POCIS Processing for Pesticides

To check for pesticides, POCIS media were extracted by using 25 mL of methanol per POCIS. Extracts were evaporated by using rotary evaporation and high-purity nitrogen blowdown prior to being sealed in amber glass ampoules at a volume of 1.0 mL. The ampoules containing the extracts were shipped to the USGS NWQL for analysis by liquid chromatograph tandem mass spectrometry. The analyses were performed using NWQL lab codes 8238 and 8239 (Furlong and others, 2001; table 1.13; https://www.usgs.gov/labs/national-water-quality-laboratory).

The calculation of time-weighted average water concentrations from POCIS data uses a first-order linear uptake model where the estimated water concentration is equal to the amount of the compound measured in the POCIS divided by the product of the sampling rate of the compound and the length of the field deployment in days (Alvarez, 2010). This calculation is as follows:

where:

Sampling rates for pesticides used in water concentration calculations are described in Alvarez and others (2004, 2007), Ahrens and others (2015), and Metcalfe and others (2016).

POCIS Processing for OWCs

To check for OWCs, POCIS media were extracted by using 25 mL of an 80:20 volume-per-volume mixture of dichloromethane:methyl-tert-butyl ether per POCIS. Extracts were evaporated by using rotary evaporation and high-purity nitrogen blowdown prior to being placed in chromatography vials. Five hundred ng of p-terphenyl-d₁₄ was added to each vial as the instrumental internal standard and the volume was adjusted to 1.0 mL. Analyses were performed using a gas chromatograph mass spectrometer (Alvarez, 2010). The calculation of time-weighted average water concentrations of OWCs uses the same calculation (eq. 1) discussed previously for the pesticides POCIS data. Sampling rates used for OWCs in water concentration calculations are described in Alvarez and others (2007), Harman and others (2008), Li and others (2010), Bartelt-Hunt and others (2011), Ahrens and others (2015), and Poulier and others (2015).

SPMD Processing for PAHs

To check for PAHs, SPMD processing consisted of dialysis to extract chemicals from the media, followed by using size exclusion chromatography to isolate the PAHs. One PRC-fortified SPMD from each site was designated for PAH analyses. Processing the samples included isolation of the PAHs from potential interferences, cleanup of the extract using a reactive tri-adsorbent gravity-flow chromatography column, and analysis using a gas chromatograph mass spectrometer in full-scan mode (Alvarez and others, 2008).

The calculation of time-weighted average water concentrations of PAHs from SPMD data uses a series of polynomial regression models that take into account site-specific environmental factors affecting uptake as determined by the loss of the PRCs (Alvarez, 2010). These calculations of PAHs from SPMD data are described by Alvarez (2010). A detailed explanation of the theoretical aspects of uptake and derivation of the models has been described by Huckins and others (2006).

Quality Control for Passive Sampling

Quality control (QC) samples for the POCIS and SPMD passive samplers consisted of fabrication blanks and field blanks. Blank samples, also called blanks, are collected and analyzed alongside the environmental samples to help detect the presence of airborne contamination of the sampling media during construction in the laboratory and handling in the field during sampling trips. Fabrication blanks were processed concurrently with the field-deployed samplers. Fabrication blanks were exposed to the air at the CERC laboratory during the fabrication of the samplers and are used to account for the interferences or contamination that occurred from the sampler components, storage, processing, and analysis (Alvarez, 2010). For each of the six deployments, one POCIS and one SPMD field blank container was opened to the air at one of the three sites (chosen at random) just prior to the passive sampler being deployed in the water and closed after the sampler was submerged. The field blank passive sampler containers were also opened before retrieving the instream passive sampler and then closed when the retrieved passive sampler was resealed in its metal can. Between deployment and retrieval of the instream passive samplers, the sealed field blank sample containers were stored in a freezer at less than 0 °C. Blanks were processed and analyzed in the same fashion as the environmental samples. Instrument verification checks and reference standards were also employed.

Laboratory reporting levels for each pesticide measured in the POCIS were established by the NWQL based on instrumental limits accounting for any sample dilutions that may have been done. The CERC established reporting levels for the OWCs in POCIS as a function of method detection limits (MDLs) and method quantitation limits (MQLs) determined for each sampling year. The MDL is determined as the mean of chemical responses in the blanks plus three times the standard deviation of the blanks. The MQL is determined as the mean of the chemical responses in the blanks plus 10 times the standard deviation of the blanks or at the level of the lowest instrumental calibration level (whichever was highest). In cases where no chemical was measured in the blanks, the MQL was set at the lowest instrumental calibration level and the MDL was set at 20 percent of the MQL (Keith, 1991). The CERC set the MDL for analysis of SPMD samples for each analyte at an assumed value equal to the low sample reject for the instrumental method (operationally defined as 20 percent of the concentration of the lowest standard concentration used for the calibration curve), and the MQL was set at the lowest standard concentration of the calibration curve.

The concentration results from field and fabrication blanks were used to recensor environmental results in some cases. Concentrations of the environmental samples were recensored at the concentrations found in the blanks if they exceeded the MDL. The estimated water concentration of the fabrication blank for a given year was used to recensor results for all sites in that same year unless the site-specific field blank had a higher estimated water concentration and then it was used for recensoring results for the site at which the field blank was collected.

Assay Development

Species-specific DNA sequences were obtained from the open-source National Center for Biotechnology Information (NCBI) GenBank database (http://www.ncbi.nlm.nih.gov) and the open-source Barcode of Life Database (https://www.boldsystems.org/). The assays developed for this study target the cytochrome oxidase subunit I (COI) gene for each species. The many individual target sequences found for each species were aligned by using the open-source software MEGA-X (https://www.megasoftware.net), and a consensus sequence was created. Each consensus sequence was entered into the open-source Eurofins qPCR Primer & Probe Design Tool (https://eurofinsgenomics.eu/en/ecom/tools/qpcr-assay-design/), and a primer and probe set was chosen for each species from the output of possibilities. This chosen assay was then imported into the NCBI Primer-BLAST software (https://blast.ncbi.nlm.nih.gov/Blast.cgi), where in silico testing was conducted to check for cross-reactivity against available sequences for all biological organisms in the GenBank database (table 3). Specifically for mussels, 38 of 39 colocated mussel species in the Big Darby Creek Basin (from the taxon list taken from the four most recent surveys; EnviroScience, 2015) have at least one COI sequence represented in the GenBank database. After passing the in silico testing, all newly developed assays were optimized and validated at the OWML.

eDNA Field Sampling and Molecular Methods

For all eDNA samples, surface-water grab samples were collected in 1 liter, bleach-sterilized polypropylene bottles following the National Field Manual protocols (Myers and others, 2007). Each sample was brought to the OWML on ice and was prefiltered through a 47-millimeter diameter, 5-micrometer cellulose nitrate filter (Whatman; Florham Park, N.J.). The full 1-liter filtrate was then filtered through a 47-millimeter diameter, 0.45-micrometer cellulose nitrate filter (Whatman; Florham Park, N.J.). Afterward, filters were folded aseptically and put into 2-mL screw-cap vials with approximately 0.3 grams of acid-washed glass beads (bead beating is part of the extraction process; Sigma-Aldrich Corp.; St. Louis, Mo.). Each day that samples were processed, negative controls (filter blanks) were processed by using sterile buffered water. After being filtered, samples were kept at −70 °C pending analysis.

DNA was extracted from the samples by using DNA–EZ extraction kits from GeneRite (Monmouth Junction, N.J.). Extraction was performed following the manufacturer's directions, apart from not using a prefilter. Negative extraction controls (extraction blanks), which are 2-mL vials containing only acid-washed glass beads, were processed in parallel with each batch of extractions. Before qPCR analysis, DNA extracts were stored at 4 °C for no more than 5 days after extraction.

Sample extracts were analyzed for kidneyshell, northern riffleshell, C. obscura (a caddisfly), and M. pulchellum (a mayfly) DNA by using qPCR. The qPCR analyses were performed by using Applied Biosystems StepOne Plus or Applied Biosystems QuantStudio 3 Real-Time PCR System and their corresponding software packages. Thresholds were set manually for each assay. Each sample was analyzed in duplicate and run for 45 cycles. Duplicates of no-template controls (qPCR blanks), which contained molecular-grade sterile water rather than DNA extract, were included on each qPCR plate. For all samples, matrix inhibition was evaluated by using matrix spikes (a known-quantity target DNA extract spiked into the sample DNA extract). Samples were considered inhibited if the average cycle threshold value (Ct) was delayed more than two cycles from the spiked no-template control. Diluted sample DNA extracts were used for qPCR if the sample showed matrix inhibition.

Quantifying eDNA by qPCR

To quantify target sequences in unknown samples, seven-point standard curves were produced in parallel with each qPCR run using plasmids containing the sequences for each of the targeted eDNA COI genes. The lowest standard was 10¹ copies of the target gene, and the highest standard was 10⁷ copies of the target gene. The slope of the log-linear component of the standard curve was used to compute amplification efficiency (Bustin and others, 2009) using the equation:

The amplification efficiency should be 90 to 110 percent; an efficiency of 100 percent means an exact doubling of the target DNA sequence at each cycle and, thus, a more accurate quantification of the target in the original sample. Guidelines for interpreting standard curve data are available in Bio-Rad Laboratories, Inc. Bulletin 5859 (Taylor and others, 2015). The qPCR result was converted to eDNA marker concentrations in copies per liter by interpolating from the standard curves. The coefficient of determination for the regression is used to assess the fit of the standard curve to the plotted data points. The closer the coefficient of determination is to 1, the better the fit. Standard curve characteristics are shown in table 4.

The limit of detection (LoD) is the lowest concentration of a target that can be identified with 95 percent confidence. The LoD is determined by performing a series of dilutions of the target with a minimum of 10 replicates per dilution. The dilutions used are based on the sensitivity of the assay. A standard curve, run alongside the dilution series, is used to validate the concentration of the dilutions, thus determining the limit of detection.

The lowest quantity of a target that can be accurately determined is known as the limit of quantification (LoQ). The LoQ is determined by performing a series of dilutions of the target with a minimum of 10 replicates per dilution. Which dilutions are used is based on the sensitivity of the assay. For the LoQ calculation, the dilution with the lowest concentration with a coefficient of variation below 35 percent and a Ct standard deviation at or below 0.5 is selected. The LoQ is calculated as the average Ct value (Ct_AVE) for the selected dilution subtracted by two times the associated standard deviation (σ_Ct), which is then used to compute a concentration (copies per reaction) using the standard curve run with the dilution series.

QC Results for Passive Samplers

Although blank samples were collected for QC, not all were able to be successfully analyzed against the environmental samples. Two of the six POCIS field blanks were reported damaged during shipment (field blanks for deployments 2 and 3 in 2021). There were 20 pesticide compounds for which a concentration could not be determined in at least one blank due to matrix interference or a ruined sample (appendix table 1.1). Fourteen pesticide compound concentrations (totaling 56 observations, all in 2020) could not be determined in blank samples due to matrix interference and 19 pesticide compound concentrations (totaling 38 observations, all in 2021) could not be determined in blank samples because the samples were ruined during analyses (appendix table 1.1). These QC issues resulted in having 13 pesticide compounds with no field blank quality-control data; however, only 3 of those 13 pesticide compounds (alachlor sulfinylacetic acid [SAA], desulfinylfipronil amide, and fipronil sulfonate) were detected in environmental samples.

There was a total of four pesticide compounds detected in at least one POCIS field blank (all in 2020), and two of those compounds (atrazine and metolachlor) were also detected in the fabrication blank (appendix table 1.2). Those four pesticide compounds were either not detected in the environmental samples or were detected at concentrations lower than concentrations found in the QC samples (appendix tables 1.2 and 1.4). No pesticide compounds were detected in the field or fabrication blanks from 2021.

For OWCs in the POCIS, 35 compounds (92 percent of the 38 OWCs analyzed) were detected in at least one field or fabrication blank, and 11 compounds (29 percent of OWCs analyzed) were detected in all field and fabrication blanks (appendix table 1.5). The majority of the OWC detections in blanks occurred in 2020, when 35 compounds were detected in both field and fabrication blanks. In 2021, only 15 OWCs (39 percent of OWCs analyzed) were detected in at least one field or fabrication blank (appendix table 1.5). The estimated water concentrations of 21 OWCs detected in blanks exceeded at least one corresponding initially uncensored concentration measured in an environmental sample (appendix tables 1.6, 1.7, and 1.8), and so the concentrations of the environmental samples were recensored as described in the “Quality Control for Passive Sampling” section.

The plasticizer diethyl phthalate (DEP) had unusually high background contamination in all fabrication and field blanks across both sampling years. The CERC laboratory could not determine the cause of this contamination, which was a factor of 10–20 times higher than the historical background contamination detected in other studies (David Alvarez, USGS, CERC laboratory, written commun., November 2023). Due to the high blank concentrations and resulting MDL/MQL levels, the CERC laboratory recommended using diethyl phthalate (DEP) results cautiously; consequently, estimated water concentrations for diethyl phthalate (DEP) have been omitted from our summary tables and plots.

Two of the six SPMD field blanks sent for analysis were reported lost or damaged during shipment (field blanks for deployment 3 in 2020 and deployment 1 in 2021). Of the 33 PAHs analyzed in the SPMDs, 28 (85 percent) were detected in one or more field blanks. Seven compounds (1-methylfluorene, 2-methylnaphthalene, 4-methylbiphenyl, fluoranthene, naphthalene, phenanthrene, and pyrene) were detected in all field blanks (appendix table 1.9). Of the 28 PAHs detected in field blanks, 13 were also detected in one or more fabrication blanks. Estimated water concentrations of 10 compounds detected in blanks exceeded initially uncensored concentrations measured in environmental samples, and so their results were re-censored as described in the “Quality Control for Passive Sampling” section (appendix tables 1.10, 1.11, and 1.12).

POCIS Results for Pesticides

POCIS extracts were analyzed for a total of 204 pesticide and pesticide degradate compounds (appendix table 1.13); 70 pesticide compounds were detected in at least one sample, and 30 of those were detected in all samples (appendix table 1.3 and 1.14). The most detections (305) occurred at the Big Darby Creek above Georgesville, Ohio, site, followed by Little Darby at West Jefferson, Ohio (287), then Little Darby Creek above Georgesville, Ohio (283). Estimated time-weighted average water concentrations for pesticide compounds detected in all deployments are shown in figure 2 (appendix table 1.4). Most of the detections were for herbicides and herbicide degradates, followed by fungicides and fungicide degradates, then by insecticides and insecticide degradates (fig. 3). Some of the highest concentrations of pesticides detected in this study were for herbicides and herbicide degradates. The three highest concentrations of herbicides detected were metolachlor, acetochlor, and atrazine (or their degradates; figs. 4–6). All three compounds (and most of their degradates) were detected in all samples. The maximum time-weighted average concentration for metolachlor was 1,304 ng/L; the degradate, metolachlor oleic acid (OA), had a maximum of 710 ng/L. The maximum time-weighted average concentration for acetochlor was 118 ng/L; the degradate, acetochlor ethanesulfonic acid (SA), had a maximum of 670 ng/L. The maximum time-weighted average concentration for atrazine was 360 ng/L (appendix table 1.3).

None of the pesticides detected in the POCIS with aquatic life benchmarks established by the U.S. Environmental Protection Agency (EPA) had concentrations that exceeded the chronic exposure benchmarks (EPA, 2023) for freshwater invertebrates. Three herbicides (atrazine, ametryn, and metribuzin) and one herbicide degradate (deethylatrazine [DEA or desethylatrazine]) were detected in at least 88 percent of samples (appendix tables 1.3 and 1.4). These same compounds have been found to strongly affect the physiology of the Pyganodon grandis (giant floater) mussels in Maumee River, Ohio (Roznere and others, 2023). The maximum concentrations detected in the Big Darby Creek Basin for selected compounds are within the range of the concentrations detected in the Maumee River study. Atrazine (detected in all samples in this study) was ranked 11 out of 44 compounds by Roznere and others (2023) as strongly correlated with changes in the mussel’s hemolymph metabolites, likely negatively impacting mussel health.

The fungicides with the highest median concentrations of detection were azoxystrobin and propiconazole (figs. 7–9), which were detected in all samples. Azoxystrobin and propiconazole are systemic fungicides frequently used to treat fungal lawn diseases, as well as some other fungal plant diseases. Bringolf and others (2007) found that propiconazole was acutely toxic to glochidia and juveniles of the Lampsilis siliquoidea (fatmucket) mussel species, with juveniles being more sensitive than glochidia.

Imidacloprid, a neonicotinoid insecticidal compound, was detected in all samples with maximum concentrations of 24.87, 12.43, and 18.33 ng/L at the three sites (figs. 10–12; appendix table 1.3). Imidacloprid has been shown to bioaccumulate in mussels and have a variety of effects on their behavior, physiology, and biochemistry; however, it generally had no negative effect on mortality (Ewere and others, 2021). Fipronil, an insecticide used in agriculture and veterinary medicine, was also detected in all samples. In a study conducted by Arslan and Günal (2023), fipronil was found to cause histopathological alterations to the gills and digestive glands of a freshwater mussel (the Unio delicatus) found in Turkey and Syria, and the authors concluded that sublethal concentrations of fipronil were toxic to freshwater mussels. Bringolf and others (2007) assessed the toxicities of various current-use pesticides to the glochidia and juveniles of several other freshwater mussel species and concluded that technical-grade fipronil was not acutely toxic to the mussels when exposed in 24-, 48-, and 96-hour tests. Fipronil is readily absorbed by sediment; however, little is known about sediment as a vector for exposure to juvenile mussels, which live one or more years buried in the substrate, nor is much known about the effects of chronic exposure to fipronil on mussels (chronic toxicity tests are 21 days; Bringolf and others, 2007).

POCIS Results for OWCs

POCIS extracts were analyzed for a total of 38 OWCs (appendix table 1.15), and 24 compounds were detected in at least one sample (appendix tables 1.6 and 1.17). The most OWC detections (71) occurred at Little Darby Creek above Georgesville, Ohio, followed by Big Darby Creek above Georgesville, Ohio (64), and Little Darby Creek at West Jefferson, Ohio (63). Estimated time-weighted average water concentrations for OWCs detected in all deployments are shown in figure 13, and boxplots showing the distribution of estimated time-weighted average water concentrations for OWCs detected at each site are shown in figures 14–16 (appendix tables 1.7 and 1.8).

One of the OWCs, DEET, was detected in all samples. The maximum concentrations of DEET detected in the Big Darby Creek Basin are within the range of the concentrations detected by Roznere and others (2023) in the Maumee River study. Roznere and others (2023) ranked DEET as 8 out of 44 compounds found to be strongly correlated with changes in metabolites in mussels (table 6). N-Butylbenzenesulfonate was detected in 94 percent of samples, and four other compounds were detected in over 80 percent of the samples. Galaxolide and tonalide were detected in 89 and 61 percent of samples, respectively. Galaxolide and tonalide are fragrance compounds known to be endocrine disruptors and for bioaccumulating (Lefebvre and others, 2017). Other detections in the majority of samples included three flame retardants (Tris(1-chloro-2-propyl)phosphate, Tris(2-chloroethyl)phosphate, and Tris(1,3-dichloro-2-propyl)phosphate).

SPMD Results for PAHs

SPMD media extracts were analyzed for 33 PAHs (appendix table 1.16); 29 compounds were detected in at least one sample, and 12 of those 29 were detected in all samples (appendix table 1.10). Of those 12 compounds detected in all samples, eight are on the EPA’s priority pollutant list (Hussar and others, 2012; EPA, 2024). The most PAH detections (133) occurred at Little Darby Creek at West Jefferson, Ohio, followed by Little Darby Creek above Georgesville, Ohio (131), and Big Darby Creek above Georgesville, Ohio (112). Estimated time-weighted average water concentrations for PAHs detected in all deployments are shown in figure 17. Of those PAHs detected in all samples, phenanthrene had the highest time-weighted average concentration (3,300 picograms per liter [pg/L]), followed by fluoranthene (2,100 pg/L). Nine other compounds had concentrations ranging between 50 and 800 pg/L (appendix tables 1.10, 1.11, 1.12; figs. 18–20). Fluoranthene is of environmental concern because of its potential toxicity to humans and other organisms and its prevalence and persistence in the environment. Some PAHs, like fluoranthene, can become several orders of magnitude more toxic in the presence of ultraviolet wavelengths of natural sunlight (Weinstein, 2002). A study involving the larval stages of a freshwater mussel (the Utterbackia imbecillis) found in some Ohio streams indicated they are very sensitive to ultraviolet-induced PAH toxicity (Weinstein, 2002).

Trends in Annual Flow Statistics

Trends in annual minimum daily flows were analyzed on a climatic year basis (April through March), and trends in all other annual statistics (mean, median, and maximum daily flows) were analyzed on a water year basis (October through September). The median trend slopes (Sen’s slopes) for these statistics for Big Darby Creek were positive (ranging from 0.4 to 0.8 percent change per year), indicating the statistics increased over time. Mann-Kendall test results indicated the trend slopes were statistically significant (indicating rejection of the zero-slope null hypothesis) for all the statistics except the annual maximum daily flow (figs. 21–24). For the Hellbranch Run and Little Darby Creek gages, trend slopes were positive for annual minimum, mean, and median daily flows and negative for annual maximum daily flows; however, none of the trend slopes at these sites were statistically significant.

Trends in Daily Nonexceedance Probabilities

Quantile-Kendall plots provide information on how flow trended over time that is different than the information provided by Mann-Kendall tests on annual flow statistics. To develop the Quantile-Kendall plot, flows associated with a series of daily nonexceedance probabilities are determined for each year of record. Trend slopes are then computed for the period of record for each nonexceedance probability. The trend slopes were evaluated for statistical significance with Mann-Kendall tests. Trend slopes were positive for all nonexceedance probabilities for the Big Darby Creek gage, indicating that the flow values associated with the daily nonexceedance probabilities are increasing over time. The trend slopes were largest, with average slopes greater than about 0.8 percent change per year, at lower nonexceedance probabilities (lower flows) and became gradually smaller but still positive at higher nonexceedance probabilities (fig. 25). The trend slopes were statistically significant except at nonexceedance probabilities greater than about 0.96. The trend slopes for the Hellbranch Run and Little Darby Creek gage records (figs. 26 and 27) for all nonexceedance probabilities were mostly positive or zero and predominantly not statistically significant. At both the Hellbranch Run and Little Darby Creek gages, some negative trend slopes were observed for higher flows with nonexceedance probabilities greater than 0.90.

Trends in Indicators of Hydrologic Alteration Statistics

The lowess smooth line fit through the annual minimum 7-day average low flows (hereafter referred to as “annual 7-day low flows”) for Big Darby Creek indicates that the annual 7-day low flows were generally higher post-1970 than pre-1970 (fig. 28). This change is even more apparent when viewing boxplots of annual 7-day low flows by decade (fig. 29). Notably, McCabe and Wolock (2002) identified increases in annual minimum and median daily flow in several eastern U.S. streams and stated that the changes in flow appeared to occur as a step change around 1970, rather than as a gradual trend. Not only were the annual 7-day low flows higher post-1970 at the Big Darby Creek gage, but the number of annual low-flow pulses (periods when the flow fell below the 25th percentile of daily flows) and extreme low-flow periods (periods when flows fell below the 10th percentile of daily flows) appear to have decreased, as indicated by the typically lower median values post-1970 (figs. 30 and 31). These changes to low pulses and extreme low-flow periods can affect the availability of floodplain habitats for aquatic organisms and alter the nutrient and organic matter exchanges between the river and its floodplain (The Nature Conservancy, 2009).

While the numbers of annual low-flow pulses and extreme low-flow periods have decreased in recent decades at the Big Darby Creek gage, the annual number of high-flow pulses (periods when the flow equals or exceeds the 75th percentile of daily flows) appear to have increased (fig. 32). Although the number of high-flow pulses has increased, there was no indication that the annual maximum 1-day flow increased over the period of record (fig. 33).

The only strong indication of temporal trend in annual flow statistics for the relatively short records of the Little Darby Creek and Hellbranch Run gages was the number of reversals between rising and falling periods. A reversal occurs when there is a change of sign in the rate of change of daily flows (when daily flows transition from falling to rising or from rising to falling). At both gages, the annual number of reversals increased between water years 1993 and 2021 (figs. 34 and 35). The number of reversals also increased over the same period at the Big Darby Creek gage (fig. 36). However, the trend over the larger period of record at the Big Darby Creek gage was more complex, suggesting a downward trend in the number of reversals from water year 1922 to about water year 1980 followed by an upward trend between water years 1980 and 2021 (fig. 36). The rate and frequency of reversals can affect drought stress on plants, entrapment of organisms on islands or floodplains, and cause desiccation stress on low-mobility stream-edge organisms (The Nature Conservancy, 2009). Freshwater mussels have been shown to be sensitive to flow alterations (Vaughn and Taylor, 1999; Galbraith and Vaughn, 2011; Allen and others, 2013), including changes to the magnitudes, duration, and timing of high and low flows. The annual number of reversals was negatively correlated with the benthic index of biological integrity scores in one study of urbanizing streams in Washington State (DeGasperi and others, 2009).