Science of the Total Environment 691 (2019) 1005–1015

Contents lists available at ScienceDirect

Science of the Total Environment

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

Incorporating functional traits to enhance multimetric index

performance and assess land use gradients
Kai Chen a,1, Abdul Razzaque Rajper a,1, Robert M. Hughes b, John R. Olson c, Huiyu Wei a, Beixin Wang a,⁎
a
Department of Entomology, Nanjing Agricultural University, Nanjing, Jiangsu 210095, PR China
b
Amnis Opes Institute and Department of Fisheries and Wildlife, Oregon State University, Corvallis, OR 97333, USA
c
School of Natural Sciences, California State University Monterey Bay, Seaside, CA 93955, USA

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

• Species traits support stream ecological

integrity.
• Functional metrics systematically co-
varied with natural variations.
• Better performance exhibited by
Functional-taxonomic than taxonomic
adjusted MMI.
• Functional-taxonomic MMI diagnoses
early changes in land use disturbance.
• Functional-taxonomic MMI facilitates
pre-emptive management and
conservation.

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

Article history: Taxonomic-based multimetric indices (MMIs) have been widely employed for assessing ecosystem status, partic-
Received 5 February 2019 ularly through the use of stream macroinvertebrate assemblages. However, the functional diversity and compo-
Received in revised form 3 July 2019 sition of assemblages is also important for maintaining stream ecosystem condition. Nonetheless, aquatic insect
Accepted 3 July 2019
functional diversity and composition have not commonly been included in MMIs. Our goal was to advance our
Available online 04 July 2019
understanding of the performance and ecological interpretation of an MMI that potentially combined functional
Editor: Yolanda Picó and taxonomic metrics. We sampled aquatic insects and natural and land-use variables at 74 temperate Chinese
streams. We selected a candidate set of 36 functional and 20 taxonomic metrics that were screened by range
Keywords: tests, natural variation, responsiveness to anthropogenic disturbance, and redundancy for subsequent inclusion
Functional diversity in MMIs. We determined if natural variation adjustments improved the performance of a functional-taxonomic
MMI MMI. Finally, we evaluated the degree to which the functional-taxonomic MMI served as an early-warning indi-
IBI cator of land use intensity. Natural variation explained between 19.62% and 71.02% of metric variability, indicat-
Random forest ing that functional metrics changed systematically along natural gradients. The final functional-taxonomic MMI
Aquatic insects
adjusted for natural variation incorporated multiple aspects of assemblage characteristics: functional richness,
Bioassessment
Rao's quadratic entropy, abundance-weighted frequency of soft bodies, abundance-weighted frequency of pred-
ators, and number of Diptera taxa. In contrast to the natural variation unadjusted MMI, the functional-taxonomic
adjusted MMI clearly distinguished least-disturbed sites from most-disturbed sites, exhibited high precision and
low bias, and showed a significant negative response to land uses. The slope of a linear regression relative to

⁎ Corresponding author.
E-mail addresses: kai.chen@njau.edu.cn (K. Chen), razarajper6@gmail.com (A.R. Rajper), hughes.bob@amnisopes.com (R.M. Hughes), joolson@csumb.edu (J.R. Olson),
2017102083@njau.edu.cn (H. Wei), wangbeixin@njau.edu.cn (B. Wang).
1
Co-first author.

https://doi.org/10.1016/j.scitotenv.2019.07.047
0048-9697/© 2019 Elsevier B.V. All rights reserved.
1006 K. Chen et al. / Science of the Total Environment 691 (2019) 1005–1015
0–10% urban and 0–20% agriculture was significantly steeper for the functional-taxonomic adjusted MMI than
that of the taxonomic adjusted MMI. We conclude that functional-taxonomic adjusted MMIs are more effective
indicators of ecological condition and risks to biota from human pressures than are purely taxonomic unadjusted
MMIs because functional-taxonomic MMIs are more sensitive to subtle anthropogenic pressures.
1. Introduction
Trait information has proven useful for characterizing ecosystem
function and structure and assessing ecological condition (Verberk
et al., 2013). Land use change is a major cause of biodiversity loss and
changes of ecosystem functions and services, and species traits ulti-
mately drive the biodiversity and ecosystem function relationships
(Flynn et al., 2009). Species traits such as life histories and physiologies
of species are expected to respond predictably to land-use caused envi-
ronmental changes (Statzner and Bêche, 2010; Dolédec et al., 2011).
Trait information based functional metrics (e.g., functional diversity
and composition) can reveal more consistent relationships between
levels of land use intensity and biological assemblage responses
than taxonomic variables (Dolédec et al., 2006; Flynn et al., 2009;
Mouillot et al., 2013). Functional diversity has also been documented
to indicate early warning signals preceding species loss (Villéger
et al., 2010; Mouillot et al., 2013). In addition, a multifaceted approach
(e.g., multimetric indices, MMIs) better represent ecological condition
and are more informative than a single metric (e.g., functional richness
and evenness) (Saito et al., 2015; de Castro et al., 2018). Incorporating
species traits in bioassessment thus offers a potential means of improv-
ing ecological condition assessments (Dolédec and Statzner, 2010).
Although increasing trait data are available for many stream and
river invertebrates (e.g., Poff et al., 2006), the use of traits in MMIs is
poorly developed (Bonada et al., 2006; Vandewalle et al., 2010;
Orlofske and Baird, 2014; Saito et al., 2015). Their uncommon adoption
reflects poor knowledge of species trait-function relationships for ap-
propriate choices of traits (Pedley and Dolman, 2014), the accuracy
and sensitivity of traits to discriminate anthropogenic stressors
(Dolédec et al., 1999; Orlofske and Baird, 2014), as well as how traits re-
spond to particular stressors (Culp et al., 2011; de Castro et al., 2018;
Leitão et al., 2018). Although trait-based functional metrics, such as
the relative abundance of feeding, habitat, and tolerance groups have
been commonly used in MMIs (e.g., Hering et al., 2006; Stoddard
et al., 2008), functional diversity and composition per se have been
rarely used (Saito et al., 2015).
MMIs based on species traits aid detection of early assemblage-level
changes along disturbance gradients, particularly those associated with
changes in ecological processes (Mouillot et al., 2013). When species
perform similar functions, high functional redundancy buffers assem-
blages against disturbances (Mouillot et al., 2014). Therefore, loss of
any given species may have relatively little effect on ecosystem func-
tioning (Petchey et al., 2007; Ricotta et al., 2016). Traditional taxonomic
composition metrics (e.g., species richness and evenness) thus are often
weak tools for signaling disturbance (Hussin et al., 2012) or ecosystem
recovery (Villéger et al., 2010). On the other hand, low disturbance in-
tensity may decrease abundance of sensitive species with vulnerable
and unique functions without modifying overall assemblage composi-
tion or affecting total species richness (Mouillot et al., 2013; Leitão
et al., 2016; de Oliveira-Junior et al., 2017).
Functional metrics based assessments may require adjustments to
partition the effects of natural variations from co-occurring anthropo-
genic pressures (Zuellig and Schmidt, 2012). Although functional com-
position was reported to be confounded by spatial extents and natural
environmental gradients (e.g., Statzner et al., 1997; Poff et al., 2010;
Pease et al., 2012), functional metrics have been rarely evaluated
along natural environmental gradients (but see Péru and Dolédec,
© 2019 Elsevier B.V. All rights reserved.
2010). However, improved understanding of the effects of natural vari-
ation on functional diversity and composition can facilitate more accu-
rate detections of ecological differences among sites and more
accurate diagnoses of anthropogenic impacts on assemblages (Dolédec
and Statzner, 2010; Menezes et al., 2010). Modeling based on natural
variation in reference condition data is an increasingly common way
to obtain site-specific expectations that can improve the performance
of MMIs (Pont et al., 2009; Hawkins et al., 2010b; Mazor et al., 2016).
Therefore, our aim was to develop a natural-variation adjusted MMI
that included functional and taxonomic metrics (hereafter functional-
taxonomic adjusted MMI) that passed our metric screening steps. We
then sought to use that MMI to assess the condition of aquatic insect as-
semblages and their associations with land use pressures. We intended
to produce a robust assessment tool that would support a wide variety
of bioassessment applications for perennial temperate streams and riv-
ers in eastern China and similar regions. We expected that adjusted
MMIs would partition the effects of natural variations from anthropo-
genic pressures on metrics, thereby improving the ability of the
functional-taxonomic MMI to discriminate least- and most-disturbed
sites (hypothesis 1). We also hypothesized that a functional-
taxonomic adjusted MMI would be a more sensitive indicator of low
levels of urban and agriculture intensity than a conventional taxonomic
adjusted MMI (hypothesis 2).
2. Methods
2.1. Study area and sampling sites
We surveyed stream sites across the middle of the Qiantang River
Basin in eastern China (Fig. 1) in April of 2010 and 2011. The Qiantang
River Basin has a drainage area N 55,000 km2. The study area is charac-
terized by a subtropical monsoon climate, with a mean annual temper-
ature of 17 °C and an average annual precipitation of 1600 mm. The
Qiantang River discharge peaks in May through September as a result
of thunderstorms (Wang et al., 2012).
We sampled a total of 90 stream sites (100-m long reach) distrib-
uted along a gradient of known anthropogenic impacts based primarily
on the amount of agricultural and urban land use in their catchments
(Wang et al., 2012). The landscapes in the region constitute mosaics of
well-established mechanized agriculture, urban areas, and undisturbed
primary forests (Liu et al., 2016). Land uses in the upper study region
were dominated by forest, but it also has been modified by some aggre-
gate mining, agricultural activities, and rural developments. The impacts
of anthropogenic activities are minimal for the northwestern region, but
the remaining area is dominated by agricultural and urban land uses.
2.2. Water quality
We collected water chemistry samples and measured physicochem-
ical variables in situ before biological sampling. We measured electrical
conductivity using a portable meter (HI 98129, HANNA, Italy). We col-
lected stream water (500 ml) from a riffle at each site in an acid-
cleaned high-density polyethylene bottle, stored it in the dark on ice,
and then analyzed it in the laboratory for total nitrogen and total phos-
phorus by UV spectrophotometer following Standard Methods for the
Analysis of Water and Wastewater (EPBC, 2002).
 K. Chen et al. / Science of the Total Environment 691 (2019) 1005–1015
Fig. 1. Locations of study sites in the middle Qiantang River Basin, eastern China.
2.3. Predictor variables and land use stressors
We characterized natural, unalterable predictor variables and
human activity land use stressors through use of a geographic informa-
tion system. We used the Geospatial Modeling Environment (Beyer,
2014) to extract predictor variables and land use stressors of the
whole catchment upstream from the site (catchment-scale) and the
local site scale. We delineated catchment boundaries for each site with
the multi-watershed delineation tool (Chinnayakanahalli et al., 2006)
from 30-m digital elevation models (http://www.gscloud.cn) provided
by the Chinese Academy of Sciences. We obtained land use information
from the Zhejiang Land Use 2007 database (http://www.zjemc.org.cn)
to calculate the % area in each catchment classified as forest, agriculture,
or urban. The predictors used in modeling were catchment area, catch-
ment shape, catchment slope, and catchment elevation range (http://
datamirror.csdb.cn), and both catchment- and site-scale land surface
temperature (Land Processes Distributed Active Archive Center, 2013),
bio-climate (Hijmans et al., 2005), air temperature, and precipitation
(Grumman, 2002; Hijmans et al., 2005; Willmott and Matsuura, 2012).
2.4. Aquatic insects
We sampled aquatic insects from each site via a multi-habitat sam-
pling technique (Barbour et al., 1999) through use of a modified Surber
sampler (0.09 m2, 250 μm mesh size) from 3 riffles and 2 pools. We
washed the substrate materials in the field and inspected the net for
aquatic insects adhering to the mesh. The composite sample in an area
of 0.45 m2 for each site was preserved with buffered 10% formalin,
and then processed in the laboratory in a water-filled shallow plastic
tray under incident illumination. The processed insects were then pre-
served in 80% ethanol. All individuals were counted and identified
under a stereomicroscope to the lowest practical taxonomic level (usu-
ally genus; Morse et al., 1994). We excluded 10 sites because some func-
tional metric (e.g., functional dispersion) calculations are not applicable
to sites with ≤2 taxa (Laliberté and Legendre, 2010). Also, metrics that
rely on abundances (e.g., taxa richness or uncommon species) show a
strong dependence on the number of individuals collected (Lorenz
et al., 2004; Chen et al., 2015). For this reason, the U.S. Environmental
Protection Agency's National Rivers and Streams Assessment limits
1007
macroinvertebrate samples to those yielding a minimum of 300 individ-
uals (Hughes and Peck, 2008; Stoddard et al., 2008). We excluded 6 sites
with b200 individuals collected (Chen et al., 2015). A total of 74 sites
with complete environmental, taxonomic, and functional data were
used for further MMI development and ecological interpretation
analyses.
2.5. Candidate biological metrics
The number of candidate metrics used in the literature often exceed
100 (e.g., Stoddard et al., 2008), we proposed a priori 86 metrics (36
functional and 50 taxonomic) for possible inclusion in the MMIs. We fo-
cused on aquatic insects (especially Coleoptera, Diptera, Ephemerop-
tera, Heteroptera, Odonata, Plecoptera, and Trichoptera) because of
their known sensitivities to anthropogenic disturbances (Wang et al.,
2012; Helson and Williams, 2013; Chen et al., 2017) and their trait infor-
mation availability.
2.5.1. Functional metrics
For each taxon, we characterized eight traits (Table 1) commonly
used in previous macroinvertebrate functional diversity studies
(Dolédec and Statzner, 2010; Schmera et al., 2017) that were considered
likely to link organism performance with environmental conditions or
ecosystem processes (Poff et al., 2006). Each trait was characterized
using a trait modality following Colzani et al. (2013) and Ding et al.
(2017). Trait information was mainly obtained from published articles
(Morse et al., 1994; Statzner et al., 1997; Usseglio-Polatera et al., 2000;
Poff et al., 2006; Tomanova and Usseglio-Polatera, 2007; Colzani et al.,
2013; Zhang et al., 2013) and online sources (http://pubs.usgs.gov/ds/
ds187/, http://www.freshwaterecology.info/index.php). Because of in-
sufficient data, we used maximum adult lengths as surrogates of larval
body size (Ding et al., 2017). We selected traits that were readily avail-
able in the literature or that could be determined confidently by expert
opinion and applied to all our taxa. Those traits have been shown to re-
spond to specific environmental gradients and stream conditions
(e.g., Poff et al., 2006; Ding et al., 2017). We defined a discrete score
for each trait modality within each trait beginning with 1, 2, 3, and so
on (Colzani et al., 2013).
1008 K. Chen et al. / Science of the Total Environment 691 (2019) 1005–1015

Table 1 functional trait-space, where the centroid is the average of the traits of
Description of traits and trait modalities used in the trait-based functional metrics analyses species present in the assemblage (Anderson, 2006; Laliberté and
(modified from Colzani et al., 2013 and Ding et al., 2017).
Legendre, 2010). High FDis reflects the differentiation among coexisting
Trait Trait modalities Score Code species, such as interspecific competition, favoring of niche segregation,
Soft body 1 Prot1 or environmental heterogeneity, selecting for species adapted to differ-
Exoskeleton or external
Lightly sclerotized 2 Prot2 ent habitats. FDiv measures the degree to which the species abundances
protection
Highly sclerotized 3 Prot3 diverge from the center of a functional space (Mason et al., 2005;
Integumentary 1 Resp1
Villéger et al., 2008). High FDiv entails a high degree of niche differenti-
Respiration type Branchial 2 Resp2
Air (spiracles, tracheae, plastrons) 3 Resp3 ation, more efficient resource use, and thus low resource competition
Small (b9 mm) 1 Size1 (Mason et al., 2005). RaoQ quantifies functional diversity as pairwise
Maximal body size Medium (9–16 mm) 2 Size2 differences in trait composition between taxa and accounts for relative
Large (N16 mm) 3 Size3 abundances of taxa in an assemblage (Botta-Dukát, 2005). RaoQ is use-
Hydrodynamic (flat, fusiform) 1 Shape1
Body shape
Not hydrodynamic (cylindrical, round) 2 Shape2
ful in explaining the relationships between assemblage assembly pro-
No refuge 1 Refu1 cesses and environmental constraints (Rao, 1982; Mouchet et al.,
Fixed network and retreat 2 Refu2 2010; Péru and Dolédec, 2010). It combines information from both
Refuge type
Sand, debris, and/or shell 3 Refu3 FRic and FDiv and simultaneously indicates if taxa are clustered in
Leaf parts or wood 4 Refu4
trait space and if there is divergence of taxa in resource use (Schleuter
Only depositional 1 Rheo1
Rheophily Depositional and erosional 2 Rheo2 et al., 2010). We also calculated the abundance-weighted frequency of
Erosional 3 Rheo3 each trait modality in the assemblage as a measure of functional
Burrower 1 Habi1 composition.
Climber 2 Habi2
Habit Sprawler 3 Habi3
2.5.2. Taxonomic metrics
Clinger 4 Habi4
Swimmer 5 Habi5 We started with over 50 candidate taxonomic metrics, which we re-
Collector-gatherer, brusher 1 Trop1 duced to 20 commonly used metrics that had previously been shown to
Collector-filterer 2 Trop2 have high disturbance-discriminating power and low redundancy in
Trophic habit Herbivore (scrapers, leaf miners) 3 Trop3
our study area (Chen et al., 2014, 2015, 2019; Table S1). Our candidate
Predator (piercers, engulfers) 4 Trop4
Shredder 5 Trop5 taxonomic metrics included all suggested metric categories (Stoddard
et al., 2008): taxonomic richness (e.g., total number of aquatic insect
taxa, number of Diptera taxa), taxonomic composition (e.g., % Ephem-
eroptera individuals), compositional diversity (e.g., Margalef richness
index), and tolerance (e.g., number of generally sensitive or intolerant
We calculated 36 functional metrics (8 functional diversity indices taxa). Some taxonomic richness and composition metrics were calcu-
and 28 functional composition metrics; supplementary data Table S1) lated because of their resistance (e.g., Diptera) or sensitivity
to describe the functional composition and structure of aquatic insect (e.g., Ephemeroptera) to disturbances (e.g., Macedo et al., 2016), using
assemblages based on the species position in multidimensional func- all taxa and abundances within that order. The selected metrics have
tional space and their relative abundances. Following Villéger et al. also been widely used in other bioassessment indices (e.g., Vander
(2008), we calculated functional diversity indices by simply lumping Laan and Hawkins, 2014; Macedo et al., 2016).
all trait modalities together without weighting. We calculated trait rich-
ness (TR) and trait diversity (TD) following Bonada et al. (2007) and 2.6. MMI development
Larsen and Ormerod (2010), respectively. TR calculates the total num-
ber of trait modalities of all species individuals occurring at each site 2.6.1. Site classification
(Bêche and Statzner, 2009). TD is similar to the Simpson taxon diversity We stratified the study region into small, medium and large
based on richness and abundance across all modalities within each trait catchment-size categories (Table 2) and then selected the least-
(Larsen et al., 2011). We used the pd function in the picante R package disturbed sites within each category as reference sites (Stoddard et al.,
(Kembel, 2010) to calculate distance-based functional diversity (dbFD; 2006; Vander Laan and Hawkins, 2014). A compromise between refer-
Faith, 1992; Petchey and Gaston, 2002). Distance-based FD measures ence site quality and representativeness of the natural environment is
the extent of complementary among-species trait values by estimating needed for site-specific condition predictions at a large and varied num-
the dispersion of species in trait space (Petchey and Gaston, 2002), ber of least-disturbed sites (Ode et al., 2016; Chen et al., 2019). Catch-
which is related to functional richness (Poos et al., 2009). Distance- ment strata ensure that we had reference sites in upland and lowland
based FD is estimated by measuring the total length of the branches of areas and across a range of stream sizes. Least-disturbed site selection
the dendrogram constructed from species functional composition thresholds for each of the screening variables were determined
(Petchey and Gaston, 2002). We used Gower's dissimilarity to develop
functional distance matrices and applied the unweighted pair group
method using arithmetic average (UPGMA) for clusters. We used the Table 2
Catchment size classes, land use, and chemical thresholds used to define least- and most-
dbFD function in the FD R package (Laliberté and Legendre, 2010) to cal-
disturbed sites.
culate functional richness (FRic), functional evenness (FEve), functional
dispersion (FDis), functional divergence (FDiv), and Rao's quadratic en- Variable Least-disturbed Most-disturbed
tropy (RaoQ). FRic is the convex-hull volume of the functional space Catchment size Catchment size
filled by all species within the local assemblage, which measures the
Small Medium Large Small Medium Large
overall spread of traits in a particular assemblage (Villéger et al., 2008)
Area (km2) ≤5.9 5.9–50 N50 ≤5.9 5.9–50 N50
with higher values implying more niches filled by taxa. We standard-
% Agriculture ≤5 ≤10 ≤15 ≥25 ≥25 ≥50
ized the FRic between 0 and 1. FEve measures the distribution of abun- % Urban ≤5 ≤10 ≤15 ≥20 ≥20 ≥40
dance in a functional trait space ranging between 0 and 1. Lower FEve Conductivity (μs/cm) ≤100 ≤100 ≤300 ≥300 ≥300 ≥500
indicates the underused niche space (Mason et al., 2005; Villéger Total nitrogen (mg/L) ≤2.0 ≤2.0 ≤2.0 ≥3.0 ≥3.0 ≥5.0
et al., 2008). FDis measures the average distance of species weighted Total phosphorus (mg/L) ≤0.1 ≤0.1 ≤0.15 – – –
Number of sites 18 4 1 7 7 2
by their relative abundance to the centroid of the assemblages in a
 K. Chen et al. / Science of the Total Environment 691 (2019) 1005–1015
differently for catchment-size categories, with the most stringent
criteria used for small catchments, and progressively less rigorous stan-
dards for medium and large catchments (Table 2). We also used land
use and available water chemistry data to identify sites with the greatest
apparent human-caused degradation. No biological information was
used to identify the least-, intermediate- or most-disturbed sites. A
site was not considered to be least- or most-disturbed if it failed any
one of the filters in Table 2. Sites that were not designated as either
least- or most-disturbed were assigned as intermediate. We identified
23 least-disturbed (reference) sites, 35 intermediate-disturbed sites,
and 16 most-disturbed sites for further analysis.
2.6.2. Metric selection & MMI development
We developed 4 MMIs: a natural variation adjusted MMI that com-
bined functional and taxonomic metrics (i.e., functional-taxonomic ad-
justed MMI), a functional-taxonomic unadjusted MMI without natural
variation adjustment of metrics, a taxonomic adjusted MMI, and a taxo-
nomic unadjusted MMI. Metrics failing any of the following screening
criteria were removed from consideration in the final MMIs (Stoddard
et al., 2008).
2.6.2.1. Range test. Metrics with a high proportion of the same values or
narrow ranges across sites are unlikely to cover a wide range of natural
variation and to sufficiently discriminate site-level difference in assem-
blage condition (Stoddard et al., 2008; Esselman et al., 2013; Chen et al.,
2014). For example, the number of Heteroptera taxa per site was usually
only one or zero, leaving no room for intermediate metric scores and in-
creasing the probability of erroneous scoring based on the detection of a
single individual or not. Low ranges in metric values across reference
sites indicate little natural variability, thereby limiting a metric's ability
to detect the potential effects of human stressors on the aquatic biota
(Stoddard et al., 2008; Chen et al., 2019). Following Tang et al. (2016),
we excluded metrics with scores ranging b15% change among reference
sites and having the same or very similar metric values at more than half
of all sites.
2.6.2.2. Natural variation adjustment. An adjusted MMI attempts to con-
trol for the effects of natural factors on metrics by determining the site-
specific expected metric values at reference sites given natural environ-
mental gradients (Pont et al., 2009; Hawkins et al., 2010b). Because bi-
ological metrics often vary along natural environmental gradients
(Hawkins et al., 2010a; Moya et al., 2011), that natural variation could
reduce our ability to discern meaningful biological degradation caused
by anthropogenic disturbances (Cao et al., 2007). To correct for this po-
tential confounding influence, we determined expected metric values
for each site as a function of natural predictor variables that were unal-
terable by humans. Residuals (observed – predicted metric values) from
random forest (RF) models of metric response to natural gradients, in-
stead of raw metrics, were used in subsequent metric selection steps.
We used residuals when RF models explained N10% of the variation in
reference metric scores; otherwise, we retained the raw metric values
(Hawkins et al., 2010a). The attractive features of RF models are their
ability to handle non-linear relationships, model interactions among
predictor variables, and conduct bootstrap subsamples (i.e., in-bag and
out-of-bag samples). In addition, RF models are insensitive to correlated
predictors and overfitting, and several diagnostic tools (e.g., variable im-
portance and partial dependence plots) are available (Cutler et al.,
2007). We implemented RF models to create 1500 trees for each
model with randomForest package (Liaw and Wiener, 2002).
In riverine ecosystems, significant spatial autocorrelation can result
from individuals moving freely among reaches (Lloyd et al., 2006),
resulting in similar assemblages simply because of spatial proximity
(Van Sickle and Hughes, 2000). Therefore, we tested for spatial autocor-
relation in species composition using Mantel tests with 1000 permuta-
tions (Pyron et al., 2008). We found no significant effects of spatial
autocorrelation in composition (Mantel r = 0.02 and p = 0.16);
1009
therefore, we did not partition spatial autocorrelation effects in our ad-
justed MMIs.
2.6.2.3. Redundancy. We used cluster analysis and t-statistics to select
non-redundant metrics to reduce the average correlation among met-
rics (Van Sickle, 2010; Chen et al., 2014). A cluster dendrogram was de-
veloped using absolute Pearson's correlation coefficients (r) as the
similarity measure and Ward linkage as the clustering method (Cao
et al., 2007).
2.6.2.4. Discrimination testing and final metric selection. The discriminat-
ing ability of each metric was evaluated through use of non-
parametric t-tests comparing mean metric values from least- and
most-disturbed sites. We selected the metric from each of the above
clusters that showed the greatest differentiation (i.e., the largest abso-
lute t-value) to include in the MMI.
2.6.3. Metric and MMI normalizing, rescaling and scoring
We standardized each metric to a score between 0 and 100 to nor-
malize the range of metric values on a common scale and, if necessary,
rescaled their response direction to calculate the final MMIs (Stoddard
et al., 2008). In practice, raw metric values have different numbers
and units and different response directions with increasing disturbance.
Also, metric values can respond either by decreasing (e.g., sensitive taxa
richness) or increasing (e.g., % tolerant individuals) with increased an-
thropogenic disturbance. To combine these individual measures into
an integrated MMI, it is essential to standardize all metrics to the
same response direction and to unitless scores on a common scale rang-
ing between 0 and 100. We scored raw metric values that decreased
with increased anthropogenic disturbance as 100*(metric value –
M5th)/(M95th – M5th), with M5th and M95th being the 5th and 95th percen-
tile of metric values. We rescaled disturbance-tolerant metrics by
subtracting the score from 100 to reverse the scale, so that all scored
metrics had the same negative direction of response to impact (Cao
et al., 2007; Stoddard et al., 2008). Each standardized metric score is
thereby finally expressed so that higher scores indicate less impairment
(Hering et al., 2006). We used 5th and 95th percentiles of metric scores
to prevent effects of extreme values, because this approach produces
an MMI with high responsiveness and lower coefficients of variation
(Blocksom, 2003). Rescaled metric scores b0 were set to 0 and those
N100 to 100 (Cao et al., 2007). The final MMI was the sum of the selected
metrics divided by the number of metrics to rescale the MMI range from
0 to 100, regardless of the final number of metrics. This step facilitates
our ability to combine different metrics into a sensitive MMI and detect
human stress on aquatic systems, without changing metrics' response
magnitudes (supplementary data Fig. S1). However, we can still exam-
ine and interpret individual metrics using the calculated metric values
before their standardization.
2.6.4. Performance evaluation
The performances of the alternative MMIs were evaluated in terms
of bias, precision, responsiveness, and sensitivity (Hawkins et al.,
2010a). We assessed systematic bias by the degree to which the remain-
ing variation within reference site MMI values were associated with
natural gradients. MMI scores at reference sites were regressed as a
function of natural environmental variables using RF models, with
higher % variance explained indicating higher bias. The observed vari-
ability (coefficient of variation, CV) of MMI scores among reference
sites was used to measure how precisely we can estimate MMI scores,
with a lower CV indicating higher precision. Responsiveness was
assessed by the difference between mean MMI scores between least-,
intermediate, and most-disturbed sites through use of a Student's t-
test. The larger absolute t-statistic indicates greater discrimination effi-
ciency. We measured sensitivity as the percentage of intermediate and
most disturbed sites inferred as being in non-reference biological condi-
tion. Samples with MMI scores below the reference score equivalence
1010 K. Chen et al. / Science of the Total Environment 691 (2019) 1005–1015
threshold from one-sided non-central interval and equivalence tests
(Bowman and Somers, 2006) were considered as being in non-
reference condition. Higher correct classification indicates higher
sensitivity.
We also evaluated the relevancy of MMIs in responding to land use
gradients to test our second hypothesis. We first used univariate and bi-
variate partial-dependence plots from RF models to assess the relation-
ships between MMI scores and % urban and % agriculture. The partial
dependence plots characterize the effect of % agriculture or % urban on
each MMI score by plotting its predicted score across the gradient of
the predictor while accounting for % urban or % agriculture (Hastie
et al., 2001). We visually inspected the partial dependence plots to iden-
tify % urban and % agriculture thresholds as the point at which a small
increase in land use variables resulted in a large change in MMI scores.
We then used linear regressions to evaluate the relevancy of the rela-
tionship between MMI scores and % agriculture and % urban (from 0%
to threshold %) land use in the catchment (Tang et al., 2016). We mea-
sured relevancy as the amount of variance in MMI scores explained by %
catchment land use and by the slope of the relationships. A steeper slope
indicates a better early warning signal of land use impacts. We used
generalized linear mixed-effects models (GLMMs) to test if different
MMIs responded significantly differently to land use gradients using
site as a random effect. We implemented GLMMs with the lmerTest
package (Kuznetsova et al., 2017). We only compared relevancy be-
tween functional-taxonomic adjusted and taxonomic adjusted MMI
scores because unadjusted MMI values were always confounded by sys-
tem biases (see Results below). All statistical analyses were conducted
in R software (version 3.2.3; R Development Core Team, 2014, http://
www.r-project.org/).
3. Results
3.1. Aquatic insect assemblages
We collected a total of 221 aquatic insect taxa, with taxa richness
ranging from 3 to 74 across all sites. We identified an average of 52
(±15) taxa and 2099 (±1001) individuals at least-disturbed sites, 15
(±11) taxa and 1135 (±1377) individuals at intermediate sites, and 5
(±2) taxa and 240 (±34) individuals at most-disturbed sites. Species
composition differed between the least- and most-disturbed sites
(Fig. S2), with the most-disturbed sites highly correlated with urban
and agriculture land use stressors, and the least-disturbed sites highly
correlated with % forest (Fig. S3).
3.2. Metric screening and selection of final metrics
The screening procedure reduced the 56 candidate metrics to 5 final
metrics for each of the 4 MMIs. Range tests excluded 5 metrics. Natural
gradients explained N10% of the variation for all remaining candidate
metrics. The residuals of those metrics were separated into five distinct
clusters. The 5 metrics with the highest t-statistic differed between
functional-taxonomic adjusted and unadjusted MMIs. The metrics in
the functional-taxonomic adjusted MMI were FRic, RaoQ, abundance-
weighted frequency of soft bodies, abundance-weighted frequency of
predators, and number of Diptera taxa, with natural variation explained
ranging between 19.62% and 71.02% (Tables 3 and S2). The metrics in-
cluded in the functional-taxonomic unadjusted MMI were FRic, dbFD,
abundance-weighted frequency of soft bodies, abundance-weighted
frequency of collector-gatherers and brushers, and number of Diptera
taxa, with natural variation explained ranging between 28.44%
and 71.02% (Tables 3 and S2). Geomorphologic (e.g., elevation and
slope), climatic (e.g., precipitation and temperature), and spatial
(e.g., latitude) were primary factors explaining 47.01–71.02% of aquatic
insect functional composition and 19.62–59.49% of functional diversity
variation (Table 3).
Table 3
Random Forest (RF) models for metrics included in functional-taxonomic adjusted and un-
adjusted MMIs. RF % Var = % variation of least-disturbed site metric values explained in RF
model. Predictors are arranged by variable importance with decreasing importance from
left to right. CatArea = catchment area (km2), Lat = latitude, ELEV_Mean_CA = mean
catchment elevation (m), SLPMAX_CA and SLPMIN_CA = maximum and minimum catch-
ment slope (%), DOY = day of year, BIO2_PT = site monthly mean diurnal range (max
temp - min temp) (°C), BIO8_PT = mean temperature of wettest quarter for site (°C),
BIO13_PT = precipitation of wettest month for site (mm), BIO15_PT = precipitation sea-
sonality (coefficient of variation) for site, BIO18_PT = precipitation of warmest quarter for
site (mm), BIO3MN_CA = isothermality (* 100) for catchment.
Metric RF % Predictors
Var
Functional richness (FRic) 28.44 CatArea+BIO15_PT
BIO8_PT+AREA
Rao's quadratic entropy (RaoQ) 19.62
+BIO18_PT
Distance-based functional diversity (dbFD) 59.49 BIO18_PT
DOY+BIO13_PT
Abundance-weighted frequency of soft body 71.02
+BIO3MN_CA
Abundance-weighted frequency of
collector-gatherer and brusher
47.01 SLPMIN_CA
LAT+SLPMAX_CA
Abundance-weighted frequency of predator 64.98
+ELEV_Mean_CA
ELEV_Mean_CA
No. of Diptera taxa 33.98
+SLPMIN_CA+BIO2_PT
The resulting 5 taxonomic metrics with the highest t-statistic also
differed between adjusted and unadjusted taxonomic MMIs. The met-
rics for the taxonomic adjusted MMI were number of Ephemeroptera
and Plecoptera taxa, number of Coleoptera taxa, number of Diptera
taxa, number of intolerant taxa, and % EPT (Ephemeroptera, Plecoptera,
Trichoptera) individuals. The metrics for the taxonomic unadjusted
MMI included number of EPT taxa, number of Coleoptera taxa, number
of intolerant taxa, % EPT individuals, and number of clinger taxa
(Table S2). Thus, the 2 taxonomic MMIs differed in only one metric: Dip-
tera taxa versus clinger taxa.
3.3. Performance of MMIs
Our functional-taxonomic adjusted MMI exhibited low bias, high
precision and responsiveness, and sensitivity to disturbance gradients
(Table 4, Fig. 2). Site-specific systematic bias of the functional-
taxonomic adjusted MMI was significantly reduced, with only 2.0% of
the variability in reference site scores associated with natural environ-
mental variation. In contrast, natural variation accounted for 12.9% of
the variability in the taxonomic adjusted MMI scores and 44.7% and
72% in the functional-taxonomic unadjusted and taxonomic unadjusted
MMI scores. Precision estimates of the functional-taxonomic adjusted
MMI (CV = 0.09) were the highest, followed by functional-taxonomic
unadjusted MMI (CV = 0.14), taxonomic adjusted MMI (CV = 0.19),
and taxonomic unadjusted MMI (CV = 0.36). Overall the functional-
taxonomic adjusted MMI performed the best in differentiating sites
under different disturbance levels, except for its ability to discriminate
intermediate- and most-disturbed sites. The percent of intermediate-
and most-disturbed sites assessed as being in non-reference condition
was 96% for the functional-taxonomic adjusted MMI. In contrast, we ob-
served the lowest sensitivity with the taxonomic adjusted MMI with
90% of test sites as being in non-reference condition.
Both functional-taxonomic adjusted and taxonomic adjusted MMIs
showed significant negative responses to the land use disturbance gra-
dient; however, the degree to which the MMIs responded to % catch-
ment land use varied (Figs. 3 and S4). Percent agriculture and urban
interactively contributed to both functional-taxonomic adjusted and
taxonomic adjusted MMI scores demonstrating the significant effect of
human land use on aquatic insect assemblage condition. However, rela-
tionships between % catchment land use was weaker with the taxo-
nomic adjusted MMI than with the functional-taxonomic adjusted
MMI. The partial dependence plot curves reveal that both functional-
 K. Chen et al. / Science of the Total Environment 691 (2019) 1005–1015 1011

Table 4
Results of MMI performance based on: bias (RF % Var: % of variation among least-disturbed site MMI scores associated with natural variables after random forest modeling), precision
(coefficient of variation [CV] of least-disturbed site MMI scores), responsiveness (mean difference for comparison between mean least-, intermediate-, and most-disturbed site MMI
scores), and sensitivity (% of intermediate- and most-disturbed sites evaluated as in non-reference condition). Adjusted MMI = MMI with natural variation adjustment, unadjusted
MMI = MMI without natural variation adjustment.

Bias CV Responsiveness Sensitivity % Relevancy

RF % Var
Least vs intermediate Intermediate vs most Least vs most % Urban % Agriculture

R2 adj. slope R2 adj. slope

Functional-taxonomic MMI
Adjusted 2.0 0.09 11.5 5.4 28.6 96 0.69 -6.6 0.48 -2.6
Unadjusted 44.7 0.14 11.4 5.9 22.4 94 - - - -
Taxonomic MMI
Adjusted 12.9 0.19 9.9 4.7 18.7 90 0.55 -5.7 0.34 -2.0
Unadjusted 72.0 0.36 10.9 4.3 13.0 92 - - - -

taxonomic adjusted and taxonomic adjusted MMIs decreased rapidly
across % urban ranging from 0 to 10% and across % agriculture ranging
from 0 to 20%, but less markedly to further disturbance (Fig. 3).
The adjusted R2 and slope of a linear regression of the functional-
taxonomic adjusted MMI relative to 0 to 10% urban and 0 to 20% agricul-
ture were higher than those for the taxonomic adjusted MMI given the
same range of % urban (adjusted R2 0.69 vs 0.55 and slope − 6.6 vs
−5.7) and % agriculture (adjusted R2 0.48 vs 0.34 and slope − 2.6 vs
−2.0) (Table 4, Fig. 3). Slopes between functional-taxonomic adjusted
and taxonomic adjusted MMIs as a function of % urban (p = 0.001)
and % agriculture (p = 0.005) were significantly different. Including
trait information strengthened the relationship between adjusted
MMIs and low levels of land use disturbance (Table 4, Fig. 3).
4. Discussion
Our functional-taxonomic adjusted MMI demonstrated that inclu-
sion of traits provided more meaningful and accurate assessment of
ecological conditions of temperate monsoonal streams than did the
other MMIs. Our results confirmed both our hypotheses that 1) the
functional metrics varied along natural gradients and adjusting metrics
for those gradients improved MMI performance, and 2) the functional-
taxonomic adjusted MMI was more sensitive to 0–10% urban and 0–20%
agriculture intensity than the taxonomic adjusted MMI. Although differ-
ence in factors (e.g., spatial extents, natural variability, candidate
metrics, and model selection) prevent direct comparisons, we believe
that our incorporation of functional metrics with natural variation ad-
justments are reasons for the better performance of our functional-
taxonomic adjusted MMI than that reported in recent published studies
(Table S3).
4.1. Natural gradient adjustment
Aquatic insect functional diversity and composition varied along
natural gradients. Similar results were reported from glacial streams
(e.g., Ilg and Castella, 2006), the tropics (e.g. Pease et al., 2012), and tem-
perate zones (e.g., Poff et al., 2010; Zuellig and Schmidt, 2012). Heino
et al. (2013) also found that assemblage traits developed from field
data sets varied along the longitudinal gradient of reference streams.
Variations in functional metrics were systematically correlated with
local- and catchment-scale geomorphology and climate variables in
our study. This indicates a high degree of natural environmental vari-
ability to which aquatic insects responded, as others have reported
(Poff, 1997; Heino, 2005). As suggested by Heino et al. (2007), Poff
et al. (2010), and Zuellig and Schmidt (2012), physiographic
(e.g., elevation), climatic (e.g., precipitation and temperature), and spa-
tial (e.g., latitude) gradients were primary factors explaining our func-
tional metric variations (Table 3). We also found that natural
environments explained higher variation in functional composition
metrics than in functional diversity indices. Those indices, which

Fig. 2. Ranges of functional-taxonomic and taxonomic adjusted and unadjusted MMIs for disturbance categories. Boxes encompass the 25th – 75th percentiles, horizontal lines are medians,
and circles are outliers. Adjusted MMI = MMI with natural variation adjustment, unadjusted MMI = MMI without natural variation adjustment.
1012 K. Chen et al. / Science of the Total Environment 691 (2019) 1005–1015
Fig. 3. Partial dependence plots from the Random Forest biota-stressor models and linear regression of MMIs on stressors showing how functional-taxonomic and taxonomic adjusted
MMIs scores vary along % urban and % agriculture gradients. Adjusted MMI = MMI with natural variation adjustment.
integrate multidimensional facets of different traits, likely diminish the
effects of natural environments to some extent, whereas functional
composition metrics are likely more sensitive to changes in natural en-
vironments. Similar results have been reported for taxonomic diversity
indices versus taxonomic composition responses to natural and anthro-
pogenic gradients (Stoddard et al., 2008; Marzin et al., 2012).
Accounting for the effects of natural variability improved the
functional-taxonomic adjusted MMI performance relative to that of un-
adjusted MMIs. This has been shown to be true for other MMIs across a
variety of stream types (e.g., tropical, subtropical, temperate), especially
regarding human alterations that co-vary with natural gradients (Cao
et al., 2007; Hawkins et al., 2010a; Chen et al., 2014; Silva et al., 2017).
Natural variation adjustment also improved the relevancy of our
functional-taxonomic adjusted MMI for detecting land use gradients,
and thus increased the possibility for causal interpretations of ecological
degradations. Similar results were reported by Vander Laan et al. (2013)
and Tang et al. (2016) for spatially extensive studies of USA streams.
Furthermore, this indicates the need to determine the degree to which
the natural gradients of reference sites represent those of intermediate
and most-disturbed sites. This is important because the regional repre-
sentativeness of the natural gradients of reference sites is necessary to
prevent the application of MMIs to inappropriate sites and invalid infer-
ences of ecological condition (Ode et al., 2016; Chen et al., 2019).
4.2. Metric selection
Functional diversity and composition are useful indicators for dis-
criminating least- and most-disturbed sites in temperate monsoonal
streams. Because traits are phenotypic characteristics linked to organ-
ism survival (Weiher, 2011), functional diversity was expected to be
lower at most-disturbed sites than at least-disturbed sites because
most-disturbed sites restrict the life forms capable of colonization. We
did observe that higher disturbance decreased functional diversity
scores (Fig. S4). Decreased FRic and RaoQ at intermediate and most-
disturbed sites also indicate that human disturbance reduced the
functional diversity of aquatic insect assemblages (Ding et al., 2017).
Decreased FRic simultaneously implies loss of suitable niches for species
(de Castro et al., 2018) as well as reduced taxonomic richness (Villéger
et al., 2008; Leitão et al., 2018) with increased disturbance. Changes in
relative abundance distributions can alter functional diversity without
changing species richness (Petchey and Gaston, 2006). RaoQ is thus a
promising functional diversity indicator for assessing effects of distur-
bance (Péru and Dolédec, 2010; Ding et al., 2017) because it measures
the pair-wise species dissimilarity in trait space weighted by relative
abundances (i.e., functional divergence) along stressor gradients (Leps
et al., 2006; Mouchet et al., 2010).
Functional composition metrics responses varied with increased dis-
turbance. For example, frequency of soft body occurrence responded
positively to increased disturbance, whereas frequency of predator oc-
currence responded negatively to increased disturbance (Fig. S4). In-
creasing farm and urban land use intensity results in more taxa having
traits of resilience or resistance to sedimentation (e.g., soft body;
Lange et al., 2014; Ding et al., 2017). The greater farm and urban land
use, the higher the sediment inputs (Allan, 2004). Therefore, taxa living
in depositional areas dominated by fine sediments tend to have soft
bodies to tolerate or thrive in unstable sediments (Stoddard et al.,
2008); on the other hand, those same conditions tend to reduce the fre-
quency of predators, which generally require coarse sediments (Fierro
et al., 2018).
4.3. Early warning signals of land use impacts
Land use intensification can greatly reduce species richness and eco-
system functioning (Davies and Jackson, 2006; Flynn et al., 2009).
Therefore, MMIs based on species traits and abundances are more likely
to be early-warning disturbance indicators than taxonomic composition
alone (Dolédec et al., 2006; Mouillot et al., 2013). Coleman et al. (2015)
concluded that traits elucidated early detection of environmental
changes in marine fish assemblages, when traditional taxonomic
MMIs did not. Our results demonstrated that our functional-
 K. Chen et al. / Science of the Total Environment 691 (2019) 1005–1015
taxonomic adjusted MMI was more sensitive to low-level land use dis-
turbance than a taxonomic adjusted MMI (Table 4, Fig. 3). Leitão et al.
(2016) reported that significant reductions in functional diversity
could be related to disproportionate extirpation of rare and sensitive
taxa. In our study, land use intensification caused early loss of species
with unique traits (Fig. S4 and Fig. 3), whereas other species with differ-
ing traits persisted, perhaps via a habitat filtering process (Flynn et al.,
2009). Thus, functional metrics change more rapidly than taxonomic
metrics because increased land use intensity greatly affects functional
composition and diversity even if few species are lost from the system
(de Castro et al., 2018). When land use intensification increases further,
functionally generalist species persist without changing functional
diversity.
Functional richness is sensitive to proportions of species losses
(Pakeman, 2014). Mouillot et al. (2013) predicted that FRic would not
be sensitive to disturbance intensity because low disturbance intensity
may deplete populations of rare species with vulnerable combinations
of traits without modifying species composition or species or functional
richness. However, our results conflict with their predictions. de Castro
et al. (2018) also reported reduced response in FRic beyond intermedi-
ate levels of disturbance gradients, but to a lesser degree than our re-
sults. We suspect that adjusting for natural variation changed the
response trajectory of FRic along land use gradients just as it did our
other trait metrics, whereas the predictions of Mouillot et al. (2013)
and the results of de Castro et al. (2018) were based on covarying effects
of natural and human disturbances.
Both traits and species abundances affect the ability to detect
early assemblage-level changes along disturbance gradients. RaoQ,
abundance-weighted frequency of soft bodies, and abundance-
weighted frequency of predators are all calculated with both trait and
abundance data. Rare or specialist species abundance with vulnerable
combinations of traits that are sensitive to low disturbance intensities
decline (Fig. S5; Mouillot et al., 2013; de Oliveira-Junior et al., 2017;
Firmiano et al., 2017). However, generalist or tolerant species that thrive
under greater levels of disturbance increase (de Oliveira-Junior et al.,
2017; Firmiano et al., 2017).
Nonetheless, functional composition metrics such as abundance-
weighted frequencies of traits are likely to vary among basins or
ecoregions when the types and degree of natural and anthropogenic
disturbances are regionally specific. More biological and ecological in-
formation from different regions with variable landscape and climate
characteristics could improve our ability to assess the functional condi-
tion of lotic ecosystems. Further research on sites that are greatly dis-
turbed (e.g., N50% urban or agriculture or mined) would greatly
strengthen our understanding of the total amount of degradation that
an MMI incorporating functional metrics could meaningfully interpret.
5. Conclusions
Incorporating functional diversity and composition metrics provides
a more comprehensive assessment of ecological condition and manage-
ment options than does an MMI based on taxonomic metrics. Our re-
sults confirmed that 1) functional composition and diversity (and
taxonomic variables) co-vary with natural gradients, and 2) functional
metrics and functional-taxonomic MMIs better distinguish least-
disturbed from intermediate-disturbed sites than purely taxonomic
metrics and MMIs. We also found that functional-taxonomic MMIs
can diagnose early changes in aquatic assemblages. This is likely be-
cause sensitive traits and sensitive species disappear early in ecosystem
degradation processes whether or not species show functional redun-
dancy. MMIs adjusted for natural variability and incorporating func-
tional metrics (i.e., functional composition and diversity) improve our
ability to: 1) identify how human pressures modify the functional struc-
ture of assemblages, 2) identify patterns and processes in freshwater
ecosystems, and 3) better plan pre-emptive management actions and
conservation initiatives.
1013
Author contributions
KC conceived the research ideas. KC and ARR led the analyses and
writing. BW provided the data. All co-authors contributed to the writing
of all versions of the manuscript.
Acknowledgements
This research was funded by the National Natural Science Founda-
tion of China (NSFC, no. 51509159 and 41771052). We thank the
Zhejiang Environmental Monitoring Center for water chemistry analy-
ses, and colleagues at the Laboratory of Aquatic Insects and Stream
Ecology of Nanjing Agricultural University for assistance with macroin-
vertebrate sampling and processing. We thank all 7 anonymous re-
viewers for the many edits, comments and suggestions that improved
the manuscript's clarity.
Appendix A. Supplementary data
Supplementary data to this article can be found online at https://doi.
org/10.1016/j.scitotenv.2019.07.047.
References
Allan, J.D., 2004. Landscapes and riverscapes: the influence of land use on stream ecosys-
tems. Annu. Rev. Ecol. Evol. Syst. 35, 257–284. https://doi.org/10.1146/annurev.
ecolsys.35.120202.110122.
Anderson, M.J., 2006. Distance-based tests for homogeneity of multivariate
dispersions. Biometrics 62, 245–253. https://doi.org/10.1111/j.1541-
0420.2005.00440.x.
Barbour, M.T., Gerritsen, J., Snyder, B.D., Stribling, J.B., 1999. Rapid Bioassessment Proto-
cols for Use in Streams and Wadeable Rivers: Periphyton, Benthic Macroinverte-
brates, and Fish. 2nd edition. U.S. Environmental Protection Agency, Office of
Water, Washington DC EPA 841-0B-99-002.
Bêche, L.A., Statzner, B., 2009. Richness gradients of stream invertebrates across the USA:
taxonomy- and trait-based approaches. Biodivers. Conserv. 18, 3909–3930. https://
doi.org/10.1007/s10531-009-9688-1.
Beyer, H.L., 2014. Geospatial Modeling Environment (Version 0.7.4). Available online at.
http://www.spatialecology.com/gme.
Blocksom, K.A., 2003. A performance comparison of metric scoring methods for a
multimetric index for mid-Atlantic highlands streams. Environ. Manag. 31 (5),
670–682. https://doi.org/10.1007/s00267-002-2949-3.
Bonada, N., Prat, N., Resh, V.H., Statzner, B., 2006. Developments in aquatic insect biomon-
itoring: a comparative analysis of recent approaches. Annu. Rev. Entomol. 51,
495–523. https://doi.org/10.1146/annurev.ento.51.110104.151124.
Bonada, N., Dolédec, S., Statzner, B., 2007. Taxonomic and biological trait differences of
stream macroinvertebrate communities between mediterranean and temperate re-
gions, implications for future climatic scenarios. Glob. Chang. Biol. 13, 1658–1671.
https://doi.org/10.1111/j.1365-2486.2007.01375.x.
Botta-Dukát, Z., 2005. Rao's quadratic entropy as a measure of functional diversity based
on multiple traits. J. Veg. Sci. 16, 533–540. https://doi.org/10.1111/j.1654-1103.2005.
tb02393.x.
Bowman, M.F., Somers, K.M., 2006. Evaluating a novel test site analysis (TSA) bioassess-
ment approach. J. N. Am. Benthol. Soc. 25, 712–727. https://doi.org/10.1899/0887-
3593(2006)25[712:EANTSA]2.0.CO;2.
Cao, Y., Hawkins, C.P., Olson, J.R., Kosterman, M.A., 2007. Modeling natural environmental
gradients improves the accuracy and precision of diatom-based indicators. J. N. Am.
Benthol. Soc. 26, 566–585. https://doi.org/10.1899/06-078.1.
de Castro, D.M.P., Dolédec, S., Callisto, M., 2018. Land cover disturbance homogenizes
aquatic insect functional structure in neotropical savanna streams. Ecol. Indic. 84,
573–582. https://doi.org/10.1016/j.ecolind.2017.09.030.
Chen, K., Hughes, R.M., Xu, S., Zhang, J., Cai, D., Wang, B.X., 2014. Evaluating performance
of macroinvertebrate-based adjusted and unadjusted multi-metric indices (MMI)
using multi-season and multi-year samples. Ecol. Indic. 36, 142–151. https://doi.
org/10.1016/j.ecolind.2013.07.006.
Chen, K., Hughes, R.M., Wang, B.X., 2015. Effects of fixed-count size on macroinvertebrate
richness, site separation, and bioassessment of Chinese monsoonal streams. Ecol.
Indic. 53, 162–170. https://doi.org/10.1016/j.ecolind.2015.01.011.
Chen, K., Hughes, R.M., Brito, J.G., Leal, C.G., Leitão, R.P., de Oliveira-Júnior, J.M., de Oliveira,
V.C., Dias-Silva, K., Ferraz, S.F.B., Ferreira, J., 2017. A multi-assemblage, multi-metric
biological condition index for eastern Amazonia streams. Ecol. Indic. 78, 48–61.
https://doi.org/10.1016/j.ecolind.2017.03.003.
Chen, K., Olson, J.R., Vander Laan, J.J., Hill, R.A., Wang, B., Hawkins, C.P., 2019. Improving
the performance of ecological indices by balancing reference site quality and repre-
sentativeness. Hydrobiologia 837, 177–194. https://doi.org/10.1007/s10750-019-
3970-3.
Chinnayakanahalli, K., Kroeber, C., Hill, R.A., Tarboton, D.G., Olson, J.R., Hawkins, C.P., 2006.
The Multi-Watershed Delineation Tool, GIS Software in Support of Regional Water-
shed Analyses. Utah State University, Logan.
1014 K. Chen et al. / Science of the Total Environment 691 (2019) 1005–1015
Coleman, M.A., Bates, A.E., Stuart-Smith, R.D., Malcolm, H.A., Harasti, D., Jordan, A., Knott,
N.A., Edgar, G.J., Kelaher, B.P., 2015. Functional traits reveal early responses in marine
reserves following protection from fishing. Divers. Distrib. 21 (8), 876–887. https://
doi.org/10.1111/ddi.12309.
Colzani, E., Siqueira, T., Suriano, M.T., Roque, F.O., 2013. Responses of aquatic insect func-
tional diversity to landscape changes in Atlantic forest. Biotropica 45, 343–350.
https://doi.org/10.1111/btp.12022.
Culp, J.M., Armanini, D.G., Dunbar, M.J., Orlofske, J.M., Poff, N.L., Pollard, A.I., Yates, A.G.,
Hose, G.C., 2011. Incorporating traits in aquatic biomonitoring to enhance causal di-
agnosis and prediction. Integr. Environ. Assess. Manag. 7, 187–197. https://doi.org/
10.1002/ieam.128.
Cutler, D.R., Edwards Jr., T.C., Beard, K.H., Cutler, A., Hess, K.T., Gibson, J., Lawler, J.J., 2007.
Random forests for classification in ecology. Ecology 88, 2783–2792. https://doi.org/
10.1890/07-0539.1.
Davies, S.P., Jackson, S.K., 2006. The biological condition gradient, a descriptive model for
interpreting change in aquatic ecosystems. Ecol. Appl. 16, 1251–1266. https://doi.org/
10.1890/1051-0761(2006)016[1251:TBCGAD]2.0.CO;2.
Ding, N., Yang, W., Zhou, Y., González-Bergonzoni, I., Zhang, J., Chen, K., Vidal, N.,
Jeppesen, E., Liu, Z.W., Wang, B.X., 2017. Different responses of functional traits
and diversity of stream macroinvertebrates to environmental and spatial
factors in the Xishuangbanna watershed of the upper Mekong River
basin, China. Sci. Total Environ. 574, 288–299. https://doi.org/10.1016/j.
scitotenv.2016.09.053.
Dolédec, S., Statzner, B., 2010. Responses of freshwater biota to human disturbances, con-
tribution of J-NABS to developments in ecological integrity assessments. J. N. Am.
Benthol. Soc. 29, 286–311. https://doi.org/10.1899/08-090.1.
Dolédec, S., Statzner, B., Bournard, M., 1999. Species traits for future biomonitoring across
ecoregions, patterns along a human-impacted river. Freshw. Biol. 42, 737–758.
https://doi.org/10.1046/j.1365-2427.1999.00509.x.
Dolédec, S., Phillips, N., Scarsbrook, M., Riley, R.H., Townsend, C.R., 2006. Comparison of
structural and functional approaches to determining landuse effects on grassland
stream invertebrate communities. J. N. Am. Benthol. Soc. 25(1), 44–60. https://doi.
org/10.1899/0887-3593(2006)25[44:COSAFA]2.0.CO;2.
Dolédec, S., Phillips, N., Townsend, C., 2011. Invertebrate community responses to land
use at a broad spatial scale, trait and taxonomic measures compared in New
Zealand rivers. Freshw. Biol. 56, 1670–1688. https://doi.org/10.1111/j.1365-
2427.2011.02597.x.
EPBC (Environmental Protection Bureau the People's Republic of China), 2002. Standard
Methods for the Monitoring and Analysis on Water and Wastewater. 4th edn. Chi-
nese Environmental Science Press, Beijing.
Esselman, P.C., Infante, D.M., Wang, L., Cooper, A.R., Wieferich, D., Tsang, Y.P., Thornbrugh,
D.J., Taylor, W.W., 2013. Regional fish community indicators of landscape disturbance
to catchments of the conterminous United States. Ecol. Indic. 26, 163–173. https://
doi.org/10.1016/j.ecolind.2012.10.028.
Faith, D.P., 1992. Conservation evaluation and phylogenetic diversity. Biol. Conserv. 61:
1–10. https://doi.org/10.1016/0006-3207(92)91201–3.
Fierro, P., Arismendi, I., Hughes, R.M., Valdovinos, C., Jara-Flores, A., 2018. A benthic mac-
roinvertebrate multimetric index for Chilean Mediterranean streams. Ecol. Indic. 91,
13–23. https://doi.org/10.1016/j.ecolind.2018.03.074.
Firmiano, K.R., Ligeiro, R., Macedo, D.R., Juen, L., Hughes, R.M., Callisto, M., 2017.
Mayfly bioindicator thresholds for several anthropogenic disturbances in neo-
tropical savanna streams. Ecol. Indic. 74, 276–284. https://doi.org/10.1016/j.
ecolind.2016.11.033.
Flynn, D.F., Gogol-Prokurat, M., Nogeire, T., Molinari, N., Richers, B.T., Lin, B.B., Simpson, N.,
Mayfield, M.M., DeClerck, F., 2009. Loss of functional diversity under land use inten-
sification across multiple taxa. Ecol. Lett. 12, 22–33. https://doi.org/10.1111/j.1461-
0248.2008.01255.x.
Grumman, N., 2002. Algorithm and Data User Manual (ADUM) for the Special Sensor Mi-
crowave Imager/Sounder (SSMIS). Technical Report. http://www.ncdc.noaa.gov/oa/
rsad/ssmi/swath/adum-ssmis-description.pdf.
Hastie, T.J., Tibshirani, R.J., Friedman, J.H., 2001. The Elements of Statistical Learning: Data
Mining, Inference, and Prediction. Springer, New York, New York.
Hawkins, C.P., Cao, Y., Roper, B., 2010a. Method of predicting reference condition biota af-
fects the performance and interpretation of ecological indices. Freshw. Biol. 55,
1066–1085. https://doi.org/10.1111/j.1365-2427.2009.02357.x.
Hawkins, C.P., Olson, J.R., Hill, R.A., 2010b. The reference condition, predicting bench-
marks for ecological and water-quality assessments. J. N. Am. Benthol. Soc. 29,
312–343. https://doi.org/10.1899/09-092.1.
Heino, J., 2005. Functional biodiversity of macroinvertebrate assemblages along major
ecological gradients of boreal headwater streams. Freshw. Biol. 50, 1578–1587.
https://doi.org/10.1111/j.1365-2427.2005.01418.x.
Heino, J., Mykrä, H., Kotanen, J., Muotka, T., 2007. Ecological filters and variability in
stream macroinvertebrate communities, do taxonomic and functional structure fol-
low the same path? Ecography 30, 217–230. https://doi.org/10.1111/j.0906-
7590.2007.04894.x.
Heino, J., Schmera, D., Erős, T., 2013. A macroecological perspective of trait patterns in
stream communities. Freshw. Biol. 58, 1539–1555. https://doi.org/10.1111/
fwb.12164.
Helson, J.E., Williams, D.D., 2013. Development of a macroinvertebrate multimetric index
for the assessment of low-land streams in the neotropics. Ecol. Indic. 29, 167–178.
https://doi.org/10.1016/j.ecolind.2012.12.030.
Hering, D., Feld, C.K., Moog, O., Ofenboeck, T., 2006. Cook book for the develop-
ment of a multimetric index for biological condition of aquatic ecosystems:
experiences from the European AQEM and STAR projects and related initia-
tives. Hydrobiologia 566, 311–324. https://doi.org/10.1007/s10750-006-
0087-2.
Hijmans, R.J., Cameron, S.E., Parra, J.L., Jones, P.G., Jarvis, A., 2005. Very high resolution in-
terpolated climate surfaces for global land areas. Int. J. Climatol. 25, 1965–1978.
https://doi.org/10.1002/joc.1276 Available online at:http://www.worldclim.org.
Hughes, R.M., Peck, D.V., 2008. Acquiring data for large aquatic resource surveys: the art
of compromise among science, logistics, and reality. J. N. Am. Benthol. Soc. 27,
837–859. https://doi.org/10.1899/08-028.1.
Hussin, W.R.W., Cooper, K.M., Froján, C.R.B., Defew, E.C., Paterson, D.M., 2012. Impacts of
physical disturbance on the recovery of a macrofaunal community: a comparative
analysis using traditional and novel approaches. Ecol. Indic. 12, 37–45. https://doi.
org/10.1016/j.ecolind.2011.03.016.
Ilg, C., Castella, E., 2006. Patterns of macroinvertebrate traits along three glacial
stream continuums. Freshw. Biol. 51, 840–853. https://doi.org/10.1111/j.1365-
2427.2006.01533.x.
Kembel, S.W., 2010. An introduction to the picante package. http://picante.r-forge.r-pro-
ject.org/picante-intro.pdf.
Kuznetsova, A., Brockhoff, P.B., Christensen, R.H.B., 2017. lmerTest package: tests in linear
mixed effects models. J. Stat. Softw. 82 (13), 1–26. https://doi.org/10.18637/jss.v082.
i13.
Laliberté, E., Legendre, P., 2010. A distance-based framework for measuring func-
tional diversity from multiple traits. Ecology 91, 299–305. https://doi.org/
10.1890/08-2244.1.
Land Processes Distributed Active Archive Center, 2013. Land Surface Temperature &
Emissivity. Available online at. https://lpdaac.usgs.gov.
Lange, K., Townsend, C.R., Matthaei, C.D., 2014. Can biological traits of stream inverte-
brates help disentangle the effects of multiple stressors in an agricultural catchment?
Freshw. Biol. 59, 2431–2446. https://doi.org/10.1111/fwb.12437.
Larsen, S., Ormerod, S., 2010. Combined effects of habitat modification on trait composi-
tion and species nestedness in river invertebrates. Biol. Conserv. 143, 2638–2646.
https://doi.org/10.1016/j.biocon.2010.07.006.
Larsen, S., Pace, G., Ormerod, S.J., 2011. Experimental effects of sediment deposition on the
structre and function of macroinvertebrate assemblages in temperate streams. River
Res. Appl. 27, 257–267. https://doi.org/10.1002/rra.1361.
Leitão, R.P., Zuanon, J., Villéger, S., Williams, S.E., Baraloto, C., Fortunel, C., Mendonça, F.P.,
Mouillot, D., 2016. Rare species contribute disproportionately to the functional struc-
ture of species assemblages. Proc. R. Soc. B 283, 20160084.. https://doi.org/10.1098/
rspb.2016.0084.
Leitão, R.P., Zuanon, J., Mouillot, D., Leal, C.G., Hughes, R.M., Kaufmann, P.R., Villéger, S.,
Pompeu, P.S., Kasper, D., de Paula, F.R., Ferraz, S.F.B., Gardner, T., 2018. Disentangling
the pathways of land use impacts on the functional structure of fish assemblages in
Amazon streams. Ecography 41, 219–232. https://doi.org/10.1111/ecog.02845.
Leps, J., de Bello, F., Lavorel, S., Berman, S., 2006. Quantifying and interpreting functional
diversity of natural communities, practical considerations matter. Preslia 78,
481–501.
Liaw, A., Wiener, M., 2002. Classification and Regression by randomForest. vol. 2. R news,
pp. 18–22.
Liu, S., Xie, G., Wang, L., Cottenie, K., Liu, D., Wang, B.X., 2016. Different roles of environ-
mental variables and spatial factors in structuring stream benthic diatom and macro-
invertebrate in Yangtze River Delta, China. Ecol. Indic. 61, 602–611. https://doi.org/
10.1016/j.ecolind.2015.10.011.
Lloyd, N.J., MacNally, R., Lake, P. S., 2006. Spatial scale of autocorrelation of assemblages of
benthic invertebrates in two upland rivers in south-eastern Australia and its implica-
tions for biomonitoring and impact assessment in streams. Environ. Monit. Assess.
115: 69–85. https://doi.org/10.1007/s10661-006-5253-5.
Lorenz, A., Kirchner, L., Hering, D., 2004. ‘Electronic subsampling’ of macrobenthic sam-
ples: how many individuals are needed for a valid assessment result? Hydrobiologia
516, 299–312. https://doi.org/10.1023/B:HYDR.0000025272.05793.00.
Macedo, D.R., Hughes, R.M., Ferreira, W.R., Firmiano, K.R., Silva, D.R., Ligeiro, R., Kaufmann,
P.R., Callisto, M., 2016. Development of a benthic macroinvertebrate multimetric
index (MMI) for Neotropical Savanna headwater streams. Ecol. Indic. 64, 132–141.
https://doi.org/10.1016/j.ecolind.2015.12.019.
Marzin, A., Archaimbault, V., Belliard, J., Chauvin, C., Delmas, F., Pont, D., 2012. Ecological
assessment of running waters: do macrophytes, macroinvertebrates, diatoms and
fish show similar responses to human pressures? Ecol. Indic. 23, 56–65. https://doi.
org/10.1016/j.ecolind.2012.03.010.
Mason, N.W., Mouillot, D., Lee, W.G., Wilson, J.B., 2005. Functional richness, functional
evenness and functional divergence, the primary components of functional diversity.
Oikos 111, 112–118. https://doi.org/10.1111/j.0030-1299.2005.13886.x.
Mazor, R., Rehn, A., Ode, P., Engeln, M., Schiff, K., Stein, E., Gillett, D., Herbst, D., Hawkins,
C.P., 2016. Bioassessment in complex environments, designing an index for consis-
tent meaning in different settings. Freshw. Sci. 35, 249–271. https://doi.org/
10.1086/684130.
Menezes, S., Baird, D.J., Soares, A.M., 2010. Beyond taxonomy: a review of macroinverte-
brate trait-based community descriptors as tools for freshwater biomonitoring.
J. Appl. Ecol. 47, 711–719. https://doi.org/10.1111/j.1365-2664.2010.01819.x.
Morse, J.C., Yang, L.F., Tian, L.X., 1994. Aquatic Insects of China Useful for Monitoring
Water Quality. HoHai University Press, Nanjing.
Mouchet, M.A., Villéger, S., Mason, N.W.H., Mouillot, D., 2010. Functional diversity
measures: an overview of their redundancy and their ability to discriminate
community assembly rules. Funct. Ecol. 24, 867–876. https://doi.org/10.1111/
j.1365-2435.2010.01695.x.
Mouillot, D., Graham, N.A., Villéger, S., Mason, N.W., Bellwood, D.R., 2013. A functional ap-
proach reveals community responses to disturbances. Trends Ecol. Evol. 28, 167–177.
https://doi.org/10.1016/j.tree.2012.10.004.
Mouillot, D., Villéger, S., Parravicini, V., Kulbicki, M., Arias-Gonzalez, J.E., Bender, M.,
Chabanet, P., Floeter, S.R., Friedlander, A., Vigliola, L., Bellwood, D.R., 2014. Functional
over-redundancy and high functional vulnerability in global fish faunas on tropical
 K. Chen et al. / Science of the Total Environment 691 (2019) 1005–1015
reefs. Proc. Natl. Acad. Sci. 111, 13757–13762. https://doi.org/10.1073/
pnas.1317625111.
Moya, N., Hughes, R.M., Dominguez, E., Gibon, F.M., Goita, E., Oberdorff, T., 2011.
Macroinvertebrate-based multimetric predictive models for measuring the biotic
condition of Bolivian streams. Ecol. Indic. 11, 840–847. https://doi.org/10.1016/j.
ecolind.2010.10.012.
Ode, P.R., Rehn, A.C., Mazor, R.D., Schiff, K.C., Stein, E.D., May, J.T., Brown, L.R., Herbst, D.B.,
Gillett, D., Lunde, K., Hawkins, C.P., 2016. Evaluating the adequacy of a reference-site
pool for ecological assessments in environmentally complex regions. Freshw. Sci. 35,
237–248. https://doi.org/10.1086/684003.
de Oliveira-Junior, J.M.B., de Marco-Junior, P., Dias-Silva, K., Leitão, R.P., Leal, C.G., Pompeu,
P.S., Gardner, T.A., Hughes, R.M., Juen, L., 2017. Effects of human disturbance and ri-
parian conditions on Odonata (Insecta) assemblages in eastern Amazon basin
streams. Limnologica 66, 31–39. https://doi.org/10.1016/j.limno.2017.04.007.
Orlofske, J.M., Baird, D.J., 2014. Incorporating continuous trait variation into biomonitor-
ing assessments by measuring and assigning trait values to individuals or taxa.
Freshw. Biol. 59, 477–490. https://doi.org/10.1111/fwb.12279.
Pakeman, R.J., 2014. Functional trait metrics are sensitive to the completeness of the spe-
cies' trait data? Methods Ecol. Evol. 5, 9–15. https://doi.org/10.1111/2041-
210X.12136.
Pease, A.A., Gonzalez-Díaz, A.A., Rodiles-Hernández, R.O.C.Í.O., Winemiller, K.O., 2012.
Functional diversity and trait–environment relationships of stream fish assemblages
in a large tropical catchment. Freshw. Biol. 57 (5), 1060–1075. https://doi.org/
10.1111/j.1365-2427.2012.02768.x.
Pedley, S.M., Dolman, P.M., 2014. Multi-taxa trait and functional responses to physical dis-
turbance. J. Anim. Ecol. 83(6), 1542–1552. https://doi.org/10.1111/1365-2656.12249.
Péru, N., Dolédec, S., 2010. From compositional to functional biodiversity metrics in bioas-
sessment, a case study using stream macroinvertebrate communities. Ecol. Indic. 10,
1025–1036. https://doi.org/10.1016/j.ecolind.2010.02.011.
Petchey, O.L., Gaston, K.J., 2002. Functional diversity (FD), species richness and com-
munity composition. Ecol. Lett. 5, 402–411. https://doi.org/10.1046/j.1461-
0248.2002.00339.x.
Petchey, O.L., Gaston, K.J., 2006. Functional diversity, back to basics and looking forward.
Ecol. Lett. 9, 741–758. https://doi.org/10.1111/j.1461-0248.2006.00924.x.
Petchey, O.L., Evans, K.L., Fishburn, I.S., Gaston, K.J., 2007. Low functional diversity and no
redundancy in British avian assemblages. J. Anim. Ecol. 76, 977–985. https://doi.org/
10.1111/j.1365-2656.2007.01271.x.
Poff, N.L., 1997. Landscape filters and species traits, towards mechanistic understanding
and prediction in stream ecology. J. N. Am. Benthol. Soc. 16, 391–409. https://doi.
org/10.2307/1468026.
Poff, N.L., Olden, J.D., Vieira, N.K., Finn, D.S., Simmons, M.P., Kondratieff, B.C., 2006. Func-
tional trait niches of North American lotic insects, traits-based ecological applications
in light of phylogenetic relationships. J. N. Am. Benthol. Soc. 25, 730–755. https://doi.
org/10.1899/0887-3593(2006)025[0730:FTNONA]2.0.CO;2.
Poff, N.L., Pyne, M.I., Bledsoe, B.P., Cuhaciyan, C.C., Carlisle, D.M., 2010. Developing link-
ages between species traits and multiscaled environmental variation to explore vul-
nerability of stream benthic communities to climate change. J. N. Am. Benthol. Soc.
29, 1441–1458. https://doi.org/10.1899/10-030.1.
Pont, D., Hughes, R.M., Whittier, T.R., Schmutz, S., 2009. A predictive index of biotic integ-
rity model for aquatic-vertebrate assemblages of western U.S. streams. Trans. Am.
Fish. Soc. 138, 292–305. https://doi.org/10.1577/T07-277.1.
Poos, M.S., Walker, S.C., Jackson, D.A., 2009. Functional-diversity indices can be driven by
methodological choices and species richness. Ecology 90, 341–347. https://doi.org/
10.1890/08-1638.1.
Pyron, M., Lauer, T.E., LeBlanc, D., Weitzel, D., Gammon, J.R., 2008. Temporal and spatial
variation in an index of biological integrity for the middle Wabash River, Indiana.
Hydrobiologia 600, 205–214. https://doi.org/10.1007/s10750-007-9232-9.
R Development Core Team, 2014. R: A Language and Environment for Statistical Comput-
ing. R Foundation for Statistical Computing, Vienna, Austria.
Rao, C.R., 1982. Diversity and dissimilarity coefficients, a unified approach. Theor. Popul.
Biol. 21, 24–43. https://doi.org/10.1016/0040-5809(82)90004-1.
Ricotta, C., Bello, F., Moretti, M., Caccianiga, M., Cerabolini, B.E.L., Pavoine, S., 2016. Mea-
suring the functional redundancy of biological communities: a quantitative guide.
Methods Ecol. Evol. 7, 1386–1395. https://doi.org/10.1111/2041-210X.12604.
Saito, V.S., Siqueira, T., Fonseca-Gessner, A.A., 2015. Should phylogenetic and functional
diversity metrics compose macroinvertebrate multimetric indices for stream bio-
monitoring? Hydrobiologia 745, 167–179. https://doi.org/10.1007/s10750-014-
2102-3.
Schleuter, D., Daufresne, M., Massol, F., Argillier, C., 2010. A user's guide to functional di-
versity indices. Ecol. Monogr. 80, 469–484. https://doi.org/10.1890/08-2225.1.
Schmera, D., Heino, J., Podani, J., Erős, T., Dolédec, S., 2017. Functional diversity, a review of
methodology and current knowledge in freshwater macroinvertebrate research.
Hydrobiologia 787, 27–44. https://doi.org/10.1007/s10750-016-2974-5.
1015
Silva, D.R., Herlihy, A.T., Hughes, R.M., Callisto, M., 2017. An improved macroinvertebrate
multimetric index for the assessment of wadeable streams in the neotropical sa-
vanna. Ecol. Indic. 81, 514–525. https://doi.org/10.1016/j.ecolind.2017.06.017.
Statzner, B., Hoppenhaus, K., Arens, M.F., Richoux, P., 1997. Reproductive traits, habitat
use and templet theory, a synthesis of world-wide data on aquatic insects. Freshw.
Biol. 38, 109–135. https://doi.org/10.1046/j.1365-2427.1997.00195.x.
Statzner, B., Bêche, L.A., 2010. Can biological invertebrate traits resolve effects of multiple
stressors on running water ecosystems. Freshw. Biol. 55, 80–119. https://doi.org/
10.1111/j.1365-2427.2009.02369.x.
Stoddard, J.L., Larsen, D.P., Hawkins, C.P., Johnson, R.K., Norris, R.H., 2006. Setting expecta-
tions for the ecological condition of streams, the concept of reference condition. Ecol.
Appl. 16, 1267–1276. https://doi.org/10.1890/1051-0761(2006)016[1267:SEFTEC]
2.0.CO;2.
Stoddard, J.L., Herlihy, A.T., Peck, D.V., Hughes, R.M., Whittier, T.R., Tarquinio, E., 2008. A
process for creating multimetric indices for large-scale aquatic surveys. J. N. Am.
Benthol. Soc. 27, 878–891. https://doi.org/10.1899/08-053.1.
Tang, T., Stevenson, R.J., Infante, D.M., 2016. Accounting for regional variation in both nat-
ural environment and human disturbance to improve performance of multimetric in-
dices of lotic benthic diatoms. Sci. Total Environ. 568, 1124–1134. https://doi.org/
10.1016/j.scitotenv.2016.03.060.
Tomanova, S., Usseglio-Polatera, P., 2007. Patterns of benthic community traits in neo-
tropical streams: relationship to mesoscale spatial variability. Fundamental and Ap-
plied Limnology/Archiv für Hydrobiologie 170, 243–255. https://doi.org/10.1127/
1863-9135/2007/0170-0243.
Usseglio-Polatera, P., Bournaud, M., Richoux, P., Tachet, H., 2000. Biological and ecological
traits of benthic freshwater macroinvertebrates, relationships and definition of
groups with similar traits. Freshw. Biol. 43, 175–205. https://doi.org/10.1046/
j.1365-2427.2000.00535.x.
Van Sickle, J., 2010. Correlated metrics yield multimetric indices with inferior perfor-
mance. Trans. Am. Fish. Soc. 139, 1802–1817. https://doi.org/10.1577/T09-204.1.
Van Sickle, J., Hughes, R.M., 2000. Classification strengths of ecoregions, basins and geo-
graphic clusters for aquatic vertebrates in Oregon. J. N. Am. Benthol. Soc. 19:
370–384. https://doi.org/https://doi.org/10.2307/1468101.
Vander Laan, J.J., Hawkins, C.P., 2014. Enhancing the performance and interpretation of
freshwater biological indices, an application in arid zone streams. Ecol. Indic. 36,
470–482. https://doi.org/10.1016/j.ecolind.2013.09.006.
Vander Laan, J.J., Hawkins, C.P., Olson, J.R., Hill, R.A., 2013. Linking land use, in-stream
stressors, and biological condition to infer causes of regional ecological impairment
in streams. Freshw. Sci. 32, 801–820. https://doi.org/10.1899/12-186.1.
Vandewalle, M., De Bello, F., Berg, M.P., Bolger, T., Dolédec, S., Dubs, F., Feld, C.K.,
Harrington, R., Harrison, P.A., Lavorel, S., da Silva, P.M., Moretti, M., Niemelä, J.,
Santos, P., Sattler, T., Paulo Sousa, J., Sykes, M.T., Vanbergen, A.J., Woodcock, B.A.,
2010. Functional traits as indicators of biodiversity response to land use changes
across ecosystems and organisms. Biodivers. Conserv. 19 (10), 2921–2947. https://
doi.org/10.1007/s10531-010-9798-9.
Verberk W.C.E.P., van Noordwijk, C.G.E., Hildrew, A.G., 2013. Delivering on a promise: in-
tegrating species traits to transform descriptive community ecology into a predictive
science. Freshw. Sci. 322, 531–547. https://doi.org/10.1899/12-092.1.
Villéger, S., Mason, N.W., Mouillot, D., 2008. New multidimensional functional diversity
indices for a multifaceted framework in functional ecology. Ecology 89, 2290–2301.
https://doi.org/10.1890/07-1206.1.
Villéger, S., Miranda, J.R., Hernández, D.F., Mouillot, D., 2010. Contrasting changes in tax-
onomic vs. functional diversity of tropical fish communities after habitat degradation.
Ecol. Appl. 20, 1512–1522. https://doi.org/10.1890/09-1310.1.
Wang, B.X., Liu, D., Liu, S., Zhang, Y., Lu, D., Wang, L., 2012. Impacts of urbanization on
stream habitats and macroinvertebrate communities in the tributaries of Qiangtang
River, China. Hydrobiologia 680, 39–51. https://doi.org/10.1007/s10750-011-0899-6.
Weiher, E., 2011. A primer of trait and functional diversity. In: Magurran, A.E., McGill, B.J.
(Eds.), Biological Diversity: Frontiers in Measurement and Assessment. Oxford Uni-
versity Press, Oxford, pp. 175–193.
Willmott, C.J., Matsuura, K., 2012. Terrestrial air temperature and precipitation:
1900–2014 gridded monthly time series (Version 3.01). Available online at. http://
climate.geog.udel.edu/~climate/html_pages/download.html.
Zhang, L.B., Liu, D. X., Liu, S.R., Zhang, Y., Tong, X.L., Wang, B.X., 2013. Responses of func-
tional diversity of aquatic insect community to land use change in middle reach of
Qiantang River, East China. Chin. J. Appl. Ecol. 24, 2947–2954. https://doi.org/
10.13287/j.1001-9332.2013.0480.
Zuellig, R.E., Schmidt, T.S., 2012. Characterizing invertebrate traits in wadeable streams of
the contiguous US, differences among ecoregions and land uses. Freshw. Sci. 31,
1042–1056. https://doi.org/10.1899/11-150.1.