Spatial and temporal variation of algal assemblages in six Midwest agricultural streams having varying levels of atrazine and other physicochemical attributes.

Science of the Total Environment 505 (2015) 65–89

Contents lists available at ScienceDirect

Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv

Spatial and temporal variation of algal assemblages in six Midwest agricultural streams having varying levels of atrazine and other physicochemical attributes J. Malia Andrus a,⁎, Diane Winter b,c, Michael Scanlan d, Sean Sullivan b, Wease Bollman b, J.B. Waggoner e, Alan J. Hosmer f, Richard A. Brain f a

Waterborne Environmental, Inc., 2001 South First Street, Suite 109, Champaign, IL 61820, United States Rhithron Associates, Inc., 33 Fort Missoula Rd., Missoula, MT 59804, United States c Algal Analysis, LLC, Missoula, MT, United States d MapTech, Inc., 3154 State Street, Blacksburg, VA 24060, United States e Inovatia, Inc., 120 East Davis Street, Fayette, MO 65248, United States f Syngenta Crop Protection, LLC, 410 Swing Rd., Greensboro, NC 27419, United States b

H I G H L I G H T S • • • • •

We monitored algal communities at 6 Midwest streams receiving atrazine in 2011 and 2012. Partitioning of CCA models of algal community by environment assessed the influence of specific variables. Overall, water chemistry and hydroclimate variables were most influential to community. Time since ≥30 μg/L atrazine pulse was more influential than other atrazine variables. Results are consistent with transitory community effects only at concentrations above 30 μg/L.

a r t i c l e

i n f o

Article history: Received 16 April 2014 Received in revised form 11 September 2014 Accepted 11 September 2014 Available online xxxx Editor: Mark Hanson Keywords: Algae Midwest Atrazine Stream Variance partitioning Multiple stressors

a b s t r a c t Potential effects of pesticides on stream algae occur alongside complex environmental influences; in situ studies examining these effects together are few, and have not typically controlled for collinearity of variables. We monitored the dynamics of periphyton, phytoplankton, and environmental factors including atrazine, and other water chemistry variables at 6 agricultural streams in the Midwest US from spring to summer of 2011 and 2012, and used variation partitioning of community models to determine the community inertia that is explained uniquely and/or jointly by atrazine and other environmental factors or groups of factors. Periphyton and phytoplankton assemblages were significantly structured by year, day of year, and site, and exhibited dynamic synchrony both between site–years and between periphyton and phytoplankton in the same site–year. The majority of inertia in the models (55.4% for periphyton, 68.4% for phytoplankton) was unexplained. The explained inertia in the models was predominantly shared (confounded) between variables and variable groups (13.3, 30.9%); the magnitude of inertia that was explained uniquely by variable groups (15.1, 18.3%) was of the order hydroclimate N chemistry N geography N atrazine for periphyton, and chemistry N hydroclimate N geography N atrazine for phytoplankton. The variables most influential to the assemblage structure included flow and velocity variables, and time since pulses above certain thresholds of nitrate + nitrite, total phosphorus, total suspended solids, and atrazine. Time since a ≥30 μg/L atrazine pulse uniquely explained more inertia than time since pulses ≥ 10 μg/L or daily or historic atrazine concentrations; this result is consistent with studies concluding that the effects of atrazine on algae typically only occur at ≥30 μg/L and are recovered from. © 2014 Elsevier B.V. All rights reserved.

1. Introduction ⁎ Corresponding author. Tel./fax: +1 217 378 4661. E-mail addresses: [email protected] (J.M. Andrus), [email protected] (D. Winter), [email protected] (M. Scanlan), [email protected] (S. Sullivan), [email protected] (W. Bollman), [email protected] (J.B. Waggoner), [email protected] (A.J. Hosmer), [email protected] (R.A. Brain).

http://dx.doi.org/10.1016/j.scitotenv.2014.09.033 0048-9697/© 2014 Elsevier B.V. All rights reserved.

The structure of algal communities in streams is affected by multiple factors. Temporally- and spatially-varying environmental parameters such as nutrient composition, pH, light intensity, salinity, wind shear, hydrology, general climate, and anthropogenic stressors can influence

66

J.M. Andrus et al. / Science of the Total Environment 505 (2015) 65–89

the composition of periphyton and phytoplankton in stream systems (Biggs, 1996; Schelske et al., 1995; Pan et al., 1999; Leira and Sabater, 2005; Julius and Theriot, 2010; Black et al., 2011). These effects occur at varying scales and include regional synchronicity, seasonality, historical influence, and interactions between factors (Pan et al., 1999; Leira and Sabater, 2005; Black et al., 2011; Allan, 2004; Soininen et al., 2004; Kent et al., 2007; Urrea and Sabater, 2009); it can be difficult to separate any individual impacts of each contributing variable. Farming practices have the potential to influence agro-ecosystems via several mechanisms which include changing the composition and concentration of sediment, addition of nutrients (fertilizers), and use of agricultural chemicals (Kroeze and Seitzinger, 1998; Malmqvist and Rundle, 2002; Foley et al., 2005). A number of these inputs are highly dynamic and can be linked to precipitation events, resulting in periodic pulses of inputs to streams adjacent to agricultural areas (Schultz, 2001; Ferenczi et al., 2002; Spalding and Snow, 1989; Neumann et al., 2003; Debenest et al., 2009; Rabiet et al., 2010); of these, herbicides have been shown in some cases to affect primary production or species composition of primary producers (e.g., (Guasch et al., 1998; Relyea, 2005; Debenest et al., 2009; Fairchild, 2011)). Atrazine is an herbicide used primarily to control broadleaf weeds in corn and sorghum via reversible inhibition of photosystem-II, and exhibits pulsed stream input behavior coincident with use and precipitation timing and intensity due to its solubility in water (Hamilton et al., 2011; Guasch et al., 1998; Giddings et al., 2005). Atrazine is used both as a pre-emergent and early post-emergent herbicide and has been used in numerous countries since the 1960s (Lakshminarayana et al., 1992; Solomon et al., 1996). Typical atrazine pulse concentrations in surface waters draining agricultural watersheds where atrazine is used range between 0.1 and 30 μg/L, with values most often reported to be below 10 μg/L; atrazine concentrations above 100 μg/L are infrequently reported (Waldron, 1974; Richard et al., 1975; Huber, 1993; Solomon et al., 1996). Atrazine inputs to small Midwestern streams generally occur in short pulses. Based on data (N150 site–year of samples collected at daily or near daily frequency) from monitored sites in Midwest US watersheds representing the upper 20th centile of atrazine concentrations, the median duration of atrazine concentrations greater than 15 μg/L is 2 days (P. Hendley, Syngenta Crop Protection, Greensboro, NC, USA, personal communication; derived from data shown in (United States Environmental Protection Agency, 2011)). Numerous studies have been conducted on the effects of atrazine on freshwater periphyton and phytoplankton, both on individual species and on algal communities in micro- or mesocosms. These studies have generally concluded that significant effects on primary producers typically only begin to occur with prolonged atrazine concentrations N30 μg/L and that subsequent to any disturbances algal populations recover (Gruessner and Watzin, 1996; Nyström et al., 2000; Baxter et al., 2011; Huber, 1993; Solomon et al., 1996; Giddings, 2012). However, mesocosm studies testing atrazine concentrations N 50 μg/L for extended periods have shown decreased activity, abundance, or diversity, or shifts in algal community structure (Kosinski and Merkle, 1984; Hamala and Kollig, 1985; Larsen et al., 1986; Krieger et al., 1988; Hamilton et al., 1988; Hamilton and Mitchell, 1997; Nyström et al., 2000; Guasch et al., 2007). Much research has been conducted using freshwater pond (lentic) micro- or mesocosms (e.g., Larsen et al., 1986; Hoagland et al., 1993; Berard et al., 1999),flowing (lotic) mesocosms (e.g., Lynch et al., 1985; Gruessner and Watzin, 1996; Nyström et al., 2000; Muñoz et al., 2001), or on natural lotic communities (Jurgensen and Hoagland, 1990; Lakshminarayana et al., 1992; Guasch et al., 1998; Dorigo et al., 2004; Laviale et al., 2011). However, to our knowledge, this initiative is the only such study to examine in situ the effects of atrazine on periphyton and phytoplankton dynamically throughout the growing season in several Midwestern agricultural stream areas where atrazine use is among the highest (Solomon et al., 1996; Andrus et al., 2013). Concurrent evaluation of native algal community structure in real

time with environmental parameters enables hypothesis testing concerning the relative contribution of measured variables to biological trends under natural environmental conditions of evaluated agroecosystems. The study reported here is an extension to the Syngenta Atrazine Ecological Monitoring Program (“AEMP”; (Prenger et al., 2009; USEPA, 2007a, 2007b, 2009a, 2009b, 2010, 2011, 2012), a program required by EPA to assess atrazine residues in small headwater streams in runoff from vulnerable watersheds (USEPA, 2007a)). The collection of AEMP watersheds as a whole was selected in part based on similar size and agricultural use; for the current study, six watersheds within the AEMP group were chosen from four different Midwestern states and differing historic atrazine concentrations. Periphyton and phytoplankton samples, along with coincident water chemistry, hydrology, climate, and geographical factors were collected from each watershed for 16 weeks, spring to summer periods in 2011 (three watersheds) and 2012 (all six watersheds). The objectives of this study were to evaluate the following: 1) Structure and dynamics of phytoplankton and periphyton communities 2) Diversity and variation in algal communities between sites and within sites 3) Extent of association between community metrics and variation in measured and unmeasured environmental metrics The tested null hypothesis was that there would be no association between environmental metrics and algal community structure and dynamics; the alternative hypothesis was that there would be. 2. Materials and methods 2.1. Study design The structure and dynamics of periphyton and phytoplankton communities were characterized weekly in situ at 6 agricultural streams sites (three in 2011 and three additional sites in 2012) in the Midwestern US over the course of the summer growing season (i.e. May through August in 2011 and April through July in 2012). The Atrazine Ecological Monitoring Program contains multiple watersheds that differ substantially in terms of land area and topography. For the purposes of this study, watersheds of similar size and site characteristics but with a range of historical atrazine concentrations were selected to enable comparison along an atrazine gradient. A variety of environmental parameters related to hydrology, geography, and water chemistry, including atrazine concentrations were measured concurrently. Statistical approaches including Canonical Correspondence Analyses (CCAs) and variation partitioning of community models were used to evaluate potential associations between biological trends and environmental metrics. Exploratory and preliminary results from the first year of the study have been published (Andrus, et al., 2013); here, the combined results from both study years are reported. To address the challenges of separating the effects of covarying natural and anthropogenic gradients and of differing temporal patterns and scales of influence, several strategies were employed. First, to more accurately describe the effects of intermittent pulses or chronic impacts of a particular parameter, a number of derived variables were incorporated into the analysis, including site averages and maxima and variables describing the time elapsed since an event of a particular threshold. Second, we used a variance partitioning methodology to assess the impact of individual environmental variables and variable groups on the composition of each algal community while controlling for coincident variables and variable groups. We chose as our ordination methodology a direct (CCA) rather than an indirect (CA or Nonmetric Mutidimensional Scaling) comparison because it allows for a more quantitative assessment of the impacts of each factor, and for more straightforward testing of significance.


2.2. Study sites Watersheds in the AEMP program were selected, in part, for their size (2330–10360 ha; 9–40 sq mi) and potential high vulnerability for atrazine runoff. Each watershed in the AEMP program is sampled at one location—the point closest to the outlet of the watershed where the stream intersects an accessible road. The watersheds monitored in the current, complementary study (three in 2011 and six in 2012) were selected to be of similar size (29–90 km2) and to be hydrologically and topographically similar, but to have different historical concentrations of atrazine (Table 1) during the corn growing season. To achieve geographically diverse samples, the sites were also selected from four states (Illinois, Iowa, Missouri, and Ohio). Characteristics of the watershed at each site are listed in Table 1, and their location is shown in Fig. 1. All sites have wellvegetated banks and are surrounded by cropland. More detailed information about the watersheds and their selection process for the AEMP study can be found in the annual AEMP reports (Prenger et al., 2009; USEPA, 2007a, 2007b, 2009a, 2009b, 2010, 2011, 2012). 2.3. Physical Habitat Assessments Physical habitat was assessed at all sites using Ohio's Qualitative Habitat Assessment Index (QHEI; (Hall et al., 2012; Rankin, 1995; Ohio EPA, 2006)), which translates qualitative observations on lotic microhabitat quality into seven principle metrics (substrate, in-stream cover, channel morphology, riparian zone and bank erosion, pool/glide and riffle/run quality, and map gradient) and then into a total weighted score. As the “map gradient” metric was identical for all sites, it was excluded from the analysis. 2.4. Environmental Factors Environmental parameters were monitored as in (Andrus et al., 2013). They are repeated in summarized form here. A weekly 1 L water grab sample from each site was taken at a depth of ~0.33 m and used to determine water chemistry properties. Half of each sample was preserved with sulfuric acid and used to determine nitrate and nitrite, and the remaining 500 mL was analyzed for total suspended solids, total phosphorus, alkalinity, and hardness. Grab samples were kept at 4 °C between collection and analysis, a duration that did not exceed one week. Temperature, dissolved oxygen, specific conductance, and pH measurements were also measured on a weekly basis using handheld water quality probes at ~ 0.33 m depth. Continuous measurements

67

(15 s intervals) of temperature, aqueous (planktonic chlorophyll-a), and specific conductivity were taken over the 16 week study interval in 2011 at two of the sites (IL-10 and MO-05) and at all six sites in 2012 using extended deployment optical probes installed at approximately half the average stream depth. In addition, time-weighted daily composite samples for measurement of atrazine (1,3,5-Triazine2,4-diamine,6-chloro-N-ethyl-N′-(1-methylethyl)) were collected over the duration of the entire AEMP study (see Table 1 for the years sampled for each site) by autosamplers (or once every four days prior to 2008 or in the rare event of autosampler malfunction). Each watershed monitored was also equipped with instrumentation attached to a solar power array and dataloggers to continuously (15 s intervals) monitor stream stage and weather conditions.

2.5. Algal Community Sampling Both periphyton and phytoplankton algal communities were sampled on a weekly basis from each site over a 16 week interval from May to August in 2011 and April through July in 2012. During the 2011 season, samples were collected in triplicate. Statistical analysis of the first year of data showed that these replicates were not significantly different (Andrus et al., 2013), and therefore in 2012 only one sample of each type was collected per week. For each community type, ash free dry mass (AFDM) and chlorophyll-a (chl-a) were measured and taxonomic determination to the lowest possible taxonomic level was performed. Water samples were collected for phytoplankton taxonomic enumeration, AFDM, and chl-a using a LaMotte grab sampler. During the 2011 season, samples were taken from three equidistant positions across the width of each stream. In 2012, the single sample was taken from the center of the stream. Water samples collected for taxonomic enumeration were preserved with 1% Lugol's solution (2% KI, 1% I2) and kept cold (~ 4oC). In April and May of 2011, these samples were 500 mL, but because of lower density at one site and to improve the accuracy of community characterization, 1 L samples were taken from June 2011 onwards at all sites. Separate water samples collected for phytoplankton AFDM determination were filtered (as much volume as possible, up to 1 L) onto glass fiber filters (Whatman 934-AH), flash frozen on dry ice and kept frozen for transport to the laboratory. To collect periphyton, custom periphytometers were used (Supplemental Fig. 1). A full description of the periphytometers appears in (Andrus et al., 2013), but in brief consist of a floating stainless steel frame holding 18 square slate tiles approximately 10 cm × 10 cm. Slate tiles were chosen as a substrate over glass slides or collection from extant substrates because they were reproducible across the

Table 1 Study site characteristics. Site

IA-03

IL-10

IL-17

MO-05

MO-07 N

OH-05

Stream name Watershed area (km2) Area in corn or sorghum (%)b Stream order Years monitored in AEMP program (prior to current study) Historical avg. daily atrazine (μg/L)a Historical max daily atrazine (μg/L)a Stream bed material Right bank height (m) Left bank height (m) Estimated thalweg depth (m) Streambed slope (%) Years monitored in the AEMP study Years biomonitored in the current study

Lick Creek 82.1 32.1 4 2009–2010 4.73 50.6 Sand 2.4 3.1 0.61 0.048 2010–2012 2011–2012

Felky Slough 87 50.1 3 2009–2010 0.57 11.4 Sand/gravel 3.1 3.1 0.61 0.091 2010–2012 2011–2012

Limestone Creek 29.3 34.9 2 2009–2010 6.93 228 Sand 5.3 5.3 1.07 0.167 2010–2012 2012

Long Branch 65.5 35.4 3 2007–2010 3.96 77.2 Sand/clay 0.46 0.3 0.52 0.001 2007–2012 2011–2012

Honey Creek 89.9 27.0 2 2009–2010 9.88 90.9 Gravel 4.6 3.1 1.5 0.095 2010–2012c 2012

Auglaize River 80.8 25.9 3 2009–2010 0.97 30.6 Sand/gravel 3.1 3.1 0.61 .011 2010–2012 2012

a b c

Across the duration of the AEMP study. At the time of the initial site survey. Monitoring location moved downstream in 2012.

68


Fig. 1. Maps of study sites. Upper-right-hand corner inset shows the region and the location of each study site. The remaining panels show maps of each watershed, with sampling location and stream system.

duration of the study, the area collected was easily measured, and the slate was more representative of the naturally-occurring gravel and rock periphyton habitats than glass. Although communities collected with artificial substrate cannot be expected to be identical to naturally-occurring assemblages, they are expected to be similar and better allow for longitudinal study. During the 2011 season, four periphytometers were deployed at each site, three for active sampling and one reserve unit in case of loss or failure. In the 2012 season, only two periphytometers were deployed at each site, one for active sampling and one reserve unit. The periphytometers were installed in a row parallel to stream flow near one bank at each stream site and allowed to equilibrate for approximately 1 month before samples were first taken. Periphyton samples were collected by removing one tile from the upstream end of a periphytometer each week for 16 weeks. Periphyton from each sample tile was divided into separate 50 ml centrifuge tubes for specific analyses as follows. During the 2011 season,

the periphyton sample from one quarter of the tile area (25.8 cm2; 4 in2) was flash frozen on dry ice and was used for chl-a analysis. The remaining sample material (77.4 cm2, 12 in2) from each tile was preserved in ~4 ml of 1% Lugol's solution (2% KI, 1% I2) with stream water added to fill completely the 50 mL centrifuge tube; this tube was kept cold (~4 °C) and used for taxonomic and AFDM analyses. However, in an effort to improve the precision of the periphyton AFDM and chl-a measurements, periphyton sample collection procedures were modified in 2012 so that one quarter of the tile area (25.8 cm2; 4 sq. in) was kept cold (~4 °C) and used for AFDM analysis. Of the remainder, the periphyton scrape from one half of the tile area (51.6 cm2; 8 sq. in) was preserved with Lugol's solution (as before, ~ 4 mL Lugol's solution and stream water were added to fill the 50 mL tube), kept cold (~ 4 °C) and used for taxonomic analysis. Finally, the periphyton sample from the remaining one quarter of the tile area (25.8 cm2; 4 sq. in) was flash frozen on dry ice and was used for chl-a analysis. For any samples, when the periphyton material on a tile was very thick and could not fit


69

Algal biomass, pigments, and taxonomy were measured as described in Andrus et al. (2013). These methods are summarized in brief below. Periphyton biomass and phytoplankton biomass (ash free dry mass, AFDM) were analyzed by filtration. Chlorophyll-a (chl-a) and pheophytin levels from frozen periphyton scrape samples were measured colorimetrically and were reported per area (periphyton). The method of duplicate sample comparison for periphytic chlorophyll yielded a mean relative percent difference (RPD) of 6.89% ± 0.06% in 2011 and of 3.32% ± 1.40% in 2012. Planktonic chlorophyll-a was measured from extended deployment optical probes (not available for the IA-03 site in 2011). Taxonomic identification was performed by microscope enumeration. Prior to analysis, planktonic algae samples were settled to concentrate cells. Both planktonic and periphytic algae samples were then divided into two sub-samples, one each for diatomic and non-diatomic algae. Diatomic algae samples were homogenized and acid cleaned, and volumes were adjusted to obtain adequate cell density. At least 300 valves or natural units (or a counting limit of one entire slide or 2 h per sample) were identified to the lowest possible level, typically species, following standard taxonomic references (see Supplemental Information). In non-diatomic algae sub-samples of both planktonic and periphytic algae, diatoms were counted as a group and used to determine the relative densities of diatom species identified in the diatom community analysis. Taxonomic analysis was repeated for 10% of all samples (24) by a second taxonomist for quality control. Bray–Curtis similarities (Bray and Curtis, 1957) were calculated between replicate identifications; all repeated samples except one (66%) had values above 83%. Based on periphyton diatom taxa, the Pollution Tolerance Index (Bahls, 1993; Lange-Bertalot, 1979) and the percentage of Cosmopolitan Taxa (Lange-Bertalot et al., 2011) were calculated.

2011; Solomon et al., 1996; Roads and Transportation Association of Canada, 1973). Because there were a large number of measured variables in this study, to simplify interpretation four descriptor categories were created: Hydroclimate, geography, water chemistry (not including atrazine) and atrazine. These categories reflected groups of major environmental factors in the structuring of algal communities, and enabled assessment of the relative contribution of each group while controlling for others. Atrazine variables were grouped autonomously since a major focus was to assess the potential impact of this herbicide solely on algal communities while controlling for other coincident factors. As the distributions of most univariate metrics (diversity and environmental parameters) were non-normal, statistical differences by site or month were assessed using the Kruskal–Wallis one-way analysis of variance (Kruskal and Wallis, 1952), followed by a multiple comparison test where the statistic showed significant differences. The correlation of variables was assessed using either Spearman correlation coefficients (Spearman, 1904) or the maximum likelihood estimator regression (Helsel, 2005) for variables containing non-detect data. Prior to modeling, univariate values (both diversity metrics and environmental parameters) were examined for normality, and transforms were made, where appropriate. Prior to analysis of community structure, raw taxa data were collected into separate taxa matrices for phytoplankton and periphyton, each divided into diatoms and non-diatom algae (NDA). Several community analysis methodologies (including those that are used in this work) are distorted by rare taxa (Austin and Greig-Smith, 1968; Orloci and Makkattu, 1973; Goff, 1975) and are typically difficult to quantify accurately. To down-weight the effect of transient population spikes and zero-inflation in community analysis, taxa not comprising more than 0.5% of total abundance in two or more samples were removed, and data were Hellinger transformed (Rao, 1995; Legendre, 2012). After this procedure, 171 of 400 diatom and 96 of 195 non-diatomic algae periphyton taxa were retained. Similarly, 131 of 214 diatom and 85 of 142 non-diatomic algae phytoplankton taxa were retained by the procedure (see Supplemental Information).

2.7. Data analysis

2.9. Analysis of relationships between community structure and environment

into the sample collection tube, the actual area of tile sampled was recorded. 2.6. Algal analysis

All statistical analyses were performed using the statistical software R, using the packages STATS (R Core Team, 2013), VEGAN (Oksanen et al., 2013), PGIRMESS (Giraudoux, 2013), PARTY (Hothorn et al., 2013) and NADA (Lopaka, 2012). Data were plotted using the VEGAN (Oksanen et al., 2013), and GGPLOT2 (Wickham, 2009) packages. The functions from these packages used for specific analyses are documented in Supplemental Table 1. Additional annotations were added using Illustrator CS5 (Adobe Systems, San Jose, CA). 2.8. Data processing and univariate statistical analysis Values for the diversity metrics taxa richness (S), evenness (Pielou's J), and total recovered cells were calculated from the taxonomic analysis of diatoms or non-diatom algae for each algae type (periphyton and phytoplankton). To quantify the time a community has had to adapt since various types of stochastic events for a number of environmental variables, the time elapsed (in days) between a selected threshold level of the variable and collection of a biological sample was calculated (see Table S2 footnote a). Daily measurements were used to calculate these descriptors when available (flow, velocity, atrazine); for other variables, measurements were only available on a weekly basis (nitrate/nitrite, total phosphorus (TP), total suspended solids (TSS)). Descriptors of flow events were first normalized to the median flow for a site (over the period 2010–2011) to account for the different hydrologies of each site. Thresholds were chosen by examination of the data and thorough consideration of literature sources (Black et al.,

Many hydroclimatic, geographical, and water chemistry (including atrazine) factors were considered as potential explanatory factors for both univariate metrics of diversity and diatom indices and multivariate community structure (Table 2). 2.10. Conditional inference forests Conditional inference forest analysis (Breiman, 2001; Strobl et al., 2007) was used to initially screen the full set of potentially explanatory factors for important relationships with community assemblages (phytoplankton and periphyton separately). For each combination of factor group and target variable or assemblage, 50 conditional inference forests were constructed and the variable importance scores for each forest calculated, using the cforest and varimp functions in the R package PARTY. To reduce computational complexity, the number of trees in each forest (“ntree”) was set to 50, and the number of variables tried at each node (“mtry”) was set to one-third of the total variables. The mean and standard deviation of variable importance scores over the 50 replicates were calculated for each group/metric pairing. For each variable group, variables, ordered by their mean variable importance scores, were plotted against their standard deviations, and this curve was modeled using a standard classification and regression tree (CART) model (via the tree function in package TREE). The minimum standard deviation value predicted by the CART model was then used as a threshold for mean importance scores, as in Genuer et al. (2010). Variables with a mean importance score below this threshold

70


Table 2 Measured environmental and diversity parameters considered in this study. Potential explanatory factors

Biological parameters

Hydroclimate

Geography

Water chemistry

Atrazine

Diversity metrics and Algal group densities diatom indices

• • • •

• • • • • • •

• • • •

• Daily composite atrazine • Days since ≥30 ppb atrazine (dS30Atz) • Days since ≥10 ppb atrazine (dS10Atz) • Avg atrazine over AEMP study a • Max atrazine over AEMP study a

• Density (diatom and NDA) • Taxa richness (diatom and NDA) • Evenness (diatom and NDA) • Chlorophyll-a • Pheophytin • AFDM • Pollution Index • % Cosmopolitan Taxa

• • • • • • •

Velocity weekly average Flow weekly avg Depth weekly avg Days since ≥5× median flow event (dS5xFlow) a Days since ≥20× median flow event (dS20xFlow) a Days since ≥50× median flow event (dS50xFlow) a Stream velocity weekly avg. Days since ≥0.33 m/s velocity event (dS033Vel) Water temp weekly avg Solar radiation weekly max. Max daily wind speed, weekly avg.

Substrate Instream cover Channel morphology Bank erosion/riparian zone Pool/glide quality Riffle/run quality % of watershed area: o Growing soybeans o Growing corn or sorghum o With claypan depth 5–25 cm and slope 1–3% o With claypan depth 5–25 cm and slope N 3% o With claypan depth 25–50 cm and slope 1–3% o With claypan depth 25–50 cm and slope N 3%

• • • • • • • • • • • • • • •

a

TSS Avg TSS over study a Max TSS over study a Days Since ≥50 mg/L TSS (dS50TSS) Days Since ≥250 mg/L TSS (dS250TSS) TP Avg TP over study a Max TP over study a Days Since ≥0.5 mg/L TP (dS05TP) NO2 + NO3 Avg NO2 + NO3 over study a Max NO2 + NO3 over study a Days since ≥1 mg/L NO2 + NO3 (dS1N) Days since ≥5 mg/L T NO2 + NO3 (dS5N) Alkalinity Hardness pH DO SpC weekly average (in situ probe)

• Bacillariophyta • Cyanophyta • Green (Chlorophyta and Charophyta) • Chrysophta • Dinophyta • Euglenophyta • Raphidophyta

By site.

were eliminated as unimportant; the logic here being that the standard deviation is a rough estimate of random variability so that mean variable importance scores smaller than this level are likely attributable to random variation. 2.11. CCA model optimization and variance partitioning Canonical correspondence (CCA) models were constructed to characterize the relationships between environmental parameters and periphyton or phytoplankton community structure. CCA models were optimized for each factor grouping individually through custom algorithms implemented in R. To optimize the model for each factor grouping, all possible sub-models of the factor set were constructed and ordered by an analogue to Akaike's Information Criterion (AIC; (Akaike, 1974; Oksanen et al., 2013)). The statistically significant (α = 0.05) model with the lowest AIC, with a variance inflation factor (VIF) less than 10, and with each included variable having a significant (Type III/marginal α = 0.05) coefficient was chosen. Variance partitioning (see (Borcard et al., 1992; Legendre, 2012)) was employed to determine the portions of variance in redundancy analysis (“RDA”, the canonical extension of PCA) or inertia in CCA models uniquely associated with large-scale gradients (site, season, or year) or to single variables or groups of variables (water chemistry, geography, hydroclimate, and atrazine). These portions were calculated by constructing partial models of the target variable while holding all other variables constant. For single variables or descriptor groups, partial models were constructed with respect to all other environmental variables in optimized models. Partial models of gradients were constructed with respect to other gradients. Portions of variance shared by more than one variable or gradient were calculated by subtracting the variance explained by the partial models of each of these variables or gradients from the variance explained by partial models containing all variables in question. CCA models were constructed using all available samples; this choice was the result of careful deliberation. Because community replicates were found in the first year to be statistically similar, to maximize the range of environmental conditions at sample sites given financial

considerations triplicate samples were discontinued and three new sites were instead added in the second year. One drawback of this decision is that samples were not equivalently distributed between site–year levels. Although aggregation of year one replicate community samples would result in the same number of samples per site–week, the number of samples per year would remain imbalanced (arguably more important due to the pronounced drought in 2012). Separate balanced ordinations to evaluate the effects of year (using only sites sampled in both years) or of site (using only 2012 data) could be constructed, however these would not allow comparison of the relative influence of each. Because CCA ordinations have been found to be robust to samples not taken regularly along, or encompassing only a range of, environmental gradients (Palmer, 1993), we concluded that using all community samples in ordinations was the least intrusive option. Statistical differences between algal assemblages were tested by analysis of similarity (ANOSIM; (Clarke and Green, 1988; Clarke, 1993)), using the Bray–Curtis similarity. Procrustes analysis (Gower, 1971, 1975, 1987; Jackson, 1995) was used to assess the concordance between correspondence analyses of algal assemblages from different sites and different algal fractions (periphyton or phytoplankton). 3. Results 3.1. Physical habitat assessments Qualitative habitat evaluation index (QHEI, Ohio EPA, 2006) scores generally reflected the topographical differences between sites. The IL-10 site has the deepest channel of the sites and steep banks. The IL-17 and MO-07N sites are somewhat shallower but also with steep bank inclines. The remaining three sites (IA-03, MO-05, and OH-05) are shallow and sandy sites, with shifting sandbars and low banks. Overall QHEI scores were correspondingly in the upper “Fair” or lower “Good” range for IL-10, IL-17, and MO-07N but in the upper “Poor” or lower “Fair” range for IA-03, MO-05, and OH-05. These differences were mostly driven by higher substrate, channel morphology, and riffle/run quality scores at the IL-10, IL-17, and MO-07N sites.


3.2. Water quality and climatological factors Measurement of water quality and climatological parameters showed that each site had distinctly different hydroclimate, geography, water chemistry, and atrazine concentrations. For each factor, at least one site exhibited persistently and significantly different levels, and for many factors there was substantial variation over the monitoring period (Fig. 2, Supplemental Table 2). However, factors were typically statistically consistent between years for sites monitored in both 2011 and 2012. Hydrological variable exhibited particular variation at the monitored sites. Flow and stream depth was greatest at IL-10, were lower at MO-07N and IL-17, and lowest at IA-03, MO-05, and OH-05. All sites experienced large flow events in the spring of each year, typical of agricultural streams in the Midwest. These flow events were unusually large during 2011 at IA-03, reflected in a significantly different flow and maximum weekly stream depth between 2011 and 2012. Additionally, 2012 was a drought year and was particularly dry in the late summer (only 20.62% of the contiguous US was not classified as in drought condition on 31 July 2012; (National Drought Mitigation Center, 2013)). Although overall flow and depth were similar between 2011 and 2012 overall at IL-10 and MO-05, during the summer (weeks of the year 25–31, i.e., only summer weeks measured during both years used) flow was significantly lower during 2012 at IA-03 and IL-10, and

71

average stream depth was lower at all three sites (multiple comparison test after Kruskal–Wallis, p b 0.05). Concentrations of atrazine at different sites, with year-to-year variations, generally followed the order MO-07N N MO-05 N IL-17 + IA-03 N OH-05 N IL-10. Atrazine concentrations exhibited pulsed dynamics, concurrent with certain flow events and greatest during the spring. In 2011, sampling began before the first pulse of atrazine (late May 2011). However, the spring of 2012 was unusually warm, and therefore atrazine applications and other preparations for the agricultural season occurred earlier than usual. Although algal and environmental sampling was initiated roughly 3 weeks earlier in 2012 (the second week of April 2012) than in 2011, the first weekly samples at IL-17, MO-05, and MO-07N were taken during the presence of atrazine runoff. Increased NO2 + NO3 concentrations were similarly episodic during springtime flow events, consistent with spring fertilizer application. However, it is possible that the peak concentrations of NO2 + NO3 occurred earlier than the beginning of the 2012 sampling at the IL-17, MO-05, and MO-07N sites. 3.3. Patterns and scales of variability of environmental and habitat parameters The following factors showed substantial seasonal components: depth and flow, NO2 + NO3, water temperature, wind speed, the

a

Fig. 2. (a and b) Environmental characteristics at study sites during study. Numbers annotating panel show the median values for each site, and single italic letters indicate the results of statistical tests for differences between sites. Different letters indicate significantly different medians by multiple comparison tests after Kruskal–Wallis, with a as the group of lowest median.

72


b

Fig. 2 (continued).

concentration of atrazine (daily composited average) and days since NO2 + NO3, TP and TSS thresholds (Fig. 3). However some of these factors also had considerable variation associated with site, reflecting different hydrology, topography, and inputs by site that were not visible on the first two principle component axes. Those factors that were predominantly described by site differences are alkalinity, hardness, and specific conductivity, DO, days since flow and velocity events, solar radiation, days since atrazine thresholds and the geographic descriptors (except pool/glide quality). Differences between the years 2011 and 2012 were associated with some variance of depth and flow, days since NO2 + NO3, TP, TSS, and atrazine thresholds, and the habitat parameters other than substrate. These results reflect the drought conditions (decreased flow and depth, and increased concentration of sediment and associated particulate phosphorus), warm spring and consequent early application of inputs (atrazine and nitrate), and evolution of habitats in 2012. Finally, the majority of variance of several inputs was shared between the three major gradients, or not explained by any of them. These variables—depth, flow, pH, nitrate/nitrite, TP, TSS, atrazine and days since associated thresholds—are ephemeral or “flashy” factors that are at least partially related to storm events, which, although more frequent and intense in the spring, are not strictly linearly associated with the Julian day of the year. 3.4. Abundance and diversity No statistically significant differences in univariate diversity metrics between sample replicates were observed. However, statistically

significant differences in measures of diversity between the six sites were observed. 3.4.1. Abundance Overall, periphyton cell density was fairly consistent across sites (Fig. 4, Supplemental Table 3). However, periphyton at IA-03 in 2011 experienced scouring due to large flooding events in late May and early June and periphyton was therefore significantly less dense for this site–year than some other site–years (IL-10 in 2011, MO-05 both years, and OH-05). At IA-03 and IL-10, periphyton density exhibited a strongly seasonal pattern, highest in the spring. Other sites showed a fairly steady density throughout the season, typically with periods of higher density: one in late June after the flow events of the spring and the other in late summer. The latter was largest at the MO-05 site (both years). At the evaluated sites, diatoms dominated periphyton communities almost entirely across all sites and throughout the season (Fig. 4). While cyanophyta were generally the most abundant group of non-diatomic algae at all sites, only at MO-05 late in the summer of 2012 were nondiatomic algae greater in number than diatoms. At this site, green algae and cyanophyta dominated the community directly prior to the stream completely drying up. Additionally, a greater proportion of cyanophyta was observed at IA-03 in 2011 directly after the large flow events during this site–year, suggesting that cyanophyta were more resilient to these conditions. Euglenophyta and cryptophyta were observed in small numbers periodically at the MO-07N, MO-05, and IA-03 (in 2012) sites. Few chrysophyta (except some early blooms),

Fig. 3. Variance of environmental factors explained by canonical axes of major gradient categories. Stacked bars represent percent of the total variance of each environmental variable that is associated with specific gradients, or not associated with any of the gradients (generally pulsed behavior).


73

Alkalinity DO Hardness NO2+NO3 Days Since 1 mg/l NO2+NO3

Chemistry

Days Since 5 mg/l NO2NO3 pH SpC TP Days Since 0.5 mg/l TP TSS Days Since 50 mg/l TSS Days Since 250 mg/l TSS % Area of Watershed in Corn+Sorghum % Area of Watershed in Soybeans Bank Erosion/Riparian Zone

Geography

Base Flow Index Channel Morphology Instream Cover Pool/Glide Quality Riffle/Run Quality Substrate Quality Depth Weekly Avg Flow Weekly Avg Days Since 5x Median Flow Days Since 20x Median Flow

Hydroclimate

Days Since 50x Median Flow Solar Radiation Weekly Max Velocity Weekly Avg Days Since 0.33 m/s Velocity Water Temp Weekly Avg Max Daily Wind Speed, Weekly Avg Daily Composite Atrazine

Atrazine

Days Since 10 ug/l Atz Days Since 30 ug/l Atz 0%

20%

40%

60%

80%

Partial (Explained) Variance by Canonical Axis

100%

Component Season (Day of Year) Year Site Other


100000

10000

10000

IA03

100000

1000

1000

100

100

1000000

1000000

100000

100000

10000

10000

IL10

#cells/m2 #cells/m2

1000000

1000

1000

100

100

#cells/m2

2012

2011 1000000

#cells/m2

74

100000

IL17

10000 1000

#cells/m2

1000000

1000000

100000

100000

10000 1000

10000 1000

#cells/m2

1000000

MO05

100

100

Jun

Jul

1000000

Aug MO07N

100000

1000 100 1000000 100000

OH05

Bacillariophyta Chlorophyta +Charophyta Chrysophyta Cryptophyta Cyanophyta Dinophyta Euglenophyta

10000

#cells/m2

May

10000 1000

#cells/m2

#cells/m2

100

100

May

Jun

Jul

Aug

Fig. 4. Spatial and temporal variability of periphyton algal groups. Black lines represent the mean overall cell density for each site and date (in 2011, three samples taken, in 2012, one sample was taken). Embedded stacked bars show the proportion of various algal groups making up each sample.

dinophyta, raphidophyta, and xanthophyta were observed in periphyton samples, but they were present in some sites. Planktonic algae were overall much less dense than periphytic algae. Cell densities for individual samples ranged between 3.80 × 100 and 2.32 × 104 cells/mL, approximately an order of magnitude lower than that of periphyton (per cm2). However, cellular abundance of phytoplankton was, as with periphyton, generally comparable across sites, although a few significant differences between site–years were observed (Fig. 6, Supplemental Table 4). At several site–years, distinct periods of bloom were observed: IL-10 2012 (April and early June, associated with cyanophyta), MO-05 2011 and 2012 and IL-17 2012 (late May and after July, associated with green algae and euglenophyta), and OH-05 2012 (late May, associated with diatoms, and late July, associated with greens and euglenophyta).

Unlike periphyton, phytoplankton were not consistently dominated by diatomic species (Fig. 6; Supplemental Table 4), although diatoms comprised a significantly higher proportion of cells (were mostly dominant) during the spring months (April, May, June) in each site–year than during the summer months (July and August; χ2 = 23.80, p b 0.0001). Additionally, the proportion of diatoms in the IL-10 2011 site–year was higher than at other sites (pariwise comparisons after Kruskal–Wallis, p b 0.05). Other major algal groups also showed seasonal trends: the percentage of total cell density comprised of green algae (chlorophyta and charophyta) and of euglenophyta were higher in the summer months (July and August) than in the spring months (χ2 = 16.14, p b 0.0001, and χ2 = 60.03, p b 0.0001, respectively). Additionally, while dinophyta were not present in many samples, samples in which they were present were collected during July and August.

Fig. 5. Spatial and temporal variability of periphyton diversity metrics and diatom indices. Ribbons are bounded by maximum and minimum values for each site and year. Numbers adjacent to plotted ribbon show the median value for each group and differing single letters following mean indicate significantly different medians by multiple comparison tests after Kruskal–Wallis, with a as the group of lowest median.


75


1000

1000

100

100

10

10

100000

100000

10000

10000

1000

1000

100

100

10

10 100000 IL17

10000 1000 100

#cells/mL

10000

IA03

100000

10000

#cells/mL

2012

IL10

#cells/mL

#cells/mL

2011

100000

#cells/mL

76

100000

10000

10000

1000 100

1000 100

#cells/mL

100000 MO05

Jul

100000

Aug

Bacillariophyta Chlorophyta +Charophyta Chrysophyta Cryptophyta Cyanophyta Dinophyta Euglenophyta

10000

MO07N

Jun

1000 100 10 100000 10000

OH05

May

#cells/mL

10

10

1000 100

#cells/mL

#cells/mL

10

10 May

Jun

Jul

Aug

Fig. 6. Spatial and temporal variability of phytoplankton algal groups. Black lines represent the mean overall cell density for each site and date (in 2011, three samples taken, in 2012, one sample was taken). Embedded stacked bars show the proportion of various algal groups making up each sample.

Although cyanophyta mostly appeared in large proportions in the spring months at certain sites, a seasonal trend was not significant across all site–years; this contrasts with cyanophyta in periphyton, which were a higher proportion during the late summer. Chlorophyll-a concentration in periphyton samples roughly followed the patterns of periphyton cell density (Fig. 5, Pearson correlation: ρ = 0.192, p = 0.003), with concentration at IA-03 in 2011 significantly lower than that at MO-05 in 2012 (Fig. 5), but with all other site–years having statistically similar concentrations. Measured biomass (ash free dry mass) increased or remained stable at all sites. However, temporary but large decreases in biomass were observed during flow events during most site–years, and for unexplained reasons at IL-17 and OH-05 during late June of 2012. No significant differences between planktonic chlorophyll-a or AFDM concentrations at different site–years were observed (Fig. 7). However, chlorophyll-a concentrations were significantly higher in April and May than in June, July, or August across all site–years (χ2 = 22.50, p b 0.0001). Diatoms grown in high light conditions have a lower chl-a concentration per cell (Strickland, 1965), so these patterns may be associated with seasonal light patterns. AFDM concentrations were more erratic across time and showed no clear patterns.

3.4.2. Diversity Because diatoms were dominant in the sites monitored, the diversity metrics of taxa richness (S) and evenness (J) were calculated for

diatoms and non-diatomic algae separately (Fig. 5; Supplemental Tables 3, 4). For periphytic diatoms, median richness for site–years were between 44 (IA-03 2011) and 83.5 (MO-07N 2012), and median evenness between 0.72 (MO-05 2012) and 0.85 (MO-07N 2012). Both richness and evenness showed similar patterns between 2011 and 2012 for sites with two years of data; exceptions to this are differing patterns in diatom evenness between years at MO-05 (though still statistically similar) and dips in diatom richness and evenness at IA-03 in 2011 (causing significant differences between years and related to high flow scouring events). Non-diatom richness was much lower in periphyton, consistent with substantial diatom dominance at the test sites. Median richness by site–year ranged between 4.5 (IL-10 2011) and 17 (MO-05 2011). NDA richness was statistically similar between 2011 and 2012 at all three sites having two years of data. NDA evenness was also fairly consistent between sites, with the notable exception of MO-05 in 2012. In this site–year, several transient increases in NDA cell density seem to be connected with decreases in evenness (Spearman ρ = − 0.855, p b 0.0001), suggesting that the decreased evenness is due to blooms or deposition of only one or a few taxa. Major increases in NDA richness and evenness were observed at MO-05 (both years) and IL-17 in late summer; other site–years had more moderate increases in these metrics toward the end of the sampling season. Consistent with the lower cell densities in the planktonic fraction of algae, planktonic diatom richness was about half that of the median


77

Fig. 7. Spatial and temporal variability of phytoplankton diversity metrics. Ribbons are bounded by maximum and minimum values for each site and year. Numbers adjacent to plotted ribbon show the median value for each group and differing single letters following mean indicate significantly different medians by multiple comparison tests after Kruskal–Wallis, with a as the group of lowest median.

richness in periphyton (Fig. 7). However, NDA taxa richness was generally higher in phytoplankton than in periphyton. Diatomic evenness was comparable to that of periphyton, as was NDA evenness in 2011; in 2012, the low periphytic evenness observed was not matched by that of phytoplankton. Between site–years, planktonic richness and evenness were relatively comparable, with most site–years being statistically similar. However, diatom evenness was also highly variable within the same site during 2012, for most sites ranging between ~0.25 and ~ 0.75. This variability was not observed during 2011. The IL-10 2012 NDA taxa evenness shows evidence of the large cyanophyta bloom that occurred between weeks 7 and 8 in this site–year. 3.5. Diatom indices The Pollution Tolerance Index is defined as the average organic pollution tolerance of the diatoms observed (1 = most pollution tolerant, 2 = pollution tolerant, and 3 = sensitive); therefore the lower the

score, the more tolerant the community to organic pollution. At the evaluated sites, median site–year values were largely between 1 and 2 (a mix of most tolerant and tolerant taxa), however some site–years (IL-10 2011 and 2012 and OH-05 2012) had median scores above 2 (generally between tolerant and sensitive species) (Fig. 5; Supplemental Table 3). The lowest Pollution Tolerance Index scores (most tolerant communities) were at IA-03 and MO-05, which were statistically lower than IL-10 and OH-05 site–years. Additionally, MO-05 2012 had a lower Pollution Index than IL-17, and MO-07N was lower than IL-10 2011. The Percentage Cosmopolitan Taxa measures the proportion of diatom taxa observed that are common (adapted to many different habitats) as opposed to rare. Diatoms at the evaluated sites were largely cosmopolitan, with median site–year values ranging from 53% (MO-05 2012) to 80% (IA-03 2011) (Fig. 5; Supplemental Table 3). IA-03 2011 had a statistically higher percentage of cosmopolitan taxa than IA-03 2012, MO-05 (both years) and MO-07N 2012. A dip in cosmopolitan taxa was observed at several site–years (IL-10 2011 and 2012, IL-17 2012, MO-05 2012 and OH-05 2012), typically in mid-season.

78


3.6. Community composition 3.6.1. Environmental factors Optimized CCA models of periphyton and phytoplankton (all sites) constructed using measured environmental factors and QHEI habitat scores are shown in the top panels of Fig. 8, and their corresponding explained inertia and variance partitioning in Table 4 and Fig. 9. Models explained 46.0% and 31.6% of total (unique + shared) periphyton and phytoplankton taxa inertia, respectively. In both models, all four descriptor groups uniquely accounted for significant portions of inertia,

a

with the order of inertia explained in the periphyton model being hydroclimate N water chemistry N geography N atrazine and with that in the phytoplankton model being water chemistry N hydroclimate N geography N atrazine. A pCCA ordination plot for each community, constrained by atrazine variables and with all other environmental variables partialled out (corresponding to the inertia uniquely associated with atrazine variables) is shown in the Supplemental Information (Supplemental Figs. 3, 4). Total shared inertia between explanatory variables was 30.9% in the periphyton model and 13.3% in the phytoplankton model, implying less redundancy in explanatory power

b

Fig. 8. Canonical correspondence analysis plots of a) periphyton, and b) phytoplankton communities. Axes represent linear combinations of environmental gradients; the composition of these axes is shown by the blue arrows, whose magnitude indicates the strength of loading onto the axes. Points represent sample algal assemblages and are coded by site or site and month sampled. In the middle panel, polygons bound samples taken in the same year. In the middle and bottom panels, black or green arrows represent the relationship of study week or diversity and abundance metrics with samples. The value of the metric increases in the direction of the arrow and the magnitude represents the overall strength of the correlation with the axes. Abbreviations are as defined in Table 2.


between variables in the phytoplankton model. Additionally, although overall the periphyton model explained a higher percent of total inertia, a larger percentage of inertia was explained uniquely by descriptor groups in the phytoplankton model than in the periphyton model (18.3% vs. 15.1%). The remaining inertia was unexplained by the measured variables. (See Table 3.) Optimized periphyton and phytoplankton models were formed of almost entirely the same variables (Fig. 9); the periphyton model did not include DO, SpC, or TP, but otherwise the compositions of the models were equivalent. No variable had a unique explained inertia, that is, the inertia explained by a variable holding all other variables constant (shown in black in Fig. 9), of more than 0.95% in periphyton models or of more than 0.86% in phytoplankton models, and mostly ranged between 0.5% and 0.8%. Inertia uniquely attributable to individual variables totaled 12.0% and 14.8% for periphyton and phytoplankton, respectively (Fig. 9). The disparity between this fraction and the total fraction uniquely explained by individual descriptor groups is the inertia shared between variables in the same descriptor group. The inertia shared between variables in the same descriptor group also accounts for the result that, by variable, inertia described by site- or site–year based variables was entirely shared (Fig. 9), but the geography descriptor group (comprised wholly of such variables) together accounted for significant unique variation (Table 4). This result also indicates that community differences between site–years could be attributed to multiple variables including geography, habitat, and nutrient and atrazine site averages and maxima. Although the range of inertia uniquely explained by variables in CCA models was fairly narrow, their rank order gives a rough assessment of relative importance to each CCA model. This rank order of variables by uniquely explained inertia differed between periphyton and phytoplankton models. However, for both models, the majority of the top 10 variables by unique inertia explained were water chemistry variables; included in both were days since ≥1 or 5 mg/L nitrate + nitrite, 0.5 mL/L TP, and 250 mg/L TSS. Additionally, flow and velocity variables were also in the set of top 10 variables for models of both periphyton and phytoplankton taxa: weekly average velocity and flow (periphyton), days since ≥50× median flow (both), 20× median flow (phytoplankton), and 0.33 m/s velocity (phytoplankton). However, examining the total percent inertia explained by hydroclimate variables, they are more important in the periphyton model. This is consistent with the variable uniquely accounting for the most inertia in each model: weekly average velocity for periphyton, and days since ≥1 mg/L nitrate/nitrite for phytoplankton. For both models, the days since ≥30 ppb atrazine variable was in the top 10 by unique inertia; 7th for periphyton (0.67%), but 2nd for

79

Table 4 Inertia explained by canonical correspondence analysis of periphyton and phytoplankton. Periphyton

Unique

Shared

Chemistry Geography Hydroclimate Atrazine Total C+H C+G H+G C+A H+A G+A C+H+G C+H+A C+G+A H+G+A C+H+G+A Total

Unexplained

Phytoplankton

Total

Proportion

Total

Proportion

0.188⁎ 0.0825⁎ 0.214⁎ 0.0523⁎ 0.537 0.119 0.239 0.0868 0.0140 0.0360 0.0249 0.252 0.0497 0.176 0.0121 0.0897 1.10 1.922

5.28% 2.32% 6.01% 1.47% 15.1% 3.34% 6.71% 2.44% 0.0039% 1.01% 0.69% 7.08% 1.39% 4.95% 0.34% 2.52% 30.9% 55.40%

0.46⁎ 0.15⁎ 0.38⁎ 0.13⁎ 1.12 0.11 0.24 0.05 0.00 0.02 0.08 0.09 0.03 0.13 0.01 0.06 0.81 4.18

7.46% 2.51% 6.19% 2.16% 18.3% 1.83% 3.91% 0.84% 0.00% 0.36% 1.33% 1.39% 0.41% 2.14% 0.11% 0.92% 13.3% 68.4%

⁎ p b 0.05.

phytoplankton (0.84%). However, other atrazine variables including daily composite atrazine concentration and days since ≥10 ppb atrazine ranked much lower in uniquely explained inertia in both models (rank 20 [0.55%]and 11 [0.33%]of 20 for periphyton and 18 [0.84%] and 12 [0.63%] of 22 for phytoplankton, respectively). Moreover, no unique inertia was ascribed to site average atrazine concentration indicating no association with longer term historical trends. Variables not included in either model were alkalinity, solar radiation, riffle/run quality, substrate quality, and % area growing soybeans and % area with claypan depth 25–50 cm and slope ≥ 3%. 3.6.2. Site, year, and seasonality gradients When the overall gradients of season (study week), year, and site are shown in the CCA plots as supplemental variables, clear patterns are apparent (Fig. 8: middle panels). In both plots, the years 2011 and 2012 are clearly separated (associated with Axis 2 in the periphyton plot and Axis 1 in the phytoplankton plot) and within each site, season gradients were consistent (seasonal chronology increased toward the upper left quadrant in the periphyton plot and toward the upper right in the phytoplankton plot). However, individual sites were distinct in the periphyton plot, but not as clearly in the phytoplankton plot. In general, the axes loadings of environmental factors in the CCA models were consistent with the RDA variation partitioning (Fig. 3). However, as

Table 3 Site–year-based environmental values for study sites. Site

IA-03

IL-17

MO-05

MO-07 N

OH-05

Year

2011

2012

IL-10 2011

2012

2012

2011

2012

2012

2012

QHEI metrics Substrate Instream cover Channel morphology Bank erosion/riparian zone Pool glide quality Riffle quality Base flow index Overall

5.7 14.3 8 2 4.3 0 23.68 Poor

5 12 9 1 4 0.3 23.68 Poor

14 13 8.7 6 2 1.3 30.84 Fair

11.3 15 11.7 3 7.3 4.3 30.84 Good

15 15.3 11 7.3 3.7 1.7 13.76 Good

1.3 12.7 9.7 5.3 2.3 0 9.11 Poor

3.7 10.7 8.7 5.3 4.7 0 9.11 Poor

14 10.3 11.7 6.7 3 1.7 13.09 Fair

7.7 11.7 7 5.3 4.3 1.7 28.06 Poor

% Area of watershed in Corn + Sorghum Soybeans Claypan depth 5–25 cm, slope 1–3% Claypan depth 5–25 cm, slope N3% Claypan depth 25–50 cm, slope 1–3% Claypan depth 25–50 cm, slope N3%

32.74% 21.10% 2.65% 24.30% 17.34% 14.95%

33.47% 26.87% 2.59% 24.02% 17.77% 14.87%

49.08% 44.02% 6.61% 1.91% 17.63% 1.04%

51.84% 43.76% 6.60% 1.91% 17.65% 1.05%

39.44% 45.51% 1.14% 0.75% 6.17% 0.84%

30.19% 48.00% 12.35% 2.22% 34.31% 0.84%

30.85% 57.69% 12.27% 2.19% 34.34% 0.83%

31.21% 33.05% 17.22% 22.27% 15.48% 3.72%

36.42% 34.42% 34.98% 8.74% 0.47% 0.05%

a

a

QHEI metrics averaged across upstream, downstream, and sampling location value.

80


Periphyton

Phytoplankton

DO

16

Hardness

10

15

NO2+NO3

14

10

Days Since 1 mg/l N

6

1

Days Since 5 mg/l N

3

5

17

20

NO2+NO3 Site Max pH

8

SpC

11

TP

2

6

Days Since 50 mg/l TSS

15

23

Days Since 250 mg/l TSS

8

3

Depth Weekly Avg

18

14

Flow Weekly Avg

5

17

16

13

12

4

4

7

Days Since 0.5 mg/l TP TP Site Max

TSS Site Avg TSS Site Max

Type Shared Unique

Velocity Weekly Avg

1

22

Days Since 0.33 m/s velocity

13

9

Water Temp Weekly Avg

9

21

Max Daily Wind Speed, Weekly Avg

19

19

20

18

Instream Cover Channel Morphology Bank Erosion/Riparian Zone Pool Glide Quality % Area of Watershed in Corn+Sorghum % Area of Watershed in Claypan Depth 25−50 cm, Slope 1−3% % Area of Watershed in Claypan Depth 5−25 cm, Slope 1−3% % Area of Watershed in Claypan Depth 5−25 cm, Slope > 3% Daily Composite Atrazine Days Since 10 ppb Atz Conc

11

12

Days Since 30 ppb Atz Conc

7

2

Atrazine Site Avg (AEMP) Conc (ppb) Atrazine Site Max (AEMP) Conc (ppb)

0%

2%

4%

6% 0%

2%

Percent Explained Inertia

4%

6%


these overall models are combinations of the four individual models, several variables are completely collinear and are therefore not included in the model or shown in the plot. ANOSIM tests confirm that these gradients are significant in structuring periphyton and phytoplankton communities. Significant differences were observed globally in periphyton and phytoplankton communities between all site–years and between months in the same site–year (R = 0.249–0.773; Supplemental Table 5). Pairwise ANOSIM tests between site–years show that phytoplankton and periphyton assemblages in all site–years were significantly different from each other (Supplemental Table 6). The lowest ANOSIM R-value for periphyton was 0.341 between IA-03 periphyton assemblages in 2011 and 2012, which would suggest a continuation of community structure between years at these sites. The highest periphyton R-values were between IL-10 2011 and MO-05 2012 (R = 0.994), MO-05 2011 and OH05 2012 (R = 0.971) and MO-05 2012 and OH-05 2012 (R = 0.969). Pairwise ANOSIM R-values for phytoplankton were lower (indicating less differentiation between sites), but still significant. The lowest ANOSIM R-value for phytoplankton is 0.094 between IA-03 and MO07N in 2012, and the largest values between MO-05 2011 and OH-05 2012 (R = 0.817), and between IL-10 2011 and OH-05 2012 (R = 0.780), MO-05 2012 (R = 0.769), and MO-07 N 2012 (R = 0.761). Additionally, periphyton and phytoplankton assemblages from different months of the same site–year were almost all significantly different (Supplemental Table 7). The number of months separating periphyton or phytoplankton samples was significantly correlated to increased R-value (Spearman correlation ρ = 0.871, p b 0.001 and ρ = 0.783, p b 0.001, respectively); that is, the farther away in time samples were taken, the greater the magnitude difference between assemblages. Succession of periphyton and phytoplankton assemblages therefore took place through the entire period of sampling. The only month pairs for which periphyton or phytoplankton assemblages were not significantly different occurred in adjacent months (periphyton: May/June IA-03, June/July MO-05, and April/May MO-07N in 2012; phytoplankton: June/July IA-03 in 2012 and IL-17 in 2013, May/June and June/July MO-05 and OH-05 in 2012). While p-values higher than α = 0.05 may be due to lower number of samples taken in 2012, R-values are also on the low side, indicating that these adjacent months may not be appreciably different in terms of assemblages. To assess to what extent these overall gradients explained community structure and to what extent the optimized explanatory environmental factors represented the gradients, variance partitioning was conducted on separate CCA models of site, season (day of year), and year, plus the optimized environmental factors, for each algal fraction (Fig. 10). For both models, the largest proportion of inertia was unexplained by the overall gradients or measured environmental factors. Large proportions of both models were likewise explained by environmental factors, but unexplained by the tested gradients. This fraction therefore corresponds to stochastic environmental variation, or to large-scale gradients we did not consider. Included in this fraction are variables having a substantial “flashy” or emphemeral characteristic, such as flow and depth, TP, TSS, solar radiation, pH, and nitrate + nitrite (see Fig. 3). A greater fraction of periphyton inertia was explained by gradients than of phytoplankton inertia. However, for both models, the gradient explaining the largest fraction of inertia was site, followed by year, and then season. Furthermore, while all inertia explained by year and site gradients was captured by the environmental factors, a fraction of inertia explained by the seasonal gradient was unrepresented by measured environmental factors. Finally, because patterns of seasonality appeared to be similar between sites and algal fractions, procrustes Protest analyses were conducted on CA (indirect gradient) ordinations (Supplemental Fig. 2) to

81

assess the similarity in overall patterns between phytoplankton or periphyton communities at different sites and also between periphyton and phytoplankton communities at each site (Supplemental Table 8). Note that each of these ordinations was constructed separately and therefore have distinct axes of inertia; however, for all ordinations, general patterns were dominated by seasonal (chronological) gradients. A large majority of site–year pairs exhibited significant concordance between their periphyton ordinations (Supplemental Table 8), which indicates similar patterns with respect to chronology (seasonality). Only the pairs of MO-05 2011/IA-03 2012, MO-07N 2012/IL-10 2012, and IA-03 2012/OH-05 2012 had statistically differing periphyton community patterns. Conversely, only about half of the site–year pairs showed concordance between their phytoplankton ordinations (Supplemental Table 8). The majority of non-significant site–year pairs contained the sites IL-17, MO-05, and OH-05 in 2012. All sites during 2011 had statistically similar chronological patterns, and pairs between these site– years (IA-03 2011, IL-10 2011, and MO-05 2011) and the sites IA-03, IL-10 and MO-07N in 2012 were also concordant. Protest analyses between periphyton and phytoplankton in the same site–year show concordance between the two fractions for the majority of site–years. Seasonal patterns were statistically different between periphyton and phytoplankton at IL-17, MO-05, and MO-07N in 2012. 3.6.3. Diversity metrics, algal groups, and indices For reference, the lower panels of Fig. 8 show diversity metrics, algal group densities, and diatom indices (for periphyton), plotted as supplemental variables onto the CCA plots for periphyton and phytoplankton, respectively. The position of these supplemental variables reflects their seasonal and annual dynamics and differences between sites as discussed previously. 4. Discussion 4.1. Community patterns The overall community patterns observed at the evaluated sites are typical of algal communities in streams and particularly in the Midwest. Diatoms were strongly dominant in periphyton and moderately so in phytoplankton at the six sites in this study; in both algal fractions, diatom dominance was strongest in the spring months. Because filamentous algae have a difficult time establishing on fine-grained substrate, diatoms are typically dominant in streams with sand or silt streambed material (Wehr and Sheath, 2003); diatom dominance has been reported for this type of stream both within the Midwest (Black et al., 2011) and elsewhere (Biggs and Smith, 2002). Additionally, as diatoms are early colonizers, they are typically dominant in streams after spring high flow events (Biggs, 1996; Hamilton et al., 1988). A switch in algal group dominance over the spring to summer months from diatoms to other groups such as chlorophyta or cyanophyta has frequently been reported in both lentic and lotic environments (Moss, 1981; Hoagland et al., 1982; Sommer, 1985; Garnier et al., 1995; Richardson et al., 2000; Peterson and Grimm, 1992). The periphyton and phytoplankton community dynamics at evaluated sites reflected this tendency to a qualified extent. Phytoplankton assemblages in most site–years in this study showed a transition to chlorophyta, euglenophyta, cyanophyta, and/or dinophyta dominance, although one site–year (IL-10 in 2011) remained dominated by diatoms at the end of the summer. Periphyton communities, on the other hand, remained dominated by diatoms throughout the study at all site–years except MO-05 in 2012 and briefly during June of 2011 at

Fig. 9. Shared and unique inertia attributable to individual factors in CCA models. Length of stacked bars represents total periphyton or phytoplankton inertia explained by each variable. The black portion of stacked bars represents the fraction of inertia uniquely associated with this variable, and the gray portion represents the fraction of inertia that is shared, or could also be explained by other variables. Numbers adjacent to black portion indicate the rank of variables by inertia uniquely explained, with 1 representing the most inertia uniquely explained.

82


1.0

Inertia Explained by: Neither gradients nor measured environmental variables (unexplained)

0.8

Proportion of Inertia

Measured environmental variables, but not gradients

Shared by Site, Season, Year

0.6

Both Year and measured environmental variables Both Season and measured environmental variables 0.4

Season, but not measured environmental variables Both Site and measured environmental variables

0.2

0.0

Periphyton

Phytoplankton

Fig. 10. Partitioning of periphyton and phytoplankton taxa inertia by gradient and environmental factors. Stacked bars represent portions of total periphyton or phytoplankton community inertia explained by combinations of gradient and environmental factors. Dark gray portion represents portion of inertia neither explained by gradients nor measured environmental factors.

IA-03. However, for all site–years in both periphyton and phytoplankton, non-diatomic algal groups increased in proportion over the study months. Algal group dynamics were also generally consistent between years for those sites sampled in both 2011 and 2012 (IA-03, IL-10, and MO-05). Studies of algal group succession in lotic systems are greatly fewer than those of lentic systems. Additionally, while a few studies have examined algal communities over time in large Midwestern rivers (Baker and Baker, 1981; Duan and Bioanchi, 2006), extant monitoring studies of algal assemblages in fine-grained streams in this region generally consist of only one time point, such as the USGS NAWQA program (USGS, 2013). This limitation makes it difficult to place the successional patterns observed at our sites in context. However, algal group patterns in this study are consistent with Leland and Porter (2000), who found in a study of sites in the upper Illinois River basin that diatoms tended to dominate throughout the summer in “minimally” and “moderately” nutrient impacted sites while chlorophytes were proportionally more abundant in “eutrophic” sites though the presence of toxic contaminants will result in lower abundances of this group. Additionally, for reference, we can compare the richness and evenness in late summer (July–September) of the periphyton and phytoplankton communities in the present study to those of stream sites of similar scale in the same Level III Ecoregions (Omernik, 1987, 1995) in the NAWQA dataset. Comparable NAWQA samples are available for periphyton in each of the five regions represented by our sites, but only available for phytoplankton in the Central and Eastern Corn Belt Plains ecoregions (in which IL-10 and OH-05 are located, respectively). The average algal taxa richness and evenness values for sites monitored in this study were generally closely in line with those from the

NAWQA dataset from the same ecoregion (Fig. 11). As in the present study, diatom taxa richness was much greater than NDA taxa richness in both periphyton and phytoplankton (approximately 2–3 times higher). Periphytic and planktonic assemblages taken from AEMP sites typically had similar or slightly greater taxa richness than the corresponding NAWQA sites. At IL-17, MO-05, and MO-07N, periphyton taxa richness was substantially higher than mean values for NAWQA sites in the same ecoregions (no comparisons were available for phytoplankton assemblages at these sites). However, planktonic diatom assemblages at the IL-10 and OH-05 sites had much higher richness and evenness than comparable NAWQA sites. Algal community evenness at the sites monitored in this study was very similar to those at the NAWQA sites, again with the exception of planktonic diatom assemblages at the IL-10 and OH-05 sites, which, although similar to other evenness values in this study, were substantially more even than communities at NAWQA sites. Despite this, these data suggest that assemblages from the sites monitored in this study are reasonably typical of local algal communities, although the number of NAWQA samples in each ecoregion varied and therefore the data may be representative of streams in each ecoregion to differing degrees. 4.2. Assemblage association with major gradients The most influential gradient in the structuring of assemblages in this study was site (Fig. 10). Characteristic differences in assemblages between sites are often observed. For example, Leland and Porter (Leland and Porter, 2000) found that variables associated with drainage basin were the most definitive structuring component to benthic algal assemblages in the upper Illinois River. Environmental variation


83

Fig. 11. Comparison of richness and evenness at study sites with NAWQA sample data in months July–September by EPA Level III Ecoregion. Circles represent mean richness or evenness at each site, with the size of circles indicating the number of samples contributing to each average. Ecoregion abbreviations: WCB = Western Corn Belt Plains, CCB = Central Corn Belt Plains, IRHV = Interior River Hills and Valleys, CIP = Central Irregular Plains, ECB = Eastern Corn Belt Plains.

associated with geographic location in this study (Fig. 3) included nutrient availability, pH, ion and mineral content, hydrology, temperature, land cover, geology, and habitat quality, variables that frequently account for large proportions of assemblage structure (Guasch et al., 1998; Stevenson, 1997; Potapova and Charles, 2002; Snyder et al., 2002). Although many of these factors exhibited variations between years or months of the study (Fig. 3), persistent differences between median values at each site were also generally observed (Fig. 2). For example, median flow at the IL-10 site was significantly higher than at other sites, and the most extreme individual flow events were observed at this site. IL-10 was also characterized by high ionic strength (e.g. pH, alkalinity, hardness and specific conductivity) and low water temperature. Significant differences between phytoplankton and periphyton assemblages of each year were also observed. Although a distinct shift in community between the two years of the study could be influenced by a number of things, the largest discernible environmental variation between the two years was hydrological. Whereas 2011 was characterized by several large rainfall and flow events, particularly at the IA-03 site, in 2012 the entire Midwest region experienced a severe drought. Flow regime and events have been shown to be especially influential for lotic systems with variable hydrology (Biggs and Close, 1989; Riseng et al., 2004; Stevenson et al., 2006). Based on the partitioning of variables along the major gradients (Fig. 3), the factors at our sites with largest variance associated with year are hydrological (depth, flow, and days since hydrological events), water chemistry variables with fluctuations typically coincident with hydrological events (nitrate/nitrite, TSS, and atrazine—particularly the “days since” variables), and habitat criteria we would expect to be influenced by hydrology (pool/glide and riffle/run quality). As only two years of data were

collected, it is unknown whether assemblages in a third year would closely resemble those of either year or whether they would be significantly different. A distinct difference in algal community observed between two years having substantially different hydrology in this study is consistent with research reporting algal community shifts with drought severity and frequency (Findlay et al., 2001; Marks et al., 2000; Ledger et al., 2008; Boix et al., 2010). Differences in hydrologic regime have also been observed to correlated to differences in the degree of effect associated with other variables such as geography (Hamilton et al., 2011) or grazing (Marks et al., 2000). However, within each year and site, the response of community structure of algal communities monitored in this study to season was significant and largely consistent (ANOSIM by month, Supplemental Table 5; procrustes analysis, Supplemental Table 8). Seasonality is typical in algal systems, both lentic (Moss, 1981; Hoagland et al., 1982; Sommer, 1985; Richardson et al., 2000) and lotic (Garnier et al., 1995; Biggs and Close, 1989). Our results imply that patterns are generally synchronous across the Midwest region from which our samples were taken. This finding is consistent with those reported by Kent et al. (2007) and Soininen et al. (2004), who argue that regional synchronicity is caused by large-scale environmental trends such as climate and hydrology. We also note that the seasonal assemblage patterns observed were generally consistent between years, suggesting that influential seasonal environmental trends were also largely consistent. In this study, variables that exhibited substantial seasonal variation included nitrogen, phosphorus, TSS, depth, flow, water temperature, wind speed and atrazine. At some levels, these factors have been observed to affect algal communities abundance and structure individually (e.g. (Stevenson et al., Algal Ecology, 1996)), and trends observed in them during this study are similar to those reported elsewhere.

84


Those site–year pairs not exhibiting concordance can be attributed to anomalous discrepancies in environment or community. For instance, in periphytic communities, site–year pairs that were not concordant also exhibited significantly different flow (Fig. 2). Lack of concordance between site–years of phytoplankton communities was much more frequent, but are likely attributable to transient blooms of individual groups. These were observed at several sites exhibiting a general lack of concordance at points that deviated from the seasonal (chronological) trend in 2012 (see Fig. 6 for algal group density data): IL-17 experienced a bloom of diatoms after a large decrease in total density in mid-May (May 15), increased density of green algae (predominantly Ankistrodesmus sp.; May 9) and euglenophyta (June 13) were observed at MO-05, and blooms of cyanophyta (April 24) and euglenophyta (July 17) occurred at OH-05. The cause of these blooms is unknown, but several of the samples that showed substantial deviations from a seasonal gradient occurred during a flow event (high water depth, Supplemental Fig. 2), which may have influenced phytoplankton community composition through resuspension, or influx of nutrients or other chemicals. These transient blooms also likely account for a lack of concordance between periphyton and phytoplankton at IL-17, MO-05, and MO-07N in 2012. Partitioning of periphyton and phytoplankton taxa inertia (Fig. 10) showed that site and year variability associated with assemblage composition was entirely represented by the factors monitored in the study. However, approximately one quarter to one-third of inertia explained by seasonality (day of year) was not also explained by the environmental parameters measured. One aspect of seasonality that was not considered in this study is the effect of interspecies relationships, including competition and predation, which may vary over the spring and summer months. Grazing pressure in particular is known to affect algal assemblage structure (Muñoz et al., 2001; Murdock et al., 2010), and therefore life cycle of aquatic invertebrates would be expected to influence algal community. However, other factors that may be influential but were not measured in this study are specific dissolved and particulate fractions of phosphorus, the dynamics of upstream algal populations and conditions, the presence of other pesticides, and other watershed management practices. 4.3. Inertia explained by variable group and by individual variables Partitioning of the periphyton and phytoplankton community inertia explained by CCA models by either variable groups or individual variables showed that a substantial proportion of explained inertia was shared between variable groups or between two or more individual variables. Based on our first year of data and on the dynamics of variables monitored in this study (see Figs. 2 and 8a and b) we expected a considerable amount of shared inertia due to multicollinearity. The variation partitioning methodology used in this study was selected to control for this multicollinearity, particularly that of atrazine concentrations and other environmental variables, and to isolate any changes in community that can be uniquely associated with particular variables in our dataset. Because it is unknown which, if any, variable or variable groups truly account for shared inertia, this is the most conservative assessment of influence. Separating any inertia that could be attributed to two or more groups or variables resulted in necessarily smaller percentages of inertia associated uniquely with factor groups (1.5–7.5%) or with individual variables (0–0.95%) in CCA models of periphyton and phytoplankton. Additionally, because the geographic variables measured in this study were all defined on site or site–year, and were therefore limited to at most 9 values, the variations of these variables taken together were entirely multicollinear, and thus no unique inertia was able to be ascribed to them. A larger difference in periphyton communities by site than phytoplankton communities is therefore also the likely origin of greater percentages of shared variation in the periphyton CCA model than the phytoplankton model. Despite these limitations, the

percentages of inertia uniquely attributable to variable groups or individual variables provide a valuable assessment of their relative influence on periphyton and phytoplankton assemblage structure. We found in this analysis that these percentages are generally consistent with the conclusions of published studies. When periphyton and phytoplankton inertias were partitioned by individual variable, many of the variables showing greatest influence (having relatively greater percentages of inertia uniquely associated with them) were variables constructed based on thresholds. The concept and estimation techniques of ecological thresholds, points at which the response of a species, community, or ecological process change abruptly, have been well-developed and accepted (May, 1977; Toms and Lesperance, 2003; Groffman et al., 2006; Huggett, 2005; Andersen et al., 2008). We would expect that variables associated with potential disturbances (such as scour, nutrients, or herbicides) would exhibit non-linear behavior and thresholds at which they become influential to the algal community. In this analysis, for nearly all parameters with multiple derived variables, at least one variable constructed based on thresholds was more influential to community models than the absolute value (the exception being weekly average velocity in the periphyton model); this suggests that responses of community structure to these parameters are non-linear or discontinuous, and supports the inclusion of thresholds in the set of potentially influential variables. Furthermore, the relative influence of variables constructed from different thresholds suggests the appropriate threshold for each variable and algal community. For example, in the periphyton model, the days since ≥5 mg/L nitrate/nitrite accounted for a greater unique percentage of inertia than did the days since 1 mg/L (0.76% and 0.67%, respectively), whereas the reverse was true for phytoplankton (0.77% and 0.86%, respectively). The most influential nitrate/nitrite threshold for phytoplankton was therefore similar to TN calculated by Black et al. with respect to several algal metrics, which ranged from 0.59– 1.5 mg/L, however the most influential threshold for periphyton in this study was higher than this range. The ability of periphytic communities to respond to nutrient concentrations has been shown to be limited by high and frequent hydrological events (Biggs and Close, 1989), which may explain this discrepancy in influence. In contrast, the days since ≥250 mg/L TSS was consistently more influential than the days since ≥ 50 mg/L for both algal communities (periphyton: 0.60% vs. 0.50%; phytoplankton: 0.83 vs. 0.40%). Similarly, the influence of the days since ≥ 30 μg/L atrazine variable was higher in models of both algal communities than of the days since ≥ 10 μg/L (the set of events ≥ 10 μg/L also including events ≥ 30 μg/L), and both were of greater influence than the daily composite atrazine concentration (periphyton: 0.67% and 0.56% vs. 0.33%; phytoplankton: 0.84% and 0.63% vs.0.54%). That the ≥30 μg/L threshold is more influential than the ≥ 10 μg/L threshold indicates that the set of pulses ≥ 10 μg/L but b30 ppb are considerably less influential to the community structure. Additionally, the collective inertia explained by the entire group of atrazine variables was low (periphyton: 1.47%; phytoplankton: 2.16%) and was the least of all variable groups. This result is consistent with numerous studies that show significant effects on algal community structure are not typically observed at concentrations below 30 μg/L atrazine. The relative influence (ranked by inertia uniquely explained) of variable groups and individual variables was likewise generally consistent with reported data. The two variable groups explaining the most unique inertia in both periphyton and phytoplankton assemblages were water chemistry and hydroclimate (periphyton: 5.28% and 6.01%; phytoplankton: 7.46% and 6.19%; Table 4). Additionally, these variable groups contained some of the most independently influential variables in the models: weekly average velocity and water temperature (periphyton), SpC (phytoplankton), hardness (periphyton), and days since a nitrate/ nitrite concentration of ≥ 1 mg/L and ≥ 5 mg/L, a TP concentration of ≥0.5 mg/L, a TSS concentration of ≥250 mg/L, or a ≥20× or 50× flow event (Fig. 9). The effects of variables in these groups on algal


community have been well-established, and are typically those which show some of the greatest correlation with community (Horner and Welch, 1981; Biggs and Smith, 2002; Biggs et al., 2005; Villeneuve et al., 2011; Biggs and Close, 1989; Riseng et al., 2004; Stevenson et al., 2006; Dodds et al., 2002; Osborne and Wiley, 1989; DeNicola, 1996). However, the group uniquely explaining the most inertia differed between periphyton assemblages (for which it is hydroclimate) and for phytoplankton (for which it is water chemistry). Likewise, overall, individual hydroclimate variables were more influential in the periphyton model and water chemistry variables more influential in the phytoplankton model (Fig. 9). The differing ecology of periphyton and phytoplankton largely accounts for the switch in variable group principally explaining inertia. Because periphyton is attached to a fixed substrate from which they may be scoured by high velocity, hydrology is particularly important to the structure of these assemblages. High flow and velocities have been associated with decreased cell density, biomass, richness, and evenness as well as assemblage shifts (Horner and Welch, 1981; Biggs and Smith, 2002; Biggs et al., 2005; Villeneuve et al., 2011). Sites with high overall flow are often characterized by species capable of withstanding scour events, such as adnate and prostrate algal forms. For lotic systems with variable flow (sites such as those observed here), flow regime is especially important to algal community structure (Biggs and Close, 1989; Riseng et al., 2004; Stevenson et al., 2006). Biggs and Close (1989) concluded that in streams with relatively constant flow, nutrient limitation and light play a greater role than hydrology in structuring periphyton assemblages, but in streams with variable flow, flow intensity and frequency are more influential. We further expect that nutrient uptake in periphyton will be moderated by mass transfer through the periphytic layer, which may explain why temperature (part of the hydroclimate group), which generally increases both mass transfer and also cellular biochemical processes (DeNicola, 1996), is of greater influence in the periphyton model than in the phytoplankton one (9th most influential vs. 21st most influential, Fig. 9). Phytoplankton, in contrast, experience flow in a Lagrangian as opposed to an Eulerian sense, and therefore while pulses of nutrients and other inputs may contact periphyton for a relatively limited time, phytoplankton travel with a pulse and may therefore be affected to a greater extent by proximate water chemistry. Additionally, some water quality measures such as sediment would be expected to be related to phytoplankton community in other direct ways; here TSS is in part a proxy measurement of phytoplankton abundance as cells are suspended solids. Both the water chemistry group and generally individual variables of this group were therefore more influential in phytoplankton models than in periphyton models. Moreover, the effects of conditions upstream of the sampling point may be influential to the sampled periphyton community; upstream effects not accounted for, such as prevalence of grazers or shading, may explain why the total community inertia explained for phytoplankton was lower than for periphyton. Some variables in the water chemistry and hydroclimate groupings accounted for unique portions of the total inertia, but were not as highly influential. These included pH (ranked 17th and 20th for periphyton and phytoplankton, respectively), days since ≥ 50 mg/L TSS, depth, DO, days since ≥ 5 × median flow, and wind speed. We hypothesize that these variables may possibly affect community by way of interaction with other variables or, as discussed above, may indicate thresholds that are non-optimal. For both periphyton and phytoplankton, geography and atrazine variable groups accounted for less explained inertia than did water chemistry and hydroclimatic variables overall. Although some QHEI habitat variables such as substrate quality or pool/glide quality would be expected to have a direct influence on algal assemblages, particularly those of periphyton, other geographic variables such as land use or base flow index would be predicted to have an indirect effect. Environmental controls on community have been hypothesized to be hierarchical, with

85

direct controls exerting more immediate and observable influence on community structure and indirect ones limiting the bounds of those influences (Poff, 1997; Naiman et al., 2000; Biggs et al., 2005; Parsons and Thoms, 2007). Because water chemistry and hydroclimatic factors are understood to typically have a more immediate effect on algal survival and metabolism than geography, the order of influence of variable groups observed in this study is expected. Additionally, as noted, the high level of multicollinearity in geographic variables and relatively small number of site–years in this study resulted in not being able to assess the influence of individual geographic parameters. However, the shared inertia explained by geographical variables (Fig. 9) suggests that land use (area of watershed growing corn and sorghum) and claypan depth/slope of watershed are relatively important. Corn and sorghum are crops on which atrazine and other agricultural chemicals are frequently used. Likewise, the watershed frequency of claypan depth of 25–50 cm with a slope of 1–3% have been linked to greater likelihood of high atrazine concentrations (Miller, 2009; Miller et al., 2012). Of the habitat variables, based on shared inertia pool/glide quality and channel morphology appear to be most influential; these variables would by hypothesized to be a measurement of areas of appropriate low velocity and depth for algal growth. As a group, the atrazine variables explain the least unique inertia. This finding is consistent both with the first year of data in this study, which suggested that other environmental variables have a greater effect on community structure relative to atrazine (Andrus et al., 2013), and with other studies that found that seasonal or other environmental changes had a larger role in structuring algal community composition than did concentrations of atrazine (Lynch et al., 1985; Berard and Benninghoff, 2001; Guasch et al., 1998). Further, the algal communities at the sites monitored in this study may be more tolerant in general to anthropogenic inputs. As noted above, these sites are in general diatom dominated, particularly periphytic assemblages and both periphyton and phytoplankton during the spring, when atrazine concentrations are frequently elevated. Diatoms and cyanophyta have been shown to be less sensitive to atrazine than other algal groups, particularly compared to chlorophyta, which typically show the highest sensitivity to atrazine (Goldsborough and Robinson, 1986; Kosinski and Merkle, 1984; Guasch et al., 1997, 1998, 2007; Dorigo et al., 2004; Guanzon et al., 1996; Tang et al., 1998; Brain et al., 2012a, 2012b). It could therefore be expected that diatomdominated communities such as those in this study would be less sensitive overall to atrazine. Additionally, those species showing a high tolerance to anthropogenic disturbance generally also are tolerant to atrazine and other agricultural chemicals (Guasch et al., 1998; Larras et al., 2012; Berthon et al., 2011; Rumeau and Coste, 1988). The mean Pollution Index (where Most Pollution Sensitive species have a score of 1 and Most Pollution Tolerant species have a score of 3; Bahls, 1993; Lange-Bertalot, 1979) at our sites reflect an overall moderate to high pollution tolerance (IA-03: 1.83; IL-10: 2.15; IL-17: 1.98; MO-05: 1.79; MO-07 N: 1.89; OH-05: 2.08). We would therefore expect atrazine concentrations to have less of an observable influence on community structure at the monitored sites as all sites had communities generally tolerant to anthropogenic disturbance. The origin of pollution tolerance at these sites is unclear, but most likely developed in response to multiple anthropogenic sources such as high nutrient loads, sediment, and chemical inputs that are common in agricultural watersheds. Although the PTI metric seems to trend inversely with the historical average concentrations of atrazine (i.e., lowest concentrations at IL-10 and OH-05), IL-17 has experienced some of the highest atrazine concentrations in the past few years (228.18 μg/L in 2011), but is not in the statistical group of lowest Pollution Index. Black et al. (2011) found that the percentage of Most Pollution Tolerant (MPT) diatom species is significantly correlated to nitrogen and phosphorus concentration; all sites evaluated in this study had maximum concentrations of nitrogen and phosphorus in excess of the thresholds of effects reported (Black et al., 2011) for MPT diatoms.

86


Moreover, that atrazine variables explained the least unique inertia of the variable groupings may also reflect an interaction between the groups. For periphyton, inertia shared between variable groups was twice that uniquely attributable to a particular group (30.9% vs. 15.1%); for phytoplankton, inertia shared between groups was slightly less than unique inertia, but still substantial (13.3% vs. 18.3%). Specifically, inertia shared between the atrazine group and other groups was considerable for both algal fraction (10.9% for periphyton and 5.3% for phytoplankton). This result is consistent with an interaction between atrazine variables and other variables on algal assemblage, but may alternatively indicate co-variation with other variables influential to community structure rather than any causality due to atrazine. Because of the concurrent variation of some parameters, it is not possible to completely assess their individual influence. However, the single variable days since ≥30 μg/L atrazine was influential in models of both periphyton and phytoplankton, ranked 7th and 2nd most influential, respectively, though these only account for 0.67% and 0.84% of explained inertia. This threshold is infrequently exceeded both in the streams monitored in this study and in the general population of agricultural streams in the Midwest (Solomon et al., 1996). Additionally, as noted above, the influence of this threshold is consistent with many studies demonstrating that significant effects on algal community metrics and structure are not realized below 30 μg/L of atrazine (Giddings, 2012; USEPA, 2012; Krieger et al., 1988; Brockway et al., 1984). Furthermore, that this variable is more influential than one using a lower threshold (days since ≥ 10 μg/L atrazine), the absolute measurement (daily composite atrazine concentration), or atrazine site averages or maxima suggests the recovery of algal communities with distance from an atypically high atrazine concentration. Implicit in the time since threshold exceedence variables used in this study is the concept of adaptation or recovery with time. Both the periphyton and phytoplankton models suggest that community shifts abate with greater time elapsed since an atrazine perturbation. The suggestion of recovery from atrazine pulses is consistent with a growing body of literature demonstrating recovery of algal species and communities after exposure to atrazine. Atrazine acts by inhibiting photosystem-II of photosynthesis through competitive and noncovalent binding of the QB site of the D1 protein (Jensen et al., 1977; Jursinic and Stemler, 1983; Shimabukuro et al., 1970). Consequently, inhibition is reversible immediately upon removal of atrazine, and recovery of both terrestrial and aquatic plant communities under a variety of conditions has been demonstrated (Jensen et al., 1977; Shimabukuro et al., 1970; Klaine et al., 1996; Brain et al., 2012a, 2012b; Brockway et al., 1984; Hughes et al., 1988; Jones et al., 1986; Juttner et al., 1995; Teodorović et al., 2012; Mohammad et al., 2008, 2010; Stay et al., 1989; Vallotton et al., 2008). In a variety of algal species, both biochemically and physiologically, recovery of photosynthesis and growth rate occur immediately, reaching control levels within hours following exposure to atrazine at concentrations much higher than typically encountered in situ (250–1000 μg/L) (Vallotton et al., 2008; Brain et al., 2012a, 2012b; Hughes et al., 1988). Several micro/mesocosm studies have also demonstrated species, community, and functional recovery of autotrophs exposed to atrazine at a variety of exposure durations and concentrations (Brockway et al., 1984; Hoagland et al., 1993; Krieger et al., 1988; Moorhead and Kosinski, 1986). Most recently, recovery of periphyton communities sampled from the sites monitored in this study within 24 h at concentrations up to 320 μg/L was demonstrated (Prosser et al., 2013). 5. Limitations Although this analysis examined the impact of many variables on algal community, several factors were not considered due to expense, time constraints or access. In particular, though numerous recent studies have focused on the effects of combined pollutants on algal communities (Faust et al., 2003; Relyea, 2005, 2009; Backhaus and Faust, 2012),

because of the frequency of sampling and associated expense, atrazine was the only pesticide measured at sites in this study. Based on land use, the presence of other crop protection compounds at the monitored sites is likely (Fairchild et al., 1998; Faust et al., 2003; Tlili et al., 2011). Our results therefore exclude possible interactions between atrazine and any other agricultural chemicals at the test sites, but may also erroneously ascribe effects to atrazine that could be associated with other coincident compounds. Additionally, the available environmental data was at several different frequencies. Although we believe that including the best available data lead to the most accurate results, having uniform frequency for each variable would have increased the power of the analysis. Furthermore, this study did not consider other trophic levels than algae. Grazing pressure has been shown to influence algal assemblage structure (Muñoz et al., 2001; Murdock et al., 2010), and we therefore would expect additional community inertia would be explained by grazing populations. Finally, this study only collected periphyton from one type of substrate (slate) near the top of the water column and from communities that were one to four months old. Although lithophytic periphyton is an important community in many streams, the results of this study may not be representative of periphyton living on wood, benthic sands or silts, or macrophytes. 6. Conclusions The current study of algal communities in six Midwestern agricultural watersheds in the years 2011 and 2012 sought to characterize algal community structure and dynamics, and environmental parameters, including atrazine concentrations, and to relate the two when possible. Periphytic communities at these sites were highly dominated by diatoms, as were planktonic assemblages in the springtime. However, phytoplankton community succession was largely consistent with typically reported patterns, with cyanophyta increasing in the months of May and June, and green algae and euglenophyta gaining dominance in the mid-to-late summer. Periphytic cyanophyta, green algae, and euglenophyta also increased over the sampling period for many site–years, but their proportion in periphyton was much less than in phytoplankton. Significant site and temporal differences were observed in univariate diversity and abundance metrics, diatom indices, and periphyton and phytoplankton assemblages. These differences were related to the large-scale gradients of site, season, and year. In community assemblages, the most influential gradient was year, which distinctly ordered samples in CCA analyses. Moreover, the seasonal patterns of temporal change in algal assemblages were largely similar between sites and between fractions (periphyton and phytoplankton) at each site, demonstrating synchrony of community reactions. Models of algal univariate metrics and assemblages show several consistent patterns. First, the inclusion of parameters with seasonal dynamics in models reflects algal community succession and known seasonal abundance patterns. Geographic variables are also prevalent in all models, and imply landscape-scale controls on algal communities. Both scour effects from stochastic events and differences in hydrology between 2011 and 2012 are visible in these models. Variation partitioning of these models show that, for diversity and abundance metrics, the amount of variation uniquely explained by individual variable groups was generally of the order shared variation N water chemistry N geography N hydroclimate N atrazine. For CCA models of periphyton assemblages, this order was: shared variation N hydroclimate N water chemistry N geography N atrazine, and for phytoplankton assemblage the order was: shared variation N water chemistry N hydroclimate N geography N atrazine. Collectively, atrazine variables uniquely accounted for no more than 2.2% of inertia in assemblage models. That the uniquely explained variance or inertia of the days since ≥ 30 ppb atrazine variable (reflecting pulsed exposures of high magnitude) in assemblage models was larger than that of other


atrazine variables is consistent with the collective laboratory and field data concluding that effects in primary producers do not begin to manifest until exposure concentrations exceed 30 ppb. Moreover, this analysis is consistent with the pulsatile behavior of atrazine exposure, where atrazine variables associated with peaks (e.g. days since ≥30 ppb) were of considerably greater consequence to algal community structure than variables representing daily or longer term exposure trends (daily composite and site average atrazine concentrations), implying a transitory nature (i.e. recovery of some form) of community changes potentially associated with atrazine, which is consistent with a growing body of literature (Brain et al., 2012a, 2012b; Vallotton et al., 2008; Laviale et al., 2011; Prosser et al., 2013). Finally, in all models, substantial variation or inertia was shared between variable groups, suggesting that considerable ambiguity exists in determining the effects of environment on algal communities in these systems. Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.scitotenv.2014.09.033. References Akaike H. A new look at the statistical model identification. IEEE Trans Autom Control 1974;19:716–23. Allan JD. Landscapes and riverscapes: the influence of land use on stream ecosystems. Annu Rev Ecol Evol Syst 2004;35:257–84. Andersen T, Carstensen J, Hernandez-Garcia E, Duarte CM. Ecological thresholds and regime shifts: approaches to identification. Trends Ecol Evol 2008;24:49–57. Andrus JM, Winter D, Scanlan M, Sullivan S, Bollman W, Waggoner JB, et al. Seasonal synchronicity of algal assemblages in three Midwester agricultural streams having varying concnetrations of atrazine, nutrients, and sediment. Sci Total Environ 2013; 458–460:125–39. Austin MP, Greig-Smith P. The application of quantitative methods to vegetation survey. II. Some methodological problems of data from rain forest. J Ecol 1968;56:827–44. Backhaus T, Faust M. Predictive environmental risk assessment of chemical mixtures: a conceptual framework. Environ Sci Tech 2012;46:2564–73. Bahls LL. Periphyton bioassessment methods for Montana streams. Helena, MT: Montana Water Quality Bureau; 1993. Baker K, Baker A. Seasonal succession of the phytoplankton in the upper Mississippi River. Hydrobiologia 1981;83:295–301. Baxter LR, Moore DL, Sibley PK, Solomon KR, Hanson ML. Atrazine does not affect algal biomass or snail populations in microcosm communities at environmentally relevant concentrations. Environ Toxicol Chem 2011;30:1689–96. Berard A, Benninghoff C. Pollution-induced community tolerance (PICT) and seasonal variations in the sensitivity of phytoplankton to atrazine in nanocosms. Chemosphere 2001;45:427–37. Berard A, Leboulanger L, Pelte T. Tolerance of Oscillatoria limnetica Lemmermann to atrazine in natural phytoplankton populations and in pure culture: influence of season and temperature. Environ Contam Toxicol 1999;37:472–9. Berthon V, Bouchez A, Rimet F. Using diatom life-forms and ecological guilds to assess organic pollution and trophic level in rivers: a case study of rivers in south-eastern France. Hydrobiologia 2011;673:259–71. Biggs BJ, Close ME. Periphyton biomass dynamics in gravel bed rivers: the relative effects of flows and nutrients. Freshw Biol 1989;22:209–31. Biggs BJ. Patterns of benthic algae in streams. In: Stevenson RJ, et al, editors. Algal ecology: freshwater benthic ecosystems. Waltham, MA, US: Academic Press; 1996. p. 31–56. Biggs BJ, Smith RA. Taxonomic richness of stream benthic algae: effects of flood disturbance and nutrients. Limnol Oceanogr 2002;47:1175–86. Biggs BJ, Nikora VI, Snelder TH. Linking scales of flow variability to lotic ecosystem structure and function. River Res Appl 2005;21:283–98. Black RW, Moran PW, Frankforter JD. Response of algal metrics to nutrients and physical factors and identification of nutrient thresholds in agricultural streams. Environ Monit Assess 2011;175:397–417. Boix D, Garcia-Berthou E, Gascon S, Benejam L, Tornes E, Sala J, et al. Response of community structure to sustained drought in Mediterranean rivers. J Hydrol 2010;383: 135–46. Borcard D, Legendre P, Drapeau P. Partialling out the spatial component of ecological variation. Ecology 1992;73:1045–55. Brain RA, Arnie J, Porch J, Hosmer A. Recovery of photosynthesis and growth rate in green, blue-green, and diatom algae after exposure to atrazine. Environ Toxicol Chem 2012a;31:2572–81. Brain RA, Hosmer AJ, Desjardins D, Kendall TZ, Krueger HO, Wall SB. Recovery of duckweed from time-varying exposure to atrazine. Environ Toxicol Chem 2012b;31:1121–8. Bray JR, Curtis JT. An ordination of upland forest communities of southern Wisconsin. Ecol Monogr 1957;27:325–49. Breiman L. Random forests. Mach Learn 2001;45:5–32. Brockway DL, Smith PD, Stancil FE. Fate and effects of atrazine on small aquatic microcosms. Bull Environ Contam Toxicol 1984;32:345–53. Clarke KR. Nonparametric multivariate analyses of changes in community structure. Aust J Ecol 1993;18:117–43. Clarke KR, Green RH. Statistical design and analysis for a biological effects study. Mar Ecol Prog Ser 1988;46:213–26.

87

Debenest T, Pinelli E, Coste M, Silvestre J, Mazella N, Madigou C. Sensitivity of freshwater periphytic diatoms to agricultural herbicides. Aquat Toxicol 2009;93:11–7. DeNicola DM. Periphyton responses to temperature at different ecological levels. In: Stevenson RJ, Bothwell ML, Lowe RL, editors. Algal ecology: freshwater benthic ecosystems. San Diego: Academic Press; 1996. p. 149–81. Dodds WK, Smith VH, Lohman K. Nitrogen and phosphorus relationships to benthic algal biomass in temperate streams. Can J Fish Aquat Sci 2002;59:865–74. Dorigo U, Bourrain X, Berard A, Leboulanger C. Seasonal changes in the sensitivity of river microalgae to atrazine and isoproturon along a contamination gradient. Sci Total Environ 2004;318:101–14. Duan S, Bioanchi TS. Seasonal changes in the abundance and composition of plant piugments in particular organic carbon in the lower Mississippi and Pearl rivers. Estuar Coasts 2006;29:427–42. Fairchild JF. Structural and functional effects of herbicides on non-target organisms in aquatic ecosystems with an emphasis on atrazine. In: Kortkamp A, editor. Herbicides and environment. Rijeka, Croatia: InTech; 2011. p. 383–404. Fairchild JF, Ruessler DS, Carlson AR. Comparative sensitivity of five species of macrophytes and six species of algae to atrazine, metribuzin, alachlor, and metalochlor. Environ Toxicol Chem 1998;17:1830–4. Faust M, Altenburger R, Backhaus T, Black H, Boedeker W, Gramatica PH, et al. Joint algal toxicity of 16 dissimilarly acting chemicals is predictable by the concept of independent action. Aquat Toxicol 2003;63:43–63. Ferenczi J, Ambrus A, Wauchope RD, Sumner HR. Persistence and runoff losses of 3 herbicides and chlorpyrifos from a corn field in the Lake Balaton watershed of Hungary. J Environ Sci Health 2002;37:211–24. Findlay DL, Kasian SE, Stainton MP, Beaty K, Lyng M. Climatic influences on algal populations of boreal forest lakes in the esperimental lakes area. Limnol Oceanogr 2001;46: 1784–93. Foley JA, DeFries R, Asner GR, Barford C, Bonan G, Carpenter SR, et al. Global consequences of land use. Science 2005;309:570–4. Garnier J, Billen G, Coste M. Seasonal succession of diatoms and chlorophyceae in the drainage nework of the Seine River—observations and modeling. Limnol Oceanogr 1995;40:750–65. Genuer R, Poggi J-M, Tuleau-Malot C. Variable selection using Random Forests. Pattern Recognit Lett 2010;31:2225–36. Giddings JM. Atrazine: review of mesocosm and microcosm studies. Lakewood, WA: Compliance Services International (CSI); 2012. Giddings JM, Anderson TA, Hall Jr LW, Hosmer AJ, Kendall RJ. Atrazine in North American surface waters: a probabilistic aquatic ecological risk assessment. SETAC 26th Annual Meeting in North America. Pensacola, FL: Society of Environmental Toxicology and Chemistry; 2005. Giraudoux P. pgirmess: data analysis in ecology. R package version 1.5.8. Retrieved from http://CRAN.R-project.org/package=pgirmess, 2013. Goff FG. Comparison of species ordinations resulting from alternative indices of interspecific association and different numbers of included species. Vegetatio 1975; 31:1–14. Goldsborough LG, Robinson GG. Changes in periphytic algal community structure as a consequence of short herbicide exposures. Hydrobiologia 1986;139:177–92. Gower JC. Statistical methods of comparing different multivariate analyses of the same data. Mathematics in the archaeological and historical sciences. Edinburgh, UK: Edinburgh University Press; 1971. Gower JC. Generalized Procrustes analysis. Psychometrika 1975;40:33–51. Gower, JC. Introduction to ordination techniques. In: Developments in numerical ecology. (Vols. NATO ASI Series, Vol. G14.). Berlin, Germany: Springer-Verlag; 1987. Groffman P, Baron JS, Blett T, Gold AJ, Goodman I, Gunderson L, et al. Ecological thresholds: the key to successful environmental management or an important concept with no practical application? Ecosystems 2006;9:1–13. Gruessner B, Watzin MC. Respons of aquatic communities from a Vermont stream to environmentally realistic atrazine exposure in laboratory microcosms. Environ Toxicol Chem 1996;15:410–9. Guanzon NG, Fukuda M, Nakahara H. Accumulation of agricultural pesticides by three freshwater microalgae. Fish Sci 1996;62:690–7. Guasch H, Muñoz I, Roses N, Sabater S. Changes in atrazine toxicity throughout succession of stream periphyton. J Appl Phycol 1997;9:137–46. Guasch H, Ivorra N, Lehmann V, Paulsson M, Real M, Sabater S. Community composition and sensitivity of periphyton to atrazine in flowing waters: the role of environmental factors. J Appl Phycol 1998;10:203–13. Guasch H, Lehmann V, van Beusekom B, Sabater S, Admiraal W. Influence of phosphate on the response of periphyton to atrazine exposure. Arch Environ Contam Toxicol 2007; 52:32–7. Hall LH, Killen WD, Anderson RD. A 2012 assessment of periphyton, aquatic macrophytes, benthic communities, and physical habitat in the Midwestern United States coinciding with varying levels of atrazine and other physiochemcial attribute. Greensboro, NC: Syngenta Crop Protection; 2012. Hamala JA, Kollig HP. The effects of atrazine on periphyton communitieis in contolled laboratory ecosystems. Chemosphere 1985;14:1391–408. Hamilton DP, Mitchell SF. Wave-induced shear stresses, plant nutrients and chlorophyll in seven shallow lakes. Freshw Biol 1997;38:159–68. Hamilton PB, Jackson GS, Kaushi NK, Solomon KR, Stephenson GL. The impact of two applications on atrazine on the plankton communities of in situ enclosures. Biomass 1988;13:123–39. Hamilton PB, Lavoie I, Ley LM, Poulin M. Factors contributing to the spatial and temporal variability of phytoplankton communities in the Rideau River (Ontario, Canada). River Syst 2011;19:189–205. Helsel DR. Nondectects and data analysis; statistics for censored environmental data. Hoboken, NJ: John Wiley and Sons; 2005.

88


Hoagland KD, Roemer SC, Rosowski JR. Colonization and community structure of two periphyton assemblages, with emphasis on the diatoms (Bacillariophyceae). Am J Bot 1982;69:188–213. Hoagland KD, Denner RW, Smith JD, Cross DR. Freshwater community responses to mixtures of agricultural pesticides: effects of atrazine and bifenthrin. Environ Toxicol Chem 1993;12:627–37. Horner RR, Welch EB. Stream periphyton development in relation to current velocity and nutrients. Can J Fish Aquat Sci 1981;38:449–57. Hothorn T, Hornik K, Strobl C, Zeileis A. party: a laboratory for recursive partytioning. R package version 1.0-13. Retrieved from http://CRAN.R-project.org/package=party, 2013. Huber W. Ecotoxicological relevance of atrazine in aquatic systems. Environ Toxicol 1993; 12:1865–81. Huggett AJ. The concept and utility of ‘ecological thresholds’ in biodiversity conservation. Biol Conserv 2005;124:301–10. Hughes JS, Alexander MM, Balu K. Evaluation of appropriate expressions of toxicity in aquatic plant bioassays as demonstrated by the effects of atrazine on algae and duckweed aquatic toxicology and hazard assessment, vol. 10. Philadelphia: American Society for Testing and Materials; 1988. Jackson DA. PROTEST: a PROcrustean randomization TEST of community environment concordance. Ecoscience 1995;2:297–303. Jensen KIN, Stephenson GR, Hunt LA. Detoxification of atrazine in three gramineae subfamilies. Weed Sci 1977;25:212–20. Jones TW, Kemp WM, Estes PS, Stevenson JC. Atrazine uptake, photosynthetic inhibition, and short-term recovery for the submersed vascular plant, Potamogeton perfoliatus L. Environ Contam Toxicol 1986;15:277–83. Julius ML, Theriot EC. The diatoms: a primer. In: Smol J, Stoermer E, editors. The diatoms: applications for the environmental and earth sciences. 2nd ed. Cambridge, UK: Cambridge University Press; 2010. p. 8–22. Jurgensen TA, Hoagland KD. Effects of short-term pulses of atrazine on attached algal communities in a small stream. Arch Environ Contam Toxicol 1990;19:617–23. Jursinic P, Stemler A. Changes in [14C]atrazine binding associated with the oxidation– reduction state of the secondary quinone acceptor of photosystem II. Plant Physiol 1983;73:703–8. Juttner I, Peither A, Lay JP, Kettrup A, Ormerod SJ. An outdoor mesocosm study to assess ecotoxicological effects of atrazine on a natural plankton community. Arch Environ Contam Toxicol 1995;29:435–41. Kent AD, Yannarell AC, Rusak JA, Triplett EW, McMahon KD. Synchrony in aquatic microbial community dynamics. ISME J 2007;1:38–47. Klaine SJ, Dixon KR, Benjamin RB, Florian JD. Characterization of Selenastrum capricornutum response to episodic atrazine exposure. Greensboro, NC: The Institute of Wildlife and Environmental Toxicology; 1996. Kosinski RJ, Merkle MG. The effect of four terrestrial herbicides on the productivity of artificial stream algal communities. J Environ Qual 1984;13:75–82. Krieger KA, Baker DB, Kramer JW. Effects of herbicides on stream Aufwuchs productivity and nutrient uptake. Arch Environ Contam Toxicol 1988;17:299–306. Kroeze C, Seitzinger SP. Nitrogen inputs to rivers, estuaries and continenetal shelves and related nitrous oxide emissions in 1990 and 2050: a global model. Nutr Cycl Agroecosyst 1998;52:195–212. Kruskal WH, Wallis WA. Use of ranks in one-criterion variance analysis. J Am Stat Assoc 1952;47:583–621. Lakshminarayana JS, O'Neill HJ, Jonnavithula SD, Leger DA, Milburn PH. Impact of atrazine-bearing agricultural tile drainage discharge on planktonic drift of a natural stream. Environ Pollut 1992;76:201–10. Lange-Bertalot H. Pollution tolerance as a criterion for water quality estimation. Nova Hedwigia 1979;64:285–304. Lange-Bertalot H, Bak M, Witkowski A, Tagliaventi N. Diatoms of Europe: diatoms of the European inland waters and comparable habitats. In: Lange-Bertalot H, editor. Eunotia and some related genera, vol. 6. Koenigstein, Germany: Gantner; 2011. Larras FL, Bouchez A, Rimet R, Montuelle B. Using bioassays and species sensitivity distributions to assess herbicide toxicity towards benthic diatoms. PLoS One 2012;7: e44458. Larsen DP, deNoyelles Jr F, Stay F, Shiroyama R. Comparisons of single-species, microcosm and experimental pond responses to atrazine exposure. Environ Toxicol Chem 1986; 5:179–90. Laviale M, Morin S, Creach A. Short term recovery of periphyton photosynthesis after pulse exposition to the photosystem II inhibitors atrazine and isoproturon. Chemosphere 2011;84:731–4. Ledger ME, Harris RM, Armitage PD, Milner AM. Disturbance frequency influences patch dynamics in stream benthic algal communities. Oecologia 2008;155:809–19. Legendre P. Numerical ecology. 3rd ed. Amsterdam: Elsevier Science BV; 2012. Leira M, Sabater S. Diatom assemblages distribution in catalan rivers, NE Spain, in relation to chemical and physicographical factors. Water Res 2005;39:73–82. Leland HV, Porter SD. Distribution of benthic algae in the upper Illinois River basin in relation to geology and land use. Freshw Biol 2000;44:279–301. Lopaka L. NADA: nondetects and data analysis for environmental data. R package version 1.5-4. Retrieved from http://CRAN.R-project.org/package=NADA, 2012. Lynch T, Johnson HE, Adams WJ. Impact of atrazine and hexachloropbiphenyl on the structure and function of model stream ecosystems. Environ Toxicol Chem 1985;4: 399–413. Malmqvist B, Rundle S. Threats to the running water ecosystems of the world. Environ Conserv 2002;29:134–53. Marks JC, Power ME, Parker MS. Flood disturbance, algal productivity, and interannual variation in food chain length. Oikos 2000;90:20–7. May RM. Thresholds and breakpoints in ecosystems with a multiplicity of stable states. Nature 1977;269:471–7.

Miller PS. Atrazine ecological exposure flowing water chemical monitoring study in corn/ sorghum watersheds—SSURGO-based characterization of shallow impervious/ restrictive soil layers. Greensboro, NC: Syngenta Crop Protection, Inc.; 2009. Miller PS, Harbourt CM, Andrus JM, Northcott W, Prenger J, Hendley P. Watershed characterization: multivariate approaches for identifying key factors contributing to potential vulnerabilities for agrochemical runoff in small headwater Midwest watersheds. American Chemical Society National Meeting, Philadelphia, PA, August 19–23, 2012. Washington, DC: American Chemical Society; 2012. Mohammad M, Itoh K, Suyama K. Comparative effects of different families of herbicides on recovery potentials in Lemna sp. J Pestic Sci 2008;33:171–4. Mohammad M, Itoh K, Suyama K. Effects of herbicides on Lemna gibba and recovery from damage after prolonged exposure. Arch Environ Contam Toxicol 2010;58:605–12. Moorhead DL, Kosinski RJ. Effect of atrazine on the productivity of artificial stream algal communities. Bull Environ Contam Toxicol 1986;37:330–6. Moss B. The composition and ecology of periphyton communities in freshwaters. II. Interrelationships between water chemistry, phytoplankton populations and periphyton populations in a shallow lake and associated experimental reservoirs (‘Lund tubes’). Br Phycol J 1981;16:59–76. Muñoz I, Real M, Guasch H, Navarro E, Sabater S. Effects of atrazine on periphyton under grazing pressure. Aquat Toxicol 2001;55:239–49. Murdock JN, Gido KB, Dodds WK, Bertrand KN, Whiles MR. Consumer return chronology alters recovery trajectory of stream ecosystem structure and function following drought. Ecology 2010;91:1048–62. Naiman RJ, Elliott SR, Helfield JM, O'Keefe TC. Biophysical interactions and the structure and dynamics of riverine ecosystems: the importance of biotic feedbacks. Hydrobiologia 2000;410:79–86. Neumann M, Leiss M, Schultz R. A qualitative sampling method for monitoring water quality in temporary channels or point sources and its application to pesticide contamination. Chemosphere 2003;51:509–13. Nyström B, Paulsson M, Almgren K, Blank H. Evaluation of the capacity for development of atrazine tolerance in periphyton from a Swedish freshwater site as determined by inhibition of photosynthesis and sulfolipid synthesis. Environ Toxicol 2000;19: 1324–31. Oksanen J, Blanchet FG, Kindt R, Legendre P, Minchin PR, O'Hara RB, et al. vegan: community ecology package. R package version 2.0-6. Retrieved from http://CRAN.R-project. org/package=vegan, 2013. Omernik JM. Ecoregions of the coterminous United States. Ann Assoc Am Geogr 1987;77: 118–25. Omernik JM. Ecoregions: a spatial framework for environmental management. In: Davis WS, Simon TP, editors. Biological assessment and criteria: tools for water resource planning and decision making. Boca Raton, FL: Lewis Publishers; 1995. Orloci L, Makkattu MM. The effect of species number and type of data on the resemblance structure of a phytosocialogical collection. J Ecol 1973;61:37–46. Osborne LL, Wiley MH. Factors influencing periphyton growth in agricultural streams of central Illinois. Nat Hist 1989;174:89–97. Palmer MW. Putting things in even better order: the advantages of canonical correspondence analysis. Ecology 1993;74:2215–30. Pan Y, Stevenson RJ, Hill BH, Kaufmann PR, Herlihy AR. Spatial patterns and ecological determinants of benthic algal assemblages in Mid-Atlantic streams, USA. J Phycol 1999; 35:460–8. Parsons M, Thoms MC. Hierarchical patterns of physical–biological associations in river ecosystems. Geomorphology 2007;89:127–46. Peterson CG, Grimm NB. Temporal variation in enrichment effects during periphyton succession in a nitrogen-limited desert stream ecosystem. J N Am Benthol Soc 1992;11: 20–36. Poff NL. Landscape filters and species traits: toward mechanistic understanding and prediction in stream ecology. J N Am Benthol Soc 1997;16:391–409. Potapova M, Charles DF. Benthic diatoms in USA rivers: distributions along spatial and environmental gradients. J Biogeogr 2002;29:167–87. Prenger JJ, Hendley P, Harbourt CM, Vamshi R, Zwilling LF, Miller PS. Atrazine ecological exposure flowing water chemical monitoring study in corn/sorghum watersheds— development of an NHDPlus analysis framework and associated tools. Greensboro, NC: Syngenta Crop Protection, Inc.; 2009. Prosser RS, Brain RA, Hosmer AJ, Solomon KR, Hanson ML. Assessing the PSII sensitivity and recovery of field-collected periphyton communities under laboratory conditions. Ecotoxicology 2013;22:1367–83. R Core Team. R: a language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing; 2013. Retrieved from http://www.R-project. org/. Rabiet M, Margoum C, Gouy V, Carluer N, Coquery M. Assessing pesticide concentrations and fluxes in the stream of a small vineyard catchment—effect of sampling frequency. Environ Pollut 2010;158:737–48. Rankin ET. The use of habitat assessments in water resource management programs. In: Davis W, Simon T, editors. Biological assessment and criteria: tools for water resource planning and decision making. Boca Raton, FL: Lewis Publishers; 1995. p. 181–208. Rao CR. A review of canonical coordinates and an alternative to correspondence analysis using Hellinger distance. Qüestiió (Quaderns d'Estadistica I Investigació Operativa), 19; 1995. p. 23–63. Relyea RA. The impact of insecticides and herbicides on the biodiversity and productivity of aquatic communities. Ecol Appl 2005;15:618–27. Reyea RA. A cocktail of contaminants: how mixtures of pesticides at low concentrations affect aquatic communities. Oecologia 2009;159:363–76. Richard JJ, Junk GA, Avery MJ, Nehring NL, Fritz JS, Svec HJ. Analysis of various Iowa waters for selected herbicides: atrazine, DDE, and dieldrin—1974. Pestic Monit J 1975;9: 117–23.

J.M. Andrus et al. / Science of the Total Environment 505 (2015) 65–89 Richardson TL, Gibson CE, Heaney SI. Temperature, growth and seasonal succesion of phytoplankton in Lake Baikal, Siberia. Freshw Biol 2000;44:431–40. Riseng CM, Wiley MJ, Stevenson RJ. Hydrologic distrubance and nutrient effects on benthic community structure in Midwestern US streams: a covariance structure analysis. J N Am Benthol Soc 2004;23:309–26. Roads and Transportation Association of Canada. Guide to bridge hydraulics. In: Neill CR, editor. Toronto, Canada: University of Toronto Press; 1973. Rumeau A, Coste M. Introduction into the systematic of freshwater diatoms for a useful generic diatomic index. Bull Fr Peche Piscic 1988;309:1–69. Schelske CL, Carrick HJ, Aldridge FJ. Can wind-induced resuspension of meroplankton affect phytoplankton dynamics? J N Am Benthol Soc 1995;14:616–30. Schultz R. Comparison of spray drift- and runoff-related input of azinphos-methyl and endosulfan from fruit orchards into the Lourens River, South Africa. Chemosphere 2001; 45:543–51. Shimabukuro RH, Frear DS, Swanson HR, Walsh WC. Glutathione conjugation: atrazine detoxification mechanism in corn. Plant Physiol 1970;46:103–7. Snyder EB, Robinson CT, Minsal GW, Rushforth SR. Regional patterns in periphyton accrual and diatom assemblage structure in a heterogeneous nutrient landscape. Can J Fish Aquat Sci 2002;59:564–77. Soininen J, Paavola R, Muotka T. Benthic diatom communities in boreal streams: community structure in relation to environmental and spatial gradients. Ecography 2004;27: 330–42. Solomon KR, Baker DB, Richards RP, Dixon KR, Klaine SJ, La Point TW, et al. Ecological risk assessment of atrazine in North American surface waters. Environ Toxicol Chem 1996;15:31–76. Sommer U. Seasonal succession of phytoplankton in Lake Constance. Bioscience 1985;35: 351–7. Spalding RF, Snow DD. Stream levels of agrichemicals during a spring discharge event. Chemosphere 1989;19:1129–40. Spearman C. The proof and measurement of association between two things. Am J Psychol 1904;15:72–101. State of Ohio Environmental Protection Agency (Ohio EPA). Methods for assessing habitat in flowing waters: using the qualitative habitat evaluation index (QHEI). Groveport, OH: State of Ohio Environmental Protection Agency Division of Surface Water; 2006. Stay FS, Katko A, Rohm CM, Fix MA, Larsen DP. The effects of atrazine on microcosms developed from four natural plankton communities. Arch Environ Contam Toxicol 1989;18:866–75. Stevenson RJ. Scale-dependent determinants and consequences of benthic algal heterogeneity. J N Am Benthol Soc 1997;16:248–62. Stevenson RJ, Bothwell ML, Lowe RL. Algal ecology. San Diego: Academic Press; 1996. Stevenson RJ, Rier ST, Riseng CM, Schultz RE, Wiley MJ. Comparing effects of nutrients on algal biomass in streams in two regions with different disturbance regimes and with applications for developing nutrient criteria. Hydrobiologia 2006;561:149–65. Strickland JD. Production of organic matter in the primary stages of the marine food chain. In: Riley JP, Skirrow G, editors. Chemical oceanography. New York: Academic Press; 1965. p. 478–610. Strobl C, Beoulesteix A-L, Zeileis A, Hothorn T. Bias in random forest variable importance measures: illustrations, sources, and a solution. BMC Bioinformatics 2007;8:25. Tang J, Hoagland KD, Siegfried BD. Uptake and bioconcentration of atrazine by selected freshwater algae. Environ Toxicol Chem 1998;17:1085–90.

89

Teodorović I, Knežević V, Tunić T, Čučak M, Lečić JN, Leovac A, et al. Myriophyllum aquaticum versus Lemna minor: sensitivity and recovery potential after exposure to atrazine. Environ Toxicol Chem 2012;31:417–26. Tlili A, Montuelle B, Berard A, Bouchez A. Impact of chronic and acute pesticide exposures on periphyton communities. Sci Total Environ 2011;409:2102–13. Toms JD, Lesperance ML. Piecewise regression: a tool for identifying ecological thresholds. Ecology 2003;84:2034–41. United States Environmental Protection Agency (USEPA). Interim reregistration eligibility decision for atrazine. Case No. 0062. Shackleford B. “Revision to January 31, 2003 IRED” mailing to registrants. Washington, DC: United States Environmental Protection Agency; 2007a. United States Environmental Protection Agency (USEPA). Preliminary interpretation of the ecological significance of atrazine stream-water concentrations using a statistically-designed monitoring program; 2007b [Prepared for the December 4–7, 2007 FIFRA Scientific Advisory Panel. EPA-HQ-OPP-2007-0934]. United States Environmental Protection Agency (USEPA). Atazine ecological exposure monitoring program data; 2009a [Washington, DC]. United States Environmental Protection Agency (USEPA). FIFRA Scientific Advisory Panel: a set of scientific issues being considered by the environmental protection agency regarding: the ecological significance of atrazine effects on primary producers in surface water streams in the corn and sorghum growing region of the United States (Part II). May 12–15, 2009 FIFRA Scientific Advisory Panel; 2009b [EPA-HQ-OPP2009-0104]. United States Environmental Protection Agency (USEPA). Atrazine ecological exposure monitoring program data. Washington, DC: USEPA; 2010. United States Environmental Protection Agency (USEPA). 2011 Atrazine ecological exposure monitoring program data—corn and sorghum areas. Washington, DC: USEPA; 2011. United States Environmental Protection Agency (USEPA). Transmittal of meeting minutes of the FIFRA SAP Meeting held June 12–14, 2012 on the scientific issues associated with “Problem Formulation for the Reassessment of Ecological Effects from the Use of Atrazine.”. Washington, DC: USEPA; 2012. United States Geological Survey (USGS). BioData—aquatic bioassessment data for the nation. Retrieved from http://aquatic.biodata.usgs.gov/, 2013. United States National Drought Mitigation Center. Drought monitor. Retrieved from http://www.droughtmonitor.unl.edu/DM_tables.htm?archive, 2013. Urrea G, Sabater S. Epilithic diatom assemblages and their relationship to environmental characteristics in an agricultural watershed (Guadiana River, SW Spain). Ecol Indic 2009;9:693–703. Vallotton N, Eggen RI, Escher BI, Krayenbühl J, Chèvre N. Effect of pulse herbicidal exposure on Scenedesmus vacuolatus: a comparison of two photosystem II inhibitors. Environ Toxicol Chem 2008;27:1399–407. Villeneuve A, Montuelle B, Bouchez A. Effects of flow regime and pesticides on periphytic communities: evolution and role of biodiversity. Aquat Toxicol 2011;102:123–33. Waldron AC. Pesticide movement from upland into Lake Erie. Washington, DC: USEPA; 1974 [EPA 660/2-74-032]. Wehr JD, Sheath RG. Freshwater algae of North America, ecology and classification. San Diego, CA: Academic Press; 2003. Wickham H. ggplot2: elegant graphics for data analysis. New York: Springer; 2009.

Temporal and spatial variation of potassium balance in agricultural land at national and regional levels in China.

Spatial and temporal variation of dung beetle assemblages in a fragmented landscape at eastern humid Chaco.

Spatial and temporal distribution of non-biting midge larvae assemblages in streams in a mountainous region in southern Brazil.

Spatial and temporal variation of malaria entomological parameters at the onset of a hydro-agricultural development in central Côte d'Ivoire.

Temporal Beta Diversity of Bird Assemblages in Agricultural Landscapes: Land Cover Change vs. Stochastic Processes.

Temporal and spatial variation of the human microbiota during pregnancy.

Amplicon pyrosequencing reveals spatial and temporal consistency in diazotroph assemblages of the Acropora millepora microbiome.

Spatial and temporal changes in bird assemblages in forest fragments in an eastern Amazonian savannah.

Assessing temporal and spatial variation in sensitivity of communities of periphyton sampled from agroecosystem to, and ability to recover from, atrazine exposure.

Musculoskeletal pain among Midwest farmers and associations with agricultural activities.

Spatial variation as a tool for inferring temporal variation and diagnosing types of mechanisms in ecosystems.

The variation of a Midwest Catholic University's drug use pattern from other Catholic University findings: why?

Diatoms on the carapace of common snapping turtles: Luticola spp. dominate despite spatial variation in assemblages.

Assessment of periphyton, aquatic macrophytes, benthic communities, and physical habitat in midwestern United States streams coinciding with varying historical concentrations of atrazine.

Spatial and Seasonal Variation of dissolved organic carbon (DOC) concentrations in Irish streams: importance of soil and topography characteristics.

Spatial and Temporal Variation of Cultivable Communities of Co-occurring Endophytes and Pathogens in Wheat.

Temporal and Spatial Variation of Soil Bacteria Richness, Composition, and Function in a Neotropical Rainforest.

Temporal variation in environmental conditions and the structure of fish assemblages around an offshore oil platform in the North Sea.

Spatial distribution, temporal variation and specificity of microhabitat of Tropisternus species (Coleoptera: Hydrophilidae) in permanent ponds.

Ten years of life in compost: temporal and spatial variation of North German Caenorhabditis elegans populations.

Temporal and spatial variation in the amount of cadmium in the phantom midge larvae (Chaoborus spp.).

Temporal and spatial variation of arbuscular mycorrhizas in salt marsh plants of the Tagus estuary (Portugal).

Analysis of spatial and temporal variation in the community structure of yeasts associated with decayingOpuntia cactus.

Impact of varying salt and fat levels on the physicochemical properties and sensory quality of white pudding.