Relation Between Selected Water-Quality Variables, Climatic Factors, and Lake Levels in Upper Klamath and Agency Lakes, Oregon, 1990

Scientific Investigations Report 2007–5117

Hypotheses Relating Water Quality to Lake Level and Climate

The hypotheses relating selected water-quality variables to lake level and climatic variables examined in Wood and others (1996) served as a starting point for this analysis, to see whether relations observed or not observed within a 5‑year dataset had changed when examining a 17-year dataset. The same caveats that applied for these hypotheses in the previous analysis apply to this report. The hypotheses are purposely stated broadly to allow exploration of the data. The direction of the relation is stated because it was presupposed that water quality can be degraded at lower lake levels, but the specific form of the relation (linear, bimodal, and so forth) is not specified. With this approach, the data, rather than preconceived ideas, influence the conclusions.

The primary focus of this analysis was to establish the existence or nonexistence of relations between certain water-quality variables and lake level. The existence of a relation, however, does not imply causality. If a relation was detected, then a specific mechanism could be proposed and tested to establish causality. In the next section addressing information-theoretic approaches to the data, previous work in the basin and information emerging from the hypothesis testing will be used to guide the selection of multiple variables to examine separately and together for their ability to explain the variance observed in the data.

Chlorophyll-a

The temporal patterns observed in chlorophyll‑a concentrations in Upper Klamath Lake during May–October differed greatly between years (fig. 7). Certain similarities can be noted in the progression of the chlorophyll‑a concentrations, however, that illustrate the onset, growth, and decline of the AFA bloom. For each year, the bloom began sometime between mid-May and mid-June, signaled by increases in chlorophyll‑a concentration. The chlorophyll‑a concentrations generally went through a downward trend sometime in July, with a great deal of variation between the years in how much of a decline this was and whether concentrations increased again or not. Many years experienced a second period of bloom growth between August and September (1991, 2000, and 2005, for example), although some years did not (1993 and 1994, for example). The highest median chlorophyll‑a concentrations were observed during periods of bloom growth but occurred at very different time periods, June 1992 and October 1996. The warm autumn in 1996 most likely contributed to the magnitude of the second AFA bloom that year. The magnitude of the first spike in chlorophyll‑a concentration does not seem to be an indicator of how long the bloom will persist. Chlorophyll‑a concentration may decrease and the bloom may not persist as in 1992, or there may be sustained growth throughout the season as in 1995 and 1997.

There is no obvious pattern in the yearly distribution of chlorophyll‑a concentrations when the years are arranged from lowest to highest lake level. When the distribution of chlorophyll‑a concentrations for each year was considered by month in relation to the lake-level ranking, Wood and others (1996) did, however, observe a potential pattern in June—the median, 25^th-, and 75^th- percentile concentrations were higher in 1992 and 1994 than in the other 3 years (1990, 1991, 1993). This pattern did not continue for the July-September distributions. In the larger 17-year dataset (fig. 8), this pattern did not seem to hold up and this hypothesis was not supported.

Further exploration of the potential relation between chlorophyll‑a and lake level during June by Wood and others (1996) led to the development of the following hypothesis:

This hypothesis was analyzed for the larger 1990–2006 dataset by plotting the median June chlorophyll‑a concentrations against the median June lake level in Upper Klamath Lake (fig. 9). No significant correlation was observed; however, 1992 stood out as a year with a much lower median June lake level than any of the other years. The median chlorophyll‑a concentration for this year was the highest of the 17 years. The median June lake level for the remainder of the years, however, fell between 4,141.5 and 4,143.5 feet above sea level. This illustrates the fact that this analysis is performed on the observational data available. Perhaps the limited fluctuation in lake levels limits the variations in bloom dynamics that would be observed over a wider range of lake levels.

Although a relation between chlorophyll‑a concentration and lake level could not be determined, Wood and others (1996) had noted an apparent progression in June chlorophyll‑a concentrations as the number of degree-days increased. This was stated as:

This relation is explored for the 1990–2006 dataset in figure 9. A significant correlation was determined between the median June chlorophyll‑a concentration and the cumulative degree-days from April 1 to May 15 (Spearman’s ρ = 0.61, p < 0.009). Intuitively, it makes sense that a warmer spring (higher cumulative degree-days) would result in increased growth of the algae, and therefore, higher chlorophyll‑a concentrations, in June.

As a natural progression, not only the magnitude of the bloom in June, but also the timing of the bloom onset, which occurred between mid-May and mid-June, was explored in relation to lake level and cumulative degree-days (fig. 10). Wood and others (1996) had observed an apparent trend toward later bloom at higher lake levels when looking at the 1990–94 dataset and therefore developed the following hypothesis:

When this analysis was expanded to the 1990–2006 dataset, however, the trend is not apparent. In fact, correlation statistics showed no significant correlation between lake level and the onset of the bloom; therefore, this hypothesis was not supported for the 1990–2006 dataset.

The timing of the bloom also was explored in relation to the cumulative degree-days (Wood and others, 1996), stated as:

The data support this hypothesis (fig. 10). A significant negative correlation was observed between the onset of the bloom and the cumulative degree-days between April 1 and May 15 (Spearman’s ρ = ‑0.65, p < 0.004). The warmest spring (highest cumulative degree-days) occurred in 1992, which also had the earliest bloom (earliest Julian day of bloom onset). In contrast, 2003 had the coolest spring temperatures and had a later onset of the AFA bloom. The year 2006 is interesting because it does not “exactly” fit this pattern. The lake-level and degree-day rankings for 2006 (table 6) fall in the middle of the distribution for the 17 years, but 2006 experienced the latest bloom onset of the 17 years. This modification to the pattern in 2006 illustrates that although the onset of the bloom is related to the warmth of the spring, there are more factors to consider.

During the day, pH fluctuates and typically reaches a maximum value late in the afternoon, whereas over a season, the highest frequency of potentially detrimental maximum pH values occurred in June and July (table 9). To examine how lake level may affect pH values, Wood and others (1996) proposed the following hypothesis:

In Wood and others (1996), 9.5 was the pH value used to delimit poor water-quality conditions; however, Reiser and others (2000) have proposed that pH values greater than 9.7 are detrimental to suckers. Therefore, this hypothesis was updated for this report to state 9.7 rather than 9.5.

As discussed previously, because the growth of the bloom results in increasing pH values, the peak pH values tend to correspond to the peak chlorophyll‑a concentrations. Wood and others (1996) examined the relation that the onset of the bloom had on when the peak pH values were measured. To explore this relation for the 1990–2006 dataset, the maximum observed pH values recorded at each visit to Wocus Bay, a site that frequently had pH values greater than 9.7 (table 10), were plotted for each year (fig. 11). The peak pH values tended to occur sometime during mid-June to mid-July (except for 1990 and 2006, both having peak pH values much later in the year). These peak pH values seem to occur roughly 4–6 weeks after the onset of the bloom, which always occurred sometime in mid-May to mid-June. For Wocus Bay, every year except 1998 had peak pH values greater than 9.7, signifying poor water-quality conditions even though all ranges of lake level, degree-days, and timing of the bloom onset are represented. Wood and others (1996) noted that although the peaks in pH and chlorophyll‑a concentrations seem to be related, there is not necessarily a correspondence between the strength of a bloom and the coincident pH.

The yearly distribution of pH values for Upper Klamath and Agency Lakes was plotted by month in order of increasing lake level in an attempt to analyze the hypothesis relating pH and lake level (fig. 12). These distributions represent all pH values recorded at all depths for all sites, and thus are intended to represent the overall condition of each of the lakes. No obvious pattern with lake level is noticeable. The June and September distributions seem to be more variable than the July and August distributions. Perhaps this is related to the greater variation in pH during times of bloom growth (June) or decline (September), whereas the pH values may be less variable during maintenance of the bloom. There were sometimes large differences between the Upper Klamath Lake and Agency Lake distributions. Those for 2006 are extreme—the pH values measured in Upper Klamath Lake were much lower than those measured in Agency Lake in June, they were similar in July, and then flip-flopped in August and September with those in Upper Klamath Lake much larger than those in Agency Lake.

A few interesting patterns appear when the maximum observed pH values are examined at the site level rather than as lakewide distributions (fig. 13). June was characterized by higher lake levels and the pH values seem to be widely distributed, with some greater than 9.7. In July, however, the lake level was a little lower than in June and the majority of pH values at each site were greater than 9.7. In August and September, the pH values were again distributed across the pH range with no apparent relation to lake level. When correlation statistics were evaluated on the entire dataset for relations between maximum observed pH values and lake level, a significant (p < 0.0001) but very weak (Spearman’s ρ = ‑0.13) negative correlation was observed. When June and July were analyzed individually, however, significant (p < 0.0001) but weak (Spearman’s ρ values between 0.3 and 0.4) correlations were observed, but they were opposing—June was negative and July was positive (fig. 13). In June, if any pattern can be observed, it appears that higher pH values coincided with lower lake levels, as the hypothesis states. In July, however, the pattern seems to show higher pH values at higher lake levels. These weak correlations and changing patterns hint that there is not an easily defined relation between pH and lake level; there are more likely other factors, like bloom dynamics, affecting this relation.

Because June is a critical month in terms of the first bloom, relations between pH values and lake level, cumulative degree-days, and the timing of the bloom onset were further explored. Wood and others (1996), finding that the more general pH hypothesis was not supported by the data, developed a narrower hypothesis.

The pH data from 1990-94 were observed to support this hypothesis, which was directly related to the fact that the chlorophyll‑a data supported the hypothesis related to the timing of the bloom (see page 22). For the larger dataset, however, neither of these hypotheses is fully supported—the correlations were significant but very weak.

When the years are ranked according to cumulative degree-days and the timing of the bloom onset instead of by lake level (fig. 14), the distribution of pH values shows a distinction between the last 6 or so years in each ranking and the rest of the time period. In general, the six years with the latest bloom onset and some of the coolest spring temperatures—2006, 1991, 2003, 1999, 1993, and 2005—had lower median pH values, whereas the distributions for the other years were fairly similar. This separation in the groups is more of a symptom of the later occurrence of the bloom than evidence of a correlation. The peak pH values typically follow about 4–6 weeks after the onset of the bloom. The earliest bloom for these 6 years did not begin until May 31; therefore, the pH values in June had not reached their peak yet, but were still increasing as the bloom was growing. Looking at the distributions for these years in July (fig. 12) shows that by July the pH values for these years was similar to the other years.

The 6 years 1990, 2002, 2004, 1995, 2005, and 1993 all had bloom onset dates of May 29-31. Comparing the pH distributions for these six years illustrates that there is no clear pattern with relation to lake level, cumulative degree-days, or bloom onset (fig. 14). The median June pH distributions for 1990, 2002, 2004, and 2005 were very similar, whereas 1995 had a narrower distribution and a higher pH median and 1993 had a broader range of pH values and lower median pH. By lake level, the years are split into two groups—2002, 2004, and 1990 had lower lake levels, and 2005, 1993, and 1995 had higher lake levels. Therefore, lake level alone does not explain the differences in pH values and distributions. Likewise, the warmth of the spring does not provide a clear pattern as the years varied across the range of cumulative degree-days covered. The more likely explanation is that the interrelations between these and other factors act together to control pH conditions in the lake.

Dissolved Oxygen

During the day, the amount of oxygen dissolved in the water column fluctuates and typically reaches a minimum value early in the morning just before photosynthesis begins for the day. Dissolved-oxygen concentrations also fluctuate over the season with the growth cycle of the algae—when the bloom is growing and the algae are undergoing photosynthesis, oxygen is produced; in contrast, oxygen is consumed by senescing cells. Additional factors that may affect the seasonal fluctuation of dissolved oxygen include the decrease in the saturation concentration of dissolved oxygen as water temperature increases over the season, the increase in the effectiveness of sediment oxygen demand as the lake level decreases through the season, and the increase in biological and chemical oxygen demand from resuspended sediments which may be more likely stirred up at lower lake levels. Wood and others (1996), however, observed that increased oxygen demand (either sediment, biological, or chemical) was not the primary controlling factor of dissolved-oxygen concentrations at lower lake levels.

To further explore the relation of dissolved oxygen and lake levels, Wood and others (1996) proposed the following hypothesis:

The yearly distributions of dissolved-oxygen concentrations for Upper Klamath and Agency Lakes were plotted by month in order of increasing lake level in an attempt to analyze this hypothesis (fig. 15). Dissolved-oxygen concentrations of less than 4 mg/L were common in most years in July and August, in about half the years in September, and in a few years in June and October. A pattern related to lake level is not obvious. In most years, less than 25 percent of the dissolved-oxygen concentrations were less than 4 mg/L. In August, however, nearly half of the years had more than 25 percent of their dissolved-oxygen concentrations less than 4 mg/L, including the lowest, middle, and highest ranking years for lake level. Three years (2002, 2000, and 1998) had dissolved-oxygen concentrations less than 4 mg/L as early as June.

These June low dissolved-oxygen concentrations were observed in the bay (Coon Point, Shoalwater Bay, Wocus Bay) and trench areas (Eagle Ridge) and in Agency Lake (Agency Lake North) (fig. 16 A). This pattern of a higher frequency of low dissolved-oxygen conditions in the northern part of the lake (Eagle Ridge, Coon Point, Shoalwater Bay, Midnorth) continues into July (fig. 16 B) and then even into August (fig. 16 C), when the sites in the southern part of the lake (North Buck Island, Pelican Marina) start to experience these low dissolved-oxygen conditions also. The occurrence of dissolved-oxygen concentrations of less than 4 mg/L in the bay and trench areas continues even later in the season into September (fig. 16 D). Even narrowing the targeted dataset to looking at one month at a time or one site and month at a time did not yield any significant, strong correlations (all Spearman’s ρ were less than 0.5). The dissolved-oxygen data from 1990–94 (Wood and others, 1996), as well as from 1990–2006, do not support a relation between dissolved-oxygen concentration and lake level.

As was previously discussed, years with an earlier bloom were also characterized by an earlier onset of low dissolved oxygen. Therefore, relations between dissolved-oxygen concentrations and cumulative degree-days and the timing of the onset of the bloom were explored (fig. 17). Because both July and August are months with a large number of dissolved-oxygen concentrations of less than 4 mg/L, and young-of-the-year fish are more susceptible to stressful conditions earlier in the season, the month of July was chosen for this further analysis. No obvious relation was revealed between the dissolved-oxygen data and the cumulative degree-days, or the timing of the bloom onset. In fact, the data distributions are actually rather similar from year to year. There were only three years that were never observed to experience dissolved-oxygen concentrations of less than 4 mg/L—1994, 1993, and 1991. Both 1991 and 1993 had later blooms but 1994 had the fourth earliest bloom, again resulting in a lack of a relation or pattern.

Although the frequency distributions of the dissolved-oxygen data do not show a relation with the onset of the bloom, there is some evidence that when the bloom is delayed, the dip in the dissolved-oxygen concentrations also is delayed and may not be as extreme. By looking at the depth-profile data at Eagle Ridge, a site that is more susceptible to low dissolved-oxygen concentrations because of its depth, for four specific years, this relation can be explored (fig. 18). To summarize the conditions in these four years again, 1992 had the earliest bloom of the 17 years, followed by 1997, whereas the bloom in 1991 started and stalled, resulting in the second latest bloom onset, followed by 2006. The onset of the low dissolved-oxygen concentrations for 1992 and 2006 both began at the end of July; even though 1992 had the earliest bloom onset and 2006 had the latest bloom onset. The low dissolved-oxygen concentrations in 1997 (in mid-July) were detected one sampling earlier than in 1992 and 2006. In contrast, low dissolved-oxygen concentrations were not measured until early October in 1991, the year with the smaller early bloom followed by the larger bloom at the end of August (fig. 7). So, for certain circumstances, the years with a later bloom also had a delayed occurrence of low dissolved-oxygen concentration (like 1991), though this did not hold for all years with later blooms (like 2006).

Total Phosphorus

Phosphorus can play an important role in Upper Klamath Lake because high concentrations of phosphorus can lead to heavy algal blooms. As previously discussed, these blooms then lead to elevated pH during the growth of the bloom and associated photosynthesis, and low dissolved oxygen when then the bloom declines. The hypothesis for analysis with regards to phosphorus from Wood and others (1996) is:

The yearly distributions of total-phosphorus concentrations for Upper Klamath and Agency Lakes were plotted by month in order of increasing lake level in an attempt to analyze this hypothesis (fig. 19). As was noted by Wood and others (1996), the data from 1992 stand out. The range of data measured in June and July of 1992 is larger than in any other year for those months. There does not appear to be a significant pattern in the data with relation to lake level. For both the 1990–94 dataset and the larger 1990–2006 dataset, the data are not consistent with this hypothesis. The distributions across the years in a particular month do not show a lot of variability—the more interesting variability to notice is between the months for certain years. Elevated total phosphorus concentrations were measured into September and October for those years that experienced blooms later in the season (1996, 1991, 1995, 1990), either as a second bloom or sustained growth from earlier in the season (fig. 7). Similar to the pattern noted in the frequency distributions of the pH data (fig. 12), and most likely because of related processes, Upper Klamath Lake and Agency Lake had quite different distributions in some years. The elevated concentrations measured in Agency Lake may be an indication of different contributing sources for the lakes.

Because of the importance of the bloom onset on delayed poor water-quality conditions and because of the effect of total-phosphorus concentrations on the algal bloom, the period of the bloom onset (June) was explored further for relations to lake level. Total phosphorus and chlorophyll‑a concentrations in June were determined to be strongly correlated (Spearman’s ρ = 0.80, p < 0.0001) (fig. 20). This indicates that the first bloom is phosphorus-limited. The years with earlier blooms (1992, 1997, 2000, 2001) had higher median concentrations of both total phosphorus and chlorophyll‑a, most likely because the bloom had gone through more growth throughout the month of June. Examining the distribution of the years in figure 20 with respect to their lake-level ranking did not reveal any relation or pattern with lake level.

These June median chlorophyll‑a and total phosphorus concentrations also were plotted by the area of the lake where the measurements were made (fig. 21). All areas retained strong, significant correlations between chlorophyll‑a and total phosphorus. This figure makes it easier to see that the total phosphorus and chlorophyll‑a concentrations in Agency Lake in June were higher than in Upper Klamath Lake. The correlation between the two, however, was the weakest in Agency Lake. The bay areas experienced a wider range of total phosphorus and chlorophyll‑a concentrations.

Water Temperature

The pattern in median surface water temperatures typically observed in Upper Klamath Lake is shown in figure 22. Temperatures typically start out between 10 and 15°C in early May, rise to around 25°C in July and then slowly decline to less than 10°C in late October. Two exceptions to this are the two years with the lowest lake level, 1992 and 1994. These two years started out in early May with surface water temperatures greater than 15°C. The warmest water temperatures of the season (greater than 25°C) were measured in July in 1990, 1996, and 2003, and in late June in 2000. In contrast, the warmest May water temperatures were measured in 1992, 1997, 2001 and 1995. Comparing this list to the degree-days ranking (table 6), based on air temperature from April 1 to May 15, is very enlightening. The years 1992 and 1997 had the first and third warmest springs based on air temperature, but 2001 and 1995 were ranked in the middle of the 17 years based on degree-days. Further exploration of the differences in these lists sheds some light on patterns observed in the timing of the bloom onset.

The years 1990 and 1994 were in the top four ranked years based on cumulative degree-days, but the water temperatures in these years did not warm until later in June and July, respectively. In contrast, water temperatures in 2001 were warming in May, but 2001 was ranked tenth in the degree-days ranking. The fact that the water temperature was on a warming trend that was not apparent in the air temperature perhaps explains why it was the year with the fourth earliest bloom onset (table 13). Likewise, 2000 had a similar lake-level ranking to 2001 and a middle-range degree-day ranking, but had the third earliest bloom onset. This may be related to the median surface water temperature reaching 25°C in late June (fig. 22). One other year that showed an interesting pattern in the rankings is 2006 (table 13). Again it had a similar lake-level ranking to 2000 and 2001 and middle-range ranking of cumulative degree-days; however, it was the year with latest bloom onset. Surface water temperatures did not get over 20°C until July in 2006. Comparing these 3 years illustrates that similar lake levels can have very different bloom onset outcomes—perhaps related to both air and water temperature effects, among other factors.

These varying patterns in air and water temperatures reveal that air temperature can be used as an estimate of water temperature but most likely does not give an accurate estimate of the absolute temperature of the lake. An overall heat budget would be needed for that determination. Generally, the largest components of the heat budget for a lake are incoming shortwave solar radiation and incoming long-wave atmospheric radiation, both of which would be affected by cloud cover. This could partly explain why there is not more correlation between air and water temperatures. The differences observed in the water-temperature patterns for these 3 years of similar lake levels and air temperatures do not support the hypothesis addressing water temperature relations (fig. 23).

Wood and others (1996) speculated that “it is unlikely that a higher lake level is effective in slowing down the spring warming of the lake.” The pattern in rankings for the year 1998 (table 12) offers support for this speculation. The year 1998 had the highest-ranked lake level, leading some to believe that the warming of the lake would be slower. The cumulative degree-days ranking for 1998 was in the cooler third of the years, yet the water warmed to greater than 20°C by the end of June and up to 25°C by the end of July (fig. 23), with the fifth earliest bloom onset. Therefore, the high lake level did not appear to have a slowing effect on the warming of the lake.

Accessibility FOIA Privacy Policies and Notices

U.S. Department of the Interior | U.S. Geological Survey
Persistent URL: https://pubs.water.usgs.gov/sir20075117
Page Contact Information: Publications Team
Page Last Modified: Thursday, 01-Dec-2016 19:46:28 EST