Study Objectives and Approach


The five main study objectives were to (1) characterize the seasonal quantity and quality of organic carbon in the Clackamas River and its primary tributaries; (2) relate the amount and composition of organic carbon to the formation of DBPs at sites throughout the watershed and in finished drinking water; (3) evaluate the major suspected sources of DBP precursors in the watershed, including tributaries, North Fork Reservoir, litter/soils, and algae; (4) assess the use of optical properties, including in-situ FDOM, for estimating DOC and DBP precursor concentrations; and (5) assess the treatability of DOC and DBP precursors by conducting “jar-test” experiments at one of the drinking-water treatment plants. 


The approach consisted of (1) continuous in-situ fluorescence monitoring of source water at the CRW DWTP in the lower river from April 2010 to September 2011; (2) monthly water sampling at four main-stem sites (table 2) for DOC, TPC and total particulate nitrogen (TPN), total and dissolved nutrients, and optical properties including absorbance and fluorescence (table 3); (3) four basin‑wide surveys: spring high flow, summer base-flow period, late‑summer reservoir drawdown, and the first “initial” autumn storm (table 4); and (4) four treatability tests on source water using standard jar-tests procedures to evaluate the potential for coagulant and powdered activated carbon (PAC) to reduce DOC and DBP precursor concentrations. During each basin-wide survey, DBP formation potentials (DBPFP) for THMs (THMFPs) and HAAs (HAAFPs) were measured to determine DBP precursor concentrations in surface water in the major lower-basin tributaries, North Fork Reservoir at the log boom, and the Clackamas River mainstem.


Study Site Descriptions


The study sites (tables 2 and 3) included 7 in the mainstem: three sites in the upper basin including North Fork Reservoir at the log boom, and 5 sites in the lower basin, including two DWTPs that obtain source water from the river in the lower 3 miles (fig. 1). Five tributaries that drain a range in land cover, from mostly forested to mostly urban, also were sampled. There are two major reservoirs in the basin—Timothy Lake in the headwaters of the Oak Grove Fork and North Fork Reservoir on the middle mainstem. Timothy Lake is typically maintained at full pool during most of summer, and then drawn down in September to accommodate autumn and winter precipitation and to augment flows during the low‑flow period. While no sampling occurred at Timothy Lake, North Fork Reservoir was sampled each year during summer blooms of blue-green algae. In contrast to Timothy Lake, North Fork is operated as a run-of-the-river reservoir, along with two other downstream diversions and dams that serve Faraday and River Mill hydroelectric facilities. Residence times are thus typically short for North Fork (on average, hours up to approximately 7 days, possibly longer depending on temperature stratification) compared to Timothy Lake (on average, approximately 8.5 months). More information on these reservoirs and the associated hydroelectric project is presented in Carpenter (2003).


Discrete Sampling


Discrete water samples were collected from within the watershed and at the DWTPs during each monthly and all basin-wide and reservoir samplings (table 3). Seven additional discrete water samples were collected at the CRW DWTP intake over a range in flow/turbidity conditions and analyzed for DOC concentration, absorbance, and fluorescence to compare laboratory measurements to the in-situ FDOM sensor response.


Watershed Samples


Water samples were collected from wadeable stream sites (table 2) using the Equal-Width Increment method, where the entire stream cross section was sampled using a depth‑integrating sampler (Edwards and Glysson, 1999). Water samples were collected from unwadable main-stem sites using a D-74 sampler in mid-stream (at bridge sites). Samples from non-wadable, bridgeless sites were collected as grab samples from the main flow directly into 1-L combusted amber glass bottles. Point samples from each site were composited and homogenized in a 16-L Teflon™ churn splitter and placed on ice prior to dispensing subsamples for nutrients, DOC, optical properties, DBPFPs, and water‑column chlorophyll-a. Samples for DOC and optical properties (absorbance and fluorescence) were filtered using 25-mm, 0.7-µm precombusted Whatman™ GF/F filters with a glass filtration unit. The filtrate was collected into combusted amber-glass bottles with Teflon™-lined caps and stored in the dark at 4°C until analyzed. All samples were analyzed within 5 days of collection.


Dissolved nutrient samples were filtered through a Pall™ 0.45 µm pore-size capsule filter. Total and dissolved nutrient samples were immediately frozen and stored at -20°C until analysis. Whole-water samples for total phosphorus and total nitrogen were obtained from the churn splitter. Samples for TPN and TPC were dispensed from the churn splitter into 125-mL baked amber-glass bottles. The contents were filtered using precombusted 25-mm Whatman™ GF/F filters (0.7-µm pore size), noting the filtrate volume. Filters were frozen and shipped to the USGS National Water-Quality Laboratory (NWQL) in Denver, Colorado, for analysis. A known volume of sample water was filtered for water-column chlorophyll-a using the same GF/F filters as above. Filters were wrapped in aluminum foil, frozen at -4°C, and analyzed within 30 days at the Oregon Graduate Institute in Portland, Oregon. Samples for DBPFP were collected into 1-L baked amber glass bottles and placed on ice. Formation potentials were evaluated on filtered and unfiltered samples. Filtered samples were passed through 142-mm GF/F filters (0.7 µm pore size) into 1-L baked amber glass bottles. DBPFP samples were acidified to a pH of 2.0 units using reagent-grade concentrated HCl to prevent possible microbial transformations in the samples prior to analysis. Samples were shipped on ice within 24 hours to the organic chemistry laboratory at the USGS California Water Science Center (CAWSC) in Sacramento, California, where they were refrigerated at 4°C until further processing. 


Source-Water and Finished-Water Samples


Raw intake “source” water was collected from both DWTPs. At the CRW DWTP, source-water samples were collected adjacent to the FDOM sensor inside the gravity-fed vault adjacent to the river, where water enters before being pumped up to the treatment plant. Samples at the LO DWTP were collected from taps inside the treatment plant. Treated or “finished” water was collected 90 minutes after source-water sampling to account for time-of-travel through each plant. 


Because DBPs commonly continue to form within the distribution system, on one occasion, as part of the regular monthly sampling, an extra DBP sample was collected from within the distribution systems of both water utilities, at the location where historical maximum DBP concentrations have occurred. These samples were collected on November 2, 2010, 1 day following the collection of the regular finished-water samples to account for time-of-travel.


Finished-water samples for determination of DBPs were collected by CRW and LO personnel into baked amber glass bottles with a Teflon™ septa to help remove bubbles. Bottles contained a quenching agent (65 mg sodium thiosulfate for THMs, final concentration 1.5 g/L, and 150 mg ammonium chloride for HAAs, final concentration of 1.18 mg/L) to fully oxidize the residual chlorine and stop further DBP formation. THM samples were collected directly into 40-mL baked amber glass vials used for collection of volatile organic compounds, using care to slowly and completely fill each vial to prevent bubble formation. HAA samples were collected into 125-mL baked amber glass bottles. Sample collection and processing of other source- and finished-water samples were conducted as described above for watershed samples.


Benthic Algae (Periphyton)


Periphyton samples were collected for biomass (chlorophyll-a) and dominant species composition at four main-stem sites and five tributary sites in July to September 2010. For the tributaries, samples were collected from 10 representative rocks/cobbles in shallow riffle areas; in the mainstem, samples were collected along wadeable portions of the channel margin. All samples were collected using USGS methods (Moulton and others, 2002) described in detail in Carpenter (2003). A known area of periphyton (approximately 4-5-in. diameter circle) was scraped from the top of each rock into a plastic dishpan using a plastic bristle brush. Multiple samples were composited and transferred to plastic 1-L bottles and placed on ice until processing (described below).


Laboratory Analytical Procedures


Organic Carbon


DOC concentrations were measured in duplicate using the platinum catalyzed persulfate wet-oxidation method on an O.I. Analytical Model 700 TOC Analyzer™ (Aiken and others, 1992), which produced standard errors of ±0.2 mg/L carbon. The standard curve, consisting of a minimum of five standards over the range of interest, was repeated for every 10–12 water samples; reported values are the average of two duplicate measurements on each sample. 


TPC samples were collected and analyzed using USEPA method 440.0. Particles from a known volume of water were filtered onto 25-mm GF/F filters (0.7-µm pore size), which were frozen at -4°C and shipped to the USGS NWQL for analysis. TOC was calculated as the sum of DOC and TPC.


Absorbance and Fluorescence


The absorption spectra was measured between 200 and 750 nm on filtered samples at constant 25°C temperature with a J&M TIDAS™ spectrophotometer, using a 1-cm quartz cell and distilled water for the blank. SUVA was calculated by dividing UVA₂₅₄ by DOC concentration and is reported in L/mg-m units (Weishaar and others, 2003). Spectral slopes were calculated using a non-linear fit of an exponential function to the absorption spectrum over specified wavelength ranges (275–295 nm and 350-400 nm, for example) as described by Twardowski and others (2004). The spectral slope ratio (SR) was calculated as the ratio of S_275–295 to S_350–400 (Helms and others, 2008) (table 1). Because of the marked decrease in UVA absorbing DOM following treatment, an exponential fit could not be applied to the absorbance curve; thus, spectral slope was not calculated for finished‑water samples.


Fluorescence EEMs were measured on filtered samples in a 1-cm cuvette at 20°C with a SPEX Fluoromax–4 spectrofluorometer (Horiba Jobin Yvon, Edison, New Jersey) using a 150W Xenon™ lamp, a 5-nm band pass, and 0.05-second integration time. Fluorescence intensity was measured at excitation wavelengths of 240 to 450 nm at 10-nm intervals and emission wavelengths of 300 to 600 nm at 2-nm intervals on room-temperature samples (25°C) in a 1-cm quartz cell. The resulting matrix consisted of 2,291 individual ex–em pairs that form the basis of the EEM diagrams. EEMs were blank corrected, instrument corrected, and normalized to the daily water Raman peak area, and the Rayleigh scatter lines were removed. The FI was calculated as the ratio of emissions at 470 to 520 nm at an excitation of 370 nm (McKnight and others, 2001; Cory and others, 2010). The Humic Index (HIX) was calculated by dividing the sum of fluorescence intensities at emission 436–480 nm by emission 300–346 nm at excitation 254 nm (Zsolnay and others, 1999). 


Disinfection By-Product Formation Potentials


DBPFP was determined on filtered and unfiltered samples to determine the relative importance of the dissolved and particulate fractions. Both THMFP and HAAFPs were determined following a version of USEPA Methods 502.2, 510.1, and 552.2 as described by Crepeau and others (2004). Briefly, the method involved a 7-day reaction time, pH buffered at 8.3, temperature held at 25°C, and final, residual‑free chlorine concentration restricted to between 2–5 mg/L. This incubation period provides information on the total DBP precursor pool and should result in concentrations of THM and HAA reflective of potential distribution-system concentrations based on residence times within the systems. For consistency, the same quenching agents used for the finished-water samples were used to quench the chlorination reaction—sodium thiosulfate for THMFP and ammonium chloride for HAAFP. However, instead of using solid material, quenching agents were dissolved in deionized water and added to the reaction vials (approximately 3 µL/mL, 0.3 percent by volume) to obtain equivalent concentrations to those used for the finished-water sample vials. Chlorine dosing and quenching were conducted at the USGS Laboratory at the CAWSC in Sacramento, California. Determination of four THMs and five HAAs was then performed by Alexin Laboratories as described below.


For quality-assurance (QA) purposes, one blank and seven samples of standard reference material (SRM) were submitted to the laboratory for analysis. As described by USEPA Method 5710B, a freshly prepared solution of anhydrous 3,5-dihydroxy-benzoic acid (DHBA, 0.039 g/L) was made to test the precision of laboratory chlorine dosing, quenching, storage, and DBP analysis. According to this method, a 7-day reaction period should result in approximately 0.119 mg/L THM as chloroform with essentially no contribution from bromide-containing THMs. With the exception of the first batch of samples sent to Alexin Laboratories, each batch of samples submitted to the lab contained at least one DHBA SRM sample. 


Disinfection By-Product Concentrations in Finished Water


DBP analyses were performed on finished-water and formation-potential samples at Alexin Laboratories (ORELAP Certification ID# OR100013) in Tigard, Oregon. Analyses included four THMs (chloroform, Cl₃CH; bromoform, Br₃CH; bromodichloromethane, Cl₂BrCH; and dibromochloromethane, ClBr₂CH) and five HAAs (monochloroacetic acid, MCAA; dichloroacetic acid, DCAA; trichloroacetic acid, TCAA; bromoacetic acid, BrAA; and dibromoacetic acid, Br₂AA) following the same methods used for DWTP compliance monitoring—USEPA method 524.2 for THMs and method 6251B for HAAs. For this study, samples were analyzed within 2–14 days for THMs and within 5–7 days for HAAs.


In this report, the sum of individual THMs (THM4) and HAAs (HAA5) is commonly used because these metrics form the basis of the DBP drinking-water regulations. Benchmark quotients (BQs), the ratio of the concentration in finished water divided by the maximum contaminant level (MCL), were calculated to compare DBP concentrations to USEPA standards. For example, BQ values of 0.5 and 1.0 would indicate concentrations at one-half the MCL and at the MCL, respectively. Although BQs provide an indication of how close individual concentrations are to the standards, the actual standards are based on the annual running averages in quarterly sampling at the DWTP and within the distribution system. 


“Specific” THMFPs and HAAFPs (STHMFPs and SHAAFPs) values were calculated for filtered and unfiltered water by dividing formation potentials by sample DOC or TOC concentration, respectively, and are reported in milligrams of DBP (THMs or HAAs) per milligram of carbon. Some studies refer to these carbon-normalized formation potentials as DBP “yields” (Summers and others, 1996). This expression indicates the average reactivity of carbon in a water sample to form DBPs during chlorination.


Water-Column Chlorophyll-a


Water-column chlorophyll-a measurements were conducted at Oregon Health and Science University using a Turner 10-AU fluorometer according to the manufacturer’s procedure. A known amount of whole water was filtered through 0.7-µm GF/F filters. The filters were stored at -20°C for no more than 30 days prior to analysis. The filters were steeped in 5-mL of 90 percent acetone for 24 hours and stored at -20°C prior to analysis. The fluorometer was zeroed with 90 percent acetone, and standard curves were generated. Standard- and regular-sample fluorescence was measured before and after addition of 3 drops of 10 percent HCL. Values were “blank corrected” by subtracting the background fluorescence of the 90 percent acetone solution. Data were entered into the manufacturer’s equation to calculate chlorophyll-a concentrations.


Benthic Algal Chlorophyll-a and General Species Composition


In the laboratory, periphyton samples were homogenized in an electric blender and transferred to an 8-L churn splitter; subsamples were removed using a 5-mL pipette. Known aliquots were transferred onto 0.7-µm 45-mm GF/F filters under a mild vacuum. Each filter was folded into quarters, placed in a plastic petri dish, wrapped in aluminum foil, and frozen at -4°C until analysis. Filters were hand ground in 90 percent acetone to extract the chlorophyll-a pigment, and samples were analyzed using standard methods (fluorometry with acid correction) at the Oregon Water Science Center (ORWSC). A Certified chlorophyll-a standard solution from the USGS NWQL was used to generate standard curves to extrapolate sample concentrations. Samples also were analyzed for dominant species composition at the ORWSC using a Leica microscope and current taxonomic references.


Nutrients


Dissolved nutrient analyses for nitrate, nitrite, soluble reactive phosphate, ammonium, and silicate were performed at Oregon Health and Science University using a 5-channel 2008 model Astoria-Pacific Segmented Continuous Flow Injection Analyzer designed for spectrophotometric analysis of nutrients in freshwater. In order to attain low detection limits, the protocols followed the manufacturer’s recommendations (detailed in Whitledge and others, 1986), with adjustments for low-concentration detection as outlined in Sakamoto and others (1990). All measurements were quality controlled using commercially purchased standard reference samples approved by the USGS. Dissolved organic nitrogen, total dissolved nitrogen, total dissolved phosphorus, total nitrogen, and total phosphorus were analyzed following an alkaline persulfate digestion followed by the colorimetric procedures outlined above. TPN samples were collected and analyzed using USEPA method 440.0. Particles from a known volume of water were filtered onto 25-mm 0.7-µm GF/F filters, which were frozen at -4°C and shipped to the USGS NWQL for analysis. 


Treatability Experiments Using Jar Tests


During each of the four basin-wide sampling events, jar tests were conducted on CRW DWTP source water using a Phipps and Bird Stirrer Model 7790-400 following standard protocols to assess treatability, defined here as the percentage of the DOC and DBP precursor pool removed from solution by coagulation. Water was collected from the intake vault and composited in an 18-L Teflon™ churn splitter, and subsamples were dispensed into test jars. Jar tests were designed to simulate the typical treatment at the CRW DWTP to remove suspended particles, primarily the addition of aluminum sulfate (alum) and aluminum chlorhydrate (ACH). The amount of coagulant added was adjusted as needed to an “optimum dose,” defined as the amount of coagulant per liter required to reach the point of zero charge as determined by an in-line streaming current monitor. Thus, when TOC (DOC + TPC) increases, higher coagulant doses were typically applied. Because CRW chlorinates during coagulation, there is no opportunity within the drinking-water treatment train to isolate the effects of coagulation alone on DOC and DBP precursor removal, making laboratory jar tests a natural choice for helping address this question. In addition, up to 5 mg/L PAC may be added during treatment to control tastes and odors. This practice might also reduce DBPs in finished water, so PAC was included as one of the treatments.


Three treatments (two replicates each) were compared: (1) coagulants (alum and ACH) applied at optimum dose; (2) coagulants (alum and ACH) applied at optimum dose along with 5 mg/L PAC; and (3) a control that received no coagulation. The amount of coagulant required to remove the maximum amount of DOC was determined during the time of source-water collection by the in-line streaming current monitor at the plant.


The jar-test protocol loosely followed that described by Shin and others (2008). Briefly, 2-L samples were placed in each of six jars and mixed at 300 revolutions per minute (rpm) for 5 seconds. Coagulants and PAC were added, and samples were mixed at 150 rpm for 3 minutes to provide rapid mixing, followed by mixing at 25 rpm for 15 minutes to enable flocculation. Samples were allowed to settle for at least 15 minutes and then filtered through 0.7-µm GF/F filters into baked amber glass containers to generate subsamples for DOC, absorbance, fluorescence, and DBPFP. The pH of coagulated sample water was checked to confirm that coagulation did not lead to a substantial drop in pH (greater than 0.5 standard units). 


On each of the four sampling dates, to verify the coagulation rates used represented the optimal dose, an additional series of six jar tests were conducted with dosing rates ranging from 25 to 200 percent of the optimal dose. Results from those tests verified the coagulant dosages used were appropriate and attained maximum DOC removal (±0.1 mg/L, data not shown). Jar-test results were compared to information obtained by comparing source-water and finished‑water DOC concentrations and changes in optical properties. 


Continuous Real-Time Measurement of Streamflow, Field Parameters, and Fluorescent Dissolved Organic Matter


The existing monitoring network of five USGS streamgages and three continuous water-quality monitors (fig. 1) provided flow, water temperature, specific conductance, DO, pH, and turbidity data which provided information on river conditions during the study in near real-time. In addition, at two main-stem sites—Estacada and Oregon City—water-column chlorophyll-a fluorescence was also measured continuously. One additional temporary site, Clackamas River near Clackamas, USGS station 14211005, was established within the CRW DWTP intake vault to monitor the quality of the source water (actual pump water) every 30 minutes for water temperature, turbidity, and FDOM. 


FDOM measurements were made using two standard fluorometers: (1) a Wet Labs™ WETStar flow-through sensor deployed from April 4, 2010, to October 20, 2011, and (2) a Turner Designs™ “open-faced” Cyclops-7 sensor (see photograph 3) deployed from April 14, 2011, to January 31, 2012. In addition, three custom-built Cyclops-7 sensors were deployed along with the standard Cyclops-7 sensor (table 5). One of these sensors was designed to detect fluorescence around ex 270/em 340 nm, a signal associated with amino acid/protein-like (peak T) fluorescence and, in the Clackamas River, expected to indicate the presence of algal-derived DOM. Two additional custom sensors with relatively narrow band-passes centered around emissions 470 and 520 nm both at excitation 370 nm were used to calculate the FI that, based on prior studies, can indicate a shift between microbial or algal-derived and terrestrial-derived sources of DOM.


All four of the Turner Designs™ Cyclops-7 fluorometers were mounted on a Cyclops 6 data logger equipped with a wiper. Water temperature, turbidity, and specific conductance also were measured at the CRW DWTP intake using a Yellow Springs Instruments, Inc., OMS multi-probe sonde. All sensors were housed within the CRW intake vault. The sensors were raised and lowered into the water in a non-reflective black plastic basket using a pulley system. The basket was adjusted to be approximately 3 ft below the water surface, which was typically about 10-15 ft off the bottom.


The WET Labs™ WETStar FDOM sensor was deployed in an unfiltered flow-through configuration using a 12-volt submersible pump (see Saraceno and others, 2009). One drawback of this configuration is that fouling of the optics is not mitigated through the use of a wiper; therefore, optics were cleaned manually every 2-4 weeks with lens paper. Prior to each measurement, the pump flushed the sensor with approximately 2–3 L of water for 10 seconds. Then while pumping continued, the measurements were recorded for 30 seconds at 1 hertz to yield a set of burst data. The burst data were reduced to a single data point using descriptive statistics. Typical variation within a burst was less than 1 percent. The median of the last 20 samples within the burst was used in the final dataset. Unlike the WETStar™, the Turner Designs™ C6 was outfitted with a wiper that cleaned each sensor before each measurement. In addition, as with the WETStar™, optics were cleaned manually every 2-4 weeks with lens paper. Data reduction was performed using burst data acquired at a rate of one sample every 11 seconds. Real-time data were recorded using a Campbell Scientific™ model CR 1000 data logger (see photograph 4) programmed for the suite of sensors deployed at the site. The data logger provided the power-distribution controls (turning sensors on and off), time stamp, and internal logging of all sensor data at predetermined sampling frequencies. The data logger was interfaced with a cellular modem (Sierra Wireless™ RAVEN XT-V) to allow remote data acquisition and troubleshooting. The data logger transmitted data to the USGS server every 2–4 hours using a Campbell Scientific™ model COM220 telephone modem. This system provided data in near real-time, which was examined using the USGS Data Grapher program (U.S. Geological Survey, 2011).


Because of the diel demand on water consumption within the service district, pumps within the intake typically turn off at night, even though the sensors continued to operate. For this reason, pump records were used to extract FDOM data only when the pumps were operating (a threshold of 7 Mgal/d or about 11 ft³/s was used to indicate when the pumps were on). This assured the measurements of FDOM within the intake vault were representative of the river and not that of possible impoundment effects in the intake of the river and not influenced by potential impoundment effects in the vault.


Water temperature and suspended particles affect fluorescence measurements (Zepp and others, 2004; Lakowicz, 2006; Saraceno and others, 2009). To correct for these effects, experimentally-derived correction factors were applied using the approach outlined in Downing and others (2012). When available, temperature and turbidity data from the co-located YSI instrument were used to make the corrections; otherwise data were obtained from the Clackamas River at Oregon City site about 1.5 mi downstream. Although temperature and turbidity values sometimes differed between these two locations, the relations were linear and reasonably well-correlated for overlapping data (r = 0.94, n = 13,112 and r = 0.99, n = 15,929, respectively). Applying temperature and turbidity corrections to raw FDOM data resulted in about an additional 8–15 percent signal recovery for baseline periods and up to an additional 23 percent signal recovery for periods characterized by high-flow storm events. High turbidity (greater than 300 NTUs) during one January storm required the most corrections that increased FDOM values by as much as 55 percent. Temperature- and turbidity-corrected FDOM data were converted to quinine sulfate dihydrate equivalents on the basis of laboratory calibration tests for quality-control purposes (Downing and others, 2012; Pellerin and others, 2012). These data were converted to DOC concentration in milligrams per liter using the near-linear relation between discrete DOC concentrations and in-situ FDOM values. 


Because the effects of temperature and turbidity on the three custom sensors were not evaluated, corrections could not be applied to these data. Data from the two Peak C custom sensors were converted into quinine sulfate dihydrate equivalents using laboratory-based calibration data for each sensor, and the ratio of these sensors was calculated as Sensor FI-A/Sensor FI-B and is referred to as FI_in-situ. Data from the Cyclops-7 Peak T sensor was also not adjusted for turbidity or temperature, but data were used in terms of changes in intensity relative to other measurements such as DOC and FDOM.


First posted February 11, 2013