Climate Change and Mountaintop-Removal Mining:
A MaxEnt Assessment of the Potential Threat to West
Virginian Fishes
Lindsey R.F. Hendrick and Daniel J. McGarvey
Northeastern Naturalist, Volume 26, Issue 3 (2019): 499–522
Full-text pdf (Accessible only to subscribers. To subscribe click here.)
Access Journal Content
Open access browsing of table of contents and abstract pages. Full text pdfs available for download for subscribers.
Current Issue: Vol. 30 (3)
Check out NENA's latest Monograph:
Monograph 22
Northeastern Naturalist Vol. 26, No. 3
L.R.F. Hendrick and D.J. McGarvey
2019
499
2019 NORTHEASTERN NATURALIST 26(3):499–522
Climate Change and Mountaintop-Removal Mining:
A MaxEnt Assessment of the Potential Threat to West
Virginian Fishes
Lindsey R.F. Hendrick1,* and Daniel J. McGarvey1
Abstract - Accounts of species’ range shifts in response to climate change are rapidly accumulating.
These range shifts are often attributed to species tracking their thermal niches
as temperatures in their native ranges increase. Our objective was to estimate the degree to
which shifts in water temperature driven by climate change may increase the exposure of
West Virginia’s native freshwater fishes to mountaintop-removal surface coal mining. We
projected midcentury shifts in habitat suitability for 9 non-game West Virginian fishes via
maximum entropy species distribution modeling, using a combination of physical habitat,
historical climate conditions, and future climate data. Modeling projections for a high-emissions
scenario (Representative Concentration Pathway 8.5) predict that habitat suitability
will increase in high-elevation streams near mountaintop mining sites for 8 of 9 species,
with increases in habitat suitability varying from 46% to 418%. We conclude that many
West Virginian fishes will be at risk of increased exposure to mountaintop mining if climate
change continues at a rapid pace.
Introduction
Quantifying and predicting species’ responses to climate change is currently
a high-priority research topic in biogeographical and conservation science (e.g.,
Angert et al. 2011, Lin et al. 2017, Pecl et al. 2017). In the Northern Hemisphere,
species are responding by shifting their ranges to the north or to higher elevations
(e.g., Chen et al. 2011, Chivers et al. 2017, Comte and Grenouillet 2013). These
latitudinal and elevational range shifts may be a result of species tracking their thermal
preferences as temperatures in their historical, native ranges increase (Comte
et al. 2013, Freeman and Class Freeman 2014, Parmesan 2006). If so, range shifts
should be most likely for vagile species that are physically capable of long-distance
movements and for ectothermic species that have narrow thermal tolerances (Calosi
et al. 2008, Deutsch et al. 2008).
Freshwater fishes of the Central Appalachian region in eastern North America
may be particularly likely to shift to higher elevations in response to a warming
climate. Like most primary freshwater fishes, they are obligate ectotherms that
may encounter stressful or lethal conditions as ambient temperatures increase (see
Farrell 2011). Furthermore, most major rivers in this region flow westward off of
the Appalachian range to the Ohio River, which is also a predominantly westward
1Center for Environmental Studies, Virginia Commonwealth University, Richmond, VA
23284. *Corresponding author - hendricklrf@gmail.com.
Manuscript Editor: Jay Stauffer
Northeastern Naturalist
500
L.R.F. Hendrick and D.J. McGarvey
2019 Vol. 26, No. 3
flowing river that originates at the confluence of the Alleghany and Monongahela
rivers, near Pittsburgh, PA (~42°N latitude), or they flow east to the Atlantic Ocean.
Thus, the topography and elevation of the Appalachian range may provide opportunities
for freshwater fishes to shift their ranges upslope, while latitudinal shifts that
extend far north will not be feasible for many populations.
Unfortunately, fishes that shift to higher elevations in central Appalachia may be
at risk of encountering another threat: increased exposure to mountaintop-removal
(MTR) surface coal mining. Mountaintop-removal mining is pervasive throughout
central Appalachia (Ferreri et al. 2004) and is particularly common in the state of
West Virginia (Fig. 1). Damages to aquatic biota may occur through acute loss of
headwater streams (via burial by valley fill) or chronic degradation of water quality
and instream habitat further downstream (Bernhardt et al. 2012). Empirical reports
of MTR impacts on native fishes take on various forms from the individual-level
toxic effects of selenium, a common byproduct of MTR that causes teratogenic
deformities (Lemly 1993, Palmer et al. 2010), to assemblage-level effects including
decreased species richness and lower population densities (Hitt and Chambers
Figure 1. Map of the study site showing the Ohio River Basin (ORB; light gray), the state
of West Virginia (outline), and the active mountaintop-removal (MTR; black) mining sites
within West Virginia.
Northeastern Naturalist Vol. 26, No. 3
L.R.F. Hendrick and D.J. McGarvey
2019
501
2014). Habitat models also suggest that MTR may have a negative effect on fish
distributions, pushing them further downstream (Hopkins and Roush 2013).
In this study, we used maximum entropy (MaxEnt; Phillips et al. 2006) species
distribution models (SDMs) to assess whether climate change and MTR may
pose an interactive threat to the native fishes of West Virginia. This was a 2-stage
process in which we first used physical habitat and historical climate data to build
SDMs for a subset of the native fishes of West Virginia. We then predicted future
habitat suitability under 2 midcentury climate-change scenarios. For each species
and future climate scenario, we assessed changes in habitat suitability for streams
in close proximity to MTR operations. However, our intent was not to model the effects
of MTR on West Virginian fishes per se. Instead, we characterized the degree
to which climate change may increase fish exposure to MTR via warming-induced
upslope range shifts. Our analyses focused on a representative subset of non-game
species. Other investigators have studied climate change and MTR effects on West
Virginian game fishes, such as Salvelinus fontinalis (Mitchill) (Brook Trout; e.g.,
Ries and Perry 1995), but little is known about the potential consequences for the
region’s diverse non-game fishes.
Specific research objectives were to (1) build SDMs for a select subset of nongame
fish species that are broadly representative of the native ichthyofauna of West
Virginia, (2) predict changes in habitat suitability under 2 midcentury climate scenarios,
and (3) use the projected habitat suitability maps to identify species that are
likely to migrate to higher elevations, thereby increasing their exposure to MTR.
Methods
Fish species selection and occurrence data
We selected 9 non-game species for inclusion in our modeling study from the
176 documented native West Virginian fishes (Stauffer et al. 1995). Selection was
guided by an iterative process. First, we removed species from the candidate list
that did not have at least 200 occurrence records within the Ohio River Basin.
We then sought to ensure that the selected species would be broadly representative
of the autecological characteristics of all native West Virginian fishes in the
Ohio River Basin (i.e., westward flowing rivers). To do so, we used a multivariate
species-traits approach. We obtained species-level descriptions for 13 functional
traits through an extensive literature review, as detailed in Woods and McGarvey
(2018), then compiled them in a species × trait matrix. Traits included multiple
indicators of body size (e.g., maximum total length, female length at maturation,
mean egg diameter), maximum longevity, degree of parental care, adult habitat,
vertical water-column position, adult feeding behavior, egg-deposition strategy,
spawning season, and migratory behavior. We used the species × trait matrix to
calculate pairwise Gower dissimilarities for all 136 native West Virginian fishes
within the Ohio River Basin. Gower dissimilarity is commonly used in functional
traits analysis because it is compatible with a combination of continuous (e.g., total
length) and categorical (e.g., feeding behavior) variables (Gower 1971). Next,
we conducted principal coordinates analysis to build a 2-dimensional trait-space
Northeastern Naturalist
502
L.R.F. Hendrick and D.J. McGarvey
2019 Vol. 26, No. 3
ordination of the 136 fishes. Finally, we used the ordination plot to visually
confirm that the selected fish species encompassed a large fraction of the total
trait-space and were therefore representative of the overall range of West Virginian
fish functional traits (Fig. 2).
We obtained presence-only occurrence records for each of the selected species
from the spatially explicit Ichthymaps digital database (Frimpong et al. 2016).
We included occurrence records distributed throughout the entire Ohio River
Basin, the parent drainage to most rivers in West Virginia. Incorporating species’
complete ranges throughout the Ohio River Basin, rather than truncated ranges
within West Virginia, ensured that the MaxEnt background samples (see Species
distribution models subsection below) would be representative of all habitats
available to the modeled species (Elith et al. 2011, Yates et al. 2018). To account
for potential spatial bias in the Ichthymaps occurrence records, we applied a
spatial thinning algorithm to the occurrence data. Using the spThin package in
R (Aiello-Lammens et al. 2015), we applied a nearest-neighbor search radius
of 10 km to each Ichthymaps occurrence record. Fewer than 2% of all occurrence
points were within 10 linear km of each other. We therefore concluded that
geographic sampling bias was not a significant concern and retained all of the
occurrence data in model development.
Figure 2. Ordination
plot of the first 2 principal
coordinate axes
from the functionaltraits
analysis. The 9
fish species modeled
in this study are shown
as solid black circles.
The remaining 127
native fishes of West
Virginia (Ohio River
Basin) are shown as
gray circles. Functional
traits that are strongly
correlated with each
axis are shown along
their respective axes,
with the direction of the
correlation indicated
by arrows. Light gray
shaded regions indicate
sections of the overall
trait space that are not well represented by the 9 modeled species. The light gray shaded
regions show that all but the largest species were included in our models and that lentic-type
species are not well represented in our models. All other regions of the overall trait space
are well represented.
Northeastern Naturalist Vol. 26, No. 3
L.R.F. Hendrick and D.J. McGarvey
2019
503
River network, physical habitat, and climate data
We used the 1:100,000 scale National Hydrography Dataset Plus, Version 2
(NHDplus V2; McKay et al. 2012) digital stream network, clipped to the Ohio
River Basin, as a common physical template for all fish occurrence records, environmental
covariates (i.e., predictor variables), and SDMs. In the NHDplus V2,
every digital stream segment has a unique “COMID” identifier that we used to
cross-reference all fish occurrence and covariate data to their respective locations
within the Ohio River Basin. We obtained physical habitat covariates from the
NHDplus V2 attribute tables and the Stream-Catchment dataset (StreamCat; Hill et
al. 2016). We selected these physical habitat covariates to represent 4 broad classes
of potential effects on fish habitat: topographic, geologic, hydrologic, or urb an.
We downloaded historical (1960–1990) and midcentury (2041–2060) climate
data from WorldClim, Version 1.4 (Hijmans et al. 2005), as 30 arc-second–resolution
grids. We captured midcentury data for 2 representative concentration pathway
scenarios (RCPs): RCP 4.5 served as a mid-range emissions scenario (Thomson et
al. 2011) and RCP 8.5 served as a high-range emissions scenario (Riahi et al. 2011).
For both RCPs, we downloaded midcentury projections for 6 general circulation
models: BCC-CSM1-1 (Wu et al. 2014), CCSM4 (Gent et al. 2011), GFDL-CM3
(Donner et al. 2011), GISS-E2-R (Schmidt et al. 2014), HadGEM2-CC (Martin et
al. 2011), and MRI-CGCM3 (Yukimoto et al. 2012). We re-projected all climate
grids to a common 1-km resolution grid spanning the entire Ohio River Basin. We
calculated ensemble mean averages for both RCP 4.5 and RCP 8.5 for monthly
air temperature and monthly precipitation in each 1-km grid cell. We performed
all grid calculations in ESRI ArcMap 10.5 software (Environmental Systems Research
Institute, Redlands, CA). We then appended gridded air temperature and
precipitation values to the NHDplus V2 stream network by superimposing the
climate grids directly onto the digital stream network, using System for Automated
Geoscientific Analyses Version 2.1.4 software (Institute of Geography, Physical
Geography Section, Hamburg University, Hamburg, DE). From these air temperature
and precipitation data, we calculated mean annual streamflow for every digital
stream segment in the Ohio River Basin using the Ohio River Basin-specific linear
regression model of Vogel et al. (1999). We then calculated mean monthly stream
temperatures using the logistic regression model of Segura et al. (2015).
Finally, we generated a Pearson correlation (r) matrix for all of the NHDplus
V2, StreamCat, and derived climate variables (streamflow and stream temperatures)
and used it to screen highly correlated covariates (|r| ≥ 0.70) from the models. We
selected a subset of 30 covariates that were potentially relevant to freshwater fishes
from the remaining variables and transformed them, as necessary, for normality
(see Table 1 for the complete covariate list with definitions, units of measure, and
transformations).
Species distribution models
We used MaxEnt to build the SDMs because we were working solely with presence-
only data (vs. presence–absence data; see Elith et al. 2006), and despite the
Northeastern Naturalist
504
L.R.F. Hendrick and D.J. McGarvey
2019 Vol. 26, No. 3
Table 1. List of the final 30 covariates that were considered in each of the Maximum Entropy species distribution models. Each covariate was calculated or
interpolated at 1 of 2 spatial extents: an average value within the local catchment, or a discrete stream segment. For each covariate, units of measurement,
data transformations (if applicable), and data sources are listed. [Table continued on following page.]
Covariate Description Spatial extent Units Transformation Source
Area Surface area of catchment Catchment km2 ln(x + 1) StreamCat
AvWetnessIndex Mean wetness (composite topographic index) within catchment Catchment na na StreamCat
BarrenLand Percent of catchment classified as barren land cover Catchment % na StreamCat
BFI Base flow index as fraction of total flow due to base flow Catchment % na StreamCat
SiO2 Percent lithological silicon dioxide content in near-surface geology Catchment % na StreamCat
within catchment
CoalMines Density of coal mines within catchment Catchment Number/km2 ln(x + 1) StreamCat
CropLand Percent of local catchment classified as row crop land cover Catchment % na StreamCat
Elevation Mean elevation within catchment Catchment m na StreamCat
Fe2O3 Percent lithological ferric oxide content in near-surface geology Catchment % na StreamCat
within catchment
ForestLoss Percent tree canopy cover loss throughout catchment and within Catchment % ln(x + 1) StreamCat
100-m buffer of stream channel
JanuaryPrecip Mean January precipitation (1960-1990) Stream mm na WorldClim
JanuaryStreamTemp Mean January stream temperature (1960–1990) Stream °C na Original
calculation
JulyStreamTemp Mean July stream temperature (1960–1990) Stream °C na Original
calculation
JunePrecip Mean June precipitation (1960–1990) Stream mm na WorldClim
Northeastern Naturalist Vol. 26, No. 3
L.R.F. Hendrick and D.J. McGarvey
2019
505
Table 1, continued.
Covariate Description Spatial extent Units Transformation Source
Mines Density of mines and mineral plants within catchment Catchment Number/km2 ln(x + 1) StreamCat
NABD Density of dams within catchment Catchment Number/km2 ln(x + 1) StreamCat
Nitrogen Percent lithological nitrogen content in near-surface geology within Catchment % na StreamCat
catchment
NPDES Density of National Pollutant Discharge Elimination System sites Catchment Number/km2 ln(x + 1) StreamCat
within catchment
OpenWater Percent of catchment classified as open water Catchment % na StreamCat
OrganicMatter Percent organic matter within catchment Catchment % ln(x + 1) StreamCat
RoadCrossings Density of road-stream crossings within catchment Catchment Number/km2 ln(x + 1) StreamCat
Roads Length of roads within catchment per unit surface area of catchment Catchment km/km2 na StreamCat
Runoff Estimated runoff with catchment (1971–2000) Catchment mm na StreamCat
Sand Percent sand in near-surface geology within catchment Catchment % na StreamCat
Slope Slope of stream channel Stream m/m ln NHDplus v2
StreamOrder Strahler stream order Stream na na NHDplus v2
Streamflow Mean annual streamflow Stream m3/s ln(x + 1) Original
calculation
Superfund Density of Superfund sites within catchment Catchment Number/km2 ln(x + 1) StreamCat
TRI Density of Toxic Release Inventory sites within catchment Catchment Number/km2 ln(x + 1) StreamCat
WaterTableDepth Mean water table depth within catchment Catchment cm na StreamCat
Northeastern Naturalist
506
L.R.F. Hendrick and D.J. McGarvey
2019 Vol. 26, No. 3
recent recognition that MaxEnt is only one of several methods to build point-process
models (Renner et al. 2015), it remains among the most flexible and accessible
software tools for doing so. Briefly, our modeling process was as follows. We used
historical species occurrence and environmental covariate data to build a MaxEnt
model for each of the 9 selected fishes. We created a background sample for each
species by randomly selecting 20% of the complete landscape (i.e., 20% of all
NHDplus V2 stream segments within the Ohio River Basin). To ease interpretation
of individual covariate effects, we constrained the MaxEnt models to simple hinge
and quadratic features (see Elith et al. 2011). We ran MaxEnt with a maximum of
10,000 background points, 500 iterations, and a convergence threshold of 0.00001.
For model development, we used prior knowledge of species’ autecology and
standard MaxEnt diagnostics to evaluate the 30 potential covariates and then selected
a subset of covariates (n ≤ 9; Moreno-Amat et al. 2015) that were the most
effective predictors of a given species’ occurrence (Fourcade et al. 2017, Petitpierre
et al. 2016). Covariate evaluation was an iterative process that included standard
MaxEnt percent contribution and permutation importance summary tables for individual
covariates as well as covariate jackknife plots (Phillips 2017).
We used MaxEnt regularized training gain as an index of model fit. Regularized
training gain is a measure of the distance between a multivariate distribution of covariates
at randomly selected background sites (i.e., a random sample of the entire
landscape that a species could potentially inhabit) and a corresponding distribution
of covariates at sites of known species occurrences (Elith et al. 2011). Hence, a
large training gain indicates an affinity for a narrow range of environmental conditions,
relative to the broader landscape, while a small training gain suggests a lack
of specialized habitat requirements (i.e., the distribution of covariates at occurrence
sites mirrors the background distribution; Merow et al. 2013). We also used
the exponential transformation of the MaxEnt regularized training gain for each
SDM to aid in model evaluation. The exponential of the regularized training gain
is the ratio of habitat suitability between sites of known occurrence and randomly
selected background sites (Phillips 2017). Exponential values much larger than
1 are indicative of species with specialized habitat requirements; because these
specialist species occupy a narrow range of habitats relative to the complete range
of available habitats, SDMs can more efficiently discriminate between suitable and
unsuitable habitat.
Notably, we did not use the replication features of MaxEnt (cross-validation,
bootstrapping, or single-split subsampling; see Phillips 2017) to assess model
generality. Our goal was only to model potential habitat suitability throughout the
study landscape, not to predict the probability of presence at any given locality.
Thus, we did not seek to estimate omission and commission error rates. Our study
was designed to characterize a potential threat to freshwater fishes within a unique
and highly context-specific scenario: upstream movement towards MTR sites in
West Virginia. Transferability to other regions, where environmental conditions and
MTR activity are almost certain to be different, was not an objective in our modelbuilding
exercises.
Northeastern Naturalist Vol. 26, No. 3
L.R.F. Hendrick and D.J. McGarvey
2019
507
Once we had specified a final MaxEnt model for each fish species in the final
subset, we projected habitat suitability to a midcentury time horizon (2041–2060)
under the RCP 4.5 and RCP 8.5 climate change scenarios. We obtained the projections
by substituting future values of the climate-driven covariates (streamflow,
stream temperature, and precipitation) for the historical values used to build and
parameterize each model. Then, by comparing aggregate distributions of MaxEnt
raw output values among historical and future landscapes, we were able to estimate
potential shifts in habitat suitability for each fish species.
Finally, we used a spatial querying process to identify stream segments that are
likely to be impacted by MTR operations in the state of West Virginia. We began
with a digital map of all active MTR permit boundaries from the West Virginia Department
of Environmental Protection (http://tagis.dep.wv.gov/home/Downloads;
downloaded on 24 October 2017). We built a 10-km–radius buffer around each of
the MTR sites in ArcMap. The 10-km buffer provided an estimate of the potential
spatial footprint of MTR effects on local aquatic ecosystems; in several instances,
significant effects of MTR on aquatic biota have been documented at downstream
distances >10 km (e.g., Bernhardt et al. 2012, Lindberg et al. 2011, Pond et al.
2008). By using the MTR buffer to query potentially impacted stream reaches from
the complete river network, we were able to test the hypothesis that climate change
is likely to increase exposure of West Virginian fishes to MTR.
We used the nonparametric one-sided Mann–Whitney U test to make comparisons
between historical and future SDM projections (Woods and McGarvey
2018). This test compared the distribution of ranks between 2 unpaired datasets.
The datasets were combined and each value was ranked from smallest to largest.
We calculated the average ranks of the members of each group from this rank
distribution; a large difference between the groups’ mean ranks suggested the
distributions were distinct. Future changes in habitat suitability were also expressed
as percentages, relative to historical suitability, according to the formula:
% change = (future median raw score - historical median raw score) ÷ historical
median raw score × 100.
Results
Species selection
Following functional trait analyses, we selected a representative subset of 9
non-game fish species from the families Catostomidae, Cottidae, Cyprinidae, and
Percidae (Table 2). These 4 families constitute the majority of native fish diversity
in West Virginia (Stauffer et al. 1995), and an abundance of occurrence records was
available for species in each family (Frimpong et al. 2016). Within families, we
selected species in proportion to the overall richness of the respective family. For
instance, we selected 4 from Cyprinidae, the most diverse family, but only 2 species
from the less diverse Catostomidae. In general, the 9 selected species spanned the
overall extent of the 2-dimensional trait-space represented by the native West Virginian
fishes. We excluded from the representative subset only the largest fishes that
reside in large, mainstem tributaries of the Ohio River (e.g., Aplodinotus grunniens
Northeastern Naturalist
508
L.R.F. Hendrick and D.J. McGarvey
2019 Vol. 26, No. 3
Rafinesque [Freshwater Drum]) or that prefer slow-flowing, lentic habitats in deep
rivers and pools (e.g., Notropis wickliffi Trautman [Channel Shiner]) (Fig. 2).
MaxEnt models
Regularized training gain varied from 0.510 to 1.613 among the final SDMs.
Though variable, the final covariate lists included in each of the models exhibited
Table 2. Summary information on the fit and structure of the Maximum Entropy (MaxEnt) species distribution
models. For each species, the number of occurrence records (n) used to build the model and
the MaxEnt regularized training gain (rtg) are shown with the exponential of the rtg in parentheses.
MaxEnt % contribution and permutation importance diagnostics are also shown for each covariate that
was retained in a species’ final model. [Table continued on following page.]
% Permutation
Species Covariate contribution importance
Catostomidae
Catostomus commersonii (Lacépède) (White Sucker)
n = 2478 Catchment area 41.8 55.7
rtg = 0.614 (1.848) Mean annual streamflow 24.4 11.8
Mean June precipitation 11.7 11.1
Catchment runoff 7.6 6.8
Catchment road crossings 4.4 0.5
Catchment BFI 4.2 2.3
Catchment elevation 4.1 8.5
Stream order 1.7 3.4
Hypentelium nigricans (Lesueur) (Northern Hogsucker)
n = 2716 Catchment area 47.7 52.5
rtg = 0.565 (1.759) Mean annual streamflow 31.6 27.6
Catchment elevation 9.6 11.1
Mean June precipitation 3.1 3.3
Catchment road crossings 2.9 1.2
Catchment BFI 2.7 2.1
Catchment water table depth 2.3 2.3
Cottidae
Cottus bairdii Girard (Mottled Sculpin)
n = 893 Catchment BFI 18.0 6.7
rtg = 1.166 (3.209) Catchment area 17.0 28.1
Catchment elevation 16.7 25.3
Catchment runoff 13.6 20.9
Mean annual streamflow 13.5 10.3
Catchment sand 9.0 4.7
Catchment road crossings 6.5 0.6
Mean June precipitation 5.7 3.3
Cyprinidae
Campostoma anomalum (Rafinesque) (Central Stoneroller)
n = 3156 Catchment area 42.1 46.6
rtg = 0.540 (1.716) Mean annual streamflow 34.7 31.7
Mean June precipitation 5.5 8.3
Catchment elevation 5.2 7.1
Catchment runoff 4.8 3.0
Catchment road crossings 3.8 0.6
Catchment BFI 2.3 1.1
Catchment Fe2O3 1.4 1.5
Northeastern Naturalist Vol. 26, No. 3
L.R.F. Hendrick and D.J. McGarvey
2019
509
Table 2, continued.
% Permutation
Species Covariate contribution importance
Notropis buccatus (Cope) (Silverjaw Minnow)
n = 1892 Catchment runoff 28.4 7.1
rtg = 0.869 (2.385) Mean annual streamflow 22.2 50.4
Catchment area 21.4 5.9
Catchment road crossings 8.1 1.2
Catchment BFI 4.8 5.1
Stream order 3.9 6.5
Catchment open water 3.9 0.1
Mean June precipitation 3.8 9.4
Mean January precipitation 3.4 14.4
Luxilus chrysocephalus Rafinesque (Striped Shiner)
n = 460 Mean annual streamflow 42.4 36.3
rtg = 1.613 (5.018) Mean January precipitation 24.3 33.3
Catchment area 8.6 6.4
Catchment wetness index 7.9 2.8
Catchment organic matter 5.9 6.1
Catchment BFI 4.6 10.3
Catchment runoff 4.4 2.4
Catchment elevation 1.9 2.3
Semotilus atromaculatus (Mitchill) (Creek Chub)
n = 3146 Catchment area 48.9 69.8
rtg = 0.510 (1.665) Mean annual streamflow 31.8 13.8
Catchment road crossings 7.0 1.1
Mean June precipitation 5.0 5.3
Catchment runoff 3.8 3.4
Stream order 2.2 4.9
Catchment elevation 1.3 1.7
Percidae
Etheostoma blennioides Rafinesque (Greenside Darter)
n = 2617 Catchment area 47.4 49.3
rtg = 0.597 (1.817) Mean annual streamflow 30.8 27.1
Catchment elevation 11.1 15.6
Catchment open water 3.2 0.2
Catchment runoff 3.1 4.5
Catchment road crossings 2.6 0.3
Mean January precipitation 1.8 2.9
Etheostoma caeruleum Storer (Rainbow Darter)
n = 2365 Catchment area 42.8 43.2
rtg = 0.625 (1.868) Mean annual streamflow 25.3 19.1
Catchment elevation 16.9 22.3
Mean January precipitation 5.4 7.3
Catchment road crossings 4.9 0.2
Catchment runoff 4.7 7.9
some clear commonalities. Catchment area (percent contribution = 41.8–48.9%)
and mean annual streamflow (percent contribution = 22.2–42.4%) were the 2
most important predictors of suitability in 6 of 9 models, and both variables were
Northeastern Naturalist
510
L.R.F. Hendrick and D.J. McGarvey
2019 Vol. 26, No. 3
included in each of the final models (Table 2). Mean January or mean June precipitation
was also included in all final models. Notably, stream temperature was not a
strong predictor of suitability for any species and was thus excluded from all models.
Complete MaxEnt results for each of the 9 modeled species, including sample
sizes, model fit diagnostics, and indices of importance for individual covariates are
shown in Table 2.
Interestingly, we found that the importance of some covariates tended to be preserved
among species within a shared taxonomic affinity. For example, suitability
was positively associated with catchment area and inversely related to streamflow
for both Etheostoma blennioides (Greenside Darter) and Etheostoma caeruleum
(Rainbow Darter). However, the shapes of the MaxEnt response curves for a given
covariate were variable. For instance, habitat suitability was negatively associated
with runoff of ~350–900 mm for the Rainbow Darter, but the runoff-response
curve was bimodal with peaks at ~325 mm and ~900 mm for the Greenside Darter.
We identified similar family-level responses to catchment area and streamflow
among Catostomidae species and the Cyprinidae species, with the exception of
Luxilus chrysocephalus Rafinesque (Striped Shiner), for which habitat suitability
was positively associated with streamflow of ~0.1–3.4 m3/s but declined at higher
streamflow. Predicted suitability among the Cyprinids generally shared a negative
association with June precipitation in the range of ~80–140 mm/y and with runoff
of ~325–675 mm/y, with the Striped Shiner again standing out as the exception;
June precipitation did not have a strong influence on Striped Shiner habitat suitability,
but runoff of ~500–625 mm/y did. Individual covariate response curves
are provided in the MaxEntReport html file for each species, available on Figshare
(DOI:10.6084/m9.figshare.6106682).
Predicted habitat suitability
Under the RCP 4.5 climate scenario, significant increases (Mann–Whitney: P <
0.001) in habitat suitability were predicted within the MTR buffer for the 2 darters
(Etheostoma), but none of the remaining species (Fig. 3). Under the RCP 8.5
scenario, however, habitat suitability was predicted to significantly increase within
the MTR buffer for 8 of 9 species. Percent increases in median habitat suitability
under the RCP 8.5 scenario varied from 46% to 418%, relative to the historical
habitat suitability values, with a grand median increase of 125% (Fig. 3). Only
Notropis buccatus (Silverjaw Minnow) was predicted to experience a decrease
in habitat suitability within the MTR buffer. Maps of MaxEnt raw scores for the
Striped Shiner within the MTR buffer under the historical, RCP 4.5, and RCP 8.5
scenarios are shown in Figure 4. Summary distributions of all MaxEnt habitat suitability
predictions are illustrated in Figure 3.
Discussion
Climate-change effects: hydrology vs temperature
Using MaxEnt SDMs for 9 representative species, we tested the hypothesis that
climate change is likely to drive directional, upslope shifts in habitat suitability for
Northeastern Naturalist Vol. 26, No. 3
L.R.F. Hendrick and D.J. McGarvey
2019
511
Figure 3. Boxplots of the distributions of MaxEnt raw scores. Box elements are standard
percentiles (see key). For each species, the distributions of raw scores across the Ohio River
Basin (ORB and within the mountaintop removal buffer (MTR) are shown for historical
data as well as the RCP 4.5 and RCP 8.5 midcentury climate-change scenarios. Percent
differences in median MaxEnt raw scores, comparing historical averages with future RCP
4.5 and RCP 8.5 averages, are shown for ORB and MTR data at right of each box. Mann-
Whitney test P-values (paired sample tests using individual stream segments as replicates)
comparing historical MaxEnt raw scores with future projections are also shown in parentheses
for the RCP 4.5 and RCP 8.5 results within the MTR buffer.
Northeastern Naturalist
512
L.R.F. Hendrick and D.J. McGarvey
2019 Vol. 26, No. 3
Figure 4. Maps of MaxEnt
raw scores for the Striped
Shiner within the MTR buffer
under the (a) historical,
(b) RCP 4.5, and (c) RCP. 8.5
scenarios. Maximum suitability
scores are shown in
blue and minimum suitability
scores are shown in red.
Northeastern Naturalist Vol. 26, No. 3
L.R.F. Hendrick and D.J. McGarvey
2019
513
West Virginian freshwater fishes, thereby placing the best habitats in close proximity
to MTR operations. In many instances, our results supported this hypothesis, but
the support was conditional, varying among species and climate-change scenarios.
Furthermore, we were surprised by the factors that ultimately drove the shifts in
habitat suitability. Ambient temperature is a fundamental regulator of niche space
for ectotherms (Coulter et al. 2014, Deutsch et al. 2008); thus, we expected stream
temperature to be a key predictor of fish habitat suitability. This was not the case,
however, and stream temperature was not included in any of the final models (see
Table 2).
The apparent lack of a strong temperature effect may indicate that the modeled
species have relatively broad thermal tolerances. At historical occurrence sites, the
widest ranges of winter and summer stream temperature values for a given species
(i.e., stream temperature ranges among all sites of known presence) spanned 2.74
°C for Semotilus atromaculatus (Creek Chub) (min. = 0.38 °C, max. = 3.12 °C) and
6.24 °C for the Silverjaw Minnow (min. = 22.72 °C, max. = 28.96 °C), respectively.
Conversely, the narrowest ranges of historical winter and summer stream temperatures
for a given species spanned 2.62 °C for the Striped Shiner (min. = 0.40 °C,
max. = 3.02 °C) and 5.53° C for Cottus bairdii (Mottled Sculpin) (min. = 23.36 ° C;
max. = 28.89 °C), respectively. In all cases, species’ historical winter and summer
temperature ranges encompassed a large fraction of the historical temperature range
across the entire Ohio River Basin; historical winter and summer stream temperatures
spanned 2.76 °C (min. = 0.36 °C, max. = 3.12 °C) and 6.32 °C (min. = 22.69
°C, max. = 29.01 °C), respectively.
Together, these observations suggest that each of the 9 modeled fishes would be
physiologically capable of occupying most or all of the streams in the Ohio River
Basin, if mean winter or summer stream temperature were the sole determinant
of habitat suitability. The fact that documented occurrences of each of the modeled
species were limited to a subset of streams within the Ohio Basin suggests
that factors other than stream temperature are fundamental in regulating fish species’
presences. With specific reference to MaxEnt, the fact that historical sample
temperatures exhibit so much overlap with the background temperatures indicates
that mean winter and summer stream temperatures may not be useful for discriminating
between suitable and unsuitable fish habitat.
Instead, species’ responses to climate change were driven primarily by hydrology.
In each model, hydrologic variables were among the best predictors of fish
occurrence (Table 2). For example, mean annual streamflow was included in every
model, and in 8 of 9 cases, it was the first or second most-influential covariate when
ranked by MaxEnt percent contribution statistics (13.5–42.4%). Other covariates
that represent hydrology or a dimension of the hydrologic cycle included summer
and winter precipitation (one of which was included in every model), runoff
(included in 8 of 9 models), and the baseflow index (included in 6 models). Importantly,
these different hydrology covariates did not provide redundant information.
When calculated across the entire Ohio River Basin, Pearson correlation coefficients
among these covariates never exceeded the collinearity threshold of |r| > 0.70
Northeastern Naturalist
514
L.R.F. Hendrick and D.J. McGarvey
2019 Vol. 26, No. 3
and in several instances were much lower. For instance, correlations between mean
annual streamflow and January precipitation and between streamflow and June
precipitation were r < 0.01 and r = -0.40, respectively. Furthermore, January and
June precipitation were not highly correlated (r = 0.35). Thus, we concluded that
the various hydrologic covariates (streamflow, precipitation, runoff, baseflow index)
represented different dimensions of the hydrologic cycle and were therefore
appropriate for inclusion in the same models.
In retrospect, the strong effect of hydrology was not surprising, given that
hydrology is widely regarded as a master variable in lotic ecosystems (Poff et al.
1997). Streamflow is a dynamic integration of many physical processes occurring
across the landscape. Though it is clearly a function of precipitation, streamflow is
also influenced by the geologic and antecedent factors that regulate surface runoff,
soil water, and groundwater dynamics (Poff et al. 1997). In this way, streamflow
becomes an efficient indicator of many different yet interrelated influences on
aquatic habitat (McGarvey and Terra 2016). Effects of these hydrologic influences
include direct, individual-level physiological and behavioral mechanisms (Mims
and Olden 2011, Poff and Allan 1995, Poff et al. 1997) as well as emergent patterns
in species’ distributions and overall richness (McGarvey 2014, Power et al. 1995,
Wenger et al. 2011). We therefore believe it is logical that streamflow, rather than
stream temperature, proved to be a primary determinant of habitat suitability in the
fish models.
It should be noted that our hydrologic variables did not account for potential
effects of MTR on streamflow (Evans et al. 2015). Disruptive effects of MTR on
local flow include large-scale deforestation and the resulting decreases in evapotranspiration
and groundwater recharge, as well as increased surface runoff via
soil compaction from heavy machinery operation (Griffith et al. 2012). These alterations
may lead to elevated peak-flow conditions downstream from MTR sites
(Nippgen et al. 2017, Wiley et al. 2001), but observed downstream effects have
so far been variable across spatial and temporal scales (Evans et al. 2015, Ross
et al. 2016, Zegre et al. 2014). Due to this variability, it is difficult to predict if
or how climate-driven changes in hydrology may simultaneously be affected by
MTR, but it is plausible that future hydrology in the immediate vicinity of MTR
sites may deviate from our predicted conditions, with unknown effects on fish
habitat suitability.
Differential responses to climate change
Of the 9 modeled species, the 2 Etheostoma spp. may be the most likely to shift
their ranges upstream in response to climate change. The Greenside Darter and
Rainbow Darter were the only species predicted to experience significant increases
in habitat suitability within the MTR buffer under both climate change scenarios
(Fig. 3). For all other species, significant increases in habitat suitability were limited
to the RCP 8.5 scenario. This responsiveness to both scenarios was driven by
the strong influence of January precipitation on darter habitat suitability. Increasing
levels of January precipitation within the MTR buffer were evident among the
Northeastern Naturalist Vol. 26, No. 3
L.R.F. Hendrick and D.J. McGarvey
2019
515
historical (mean = 83.8 mm; sd = 6.3 mm), RCP 4.5 (mean = 92.9; sd = 7.3), and
RCP 8.5 (mean = 99.2; sd = 8.3) datasets. Predicted suitability for both Etheostoma
spp. was, in turn, positively associated with January precipitation, particularly with
totals of ~40–100 mm/month. Should January precipitation exceed 100 mm/month,
however, habitat suitability is likely to decrease (see MaxEnt response-curve plots
in the MaxEntReport html file for each species, available on Figshare; DOI:10.6084/
m9.figshare.6106682.) We therefore suggest that the projected moderate to large increases
in winter precipitation are likely to make habitat in high-elevation streams
in West Virginia more suitable for Greenside Darter and Rainbow Darter.
Currently, we do not know what specific mechanism might link winter precipitation
to Etheostoma habitat. Both darters in this study lack air bladders and reside
in benthic, riffle habitats, often sheltering from fast currents behind large rocks or
woody debris (Fahy 1954, Harding et al. 1998). This tendency to avoid suspension
in fast water is consistent with our observation that Etheostoma habitat suitability
decreased with increasing mean annual streamflow (each species’ response curve
for streamflow is available in the respective MaxEntReport, available, as above, on
Figshare; DOI:10.6084/m9.figshare.6106682). This result seems counterintuitive,
in light of the seeming preference for higher winter precipitation, which should
increase mean annual streamflow. One parsimonious explanation is that pulses in
winter flow may prevent coarse riffle substrates from becoming embedded with
fine sediments, thereby maintaining the critical physical habitats that Etheostoma
darters utilize year-round. This dynamic was documented for a Greenside Darter
population in the Grand River, ON, Canada (Bunt et al. 1998) and, if it is similarly
applicable in streams in West Virginia, it is cause for heightened concern because
increased fines and sedimentation are common byproducts of MTR (Griffith et al.
2012, Nelson et al. 1991).
January precipitation and mean annual streamflow were also key predictors of
habitat suitability for the Striped Shiner. However, the effect of streamflow differed
from the effect on habitat of Etheostoma spp. Striped Shiner habitat suitability increased
rapidly with increasing streamflow (see MaxEntReport on Figshare), rather
than decreasing from a low modal streamflow value as it did for Etheostoma spp.
This finding accounts for the dramatic increase in Striped Shiner habitat suitability
under RCP 8.5 (Fig. 3); a large increase in mean annual streamflow is expected
within the MTR buffer under RCP 8.5 (mean = 0.18 m3/s; sd = 0.51 m3/s; streamflow
values are transformed (ln[x+1]), relative to historical streamflow (mean = 0.06; sd
= 0.22). However, a comparable increase in streamflow is not expected under RCP
4.5 (mean = 0.06; sd = 0.24). Thus, a significant increase in Striped Shiner habitat
suitability was not predicted for RCP 4.5 (Fig. 3).
Predicted increases in habitat suitability for 2 of the remaining species of Cyprinidae,
the Creek Chub and Campostoma anomalum (Central Stoneroller), as
well as all species of Catostomidae and Cottidae, were also driven mean annual
streamflow. For each species, maximum habitat suitability occurred at a relatively
low streamflow value, then decreased to a stable, asymptotic level with increasing
streamflow (see streamflow response curves in MaxEntReports on Figshare;
Northeastern Naturalist
516
L.R.F. Hendrick and D.J. McGarvey
2019 Vol. 26, No. 3
DOI:10.6084/m9.figshare.6106682).For example, habitat suitability for the
Creek Chub peaked at a mean annual streamflow value of ~0.02 (in natural log
transformed units) then rapidly decreased at higher streamflows. In these cases,
the most suitable streamflow values were similar to the average streamflow values
predicted within the MTR buffer for RCP 8.5 (median streamflow = 0.012 in
natural log units), but not for RCP 4.5 (median streamflow = 0.002). Thus, significant
increases in habitat suitability within the MTR buffer were limited to the
RCP 8.5 scenario.
Decreasing habitat suitability within the MTR buffer was predicted only for the
Silverjaw Minnow (Fig. 3) and was due to the strong, negative associations that
mean annual streamflow and precipitation had with habitat suitability. For each
of these covariates, suitability peaked at a low value then rapidly decreased (see
response curves on Figshare). These modeling observations are consistent with
field studies that showed the Silverjaw Minnow is often abundant in small streams
but highly sensitive to periods of low-flow drought and to high-flow events (Toth
et al. 1982, Wallace 1972). In theory, the decreasing suitability of stream habitat
within the MTR buffer might serve to protect the Silverjaw Minnow; this species
is unlikely to be impacted by MTR if it does not inhabit streams near MTR operations.
Unfortunately, the Silverjaw Minnow will experience no net benefit if the
availability of small streams that it has historically occupied is greatly diminished
by climate change.
Are the model predictions cause for concern?
Our results suggest that, in a warming climate, habitat suitability for 8 of the
9 modeled species is likely to increase in high-elevation streams near MTR operations.
However, we cannot prove that any of our predicted changes in habitat
suitability will come to pass, or that the study species will in fact migrate to streams
within the MTR buffer. We therefore conclude with some general thoughts on the
relevance of our modeling process and findings.
First, we emphasize that the RCP 4.5 and 8.5 climate change scenarios, though
heuristic in nature, are broadly recognized by the scientific community as valid and
entirely plausible. Indeed, Smith et al. (2011) have shown that global warming of
2 °C beyond pre-industrial levels may be achieved as early as 2030 and that 4 °C
warming may occur as soon as 2060. Similarly, Betts et al. (2011) estimate global
mean temperature will increase by 4 °C above pre-industrial levels between 2060
and 2070. Others propose global mean temperature is highly likely to exceed the
benchmark of 2 °C by 2030, citing a likely temperature increase of from 2 °C to
4.9 °C by 2100 (Raftery et al. 2016). Collectively, these reports show, despite uncertainty,
that the range of outcomes bracketed by the RCP 4.5 and RCP 8.5 climate
scenarios are plausible and could be reached by midcentury.
Second, there are relatively few migration barriers in the rivers and streams
of West Virginia that would categorially prevent fishes from migrating to higher
elevations near MTR sites. Numerous lock-and-dam structures that may constrain
movement and reduce population connectivity exist along the mainstem Ohio,
Northeastern Naturalist Vol. 26, No. 3
L.R.F. Hendrick and D.J. McGarvey
2019
517
Kanawha, and Monongahela rivers in West Virginia, but these structures are at
least semi-permeable to fish movement (Argentina et al. 2018). In general, the
number of large, impassible dams on westward-flowing Ohio River tributaries
is modest in comparison to other eastern US states (USACE 2016). Using a GIS,
we performed a manual search for large, impassible dams that would prevent
upstream fish movement and identified 10: Hawks Nest (inclusive of all upstream
dams on the mainstem New River and Kanawha Falls, ~5 km downstream of
Hawks Nest Dam), Summersville, Sutton, Taylor Fork, Shannonpin Mine, Cheat
Lake, Tygart, R.D. Bailey, Upper Mud River No. 2A, and East Lynn. The total
length of stream channel that was upstream of any of these barriers and within the
MTR buffer was 1445 km, or ~9% of the 15,732 km of total stream channel within
the buffer.
Less-conspicuous barriers could also constrain future fish movement. For
instance, road crossings and culverts often impede fish movement (Januchowski-
Hartley et al. 2013, Warren and Pardew 1998). This barrier is a point of concern
because the density of road crossings was a good predictor of fish occurrence,
and therefore selected as a final predictor variable, for 8 of the 9 modeled species
(Table 2). Currently, we do not have comprehensive data that could be used
to incorporate road crossings into our analyses in a spatially explicit manner, but
we do note that most of the fishes in this study have broad ranges that historically
include some mid- to high-elevation streams. Thus, it is likely that even in a landscape
that is highly fragmented by road crossings, some potential colonists are already
present near the MTR buffer sites and therefore capable of moving to them
in a changing climate.
Finally, we submit that our specific results should be broadly representative of
the native stream ichthyofauna of West Virginia. Functional-trait analysis indicated
that the selected model species covered much of the functional trait space encompassed
by all Ohio River Basin fishes of West Virginia (Fig. 2). Furthermore, the 4
families represented by our selected subset—Catostomidae, Cottidae, Cyprinidae,
and Percidae—include 114 non-game fish species and represent 65% of all native
fishes in West Virginia (Stauffer et al. 1995). As noted above, the model-predicted
shifts in high-elevation habitat suitability were generally positive (i.e., increasing
suitability) for 8 of 9 species. Thus, we believe it is logical to predict that habitat
suitability for many of the remaining fishes will respond in a s imilar manner.
Although we did not model MTR effects on fishes per se, we posit that the
predicted tendency for habitat suitability to increase near MTR sites is, of itself,
legitimate cause for concern. The most acute, negative effect of MTR on freshwater
fishes will be direct habitat loss as MTR overburden is dumped as valley-fill,
effectively eliminating headwater streams. Further downstream, chemical contaminants
will accumulate through leaching and as coal is washed to lower its sulfur
content. Toxicants from MTR are known to cause infertility (Palmer et al. 2010),
teratogenic deformities (Palmer et al. 2010), and death among individual fishes
(Ferreri et al. 2004), as well as population- and assemblage-level declines in fish
abundance and diversity (Ferreri et al. 2004, Hitt and Chambers 2014). In southern
Northeastern Naturalist
518
L.R.F. Hendrick and D.J. McGarvey
2019 Vol. 26, No. 3
West Virginia, more than 750 km of high-elevation streams have already been
buried by MTR waste, and chronic effects of MTR are now impacting 2800–4300
km of additional stream habitat (Bernhardt et al. 2012). Thus, we conclude that the
combined effects of climate change and MTR are likely to pose very real and significant
threats to many of West Virginia’s native freshwater fishes by midcentury.
Acknowledgments
We thank Catalina Segura for sharing the complete results and parameters from the
stream-temperature model (Segura et al. 2015). Taylor Woods assembled and shared the fish
functional-trait database. Lindsey Hendrick received financial support through the Virginia
Commonwealth University Rice Rivers Center. Daniel McGarvey received financial support
through the National Science Foundation (DEB-1553111) and the Eppley Foundation for
Scientific Research.
Literature Cited
Aiello-Lammens, M.E., R.A.Boria, A. Radosavljevic, B. Vilela, and R.P. Anderson. 2015.
spThin: An R package for spatial thinning of species occurrence records for use in ecological
niche models. Ecography 38(5):541–545.
Angert, A.L., L.G. Crozier, L.J. Rissler, S.E. Gilman, J.J. Tewksbury, and A.J. Chunco.
2011. Do species’ traits predict recent shifts at expanding range edges? Ecology Letters
14:677–689.
Argentina, J.E., P.L. Angermeier, E.M. Hallerman, and S.A. Welsh. 2018. Spatial extent of
analysis influences observed patterns of population genetic structure in a widespread
darter species (Percidae). Freshwater Biology 63(10):1185–1198.
Bernhardt, E.S., B.D. Lutz, R.S. King, J.P. Fay, C.E. Carter, A.M. Helton, and J. Amos.
2012. How many mountains can we mine? Cumulative impacts of surface coal mining
on Central Appalachian rivers. Environmental Science and Technology 46:8115–8122.
Betts, R.A., M. Collins, D.L. Hemming, C.D. Jones, J.A. Lowe, and M.G. Sanderson. 2011.
When could global warming reach 4 °C? Philosophical Transactions of the Royal Society
A: Mathematical, Physical, and Engineering Sciences 369(1934):67–84.
Bunt, C.M., S.J. Cooke, and R.S. Mckinley. 1998. Creation and maintenance of habitat
downstream from a weir for the Greenside Darter, Etheostoma blennioides: A rare fish
in Canada. Environmental Biology of Fishes 51(3):297–308.
Calosi, P., D.T. Bilton, and J.I. Spicer. 2008. Thermal tolerance, acclimatory capacity, and
vulnerability to global climate change. Biology Letters 4(1):99–102.
Chen, I.C., J.K. Hill, R. Ohlemüller, D.B. Roy, and C.D. Thomas. 2011. Rapid range shifts of
species associated with high levels of climate warming. Science 333(6045):1024–1026.
Chivers, W.J., A.W. Walne, and G.C. Hays. 2017. Mismatch between marine plankton
range movements and the velocity of climate change. Nature Communications 8:14434.
DOI:10.1038/ncomms14434.
Comte, L., and G. Grenouillet. 2013. Do stream fish track climate change? Assessing distribution
shifts in recent decades. Ecography 36(11):1236–1246.
Comte, L., L. Buisson, M. Daufresne, and G. Grenouillet. 2013. Climate-induced changes
in the distribution of freshwater fish: Observed and predicted trends. Freshwater Biology
58(4):625–639.
Coulter, D.P., M.S. Sepúlveda, C.D. Troy, and T.O. Höök. 2014. Thermal habitat quality of
aquatic organisms near power plant discharges: Potential exacerbating effects of climate
warming. Fisheries Management and Ecology 21(3):196–210.
Northeastern Naturalist Vol. 26, No. 3
L.R.F. Hendrick and D.J. McGarvey
2019
519
Deutsch, C.A., J.J. Tewksbury, R.B. Huey, K.S. Sheldon, C.K. Ghalambor, D.C. Haak, and
P.R. Martin. 2008. Impacts of climate warming on terrestrial ectotherms across latitude.
Proceedings of the National Academy of Sciences 105(18):6668–6672.
Donner, L.J., B.L. Wyman, R.S. Hemler, L.W. Horowitz, Y. Ming, M. Zhao, et al.. 2011.
The dynamical core, physical parameterizations, and basic simulation characteristics of
the atmospheric component AM3 of the GFDL global coupled model CM3. Journal of
Climate 24(13):3484–3519.
Elith, J., C.H. Graham, R.P. Anderson, M. Dudik, S. Ferrier, A. Guisan, et al.. 2006. Novel
methods improve prediction of species’ distributions from occurrence data. Ecography
29(2):129–151.
Elith, J., S.J. Phillips, T. Hastie, M. Dudík, Y.E. Chee, and C.J. Yates. 2011. A statistical
explanation of MaxEnt for ecologists. Diversity and Distributions 17(1):43–57.
Evans, D.M., C.E. Zipper, E.T. Hester, and S.H. Schoenholtz. 2015. Hydrologic effects of
surface coal mining in Appalachia (US). Journal of the American Water Resources Association
51(5):1436–1452.
Fahy, W.E. 1954. The life history of the Northern Greenside Darter, Etheostoma Blennioides
Blennioides Rafinesque. Journal of the Elisha Mitchell Scientific Society 70(2):139–205.
Farrell, A.P. 2011. Encyclopedia of Fish Physiology: from Genome to Environment, 1st Edition.
Elsevier, London, UK. 2272 pp.
Ferreri, C.P., J.R. Stauffer, and T.D. Stecko. 2004. Evaluating impacts of mountaintop
removal/valley-fill coal mining on stream fish populations. Pp. 576–592, In R.I. Barnhisel
(Ed.). Proceedings of a Joint Conference of the American Society of Mining and
Reclamation 21st Annual National Conference and the 25th West Virginia Surface Mine
Drainage Task Force Symposium, American Society of Mining and Reclamation, Lexington,
KY.
Fourcade, Y., A.G. Besnard, and J. Secondi. 2018. Paintings predict the distribution of
species or the challenge of selecting environmental predictors and evaluation statistics.
Global Ecology and Biogeography 27(2):245–256.
Freeman, B.G., and A.M. Class Freeman. 2014. Rapid upslope shifts in New Guinean birds
illustrate strong distributional responses of tropical montane species to global warming.
Proceedings of the National Academy of Sciences 111(12):4490–4494.
Frimpong, E.A., J. Huang, and Y. Liang. 2016. IchthyMaps: A database of historical distributions
of freshwater fishes of the United States. Fisheries 41( 10):590–599.
Gent, P.R., G. Danabasoglu, L.J. Donner, M.M. Holland, E.C. Hunke, S.R. Jayne, D.M.
Lawrence, R.B. Neale, P.J. Rasch, M. Vertenstein, P.H. Worley, Z-L. Yang, and M.
Zhang. 2011. The community climate system model version 4. Journal of Climate
24(19):4973–4991.
Gower, J.C. 1971. A general coefficient of similarity and some of its properties. Biometrics
27(4):857–871.
Griffith, M.B., S.B. Norton, L.C. Alexander, A.I. Pollard, and S.D. Leduc. 2012. The effects
of mountaintop mines and valley fills on the physicochemical quality of stream
ecosystems in the central Appalachians: A review. Science of the Total Environment
417–418:1–12.
Harding, J.M., A.J. Burky, and C.M. Way. 1998. Habitat preferences of the Rainbow
Darter, Etheostoma caeruleum, with regard to microhabitat velocity shelters. Copeia
1998(4):988–997.
Hijmans, R.J., S.E. Cameron J.L. Parra, P.G. Jones, and A. Jarvis. 2005. Very high resolution
interpolated climate surfaces for global land areas. International Journal of Climatology
25(15):1965–1978.
Northeastern Naturalist
520
L.R.F. Hendrick and D.J. McGarvey
2019 Vol. 26, No. 3
Hill, R.A., M.H. Weber, S.G. Leibowitz, A.R. Olsen, and D.J. Thornbrugh. 2016. The
Stream-Catchment (StreamCat) dataset: A database of watershed metrics for the
conterminous United States. Journal of the American Water Resources Association
52(1):120–128.
Hitt, N.P., and D.B. Chambers. 2014. Temporal changes in taxonomic and functional diversity
of fish assemblages downstream from mountaintop mining. Freshwater Science
33(3):915–926.
Hopkins, R.L., and J.C. Roush. 2013. Effects of mountaintop mining on fish distributions
in central Appalachia. Ecology of Freshwater Fish 22(4):578–586.
Januchowski-Hartley, S.R., P.B. McIntyre, M. Diebel, P.J. Doran, D.M. Infante, C. Joseph,
and J.D. Allan. 2013. Restoring aquatic ecosystem connectivity requires expanding
inventories of both dams and road crossings. Frontiers in Ecology and the Environment
11(4):211–217.
Lemly, A.D. 1993. Teratogenic effects of selenium in natural populations of freshwater fish.
Ecotoxicology and Environmental Safety 26(3):181–204.
Lin, H.-Y., A. Bush, S. Linke, H.P. Possingham, and C.J. Brown. 2017. Climate change
decouples marine and freshwater habitats of a threatened migratory fish. Diversity and
Distributions 23(7):751–760.
Lindberg, T.T., E.S. Bernhardt, R. Bier, A.M. Helton, R.B. Merola, A. Vengosh, and R.T. Di
Giulio. 2011. Cumulative impacts of mountaintop mining on an Appalachian watershed.
Proceedings of the National Academy of Sciences 108(52):20929–20934.
Martin, G.M., N. Bellouin, W.J. Collins, I.D. Culverwell, P.R. Halloran, S.C. Hardiman, et
al.. 2011. The HadGEM2 family of Met Office Unified Model climate configurations.
Geoscientific Model Development 4:723–757.
McGarvey, D.J. 2014. Moving beyond species: Discharge relationships to a flow-mediated,
macroecological theory of fish species richness. Freshwater Scie nce 33(1):18–31.
McGarvey, D.J., and B. de Freitas Terra. 2016. Using river discharge to model and deconstruct
the latitudinal diversity gradient for fishes of the Western Hemisphere. Journal of
Biogeography 43(7):1436–1449.
McKay, L., T. Bondelid, T. Dewald, J. Johnston, R. Moore, and A. Rea. 2012. NHD Plus
Version 2: User Guide. Available online at ftp://ftp.horizon-systems.com/NHDplus/
NHDPlusV21/Documentation/NHDPlusV2_User_Guide.pdf.
Merow, C., M.J. Smith, and J.A. Silander. 2013. A practical guide to MaxEnt for modeling
species’ distributions: What it does, and why inputs and settings matter. Ecography
36(10):1058–1069.
Mims, M.C., and J.D. Olden. 2012. Life-history theory predicts fish-assemblage response
to hydrologic regimes. Ecology 93(1):35–45.
Moreno-amat, E., R.G. Mateo, D. Nieto-lugilde, N. Morueta-holme, J. Svenning, and I.
García-amorena. 2015. Impact of model complexity on cross-temporal transferability in
Maxent species distribution models: An assessment using paleobotanical data. Ecological
Modelling 312:308–317.
Nelson, R.L., M.L. McHenry, and W.S. Platts. 1991. Mining. American Fisheries Society
Special Publication 19:425–457.
Nippgen, F., M.R.V. Ross, E.S. Bernhardt, and B.L. McGlynn. 2017. Creating a more perennial
problem? Mountaintop removal coal mining enhances and sustains saline baseflows
of Appalachian watersheds. Environmental Science and Technology 51:8324–8334.
Palmer, M.A., E.S. Bernhardt, W.H. Schlesinger, K.N. Eshleman, E. Foufoula-Georgiou,
M.S. Hendryx, A.D. Lemly, G.E. Likens, O.L. Loucks, M.E. Power, P.S. White, and P.R.
Wilcock. 2010. Mountaintop mining consequences. Science 327:148–149.
Northeastern Naturalist Vol. 26, No. 3
L.R.F. Hendrick and D.J. McGarvey
2019
521
Parmesan, C. 2006. Ecological and evolutionary responses to recent climate change. Annual
Review of Ecology, Evolution, and Systematics 37(1):637–669.
Pecl, G.T., M.B. Araújo, J.D. Bell, J. Blanchard, T.C. Bonebrake, I. Chen, et al.. 2017.
Biodiversity redistribution under climate change: Impacts on ecosystems and human
well-being. Science 31(355):1389.
Petitpierre, B., O. Broennimann, C. Kueffer, C. Daehler, and A. Guisan. 2017. Selecting
predictors to maximize the transferability of species distribution models: Lessons from
cross-continental plant invasions. Global Ecology and Biogeography 26:275–287.
Phillips, S.B., V.P. Aneja, D. Kang, and S.P. Arya. 2006. Maximum entropy modeling of
species geographic distributions. Ecological Modelling 190:231–259.
Phillips, S.J. 2017. A brief tutorial in Maxent. Available online at https://biodiversityinformatics.
amnh.org/open_source/maxent/. Accessed February 2018.
Poff, N.L., and J.D. Allan. 1995. Functional organization of stream fish assemblages in relation
to hydrological variability. Ecology 76(2):606–627.
Poff, N.L., J.D. Allan, M.B. Bain, J.R. Karr, K.L. Prestegaard, B.D. Richter, B.D., R.E.
Sparks and J.C. Stromberg. 1997. The natural flow regime. BioScience 47(11):769–784.
Pond, G.J., M.E. Passmore, F.A. Borsuk, L. Reynolds, and C.J. Rose. 2008. Downstream
effects of mountaintop coal mining: Comparing biological conditions using family- and
genus-level macroinvertebrate bioassessment tools. Journal of the North American Benthological
Society 27(3):717–737.
Power, M.E., A. Sun, G. Parker, W.E. Dietrich, J.T. Wootton, and J.T. Woott. 1995. Hydraulic
food-chain models: An approach to the study of food-web dynamics in large rivers.
BioScience 45(3):159–167.
Raftery, A.E., A. Zimmer, D.M.W. Frierson, R. Startz, and P. Liu. 2017. Less than 2 °C
warming by 2100 unlikely. Nature Climate Change 7(9):637–641.
Renner, I.W., J. Elith, A. Baddeley, W. Fithian, T. Hastie, S.J. Phillips, G. Popovic, and D.I.
Warton. 2015. Point-process models for presence-only analysis. Methods in Ecology
and Evolution 6(4):366–379.
Riahi, K., S. Rao, V. Krey, C. Cho, V. Chirkov, G. Fischer, G. Kindermann, N. Nakicenovic,
and P. Rafaj. 2011. RCP 8.5-A scenario of comparatively high greenhouse gas emissions.
Climatic Change 109(1):33–57.
Ries, R., and S. Perry. 1995. Potential effects of global climate change on Brook Trout
growth and prey consumption in central Appalachian streams, USA. Climate Research
5(3):197–206.
Ross, M.R.V., B.L. McGlynn, and E.S. Bernhardt. 2016. Deep impact: Effects of mountaintop
mining on surface topography, bedrock structure, and downstream waters. Environmental
Science and Technology 50:2064–2074.
Schmidt, G.A., M. Kelley, L. Nazarenko, R. Ruedy, G.L. Russell, I. Aleinov, et al. 2014.
Configuration and assessment of the GISS ModelE2 contributions to the CMIP5 archive.
Journal of Advances in Modeling Earth Systems 6:141–184.
Segura, C., P. Caldwell, G. Sun, S. Mcnulty, and Y. Zhang. 2015. A model to predict stream
water temperature across the conterminous USA. Hydrological Processes 29:2178–2195.
Smith, M.S., L. Horrocks, A. Harvey, and C. Hamilton. 2011. Rethinking adaptation for a
4 °C world. Philosophical Transactions of the Royal Society A: Mathematical, Physical,
and Engineering Sciences 369(1934):196–216.
Stauffer, J.R., Jr., J.M. Boltz, and L.R. White. 1995. The Fishes of West Virginia. Proceedings
of the Academy of Natural Sciences of Philadelphia 146:1–389.
Thomson, A.M., K.V. Calvin, S.J. Smith, G.P. Kyle, A. Volke, P. Patel, S. Delgado-Arias, B.
Bond-Lamberty, M.A. Wise, L.E. Clarke, and J.A. Edmonds. 2011. RCP4.5: A pathway
for stabilization of radiative forcing by 2100. Climatic Change 109(1):77–94.
Northeastern Naturalist
522
L.R.F. Hendrick and D.J. McGarvey
2019 Vol. 26, No. 3
Toth, L.A., D.R. Dudley, J.R. Karr, and O.T. Gorman. 1982. Natural and man-induced
variability in a Silverjaw Minnow (Ericymba buccata) population. American Midland
Naturalist 107(2):284–293.
US Army Corps of Engineers (USACE). 2016. National inventory of dams. Available online
at http://nid.usace.army.mil/cm_apex/f?p=838:12. Accessed March 2018.
Vogel, R., I. Wilson, and C. Daly. 1999. Regional regression models of annual streamflow
for the United States. Journal Of Irrigation and Drainage Engineering 125(3):148–157.
Wallace, D.C. 1972. The ecology of the Silverjaw Minnow, Ericymba buccata Cope.
American Midland Naturalist 87(1):172–190.
Warren, M.L., and M.G. Pardew. 1998. Road crossings as barriers to small-stream fish
movement. Transactions of the American Fisheries Society 127(4):637–644.
Wenger, S.J., D.J. Isaak, C.H. Luce, H.M. Neville, K.D. Fausch, J.B. Dunham, D.C.
Dauwalter, M.K. Young, M.M. Elsner, B.E. Rieman, A.F. Hamlet, and J.E. Williams.
2011. Flow regime, temperature, and biotic interactions drive differential declines of
trout species under climate change. Proceedings of the National Academy of Sciences
of the United States of America 108(34):14175–14180.
Wiley, B.J.B., R.D. Evaldi, J.H. Eychaner, and D.B. Chambers. 2001. Reconnaissance of
stream geomorphology, low streamflow, and stream temperature in the mountaintop
coal-mining region, southern West Virginia, 1999–2000. Water-Resources Investigations
Report 01-4092. US Geological Survey, Charleston, WV. 38 pp.
Woods, T.E., and D.J. McGarvey. 2018. Assessing the relative influences of abiotic and biotic
factors on American Eel, Anguilla rostrata, distribution using hydrologic, physical
habitat, and functional trait data. Ecography 41(12):2067–2079.
Wu, T., L. Song, W. Li, Z. Wang, H. Zhang, X. Xin, Y. Zhang, et al. 2014. An overview of
BCC Climate System Model development and application for climate change studies.
Journal of Meteorological Research 28(1):34–56.
Yates, K.L., P.J. Bouchet, M.J. Caley, K. Mengersen, C.F. Randin, S. Parnell, et al. 2018.
Outstanding challenges in the transferability of ecological models. Trends in Ecology
and Evolution 33(10):790–802.
Yukimoto, S., Y. Adachi, M. Hosaka, T. Sakami, H. Yoshimura, M. Hirabara, et al. 2012.
A new global climate model of the Meteorological Research Institute: MRI-CGCM3—
Model description and basic performance. Journal of the Meteorological Society of
Japan 90A:23–64.
Zegre, N.P., A.J. Miller, A. Maxwell, and S.J. Lamont. 2014. Multiscale analysis of hydrology
in a mountaintop mine-impacted watershed. Journal of American Water Resources
Association 50(5):1257–1272.