Scientific Investigations Report 2006–5209
U.S. GEOLOGICAL SURVEY
Scientific Investigations Report 2006–5209
The weekly water samples, while collected at only a subset of the sites in the study area, provide valuable context for understanding the data from the continuous water-quality monitors. Trends and fluctuations in water-quality parameters often are associated with bloom dynamics, as reflected in the chlorophyll a data. The basic reasons are straightforward: a sharp reduction in chlorophyll a indicates a crash in the bloom. The spatial extent of the crash may be small enough to be captured at only one site, or it may be large enough that the crash is evident at several sites. A crash in the bloom is characterized by a shutdown or sharp reduction in oxygen production through photosynthesis, which is manifested as a decrease in dissolved oxygen, generally to values below saturation as ongoing sediment and water column demands continue to operate unabated. Periods of growth in the bloom are, in contrast, generally manifested as supersaturated dissolved oxygen and high pH, as photosynthesizing algae consume carbon dioxide and produce oxygen.
The chlorophyll a data collected in all 3 study years are presented in figure 9. These 3 years present interesting contrasts and demonstrate that the 3 study years were quite different. In 2002, the chlorophyll a in samples collected at three of the four weekly sampled sites indicated that a crash in the bloom occurred around mid-July; the chlorophyll a concentration at site UKL06 decreased from 630 µg/L on July 15 to 29 µg/L on July 22. At all sites except UKL08, chlorophyll a concentration remained low for two weekly samples. The best estimate of the dates encompassing the bloom crash are June 19 to August 1, which was arrived at by projecting backward 3 days (approximately one-half of the weekly sampling interval) from the first sample date on which the crash was observed, and forward 3 days from the last sample date on which the bloom was observed. A second crash, evident on only one sample date, occurred later in the season, and the best estimate of the dates encompassing that crash are September 6–12.
The sample collected on July 29, 2002, in Shoalwater Bay was an anomaly—a value of 1,410 µg/L was measured there at the same time that the bloom appears to have crashed at the other sites. Such exceptionally high values have, on occasion, been measured in Shoalwater Bay and Howard Bay in the past 15 years (Klamath Tribes, unpub. data), and may be an indication that AFA can become physically concentrated in these embayments by the action of wind and currents. The high chlorophyll a concentration in Shoalwater Bay when concentrations elsewhere in the lake were at a minimum also is evidence that the blooms in the embayments tend be localized, and that the bloom dynamics in the embayments are often independent of the bloom dynamics in the open waters of the lake.
Chlorophyll a in samples collected in 2003 at all three weekly sampled sites indicated that a crash in the bloom occurred near the end of July, when chlorophyll a concentration decreased from 397 µg/L on July 15 to 17.5 µg/L on July 23. The estimated dates of the crash are July 19 to August 1. Secondary crashes lasting only one sample date were observed in the second and fourth weeks of September; estimated dates are September 6–12 and September 20–26. There is no evidence that the fact that the dates of the July and first September crash coincide exactly with those in 2002 is anything other than a coincidence, although it is possible that some natural cycle in the bloom was responsible.
Data from 2004 provide an interesting contrast to the other 2 years. The chlorophyll a data did not indicate a dramatic bloom crash at any point in the season that affected all sites, although very low chlorophyll a values were measured at individual sites at various times throughout the season. On the basis of low chlorophyll a concentrations at four of the five sites, a minor bloom crash may have occurred between August 7 and 13.
The crashes in the AFA blooms are not prominent in the total phosphorus data (fig. 10), indicating that when the cells senesce, they tend to recycle nutrients to the water column, rather than carrying nutrients to the lake bottom. This is also evident in the plots of dissolved nutrients (orthophosphate and ammonia), which show a roughly inverse relation overall with chlorophyll a (figs. 11 and 12).
When high ammonia concentrations occur coincident with high pH and high temperatures, a significant fraction of the total concentration will be present in the un-ionized form, which is particularly toxic to aquatic life. The dependence on pH is stronger than the dependence on temperature—at 22°C and a pH of 9, 31 percent of the ammonia will be in un-ionized form, but at a pH of 9.5 that fraction increases to 60 percent (U.S. Environmental Protection Agency, 1998). Because ammonia tends to peak when the bloom declines, the peaks in ammonia are often mismatched in time with the highest pH values. That is why the peak in un-ionized ammonia preceded the peak in total ammonia concentration at most sites in the 3 study years (fig. 13). The highest concentration of un-ionized ammonia occurred in 2003 just prior to the bloom crash at the end of July. Concentrations peaked at nearly 300 µg/L at two sites. Experimentally derived 96-hr LC50 (median lethal concentration) values for un-ionized ammonia were 780 and 530 µg/L for Lost River and shortnose suckers, respectively (Saiki and others, 1999), well above the peak concentrations observed in this study. Nonetheless, exposure to lower concentrations cannot be dismissed as a source of chronic stress to fish.
The lack of detectable nitrate/nitrite concentrations through much of 2002 and 2003 and very low nitrate/nitrite concentrations overall in comparison to ammonia, suggest that, in spite of very high ammonia concentrations, nitrification is not a rapid or effective process for removal of ammonia from the water column (fig. 14).
In addition to the temporal patterns through the June to October season, the chlorophyll a and nutrient data showed site-specific trends (fig. 15). The median chlorophyll a and total phosphorus values were higher at site UKL07 in all 3 years than at sites UKL05, UKL02 (2004 only), or UKL06 (2002 only). Thus, there is a tendency for the deeper part of the study area to the south to experience higher concentrations of chlorophyll a, total phosphorus, and ammonia than the shallower part of the study area to the north and northeast. There also is a slight tendency for concentrations at site UKL08, which is in Shoalwater Bay, to be lower than those at site UKL07, which is centrally located in the study area.
In order to provide a sense of the overall seasonal and interannual variability in water quality in the study area, it was useful to consolidate the datasets into a single statistical median that would be broadly representative of the entire study area. Because the sampling design changed between 2002 and 2003, this median was computed using only sites that were common (or nearly so) to all 3 years: UKL01–UKL11. Two sites, UKL07 and UKL08, were located at the same latitude and longitude in all 3 years; the rest of the sites included in the median were located at a different latitude and longitude between 2002 and 2003, but were judged to represent generally similar locations in the lake. A study-area daily median was only computed when a daily median could be computed for 8 of the 11 individual sites. Daily medians were calculated at a site only for those days on which at least 20 hours of data were available.
The study-area daily median of dissolved oxygen, percent saturation of dissolved oxygen, temperature, and pH for 2002–04 is presented in figure 16. The percent saturation of dissolved oxygen provides a convenient way to assess the seasonal trends in each year and to make year-to-year comparisons. A value of 100 percent saturation indicates that the water column is in equilibrium with the atmosphere. Values greater than 100 percent indicate that oxygen production through photosynthesis is in excess of oxygen consumption through respiration and decay processes. Values less than 100 percent saturation indicate the opposite situation—oxygen consumption is in excess of oxygen production. Surface exchange processes always act to move the system toward 100 percent saturation from either a supersaturated or an undersaturated condition.
Over much of the summer and autumn in all 3 years the rate of photosynthetic production of oxygen in the water was enough to keep the study-area daily median dissolved oxygen concentrations well above saturation, even though chlorophyll a concentrations indicate that the bloom was highly variable, both spatially and temporally. This was the case even though concurrent oxygen demanding processes such as sediment oxygen demand and water column biological oxygen demand can consume large amounts of oxygen (Wood, 2001; Lieberman and others, 2003). In each year, however, there were periods when the balance changed and photosynthetic production abated or stopped and/or oxygen demands increased, and the study-area daily median concentration of dissolved oxygen in the water column decreased to well below saturated conditions. Understanding what happens during these periods is of interest because dissolved oxygen can become low enough to produce chronic or acute stress on fisheries. These conditions were more common in 2003 than in 2002 or 2004, and notably the percent saturation and concentration of dissolved oxygen trended downward through the entire month of July 2003. The study-area median concentration approached 2 mg/L and 20 percent saturation at the end of July 2003.
Study-area median pH conditions for 2002 and 2004 had a similar pattern. The pH increased early in the season and slowly decreased until mid-to-late August, following the same pattern of expansion and decline in the AFA bloom that is evident in the dissolved oxygen data. The pH then increased to greater than 9.0 and remained there for the rest of these field seasons. The study-area daily median pH for 2003 had the same pattern of higher pH in the early field season, but then showed large departures from those in the other 2 years. The pH decreased more abruptly in mid-to-late July than in the other 2 years of the study, mirroring the steep decrease in dissolved oxygen during this time. The study-area daily median pH then increased sharply in early August, hovered around 9.5 until late August, and then decreased for the rest of the field season to values in October that were much less than those in the other years of the study.
Temperatures increased from late June through late July, and then began a steady, decreasing trend from early August until the end of the field season. This general pattern is seen in all years, with the result that the highest temperatures in each of the 3 years of the study occurred around the end of July. Notable, however, is the fact that the highest temperatures of the 3 study years were observed at the end of July 2003. These temperatures exceeded the peak temperatures in the other 2 years by 2 to 3°C. This may be significant because the July 2003 time period coincided with the lowest study-area median dissolved oxygen. High temperatures accelerate oxygen demanding processes and may have contributed to the low dissolved oxygen during this time.
In order to identify periods when water-quality conditions potentially harmful to fish were present in the study area, and the spatial extent of these conditions, the hours when dissolved oxygen was less than 4 mg/L, temperature was greater than 28ºC, or pH was greater than 9.7 were enumerated at each site. These values were based on high stress thresholds established to calculate stress indices for Upper Klamath Lake suckers (Reiser and others, 2000). The numbers of hourly occurrences of these conditions per day were summed over all of the sites to provide a count of “site-hours” per day of these conditions. To compare these counts of site-hours across years, the counts per day were divided by the total possible site-hours in the day (24 hours multiplied by the number of sites), giving a percentage of site-hours in the day that had potentially harmful conditions. This statistic provided a concise measure of the severity of the water-quality conditions across the study area during each day. To get a measure of the spatial extent of these potentially harmful conditions, the percentage of sites at which a potentially harmful condition was recorded at least 1 hour in a day was calculated. The study-area statistic was only computed when a daily statistic could be computed for 8 of the 11 individual sites, and the daily statistics were calculated at a site only for days on which at least 20 hours of data were available (figs. 17 and 18). Temperature is not included in the plots because temperature did not exceed 28°C in 2002 or 2004, and exceeded 28°C at only one site for a few hours on 3 days in 2003.
During several periods in 2002, at least one instance of potentially harmful dissolved oxygen conditions per day was recorded at 10 percent or more of the sites in the study area. Before mid-August, there were several days during which almost 20 percent of the sites recorded a dissolved oxygen concentration less than 4 mg/L; on one day in early August, 40 percent of the sites recorded dissolved oxygen concentration less than 4 mg/L (fig. 17A). During the second week in July and again during the second week in August, the percentage of site-hours during which concentrations less than 4 mg/L were recorded exceeded 60 percent, indicating that for a few days at a time the occurrence of these conditions was widespread and persistent throughout the day (fig. 18A). Occurrences of potentially harmful pH conditions were less common, but during a few days in late July, 10 percent or more of the sites recorded at least one pH reading greater than 9.7 (fig. 17B). There were, however, relatively few site-hours for which conditions potentially harmful to fish were recorded for any parameter in 2002, indicating that the duration of these conditions at any one site generally was only a few hours during the day (fig. 18B).
The year 2003 was unique among the 3 study years in both the extent and daily duration of low dissolved oxygen conditions. From July 22 through July 31, 2003, a large fraction of the study area (as much as 60 percent of sites) had at least one dissolved oxygen measurement less than 4.0 mg/L in a day. During the same period, consistently high percentages (more than 60 percent) of the daily site-hours were characterized by concentrations less than 4.0 mg/L. The combination of a large fraction of the study area having at least one dissolved oxygen reading per day less than 4.0 mg/L and a high percentage of daily site-hours having these conditions indicates that the spatial extent of this low dissolved oxygen event was large and that potentially harmful dissolved oxygen conditions persisted through most of the day during the event. The large percentage of sites (40–50 percent) and the consistently high percentage of site-hours (50 percent or greater for nearly 2 weeks) recording dissolved oxygen concentrations less than 4 mg/L in late September and early October also was unique (figs. 17A and 18A). Figures 17B and 18B show that potentially harmful pH conditions in July and again in August and early September were, like the low dissolved oxygen conditions in late July, widespread throughout the study area and persistent throughout the day during these periods.
The 2003 statistics show that low dissolved oxygen conditions were more widespread throughout the study area and more persistent throughout the day for about 10 consecutive days between mid-July and early August than during any period in any of the other years. These water-quality conditions coincided with the beginning of a significant fish die-off event in 2003. In the period from July 22 to September 30, 2003, 108 endangered suckers were found dead in Upper Klamath Lake (B.J. Adams, U.S. Geological Survey, written commun., 2006). Widespread and persistent pH conditions greater than 9.7 through much of the rest of the field season provided further potential to stress fish during 2003 and likely contributed to fish mortality in the latter part of the summer of that year.
Potentially harmful dissolved oxygen conditions were recorded at more than 20 percent of sites in 2004 during a few days in mid-August (fig. 17A), and the percentage of site-hours indicates that the conditions were widespread during those days (fig. 18A). A pH greater than 9.7 was recorded during more than 20 percent of the site-hours nearly every day over a 2-week period starting in late June of 2004, and the percentage peaked at more than 40 percent (fig. 17B). The potentially harmful pH conditions were widespread in the study area during this period, as indicated by the high percentage (more than 60 percent) of site-hours affected (fig. 18B).
The previous discussions have provided an overview of the occurrence and daily duration of poor water-quality conditions across the study area. It is also of interest to determine more specifically where these conditions occur, whether they tend to occur at the same sites in every year, whether the spatial extent is defined by a single, contiguous area or whether occurrences are scattered around the study area, and to get a quantitative estimate of the spatial extent.
To answer these questions, two-dimensional spline interpolations of the weekly median values at each site in each study year were generated using ArcGIS Spatial Analyst. The weekly time step was used because it allowed enough averaging to smooth the data while still capturing the important temporal changes during the season. The interpolations have limitations, however. First, the spacing of the sites in the network provides a lower limit on the length scale of spatial variability that can be resolved. At various times in each year, an effort was made to complete higher resolution transects between established sites in order to determine whether the dominant spatial scales were being resolved by the network. The results were inconclusive, as sometimes the transects showed little small-scale variability between network sites, although at other times there was more small-scale variability. A complicating factor seems to be that it is impossible to get a truly synoptic set of measurements, and the temporal variability in lake conditions is sometimes so great that what appears to be spatial variability is simply temporal variability over the traveltime along the transect. Nonetheless, the fact that the interpolations generally show smooth trends over the dimensions of the study area suggests that the dominant patterns are being resolved by the network. Second, the spline interpolation is a mathematical algorithm that fits a smooth surface through known values at known locations, but it does not incorporate equations describing the physics of the situation. For example, geographic features like Eagle Ridge that are real physical barriers are not seen as such by the interpolation. Another example is that physical features of the lake bathymetry that influence the water circulation and the distribution of water quality, like the trench, do not affect the interpolation. In this sense, the GIS interpolations are different from, for example, the visualization of the output of a numerical transport model. Even with these limitations, however, the GIS interpolations are a useful means of visualizing the spatial patterns in water quality and estimating the spatial extent of extreme conditions.
On the basis of GIS coverages of the interpolations, the spatial extent of dissolved oxygen concentrations less than 4 mg/L was estimated for each Julian week in each study year. From these results, it was determined that the maximum spatial extent of these conditions occurred between week 31 and 34 in all 3 years. (The dates corresponding to Julian weeks in each year are provided in table 2 for reference.) The results from these 4 weeks show how uniquely severe the conditions were in 2003, when the spatial extent of the weekly median of dissolved oxygen less than 4 mg/L reached a maximum of 39 km2 in Julian week 31 (table 3). In comparison, the maximum spatial extent of these conditions in 2004 was 2 km2 and in 2002 was 11 km2.
Even though there was a big difference in the severity of the lowest dissolved oxygen conditions between the 3 years of the study, the interpolations show consistency in the patterns of weekly median dissolved oxygen (fig. 19). Recall that the water-quality monitors are located at 1 m from the bottom, so the interpolations represent the concentrations on a surface that follows the bathymetry of the lake, 1 m from the bottom, rather than concentrations at or near the surface of the lake. The patterns that are evident, therefore, are a result of vertical as well as geographic differences in the location of the monitors. The highest concentrations generally occur toward the northern boundary of the study area, although the lowest concentrations tend to occur toward the southern end of the study area, along the segments of the trench that cross the entrances of Ball and Shoalwater Bays, and sometimes extend into the bays themselves. The patterns demonstrate that the lowest concentrations occur most often at the deeper sites (table 4).
Spatially, the range in values is quite large, from highly supersaturated (indicating photosynthetic production of dissolved oxygen in excess of consumption) to very undersaturated (indicating that respiratory and other consumptive processes exceed production). The 2003 map for week 31 shows a particularly large range in dissolved oxygen from north to south (from the shallower areas to the deeper areas) in the study area. This period coincides with the major bloom decline in that year and reflects a steep decrease in dissolved oxygen in the deeper part of the study area that is associated with that decline. In all 3 years, the blooms tended to peak at higher values at the deeper sites, as evidenced by the general increase in median seasonal chlorophyll a concentrations from the northern to the southern sites (fig. 15), and the effects of the bloom declines on dissolved oxygen were more severe at the deeper sites as well (fig. 19). In contrast, both the peaks of the bloom and the decreases in dissolved oxygen associated with a crash were moderated at the shallower sites toward the north and east in the study area.
The spatial extent of pH greater than 9.7 was at a maximum in late June in each study year, but among the 3 study years the extent was by far the greatest in 2004, encompassing as much as 47 km2 (table 3). Overall, bloom intensity was moderate in 2004 compared with that in the other 2 years (fig. 9), but the June peaks in chlorophyll a were higher and apparent at more of the sites in 2004 than in 2002 or 2003. As a result, the spatial extent of extreme pH in June, 2004, was large, and the extreme values occurred with nearly the same frequency and daily duration at deep and shallow sites (table 5). At times, the spatial pattern in pH trended from higher values toward the north to lower values toward the south (fig. 20). This trend was particularly evident in 2004, and reflects the greater photosynthetic activity in the shallower areas to the north, where the entire water column is in the photic zone.
Temperatures greater than 28°C did not have a measurable spatial extent in any week of any year (fig. 21). Although there were a few periods in 2003 when temperatures exceeded 28°C, those periods were extraordinary and of short duration, lasting less than half a day. The fact that temperatures greater than 28°C are rarely observed suggests that extreme temperature is not a significant contributor to chronic stress on fish health in at least the northern part of Upper Klamath Lake.
Histograms showing the timing of daily minimum dissolved oxygen, daily maximum pH, and daily maximum temperature at a representative deep site (defined as having a full pool depth greater than 2.5 m) and shallow site (defined as having a full pool depth less than 2.5 m) are shown in figure 22. These histograms display the normalized frequency of occurrence of the daily extreme water-quality reading as a function of the hour of day the most extreme water-quality reading occurred. Data from all 3 years were combined to create these histograms after determining that data from each of the 3 years showed the same patterns. Histograms for shallow sites were similar to each other, as were the histograms for the deeper sites. A representative shallow site (UKL02) and a representative deep site (UKL07) were chosen to illustrate the results. It is helpful to recall that the monitors used in this study were located at 1 m off the bottom, so at shallow sites they are located within the photic zone, whereas at deep sites they are located below the photic zone.
Daily minimum dissolved oxygen concentration at the shallow sites tended to occur around 7 a.m. (fig. 22A). Daily maximum pH and temperature occurred most frequently in the early evening between the hours of 4 p.m. and 8 p.m. (fig. 22B and 22C). This pattern reflects the strong influence of the diel cycle of photosynthesis and respiration on the water column, which is almost entirely within the photic zone at these sites.
At deep sites, however, there was no distinct period of the day when daily minimum dissolved oxygen conditions tended to occur (fig. 22D). The maximum pH at the deep sites tended to occur in late afternoon, but over a much broader window of time than at the shallow sites, lasting from the hours of about 4 p.m. to 2 a.m. the next morning (fig. 22E). Temperature at the deep sites behaved much like temperature at the shallow sites, usually peaking in a narrow window between the hours of about 5 p.m. and 9 p.m. in the late afternoon (fig. 22F).
At the deep sites, only the surface (roughly the top 2 m) of the water column is in the photic zone. In the lower part of the water column, below the photic zone, the rate of consumption of dissolved oxygen by metabolic respiration and other oxygen demanding processes within the water column is greater than the rate of production of oxygen by photosynthesis. Near the bottom, sediment oxygen demand also consumes dissolved oxygen from the water. Temperature differences between the upper and lower water column can produce some degree of thermal (density) stratification, which isolates the lower from the upper water column. If the water column develops even a small degree of stratification, it then becomes possible for dissolved oxygen to reach minimum daily conditions in the lower part of the water column (where site monitors were located) at any time during the day, even during daylight hours. This is because dissolved oxygen below the photic zone will continue to decrease until a mixing event erodes the thermal stratification and mixes the lower layer with the upper layer, which is at a relatively higher concentration. At a deep site, therefore, the minimum in dissolved oxygen does not always occur in the early morning hours just before dawn, but rather can occur throughout the day depending on when the wind picks up and mixes the water column.
At the shallow sites, a secondary maximum in pH and temperature occurs at midnight (figs. 22B and 22C). On close inspection of individual time series, it was determined that this secondary maximum occurs on a significant subset of days when the parameter (pH or temperature) is trending strongly downward. In that case, the change in the parameter over the day from midnight to midnight is greater than the upward and downward movement due to solar radiation during the day, with the result that the maximum occurs at the beginning of the day instead of some other time during the day. Similarly, when dissolved oxygen is trending strongly downward, the minimum of the day occurs at 11 p.m. for the same reason (fig. 22A). When the parameters are trending strongly upward, the opposite (maximum in pH and temperature at 11 p.m.; minimum in dissolved oxygen at midnight.) does not occur as often because of the asymmetry in the diel cycle, which is heavily skewed toward the late afternoon, thus making it less likely that a value at the end of the day will end up being an extreme value. A strong secondary maximum in temperature at midnight also is seen at the deep site (fig. 22F). The trends and diel cycle in temperature in the lower layer of the deep water column develops in response to the same atmospheric forcing that operates at the shallow sites, so the timing of extreme values is similar, but the magnitude of the diel fluctuations generally is dampened in comparison to that at the shallow site.
U.S.
Department of the Interior | U.S. Geological
Survey
Persistent URL: https://pubs.water.usgs.gov/sir20065209
Page Contact Information: Publications Team
Page Last Modified: Thursday, 01-Dec-2016 19:13:00 EST
