USGS Home
Contact USGS
Search USGS

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

Back to Table of Contents

Characterization of Years

To help in evaluating relations in water-quality conditions between the years, three factors were used to rank the years in relation to each other—end-of-month lake level, degree-days, and timing of the beginning of the AFA bloom.

Lake Level

To explore how each year from 1990 to 2006 compared to each other and to the historical record, the end-of-month lake levels were compared (fig. 3). Historically, the gage height at USGS site 11507000, at the southeastern end of the lake near the outflow, was used as a measure of lake level, so the gage height is used for the entire 1922–2006 period for this comparison, rather than the calculated spatially averaged lake level, which was not developed until 1974. It is obvious that 1992 (May–August) and 1994 (July and August) are outliers for the 1990–2006 dataset, and that lake levels in these years were low not only for this dataset but also for the historical record. It is difficult to distinguish the ranking of the rest of the 1990–2006 years from figure 3; however, it can be observed that a large overall range in lake level is represented in the 1990–2006 dataset with most years falling within the middle 50 percent of the historical data. In fact, the only years falling below the 25^th percentile of the historical data are 1992, 1994, 1991, and 2002, respectively. A few years were above the 75^th percentile for the historical data, but there is no consistent pattern to these years.

A statistically based convention was used to assign a ranking to each year based on lake level. Percentiles were assigned to the end-of-month lake levels for each year by month based on the corresponding distribution of the end-of-month lake levels for the historical 1922–2006 dataset (table 2). For each year, these monthly percentiles were summed for May through August. These sums were then ordered to develop an overall lake-level ranking for each year. From this ranking, once again the years 1992, 1994, and 1991 are recognized to be notably low lake-level years.

Degree-Days

Recognizing that the growth of the AFA bloom is related to the water-quality conditions in the lake, and that climatic conditions also affect the growth of the bloom, climatic factors also were considered in making year-to-year comparisons. Factors such as air temperature and cloud cover are expected to have an effect on when the bloom begins to grow and the rate at which it develops. One way of trying to quantify these conditions is by comparing the warmth of the springtime conditions in each year. Degree-days can be used as a measure of heating or cooling for a system. For this analysis, degree-days were calculated as the difference between the mean temperature (in degrees Celsius [°C]) for a given day and a reference temperature (0°C in this case). The mean temperature for the day was calculated as the average of the minimum and maximum air temperature for each day at the Klamath Falls airport. These degree-days were then summed for each year from April 1 to May 15. This summation then provides a way to compare how warm the spring was for each year in relation to other springs and another way to rank the 1990–2006 dataset (table 3).

Timing of Bloom

Determining the onset of the AFA bloom can be tricky. Individual chlorophyll-a measurements ranged from less than 2 µg/L, the approximate limit of detection, to more than 1,000 µg/L. The median chlorophyll-a concentration for samples collected prior to May 15 of each year was 6 µg/L with a maximum concentration of 16 µg/L. Once the bloom begins to develop each year, individual values rise quickly, and concentrations at any time vary widely throughout the lake. When more than one individual measurement was made in a day, the median of the individual measurements was selected to represent the overall condition of the lake on that date. The date the median concentration exceeded 20 µg/L was taken to represent the beginning of each year’s AFA bloom. Problems arose when the onset of the bloom was determined based on the first sampling date with a median chlorophyll‑a concentration greater than 20 µg/L. Examining the concentrations measured in 1997 illustrates this point (fig. 4). The median chlorophyll‑a concentration measured on May 20 was 18 µg/L, closely approaching the defined onset of the bloom. The next sampling day was not until June 16, almost a month later, when the bloom was near its peak with a median chlorophyll‑a concentration of 197 µg/L, the highest median concentration measured that year. By the method described above, the onset of the bloom would be defined as June 16 (or Julian day¹ 167), clearly much later than the “true” onset of the bloom.

¹Julian days are a way of expressing dates, such that days in the year are numbered consecutively from 1 to 364 or 365. It is easier to make comparisons between years when using Julian days.

Therefore, the method used to define the onset of the bloom for this analysis was to interpolate between the sampling dates to determine when the median chlorophyll‑a concentration would have exceeded 20 µg/L. The inherent problem with this method lies in the assumption that there is a linear relation between the sampled concentrations. Considering the pitfalls of this assumption, the data were examined for potential problems. For many of the years, one of the two samplings with chlorophyll‑a concentrations bracketing the 20 µg/L value had concentrations fairly close to 20 µg/L; therefore the error associated with the interpolation was minimized. For a few years, there was a larger range between the chlorophyll‑a concentrations but the samplings were closer together, again minimizing the error associated with the interpolation. So, for 1997, the onset of the bloom was defined as May 20 or Julian day 140.5 (fig. 4). Based on the Julian days assigned as the onset of the bloom from the interpolation method, a ranking of the years for the timing of the bloom was developed (table 4).

The onset of the bloom for the majority of the years occurred during the last week of May and the first week of June, with the earliest bloom starting in mid-May in 1992 and the latest bloom starting in mid-June in 2006. One year worth noting is 1991. The chlorophyll‑a concentrations began to increase in early June with a median concentration of 26 µg/L, but by the next sampling date on June 19, the bloom had clearly stalled (median concentration of 13 µg/L). By the next sampling date (July 1), however, the bloom was in full production with a median concentration of 120 µg/L. The second time that the chlorophyll‑a concentration was interpolated to exceed 20 µg/L was used as the onset of the bloom (Julian day 170.8).

Wind Speed

A ranking of the years based on median wind speed during the July–August period also was considered a way to characterize climatic conditions for the lake. As Wood and others (2006) discussed, it is the peak wind speeds of the day that are of potential concern in relation to water-quality conditions. Wind speeds throughout the day generally are low with brief periods of peak conditions. In an effort to examine these higher wind speeds and minimize the effects of the low wind speeds that dominate the dataset, only wind speeds that exceeded the median wind speed for each month as determined from the entire 17-year dataset were considered. The median wind speed of this upper half of the data was then reported for July and August (table 5).

Comparison of Rankings

Some interesting observations can be made when comparing the rankings resulting from these three methods of characterizing the years (table 6). The year 1992 is notable because it is at the top of all three rankings—the lowest overall lake level, the warmest spring, and the earliest onset of the bloom. Similar in lake level, 1991 has already been shown to be an unusual year. With a lower lake level and cooler spring, the bloom seemed to stall as it was ramping up, resulting in a later onset of the bloom. In 2006, lake level and degree-day rankings were in the middle of the grouping; however, the bloom had the latest onset of the 17 years. In contrast, 1997, which had a higher lake level but a warm spring, had the second earliest onset of the bloom. By examining the rankings of these variables, it would seem that a warmer spring may have a greater influence on the timing of the bloom and therefore the related poor water-quality conditions. It should be noted, however, that the pattern observed in 2006 hints that there may be more to the story.

Back to Table of Contents