN

Accuracy

0.90 0.90 0.87 0.89 0.89 0.87 0.65 0.89 0.92 0.93 0.93 0.92 0.92 0.91 0.94 0.89 0.90 0.90 0.92 0.93 0.86 0.88

Kappa

0.60 0.62 0.74 0.75 0.56 0.73 0.27 0.75 0.35 0.74 0.51 0.79 0.50 0.66 0.62 0.51 0.77 0.80 0.81 0.83 0.72 0.76

Important variable 1st

2nd

PrcpJAN PrcpAVE Altitude Altitude PrcpJAN Order Order TempDIF TempDIF Order PrcpJAN Order Order Altitude TempDIF Order TempOCT Altitude TempDIF TempJAN TempOCT TempOCT

TempDIF PrcpJAN Order Order TempDIF Altitude Altitude TempJAN TempJAN TempOCT TempDIF TempJAN TempOCT TempOCT Altitude TempAPR Order TempJUL TempJAN TempOCT TempJAN TempJUL


351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373

375 376 377 378 379 380 381 382 383 384 385 386 387

6


100

ACGR

RHUY

SQGR

COHE

PSKO

MIYA

ODPL

ODIN

LIME

ACYA

SANI

SQCH

SAVA

HEEI

ACKO

COSP

ZAKO

KORO

LIAN

50 0 100 50 0

50

50

F

0 100

SQJA

O

Relative importance

100

IKKO

MILO

PrcpDIF PrcpSEP PrcpJUL PrcpMAR PrcpJAN PrcpAVE TempDIF TempSEP TempJUL TempMAR TempJAN TempAVE Altitude Slope Order


R O

0 100

Variables

Variables

50 0 PrcpDIF PrcpSEP PrcpJUL PrcpMAR PrcpJAN PrcpAVE TempDIF TempSEP TempJUL TempMAR TempJAN TempAVE Altitude Slope Order

Variables

T

E


Variables

D

0 100

P

50

C

Fig. 3. Importance of environmental variables for each species in the random forest models. Acronyms for environmental variables and species are given in Tables 2 and 4, respectively.

t5:1 t5:2 t5:3 t5:4

Table 5 Changes in the number of sub-watersheds predicted for each species to habit under the climate change. Presence indicates the total number of sub-watersheds where each species had habitats in the 2000s, and addition and disappearance represent the number of sub-watersheds where each species was predicted to have new habitats and disappeared in each decade, respectively. Species acronyms are 4.

R

Species

t5:6

N

Acanthorhodeus gracilis Rhodeus uyekii Squalidus gracilis majimae Coreoperca herzi Pseudobagrus koreanus Microphysogobio yaluensis Odontobutis platycephala Odontobutis interrupta Liobagrus mediadiposalis Acheilognathus yamatsutae Sarcocheilichthys nigripinnis Squalidus chankaensis Sarcocheilichthys variegatus wakiyae Hemiculter eigenmanni Acheilognathus koreensis Squalidus japonicus coreanus Coreoleuciscus splendidus Zacco koreanus Iksookimia koreensis Microphysogobio longidorsalis Koreocobitis rotundicaudata Liobagrus andersoni

U

t5:7 t5:8 t5:9 t5:10 t5:11 t5:12 t5:13 t5:14 t5:15 t5:16 t5:17 t5:18 t5:19 t5:20 t5:21 t5:22 t5:23 t5:24 t5:25 t5:26 t5:27 t5:28

change conditions. Of the six SDMs, RF showed the best predictability for the current distributions of endemic fish species (Fig. 2). Fukuda et al. (2013) reported the similar results, stating that RF had the best performance in the prediction of fish habitats, followed by SVM and CART. Of the SDMs, each modeling approach has strengths and

R

O

t5:5

2000s

2040s

Presence

Addition

Disappearance

No change

Addition

Disappearance

No change

42 55 88 70 43 78 78 49 20 51 20 63 34 45 28 27 56 77 40 20 21 21

13 22 9 3 7 3 3 1 6 1 0 10 3 8 5 3 4 15 1 0 0 0

28 17 2 14 28 3 4 29 11 36 20 9 28 26 24 24 11 0 18 18 7 10

14 38 86 56 15 75 74 20 9 15 0 54 6 19 4 3 45 77 22 2 14 11

4 17 20 0 1 4 6 4 3 0 1 10 0 4 1 0 2 14 0 0 0 0

38 41 1 40 36 5 7 26 20 51 20 24 34 40 28 27 27 0 37 20 19 21

4 14 87 30 7 73 71 23 0 0 0 39 0 5 0 0 29 77 3 0 2 0

C

390 391

E

392

climatic and geo-hydrological variables, of 22 endemic fish species using six different SDMs to assess the potential effects of climate change on stream fish species. We compared the prediction powers of the six different SDMs for predicting the distribution of 22 endemic fish species under climate

388 389

2080s


393 394 395 396 397

7

E

D

P

R O

O

F


405 406 407 408 409 410 411 412 413 414 415

E

R

403 404

R

402

N C O

400 401

weaknesses (Heikkinen et al., 2006), and they also differ in their abilities to summarize the responses of species distributions to climatic predictors (Segurado and Araújo, 2004). Moreover, in assessing the potential effects of climate change, seemingly small differences in the accuracies of the models in predicting the current distributions may result in disturbingly dissimilar projections of future distributions (Thuiller, 2003, 2004). Considering the differences in the objective functions and optimization algorithms across the SDMs, the use of multiple criteria is useful for illustrating differences in predictive performance (Fukuda et al., 2013; Mouton et al., 2010). Consequently, we used three criteria – accuracy, AUC, and Cohen's kappa – to solve these problems and to select the most suitable model for predicting fish species distributions in this study. The Earth's climate has warmed by approximately 0.61 °C over the past 100 years (Walther et al., 2002), and the temperature is expected to continue rising. In Korea, the National Institute of Meteorological Research (NIMR;, 2008) has predicted increases of 3.8 °C and 15% in the mean air temperature (in 2071–2100 from the 1971–2000 levels) and

U

398 399

C

T

Fig. 4. Examples of spatial distribution of species at different decades. (a) Coreoleuciscus splendidus and (b) Hemiculter eigenmanni. The spatial distribution of each species was predicted based on the random forest models with a climate change scenario (RCP 8.5). The color maps indicate the probabilities of the occurrence of each species and the gray circles indicate the sampling sites with occurrence probabilities greater than 0.5. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

precipitation (in 2079–2100 from the 1979–2000 levels), respectively. According to these predictions, most of the Korean Peninsula would have a subtropical climate by the end of the 21st century if the global temperature continues to increase at the current rate (Kwon et al., 2008); subsequent changes of species distribution range and community composition are expected to follow. Recently, Li et al. (2014) predicted that the impact of global warming on stream insects would lead to species extirpation of up to 20% in the highland areas and 2% in the lowland areas by the 2080s, and stream insect communities would become more homogenous under climate change. The increase in temperature has strong effects on the distributions, abundances, and assemblage compositions of fish (Attrill and Power, 2002; Cushing and Dickson, 1976; Genner et al., 2004; Kirby et al., 2006; Perry et al., 2005) that are mediated through changes in growth, survival, and reproduction. Freshwater fish cannot disperse across terrestrial areas and are limited to the river basin (Buisson et al., 2008). Therefore, climate change has been predicted to lead to species extinctions (Thomas et al., 2004) and to poleward shifts in the latitudinal

Fig. 5. Occurrence probability predicted in the random forest models across stream order under the climate change scenario.


416 417 418 419 420 421 Q9 422 423 424 425 426 427 428 429 430 431 432 433

8


2000s

2040s

2080s

ACGR RHUY SQGR COHE PSKO MIYA ODPL ODIN LIME

F

ACYA SANI

O

SQCH SAVA

R O

HEEI ACKO SQJA COSP IKKO

0.8 0.6 0.4

D

MILO KORO >7 6 5 4 3 2 1

>7 6 5 4 3 2 1

Up Down

>7 6 5 4 3 2 1

Up Down

0.0

Up

T

Down

E

LIAN

0.2

Occurrence level

P

ZAKO

1.0

C

Stream order

441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462

R

R

O

439 440

C

437 438

N

436

distribution ranges of species (Parmesan and Yohe, 2003; Perry et al., 2005). Our results are in line with these predictions. The mean occurrence probabilities of endemic species were decreased slightly in midstream regions (stream order levels from 3 to 6), although upstream (stream order levels 1 and 2) and downstream (above stream order level 7) regions have increased probabilities under the climate change conditions (Fig. 5). It was also notable that our model predicted the extirpation of some endemic species affected by the climate change. For example, five species (A. yamatsutae, S. variegatus wakiyae, S. japonicus coreanus, M. longidorsalis, and L. andersoni) were predicted to disappear from all of their habitable sub-watersheds (Table 5). The occurrence probability of Squalidus chankaensis tsuchigae in small streams (smaller than stream order 3) increased by more than 5% (1st: 11%, 2nd: 11%, 3rd: 6%) from the 2000s to the 2080s, though it fell off in large streams (higher than stream order 4) in the 2080s (Fig. 6). Overall, the decline of endemic fish species richness and their occurrence probability due to changes in habitat conditions induced by climate change lead to the poleward and upward shifts of stream fish species and species extirpation. Therefore, more attention and conservation strategies are needed to conserve these endemic species. In this study, we considered climatic variables and geo-hydrological variables for the prediction of species distribution. However, species distributions are also influenced by various other factors, such as water quality, biological factors (including food resources and predators), micro-environmental factors, and ecological factors (Angermeier and Karr, 1983; Hughes and Gammon, 1987; Reyjol et al., 2001). We did not consider these aspects in this study due to the lack of ecological information, but they may affect the probability of some endemic species adapting or shifting their habitats in Korean streams. In this context,

U

434 435

E

Fig. 6. Change in occurrence probability of each species predicted in the random forest models across stream order under the climate change scenario. Species acronyms are given in Table 4.

further studies on the biology and ecology of these species are required 463 before setting up a conservation plan. 464 5. Conclusions

465

In this study, we evaluated the potential impacts of climate change on 22 endemic freshwater fish using species distribution models (SDMs). To identify the most appropriate methods for the prediction of species occurrence, we compared the results of six different SDMs (LDA, RF, CART, GLM, SVM, and MARS). Among them, the RF method showed the best prediction power for the current distributions of species. Therefore, the RF model was selected to estimate the occurrence probabilities of the 22 endemic fish in terms of the climate change scenario. The model predicted the extirpations of five species (A. yamatsutae, S. variegatus wakiyae, S. japonicus coreanus, M. longidorsalis, and L. andersoni) from their habitats in the 2080s, showing impacts on endemic species and their occurrence probabilities. According to the predictions of the models, the endemic species richness and occurrence probabilities were decreased, and their distributions were shifted to the poleward and upstream shifts. Finally, the SDMs can help to understand the impacts of climate change on endemic freshwater fish and their interpretable information is useful for planning and management of target species under climate change conditions.

466 467

6. Uncited references

484 Q11

Clark et al., 2001 Eaton and Scheller, 1996 Shuter and Post, 1990


468 469 470 471 472 473 474 475 476 477 478 479 480 Q10 481 482 483

485 486 487


488

Acknowledgments

This study was conducted through the grants “National Aquatic Ecosystem Health Monitoring Program” by the Ministry of Environment 491 Q12 and the National Institute of Environmental Research (Korea).

O

R O

P

Angermeier, P.L., Karr, J.R., 1983. Fish Communities Along Environmental Gradients in a System of Tropical Streams. Evolutionary ecology of neotropical freshwater fishes Environmental Biology of Fishes. Springer, Netherlands, pp. 39–58. Araújo, M.B., Cabeza, M., Thuiller, W., Hannah, L., Williams, P.H., 2004. Would climate change drive species out of reserves? An assessment of existing reserve-selection methods. Glob. Chang. Biol. 10, 1618–1626. Attrill, M.J., Power, A.J., 2002. Climatic influence on a marine fish assemblage. Nature 417, 275–278. Austin, M.P., Meyers, J.A., 1996. Current approaches to modelling the environmental niche of eucalypts: implication for management of forest biodiversity. Forest Ecol. Manag. 85, 95–106. Balshi, M.S., McGuire, A.D., Duffy, P., Flannigan, M., Walsh, J., Melillo, J., 2009. Assessing the response of area burned to changing climate in western boreal North America using a Multivariate Adaptive Regression Splines (MARS) approach. Glob. Chang. Biol. 15 (3), 578–600. Begon, M., Townsend, C.R., Harper, J.L., 2006. Ecology: From Individuals to Ecosystems. Blackwell Publishing, Oxford, UK. Berry, P.M., Dawson, T.P., Harrison, P.A., Pearson, R.G., 2002. Modelling potential impacts of climate change on the bioclimatic envelope of species in Britain and Ireland. Glob. Ecol. Biogeogr. 11, 453–462. Breiman, L., 2001. Random forests. Mach. Learn. 45, 5–32. Breiman, L., Friedman, J.H., Olshen, R.A., Stone, C.J., 1984. Classification and Regression Trees. Wadsworth, Belmont, p. 368. Brito, J.C., Crespo, E.G., Paulo, O.S., 1999. Modelling wildlife distributions: logistic multiple regression vs overlap analysis. Ecography 22, 251–260. Brooker, R.W., Travis, J.M.J., Clark, E.J., Dytham, C., 2007. Modelling species' range shifts in a changing climate: the impacts of biotic interactions, dispersal distance and the rate of climate change. J. Theor. Biol. 245, 59–65. Buisson, L., Blanc, L., Grenouillet, G., 2008a. Modelling stream fish species distribution in a river network: the relative effects of temperature versus physical factors. Ecol. Freshw. Fish 17, 244–257. Buisson, L., Thuiller, W., Lek, S., Lim, P., Grenouillet, G., 2008b. Climate change hastens the turnover of stream fish assemblages. Glob. Chang. Biol. 14, 2232–2248. Buisson, L., Grenouillet, G., Casajus, N., Lek, S., 2010. Predicting the potential impacts of climate change on stream fish assemblages. Am. Fish. Soc. Symp. 73, 327–346. Carpenter, S.R., Kitchell, J.F., Hodgson, J.R., 1985. Cascading trophic interactions and lake productivity. Bioscience 35, 634–639. Chen, S.Q., Fath, B.D., Chen, B., 2011. Information-based network environ analysis: a system perspective for ecological risk assessment. Ecol. Indic. 11 (6), 1664–1672. Chen, S.Q., Chen, B., Fath, B.D., 2013. Ecological risk assessment on the system scale: a review of state-of-the-art models and future perspectives. Ecol. Model. 250, 25–33. Clark, M.E., Rose, K.A., Levine, D.A., Hargrove, W.W., 2001. Predicting climate change effects on Appalachian trout: combining GIS and individual-based modeling. Ecol. Appl. 11 (1), 161–178. Cohen, J., 1960. A coefficient of agreement for nominal scales. Educ. Psychol. Meas. 20, 37–46. Comte, L., Grenouillet, G., 2013. Do stream fish track climate change? Assessing distribution shifts in recent decades. Ecography 36, 1236–1246. Cristianini, N., Shawe-Taylor, J., 2000. An Introduction to Support Vector Machines. Cambridge University Press, Cambridge, UK. Cushing, D.H., Dickson, R.R., 1976. The biological response in the sea to climatic changes. Adv. Mar. Biol. 14, 1–122. Daufresne, M., Boët, P., 2007. Climate change impacts on structure and diversity of fish communities in rivers. Glob. Chang. Biol. 13, 2467–2478. Eaton, J.G., Scheller, R.M., 1996. Effects of climate warming on fish thermal habitat in streams of the United States. Limnol. Oceanogr. 41 (5), 1109–1115. EEA (European Environment Agency)., 2004. Impacts of Europe's Changing Climate: An Indicator-based Assessment, (Copenhagen, Denmark). Engler, R., Gnguisan, A., Rechsteiner, L., 2004. An improved approach for predicting the distribution of rare and endangered species from occurrence and pseudo-absence data. J. Appl. Ecol. 41, 263–274. ESRI., 2008. ArcGIS 9.3. Environmental Systems Research Institute Inc. Fielding, A.H., Bell, J.F., 1997. A review of methods for the assessment of prediction errors in conservation presence/absence models. Environ. Conserv. 24, 38–49. Friedman, J.H., 1991. Multivariate adaptive regression splines. Ann. Stat. 19, 1–141. Fukuda, S., de Baets, B., Waegeman, W., Verwaeren, J., 2013. Habitat prediction and knowledge extraction for spawning European grayling (Thymallus thymallus L.) using a broad range of species distribution models. Environ. Model Softw. 47, 1–6. Garzon, M.B., Blazek, R., Neteler, M., de Dios, R.S., Ollero, H.S., Furlanello, C., 2006. Predicting habitat suitability with machine learning models: the potential area of Pinus sylvestris L. in the Iberian Peninsula. Ecol. Model. 197 (3), 383–393. Genner, M.J., Sims, D.W., Wearmouth, V.J., Southall, E.J., Southward, A.J., Henderson, P.A., Hawkins, S.J., 2004. Regional climatic warming drives long-term community changes of British marine fish. Proc. R. Soc. B 271, 655–661. Gillanders, B.M., Elsdon, T.S., Halliday, I.A., Jenkins, G.P., Robins, J.B., Valesini, F.J., 2011. Potential effects of climate change on Australian estuaries and fish utilising estuaries: a review. Mar. Freshw. Res. 62, 1115–1131.

D

493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568

E

References

U

N C O

R

R

E

C

T

492

Guisan, A., Edwards Jr., T., Hastle, T., 2002. Generalized linear and generalized additive models in studies of species distributions: setting the scene. Ecol. Model. 157, 89–100. Hampe, A., Petit, R.J., 2005. Conserving biodiversity under climate change: the rear edge matters. Ecol. Lett. 8, 461–467. 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 (6), 1–27. Hijmans, R.J., Elith, J., 2013. Species Distribution Modeling with R. R CRAN Project. Hill, J.K., Thomas, C.D., Huntley, B., 2003. Modeling Present and Potential Future Ranges of European Butterflies Using Climate Response Surfaces. In: Bogs, C., Watt, W., Ehrlich, P. (Eds.), Butterflies. Ecology and evolution taking flight. The University of Chicago Press, Chicago, pp. 149–167. Hughes, R.M., Gammon, J.R., 1987. Longitudinal changes in fish assemblages and water quality in the Willamette River. Or. Trans. Am. Fish. Soc. 116 (2), 196–209. Iverson, L.R., Prasad, A.M., 1998. Predicting abundance of 80 tree species following climate change in the eastern United States. Ecol. Monogr. 68, 465–485. Iverson, L.R., Prasad, A.M., 2002. Potential redistribution of tree species habitat under five climate change scenarios in the eastern US. Forest Ecol. Manag. 155, 205–222. Jenkins, M., 2003. Prospects for biodiversity. Science 302, 1175–1177. Karr, J.R., Freemark, K.E., 1985. Disturbance and Vertebrates: An Integrative Perspective. In: Pickett, S.T.A., White, P.S. (Eds.), Ecology of Natural Disturbance and Patch Dynamics. Academic Press, New York, NY, USA, pp. 153–168. Kirby, R.R., Johns, D.G., Lindley, J.A., 2006. Fathers in hot water: rising sea temperatures and a Northeastern Atlantic pipefish baby boom. Biol. Lett. 2, 597–600. Kouamélan, E.P., Teugels, G.G., N'Douba, V., Bi, G.G., Koné, T., 2003. Fish diversity and its relationships with environmental variables in a West African basin. Hydrobiologia 505, 139–146. Kwon, Y.A., Kwon, W.T., Boo, K.O., 2008. Future projections on the spatial distribution of onset date and duration of natural seasons using SRES A1B data in South Korea. Korean Geogr. Soc. 43, 36–51. Kwon, Y.-S., Li, F., Chung, N., Bae, M.-J., Hwang, S.-J., Byoen, M.-S., Park, S.-J., Park, Y.-S., 2012. Response of fish communities to various environmental variables across multiple spatial scales. Int. J. Environ. Res. Public Health 9. Landis, J., Koch, G., 1977. The measurement of observer agreement for categorical data. Biometrics 33, 159–174. Leathwick, J.R., Whitehead, D., McLeod, M., 1996. Predicting changes in the composition of New Zealand's indigenous forests in response to global warming: a modelling approach. Environ. Softw. 11, 81–90. Li, F., Chung, N., Bae, M.-J., Kwon, Y.-S., Kwon, T.-S., Park, Y.-S., 2013. Temperature change and macroinvertebrate biodiversity: assessments of organism vulnerability and potential distributions. Clim. Chang. 119, 421–434. Luoto, M., Pöyry, J., Heikkinen, R.K., Saarinen, K., 2005. Uncertainty of bioclimate envelope models based on geographical distribution of species. Glob. Ecol. Biogeogr. 14, 575–584. Marschner, I., 2013. glm2: fitting generalized linear modelsR Package Version 1.1.1. http://www.r-project.org. Matthews, W.J., 1998. Patterns in Freshwater Fish Ecology. Chapman & Hall, New York. McCauley, R.W., 1990. Determining the health of fish communities — parallels with human medicine. J. Great Lakes Res. 16, 635–636. McCullagh, P., Nelder, J.A., 1989. Generalized Linear Models (Monographs on Statistics and Applied Probability 37). Chapman Hall, London. McLachlan, G.H., 2004. Discriminant Analysis and Statistical Pattern Recognition. Wiley, New York, p. 552. Meyer, D., Dimitriadou, E., Hornik, K., Weingessel, A., Leisch, F., Chang, C.-C., Lin, C.-C., 2014. e1071: Misc Functions of the Department of Statistics (e1071). TU Wien, (R Package Version 1.6-3. http://www.r-project.org.). Milborrow, S., Tibshirani, R., 2014. Earth: multivariate adaptive regression spline modelsR Package Version 3.2-7. http://www.r-project.org. Milly, P.C.D., Dunne, K.A., Vecchia, A.V., 2005. Global pattern of trends in streamflow and water availability in a changing climate. Nature 438, 347–350. MOE/NIER (The Ministry of Environment/National Institute of Environmental Research, Korea), 2008. The Survey and Evaluation of Aquatic Ecosystem Health in Korea. MOE/NIER, Inchon, Korea. Moisen, G.G., Frescino, T.S., 2002. Comparing five modelling techniques for predicting forest characteristics. Ecol. Model. 157, 209–225. Moisen, G.G., Freeman, E.A., Blackard, J.A., Frescino, T.S., Zimmermann, N.E., Edwards Jr., T.C., 2006. Predicting tree species presence and basal area in Utah: a comparison of stochastic gradient boosting, generalized additive models, and tree-based methods. Ecol. Model. 199 (2), 176–187. Montgomery, D.C., Peck, E.A., Vining, G.G., 2001. Introduction to Linear Regression Analysis, 3rd ed. Wiley Interscience, New York, NY. Moore, P.D., 2003. Back to the future: biogeographical responses to climate change. Prog. Phys. Geogr. 27, 122–129. Mouton, A.M., De Baets, B., Goethals, P.L.M., 2010. Ecological relevance of performance criteria for species distribution models. Ecol. Model. 221, 1995–2002. NIMR (National Institute of Meteorological Research)., 2008. The Application of Regional Climate Change Scenario for the National Climate Changes Report (IV). Korea Meteorological Administration (KMA), Seoul ((In Korean)). Parmesan, C., 1996. Climate and species range. Nature 382, 765–766. Parmesan, C., 2006. Ecological and evolutionary responses to recent climate change. Annu. Rev. Ecol. Evol. Syst. 37, 637–669. Parmesan, C., Yohe, G., 2003. A globally coherent fingerprint of climate change impacts across natural systems. Nature 421, 37–42. Pearce, J., Ferrier, S., 2000. Evaluating the predictive performance of habitat models developed using logistic regression. Ecol. Model. 133, 225–245.

F

489 490

9


569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654

F

O

R O

U

N

C

O

R

R

E

C

T

734

Somero, G.N., 1997. Temperature Relationships: From Molecules to Biogeography. In: Danztler, W.H. (Ed.), Handbook of physiology: section 13 comparative physiology. Oxford University Press, Oxford, pp. 1391–1444 (vii.). Staudt, A., Leidner, A.K., Howard, J., Brauman, K.A., Dukes, J., Hansen, L., Paukert, C., Sabo, J., Solórzano, L.A., Johnson, K., 2012. Chapter 5: Impacts of Climate Change on Already Stressed Biodiversity, Ecosystems, and Ecosystem Services. In: Staudinger, M.D., Grimm, N.B., Staudt, A., Carter, S.L., Chapin III, F.S., Kareiva, P., Ruckelshaus, M., Stein, B.A. (Eds.), Impacts of Climate Change on Biodiversity, Ecosystems, and Ecosystem Services: Technical Input to the 2013 National Climate Assessment. Cooperative Report to the 2013 National Climate, Assessment (5-1 – 5-36). Swets, J.A., 1988. Measuring the accuracy of diagnostic systems. Science 240, 1285–1293. Therneau, T., Atkinson, B., Ripley, B., 2014. rpart: recursive partitioning and regression trees. R package version 4.1-8. http://www.r-project.org. Thomas, C.D., Cameron, A., Green, R.E., Bakkenes, M., Beaumont, L.J., Collingham, Y.C., Erasmus, B.F.N., de Siqueira, M.F., Grainger, A., Hannah, L., Hughes, L., Huntley, B., van Jaarsveld, A.S., Midgley, G.F., Miles, L., Ortega-Huerta, M.A., Peterson, A.T., Phillips, O.L., Williams, S.E., 2004. Extinction risk from climate change. Nature 427, 145–148. Thuiller, W., 2003. BIOMOD — optimizing predictions of species distributions and projecting potential future shifts under global change. Glob. Chang. Biol. 9, 1353–1362. Thuiller, W., 2004. Patterns and uncertainties of species' range shifts under climate change. Glob. Chang. Biol. 10, 2020–2027. Thuiller, W., 2007. Biodiversity: climate change and the ecologist. Nature 448 (7153), 550–552. Thuiller, W., Lavorel, S., Araújo, M.B., Sykes, M.T., Prentice, I.C., 2005. Climate change threats to plant diversity in Europe. Proc. Natl. Acad. Sci. 102, 8245–8250. UNESCO., 2003. The United Nations World Water Development Report: Water for People. Water for Life. UNESCO & Berghahn Books, Paris. Walther, G.R., Post, E., Convey, P., Menzel, A., Parmesank, C., Beebee, T.J.C., Fromentin, J.M., Hoegh-Guldberg, O., Bairlein, F., 2002. Ecological responses to recent climate change. Nature 416, 389–395. Wehrly, K.E., Wiley, M.J., Seelbach, P.W., 2003. Classifying regional variation in thermal regime based on stream fish community patterns. Trans. Am. Fish. Soc. 132, 18–38. Wood, C.M., McDonald, D.G., 1997. Global Warming: Implications for Freshwater and Marine Fish. Cambridge University Press, Cambridge. Wootton, J.T., Power, M.E., 1993. Productivity, consumers, and the structure of a river food chain. Proc. Natl. Acad. Sci. U. S. A. 90, 1384–1387.

P

Perry, A.L., Low, P.J., Ellis, J.R., Reynolds, J.D., 2005. Climate change and distribution shifts in marine fishes. Science 308, 1912–1915. Poff, N.L., Brinson, M.M., Day, J.W.J., 2002. Aquatic Ecosystems and Global Climate Change: Potential Impacts on Inland Freshwater and Coastal Wetland Ecosystems in the United States. Pew Center on Global Climate Change, Arlington, VA, USA. Pont, D., Hugueny, B., Oberdorff, T., 2005. Modelling habitat requirement of European fishes: do species have similar responses to local and regional environmental constraints? Can. J. Fish. Aquat. Sci. 62, 163–173. Pop, H.F., Sârbu, C., 2013. A new fuzzy discriminant analysis method. MATCH Commun. Math. Comput. Chem. 69, 391–412. Pouteau, R., Meyer, J.-Y., Taputuarai, R., Stoll, B., 2012. Support vector machines to map rare and endangered native plants in Pacific islands forests. Ecol. Inform. 9, 37–46. Power, M.E., Matthews, W.J., Stewart, A.J., 1985. Grazing minnows, piscivorous bass, and stream algae: dynamics of a strong interaction. Ecology 66, 1448–1456. Prasad, A.M., Iverson, L.R., 2000. Predictive Vegetation Mapping Using a Custom Built Model-chooser: Comparison of Regression Tree Analysis and Multivariate Adaptive Regression Splines. 4th International Conference on Integrating GIS and Environmental Modeling: Problems, Prospects and Research Needs. Banff, Alberta, Canada. Revenga, C., Campbell, I., Abell, R., de Villiers, P., Bryer, M., 2005. Prospects for monitoring freshwater ecosystems towards the 2010 targets. Phil. Trans. R. Soc. B 360, 397–413. Reyjol, Y., Lim, P., Belaud, A., Lek, S., 2001. Modelling of microhabitat used by fish in natural and regulated flows in the river Garonne (France). Ecol. Model. 146 (1), 131–142. Ripley, B., Venables, B., Bates, D.M., Hornik, K., Gebhardt, A., Firth, D., 2014. MASS: support functions and datasets for Venables and Ripley's MASS. R Package Version 7.3-31. http://www.r-project.org. Robnik-Sikonja, M., 2004. Improving random forests. Mach. Learn. 3201, 359–370. Robnik-Sikonja, M., Savicky, P., 2013. CORElearn: CORElearn — classification, regression, feature evaluation and ordinal evaluation. R package version 0.9.42. http://www.rproject.org. Rodríguez, M.A., 1995. Habitat-specific estimates of competition in stream salmonids: a field test of the isodar model of habitat selection. Evol. Ecol. 9, 169–184. SCBD (Secretariat of the Convention on Biological Diversity)., 2003. Inter Linkages Between Biological Diversity and Climate Change and Advice on the Integration of Biodiversity Considerations into the Implementation of the United Nations Framework Convention on Climate Change and its Kyoto protocol (CBD Technical Series no. 10). SCBD, Montreal, Canada. Segurado, P., Araújo, M., 2004. An evaluation of methods for modelling species distributions. J. Biogeogr. 31, 1555–1569. Shuter, B.J., Post, J.R., 1990. Climate, population viability, and the zoogeography of temperate fishes. Trans. Am. Fish. Soc. 119, 314–336.

D

655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695


E

10


696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733