Terence D. Cooke, Consulting Scientist

Woodward-Clyde International Americas, 500 12th Street, Oakland, CA 94067-4014

Phone: (510) 874-1736

David D. Drury, Senior Civil Engineer

Santa Clara Valley Water District, 5750 Almaden Expressway, San Jose, CA 95118-3686

Phone: (408) 927-0710

The Santa Clara Valley Urban Runoff Pollution Prevention Program (Program) has been conducting investigations of pollutant sources to urban stormwater runoff since 1987. Characterization of pollutants in runoff indicated many of the metals and other pollutants are present as suspended sediment in receiving streams during storm events. To better focus program resources a pilot sediment sampling study was conducted to develop techniques for bed sediment characterization and examine the impact of urban drainage on local waterways. The study developed field sampling and laboratory analysis methods to estimate how sediment chemistry changes throughout the watershed and to relate these changes to different drainage area characteristics (soil type, size, local sources, landuse). Prospective statistical power analysis was used to ensure sufficient samples were collected for statistical comparisons. Sample compositing, replication, and field sieving was used to minimize variability and enable measurements of metals in the fine sediment fraction. The laboratory testing program included techniques to distinguish labile metal (likely associated with pollution inputs) from residual metals (likely associated with erosion of soil minerals). Sampling and analysis protocols were tested by application to six stream reaches throughout the watershed. The results were used to refine sediment sampling methods and to develop data analysis methods for evaluation of the impacts of urban drainage on bed sediment concentrations, and to evaluate disposal options for maintenance dredging of improved channels.

The Santa Clara Valley Urban Runoff Pollution Prevention Program (Program) has been conducting investigations of pollutant sources to urban stormwater runoff since 1987. Characterization of pollutants in runoff indicated many of the metals and other pollutants are present as suspended sediment in receiving streams during storm events. To better focus program resources a pilot sediment sampling study was conducted to develop techniques for bed sediment characterization and examine the impact of urban drainage on sediments in local waterways. The goals of the pilot sediment sampling study were to:

• Develop sampling and analysis methods estimate how sediment chemistry changes throughout a watershed;
• Relate these changes to natural and man-made sources of pollution;
• Determine "Enrichment factor" = (concentration in the sediment)/(background or natural concentrations).

Background is assumed to be represented by sediment samples taken in the upland portions of streams where the catchment is primarily open space. The study results are presented in four sections: Stream Reach Selection, Field Sampling, Laboratory Analysis and Data Analysis.

The pilot study was implemented in the Calabazas Creek watershed. Calabazas Creek watershed is highly developed (80% urbanized) and has significant flood control improvements in the most heavily urbanized areas of the watershed. For this initial pilot study is was decided to gain a general overview of sediment chemistry in the main channel (rather than urban tributaries) and to test the field and laboratory approach. For these reasons stream reaches were selected in the main stem of the creek rather than in tributaries.

Several different watershed conditions were considered during the stream reach selection process. The reach selection process was used to identify specific stream reaches within the pilot watershed for sampling. Stream reaches rather than specific locations within the stream were selected to enable field verification of places within the stream where sedimentation occurs and therefore where sampling was feasible. Stream reaches are defined as sections of the stream channel 50’ to 200’ in length. Within each reach, specific stations were chosen for sampling during field reconnaissance. The general stream reach selection process included the following steps:

1. Define and qualitatively classify the major subwatersheds within each watershed based on general land use and topography as upland and urbanized.
1. Delineate major storm drains and tributaries
2. Determine important characteristics for each subwatershed which can influence sediment composition and quantity including :

Six stream reaches were selected for characterization. Five of the reaches were chosen in the nontidal portion of the stream. A sixth reach was selected in the tidally influenced portion of the stream. Stream reaches that drain different types of watershed areas (upland, and urban) were selected for characterization in the nontidal areas. Upland areas were characterized as having steep slopes and primarily open (undeveloped) land-use. Urban areas generally have lower slopes with a high percentage of residential, commercial or industrial land-use. Figure 1 shows the location of the selected stream reaches.

Two reaches were selected for characterization of Upland sediments. Reach C1 is located on Calabazas Creek below Comers Debris Basin. The reach below the basin was selected because sediments settled in the basin are removed by the District and therefore not a source to the Bay. Reach C2 is located on Prospect Creek above its confluence with Calabazas. Prospect Creek is a major tributary to Calabazas, receiving drainage largely from open land-use areas. The Prospect Creek tributary rather than Calabazas after the confluence with Prospect was selected to enable comparison of sediment metals in the two major upland tributaries.

Four reaches were selected along Calabazas Creek in the urban land-use area to determine if a gradient in enrichment factors occurs as land use becomes more urbanized. The reaches were selected downstream from major tributaries which drain urban land-use areas. Sites were selected such that the incremental drainage areas were comparable. Reach C3 is located in Calabazas Creek downstream from the confluence with Rodeo and Regnart Creek tributaries. These two tributaries drain primarily residential land-use areas. Reach C4 is located in Calabazas Creek near Homestead Ave. This station is located downstream of Interstate 280 and receives drainage from the Junipero Sierra storm drain and the industrial and residential areas located between the two reaches. Reach C5 contains the Program storm water quality sampling station (S3) located at Wilcox High School. Between Reach C4 and C5 Calabazas Creek receives drainage from Residential, Commercial, and Industrial land-use areas. Reach C6 is located in Calabazas Creek at Alviso. This station is located in the tidally influenced area and has been previously characterized by the District. Between reaches C5 and C6 Calabazas Creek receives drainage from residential, commercial, and industrial areas.

The primary objective of the sediment sampling design was to provide a method for collection of sediment from stream beds which would enable estimation of mean metals concentrations in fine and coarse grain sized sediments. The study design focused on estimation of metals concentrations within stream reaches which were defined as sections of the stream channel 50’ to 200’ in length.

To estimate how many samples would be necessary to detect a given change in mean metal concentration a statistical technique called a prospective power analysis was used. A power analysis predicts the probability that a specific difference in concentration will be statistically significant, if present. The number of samples necessary to prove significance depends on the variability of the data and the magnitude of the difference. To estimate the number of samples necessary prior to collecting any data the expected variability was estimated by analyzing existing sediment data collected by the program and USGS. Variability can be characterized as the coefficient of variation (CV) which is:

where:

s = standard deviation, and

x = mean

None of the existing data contained copper concentrations from multiple samples within one stream reach. Therefore, data from all samples collected in Calabazas Creek were pooled and used to estimate mean and standard deviation and calculate a CV. The CV calculated from the pooled data is higher than what would be expected within each reach because data were collected over several years and from many different reaches. Using the pooled data a CV of 0.5 is estimated. Using this CV in a power analysis with a required confidence interval of 95% (a = 0.05) and a required power of 80% (b =0.20) indicated that a 40% difference in mean concentration could be detected if seven samples per reach were collected. For comparison to detect a 30% difference between reaches would require 12 samples per reach while a 60% difference in concentration could be detected with three samples per reach. Table 1 shows the results of the power analysis for different values of CV and different minimum percent differences. Because the estimated CV was thought to be higher than what was expected in the current study, five stations were selected to be characterized within each reach.

Chemicals in sediments can be highly variable on small spatial scales. Compositing was used to minimize the effects of small scale variability. Field sieving was used to enable collection of sufficient coarse grainsized sediment to enable separate laboratory analysis of the fine fraction. Figure 2 presents a schematic of the field and laboratory sampling and compositing scheme. Briefly, six stream reaches were sampled. Within each reach five stations were selected, generally separated by 20’ to 50’. In each station, three substations were sampled. Sampling each substation consisted of combining three or more scoops of sediment into the field sieve (2 mm), wet or dry sieving (depending on the sediment moisture and composition), and collection and transfer of the < 2 mm fraction into sample jars for transfer to the laboratory. In the laboratory, composite samples from individual substations (e.g., C1-1.1.comp, C1-1.2.comp and C1-1.3.comp) were composited by mixing equal volumes in an acid rinsed plastic container to produce each station composite (C1-1.comp). The station composite was then split into four subsamples. Subsample 1 was wet-sieved in the laboratory to separate the fine (< 63µm) and coarse (63 µm -2000 µm) fractions and each fraction was analyzed for total recoverable and 0.5 N HCl leachable metals and total volatile solids (fines only). Station composite subsample 2 was analyzed directly for total recoverable metals, TOC, grain size, pH, moisture, and total sulfides.

Station composite subsample 3 was combined with subsamples from the five other station composites collected from the same stream reach (e.g. C1-2.comp, C1-3.comp, C1-4.comp, C1-5.comp, C1-6.comp) to form one reach composite (C1.comp). The reach composite was analyzed for Organics, BOD, Nutrients, Volatile Solids. Station composite subsample 4 was frozen and archived for future analysis.

The primary objectives of the laboratory analysis were:

1. To provide a method for preparation and analysis of sediment samples that would enable a technically defensible comparison of metals content of fine and coarse sediments.
1. To determine if a weak acid leaching method can be used in freshwater sediments to provide an indication of potential biologically available metals and how these measurements compare to total recoverable metals concentrations.
2. To determine if other constituents should be analyzed along with metals to aid in the interpretation of the data.

A wet-sieving procedure was used in the laboratory to separate the fine (< 63 µm) and coarse (63-2000 µm) fractions. The wet-sieving procedure consisted of placing a subsample of station composite onto a 63 µm mesh stainless steel sieve and washing the fine sediment through the sieve using reagent water. The coarse sediment from the top of the sieve was transferred into sample containers using acid rinsed plastic scoops. The rinsate water and fine sediment were collected in an acid rinsed plastic tub and transferred into a acid cleaned glass beaker. The beaker was then placed onto a hot plate and heated at low temperatures to evaporate most of the rinsate, leaving fine sediment and metals associated with the rinsate water in the beakers. The rinsate water was included with the fine sediment to prevent loss of metals that were weakly associated with the sediment which may have gone into solution as a result of the wet-sieving. It should be noted metals weakly associated with the coarse fraction would also be included in the fine fraction. Loss of volatile solids (analyzed in the fine fraction) due to the evaporation procedure was minimized through the use of low heat. However, some loss may have occurred.

A weak acid leaching procedure used by researchers at University of California Santa Cruz (UCSC) conducting sediment metals characterization studies was performed on the fine and coarse fractions of the station composite samples to determine how the results compared to the total recoverable fraction and provide an approximation of the potentially bioavailable metals fraction. ("Sediment Extraction for Analysis of Bio-Available Metals Dry Leach Method, 4/1/91" Dr. Khalil Abu-Saba). The procedure involves a mild acid digestion (0.5 N HCl, 24-hour, room temperature) designed to solubilize only ‘labile’ metals.

Three types of data analysis procedures used in the pilot sediment sampling study are described below. The three types are: reach average enrichment factor calculations; enrichment factor analysis with normalization to aluminum, and statistical hypothesis testing.

The primary objective of the sediment sampling study was to estimate how sediment chemistry changes throughout a watershed and to relate these changes to natural and man-made sources of pollution. One measure of changes in sediment chemistry is defined by the "enrichment factor" which is the ratio of the concentration of a given pollutant in the sediment (measured as milligrams of pollutant per kg of sediment - mg/kg) at a given location within the watershed to the corresponding concentration at a location that represents background or natural concentrations. In this case background is assumed to be represented by sediment samples taken in the upland portions of streams where the catchment is primarily open space. An alternative method which may be used to calculate enrichment is to ‘normalize’ the measured pollutant concentrations in the sediment to the amount of native minerals. The advantage of the normalization procedure is that variations in the amount of native mineral are accounted for, thereby reducing apparent ‘false enrichment’ due to sediments with high mineral content. Both types of enrichment calculations are described below.

Reach average enrichment factors can be calculated for each stream reach. The enrichment factors for the reach are calculated as the average of the enrichment factors for each station within the reach as follows:

where:

EF _D = average enrichment factor for the downstream reach of interest

M _{D.1.comp -D.5.comp} = are metal concentration at downstream stations 1 through 5 in reach D

avg (M _U) = average metal concentration of upstream stations

In the case of the pilot study only reach C1 was used as representative of the upstream areas of the watershed based on apparent differences between copper concentrations in fine sediment from reach C2 as compared to reach C1. Figure 3 shows the results of the calculations for copper. The highest enrichments were found for the fine grained sediment in the urban reaches. However, fine grain sediment was less than two percent of the nongravel sediment in the urban reaches. Whole sediment was not enriched in urban reaches as compared to the upland reach. Sediments in the tidally influenced reach (C6) were 98 % fine grain sized. Fine sediment was only slightly enriched as compared to upland sediments. However, because fines accounted for the majority of the sediment whole sediment in this reach showed an apparent enrichment factor of 1.8. These results indicate enrichment factors for similar grain sizes should be compared rather than whole sediment.

Metal concentrations can also be normalized to other factors which are measured in the same sample. Normalization allows determining how much a given metal is enriched beyond what might be expected due to the erosion of minerals from the upland soils. Aluminum is a major component of clay minerals. Iron is also a major mineral phase and many metals are known to adsorb onto iron oxide mineral phases. Normalization to upstream aluminum or iron to metal ratios provides an indication if elevated metal concentrations in downstream reaches are due to accumulation of upstream clay or iron minerals in depositional environments or are actually enriched above what would be expected from upstream minerals. Normalization is done using the following steps:

1. Perform a linear regression analysis with the upstream station data only using the mineral indicator metal (aluminum or iron) as the x variable and the metal of interest as the y variable.
1. Plot the results of the regression.
2. Show the confidence limits of the linear regression fit on the plot.
3. Plot the downstream data on the plot.
4. Visually inspect the plot to determine if the downstream station data contain higher metal concentrations than the upstream ratio data could support (data points are above the upper confidence limits of the fit) or are depleted in metals as compared to the upstream data (data points are below the lower confidence limit of the fit).

An example of the normalization procedure conducted for copper is presented in Figure 4. The normalization procedure was conducted using reaches C1 and C2 as representative of the upstream ratios (line on the linear regression plot) and by plotting the remaining reach stations individually. The figure shows that all downstream reaches, with the exception of the intertidal reach (C6), contain stations with elevated copper concentrations. The intertidal reach actually contained less copper than is expected based on the aluminum data.

Statistical hypothesis testing is used to determine if observed differences between reaches and fractions are significantly different within a desired confidence interval. Null hypothesis are tested to determine if they can be rejected or accepted depending. An example of a null hypothesis is

If the hypothesis is rejected the implication is that the copper concentrations in different size fractions are significantly different.

Comparison of total and weak acid leachable metals concentrations in different size fractions and among different stream reaches was accomplished using a one-way analysis of variance (ANOVA), if multiple reaches were compared, or a t-test, if only two reaches were compared. Nonparameteric statistical tests were used because the results of Levene’s test showed that the variances were not equal (P < 0.05). Figure 5 shows the results of the Wilcoxon Kruskal-Wallis Test performed using the copper data from nontidally influenced reaches. The results of the testing indicate the copper concentrations in the different sediment size fractions were different (p < 0.001). Examination of the percentile box plots and score sum statistic in the Kruskal-Wallis Test shows that the < 63 um fraction is much higher than either the >63 um or whole sediment copper concentrations.

The following conclusions were used to develop sediment sampling programs for addition watersheds in Santa Clara Valley:

• Field and laboratory compositing methods proved useful in averaging out small scale spatial variability within a given stream reach.
• Three to four composite samples per reach were necessary to detect 30% differences in metals concentrations between reaches with 80% power.
• Tidally influenced reaches showed significantly different chemistry as compared to nontidally influenced stream reaches.
• Fine sediment generally comprises a small fraction of the whole sediment in nontidal stream reach segments.
• Fine sediment can have high enrichment factors in nontidal reaches that receive urban drainage.
• Aluminum and iron are useful indicators of mineral enrichment and can be used to normalize metals concentrations to account for these differences.

The Santa Clara Valley Water District used the general field and laboratory testing methods developed during the conduct of this study to improve the reliability of the chemical characterization of sediments in stream channels that require maintenance dredging or realignment. Results from the analysis of other watersheds confirmed the larger enrichment in tidally influenced stream reaches as compared to nontidally influenced reaches.

Figure 4. Regression enrichment analysis for copper in fine sediments (mg/kg dry weight).
Linear regression shows copper/aluminum ratio in upstream reaches.