Skip Navigation Links

USGS - science for a changing world

USGS Data Series 423

Selected Physical, Chemical, and Biological Data Used to Study Urbanizing Streams in Nine Metropolitan Areas of the United States, 1999–2004

National Water-Quality Assessment Program

By Elise M.P. Giddings, Amanda H. Bell, Karen M. Beaulieu, Thomas F. Cuffney, James F. Coles, Larry R. Brown, Faith A. Fitzpatrick, James Falcone, Lori A. Sprague, Wade L. Bryant, Marie C. Peppler, Cory Stephens, and Gerard McMahon

Instream Characteristics

Once the sampling locations were identified, instream characteristics of the streams were determined at or near the sampling locations. These characteristics include stream level (or stage) and temperature conditions, stream chemistry, and instream aquatic habitat. All of these characteristics were measured within a specified time frame that coincided with the biological sampling of the streams. The following sections discuss the collection and characterization of these measurements, and the data are summarized in table 3.


Continuous stream-stage data were collected by using a submersible pressure transducer with an internal data logger and temperature sensor attached (Greenspan Technology Pty Ltd, 2002). Standard USGS streamgaging techniques for collecting streamflow data were not used because of the short data-collection period (equal to or less than (≤) 1 year) and the cost of developing a stage-to-discharge rating at each site. The pressure transducer has a range of 0 to 30 m and an accuracy of plus or minus (±) 0.036 m. This level of accuracy does not meet the USGS standard for accuracy of stage data, which is 0.003 m (Sauer, 2002); therefore, the unit values are not published here. A submersible pressure transducer records the pressure of water above the transducer membrane. This value is then converted to water depth from field measurements. The transducer model used in this study recorded changes in stream stage as a result of both water-level changes and atmospheric-pressure changes. As a result, the data were corrected for fluctuations in atmospheric pressure by using barometric pressure data from nearby airports and adjusting for differences in altitude.

Stream-stage values collected at 15-minute intervals were converted subsequently to cross-sectional area in seven of the nine studies (excluded in Birmingham and Boston). A relation between stream stage and cross-sectional area was established by surveying the cross section at the location of the transducer installation.

Stream stage and cross-sectional area values were summarized by using hydrologic-condition metrics (McMahon and others, 2003; table 3-A, 278 KB). An hourly dataset was created by using the data point collected at the beginning of each hour; this dataset was used in further analysis to reduce necessary computing resources. Metrics were calculated to summarize overall hydrologic variability; the rate of change-of-water levels (or flashiness); and the magnitude, frequency, and duration of high and low stage. Calculations generally were based on a 1-year data record that included the biological sampling event. Missing values were not estimated, and the amount of missing data is given as a percentage for each site. The Birmingham area was undergoing drought conditions during much of the study duration.

Stream Temperature

Data loggers were used to record temperature at measurement intervals that ranged from 1 to 60 minutes depending on the equipment, site, and time of year. Equipment malfunctions, floods, droughts, and winter ice led to loss of data and the generation of incomplete annual stage (238 of 265 sites) and temperature (252 of 265 sites) records. Temperature data were processed to produce temperature records and data summaries that could be used to assess the effects of urbanization on water temperature and compare temperature regimes among and across the urban study watersheds and study areas (Cuffney and Brightbill, 2008).

Summary temperature statistics for annual and summer time periods were developed using temperature data collected at each site (table 3-B, 150 KB). Summary statistics consisted of average, minimum, maximum, range, and standard deviation. The rates of change in water temperature (degrees Celsius per hour [°C/h]) also were summarized (average, minimum, maximum, range, standard deviation, and number of observations) for 1-hour intervals. Annual degree-days also were calculated after estimating missing daily average temperature values.

Fluctuations in temperature over the annual and summer periods of record (PR) also were assessed based on the number of hourly rate-of-change values that fall within multiples of the standard deviation (SD) of rates observed over the PR. Rate values were converted to absolute values before calculating the SD over the PR and counting the number of rates that fall within six multiples of the SD for the PR (0 ≤ SD < 1; 1 ≤ SD < 2; 2 ≤ SD < 3; 3 ≤ SD < 4; 4 ≤ SD < 5; ≥ 5 SD).

Stream Chemistry

Stream chemistry conditions were measured by collecting water samples and deploying SPMDs (Bryant and others, 2007; Sprague and others, 2007). Water samples were collected at equal-width increments across the stream channel and processed on site in accordance with standard USGS protocols (Wilde and others, 1999, 2002). Water samples were analyzed at the USGS National Water Quality Laboratory (NWQL) in Denver, Colorado. SPMDs were placed at each site in six of the study areas for a period of approximately 6 weeks during April and May 2003.

Nutrients and Pesticide Compounds

The NWQL has established two detection-level values for nutrient and pesticide analyses—a lower method detection level, which is set to avoid a false negative reading (not detecting a compound when it actually is present), and a higher reporting level to avoid a false positive reading (detecting a compound when it actually is not present). If a compound is identified at a concentration between these two levels, the result is noted with an “e” to indicate that the concentration was estimated (Childress and others, 1999). The estimated values are greater than zero but are known with less confidence than values above the reporting level. Values also may be noted as estimated when the detected concentration is outside of the calibration range for the instrument, when the average recovery for the analyte in quality-assurance samples is less than 60 percent, or when the analyte is regularly detected in laboratory blank samples. Estimated concentrations must be interpreted with caution. Values reported with a less than (<) symbol were not detected at the lower method detection level and are presented as less than the (higher) reporting limit (Childress and others, 1999).

Two chemistry datasets were created from samples collected during the synoptic water sampling, one for nutrients and other nonpesticide analytes (tables 3-C, 3-D, and 3-E) and the other for pesticide compounds (tables 3-F, 3-G, and 3-H). For each dataset, the data were summarized into three periods—a high-flow period (one sample per site; table 3-C, 240 KB, for nutrients and table 3-F, 455 KB, for pesticides), a low-flow period (one sample per site; table 3-D, 232 KB, for nutrients and table 3-G, 436 KB for pesticides), and a multiple sample period (more than two samples collected per site; table 3-E, 2.4 MB for nutrients and table 3-H, 622 KB, for pesticides). The following censored-data rules were applied in preparing the chemical concentration values reported in the tables. For nonpesticide constituents, half the less-than value was reported. For pesticide constituents, zero was reported for less-than values. For all estimated constituents, the values were reported. The adjusted data for each study site then were recombined into the appropriate flow and constituent spreadsheet (Sprague and others, 2007).

Several summary variables were created for pesticide compounds. The number of detections and total concentrations of different pesticide groups, such as insecticides and herbicides, were compiled. In addition, a pesticide toxicity index (PTI) was determined. The PTI combines information on exposure of aquatic biota to pesticides with toxicity estimates for multiple pesticides in each sample and produces a relative index value for a sample or stream (Munn and Gilliom, 2001). The PTI value was computed for each sample of streamwater by summing the toxicity quotients for all pesticides detected in the sample. The toxicity quotient is the measured concentration of a pesticide in a stream sample divided by the median toxicity concentration from bioassays, such as 50-percent lethal concentration (LC50) or 50-percent effect concentration (EC50). Separate PTI values were computed for fish, cladocerans, and benthic invertebrates by using median toxicity concentrations from Munn and Gilliom (2001) supplemented with updated toxicity data (L.H. Nowell and P.W. Moran, U.S. Geological Survey, written commun., May 13, 2005).

The PTI approach can be a useful tool for examining pesticide mixtures in streams; however, it has several important limitations that must be considered in applications to water-quality data (L.H. Nowell, U.S. Geological Survey, written commun., December 2005). First, the PTI approach assumes that toxicity is additive and combines toxicity-weighted concentrations of pesticides from multiple chemical classes without regard to mode of action. This approach does not account for synergistic or antagonistic effects. Moreover, toxicity values are based on bioassays of acute exposure and do not incorporate the effects of chronic exposure. Environmental factors that can affect bioavailability and toxicity, such as dissolved organic carbon and temperature, are not accounted for. Second, the PTI approach is limited to pesticides measured in the water column—hydrophobic pesticides may be underrepresented in terms of potential toxicity, especially to benthic organisms. Because toxicity values from different sources vary, there is considerable uncertainty in the relative toxicity of pesticides when only a few bioassays are available. The number of bioassays varied among pesticides from 1 to 165 for a given taxonomic group. Finally, not all potentially important local species were included in the bioassays. The PTI does not indicate whether water in a stream sample is toxic; however, the PTI can be used to rank or compare the relative potential toxicity of different samples or different streams.

Hydrophobic Organic Compounds

Semipermeable membrane devices (SPMD) are passive samplers that concentrate trace levels of hydrophobic organic compounds in the water column. The samplers are designed to mimic the bioaccumulation of organic compounds in the tissues of aquatic organisms. To examine concentrations of hydrophobic organic compounds over time, two 6-inch long SPMDs were placed at each site for approximately 6 weeks during the period 4-6 weeks prior to sampling invertebrates, algae, and fish in each stream (Bryant and others, 2007; table 3-I, 636 KB). SPMD data were not collected at the Boston and Salt Lake City study watersheds.

At the end of the 6-week deployment period, compound residues concentrated in the SPMDs were recovered, and three assays were run on the dialysates from each site—an ultraviolet (UV) fluorescence scan (Johnson and others, 2004), a Microtox® bioassay (Johnson, 1998), and a P450RGS test (Bryant and others, 2007). The UV fluorescence scan provided a semiquantitative screen for polycyclic aromatic hydrocarbons (PAH). The Microtox® bioassay measured the light production of photo-luminescent bacteria when exposed to the SPMD residues; the biochemical pathway for light production is lowered by a wide range of compounds sequestered by the SPMDs. The P450RGS test provides a rapid screen for aryl hydrocarbon receptor (AhR) compounds that include polychlorinated biphenyls (PCB), PAHs, dioxins, and furans (Ang and others, 2000). A portion of each SPMD dialysate also was sent to the NWQL for identification and quantification of the target compounds (Tom Leiker, U.S. Geological Survey, written commun., 2005). Results of the bioassays and chemical analyses also were normalized for time of exposure, because the time of exposure directly affects the concentrations in the SPMD. Bioassay values were divided by time of exposure and multiplied by 30 days. Therefore, the values reported have the appropriate units described in the respective analytical methods per 30 days of exposure. This allowed values for all endpoints to be comparable among all sites.


Stream habitat characteristics were measured at all EUSE sites during late-summer low flows by using standard NAWQA protocols for wadeable streams (Fitzpatrick and others, 1998). Habitat data consisted of GIS-derived characteristics for stream segments upstream from the sampling location and field-based measurements collected from a sampling reach immediately upstream or downstream from the water-chemistry sampling location. Reach lengths sampled for habitat corresponded to the same reach lengths that were sampled for fish, invertebrates, and algae. Reach lengths optimally were at least 20 times the average wetted stream width and included at least one full meander bend (Leopold and others, 1964). A reach length of at least 150 m and up to 300 m was preferred for wadeable streams. For some studies, reaches were less than 150 m because of the abundance of channel modifications and road crossings in urban areas.

Segment Characteristics

A segment is a length of stream with relatively homogeneous physical, chemical, and biological characteristics (Fitzpatrick and others, 1998). The stream segment length for collecting segment data upstream from the biological sampling location was equal to approximately the log-10 distance of the watershed area (for example, the segment length for a 10-km2 watershed is 1,000 m; for a 100-km2 watershed, the segment length is 2,000 m). Segment data include riparian land-cover characteristics, physical characteristics (gradient and sinuosity), and the number of road crossings (table 3-J, 92 KB).

Reach Characteristics

Qualitative and quantitative data on channel geometry and hydraulics, streambed substrate, habitat volume and flow, habitat complexity and cover, and bank and riparian conditions were collected at 11 equally spaced transects along the sampling reach (Fitzpatrick and others, 1998). Reach habitat measurements were summarized for each reach to include minimum, maximum, and mean values, and coefficients of variation for wetted channel width, depth, velocity, and canopy cover. Minimum, maximum, and mean values were calculated for bankfull width, bankfull depth, width-to-depth ratios, bank vegetative cover and shading, and embeddedness. Measurements made at transect points were summarized for bank erosion, substrate size classes, silt covering of substrate, and habitat cover types. To facilitate comparisons of habitat characteristics among sites, average bankfull width, depth, and channel area of geomorphic channel units (riffle, run, and pool) within a reach were standardized by dividing by watershed drainage area. Morphologic indicators were used to estimate bankfull stage and included variations in bank slope and riparian vegetation, undercut banks, and substrate changes associated with point bars (Fitzpatrick and others, 1998). Bankfull data from riffle and run transects only (no pools) were used to calculate reach-averaged dimensions. Transect measurements from pool units can overestimate bankfull dimensions and were, therefore, excluded from the bankfull reach-averaged dimension calculations. Bankfull dimensions then were normalized by drainage area for consistency across the study area. Original drainage areas were smaller for some Salt Lake City streams than in other studies because of issues with a nested design and drainage-network alterations. Drainage areas based on topographic divides were additionally calculated for these streams and used for the drainage-area normalized variables.

Habitat metric and summary data were retrieved from the Habitat Data Analysis System (HDAS, version 4, M.C. Peppler, U.S. Geological Survey, written commun., 2009) for each study area (table 3-K, 468 KB). Several habitat metrics were calculated by HDAS to summarize channel characteristics in the sampling reach. Wetted cross-sectional area, wetted perimeter, and channel shape were calculated at each transect and summarized for the reach by using mean, minimum, and maximum values. Hydraulic conditions were summarized by calculating the mean Froude number, mean Manning’s roughness, and a flow-stability index (an indicator of flashiness). Habitat volume measurements included wetted width and depth, reach area and volume, maximum depth, discharge at the time of sampling, and average velocity and coefficient of variation in velocity. Habitat-cover data were not used because many of the streams had water depths of less than 30 centimeters (cm), a limitation of the protocol for small streams selected for the EUSE study.

Raw point and transect-level data were retrieved from internal databases and used to calculate several additional metrics. Point observations of dominant substrate size (categories) were summarized into percentage of transect points with fine (sand-sized or smaller), gravel, cobble, and boulder or bedrock substrate. Other additional substrate characteristics included median particle size (D50) and percentage of embedded particles (to the nearest 10 percent). The definitions tab in the spreadsheet defines the particle-size categories. The percentage of disturbed riparian land cover within a 30-m buffer was calculated on the basis of transect endpoint description of land cover. Disturbed land cover (RipLUDis) included cropland, pasture, farmsteads, residential, commercial, or transportation. Undisturbed land cover included grassland, shrubs and woodland, or wetland. Water-surface slope (ReachSlope) was measured at reaches in eight of the nine study areas.

Additional information was collected on local channel modifications and natural controls that might affect habitat and geomorphic responses to urbanization. Using a combination of remarks on field forms, notes on field-drawn reach diagrams and maps, photographs of the reach taken during the time of sampling, topographic maps and aerial photographs, estimates were made of the percentage of the reach with bank stabilization or channelization, number of grade-control structures (weirs, low-head dams, culverts) within the reach and within a distance of one reach length upstream or downstream, presence or absence of bedrock cropping out in the channel, and presence or absence of depositional features (lateral, mid, or point bars). The degree of confidence in these techniques varied by metropolitan area, depending on the level of details included on field forms and maps and the number of field photographs (Faith Fitzpatrick, U.S. Geological Survey, written commun., 2009). The calculated values represent a minimum for each type of control (table 3-L, 19 K).

NAWQA Home Page EUSE NAWQA Home page DS 423 Home Page

Accessibility FOIA Privacy Policies and Notices

Take Pride in America logo logo U.S. Department of the Interior | U.S. Geological Survey
Page Contact Information: Contact USGS
Page Last Modified: Monday, 28-Nov-2016 14:29:40 EST