USGS - science for a changing world

Scientific Investigations Report 2006–5101–D

Scientific Investigations Report 2006–5101–D

Back to Table of Contents


Physical Characteristics

Geographic Setting

Urban and agricultural watershed development in the Willamette River basin and surrounding area followed the prevailing natural regional topography: most development was in the flat valley lowlands rather than in the higher elevation foothills and mountains. This was evident by the strong correlation between the natural environmental setting metrics (mean and minimum watershed elevation) and the urban indicator metrics (UII, percentage impervious surface, POPDEN00, percentage urban + agriculture, and ROADDEN (table 2). Environmental setting affected the natural characteristics of streams through variations in precipitation, erosion, and instream habitat as mediated by natural channel geomorphology and geology. Therefore, even without human influence, there were minor to moderate differences from the higher gradient foothill streams to the low-gradient valley streams. However, many of the environmental setting metrics overemphasized these differences because some variables were calculated for the whole watershed and not just the local area surrounding the sampling site. A good example was mean watershed slope and watershed elevation. Because many streams in the Willamette River basin and surrounding area originated in the foothills or mountains, calculated watershed metrics included parts of the higher elevation and higher gradient reaches even though the characteristics of the stream at the sampling site reflected the location of the stream with a low-gradient valley. The environmental setting metrics were calculated this way to provide measures of watershed characteristics that were consistent nationally and simple to calculate. There was an effort to minimize the natural differences among sites by selecting stream sampling reaches that were within the low-gradient valley, even though a large part of the upper watershed may be in a different ecoregion. For example, 75 percent of the sites were within 80 m of the overall mean elevation of 220 m; sites ranged between 50 m and 620 m in elevation. Correlation of minimum watershed elevation to the urban indicator metrics, although still moderately statistically significant, was somewhat lower than the correlation of mean watershed elevation to the urban indicator metrics (table 2).

Stream Hydrology

Increased flow variability, or stream “flashiness” in the form of frequent high peaks and low troughs, is considered a key effect of urbanization on streams (Paul and Meyer, 2001; McMahon and others, 2003; Konrad and Booth, 2005; Roy and others, 2005). Konrad and Booth (2005), after reviewing the literature and analyzing a small number of sites with streamflow gaging stations from reference and urban dominated land use watersheds, determined that the frequency of “high-flow events as measured as the number of events three times above the median flow” and the “percent daily change” (flashiness) were the two most sensitive measures of changes in hydrographs due to urbanization. In their analysis, these two variables also were significantly correlated with algae, macroinvertebrates, and fish assemblage metrics. Data from this EUSE study corroborates these findings. For example, the four hydrologic variability metrics PeriodF5, PeriodF9, PeriodR5, and Richards-Baker Flashiness Index (Rb-flash), which related to rate in streamflow change, had significant correlation with the urban indicator metrics (table 3). PeriodF5, PeriodF9 and PeriodR5 were metrics that summarized the frequency of periods of falling (F) or rising (R) stream-discharge events, where hourly stream-discharge change was greater than or equal to 5 or greater than or equal to 9 multiplied by the median decrease or increase over the period of record (table A5). For example, PeriodF5 referred to the number of hours when streamflow fell over the period of record by at least five times the median flow for that site. The Rb-flash characterizes the degree that streamflow changed relative to the daily median. PeriodF9, the metric that documented the number of falling hydrologic events greater than nine times the site median, had the strongest correlation values—rho 0.69 with the UII (fig. 5) and 0.71 with road density (ROADDEN) (table 3). Associations among pairs of variables are shown as regression graphs (figs. 5 and 6) with simple linear or curvilinear trend lines added to aid interpretation.

Water Temperature

Water temperature metrics generally did not correlate strongly with any urban indicator metrics; however, the minimum water temperature metric (95th percentile) was positively correlated with the UII (0.56) (table 3) and negatively correlated with pollution sensitive diatoms (presented in the algae section, below). Because site selection was restricted to valley streams, the natural range in water temperature was narrow compared with the range in larger geographic areas or other ecosystems. Temperature data were lost during the hot summer months at three sites (Silk, Chehalem, and South Scappoose Creeks) due to transducer failure.

Stream Habitat

Stream habitat metrics did not have particularly strong correlation values with any urban indicator metrics; the strongest correlation was between percentage of riffle habitat and UII and POPDEN00, rho = -0.63 (table 3). The weak correlations among habitat metrics and urban indicator metrics likely were due to the study design; sites were selected to minimize natural differences to increase the chances of isolating the effects of urbanization (Short and others, 2005). Therefore, habitat measurements may have a better relation to changes in urbanization than was revealed in this study. Although not strongly correlated to urban indicators, certain habitat metrics did correspond well to water-chemistry metrics. For example, percentage of riffle habitat was strongly correlated with the summer DO concentrations (rho = 0.84; fig. 6).

Chemical Characteristics

Pesticide Occurrence

Ninety-six stream samples were analyzed for 64 pesticides and degradation compounds. Among the samples, 28 pesticides or degradates were detected including 12 herbicides, 8 insecticides, 2 fungicides, and 6 degradates (fig. 7). At least one pesticide or degradation product was detected in 83 percent of the samples. Among all samples, the six most frequently detected pesticides were herbicides and herbicide degradates: atrazine, deethylatrazine (degradate of atrazine), simazine, hexazinone, prometon, and metolachlor. The highest frequency of occurrence was for atrazine, detected in 49 percent of all samples. Other pesticides with at least 10 detections include 3,4-dichloroaniline (degradate of diuron, and other phenylurea herbicides), tebuthiuron, trifluralin, carbaryl, diazinon, chlorpyrifos, metalaxyl, and myclobutanil. Ten or more pesticides were detected at 7 sites in either the spring or summer sampling; North Fork Deep Creek had 10 or more detections in both samplings (table 4).

Generally, individual pesticide concentrations were relatively low (fig. 7). The median concentration for any pesticide was 0.02 µg/L or less, and only seven pesticide concentrations exceeded 0.1 µg/L. The highest concentration of any sample was 1.72 µg/L for atrazine, and the highest combined pesticide concentration for a sample was 2.08 µg/L, occurring at Battle Creek during the spring 2004 sampling.

Seasonal Variability

On average, there were more pesticide detections and higher pesticide concentrations during the spring sampling than during the summer (tables 4 and 5). For example, nearly twice as many pesticides were detected in spring, including 116 herbicides (including 24 degradates), 27 insecticides (including 4 degradates) and 8 fungicides, compared to summer when 61 herbicides (including 16 degradates), 16 insecticides (including 2 degradates) and 7 fungicides were detected (table 4). During spring sampling, 5 or more pesticides were detected at 16 sites, whereas during summer, at least 5 pesticides were detected at only 7 sites. Salmon Creek and Iler Creek, two minimally effected sites, were the only two streams with no pesticide detections in either spring or summer samplings. Most pesticide detections were in North Fork Deep Creek in both sampling periods likely because this area includes the highest amount of agricultural land use (48 percent) of any of the watersheds (U.S. Geological Survey, 2005). In addition to higher frequency of detections in the spring, total pesticide concentration for spring was more than 3 times greater than in summer, although a large part of this difference was due to one large concentration at Battle Creek, Oregon, during the spring sampling (table 5).

Herbicides were detected at 25 sites during the spring sampling and 19 sites in the summer (table 4). Atrazine, hexazinone, deethylatrazine, simazine, and prometon were the five most commonly detected herbicides in the spring, all with a detection frequency greater than 46 percent. During summer sampling the highest detection frequencies were for deethylatrazine, simazine, prometon, 3,4-dichloroaniline, and metolachlor, ranging between 25 and 43 percent, with prometon being the highest. Unlike atrazine, which primarily is used for agricultural purposes, and simazine, which is used in both urban and agricultural applications, prometon is used mostly for nonagricultural purposes, such as domestic and commercial applications to driveways, fence lines, lawns, and gardens. Prometon also can be used as an asphalt additive (Gilliom and others, 2006). Previous research documented a direct relation between urban land use and prometon detection frequency in surface water and ground water (Koplin and others, 1998), so it was not surprising to see such a high frequency of detection in our study. Where insecticides were detected, carbaryl and diazinon were predominant. Carbaryl, an agricultural and urban insecticide, was detected 39 percent of the time in spring and 18 percent in summer. Diazinon was detected slightly less than carbaryl in spring at 29 percent, yet slightly more in summer at 21 percent. Due to changes in pesticide regulations, residential uses of diazinon were cancelled in 2004, but use is still approved for agriculture.

Temporal Variability

During the 10 site “high frequency” sampling effort, pesticides were detected in all samples (6 sample times) for the 3 most highly-urbanized sites (Claggett, Pringle, and Kellogg Creeks) and in 2 mixed agricultural-urban sites (North Fork Deep and Tickle Creeks) (table 4). Pesticides were detected most frequently at North Fork Deep Creek, with at least eight pesticides detected in each sample. The fewest pesticides were detected in Salmon Creek, a predominantly forested watershed (UII = 20), with only one detection in March 2004. This relatively low pesticide detection frequency in Salmon Creek likely was due to the low amount of agricultural land in this watershed (2 percent). Three pesticides were detected in three of the six samplings (50 percent) in the East Fork Dairy Creek watershed, even though it had only 1 percent combined urban plus agriculture land use (the site with the UII = 0). This probably was due to the close upstream proximity to the sampling site of a variety of agricultural activities (for example, Christmas tree plantations and nursery operations) even though they were of small acreage and therefore did not add significantly to the total of agricultural land use summarized as a percentage of the total watershed area.

Overall, 44 of the 60 samples collected contained 2 or more pesticides and 11 of the 60 samples contained 10 or more pesticides. On average, between 5 and 11 pesticides were detected at the 3 most highly-urban sites during the 6 high frequency samplings, and between 1 to 3 pesticides were detected at the 3 lowest-urban sites. Streams draining predominantly urban watersheds have been shown to have higher detection frequencies and concentrations of some insecticides than other types of land uses (Anderson and others, 1997; Gilliom and others, 2006), and the results for this study followed this pattern. High frequency samples collected at the 3 most urban sites (UII ≥ 88) (table 1) had a detection of at least one insecticide in 58 percent of samples; whereas high frequency samples collected at the 3 least urban sites (UII ≤ 20) had insecticide detections in only 8 percent of samples.

Pesticide Metrics in Relation to Urban Intensity Index

Relations between the UII and pesticide occurrence were strongest when considering the total number of pesticides and total concentration of all pesticides in a sample. For the spring and summer samplings a high UII was associated with a large number of pesticides detected in a sample. Comparison among groupings of sites based on the four levels of UII shaded in tables 4 and 5 (low: less than 10; medium: 10 to 25; high: 25 to 70; and very high: greater than 70) reveals some interesting patterns. For example, when summed across the spring and summer samplings, an average of 12 pesticides were detected in both the high and very high UII groups (table 4). The number of detections dropped substantially, four pesticides detected on average, when only looking at medium UII sites and only one pesticide was detected on average in low UII sites (UII less than 10) (table 4). The pattern of herbicide concentrations varied from this, as average total herbicide concentrations were higher for high UII sites than when compared to very high UII sites (UII greater than 70), even though the number of herbicide detections were the same between these two groups of sites. The higher herbicide concentrations of the high UII group compared to the very high UII group remained even after the extreme herbicide value from Battle Creek was removed. However, for insecticides, the average concentration was more than 2.5 times greater in the very high UII group of sites than the high UII group.

The fact that the high UII group averaged as many pesticide detections as the very high group, likely was due to the influence of agricultural land use in the watersheds. The amount of agricultural land in the watersheds in the high UII group ranged from 16 to 48 percent (an average of 31 percent), and the amount of urban ranged 7 to 72 percent (an average of 30 percent) (table 1; fig. 8). Therefore, many high category UII watersheds had about the same amount of influence from agricultural land use as urban land use. On the other hand, the very high UII category, which was dominated by urban land use (60–98 percent urban), had a relatively minor influence from agricultural land use (0–8 percent agriculture). In terms of the number of pesticides detected in streams, little difference was observed between agricultural and urban land; however, the composition of the pesticide mixture and the timing of delivery to the stream varied considerably between agricultural and urban sites in the study. These differences in pesticide detection frequency and types of pesticides between agriculture and urban land use are similar to those reported in the Willamette Valley by Anderson and others (1997) and in streams across the country by Gilliom and others (2006).

Among individual pesticides detected during this study, only prometon showed a significant correlation with UII (rho = 0.70), and then only during the spring sampling. Total pesticide and insecticide concentrations (log transformed due to extreme values that skew the distribution; log [X + 0.0001] summed across spring and summer) were strongly correlated with the UII (rho = 0.68 and 0.69, respectively) (figs. 9, 10, and table 6). The correlation of log total pesticide concentration increased slightly when correlated to percentage urban plus agricultural land (rho = 0.72), yet the correlation decreased dramatically when related to only percent agricultural land (rho = 0.41). Conversely, the correlation of log total insecticide concentration decreased when related to urban plus agricultural land compared to its correlation to UII (rho = 0.63 and 0.69, respectively) (table 6). This suggests that many insecticide detections originated from applications in urban areas, not from the agricultural uses.

Pesticide Toxicity Index in Relation to Urban Intensity Index

PTI scores at the 28 sites typically were greater in spring than summer (18 greater, 8 less, 2 the same). The difference between most pairs of PTI scores was small, as 20 of the 28 scores changed by one order of magnitude or less between the two samplings. The sum of the spring and summer PTI values was used to estimate the potential pesticide toxicity among sites and to follow our summary of actual pesticide detections and concentrations presented previously in this report. The sum of the PTI was significantly correlated with the UII (rho = 0.63, fig. 11A), yet had a stronger correlation to ROADDEN at rho = 0.69 (fig. 11B). The relation of PTI to ROADDEN was curvilinear and revealed two basic groups of sites with relatively high PTI values. One group of sites with the highest road density (ROADDEN greater than 10) also had the highest percentage of urban land use or highest UII values. Another group of sites with moderate road density had a combination of moderate percentage of urban land and substantial amounts of agricultural land (ROADDEN of 3.5 to 8.5 and PTI values greater than 4). This pattern followed the results of a number of pesticide detections stated above, such that high detection frequencies occurred at sites with high urban land use and at sites with lower amounts of urban land use but with moderate amounts of agricultural land.

Semipermeable Membrane Device Assays in Relation to Urban Intensity Index

Of the three assays run on the SPMDs, the toxic equivalents (TEQ) index (P450 RGS assay for aryl hydrocarbon receptor agonists) and Pyrene Index (fluoroscan for total PAHs) provided consistent and reliable results. No interpretable results were achieved from the Microtox® assay and are not discussed (Bryant and others, 2007). The TEQ and Pyrene Index assays were correlated to the five urban indicator metrics, with TEQ having the strongest correlation to both UII and POPDEN00 at rho = 0.81 (table 6; fig. 12). Bryant and others (2007) determined similar strong correlations of the TEQ and Pyrene indices compared to the individual UIIs of other USGS EUSE studies in Atlanta, Georgia; Raleigh-Durham, North Carolina; Denver, Colorado; Dallas-Fort Worth, Texas; and Milwaukee-Green Bay, Wisconsin. They also concluded that the strong correlation of UII with pentachloroanisole and pyrogenic PAHs in the other study areas was evidence that these compounds were an important part of urbanization regardless of geographic location.

Along with the three assays, part of each SPMD dialysate was analyzed for hydrophobic chemical compounds. Of the 141 compounds targeted for identification by gas chromatography and mass spectrometry analysis, 39 were detected in the Willamette River basin and surrounding area. In comparison, detection in the other 5 EUSE studies ranged from 49 compounds detected in Raleigh-Durham to 36 in Dallas-Fort Worth (Bryant and others, 2007). Only three PAH compounds detected in the Willamette River basin and surrounding area were significantly correlated to the UII, and this was the lowest number of significant correlations among the six EUSE studies (high of 21, Raleigh-Durham).

Nutrients and Field Parameters

For spring and summer samplings, total nitrogen (TN), total phosphorus (TP), and orthophosphorus (soluble reactive phosphorus; SRP) had positive correlations with the UII (table 6). The highest two TN concentrations of all sampling sites were in Curtin Creek during spring and summer (4.8 and 3.9 mg/L, respectively). Curtin Creek was considered an outlier due to the relatively high TN values measured, which were probably caused by the large amount of ground-water inflow (ground-water that is high in DO and TN due to the natural coarse grain geology, which likely minimizes the amount of denitrification) just upstream of the sampling site. TN concentrations averaged for the spring and summer samplings and nutrient index were positively correlated to the UII (TN: rho = 0.79; fig. 13A) (nutrient index: rho = 0.85; table 6). The highest TP concentration (0.18 mg/L) was in Beaverton Creek in spring, and in Claggett Creek in summer (TP = 0.28 mg/L).

Some of the highest TN values were in the medium to high UII groups of sites (UII 25 to 70), likely due to the increased amount of agricultural land at these moderately urban sites (fig. 8). As a result, the correlation of TN increased further when percentage of agricultural land was included with percentage of urban land as the correlative variable (rho = 0.86; fig. 13B). Nevertheless, lower TN concentrations (generally less than 0.5 mg/L) were detected in sites with relatively low urban development (UII less than 25) , whereas relatively higher TN concentrations ranging from 0.7 to 2.3 mg/L were commonly detected in higher UII sites (greater than 25), (not including the outlier value for Curtin Creek). The pattern for TP was not as consistent as shown for TN and its correlation decreased when the amount of agriculture was included (table 6), nevertheless, high UII sites generally had the highest TP concentrations ranging from 0.08 to 0.28 mg/L.

Phosphorus concentrations (TP and SRP) in many stream sites increased from spring to summer as streamflow decreased towards base-flow. This likely was due to inputs of phosphorus in ground water to streams, though other possible explanations include increased water use and increased influences of wastewater inputs from septic systems or treatment plants. Twenty-two of the 28 sites showed this pattern of increased SRP from summer to spring. Nitrogen concentrations, however, decreased from spring to summer and were more variable than phosphorus concentrations, possibly reflecting inputs from runoff of spring fertilizer applications in urban and agricultural land that may have subsided during the dry summer months. Most streams showed decreased concentrations of dissolved inorganic nitrogen (DIN) through the growing season, which may reflect the tendency for nitrogen to be in relatively short supply compared with phosphorus in some Northwest streams, particularly during summer (Carpenter, 2003). About one-third of the streams had DIN concentrations that were 0.5 to 1.7 mg/L lower during summer compared with spring. The greatest change in the DIN concentration from spring to summer was in North Fork Deep Creek, where nuisance levels of filamentous green algae (Cladophoraglomerata) contributed to relatively high chlorophyll-a levels (157 mg/ m2).

Among the water chemistry variables, strongest correlations to the UII (rho greater than 0.70) was for DOC and dissolved sulfate (SO4; table 6) whereas, specific conductance, bicarbonate alkalinity, chloride, and summer dissolved oxygen (DO) (negative) also were correlated, but slightly less significantly (rho greater than 0.50). Sulfate sources include fertilizers, road pavement amendments, and certain algicides (copper sulfate, for example), and is often produced during combustion. Sulfate also is produced when bacteria in organic soils oxidize hydrogen sulfide (H2S). Potential sources of bicarbonate in urban areas include the slow erosion of concrete structures, sidewalks, and roadways, and calcium carbonate based lime products applied to lawns for pH control.

DO is a critical parameter for aquatic life in streams, and is affected by a number of processes, including water temperature, atmospheric pressure, and the activity of bacteria, algae, and other aerobic organisms that consume DO, and processes that produce it (aeration in riffles, for example, and photosynthesis by algae). All DO data used in this report were instantaneous measurements collected during midday, and do not reflect the daily cycle of DO that often occurs in nutrient-enriched streams with high algal production. It is likely that in many of these urban nutrient-enriched streams the DO may show large diurnal swings; very low DO in early morning after nighttime anerobic activity and super-saturated DO in late afternoon after photosynthesis by abundant algae.

Biological Characteristics

Algae Assemblages

Algal assemblages were dominated by pennate diatoms (Pennales Order), which comprised 214 of the 254 algal taxa identified in RTH (richest target habitat) riffle samples from the 28 sites (table 7; table A9). Based on biomass, however, Chlorphytes (green algae) were dominant, contributing on average about 70 percent of the total algal biovolume, whereas diatoms comprised 17 percent of the total algal biovolume. The most common diatoms in RTH samples were Achnanthidium minutissimum, Rhoicosphenia abbreviata, Cocconeis placentula var. euglypta, Planothidium lanceolatum, Navicula minima, Gomphonema kobayasii, Sellaphora seminulum and Achnanthes subhudsonis var. kraeuselii (table 7).

Based on cell density (number of cells/cm2) blue-green and red algae were the dominant taxa at all but one site (Deep Creek), with the blue-green Homeothrix janthina dominating 11 sites and unidentified red algae (vegetative “chantransia” stage) six sites (table 8). The red and blue-green algae have relatively small cells, and, therefore, tended to dominate cell densities. Many dominant diatoms, particularly at sites high on the UII, were high-nutrient (eutrophic) taxa, or preferred high TN concentrations, and were tolerant of moderate levels of DO (greater than 75 percent saturation) (table 8). Although many sites lower on the UII also were dominated by eutrophic diatom taxa, several were dominated by Achnanthes and Achnanthidium species whose water-quality preferences have not yet been established.

Sixty-seven percent of the total algal biovolume (for all RTH samples combined) was comprised of filamentous green algae, including Cladophora glomerata, Stigeoclonium, Odeogonium, and Spirogyra (table 7). The occurrence of these high-biomass forming filamentous green algae was sporadic along the UII, as they were detected at few sites despite relatively high nutrient levels. In addition to requiring high nutrients, these taxa also prefer relatively high light levels, which was limited in some streams where riparian vegetation or topographic relief provided shading. High sediment concentrations in some streams also may have limited light availability. Potapova and others (2005) determined that light conditions affected algal assemblages in streams around Salt Lake City, Utah, due to riparian vegetation, stream size, and suspended sediment. Carpenter and Waite (2000) determined filamentous blue-green algae, such as Oscillatoria, to be common in silt-laden agricultural streams in the Willamette Valley, possibly due to their ability to move and unbury themselves after siltation events, or from an inherent ability to grow under low light conditions.

Response in Algal Biomass to Urban Intensity Index

Benthic algal biomass was highly variable along the urban gradient, with chlorophyll-a values ranging from 5 to 212 mg/m2, and showed no obvious response to urbanization (fig. 14A). AFDM, a measure of the organic matter present, ranged from 2.4 to 70 g/m2, and was positively correlated with the UII (rho = 0.56; fig. 14B) and the nutrient index (rho = 0.72; fig. 15A). All but one site less than 25 on the UII had an AFDM value that indicated at least a moderate degree of organic enrichment, and many sites higher on the UII (greater than 25) exceeded the criterion to be considered organically enriched (fig. 15A; Biggs, 1996). Twelve of 28 streams had chlorophyll-a concentrations exceeding 50 mg/m2, a low-end threshold suggested to protect recreational and aesthetic qualities of streams (Biggs, 1996). The highest chlorophyll-a concentrations occurred in North Fork Deep and Amazon Creeks (157 and 212 mg/m2, respectively) due to high abundances of Cladophora glomerata, Oscillatoria princes, and Sellaphora seminulum (fig. 14A). The chlorophyll-a concentrations in these streams also exceeded common nuisance indicator levels for benthic algae 100–150 mg/m2 (Horner and others, 1983; Welch and others, 1988, 1989; Biggs, 1996; Dodds and others, 1997, 1998). Proliferations of algae may develop quickly during periods of stable streamflow, especially in streams receiving nutrients. Newall and Walsh (2005) found that repeated rainfall events can stimulate algal growth in streams by providing pulses of nutrients, an effect that was enhanced by the amount of impervious surface and the degree of drainage connection within the storm-water network.

Although there was considerable variation between AFDM and DOC (fig. 15B), the highest AFDM values occurred when the DOC exceeded about 4 mg/L, and DOC was negatively correlated with DO concentrations (fig. 15C). Taken together, the relations among algal biomass, DOC, and DO indicate that algal biomass may be affecting DO levels through bacterial decomposition processes involving the production of DOC. DO also is affected, however, by water temperature and the amount of riffle habitat that aerates the water (fig. 6). Because diurnal fluctuations in DO and pH can occur from algal photosynthesis, however, the one-time instantaneous midday measurements collected for this study likely do not fully reflect the processes of photosynthesis and respiration that may occur in these streams.

Elevated levels of dissolved nutrients can stimulate the growth of benthic and planktonic algae in streams. In some cases, high-biomass forming benthic algae such as filamentous Chlorophytes (green algae) may cover streams and foul substrates when high amounts of light are available for photosynthesis (Carpenter and Waite, 2000). In addition to the prolific growths of green algae described above, another high-nutrient indicator alga—Melosira varians—was detected at more than 50 percent of the sites (16 sites; table 7), making up about 10 percent of the total biovolume for all RTH samples. This eutrophic diatom is also a N-heterotroph because it may use organic forms of nitrogen for energy and growth. Melosira is considered a high quality food item for benthic macroinvertebrates because of its high nutrient and fatty acid content. This filamentous diatom has a morphology of relatively loosely connected cells that make it susceptible to removal by disturbance such as repeated scouring flows, high water velocity, or grazing by herbivorous benthic macroinvertebrates. Because of its tendency to fragment, Melosira does not tend to reach as high densities or biomass as other more resistant types (for example, Cladophora or Stigeoclonium) in disturbed habitats.

Multivariate Analysis of Diatom Assemblages

A number of environmental factors such as climate, geology, water-quality, habitat conditions, and anthropogenic disturbances (Biggs, 1990) can influence algal assemblages. In this study, diatom assemblage structure (axes scores from nMDS ordinations) were not significantly correlated (rholess than 0.2) to any of the urban indicators including road and population density, percentage of impervious area and urban land, or the UII. Additional multivariate analyses were conducted to identify which environmental variables (or combinations) explained the most variation in the diatom assemblage structure using the BEST routine in PRIMER. Variables included habitat parameters (water depth, velocity, and embeddedness), disturbance indicators (benthic macroinvertebrate grazers and a hydrologic variability index), light availability (open canopy), and water-quality measures (nutrients, DOC, pH, DO, and specific conductance). The best combination of variables—DOC, pH, Rb-flash and benthic macroinvertebrate scraper density—explained 68 percent of the variation in the diatom assemblage.

DOC, which was negatively correlated with DO (fig. 15), explained 44 percent of the variation in the diatoms assemblage structure among all sites. The highest individual rho values were those associated with habitat and channel hydraulics (percentage run habitat, Froude number, and maximum depth, rho = 0.47; average substrate embeddedness, rho = 0.32) and water chemistry (high-flow period specific conductance, rho =0.43; summer total phosphorus, rho = 0.33; summer particulate nitrogen, rho = 0.33; summer minimum water temperature, rho = 0.33). Habitat and hydraulic conditions can alter the velocity regime for benthic algae, which can affect its overall growth form and profile (Hoagland and others, 1982), which is consistent with the influence of Froude number and Rb-flash on the diatom assemblage structure. In these streams, the higher amount of run habitat (and gradient) also might be contributing to higher sedimentation, leading to higher average substrate embeddedness. The water-quality variable with the highest rho value was specific conductance. Specific conductance often is used as a broad measure of anthropogenic influence, but is also affected by dilution (and watershed size), as well as natural factors such as soil and geology. Specific conductance has been shown to correlate well with algal assemblages in other studies in Oregon (Walker and Pan, 2006), other EUSE study areas (Potapova and others, 2005), and in Australia (Newall and Walsh, 2005). Correlations between specific conductance and anthropogenic influences can be stronger than nitrogen and phosphorus concentrations in streams with significant algal growths because of the nutrient uptake effect, which can lower nutrient levels substantially during periods of active growth.

Response in Algal Metrics to Urban Intensity Index and Select Environmental Variables

Two algal metrics were correlated to the UII, including percentage of diatoms requiring high DO concentrations (nearly saturated) and, to a lesser degree, percentage of eutrophic (high-nutrient indicator) algae (fig. 16). Algal metrics also indicate that nutrient and sediment enrichment have measurable effects on the diatom assemblage structure in these streams, with eight metrics having significant correlations with either the UII or other environmental variable (table 9). The highest correlation coefficients (rho values) occurred between algal metrics and water-quality variables, including nutrients, DOC, and DO, and measures of algal biomass, especially AFMD (table 9).

The percentage of high-nutrient indicators (eutrophic diatoms) was strongly correlated to specific conductance (rho = 0.67) and moderately correlated to soluble reactive phosphorus (rho = 0.65), total phosphorus (rho = 0.55), and the nutrient index (rho = 0.52) (table 9). Another high-nutrient indicator metric (nitrogen heterotrophic taxa)—those that can use organic forms of nitrogen (organic nitrogen)—also were significantly correlated with concentrations of total nitrogen and total phosphorus (rho = 0.62 and 0.69, respectively) and algal biomass (AFDM: rho = 0.73; chlorophyll-a: rho = 0.69) (table 9).

Algal metrics also showed the effects of depressed levels of DO in these streams that can result from bacterial respiration associated with decomposition of organic matter. For example, low-oxygen indicating taxa tolerant of depressed DO (10–30 percent saturation, or less [van Dam and others, 1994]) were positively correlated with total phosphorus (rho = 0.62), DOC (rho = 0.69), and benthic algal biomass (AFDM, rho = 0.63) (table 9). In contrast, high oxygen indicator diatoms had negative correlations for most water-quality variables, particularly specific conductance and soluble reactive phosphorus (rho = -0.74 and -0.62, respectively). Additionally, the percentage of taxa associated with high levels of organic enrichment (a-mesosaprobic diatoms) was significantly correlated with concentrations of total phosphorus (organic nutrient: rho = 0.64) (table 9). These taxa tolerate depressed DO levels (10–70 percent saturation) and are associated with biological oxygen demand (BOD) levels of 4–22 mg/L (van Dam and others, 1994).

The percentage of diatoms tolerant of nutrient and organic pollution (Bahls, 1993) were similarly positively correlated to total phosphorus and total nitrogen concentrations (tolerant taxa: rho = 0.67 and 0.71, respectively) and algal biomass (AFDM: rho = 0.66; chlorophyll-a: rho = 0.62). The percentage of pollution sensitive diatoms, however, showed the opposite pattern and was negatively correlated with TP and TN (sensitive taxa: rho values = -0.68; and -0.62, respectively) and algal biomass (AFDM: rho = -0.69; chlorophyll-a: rho = ‑0.56). Lastly, the Silt Index—the percentage of motile diatom genera Navicula and Nitzschia—was positively correlated with total phosphorus and AFDM (table 9). These organisms can thrive in streams affected by siltation because they can move out of the sediments to the surface where light levels are higher.

Taken together, the algal data show that diatom assemblages are affected by variations in streamflow, grazing by herbivorous benthic invertebrates, and processes relating to DOC (organic matter formation from excessive nutrients and high water temperature and decomposition, and subsequent effects of algal and bacterial metabolism on concentrations of DO). The positive correlation between benthic organic matter (AFDM) and the UII indicate that urbanization increases the amount of algae and other organic matter in streams through nutrient and (or) organic enrichment. DO is an important factor for important fish such as trout and salmon, which require relatively high levels of DO for survival and reproduction.

Benthic Macroinvertebrate Assemblages

One-hundred thirty-nine unique benthic macroinvertebrate taxa were identified in the 28 RTH samples (table 10 and table A10). The most taxa (52) was for the insect order Diptera, with 37 from one dipteran family (Chironomidae), commonly known as midges. Diptera made up one-quarter of the total number of taxa collected, and had the highest taxa richness per family or order, by far. EPT orders—Ephemeroptera (mayflies), Plecoptera (stoneflies), and Trichoptera (caddisflies) had 13, 10, and 19 taxa per respective order. There were 12 taxa within the Coleoptera order (beetles), 2 taxa in Odonata (dragonflies and damselflies), and 1 each in Lepidoptera (butterflies and moths) and Megaloptera (dobsonflies). In addition to these insects, 29 noninsect taxa were spread among 17 orders, including snails, clams, aquatic worms, amphipods, and mites (table 10).

Only Simulium caadense (a dipteran blackfly) and Fluminicola (a gastropod snail) reached maximum single-sample abundances greater than 11,000 specimens per m2. Other major insect and noninsect orders had maximum abundances of between 1,300 and 4,100 specimens per sample. The noninsect Acari (Hydracarina, or water mites) and Simulium canadense (Diptera) were the most common taxa collected, occurring at 27 out of 28 sites (96 percentage occurrence; table 11), in addition, four of the top six most common taxa were other noninsects: Acari, Juga sp. (a snail), Dero sp. (Oligochaete worm), and Lumbriculidae (Oligochaete worm). Of the 22 taxa with at least 50 percent occurrence, 9 were noninsects, 6 were Diptera (5 chironomid midges and one blackfly, Simulium), 3 were Ephemeroptera, and 2 each were from Trichoptera and Coleoptera orders (table 11). Eighteen of these 22 taxa were considered moderately to highly tolerant of poor water-quality conditions, although 4 taxa (Paraleptophlebia sp., Zapada cinctipes, Rhithrogena sp., and Ceratopsyche cockerelli) were considered moderately to relatively sensitive.

Benthic Macroinvertebrate Metrics in Relation to Urban Intensity Index

The benthic macroinvertebrate metric “percent dominance” (percentage abundance of the maximum single taxon) has been considered a good bioindicator by some researchers (Barbour and others, 1999); however, in this and other recent studies, it does not correlate well with disturbance in this geographic region. In this study, percent dominance ranged from 14 percent to 67 percent, yet it only had a correlation to the UII of rho = 0.31. However, the tolerance values of the single dominant taxa did show a good relation to the UII. A list of sites, sorted by UII (table 12), showed that the dominant taxa of the 17 sites with a UII greater than 25 had tolerance values between 7 and 10 (average 8), as determined by the USEPA tolerance scale (least to most tolerant—0 to 10). Additionally, those sites with UII greater than 25 had, on average, EPT richness of 4 taxa totaling 15 percent. On the other hand, sites with a UII less than 25 had tolerance values for their dominant taxa of between 3 and 8 (average 5), with much greater numbers of EPT taxa (average of 15 EPT taxa or a percent EPT richness of 44 percent). As a result, although percent dominance as a metric by itself did not correlate well to disturbance, interpreting ecological characteristics of the single dominant taxa among sites, such as USEPA tolerance values and optima, was useful (table 12). Looking at all sites instead of just the end members, there also was a strong negative correlation between EPT richness and UII (rho = -0.75; fig. 17) that reflects this large decrease in EPT taxa from the low to high urban sites.

Cuffney and others (2005) compared benthic macroinvertebrate metrics from three urban studies for two site groupings with different urban intensities: (1) most highly urban sites sampled (UII ≥ 70) and (2) sites near reference condition (UII ≤ 10). They detected large differences between these two groupings in the richness and density of major benthic macroinvertebrate metrics including all taxa, EPT taxa only, Diptera, Chironomidae, and noninsects. Results from this study were similar, with an average increase of 12 EPT taxa and 32 percent EPT richness from the high urban group to the near reference condition group (table 13). However, unlike Cuffney and others (2005), who showed a large increase in total richness (+25 taxa), we detected only a small increase in total taxa richness (an average increase of 5 taxa). In this study, the large increase in “EPT percent richness” was in addition to a large increase in average abundance of intolerant macroinvertebrates (+3,848) and Ephemeroptera (+3,111). There also was a large decrease in percentage of noninsect richness (-31 percent) and abundance of tolerant taxa (-48 percent). Plecoptera are one of the most sensitive aquatic insect orders and, on average, almost no Plecoptera were detected at the highly urbanized sites, yet on average, 14 percent abundance was detected at the sites near reference condition. Total taxa richness did not show a strong relation to urbanization in the Willamette Valley, even though richness of individual groups like EPT and noninsects had a strong relation to the UII. This indicates that there is specie replacement along the gradient, such that as sensitive EPT taxa drop out as urbanization increases noninsects and chironomid taxa take their place and total taxa richness remains relatively the same. Therefore, like the metric percent dominance mentioned previously, total taxa richness does not work well as a metric indicative of disturbance in this region even though it often is useful in other geographic regions.

Many benthic macroinvertebrate metrics had strong correlations to the UII. The three greatest rho values were for Ephemeroptera richness (negative), Plecoptera richness (negative), and the abundance of tolerant taxa (positive; table 14). Spearman correlation coefficients for these three benthic macroinvertebrate metrics and the five urban indicators were all greater than ±0.69, with the strongest positive correlation between tolerant taxa and the UII (0.79) and the strongest negative between EPEM richness and POPDEN00 (-0.79; table 14). Many benthic macroinvertebrate metrics also had relatively strong correlation values with selected environmental setting metrics, including measures of soil erosion potential, elevation, precipitation, watershed slope, and percentage of low elevation flat land in the watershed (table 14). Urban and agricultural land use development follows the natural topography in the Willamette Valley; a higher percentage of development is in the flatter low-elevation valley and less in the higher elevation foothills. Cuffney and others (2005) determined strong correlations for similar environmental setting variables with benthic macroinvertebrate metrics, but only for the more mountainous Salt Lake City region and not for the Boston or Birmingham areas. In the Willamette River basin and surrounding area, a number of benthic macroinvertebrate metrics had their strongest correlation coefficients to water-chemistry metrics, such as TP, total insecticide and total pesticide concentration, PTI, TEQ, and the Pyrene Index. Most Spearman correlation coefficients between benthic macroinvertebrate richness metrics and the aforementioned water-chemistry metrics were greater than 0.70 (table 14). The four strongest correlations were between percent Diptera richness (not including chironomid midges) and the Pyrene Index (rho = 0.88) and the TEQ (0.85), and between percent richness of tolerant taxa and the TEQ (0.87), and between the EPT: Chironomid ratio and the total insecticide concentration (-0.85; table 14). These strong correlation values between water-chemistry metrics and benthic macroinvertebrate metrics are similar to those published by Cuffney and others (2005), and indicate that one effect of pesticides and other potentially toxic compounds is a reduction in the number of sensitive insect taxa in favor of more tolerant chironomid midge larvae.

A shift in the benthic macroinvertebrate assemblage toward less palatable organisms such as worms (Oligochaetes), or snails potentially could affect fish assemblages. Any reduction in EPT insect abundance or taxa richness could have implications for salmonids and other fish in this region because EPT taxa are important contributors to aquatic food webs, linking algae with fish. EPT taxa include herbivorous caddisflies and mayflies, which play an important role in food webs by grazing algae. The EPT taxa typically emerge into flying adults in a chronologic sequence that lasts nearly year-round, in a pattern predictable to the local angler and resident fish. Reductions in EPT taxa could, reduce food available for fish that may lead to reductions in production or changes in species composition indirectly, in addition, stream conditions such as temperature, DO, or other factors that affect macroinvertebrates may also directly affect the natural fish assemblage.

Multivariate Analysis of Benthic Macroinvertebrate Assemblages

Ordination analysis took advantage of the full species assemblage at each site to determine patterns among sites based on the biological data. This approach provided a more complete picture compared to analysis of individual benthic macroinvertebrate metrics, which examined selected components of the assemblage. The first nMDS ordination axis scores summarizes the major variation among sites as revealed by the full benthic macroinvertebrate species data (see Methods: Data Reduction and Analysis for explanation of ordinations), the scores reveal how the sites spread in a 2-dimensional plot based on the species occurring at each site. Therefore, sites that plot close to each other are similar in species composition, whereas sites that plot far apart from one another are very different in species composition. The first ordination axis scores were negatively correlated to the UII with a consistent and fairly even distribution of points, whereas the percentage of low elevation flat land in the watershed, shown to have relatively high correlations only to a few invertebrate metrics, did not correlate as well (figs, 18A and C), nor did the points spread consistently over the range of ordination scores. The ordination axis scores on the other hand, also had strong correlations to many of the same water-quality parameters as did the individual metrics (fig. 18B) and the correlation values were within a similar range as the metrics (TEQ; rho = -0.87; fig. 18D). The PRIMER BEST routine identified six variables—TEQ, sum of total pesticides, average embeddedness, DO or percent riffle habitat (surrogates for each other), 7-day average water temperature, and the UII or percentage of urban plus agricultural land (surrogates for each other)—that explained about 65 percent of the variation in the benthic macroinvertebrate assemblages among all sites.

Although the first ordination axis used data from the complete benthic macroinvertebrate assemblage, the richness of EPT or Ephemeroptera metrics were strongly correlated with the nMDS axis score (rho greater than 0.90), which indicates that the first ordination axis largely described the same variation as the individual metrics, namely the change in richness of EPT or Ephemeroptera. This indicates that the disturbance gradient (UII) in the Willamette Valley was strong enough that differences among sites could be detected with only a part of the benthic macroinvertebrate assemblage. There is enough taxa diversity in the benthic macroinvertebrate data that part of the information is redundant. For example, EPT richness, a subset of the full species data set, could explain as much variation among sites as the full data (such as ordination axis).

Most benthic macroinvertebrate metrics showed linear responses to urbanization with no apparent threshold except, possibly, for EPT richness. EPT richness plotted against the UII showed a strong negative trend as urbanization increased and may have exhibited a threshold response near UII = 25 (fig. 17). All sites less than UII 25 had greater than 12 or more EPT taxa (average 15) except for South Scappoose Creek, whereas all sites greater than UII 25 had less than 9 EPT taxa (average 4), except for Curtin Creek (table 12). South Scappoose Creek likely has a lower EPT richness because it is a low gradient stream with minimal riffle habitat with potentially more urban influence than reflected by the UII score. Conversely, Curtin Creek had a greater abundance of EPT taxa than other streams greater than UII 25 because it had cold, clear summer flows due to a large amount of ground-water discharge upstream of the sampling site, thus offering better water quality habitat than what the UII would suggest. In addition, no insecticides were detected in Curtin Creek (table 4) and it had low TEQ and Pyrene Index values. As a result, Curtin Creek had remarkable numbers of EPT taxa and low percent taxa dominance even with little in-stream habitat. Cuffney and others (2005) determined that responses of benthic macroinvertebrate metrics in the pilot USGS EUSE study areas generally were linear and without thresholds, except for a few selected metrics for the Boston-area. No thresholds or initial resistance to the effects of urbanization for full assemblage measures or ordination axes were detected in any region (Cuffney and others, 2005).

Although there appeared to be a possible threshold in EPT richness at UII equal to 25, the apparent threshold was likely due to added agricultural land use in all sites greater than UII 25 than to any actual urban threshold. For example, at UII less than 25 the total percentage of urban plus agricultural land in the watershed was less than 16 percent (except for Deep Creek at 34 percent), yet immediately greater than a UII of 25, urban plus agricultural land increased markedly to between 53 and 98 percent (except for Oak Creek at 26 percent, table 1 and fig. 8). Therefore, any apparent threshold along the UII gradient was likely a threshold of the total of urban plus agricultural land use. However, the exact threshold is unknown because only two sites (Oak and Deep Creek) had urban plus agricultural land use percentages between 16 and 52 percent; therefore, there was not enough information in this data range to more fully evaluate thresholds. The plot of EPT richness and UII (fig. 17) indicates that if it exists, a threshold is at low values of urban plus agricultural land use, perhaps as low as 10 percent combined land use.

Although the streamflow, water temperature, and habitat measurements did not have as strong correlations as water-chemistry metrics to benthic macroinvertebrate metrics, they did have statistically significant values (greater than 0.60) for correlations of selected variables to a few metrics (table 14). The four hydrologic variables (PeriodF5, PeriodF9, PeriodR5 and Rb-flash) had correlations to a few benthic macroinvertebrate metrics greater than 0.60. For habitat measures, correlation values were this strong only for percentages of riffle and embeddedness correlated with Ephemeroptera abundance and tolerant abundance. Water temperature had only one correlation to benthic macroinvertebrate metrics greater than 0.60, with Oligochaete percent richness.

Fish Assemblages

Fish Metrics in Relation to Urban Intensity Index

Total fish richness ranged from 2 to 12 species, total abundance ranged from 52 to 672, and maximum relative abundance or percent dominance of any single species ranged from 20 to 98 percent among all sites (table 15). Sixty percent of the sites had six species or fewer, yet there was no strong correlation of number of fish species to total abundance. For example, the site with the highest number of species (12) had a total abundance of 163 (North Fork Deep Creek) and the site with the lowest numbers of species (2) had an abundance of 380 (Tyron Creek). Western streams naturally have relatively low fish species richness compared to streams east of the Rocky Mountains (Simon and Lyons, 1995; Meador and others, 2005), and as a result, fish species richness from western streams generally have not been a good bioindicator. For example, a poor relation of fish species richness to UII was determined in this study, yet Meador and others (2005) determined a strong relation between fish species richness and urbanization in the Boston and Birmingham areas. On the other hand, nonnative or invasive species are a more serious problem in western than eastern streams and nonnative fish were in approximately 50 percent of sites greater than a UII of 25, although only one occurrence was at a site less than 25 (table 15). Amazon Creek, with a UII of 77, contained the most nonnative species (five species) and highest percent abundance of nonnatives (98 percent) of any site. Claggett Creek, with a UII of 100, was the next highest with four nonnative species (19 percent abundance).

Although total species richness was not different between low and high UII sites, percentage of salmonids and nonnatives were different. Salmonids were present at 10 of 11 sites less than UII 25, whereas salmonids were present in only 4 of 17 sites greater than UII 25. No sites with salmonids had any nonnatives except one site, North Fork Deep Creek with 33 percent salmonids and 5 percent nonnatives. In this study, percent dominance (relative abundance) calculated from fish assemblage data had a strong curvilinear or possible threshold relation to the UII (rho = 0.67) where low urban sites (less than UII 25; table 15) had low percent dominance (average of 46 percent), although sites greater than UII 25 had average dominance values greater than 80 percent (fig. 19). The threshold for percent dominance at UII of 25 was equivalent to about 3–5 percent impervious surface, which was lower than most other thresholds previously reported for fish community metrics (Lyons and others, 1996; Wang and others, 2001). The apparent urban threshold likely was due to added agricultural land use in sites greater than UII 25 and therefore the threshold likely represents the effect from total watershed disturbance of urban plus agricultural land use and not just urbanization.

The fish index, which combined individual metrics of percentages of salmonids, reticulate sculpins, nonnatives, and natives (with salmonids and reticulates removed), had high correlation values to urban indicators (UII: rho = -0.68; table 16). The fish index also was strongly correlated (rho greater than 0.60) to PeriodF5 hydrology, embeddedness, width-depth ratio, riffle percentage, summer DO, DOC, specific conductance, sulfate concentration, and TEQ (table 16). The highest correlation values were between percentage of salmonids and summer DO (rho = -0.81) and percentage of riffle (rho = ‑0.78).

Multivariate Analysis of Fish Assemblages

Bioindicators, such as ordination axis scores and individual metrics, were correlated against individual environmental variables to gain insight into what was structuring or potentially affecting the fish assemblages. The full fish assemblage, as represented by the scores of the first nMDS ordination axis, was negatively correlated to the UII with no apparent threshold (rho = -0.76; fig. 20A). With the exception of the percentage of salmonids metric (which had some high rho correlation values; table 16), the nMDS ordination axes scores had stronger correlations to the environmental variables than the fish index or individual metrics.

The relations between the ordination axis 1 scores and summer DO, TEQ, and PeriodF5 are shown in figures 20B–D. A strong linear response was noted in the fish assemblages along the UII and in response to DO concentrations. The response in the TEQ, however, may indicate a potential threshold between 300 and 500 picograms TEQ (fig. 20B). The axis 1 ordination scores also were strongly related to three hydrology variation metrics (PeriodF5; fig. 20D), percent riffle, minimum temperature, DOC, and sum of insecticides and pesticides (table 16). The BEST routine in PRIMER indicated that summer DO, percent riffles and PeriodF5 could explain about 60 percent of the variation in full fish assemblage data among the sites. The percentage of salmonids compared to the summer DO concentration may also be indicative of a possible threshold response (fig. 21). No salmonids were sampled at sites with midday summer DO concentrations less than 8 mg/L, except at Curtin Creek. Because salmonids require cool waters with high DO to thrive, few were observed during the summer in these Willamette Valley ecoregion streams that often can have high summer temperatures with low dissolved oxygen. The Oregon Department of Environmental Quality standard for DO is 5.5 mg/L and 18 degrees Celsius (7-day moving average for the minimum summer water temperature). Curtin Creek was a unique site and although little high-quality fish habitat was within the sampled reach, it offered salmonids cold clear water during the summer with abundant instream habitat cover in the form of macrophytes and overhanging riparian vegetation. As a result, salmonids probably immigrated from nearby streams into Curtin Creek to take advantage of the cold water during summer.

Back to Table of Contents

AccessibilityFOIAPrivacyPolicies and Notices

Take Pride in America logoU.S. Department of the Interior | U.S. Geological Survey
Page Contact Information: Publications Team
Page Last Modified: Thursday, 01-Dec-2016 19:24:19 EST