Sei sulla pagina 1di 16

Science of the Total Environment 621 (2018) 1208–1223

Contents lists available at ScienceDirect

Science of the Total Environment

journal homepage: www.elsevier.com/locate/scitotenv

Identifying non-reference sites to guide stream restoration and


long-term monitoring☆
Ryan A. McManamay ⁎, John G. Smith, Robert T. Jett, Teresa J. Mathews, Mark J. Peterson
Environmental Sciences Division, Oak Ridge National Laboratory, Oak Ridge, TN 37831, United States

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

• Reference sites have served as a stan-


dard to measure ecosystem restoration
success but do not elucidate impaired
processes
• Comparison of reference sites to bearing
sites, i.e. altered ecosystems, help iden-
tify constraints to effective restoration
• A spatial framework and case study are
used to identify bearing streams that
match biophysical conditions at re-
stored sites
• Out of N13,000 stream sites, a subset of
bearing sites were prioritized for use in
future long-term monitoring
• Bearing streams revealed disrupted pro-
cesses at play in restored sites and ex-
posed limitations of current reference sites

a r t i c l e i n f o a b s t r a c t

Article history: The reference condition paradigm has served as the standard for assessing the outcomes of restoration projects,
Received 15 June 2017 particularly their success in meeting project objectives. One limitation of relying solely on the reference condition
Received in revised form 11 October 2017 in designing and monitoring restoration projects is that reference conditions do not necessarily elucidate impair-
Accepted 12 October 2017
ments to effective restoration, especially diagnosing the causal mechanisms behind unsuccessful outcomes. We
Available online 1 November 2017
provide a spatial framework to select both reference and non-reference streams to guide restoration planning
Editor: Ouyang Wei and long-term monitoring through reliance on anthropogenically altered ecosystems to understand processes
that govern ecosystem biophysical properties and ecosystem responses to restoration practices. We then applied
Keywords: the spatial framework to East Fork Poplar Creek (EFPC), Tennessee (USA), a system receiving 30 years of reme-
Restoration diation and pollution abatement actions from industrialization, pollution, and urbanization. Out of N13,000
Anthropogenic disturbance stream reaches, we identified anywhere from 4 to 48 reaches, depending on the scenario, that could be used in
Contamination restoration planning and monitoring for specific sites. Preliminary comparison of fish species composition at
Stream these sites compared to EFPC sites were used to identify potential mechanisms limiting the ecological recovery
Fish communities
following remediation. We suggest that understanding the relative role of anthropogenic pressures in governing
Landscape alteration
ecosystem responses is required to successful, process-driven restoration.
© 2017 Elsevier B.V. All rights reserved.

☆ This manuscript has been authored by UT-Battelle, LLC under Contract No. DE-AC05-00OR22725 with the U.S. Department of Energy.The United States Government retains and the
publisher, by accepting the article for publication, acknowledges that the United States Government retains a non-exclusive, paid-up, irrevocable, world-wide license to publish or
reproduce the published form of this manuscript, or allow others to do so, for United States Government purposes.The Department of Energy will provide public access to these results
of federally sponsored research in accordance with the DOE Public Access Plan(http://energy.gov/downloads/doe-public-access-plan).
⁎ Corresponding author at: Environmental Sciences Division, Oak Ridge National Laboratory, Bldg 1504, MS-6351, Oak Ridge, TN 37831-6351, United States.
E-mail address: mcmanamayra@ornl.gov (R.A. McManamay).

https://doi.org/10.1016/j.scitotenv.2017.10.107
0048-9697/© 2017 Elsevier B.V. All rights reserved.
R.A. McManamay et al. / Science of the Total Environment 621 (2018) 1208–1223 1209

1. Introduction geospatial approaches to characterize landscape conditions and biolog-


ical survey information to identify best representatives of reference con-
The reference condition is considered by many as the standard or ditions within a region. Developing reference standard criteria is
benchmark for ecological assessments, including measuring the out- completely appropriate for regional biological assessments, i.e. Indica-
comes of restoration projects (Stoddard et al., 2006; Hawkins et al., tors of Biological Integrity (Ode et al., 2016); however, these criteria
2010). Clearly defined goals are essential to objectively guiding and are typically represented by hundreds of reference sites where biologi-
measuring the success of restoration actions (Tear et al., 2005). Natural- cal information is available (e.g., Lunde et al., 2013) and not necessarily
ly, conservation goals and the metrics by which they are evaluated are selected based on the specific physiochemical properties or anthropo-
typically based upon the selection of reference sites and the characteris- genic disturbance regimes that approximate conditions at individual
tics they possess. Even so, there continues to be disagreement on how restoration sites. Indeed, restoration practices require very specific
best to evaluate a restoration project's success or failure (e.g., Palmer guidance to set realistic expectations and appropriate actions (Suding,
et al., 2005; Bernhardt et al., 2007; Brewer and Menzel, 2009; Morandi 2011). Additionally, we suggest that non-reference sites help provide
et al., 2014; Suding, 2011). some of this guidance. Programs that monitor the effectiveness of resto-
To aid in the process of reference selection, Stoddard et al. (2006) ration may arbitrarily select reference sites that either match pre-con-
recommended some standard terminology for use in defining reference ceived preferences or are convenient (e.g., easily sampled, accessible,
conditions. Stoddard et al. (2006) urge that the term “reference” be pre- part of existing sampling regimes, etc). For example, the justification
served for systems absent of human disturbance and then suggest terms for selecting restoration comparison sites (reference or non-reference)
for several alternative conditions, i.e. “minimally disturbed condition”, is typically based on access or priority from social emphasis (e.g.,
“historical condition”, “least disturbed condition”, and “best attainable Violin et al., 2011; Kondolf et al., 2008). To avoid human-generated
condition”. While their terminology has been adapted by some, what biases, a comprehensive geographic approach to selecting reference
constitutes an appropriate reference site remains an important yet un- and non-reference sites (i.e., bearing streams) is recommended, espe-
settled issue, as evidenced by the focus of several recent major reviews cially not limiting the selection to only sites biologically sampled.
(e.g., Dallas, 2013; Hawkins et al., 2010; Morandi et al., 2014) and re- Herein, we present a spatial framework as a 1st-step screening
search efforts (e.g., Feio et al., 2013; Kosnicki et al., 2014; Ode et al., criteria to identify bearing streams (both reference and non-reference
2016). The debate over defining “reference condition” stems, in part, condition) to guide stream restoration. The framework is structured as
from many regions across the globe having little or no populations of a step-wise procedure that requires managers clearly define and revisit
undisturbed sites for adequate representation in biological integrity goals and objectives of restoration while also evaluating all potential
analyses (Brewer and Menzel, 2009; Dallas, 2013; Feio et al., 2013; streams that could approximate those goals. First, we provide a brief
Kosnicki et al., 2014). overview of the spatial framework and its elements. Then we apply
Regardless of how reference streams are defined, the main precept the framework to a case study of East Fork Poplar Creek, the focus of a
of the reference stream paradigm suggests that context is necessary to 30-year pollution remediation and biological monitoring program.
assess biophysical conditions at non-reference sites and interpret Multi-dimensional impacts from large-scale chemical contamination
what those values mean relative to a given standard (Hawkins et al., and urbanization and the sheer complexity of the ecosystem present
2010). Reference conditions, however, are not the only benchmark an ideal situation requiring the additional context of non-reference
that provides context in interpreting the outcomes of ecological assess- streams to isolate impaired physical and ecological mechanisms still at
ments (Brewer and Menzel, 2009). Streams of varying disturbance re- play in the stream.
gimes, and the ecological communities they support, also provide
additional dimensions in assessing the ecological status of a restoration 2. Spatial framework
project and understanding the nature of disrupted mechanisms (or pro-
cesses) influencing ecological patterns. We term these systems “bearing The proposed framework includes six main elements, which can be
streams”, as they provide a point of orientation that bears witness to the used to incrementally screen and reduce the potential number of bear-
desired endpoint. “Bearing trees”, as defined by the US Department of ing streams (Fig. 1). The elements are collectively drawn from our re-
Agriculture Forest Service, do not mark the true property corner, but view of the literature related to: developing candidate reference sites
bears witness to that location (USFS, 2013).In other words, they serve (e.g., Hughes et al., 1986), lack of suitable reference sites (e.g., Dallas,
as a bearing and distance to the true property corner. 2013), understanding watershed disturbances to prioritize restoration
Bearing streams provide a point-of-reference rather than serve as efforts (e.g., Beechie et al., 2008), using typologies to infer reference
the reference condition, as defined by Stoddard et al. (2006). Identifying conditions (Hawkins et al., 2010), evaluating local and landscape con-
both reference and disturbed streams in post-restoration monitoring straints to assess ecosystem recovery following restoration (e.g.,
(e.g., Violin et al., 2011) help to elucidate the impaired processes, and Suding, 2011), principles of successful restoration outcomes (Palmer
by doing so, they assist in appropriately setting goals and guiding restora- et al., 2005), and adaptive management (Murcia and Aronson, 2014).
tion practices to achieve a given endpoint (Roni et al., 2008). In many Spatial extent, disturbance regime, physical template, and ecological
cases, restoration occurs in areas of intense cumulative disturbance aris- community (all described below) are more essential elements in identi-
ing from multiple independent and interacting agents, all of which act fying potential reference sites, whereas, the remaining elements (previ-
upon the physiochemical and biological processes that shape the ecolog- ous sampling and accessibility) are more a matter of preference (Fig. 1).
ical community. Teasing apart the relative roles of each disturbance agent An important point to illuminate, however, is that the order of steps or
is a difficult, but necessary, step in guiding restoration practice; however, elements may not necessarily reflect the priority of restoration goals.
elucidating roles of disturbance is unlikely via direct comparison with ref- For instance, goals for restoration commonly have an ecological focus
erence sites lacking any disturbance. In situations where some distur- and include increasing species richness and population recruitment, re-
bances can be ameliorated via restoration while others will remain instating ecosystem services, or establishing unrepresented functional
active agents, selecting bearing streams impacted by individual distur- groups, all of which are pertinent. While these goals most closely align
bances may be advantageous in understanding disturbance-mediated with the ecological community element (step 3, Fig. 1), ecological con-
processes, which only become evident after long-term monitoring. ditions are contingent upon the previous three steps. In particular, the
A great deal of effort has been expended in identifying regional pools anthropogenic disturbance regime and physical characteristics can
of reference streams to provide standards or benchmarks for biological serve as surrogates for ecosystem state and structure, respectively,
assessments (Hughes et al., 1986; Whittier et al., 2007; Dallas, 2013: with each exerting progressively less control on subsequent elements.
Lunde et al., 2013; Ode et al., 2016).Many of these studies utilize While the selection of bearing streams should be aligned with clearly-
1210 R.A. McManamay et al. / Science of the Total Environment 621 (2018) 1208–1223

shifts attention away from static endpoints to a mechanistic restoration


of processes that control those endpoints.Because restoration is aimed
at improving biophysical conditions, typical reference sites are those
that have little to no anthropogenic influence, as this is representative
of the end-goal. This gives the term “reference” a connotation of
“good” or “better”, as opposed to serving its true definition as a “base-
line” or “point of orientation”. This does not insinuate that reference
site selection should avoid pristine areas, but that the selection of refer-
ence sites need not always be confined to low-disturbance conditions.
Identifying streams of varying individual disturbance levels can be
achieved at a high-level by screening sites based on landscape-level an-
thropogenic disturbance regimes or potential chemical disturbances
(Supplementary material 1).

2.3. Physical template

The physical template determines the process by which disturbance


regimes influence ecological communities. Ideally, bearing streams will
be selected based on possessing similar (or desired) physical attributes
to that of the restoration project.Determining the physical properties of
streams, however, is no small task, typically requiring intensive field in-
vestigation. While this is especially true in cases of gathering detailed
measurements to properly design stream channels, this amount of
work is unrealistic for screening many potential candidates.Large-scale
stream classification systems provide an inventory of different physical
types of streams for a given region.For example, Olivero Sheldon et al.
(2015) developed a 6-layer stream classification for NHD stream
reaches for the Mid-Appalachian Region of the United States, which in-
cludes the Tennessee River Basin (Fig. 2). The six layers included stream
size, gradient, temperature (mean summer), hydrology, buffering ca-
pacity, and confinement (Fig. 2, for a description see Supplementary
material). The advantage of layered approaches is that investigators
can select any number of physical properties that have relevance to
the restoration objectives of interest. Additionally, each of the stream
layers represent reference conditions as they were developed based
on natural stream reach attributes (e.g., size, gradient) or were modeled
from empirical observations collected from streams with minimal an-
thropogenic disturbances (Olivero Sheldon et al., 2015). Thus, stream
classifications provide a template to establish the reference condition
for streams undergoing restoration (Hawkins et al., 2010) while also
Fig. 1. Six elements of the proposed spatial framework to incrementally screen potential
reference or bearing streams to meet desired conditions. The first four elements are providing a framework to select surrounding streams with similar phys-
essential whereas the last two are a matter of preference. ical and chemical properties.

2.4. Ecological condition


defined conservation objectives to guide and measure the success of
restoration actions (Tear et al., 2005), the current state, structure, and As mentioned previously, objectives for restoration are commonly
ecological composition of an ecosystem reflects past and present pertur- ecologically related. Ecological objectives may include increasing popu-
bations (Beisner et al., 2003). In this case, the disturbance regime is the lation size or habitat for a species, increasing the species richness of
most import determinant of ecosystem state. entire communities, and/or establishing functional groups, functional
processes, and ecosystem services (Bernhardt et al., 2005). Past ap-
2.1. Spatial extent proaches to identifying pools of reference sites excluded streams not
previously sampled for ecological communities (e.g., Feio et al., 2013;
The spatial extent determines the population of streams from which Lunde et al., 2013; Ode et al., 2016). Given that b5% of the nation's
potential bearing streams are selected. Ecoregions provide an appropri- area has been systematically surveyed for aquatic taxa (Troia and
ate and convenient geographic context to organize streams because wa- McManamay, 2016), this screening criteria would exclude the majority
tersheds within these areas have similar landforms, soil, vegetation, of streams from further consideration. To consider all potential streams
land use, and ecological communities (Omernik, 1987; Hughes et al., as candidates, modeling estimated ecological conditions using species
1986).Even still, watersheds within ecoregions tend to have their own distribution models (Huang and Frimpong, 2015) or estimating ecolog-
chemical, physical, and biological signatures (Hughes et al., 1986) and ical processes within unsampled stream reaches may be required
may provide an additional template to determine a region of potential (Kosnicki et al., 2014).This modeling process should not be done in iso-
stream candidates. lation from the physical template. One mechanism for ensuring ecolog-
ical conditions are ‘physically relevant’ is to use the physical template as
2.2. Anthropogenic disturbance regime predictors for ecological models.Based on pre-determined objectives,
streams that meet the ecological criteria can be selected as candidates;
Placing the immediate focus of reference site identification on the however, this is a challenging exercise and requires expert input. For in-
anthropogenic disturbance regime rather than desired future condition stance, investigators must identify specific species, functional groups,
R.A. McManamay et al. / Science of the Total Environment 621 (2018) 1208–1223 1211

Fig. 2. Stream classification layers used to characterize the natural physical template of streams in the Tennessee River Basin, USA.The six-layer stream classification was produced by
Olivero Sheldon et al. (2015). Buffering capacity is not shown. Descriptions of stream classes are provided in the Supplementary material.

guilds, or ecological processes as targets for restoration. Bearing candi- by foot has to also be taken into consideration, such as determining
dates are then selected that best approximate those targets. road crossings or land ownership.
Additionally, many ecological processes require several years to de-
2.5. Previous sampling and accessibility cades to fully capture; hence, determining the success of restoration ef-
forts requires long-term monitoring and consideration of the extent and
When planning a monitoring regime, the logistics of including addi- nature of previous sampling. However, as noted above, selecting only
tional monitoring efforts will likely be a factor included in prioritizing sites previously sampled for ecological communities, especially under
sites, especially if resources are limited. Preferences are usually shown rigorous sampling procedures, will limit the pool of candidate streams.
for sites that are easily accessible, which considers proximity to a Thus, we re-emphasize that such criteria should be optional. Studies
home base and ease of access. Driving times can be estimated from evaluating the effectiveness of restoration have shown the importance
road network analysis to determine the feasibility of sampling various of proper control designs in detecting changes (Gowan and Fausch,
locations based on their distance. The potential to access the stream 1996) as well as the effect of study duration on findings (White et al.,
1212 R.A. McManamay et al. / Science of the Total Environment 621 (2018) 1208–1223

2011). Of the many experimental designs, the control-treatment pairing remediation of the floodplain in lower EPFC, and a bank stabilization
(CTP) design (Eberhardt, 1976) and subsequent Before-After-Control- project to reduce erosion of mercury-contaminated soils (Loar et al.,
Impact (BACI) and paired BACI (BACIP) designs (Stewart-Oaten et al., 2011).For a more comprehensive review of the impacts and remedia-
1986) have been considered among the most robust sampling strategies tion to the system, please see Loar et al. (2011) and Brooks and
to evaluate restoration effectiveness, especially if multiple controls are Southworth (2011).
used (Underwood, 1994). However, anticipating the implementation As a condition of the NPDES permit, the Biological Monitoring and
of restoration practices years in advance and adequately preparing an Abatement Program (BMAP) was initiated in 1985 to evaluate and doc-
appropriate experimental design is typically an exception, not generally ument the ecological effects of implementing a Water Pollution Control
a reality. Presumably, this may be a result of limited resources. Prioritiz- Program and remediation actions (Peterson, 2011). One of the tasks in-
ing bearing streams based on pre-existing sampling would make full cluded in the BMAP to help fulfill the requirements of the program was
use of existing resources and assist in preparing an analysis. Obviously, monitoring the fish community in the stream. Five fish community sam-
the type of sampling regimen should be considered as more information pling sites were established at locations immediately below the Y-12
will be available as discrete singular sampling events as opposed to sys- Complex and extending 19 km downstream (Ryon, 2011) (Fig. 3).
tematic sampling with repeated surveys. Drainage areas for sampling sites ranged from 1.6 to 49 km2. Two near-
by reference streams, Brushy Fork (40 km2) and Hinds Creek (45 km2),
3. Case study were selected as representative of conditions in EFPC in the absence of
the Y-12 Complex and industrial effluents (Ryon, 2011). Topographic
We applied the spatial framework to identify potential bearing maps were scanned to identify a pool of potential sites, and then a re-
streams for East Fork Poplar Creek (EFPC) Tennessee, a system impacted connaissance survey of each prospective site was completed to deter-
by substantial chemical contamination, channel alteration, and urbani- mine each site's suitability. These reference streams did not represent
zation followed by intense remediation and pollution abatement efforts. pristine conditions, but most closely approximated the “historical con-
For over 30 years, EFPC has been the focus of large-scale remediation ef- dition” (Stoddard et al., 2006). The sites are indicative of rural-agricul-
forts and a long-term biological monitoring program (Peterson, 2011; ture influenced systems typical of the Ridge and Valley Ecoregion and
Loar et al., 2011), a sampling regimen that included reference streams provide an estimate of likely conditions had the Y-12 Complex not
as comparisons to impacted sites. Given that reference sites were select- been built.
ed three decades ago based on expert judgement, we questioned
whether a spatial approach might yield comparable or different results. 3.2. Applying the spatial framework

3.1. Study site and background We used three scenarios to identify potential bearing stream candi-
dates for each of the five EFPC fish community sites. The first scenario
East Fork Poplar Creek originates within the U.S. Department of (S1) selected candidate streams impacted by agricultural land use sim-
Energy's Y-12 National Security Complex in eastern Tennessee, USA ilar to the criteria used for the current set of “reference sites”. This pro-
(Fig. 3). EFPC drains a 55 km2 watershed before joining Bear Creek vides a comparison of the degree to which existing reference sites
and eventually flowing to Poplar Creek, an arm of Watt's Bar Reservoir match their intended purpose. In contrast to S1, the second scenario
on the Clinch River System (Fig. 3). After exiting the Y-12 Complex (S2) selected sites experiencing urbanization, as to provide a baseline
boundary ~2 km downstream of its origin, EFPC flows ~15 km through comparison that isolates the conflicting effects of contaminant amelio-
the city of Oak Ridge before reentering the Department of Energy's Oak ration from the remaining urban impacts. Additionally, S1 and S2
Ridge Reservation where it flows an additional 8 km before its conflu- sought sites having a fish species assemblage very similar to the desired
ence with Poplar Creek. The EFPC watershed is highly urbanized, with EFPC ecological community and that were located within a 45 min drive
developments exceeding 80% developed land cover in the upper (~55 km radius) from home base (Oak Ridge National Laboratory). Fi-
reaches and 45% in lower reaches. nally, the third scenario (S3) selected “least disturbed” streams having
As part of the Manhattan Project, the Y-12 Complex was constructed little anthropogenic disturbance and/or falling within protected areas
in 1943 to produce uranium and help fabricate nuclear weapons (Loar and, to some extent, relaxed the criteria for similarity in desired species
et al., 2011). Because the Y-12 Complex is situated in the headwaters composition. S3 also increased the acceptable driving time to include
of EFPC, operations have released significant contaminant loads to the potential candidate streams within a 90 min drive (~ 110 km radius)
stream.In 1985, at least 250 different effluent and storm water outfalls from home base. We then compared bearing streams identified through
discharged into the headwaters of EFPC. Effluent discharges such as the S1–3 scenarios to the existing reference streams.
cooling tower blowdown and process wastewaters contributed signifi-
cant loads of chlorine to the stream, while other contaminants entered 3.2.1. Spatial extent
EFPC through erosion and surface runoff (Kasten, 1986; Loar et al., EFPC is located on the edge of the Ridge and Valley (RV) Ecosystem
2011). Byproducts of operations in wastewaters included substantial (in proximity to the Central Appalachian (CA) and Southwestern Appa-
quantities of heavy metals, organic solvents, polychlorinated biphenyls lachian (SWA) Ecoregions (Fig. 4A). Fish assemblages within EFPC are
(PCBs), and mercury. It has been estimated that N 100,000 kg of mercury influenced by neighboring ecoregions as evidenced by occasional detec-
was released into the stream causing extensive contamination of the tion of species that are hypothesized to immigrate from the CA and SWA
surface water, sediments, and floodplain soils (Loar et al., 2011). (closest known source populations for those species occur in CA and
Over a two-decade period beginning in 1985, several major remedial SWA). We considered the spatial extent to be an area bounding the in-
actions, pollution abatement projects, and operational changes were tersection of the RV, CA, and SWA Ecoregions within the middle water-
completed at the Y-12 Complex under several federal regulations in- sheds of the greater Tennessee River Basin (Fig. 4A). Potential reference
cluding a renewed National Pollutant Discharge and Elimination System streams were spatially represented as stream reach segments within
permit (Clean Water Act), the Resource Conservation and Recovery Act, the NHDplus V2 dataset (HSC, 2017). This area provided 13,519
and the Superfund program established by the Comprehensive Environ- NHDplus stream reaches that serve as the population of potential refer-
mental Response, Compensation, and Liability Act (CERCLA). Some of ence stream candidates.
the major remedial actions included construction of a Central Pollution
Control Facility, relining drains from sanitary and storm sewers, dechlo- 3.2.2. Anthropogenic disturbance regime
rination of cooling water discharges, implementation of flow manage- Disturbance variables were summarized for local stream reach
ment program, construction of a mercury treatment system, mercury catchments and for the entire upstream network contributing to each
R.A. McManamay et al. / Science of the Total Environment 621 (2018) 1208–1223 1213

Fig. 3. Study site map of East Fork Poplar Creek (EFPC) within the Tennessee River Basin, USA (upper left panel). The EFPC watershed (grey region) is shown in relation to Department of
Energy sites, including the Y-12 National Security Complex (Y12), Oak Ridge National Laboratory (ORNL), and the East Tennessee Technology Park (ETTP) (top left panel). The five EFPC
study sites occur within four different natural stream class typologies (combination of stream size, gradient, hydrology, buffering capacity, and valley confinement).

local stream reach based on the National Fish Habitat Partnership et al. (2009) found that invertebrate communities showed significant
(NFHP) (Esselman et al., 2013; NFHP, 2017). Examples of variables shifts beyond 25% upstream agriculture within the Southeastern Coastal
within the NFHP included urban and agricultural land cover (e.g., Sup- plain whereas Midway et al. (2015) found no negative relationships be-
plementary material), dams, road crossings, and National Pollution Dis- tween fish richness and agriculture within the same region. In concur-
charge Elimination System permits, among several other attributes rence with Utz et al. (2009), we used 25% upstream agriculture as a
(Esselman et al., 2013). These variables were used to create a cumula- threshold for S1 (% agriculture in streams depicted in Fig. 4E). As men-
tive disturbance index, which provides a surrogate of landscape alter- tioned previously, EFPC's watershed ranges from 45 to 80% developed
ation risk to fish habitats (e.g., Supplementary material, Esselman et land cover. Based on several studies, ecological degradation has been
al., 2013; NFHP, 2017). To determine the protected status of each stream observed with as little as 10% upstream urbanization (King et al.,
reach, we summarized the percentage of riparian floodplain area classi- 2005), but consistently observed beyond 25% upstream urbanization
fied as protected land (Fig. 4D). Floodplain areas were delineated using (Paul and Meyer, 2001; Walsh et al., 2005; Utz et al., 2009). For S2, we
the Valley Confinement Algorithm (Nagel et al., 2014), which identifies conservatively selected all streams having at least 25% developed land
floodplains as areas inundated at 5 × bankfull width height (Fig. 4D). cover in their upstream watersheds (% urbanization in streams depicted
Protected land coverages were available through the Protected Area in Fig. 4C). For S3, streams deemed “least disturbed condition” were se-
Database (USGS, 2016). lected that had “very low” or “low” NFHP cumulative disturbance indi-
Identifying potential candidate bearing streams that isolate effects of ces and/or had at least 50% of their riparian floodplain overlapping
individual disturbances is challenging, as this requires establishing suit- protected conservation lands (see Fig. 4D).
able thresholds. Thresholds for agriculture land cover have been docu-
mented far less than those for urbanization. Burcher and Benfield 3.2.3. Physical template
(2006) found that agriculturally impaired streams with an average of To characterize physical stream types, we used the 6-layer stream
35% agricultural land cover in their upstream watersheds had signifi- classification developed by Olivero Sheldon et al. (2015). Among the
cantly less fish species richness than comparison sites. However, Utz 13,519 stream reaches within the defined spatial extent, all six physical
1214 R.A. McManamay et al. / Science of the Total Environment 621 (2018) 1208–1223

Fig. 4. Examples of geospatial data layers representing different elements of the spatial framework applied to screening candidate bearing streams for East Fork Poplar Creek. A) The spatial
extent used to determine the population of candidate bearing streams included three Ecoregions within watersheds of the middle Tennessee River Basin. B) The diversity of simplified
stream class typologies (47 unique types) among 13,519 stream reaches were used to screen potential bearing stream candidates. C) The percentage of developed lands within the
cumulative upstream area of each stream reach was used in the urban-impact scenario, S2. D) The percentage of protected lands within each stream reach's floodplain was used to
screen candidate sites in the least disturbed condition scenario, S3. E) The percentage of agricultural lands within the cumulative upstream area of each stream reach was used in the
agriculture-impact scenario, S1. F) The distribution and type of survey of locations of previous fish sampling. An inset shows the EFPC sites relative to current reference sites (Hinds
Creek, HCK20.6, and Brushy Fork, BFK7.6).

layers collectively yielded 767 unique stream type combinations. manageable number of typologies to identify potential reference
Because the number of physical typologies was overtly complex to candidates (Fig. 4B). Interestingly, both the complex and simplified
inform meaningful decision making, Olivero Sheldon et al. (2015) stream classification schemes yield four different types of streams for
simplified stream types by consolidating the size, gradient, hydrology, the EFPC sites (Fig. 3). For each EFPC site, we selected streams that
and temperature classes and excluding confinement and buffering matched their respective simplified stream types under all three
layers. This yielded 47 unique stream types, which provide a more scenarios.
R.A. McManamay et al. / Science of the Total Environment 621 (2018) 1208–1223 1215

3.2.4. Ecological condition


For each EFPC site, we generated lists of resident and infrequent fish
species, along with fish hypothesized to historically occur at that site but
are currently absent (Supplementary material 1). All fish species detect-
ed during the last 2 years (2 sampling periods per year) were consid-
ered resident whereas all other species detected over the course of all
sampling events were considered infrequent. List of hypothesized his-
toric residents were generated according to current and historic occur-
rence records from the greater Oak Ridge Reservation (Ryon and Loar,
1988; Krumholz, 1954) and occurrence records in the proximate region
and species habitat descriptions (Etnier and Starnes, 1993). The resul-
tant list of all resident, intermittent, and historic resident species across
all EFPC sites totaled 50 fish (Supplementary material 1).
The objective was to select bearing sites that best approximate the
full list of potential fish species for each site; however, only a subsample
of streams have been sampled. Furthermore, sampling records vary
considerably from discrete single-species occurrences to routinely sur-
veyed community records. For all stream reaches to be considered as
potential bearing site candidates, the probability of a species presence
must be estimated using species distribution models (SDMs) (Fig. 5).
Occurrence records were compiled from multiple open-access reposito-
ries for the region of interest (Troia and McManamay, 2016). All records
were summarized by NHDplus stream reaches. Because of widely vary-
ing sampling methods, detection probabilities based on survey methods
were estimated for each species occurrence within a stream reach. Oc-
currence records within each stream reach were classified as single-spe-
cies occurrence (SS), community survey (CS), or a combination of both
(SCS). All stream reach occurrences were then summarized into 12-
digit hydrologic unit (HUC12) watersheds, which typically contain mul-
tiple NHDplus reaches. Species detection probabilities were estimated
for SS, CS, and SCS survey types based on occurrence in HUC12 water-
sheds using survey-specific patch occupancy modeling in PRESENCE
2.0 (USGS, 2017). Patch occupancy models included survey type and
the number of occurrences according to each survey type (survey covar-
iate). PRESENCE software is flexible in that not all surveys have to have
been conducted within all HUC12 watersheds in order for the models to
estimate patch occupancy and detection probabilities. Survey-specific
detection probabilities, once estimated, were then back applied to the
stream reach scale based on the type of survey that was conducted
within that reach. Detection probabilities were then included as vari- Fig. 5. An example of mapping desired ecological conditions in stream reaches. Species
ables in SDMs to control for imperfect detections. SDMs were construct- distribution models, such as that for snubnose darter (Etheostoma simotreum), were
ed for each of the 50 fish species using their detection probabilities and developed for 50 species. Lists of potential and desired species were compiled for each
105 landscape-scale variables within boosted regression trees (Elith et of the five East Fork Poplar Creek (EFPC) sites and transformed into binary matrices of
species presence-absences.Bray-Curtis dissimilarity values were calculated from the
al., 2008) to model species probability of occurrence within NHDplus comparison of EFPC species-presence-absence matrices and the matrix of species-
stream reaches (Fig. 5) as binomial distributions in the Generalized occurrence probabilities from SDMs. Streams with low dissimilarity values suggest
Boosted Regression Model (gbm) package in the R programming envi- predicted fish communities most closely approximate desired ecological conditions.
ronment (Ridgeway, 2017). Variables used within SDMs are provided
in the Supplementary material. as such. In all scenarios, we provided binary indications of whether
Species occurrence probabilities for all stream reaches and lists of stream reaches have been previously sampled, had access from a road,
presence/absences of all potential species within EFPC sites were orga- or were located within a given driving time from home base. Previous
nized in a stream reach-by-species matrix. Non-metric multi-dimen- sampling was summarized by reach as either single-species occurrences
sional scaling (NMDS) based on Bray-Curtis dissimilarities was used to or community surveys (Fig. 4F). Road access was determined if a road
ordinate stream reaches in multivariate space based on community crossed a given stream reach.
composition. Pairwise dissimilarity values from each EFPC site to all To determine driving distances (and estimated driving time), we
stream reaches were extracted from the dissimilarity matrix (Fig. 5). compiled all primary and secondary roads for counties intersecting
Dissimilarity vectors for each EFPC site were partitioned into five cate- the spatial extent (Fig. 6). All roads were then used to build a geometric
gories of incrementally increasing values based on the Jenks natural network of edges (road lines) and their ends or intersections with other
breaks classification method (Jenks, 1967) (Fig. 5). For the S1 and S2 roads (nodes). Our location at Oak Ridge National Laboratory was used
scenarios, reaches with dissimilarities falling within the lowest Jenks as the starting point, each stream reach's centroid was used as the des-
category were selected. For the S3 scenario, we relaxed the selection tination. Using the Origin-Destination Cost Matrix tool in Arc GIS (10.2),
of community similarity to include reaches falling within the 1st and we calculated pairwise road network distances between the starting
2nd categories (lowest and second-to-lowest values). point and all stream reach destinations (Fig. 6). Assuming an average
speed of 45 mile h−1, we then calculated travel times to all stream
3.2.5. Previous sampling and accessibility reaches. For S1–S2, we identified stream reaches falling within a
Both previous sampling and accessibility are considered preferential 45-min driving radius (55 km) whereas for S3, we expanded the accept-
attributes of potential bearing stream candidates and thus, were treated able range to a 90-min driving radius (110 km).
1216 R.A. McManamay et al. / Science of the Total Environment 621 (2018) 1208–1223

3.2.6. Bearing stream candidates such as environmental filters or anthropogenic constraints, influencing
Locations of bearing stream candidates and their characteristics are the composition of fish communities. We organized species by fish life
provided in an interactive map (Supplementary material, Interactive histories, which represent a continuum of tradeoffs among parental
Map). Depending on the EFPC site, the S1 scenario identified anywhere care, fecundity, spawning periodicity, body size, and longevity
from 9 to 48 potential bearing stream candidates that met the criteria (Winemiller and Rose, 1992). Fish life-history groups represent three
for disturbance regime, physical template, and ecological condition divergent strategies: equilibrium, opportunistic, and periodic species.
(Appendix 1). In contrast, S2 identified fewer candidates (2 to 16) as Equilibrium species are small-bodied, have small clutches of eggs, and
did S3 (10 to 28 reaches), depending on EFPC site (Appendix 1). This have high parental care (e.g., Banded sculpin, Cottus carolinae). Oppor-
makes sense as agriculture is wide-spread through the region. Candi- tunistic fishes are also typically small-bodied fish that mature early
date bearing streams were more proximate to EFPC and displayed and have little parental investment, moderate fecundity, and multiple
more spatial clustering (less spatial variation) in the S2 scenario com- spawning bouts (e.g., Spotfin shiner, Cyprinella spiloptera). Periodic
pared to the S1 and S3 scenarios (Appendix 1) (note that driving dis- species are large fish with late maturation, large clutch sizes, and
tance was not included in screening sites, but as an optional variable). low parental investment (e.g., Black redhorse, Moxostoma duquesnei).
In all three scenarios, bearing stream candidates varied according to All fish species under consideration were assigned to one of the three
whether they were previously sampled, had road access, or occurred life history groups (McManamay and Frimpong, 2015, Supplementary
within the desired driving-time window (Appendix 1). Had we also material 1).
used these attributes to screen potential sites, the selection of candidate We evaluated trends (1985–2015) in fish species richness according
reaches would have been more restrictive and would have resulted in to life history strategies at all EFPC sites. We then compared these values
far less, if any, reaches for some sites. to patterns in fish richness at current reference sites and potential bear-
None of the bearing stream candidates were identical to the individ- ing sites under the three scenarios. At candidate sites, fish richness
ual stream reaches currently used as reference streams (Brushy Fork within each life history group was estimated using the sum of species
and Hinds Creek); however, bearing stream candidates were identified occupancy probabilities from SDMs. This method reduces the need to
within stream reaches located within those same river systems, at times define thresholds for species presence while accounting for uncertainty
directly upstream and downstream of the reference reaches. For exam- in species occurrence, but also avoids overpredicting richness (Dubuis
ple, of the bearing stream candidates identified in the agriculture et al., 2011; Calabrese et al., 2014).
scenario (S1) for EFK24.4, EFK23.4, and EFK6.3 sites, 48%, 100%, and Fish richness increased in an upstream-to-downstream direction
4% of those sites were located on the Hinds Creek or Brushy Fork and increased with time at all EFPC sites, except EFK 24.4 (Fig. 7). Typi-
river systems. Additionally, one site on Brushy Fork was found as a cally, bearing stream candidates had higher total fish richness than re-
bearing stream candidate for EFK24.4 for the least disturbance (S3) spective EFPC sites, specifically opportunistic and equilibrium life
scenario.None of the bearing stream candidates for the urban-impacted histories at EFK 6.3 and 13.8 and all life history strategies at remaining
scenario (S2) overlapped with existing reference stream reaches. EFPC sites. No periodic species have been detected at EFK 24.4 and op-
portunistic species have not been collected since 1988 at this site.
3.3. Hypothesized mechanisms of disturbance There were cases in which richness was lower at some candidate sites
than EFPC sites within a given scenario; however, this primarily oc-
Further consideration of potential bearing sites may benefit from a curred at EFK 6.3 and EFK 13.8. Interestingly, candidate sites within
comparison of ecological conditions at restoration sites to that of poten- the S3 scenario (least disturbed condition) had relatively low values
tial stream candidates. Additionally, evaluating ecological conditions for richness. Existing reference streams had similar richness values to
through a trait lens can elucidate potential underlying mechanisms, that of bearing stream candidates.

Fig. 6. Distances were calculated from the start position (home base) to all possible stream reach destinations using a road network. Examples of distances for two random stream reaches
(Site A and B) are provided.
R.A. McManamay et al. / Science of the Total Environment 621 (2018) 1208–1223 1217

4. Discussion streams to guide restoration actions at a given location via long-term


monitoring (Supplementary material, Interactive Map). These reaches
Our spatial framework identified both low-disturbance and non-ref- could further be refined based on other criteria, such as previous sam-
erence sites to be used as comparisons to restoration sites in order to pling, accessibility, and proximity. This level of specificity is needed to
guide restoration actions or long-term monitoring. Rather than develop appropriately guide restoration, adaptive management, and long-term
a substantive pool of reference sites within a region (e.g., Ode et al., monitoring efforts.
2016), our framework narrowed the selection of potential bearing There are two general approaches for selecting reference or non-ref-
streams down to a handful of sites that mirrored well-defined end- erence sites that are important to distinguish from our framework. As
points of disturbance regimes, physiochemical properties, and ecologi- mentioned previously, the first approach develops a pool of reference
cal conditions. For example, out of 13,519 potential stream candidates, sites to provide regional standards, primarily water quality criteria
we identified anywhere from 2 to 48 reaches that could serve as bearing (e.g., Hughes et al., 1986; Ode et al., 2016). In recent years, the vast

Fig. 7. Comparison of fish richness within life history groups at East Fork Poplar Creek (EFPC) sites, candidate bearing streams under the three scenarios, and at existing reference sites.
Richness values for EFPC sites and existing reference sites represent maximum species detected for each year (1986–2015).Richness values for candidate bearing streams represent
values for multiple independent streams. A richness value for each stream was calculated as the sum of species occupancy probabilities for that stream reach.
1218 R.A. McManamay et al. / Science of the Total Environment 621 (2018) 1208–1223

majority of these studies have used Geographic Information Systems and hydrologic regimes ensure pools of lead disturbed sites had repre-
(GIS) to characterize the natural and anthropogenic characteristics of sentation across river types. Likewise, Harris (1999) used cluster analy-
sites with existing biological monitoring data (e.g., Whittier et al., sis to group riparian vegetation within different floodplain landforms as
2007; Feio et al. 2013; Ode et al., 2016). While the reference-pool ap- a strategy to identify context-specific reference riparian communities.
proach can identify a large number of sites representative of the natural Hawkins et al. (2010) suggests that classifications or typologies based
environmental gradients present within a region (Ode et al., 2016), this on natural patterns provide an estimation of reference conditions as a
approach only yields a temporal snapshot of biophysical conditions. benchmark for ecological assessments. The advantage of developing
Furthermore, the biological criteria developed from reference pools classifications based on natural patterns is that these can be easily ex-
may be influenced by various sampling regimes with variable levels of tended to infer reference conditions while providing a guide to select
sampling effort and quality control (Ode et al., 2016). Moreover, refer- comparison sites sharing a similar geophysical context. In accordance
ence pools may not provide the level of biophysical specificity or tempo- with this approach, we use stream classifications to identify bearing
ral extent needed for monitoring an individual restoration project. streams, but we also suggest that natural classes should be considered
In contrast to developing regional reference pools, the second ap- in conjunction with anthropogenic disturbance regimes to appropriate-
proach entails handpicking one to several reference or non-reference ly understand biophysical conditions. As an example, the Ecological
sites as comparisons in field studies monitoring the outcomes of resto- Limits of Hydrologic Alteration framework (ELOHA) uses natural hydro-
ration. We use the term “approach” loosely here as there is very little lit- logic classes to understand ecological responses to human-induced al-
erature formalizing a procedure for identifying reference sites for terations in streamflow regimes (Poff et al., 2010). The assumption is
monitoring restoration outcomes (Whittier et al., 2007). Ideally, sites that streams of similar hydrologic character would respond similarly
used as comparisons are identified prior to restoration taken place and to hydrologic alteration (Arthington et al., 2006).
are monitored jointly with restored sites in BACI or BACIP-like designs
(Stewart-Oaten et al., 1986; Underwood, 1994). In many cases, 4.1. Utility of including non-reference streams
however, the temporal extent of monitoring is truncated because resto-
ration actions have already ensued or resource limitations preclude Non-reference systems have been an under-utilized resource in res-
more robust sampling; thus, space is replaced for time and restored toration planning and long-term monitoring, despite the growing
sites are compared to unrestored (i.e., non-reference) and reference awareness that the ‘reference condition’, as defined by Stoddard et al.
(i.e., pristine) sites (e.g., Violin et al., 2011). In these studies, selection (2006), is becoming quite rare. A very common limitation in forming
of comparison sites tends to be somewhat haphazard as very little appropriate standards or benchmarks for evaluating restoration out-
justification is provided beyond generic similarities in basin attributes comes is the absence of suitable reference sites (Brewer and Menzel,
(e.g., size), description of disturbance (or lack thereof), or spatial prox- 2009; Dallas, 2013; Feio et al., 2013; Kosnicki et al., 2014). Within the
imity. Potentially, this leads to selecting sites that are biased by personal US, at least 39% of stream habitats for fish are jeopardized by landscape
preferences and not reflective of standard criteria. For instance, Whittier alterations (Esselman et al., 2013) and as much as 80% of streams show
et al. (2007) found that reference sites handpicked by state and federal some sign of hydrologic alteration (McManamay et al., 2017).
resource agencies were more representative of regional disturbed con- In some cases, non-reference systems may better characterize the
ditions than least-disturbed conditions. biophysical properties that managers desire to restore. For example,
Our framework is similar but divergent from past efforts in a few within the upper Tennessee River System (USA), the largest river sys-
ways. First, we focus considerable attention on identifying non- tems are regulated by dams, but yet contain the richest and most di-
reference sites, whereas most efforts have arguably focused on defining verse fish faunas (McManamay et al., 2015). In other situations,
reference conditions (we discuss the benefits of non-reference sites systems of reference quality are not desired or warranted. For example,
more thoroughly in subsequent sections). Akin to the intent of the sec- studies examining the biological impacts of a fly ash spill into Watts Bar
ond approach mentioned above, our principle objective is to provide an Reservoir, TN due to a dike failure at Tennessee Valley Authority's Kings-
approach to identify a few sites with desired biophysical specificity to ton Fossil plant used a non-impacted arm of the reservoir as the “refer-
guide restoration actions through long-term monitoring. Like the refer- ence” site as comparison to downstream “impacted” sites (Mathews et
ence-pool approach, however, we use very rich geospatial analyses to al., 2014; Smith et al., 2016). In this case, the reference condition was
consider thousands of potential sites thereby avoiding human-biases represented by a completely unnatural and man-made system, a reser-
in the selection of comparison sites. Another distinction is that our voir, which was far more appropriate than a free-flowing stream in eval-
framework isolates sites for future monitoring, rather than selecting uating potential background contaminant levels.
sites for their static pre-assessed conditions; however, SDMs and The inclusion of bearing streams in our analysis provided a temporal
existing biological monitoring information, where available, can pro- snapshot of potential mechanisms limiting restoration; however, the
vide a snapshot of biological conditions at potential comparison sites full utility of bearing streams become realized when they are included
to identify potential impaired processes. We used synthesized biological in long-term monitoring efforts. Long-term patterns promote the devel-
information to depict ecological conditions at bearing streams rather opment of sophisticated models to tease apart constraints to effective
than rely on empirical biological survey information. The benefit of restoration and promote adaptive management. We suggest that the in-
using SDMs to predict likely fish assemblage composition is that our as- clusion of carefully selected, non-reference sites can improve restora-
sessment considered all streams, not just those with previous commu- tion planning, prioritization, and monitoring in three main ways. First,
nity sampling. However, the obvious limitation is that ecological non-reference streams serve as a control to explain potential variation
conditions are estimated and hence, prone to uncertainty. Ideally, the of disturbances still at play in unrestored sites (e.g., Violin et al., 2011).
selection of sites should be conducted prior to restoration taking place These comparisons deepen the understanding of biophysical processes
to appropriately inform monitoring designs. We reiterate the impor- impaired by watershed disturbances (e.g., Violin et al., 2011)
tance of long-term monitoring in detecting and understanding impaired (Appendix 1), which is unlikely via direct comparison with undisturbed
biophysical dynamics to appropriately guide restoration (Muotka et al., reference sites. In situations where some biophysical properties are
2002). ameliorated via restoration while others remain actively disrupted, iso-
A noteworthy similarity of our framework and that of other ap- lating the role of individual disturbances requires careful selection of
proaches is the use of ecosystem classifications to stratify site selection. bearing systems. For instance, EFPC sites experience compounded dis-
Stream typologies have commonly been used to stratify site selection turbances primarily from intense urbanization, channelization, and ri-
for environmental monitoring efforts (e.g., Wollock et al., 2004). For ex- parian modification while the main focus of restoration has been
ample, Feio et al. (2014) used stream typologies including size, geology, reducing chemical contamination (e.g., mercury) (Loar et al., 2011).
R.A. McManamay et al. / Science of the Total Environment 621 (2018) 1208–1223 1219

Comparing physical and ecological patterns in EFPC to sites impacted by mirror the objectives of investigators, assuming the original selection
urbanization could help isolate impacts of chemical contamination on of the two reference sites was objective and unbiased by personal pref-
ecological recovery (Appendix 1). Additionally, comparing species and erences. This also suggests the framework offers significant time savings
functional community composition of EFPC to surrounding agricultural- given that the original reference site candidates were preliminarily
ly-impaired sites (majority of streams in region) yields information on identified on topographic maps and followed up by site visits with man-
common species and traits that would be expected in EFPC had industri- ual field measurements (Loar et al., 1992).
alization not taken place (Appendix 2). As would be expected, bearing stream candidates within the urban-
Bearing streams also assist in identifying thresholds and constraints impacted scenario (S2) and the least disturbance scenario (S3) were not
of ecosystem responses to disturbance and restoration, respectively. As found on the same river systems as current reference sites (an exception
is often the case, local conditions are an artifact of larger-scale distur- is that one bearing site for EFK24.4 was located on upper Brushy Fork for
bances at play within entire watersheds or surrounding landscapes the S3 scenario). While not a surprising outcome, the number of bearing
(Booth, 2005); hence, the scope of restoration projects should align stream candidates identified under these scenarios implies that a diver-
with the scales at which disturbances operate (Bernhardt and Palmer, sity of non-reference systems could, in theory, be selected to inform the
2011). Systems severely impacted by acute or chronic disturbances or monitoring of restoration outcomes. Furthermore, the objectives used
altered by unprecedented human influences may have reached a new to select these bearing streams will yield highly divergent results.
state of equilibrium (Beisner et al., 2003), undergone regime shifts Bearing stream candidates also allowed to us to examine the ade-
(e.g., Hilt et al., 2011), exceeded thresholds in which ecosystems are quacy of existing reference streams in representing desired ecological
likely responsive to restoration (Miller and Bestelmeyer, 2016), or conditions. Based on species distribution models, we compared
completely emerged as self-organized novel ecosystems (Hobbs et al., 30 years of ecological recovery in fish communities (or lack thereof) at
2009). Furthermore, these new states of self-organization may be irre- EFPC sites to observed and predicted fish species assemblages at
versible (Clewell et al., 2005), due in large part to surpassed threshold existing reference sites and bearing streams, respectively. Values for
exceedances (Miller and Bestelmeyer, 2016). Comparisons to non-ref- total richness of fish species were similar between existing reference
erence sites experiencing individual, rather than compounded, distur- streams and bearing streams for the EFK6.3, 13.8, and 18.6 sites
bances help to identify thresholds and their exceedances. These (Fig. 7); however, for EFK24.4 and EFK23.4 sites, fish species richness
“gradients of alteration” (Miller and Bestelmeyer, 2016) isolate poten- predicted by SDMs for bearing stream candidates was noticeably
tial disturbance thresholds, while simultaneously identifying con- lower than observed richness values in HCK20.6. Relying on HCK20.6
straints to effective restoration (e.g., limits to colonization, Bond and as reference stream for the two upper EFPC sites may be misleading
Lake, 2003; Miller and Hobbs, 2007). The nature of cause-effect, stress- and overestimate the potential species richness that could be attained
or-ecosystem-response relationships (Davies and Jackson, 2006), at these sites.Indeed, most of the bearing stream candidates identified
specifically the linearity or non-linearity (Walsh et al., 2005), are impor- for EFK24.4 and EFK23.4 had smaller drainage areas than HCK20.6.
tant in understanding the stages at which ecosystem transition to alter- Beyond assessing the adequacy of existing reference sites, compari-
nate states or in identifying thresholds, beyond which new ecosystems sons of ecological conditions at bearing stream candidates and EFPC
emerge. sites provided additional dimensionality in teasing apart the relative in-
Lastly, bearing systems provide a means to promote adaptive man- fluence of many compounded disturbances in constraining the ecologi-
agement by generating hypotheses on how disturbances and restora- cal effectiveness of restoration (Appendix 2). A general observation was
tion organize ecosystem function (Appendix 1), thereby reducing that fish species richness remained lower in three of the EFPC sites com-
uncertainty and risk in making restoration decisions (Walters, 1997). pared to reference streams and bearing streams (Fig. 7). Ecological re-
Bearing systems help to reveal causal mechanisms governing ecosystem covery was more pronounced at downstream sites, which were closer
processes under various disturbance regimes, and thus, identify alterna- to sources of colonists (e.g., McManamay et al., 2016) and removed
tive hypotheses. When viewed within the context of adaptive manage- from sources of physiochemical disturbances that ameliorated with
ment, alternative hypotheses translate into different restoration actions, distance downstream. EFPC consistently has lower numbers of opportu-
which are analogous to ecosystem-level experimental manipulations to nistic fish species than surrounding urban, agriculture, and low-distur-
inform subsequent decision-making (Palmer and Bernhardt, 2006; bance streams. We surmise this is the result of artificially-induced
Konrad et al., 2011). ecosystem stability arising from intense and chronic disturbances
(Beisner et al., 2003; Hilt et al., 2011), whereas opportunistic fish prefer
4.2. Applying the spatial framework to EFPC: a case study naturally variant environments (Olden and Kennard, 2010). Additional-
ly, many opportunistic fish are lithophilic, broad-cast spawners
We applied the spatial framework to identify bearing streams as ad- (Winemiller, 2005), which require clean and porous sand and gravel
ditional monitoring sites for comparison to restoration efforts in EFPC. (Balon, 1975). Intense urbanization and dramatic flood pulses in up-
This afforded an opportunity to examine the degree to which our frame- stream portions of EFPC have likely removed some of these finer mate-
work identified bearing streams similar to existing reference sites, rials or induced embeddedness of bed sediments in other areas.
based on following the same objectives for their selection. To provide Using species richness values from disturbed systems when setting
a reference scenario comparable to the criteria used to identify the cur- objectives for restoration creates some challenges. For example, ecolog-
rent set of reference sites, we identified bearing stream candidates im- ical endpoints for restoration in EFPC are likely to be an artifact of spe-
pacted by agriculture (e.g., S1 scenario) that mirrored the physical cies pools present in the surrounding impaired landscape, which
template and desired ecological conditions of EFPC sites. Given that surprisingly may be artificially high. For example, fish richness was typ-
36% of streams in the region have at least 25% upstream agricultural ically higher for bearing streams in the agriculture and urbanization sce-
land cover, it was not surprising that the S1 scenario yielded the most narios than in the least disturbance scenario – a somewhat
candidate bearing streams compared to other scenarios. Indeed, agricul- counterintuitive result. We conjecture this pattern is not an artifact of
tural impairment would be the expected norm for EFPC had industrial- our SDMs. While agriculture and urbanization land use practices have
ization of the Y12 Complex not taken place. While S1 bearing stream contributed to degradation of many streams in the ecoregion (Zheng
candidates were not identical to the stream reaches currently used as et al., 2008; Brown et al., 2005), we attribute higher richness to in-
reference sites, our framework identified several candidate stream creased numbers of generalists (Midway et al., 2015; Meador and
reaches immediately up and downstream from existing reference sites Goldstein, 2003), native-invasive species (Scott and Helfman, 2001),
and many more were identified within the same river systems. These but also endemic fluvial specialists, such as lamprey species
results suggest that the framework performs well at finding sites that (Ichthyomyzonspp.) or River chubs (Nocomis micropogon), whose
1220 R.A. McManamay et al. / Science of the Total Environment 621 (2018) 1208–1223

habitat needs are met in other impaired systems, but apparently not in outcomes are rarely documented (Palmer et al., 2010). Ultimately, this
EFPC. Appropriate restoration will require understanding whether suggests that current restoration practice is failing to address the most
members of ecological communities in impacted systems are the result important processes governing ecosystem impairments, or in the
of artificial inflation in species richness or remnants of species-rich eco- least, the nature of those processes and the scale at which they operate.
systems of the past. We suggest that more studies evaluate gradients of ecosystem
We emphasize that our 1st order comparisons between EFPC sites alteration (Miller and Bestelmeyer, 2016) or incorporate systems
and bearing sites should be further evaluated through long-term moni- representing a diversity of disturbances in restoration monitoring and
toring and analyses (Peterson, 2011). The hypothesized causal mecha- assessments to identify the causal mechanisms behind ecosystem im-
nisms generated from these preliminary comparisons provide grounds pairment, appropriately design restoration projects, or learn from and
for fostering adaptive management by suggesting new restoration ac- iteratively adapt to restoration actions that fail to meet project objec-
tions. Specifically, the absence of river chub and lamprey species suggest tives. However, we suggest that these disturbed systems be selected
multiple factors including limited recolonization potential and habitat under rigorous approaches to meet the site specifications needed to ad-
limitations, possibly a lack of suitable habitat. Continued evaluation of equately guide restoration. By using comprehensive and rich geospatial
bearing streams could help elucidate the likelihood of alternative man- coverage of all streams for consideration, the process minimizes
agement approaches, such as reintroduction efforts or habitat restora- human-generated biases in the arbitrary selection of potential streams
tion efforts. that do not align with prioritized objectives. Hence, the structure of
the framework is meant to provide a reality-check as to ensure the de-
5. Conclusions sired ecological outcomes match the constraints imposed by the condi-
tion of the landscape and physical properties of streams.
We provided a framework to identify specific low-disturbance and
non-reference sites as comparisons to restoration sites. Substantial liter- Acknowledgements
ature exists on defining reference conditions (Hawkins et al., 2010;
Dallas, 2013; Ode et al., 2016), prioritizing restoration actions (Bohn The study was authored by employees of UT-Battelleunder contract
and Kershner, 2002; Beechie et al., 2008), or developing appropriate DE-AC05-00OR22725 with the US Department of Energy. This research
restoration endpoints (Palmer et al., 2005; Tear et al., 2005; Miller and was sponsored by the Environmental Compliance Department of the
Hobbs, 2007). In contrast, the scientific literature is deficient of exam- Y-12 National Security Complex and by the Oak Ridge National
ples where reference or altered ecosystems are identified, prioritized, Laboratory's Environmental Protection Services Division's Water
and specifically tailored to guide restoration actions. Quality Programs. We thank Allison Fortner and two anonymous re-
Isolating, understanding, and repairing impaired processes are keys viewers for providing valuable comments on earlier versions of this
to successful restoration (Roni et al., 2008), yet successful restoration manuscript.

Appendix 1
Bearing stream candidates identified under the (A) agriculture impacts scenario, (B) urban impacted scenario, and (C) low-disturbance scenario.
Bearing candidates (pies) are color-coded to indicate whether they provide road access, are located within 45-m drive, have been previously sam-
pled, or none of the above. Existing reference sites (stars) are also provided as a comparison to bearing candidates within the agriculture impacts
scenario.

Appendix 2. Hypotheses regarding observed response to restoration generated from comparisons with predicted fish species compositions at
bearing sites

Scenario General observation Hypothesized mechanisms behind ecological responses to restoration

All Total fish species richness was generally lower at EFPC sites than bearing streams, Multivariate and cumulative disturbances are at play in EFPC as opposed to
especially EFK18.7, EFK23.4, and EFK24.4 sites. Opportunistic species richness singular disturbance agents acting on bearing stream communities. Several
was more noticeably lower in EFPC sites than bearing streams. cyprinid species, primarily opportunists, are missing from EFPC. Opportunistic
species favor naturally variable environments that promote competition over
niche space (e.g., larger river systems or alternatively, systems with flashy flows
followed by periodic intermittency) (Olden and Kennard, 2010); however,
maintained high levels of anthropogenic disturbance create an artificially stabile
system (Beisner et al., 2003), which is favored by equilibrium species
(Winemiller, 2005). Additionally, members of Nocomis, a genus of
mound-building chubs, are absent from EFPC sites but found in surrounding
streams. Nocomis chubs are considered keystone species are their mounds are for
utilized for spawning by many cyprinid opportunists (Peoples and Frimpong,
2015).
All Fish communities at downstream EFPC sites display stronger recoveries than Disturbance agents, such as sources of contaminants, are higher upstream (closer
upstream sites. Differences in species richness between EFPC and bearing streams to Y12-Complex). Urban landcover exceeds 80% of watershed area in upper
are more pronounced at upstream sites. sections of EFPC. Channelization and entrenchment is extensive in the upper
reaches. Downstream EFPC sites are closer to sources of colonists from adjacent
watersheds with lower disturbances. Connectivity has dramatic influence on
ecological recovery of restored ecosystems (McManamay et al., 2016)
All Species richness was extremely low at EFK24.4 (only 4 species) and the The entirety of upper EFPC's flow emerges as industrial effluent from the Y-12
community was comprised of only equilibrium species. Complex b1 km upstream of EFK24.4.Additionally, an impoundment is located no
N1 km downstream of EFK24.4, which prevents colonists from migrating to the
site. EFK24.4 is characterized as having high impervious surface, highly controlled
flows from industrial operations, channelized stream banks, and lack of finer
substrates. Such extreme cases of anthropogenic disturbances transition into
alternative ecosystem states of artificial stability (Beisner et al., 2003). Stable and
resource limiting environments are highly conducive to dominance by
equilibrium strategists (Winemiller, 2005; Olden and Kennard, 2010).
S1 Agriculturally-impaired streams represented the most numerous candidate Within the Ridge and Valley Ecoregion (RVE), most low-gradient river systems
R.A. McManamay et al. / Science of the Total Environment 621 (2018) 1208–1223 1221

(continued)

Scenario General observation Hypothesized mechanisms behind ecological responses to restoration

bearing streams and generally, had higher species richness than multiple EFPC are influenced by agricultural landcover (Zheng et al., 2008). Almost 36% of
sites and other bearing streams. streams in our region of interest had at least 25% upstream agricultural landuse.
In nearby ecoregions, agriculturally-impacted systems have higher fish richness
than forested streams and have displayed increases in richness over the last 50
years, even when excluding introduced species (Midway et al., 2015). Fish
community condition has been documented as high (Index of Biotic Integrity) in
areas of intense (50%) agricultural land use (Meador and Goldstein, 2003).
Agriculturally-impaired systems tend to have higher numbers of generalist
species, e.g., centrarchids (Midway et al., 2015), whose strategies are balanced
between equilibrium and opportunistic endpoints.
S2 Streams impacted by urbanization represented the smallest number of bearing Urbanized streams are numerically rare in the RVE (Zheng et al., 2008), especially
streams compared to other scenarios. Similar to agriculture, urbanized bearing those meeting the desired ecological community criteria. Only 11% of streams in
streams had higher species richness than many EFPC sites, had the highest our region of interest had at least 25% urban land cover. Decreases in richness
number of total species at 2 EFPC sites, and the highest number of opportunistic have commonly been observed with increased urbanization under various
species among all scenarios at 3 EFPC sites. stressor-response relationships (Carlisle et al., 2003; Meador et al., 2005);
however, in some cases, fish richness increases under intermediate levels of
urbanization from colonization by native-invasive species (Scott and Helfman,
2001). Species that dominate highly urbanized streams are generalists, primarily
equilibrium species, but also species with hedge betting strategies along the
equilibrium-periodic and equilibrium-opportunistic endpoints (Meador et al.,
2005). Increased opportunists in urbanized streams may again be an artifact of
the presence of Nocomis nest associates.
S3 Low-disturbance streams consistently had less species than other bearing stream Low disturbance systems may not represent the most species-rich systems
scenarios, but still had higher species richness than 3 of the respective EFPC sites. (Midway et al., 2015) for numerous reasons. First, streams falling under
“protected land” status were included, yet these streams may still be impacted by
anthropogenic disturbances. Secondly, while agriculture and urbanization usually
lead to endemic species loss, habitat changes that accompany these disturbances
usually provide new niches for occupation by generalist species, especially
centrarchids. Lastly, due to the nature of human settlement, streams with lower
disturbance levels may be further removed from human populations and hence,
not direct tributaries of larger systems, which can serve as sources of colonists
(McManamay et al., 2016).

Appendix 3. Supplementary data


Supplementary data associated with this article can be found in the online version, at https://doi.org/10.1016/j.scitotenv.2017.10.107. These data
include the Google map of the bearing stream candidates identified in the case study for each of the EFPC sites and each of the scenarios.

References Calabrese, J.M., Certain, G., Kraan, C., Dormann, C.F., 2014. Stacking species distribution
models and adjusting bias by linking them to macroecological models. Glob. Ecol.
Arthington, A.H., Bunn, S.E., Poff, N.L., Naiman, R.J., 2006. The challenge of providing Biogeogr. 23, 99–112.
e-flow rules to sustain river systems. Ecol. Appl. 16, 1311–1318. Carlisle, D.M., Stewart, P.M., Butcher, J.T., 2003. Macroinvertebrate assemblages associated
Balon, E.K., 1975. Reproductive guilds in fishes: a proposal and definition. J. Fish. Res. with patterns in land use and water quality. In: Simon, T.P. (Ed.), Biological Response
Board Can. 32, 821–864. Signatures: Indicator Patterns Using Aquatic Communities. CRC Press, New York,
Beechie, T., Pess, G., Roni, P., Giannico, G., 2008. Setting river restoration priorities: a re- pp. 271–285.
view of approaches and a general protocol for identifying and prioritizing actions. Clewell, A., Rieger, J., Munroe, J., 2005. Guidelines for Developing and Managing Ecological
N. Am. J. Fish Manag. 28, 891–905. Restoration Projects. 2nd edition. Society for Ecological Restoration International,
Beisner, B.E., Haydon, D.T., Cuddington, K., 2003. Alternative stable states in ecology. Tucson, Arizona http://www.ser.org/page/SERDocuments, Accessed date: 3 June
Front. Ecol. Environ. 1, 376–382. 2017.
Bernhardt, E.S., Palmer, M.A., 2011. River restoration: the fuzzy logic of repairing reaches Dallas, H.F., 2013. Ecological status assessment in Mediterranean rivers: complexities and
to reverse catchment scale degradation. Ecol. Appl. 21, 1926–1931. challenges in developing tools for assessing ecological status and defining reference
Bernhardt, E.S., Palmer, M.A., Allan, J.D., Alexander, G., Barnas, K., et al., 2005. Synthesizing conditions. Hydrobiologia 719, 483–507.
U.S. river restoration efforts. Science 308, 636–637. Davies, S.P., Jackson, S.K., 2006. The biological condition gradient: a descriptive model for
Bernhardt, E.S., Sudduth, K.B., Palmer, M.A., Allan, J.D., Meyer, J.L., et al., 2007. Restoring interpreting change in aquatic ecosystems. Ecol. Appl. 16, 1251–1266.
rivers one reach at a time: results from a survey of U.S. river restoration practitioners. Dubuis, A.J., Pottier, J., Rion, V., Pellissier, L., Theurillat, J., Guisan, A., 2011. Predicting spa-
Restor. Ecol. 15, 482–493. tial patterns of plant species richness: a comparison of direct macroecological and
Bohn, B.A., Kershner, J.L., 2002. Establishing aquatic restoration priorities using a water- species stacking modelling approaches. Divers. Distrib. 17, 1122–1131.
shed approach. J. Environ. Manag. 64, 355–363. Eberhardt, L.L., 1976. Quantitative ecology and impact assessment. J. Environ. Manag. 4,
Bond, N.R., Lake, P.S., 2003. Local habitat restoration in streams: constraints on the effec- 27–70.
tiveness of restoration for stream biota. Ecol. Manag. Restor. 4, 193–198. Elith, J., Leathwick, J.R., Hastie, T., 2008. A working guide to boosted regression trees.
Booth, D.B., 2005. Challenges and prospects for restoring urban streams: a perspec- J. Anim. Ecol. 77, 802–813.
tive from the Pacific Northwest of North America. J. N. Am. Benthol. Soc. 24, Esselman, P.C., Infante, D.M., Wang, L., Cooper, A.R., Wieferich, D., et al., 2013. Regional
724–737. fish community indicators of landscape disturbance to catchments of the contermi-
Brewer, J.S., Menzel, T., 2009. A method for evaluating outcomes of restoration when no nous United States. Ecol. Indic. 26, 163–173.
reference sites exist. Restor. Ecol. 17, 4–11. Etnier, D.A., Starnes, W.C., 1993. Fishes of Tennessee. Second edition. The University of
Brooks, S.C., Southworth, G.R., 2011. History of mercury use and environmental contam- Tennessee Press, Knoxville, Tennessee, USA.
ination at the Oak Ridge Y-12 Plant. Environ. Pollut. 159 (1), 219–228. Feio, M.J., Aguiar, F.C., Almeida, S.F.P., Ferreira, J., Ferreira, M.T., et al., 2013. Least disturbed
Brown, D.G., Johnson, K.M., Loveland, T.R., Theobald, D.M., 2005. Rural land-usetrends in condition for European Mediterranean rivers. Sci. Total Environ. 476–477, 745–756.
the conterminous United States, 1950–2000. Ecol. Appl. 15, 1851–1863. Gowan, C., Fausch, K.D., 1996. Long-term demographic responses of trout populations to
Burcher, C.L., Benfield, E.F., 2006. Physical and biological responses of streams to suburban- habitat manipulation in six Colorado streams. Ecol. Appl. 6 (3):931–946. https://
ization of historically agricultural watersheds. J. N. Am. Benthol. Soc. 25, 356–369. doi.org/10.2307/ 2269496.
1222 R.A. McManamay et al. / Science of the Total Environment 621 (2018) 1208–1223

Harris, R.R., 1999. Defining reference conditions for restoration of riparian plant commu- NFHP (National Fish Habitat Partnership), 2017. Through a fish's eye: the status of fish
nities: examples from California, USA. Environ. Manag. 24, 55–63. habitats in the United States 2015. http://assessment.fishhabitat.org/, Accessed
Hawkins, C.P., Olson, J.R., Hill, R.A., 2010. The reference condition: predicting benchmarks date: 6 February 2017.
for ecological and water-quality assessments. J. N. Am. Benthol. Soc. 29, 312–343. Ode, P.R., Rehn, A.C., Mazor, R.D., Schiff, K.C., Stein, E.D., et al., 2016. Evaluating the ade-
Hilt, S., Köhler, J., Kozerski, H.-P., van Nes, E.H., Scheffer, M., 2011. Abrupt regime shifts in quacy of a reference-site pool for ecological assessments in environmentally complex
space and time along rivers and connected lake systems. Oikos 120, 766–775. regions. Freshw. Sci. 35, 237–248.
Hobbs, R.J., Higgs, E., Harris, J.A., 2009. Novel ecosystems: implications for conservation Olden, J.D., Kennard, M.J., 2010. Intercontinental comparison of fish life history strategies
and restoration. Trends Ecol. Evol. 24, 599–605. along a gradient of hydrologic variability. In: Gido, K.B., Jackson, D.A. (Eds.), Commu-
HSC (Horizon Systems Corporation), 2017. NHDPlus Version 2. NHDPlus Home. http:// nity Ecology of Stream Fishes: Concepts, Approaches, and Techniques. American Fish-
www.horizon-systems.com/nhdplus/, Accessed date: 14 June 2017. eries Society Symposium 73, Bethesda, Maryland, USA, pp. 109–136.
Huang, J., Frimpong, E.A., 2015. Using historical atlas data to develop high-resolution Olivero Sheldon, A., Barnett, A., Anderson, M.G., 2015. A stream classification for the Ap-
distribution models of freshwater fishes. PLoS ONE https://doi.org/10.1371/ palachian region. The Nature Conservancy, Eastern Conservation Science. Eastern Re-
journal.pone.0129995. gional Office, Boston, MA.
Hughes, R.M., Larsen, D.P., Omernik, J.M., 1986. Regional reference sites: a method for Omernik, J.M., 1987. Ecoregions of the conterminous United-States. Ann. Assoc. Am.
assessing stream potentials. Environ. Manag. 10, 629–635. Geogr. 77, 118–125.
Jenks, G.F., 1967. The data model concept in statistical mapping. International Yearbook of Palmer, M.A., Bernhardt, E.S., 2006. Hydroecology and river restoration: ripe for research
Cartography. 7, pp. 186–190. and synthesis. Water Resour. Res. 42. https://doi.org/10.1029/2005WR004354.
Kasten, J.L., 1986. Resource management plan for the Oak Ridge reservation. Water Con- Palmer, M.A., Bernhardt, E.S., Allan, J.D., Lake, P.S., Alexander, G., et al., 2005. Standards for
servation Plan for the Oak Ridge Reservation ORNL/ESH-1/V21. vol. 21. Oak Ridge ecologically successful river restoration. J. Appl. Ecol. 42, 208–217.
National Laboratory, Oak Ridge, Tennessee, USA. Palmer, M.A., Menninger, H.L., Bernhardt, E., 2010. River restoration, habitat hetero-
King, R.S., Baker, M.E., Whigham, D.F., Weller, D.E., Jordan, T.E., 2005. Spatial consider- geneity and biodiversity: a failure of theory or practice? Freshw. Biol. 55,
ations for linking watershed land cover to ecological indicators in streams. Ecol. 205–222.
Appl. 15 (1), 37–153. Paul, M.J., Meyer, J.L., 2001. Streams in the urban landscape. Annu. Rev. Ecol. Syst. 32,
Kondolf, G.M., Angermeier, P.L., Cummins, K., Dunne, T., Healey, M., et al., 2008. 333–365.
Projecting cumulative benefits of multiple river restoration projects: an example Peoples, B.K., Frimpong, E.A., 2015. Biotic interactions and habitat drive positive co-occur-
from the Sacramento-San Joaquin River system in California. Environ. Manag. 42, rence between facilitating and beneficiary stream fishes. J. Biogeogr. 43, 923–931.
933–945. Peterson, M.J., 2011. Introduction to the biological monitoring and abatement program.
Konrad, C.P., Olden, J.D., Lytle, D.A., Melis, T.S., Schmidt, J.C., et al., 2011. Large-scale flow Environ. Manag. 47, 1005–1009.
experiments for managing river systems. Bioscience 61, 948–959. Poff, N.L., Richter, B.D., Arthington, A.H., Bunn, S.E., Naiman, R.J., et al., 2010. The ecological
Kosnicki, E., Sefick, S.A., Paller, M.H., Jarrell, M.S., Prusha, B.A., et al., 2014. Defining the ref- limits of hydrologic alteration (ELOHA): a new framework for developing regional e-
erence condition for wadeable streams in the Sand Hills subdivision of the Southeast- flow standards. Freshw. Biol. 55, 147–170.
ern Plains ecoregion. Environ. Manag. 54, 494–504. Ridgeway, G., 2017. Package ‘gbm’. Generalized boosted regression models. March 21,
Krumholz, L.A., 1954. An Ecological Survey of White Oak Creek, 1950–1953. ORO-587. 2017. https://cran.r-project.org/web/packages/gbm/index.html, Accessed date: 6 Jan-
Oak Ridge National Laboratory, Oak Ridge, Tennessee. uary 2017.
Loar, J.M., Adams, S.M., Allison, L.J., Black, M.C., Boston, H.L., et al., 1992. First Report on the Roni, P., Hanson, K., Beechie, T., 2008. Global review of the physical and biological effec-
Oak Ridge Y-12 Plant Biological Monitoring and Abatement Program for East Fork tiveness of stream habitat rehabilitation techniques. N. Am. J. Fish Manag. 28,
Poplar Creek. Y/TS-886. Oak Ridge Y-12 Plant, Oak Ridge, TN. 856–890.
Loar, J.M., Stewart, A.J., Smith, J.G., 2011. Twenty-five years of ecological recovery of East Ryon, M.G., 2011. Recovery of fish communities in a warm water stream following pollu-
Fork Poplar Creek: review of environmental problems and remedial actions. Environ. tion abatement. Environ. Manag. 47, 1096–1111.
Manag. 47, 1010–1020. Ryon, M.G., Loar, J.M., 1988. A checklist of fishes on the department of energy Oak Ridge
Lunde, K.B., Cover, M.R., Mazor, R.D., Sommers, C.A., Rash, V.H., 2013. Identifying reference reservation. J. Tenn. Acad. Sci. 63, 97–102.
conditions and quantifying biological variability within benthic macroinvertebrate Scott, M.C., Helfman, G.S., 2001. Native invasions, homogenization, and the mismeasure of
communities in perennial and non-perennial northern California streams. Environ. integrity of fish assemblages. Fisheries 26 (11), 6–15.
Manag. 51, 1262–1273. Smith, J.G., Baker, T.F., Murphy, C.A., Jett, R.T., 2016. Spatial and temporal trends in
Mathews, T.J., Fortner, A.M., Jett, R.T., Morris, J., Gable, J., Peterson, M.J., Carriker, N., 2014. contaminant concentrations in Hexagenia nymphs following a coal ash spill at the
Selenium bioaccumulation in fish exposed to coal ash at the Tennessee Valley Au- Tennessee Valley Authority's Kingston Fossil Plant. Environ. Toxicol. Chem. 35,
thority Kingston Spill Site. Environ. Toxicol. Chem. 33, 2273–2279. 1159–1171.
McManamay, R.A., Frimpong, E.A., 2015. Hydrologic filtering of fish life history strategies Stewart-Oaten, A., Murdoch, W.W., Parker, K.R., 1986. Environmental impact assessment:
across the United States: implications for streamflow alteration. Ecol. Appl. 25, pseudoreplication in time? Ecology 67, 929–940.
243–263. Stoddard, J.L., Larsen, D.P., Hawkins, C.P., Johnson, R.K., Norris, R.H., 2006. Setting expecta-
McManamay, R.A., Peoples, B.K., Orth, D.J., Dolloff, C.A., Matthews, D.C., 2015. Isolating tions for the ecological condition of streams: the concept of reference condition. Ecol.
causal pathways between flow and fish in the regulated river hierarchy. Can. J. Fish. Appl. 16, 1267–1276.
Aquat. Sci. 72, 1731–1748. Suding, K.N., 2011. Toward an era of restoration in ecology: success, failures, and oppor-
McManamay, R.A., Jett, R.T., Ryon, M.G., Gregory, S.M., Stratton, S.H., Peterson, M.J., 2016. tunities ahead. Annu. Rev. Ecol. Evol. Syst. 42, 465–487.
Dispersal limitations on fish community recovery following long-term water quality Tear, T.H., Kareiva, P., Angermeier, P.L., Comer, P., Czech, B., et al., 2005. How much is
remediation. Hydrobiologia https://doi.org/10.1007/s10750-015-2612-7. enough? The recurrent problem of setting measurable objectives in conservation.
McManamay, R.A., Surendran Nair, S., DeRolph, C.R., Ruddell, B.L., Morton, A.M., Stewart, Bioscience 55, 835–849.
R.N., Troia, M.J., Tran, L., Kim, H., Bhaduri, B.L., 2017. US cities can manage national hy- Troia, M.J., McManamay, R.A., 2016. Filling in the GAPS: evaluating completeness and cov-
drology and biodiversity using local infrastructure policy. Proc. Natl. Acad. Sci. 114: erage of open-access biodiversity databases in the United States. Ecol. Evol. https://
9581–9586. https://doi.org/10.1073/pnas.1706201114. doi.org/10.1002/ece3.2225.
Meador, M.R., Goldstein, R.M., 2003. Assessing water quality at large geographic scales: Underwood, A.J., 1994. On beyond BACI: sampling designs that might reliably detect en-
relations among land use, water physicochemistry, riparian condition, and fish com- vironmental disturbances. Ecol. Appl. 4, 3–15.
munity structure. Environ. Manag. 31, 504–517. USFS (United States Department of Agriculture Forest Service), 2013. Bearing trees very
Meador, M.R., Coles, J.F., Zappia, H., 2005. Fish assemblage responses to urban intensity important for maintaining property corners. https://www.fs.usda.gov/detail/
gradients in contrasting metropolitan areas: Birmingham, Alabama and Boston, Mas- blackhills/news-events/?cid=STELPRDB5410085, Accessed date: 6 January 2017.
sachusetts. Am. Fish. Soc. Symp. 47, 409–423. USGS (United States Geological Survey, Patuxent Wildlife Research Center), 2017. Pres-
Midway, S.R., Wagner, T., Tracy, B.H., Hogue, G.M., Starnes, W.C., 2015. Evaluating changes ence. https://www.mbr-pwrc.usgs.gov/software/presence.html, Accessed date: 5 Oc-
in stream fish species richness over a 50-year time-period within a landscape con- tober 2017.
text. Environ. Biol. Fish 98, 1295–1309. USGS (United States Geological Survey), United States Department of the Interior,
Miller, J.R., Bestelmeyer, B.T., 2016. What's wrong with novel ecosystems, really? Restor. 2016. National gap analysis program: protected areas database. http://
Ecol. 24, 577–582. gapanalysis.usgs.gov/padus/, Accessed date: 6 January 2017.
Miller, J.R., Hobbs, R.J., 2007. Habitat restoration: do we know what we're doing? Restor. Utz, R.M., Hildebrand, R.H., Boward, D.M., 2009. Identifying regional differences in
Ecol. 15, 382–390. threshold responses of aquatic invertebrates to land cover gradients. Ecol.
Morandi, B., Piégay, H., Lamouroux, N., Vaudor, L., 2014. How is success or failure in river Indic. 9, 556–567.
restoration projects evaluated? Feedback from French restoration projects. J. Environ. Violin, C.R., Cada, P., Sudduth, E.B., HassettSor, B.A., Penrose, D.L., Bernhardt, E.S., 2011. Ef-
Manag. 137, 178–188. fects of urbanization and urban stream restoration on the physical and biological
Muotka, T., Paavola, R., Haapala, A., Novikmec, M., Laasonen, P., 2002. Long-term recovery structure of stream ecosystems. Ecol. Appl. 21, 1932–1949.
of stream habitat structure and benthic invertebrate communities from in-stream Walsh, C.J., Fletcher, T.D., Ladson, A.R., 2005. Stream restoration in urban catchments
restoration. Biol. Conserv. 105, 243–253. through redesigning stormwater systems: looking to the catchment to save the
Murcia, C., Aronson, J., 2014. Intelligent tinkering in ecological restoration. Restor. Ecol. 22, stream. J. N. Am. Benthol. Soc. 24, 690–705.
279–283. Walters, C., 1997. Challenges in adaptive management of riparian and coastal ecosystems.
Nagel, D.E., Buffington, J.M., Parkes, S.L., Wenger, S., Goode, J.R., 2014. A Landscape Scale Conserv. Ecol. 1:1. http://www.consecol.org/vol1/iss2/art1/ (accessed 23 February
Valley Confinement Algorithm: Delineating Unconfined Valley Bottoms for Geomor- 2017).
phic, Aquatic, and Riparian Applications. U.S. Department of Agriculture, Forest Ser- White, S.L., Gowan, C., Fausch, K.D., Harris, J.G., Saunders, C.W., 2011. Response of trout
vice, Rocky Mountain Research Station, Fort Collins, CO (General Technical Report populations in five Colorado streams two decades after habitat manipulation. Can.
RMRSGTR-321, 42 pp.). J. Fish. Aquat. Sci. 68, 2057–2063.
R.A. McManamay et al. / Science of the Total Environment 621 (2018) 1208–1223 1223

Whittier, T.R., Stoddard, J.L., Larsen, D.P., Herlihy, A.T., 2007. Selecting reference sites for Zheng, L., Gerritsen, J., Beckman, J., Ludwig, J., Wilkes, S., 2008. Land use, geology,
stream biological assessments: best professional judgement or objective criteria. enrichment, and stream biota in the eastern ridge and valley ecoregion:
J. N. Am. Benthol. Soc. 26, 349–360. implications for nutrient criteria development. J. Am. Water Resour. Assoc. 44,
Winemiller, K.O., 2005. Life history strategies, population regulation, and implications for 1521–1536.
fisheries management. Can. J. Fish. Aquat. Sci. 62, 872–885.
Winemiller, K.O., Rose, K.A., 1992. Patterns of life-history diversification in North Ameri-
can fishes: implications for population regulation. Can. J. Fish. Aquat. Sci. 49, Further reading
2196–2218.
Wollock, D.M., Winter, T.C., McMahon, G., 2004. Delineation and evaluation of hydrologic- Smith, J.G., Brandt, C.C., Christensen, S.W., 2011. Long-term benthic macroinvertebrate
landscape regions in the United States using geographic information system tools and community monitoring to assess pollution abatement effectiveness. Environ.
multivariate statistical analyses. Environ. Manag. 34, 71–88. Manag. 47, 1077–1095.

Potrebbero piacerti anche