USGS Home
Contact USGS
Search USGS

U.S. GEOLOGICAL SURVEY
Scientific Investigations Report 2007–5117

Back to Table of Contents

Description of Datasets and Methods

In 1987, the Klamath Tribes began a water-quality monitoring program in Upper Klamath and Agency Lakes and contributing tributaries. The goal of this sampling program is to develop models to predict and prevent nutrient loading into Upper Klamath Lake, with the ultimate goal of “…reducing toxic algal effects…and returning endangered species to harvestable levels for all residents of the basin” (The Klamath Tribes, 2006). Among other sampling efforts that make up the larger program, depth-profile measurements of specific conductance, pH, water temperature, and dissolved-oxygen concentration were made at established sites throughout Upper Klamath and Agency Lakes every two weeks from May through October. Additionally, composite samples were collected for the analysis of nutrient, chlorophyll‑a, and phaeophytin concentrations. During 1987–89, these samples were collected with a grab sampler and the sampling was not as frequent. Beginning in 1990, however, depth-integrated samples were collected and the sampling frequency was more consistent. Therefore, for this analysis, data prior to 1990 were not included.

Field-collection, analytical, and quality-assurance methods are described in The Klamath Tribes Quality Assurance Project Plan (The Klamath Tribes, 2006). For the most part, only small changes have been made to these techniques during the 17-year period (Kris Fischer, The Klamath Tribes, written commun., May 2007). The same laboratory has been used for the analyses, the sample-collection process has remained unchanged, and most of the sites have been maintained. The depth-profile measurements were made with multiparameter water-quality instruments, and typically were taken at the surface, at 0.5 m, and then at 1‑m intervals descending the water column, with the last measurement taken at 0.1 m off the bottom. Depth-integrated samples, which were collected using a tube sampler lowered to within 6 inches of the sediment in an effort to represent the full water column, were composited into a churn splitter and then subsampled and preserved for the desired analyses. The types of quality-assurance samples collected by The Klamath Tribes are consistent with USGS protocols and include at least one trip blank, one equipment blank, one replicate sample, and one spike or reference sample per sampling event. These data are evaluated on a yearly basis with adjustments to sampling or analytical protocols made as needed.

Because this analysis involves examining seasonal and year-to-year variability, the desired dataset would include longer-term, regularly and consistently sampled sites. Therefore, sites with less than 10 years of data were not included. The 10 sites that met these criteria were Coon Point, Shoalwater Bay, Wocus Bay, Midnorth, Midlake, Eagle Ridge, North Buck Island, Pelican Marina, Agency Lake North, and Agency Lake South (fig. 1). For some analyses in this report, these sites are grouped together into areas—Agency Lake area (Agency Lake North and Agency Lake South), bay areas (Coon Point, Shoalwater Bay, and Wocus Bay), trench area (Eagle Ridge), open water areas (Midlake and Midnorth), and outflow areas (North Buck Island and Pelican Marina). Since 1990, most sites were sampled generally every two weeks from sometime in May through October (table 1). In some years, additional samples were collected during the rest of the year but much less frequently and not at all sites. In 1990, however, four sites were sampled almost weekly, whereas the other four sites were visited roughly monthly.

Lake level is recorded by the USGS at 30-minute intervals at three gaged sites around the lake (Rocky Point, Rattlesnake Point, and near the city of Klamath Falls). From these values a daily average is determined for each site and then these daily averages are weighted to determine a spatially averaged lake level for the entire lake. The spatially averaged lake level reported in the USGS Water Resources Data for Oregon 1974–2006 was used in this report (fig. 2). Over the 17 years discussed in this report, lake level was most variable during the first 5 years—the years discussed in the first report (Wood and others, 1996). There has been little variation in the maximum lake levels in the last 12 years, however, and little variation in the minimum lake levels in the last 5 years.

Air temperature and wind speed were the only climatic variables with long-term consistent datasets for 1990–2006. Air-temperature and wind-speed values used in this report were recorded at the Klamath Falls airport and were provided by the Oregon Climate Service (more recent data) and the National Climatic Data Center (an archive of longer-term data). Agrimet, an agricultural weather network maintained by the Bureau of Reclamation, also has a weather station (AGKO) on the banks of Agency Lake. Wind data were available for this location from 2000 to present; however, the limited time period covered by this dataset hampered the use of these data for this analysis. Besides air temperature and wind speed, the use of climatic data in this analysis was limited because the climatic datasets for the area for 1990–2006 were sporadic and not necessarily representative of conditions on the lake itself.

Strengths and Limitations of Datasets

The primary limitation of the Klamath Tribes water-quality dataset for this report is that the sampling program was not designed to address the questions being asked in this report. The spatial and temporal coverage of the Klamath Tribes sampling program meets the needs of their program, but the two-week sampling interval is not ideal for assessing water-quality conditions and potentially biologically significant events over the course of the algal bloom growth and decline. To truly capture all of the variation in water-quality conditions during this period, continuous monitoring of specific conductance, pH, water temperature, and dissolved-oxygen concentration along with more frequently collected water-column samples for the analysis of nutrient and chlorophyll‑a concentrations would be needed.

For the purpose of this report, the main and overriding strength of the Klamath Tribes water-quality dataset is the period of record and its consistency. It is the only 17-year dataset with a consistent group of sites and little change in sampling and analytical techniques available for Upper Klamath Lake. For these reasons, it is an adequate dataset to be used for looking for long-term relations between water-quality variables, climatic factors, and lake level. These same type of statements about the strengths and limitations of the water-quality dataset can be made about the climatic datasets. The climatic factors considered in this report were limited to air temperature and wind speed because these were the only datasets available for 1990–2006. The wind-speed data collected at the Klamath Falls airport were not ideal for this analysis because they were not deemed to be very representative of conditions on the lake itself, but this is discussed further in the “Wind Speed” section of this report.

Statistical Methods

When it is suspected that a dataset is not from a normally distributed population, as was assumed for this analysis, nonparametric techniques may be more appropriate for examining correlations. Nonparametric statistics use rankings of the data rather than the actual values. A Spearman test, which calculates a Spearman rho correlation coefficient (ρ) and a probability level (p), was used to examine correlations between datasets in this analysis. Spearman’s ρ has values from -1 to +1 for negative and positive associations and values close to 0 for little or no association. In this report, a correlation is considered “statistically significant” when the probability of two variables appear to be correlated when, in fact, they are not, is less than 0.05 (a Type 1 error, greater than 95 percent confidence level). Furthermore, a correlation is considered “strong” when the correlation is significant and the Spearman’s ρ value is greater than 0.5 (positive or negative).

For the multivariable investigations presented in this report, multiple linear regressions were calculated and then evaluated with the Akaike Information Criterion (AIC_c) to determine the “best fitting” models. This is discussed in greater detail in the “Information-Theoretic Approach to Multivariable Analysis of Water-Quality Conditions” section of this report. All statistical tests were run using P‑STAT, version 2.21, rev. 8 (P‑STAT, Inc., 1990).

Back to Table of Contents