Methods 


Water Quality Data Collection


Water column samples were collected from different sites and at various sampling intervals between 2007 and 2009 to create overlap between water quality data collection and the U.S. Geological Survey’s juvenile sucker sampling effort; resource constraints also have limited sampling in some years. In 2007, water samples were collected at six sites on Upper Klamath Lake that were co-located with continuous water-quality monitoring sites according to established collection and quality control protocols (U.S. Geological Survey, variously dated) for the analysis of cyanotoxins (microcystins and cylindrospermopsins), chlorophyll a, dissolved nutrients (orthophosphate [DIP], ammonia and nitrite plus nitrate [DIN]), total phosphorus (TP), and total nitrogen (TN). Samples for cyanotoxin analysis were collected monthly (July 9 and 10, July 31 and August 1, September 4 and 5, and October 16 and 17) at sites MDN, WMR, EPT, MDT, RPT, and HDB (table 1; fig. 1); nutrient and chlorophyll a samples were collected weekly from these sites from mid-May to mid-November. In 2008, the sampling frequency for cyanotoxin analysis was increased from monthly to biweekly and occurred between June 30 and September 22 at fewer sites—MDN, WMR, MDL, and NBI—than in the previous year. Also in 2008, samples for nutrient and chlorophyll a analyses were collected weekly from mid-May to early October, but at only two of the four cyanotoxin sample sites, MDN and WMR. In 2009, weekly sampling for nutrient and chlorophyll a analyses began in mid-May (as in the previous 2 years) and occurred concurrently with cyanotoxin sample collection (also weekly) between June 16 and September 14 (weekly sampling for nutrient and chlorophyll a analyses continued through September 28.) All samples were collected from five sites in 2009—MDN, WMR, EPT, MDT, and RPT (table 1; fig. 1). Additional samples were collected between June 16 and August 31, 2009, for analyses of total particulate carbon (TPC), total particulate phosphorus (TPP), total particulate nitrogen (TPN), and particulate inorganic phosphorus (PIP). 


Water samples for chlorophyll a, TN, and TP analyses were integrated by collecting lake water in two 1-L vented bottles held in a weighted cage that was dropped at a constant rate from the surface to 0.5 m from the sediment at shallow sites (< 10.5 m), and to 10 m from the surface at deep sites (> 10.5 m). Samples were mixed in a churn splitter and divided into separate fractions for each type of analysis. Samples analyzed for dissolved nutrients (ammonia, orthophosphate, and nitrite plus nitrate) were collected at discrete depths to identify sources (benthic or water column) of these nutrients for the long-term monitoring project. To do this, a hose was lowered to one-half the water column depth at shallow sites and at two points, one-quarter and three-quarters the water column depth, at deep sites. Lake water was drawn into the hose by a peristaltic pump, passed through a 0.45-µm capsule filter attached to the hose end, and collected into sample bottles. Total particulate carbon and nutrient (TPC, TPN, and TPP) samples were also collected at discrete depths and in the same manner as the dissolved nutrient samples, but with the capsule filter removed. Water samples for particulate carbon and nutrient analyses were filtered at the U.S. Geological Survey Klamath Falls Field Station on 25-mm (precumbusted for TPC and TPN) or 47-mm (for TPP and PIP), 0.7-µm pore size, glass microfiber filters (GF/F, Whatman, Inc., Piscataway, N.J.) and shipped overnight to the University of Maryland Chesapeake Bay Laboratory (CBL), Solomons, Maryland. Once received, samples for TPC and TPN analyses were processed according to EPA Method 440.0 (U.S. Environmental Protection Agency, 1997), and TPP concentrations were analyzed according to Aspila and others (1976). Samples for the measurement of PIP were extracted in an acidic medium and analyzed according to the method of Aspila and others (1976). Further processing and analyses of chlorophyll a, total nutrients, and dissolved nutrients are described in Kannarr and others (2010). Nutrient (total and dissolved) and chlorophyll a samples collected in 2009 were processed as in 2008. 


From June 16 to August 3, 2009, and during all samplings in 2007 and 2008, cyanotoxin samples were collected by depth integration (described above), transferred from the churn splitter to 1-L amber, high density polyethylene (HDPE) bottles, and immediately stored on ice. Samples collected on August 10, 2009, and thereafter were collected by depth integration at the shallower sites, MDN, WMR, and RPT, and at two points, one-quarter and three-quarters the water column depth, at the deeper sites, MDT and EPT, using a hose and peristaltic pump (the same method as for particulate nutrient sample collection). Samples collected in all years were shipped overnight to the U.S. Geological Survey, Columbia Environmental Research Center (CERC), Columbia, Mo. Once received, samples were kept on ice in a walk-in cooler and processed within 24 hours. Samples for quality-assurance determination also were collected, and the results of this analysis are presented in appendix A. 


Continuous water-quality monitors (YSI, Inc., Yellow Springs, Ohio) were deployed each year at sites shown in table 1 to measure water temperature, dissolved oxygen, pH, and specific conductance according to Lindenberg and others (2009). Monitors were positioned vertically 1 m from the lake bottom at all sites, except site WMR, and an additional sonde was placed on the same mooring 1 m from the lake surface at the deeper sites, MDN, EPT, and MDT. Water column stability was calculated as the median relative thermal resistance to mixing (RTRM) using data collected by the upper and lower sonde at site MDN. RTRM was determined by comparing the water column density gradient (based on the temperature difference between the upper and lower sondes) to the density difference between 4 and 5°C (Jones and Welch, 1990; Kann and Welch, 2005). In 2009, only the lower sonde (1 m from the bottom) was deployed at site EPT. The depth at site WMR is less than 2 m, so the sonde at this site was placed horizontally at one-half the water column depth. Calibration, maintenance, and data handling were performed following standard procedures outlined in Wagner and others (2006). Meteorological data were collected from the floating station at site MDL as described in Lindenberg and others (2009) and Kannarr and others (2010). 


Cyanotoxin Analysis


Water samples were fractionated (filtered) at CERC to determine the relative contributions of different phytoplankton size classes to the total cyanotoxin concentration in each sample. Samples were first filtered with a 63 µm sieve to isolate large cyanobacterial filaments and colonies. The small fraction, 1.5–63 µm, representing smaller forms, was collected by filtering the 63-µm fraction filtrate onto pre-weighed ProWeigh^TM glass fiber filters (Environmental Express, Mt. Pleasant, S.C.). The filtrate from this step was considered the dissolved fraction and retained for determination of dissolved microcystin (< 1.5 µm) concentrations. Prior to extraction, both the large and small particulate fractions were freeze-dried and weighed. In 2007, samples were extracted three times with 5 mL of 50 percent aqueous methanol and 0.1 percent trifluoroacetic acid for 5 minutes by ultrasonication. The extracts were centrifuged at 10,000 revolutions per minute (rpm) for 15 minutes, and the supernatant was filtered through a 0.45-µm nylon syringe filter or a 0.45-µm UniPrep^TM syringeless glass microfiber (GMF) filter (Whatman Inc., Piscataway, N.J.). In 2008 and 2009, freeze-dried biomass (containing the large particulate fraction) and filters (containing the small particulate fraction) were extracted using a Dionex Accelerated Solvent Extraction System, ASE 200 (Dionex Corporation, Sunnyvale, Calif.) in 5-mL sample cells and using the parameters reported in Aranda-Rodriguez and others (2005). The cells were prepared by tamping a glass fiber filter (Fisher G2; 1 µm, cut to size with a cork borer) into the exit end of each cell and filling it half full of glass beads (Kimble Kimax KG-33; 3 mm) for filter extraction or with Hydromatrix for extraction of raw water filtrate. 


The samples were dried under nitrogen to remove methanol and resuspended in 10 mL deionized water. Diluted (1/100) extracts were analyzed in 96-well microtiter plates for determination of microcystin and cylindrospermopsin concentrations using congener-independent enzyme-linked immunosorbent assays (ELISA; kit 520011, Abraxis, LLC, Warminster, Pa.) following the manufacturer’s protocol. Absorbances of the samples at 450 nm were determined within 15 minutes after addition of the final solution. Concentrations were determined from a regression of the mean absorbance of calibration standards, and were analyzed in duplicate (duplicate blanks were included) at the same time as the samples. Total microcystin or cylindrospermopsin concentrations were determined by summing the particulate and dissolved concentrations (Graham and Jones, 2007). Concentrations in all size fractions were calculated volumetrically to facilitate comparisons and to calculate total microcystin concentrations. However, because the route of exposure to affected suckers in Upper Klamath Lake is likely through ingestion, concentrations of cell associated microcystins also were expressed as mass per dry weight of suspended solids. Dissolved fractions of both toxins were not analyzed in 2007, and cylindrospermopsin concentrations were not measured in 2009. Therefore, total concentrations described for 2007 may be underestimates. Samples collected in 2007 and 2008 contained cylindrospermopsins near or less than the detection limit and are not described further in this report. The detection limit for the microcystin assay is 0.10 ppb (µg/L), and the detection limit for the cylindrospermopsin assay is 0.05 ppb. Values less than 0.1 µg/L that are not censored (as < MQL) are considered detections, although they appear to be less than the detection limit since the aqueous concentrations were calculated from the total extracted biomass and sample volumes. Cyanotoxin data are available in appendix B.


Data Analysis


Maximum-likelihood estimation (MLE) of summary statistics for datasets with censored values have been shown to produce estimates with large bias and poor precision for small (n < 15) sample sizes (Gleit, 1985). Therefore, for this reason and for simplicity, values equal to or less than the detection limit were considered equal to the detection limit for determination of median values. This method also was used for statistical analysis, because only 13 percent of the samples (all years combined) collected between July 7 and September 14 (the sampling period common to all study years, hereafter referred to as July–September) contained dissolved microcystins less than the detection limit, and no samples contained microcystins in the large particulate fraction at concentrations less than the detection limit during that time; small particulate (cell associated) microcystin data were not used in statistical analysis. 


Changes in cyanotoxin, nutrient, and chlorophyll a concentrations over time were compared by calculating the daily median values of these variables measured in water samples from all sites on each sample date. Analyses of water temperature, pH, and dissolved oxygen concentration or percent saturation were based on the median value of 24 hourly measurements made at all sites on each sample date. Interannual comparisons of cyanotoxin data were based on daily median values of concentrations determined in samples from all sites on each sample date between July and September. Inter- and intra-annual comparisons of cyanotoxin concentrations measured between sites were based on daily median values at each site between July and September.


Correlations between microcystin concentrations in the dissolved and large particulate (> 63 µm) fractions and environmental variables were determined using Spearman rank order correlation analysis; correlations were considered significant at p < 0.05 and near significance at p < 0.1. Due to the low concentration (maximum values between 0.03 and 0.07 µg/L), relative to those in the dissolved and large particulate fractions, microcystins within the small particulate (1.5–63 µm) fraction and cylindrospermopsins were not included in comparisons with environmental variables. All microcystin, chlorophyll a, and nutrient data collected at sites MDN and WMR, the only sites commonly sampled from 2007 to 2009, were used in interannual correlations. Interannual correlations with data collected from continuous monitors (water temperature, pH, and dissolved oxygen concentration or percent saturation) were based on the median of 24 hourly measurements collected on each sample date. For interannual correlations, all data collected in 2007 and 2008 were included, but only data collected on corresponding sample dates in 2009 were used in order to weight the years evenly. Intra-annual correlations were based on median values of data collected at all sites on each sample date in 2009. Correlations with wind speed were based on the median of 24 hourly measurements made on each sample date at meteorological station MDL.


First posted May 30, 2012