Purpose and Scope

WQ data have been collected biweekly in Upper Klamath Lake by the Klamath Tribes (Klamath Tribes, 2006) since the early 1990s and continuously with weekly sampling by the U.S. Geological Survey (USGS) since 2002 when monitoring stations were first placed in the northern part of the lake (Wood and others, 2006). The USGS monitoring program expanded to be lake-wide in 2005 (Hoilman and others, 2008), when WQ sondes still in operation today were installed at the four core continuous monitoring sites (fig. 1). Previous studies examined the status and trends of WQ in Upper Klamath Lake from the available data (Jassby and Kann, 2010; Eldridge and others, 2014) and determined that continued long-term monitoring would be necessary to identify statistically significant trends in extreme water quality that could be harmful to fish health. For this study, relations between the continuous WQ and discrete ammonia measurements from the four core USGS sites during 2005–19 were examined graphically to visualize the spatial and temporal characteristics of extreme WQ in Upper Klamath Lake. This study does not examine the long-term trends of extreme WQ in Upper Klamath Lake nor does it incorporate fish health or population data from Upper Klamath Lake.

Processing Continuous Data

The continuous WQ data were interpolated to an hourly time step filling gaps of 2 hours or less. Daily WQ time series were created by averaging hourly values for each day. Days with data gaps larger than 2 hours were set to not applicable for the daily set. The same process was used for the continuous lake-surface elevation data to produce hourly and daily sets. The daily average datasets will be referred to as WT_avg, pH_avg, DO_avg, SC_avg, and ELEV_avg. Some analyses considered deviations in WQ and lake-surface elevation from expected values by calculating the difference between daily average values and Julian day average values. The daily deviation datasets will be referred to as WT_dev, pH_dev, DO_dev, SC_dev, and ELEV_dev.

Extreme Water-Quality Days

Days with persistent conditions of extreme and potentially harmful WQ were identified over the study period. Extreme WQ days based on continuous data were defined as those during which at least 12 hours of measurements exceeded, or were less than in the case of DO, an extreme and potentially harmful WQ threshold. These days will be referred to as WT_ext, pH_ext, and DO_ext. For NH₃, extreme WQ days were defined as those days during which the weekly estimated NH₃ concentration exceeded the extreme WQ threshold.

Statistically extreme WQ thresholds were calculated from the combined WT_avg, pH_avg, and DO_avg time series for all core monitoring sites from 2005 to 2019. For water temperature and pH, which can be harmful to fish at higher values, the extreme WQ threshold was calculated as the 99th percentile of the long-term daily averages with values of 24.38 °C for water temperature and 10.04 standard units for pH (table 2). For DO, which can be harmful to fish at lower values, the extreme WQ threshold was calculated as the 1st percentile of the long-term daily average concentration at a value of 1.76 mg/L (table 2). An extreme WQ threshold also was calculated for NH₃, which can be harmful to fish at high concentrations, from the combined weekly set of NH₃ estimated from weekly measured total ammonia at the core sites from 2005 to 2019. The 99th percentile of the long-term weekly NH₃ set was 264 μg/L (table 2).

Graphical and Tabular Data Summary

This study focused on graphical representations and tabular summaries of data using R (Wickham, 2016; R Core Team, 2021) to examine relations between WQ variables and better understand the drivers of extreme water quality. The non-parametric Spearman rank correlation (ρ) was used to evaluate the magnitude of relations between datasets and a significance level (α) of 0.05 was used to reject the null hypothesis that ρ was not significantly different than zero. Daily average and hourly values of water temperature, pH, DO, and SC, and weekly estimates of NH₃ from the four core sites in Upper Klamath Lake from 2005 to 2019 were considered. Most data were collected from June to September; however, in some years, data collection included parts of May, October, and November (fig. 2). For some analyses, WQ variables were considered with respect to daily average lake-surface elevation, which had a consistent seasonal pattern throughout the study period (fig. 3). One analysis used deviations from expected values of data that were created by subtracting the Julian day average values, calculated from 2005 to 2019, from the daily average values on matching Julian days. The graphical representations included the following:

Comparisons of Daily Average Water Quality and Extreme Water-Quality Days

The combined WT_avg and DO_avg for the core monitoring sites had a weak, negative relation (ρ = -0.21 [p< .001]), where the lowest DO_avg (< 1.76 mg/L) occurred in July and August (fig. 4A) when WT_avg was high (> 20 °C); however, many days showed that high DO_avg co-occurred with high WT_avg. July and August also had the most days of extreme water temperature and DO (table 3). DO_avg and pH_avg had a weak, positive relation (ρ = 0.38 [p < .001]; fig. 4B) where the highest pH_avg values (> 10.04) occurred with high DO_avg values (> 10 mg/L) and the lowest DO_avg values (< 1.76 mg/L) occurred with pH_avg values that ranged from 7 to 9.25.

SC_avg had negligible relations with pH_avg (ρ = –0.03 [p < .001]) and DO_avg (ρ = –0.16 [p < .001]) during the complete sampling season (fig. 4C, 4D); however, seasonal relations were considered. The lowest SC_avg values occurred earlier in the sampling season and in May and June, SC_avg had a weak, positive relation with both pH_avg (ρ = 0.34 [p < .001]) and DO_avg (ρ = 0.22 [p < .001]). The highest SC_avg concentrations (> 120 microsiemens per centimeter [μS/cm]) occurred in all sampling months and corresponded to a range of pH_avg values and DO_avg concentrations. The pH_avg values that corresponded to high SC_avg concentrations had a bimodal distribution wherein one group of values clustered above a pH_avg of 9.0 (with a median pH_avg of 9.7) and were measured in June through August, and the second group of values clustered below a pH_avg of 9.0 (with a median pH_avg of 7.8) and were measured predominantly in September and October. In September and October, SC_avg had a moderate, negative relation with both pH_avg (ρ = –0.56 [p < .001]) and DO_avg (ρ = –0.44 [p < .001]).

Weekly estimated NH₃ had a moderate, positive relation with pH_avg (ρ = 0.53 [p < .001]), which was one of the variables used in its calculation, and a weak, negative relation with DO_avg (ρ = –0.24 [p < .001]; fig. 4E, 4F). The highest concentrations of NH₃ (> 264 μg/L) mostly occurred in July and August (table 3), with pH ranging from 8.8 to 9.8 and DO ranging from 4.1 to 9.4 mg/L.

Distribution of Water Quality with Respect to Lake-Surface Elevation and Month

Daily average lake-surface elevation (ELEV_avg) in Upper Klamath Lake had a consistent seasonal pattern throughout the summer sampling period from 2005 to 2019. ELEV_avg were highest in May, averaging 4,142.7 ft (1,262.7 m), and steadily decreased until October when elevation stabilized to an average 4,138.7 ft (1,261.5 m; fig. 3).

The distribution of WT_avg with respect to ELEV_avg and month had a seasonal pattern wherein temperatures increased through July as lake-surface elevation decreased, then decreased through the end of the sampling period as lake-surface elevation decreased (fig. 5A). The highest WT_avg occurred in July and August, when ELEV_avg was 4,139.5– 4,141.5 ft (1,261.7–1,252.3 m), and 60 days of WT_ext were measured in July, from 2005 to 2019 (table 3).

The distribution of pH_avg with respect to ELEV_avg and month had a similar seasonal pattern as WT_avg where pH increased through July as lake-surface elevation decreased, then decreased through the end of the sampling period as lake-surface elevation decreased, although there were differences (fig. 5B). The highest pH_avg values occurred in June, July, or August, when ELEV_avg was 4,139.5– 4,142.5 ft (1,261.7– 1,262.6 m), and 96 days of pH_ext were measured in June and July, from 2005 to 2019 (table 3). The distributions of pH_avg in July and August at 4,139.5– 4,141.5 ft (1,261.7–1,262.3 m) were negatively skewed and many low pH_avg outliers fell below the lower whisker which represented the smallest value within 1.5 times the interquartile range.

The distribution of DO_avg with respect to ELEV_avg and month showed little change in the central tendency with median DO_avg concentrations ranging from 7.5 to 10 mg/L (fig. 5C). Like high WT_avg, the lowest DO_avg concentrations were observed in July and August when ELEV_avg ranged from 4,138.5 to 4,141.5 ft (1,261.4– 1,252.3 m). The lowest DO_avg concentrations occurred in the months of July, August, and September across a broad range of ELEV_avg, and those months combined had 153 days of DO_ext from 2005 to 2019 (table 3). The 153 days of DO_ext occurred with WT_avg of 18.7–24.3 °C on all but 3 days, which corresponded to a maximum daily average level of 24.4 percent oxygen saturation. The highest DO_avg concentrations (> 12.5 mg/L) were measured primarily in July for the 4,140.5– 4,141.5 ft (1,262.0–1,262.3 m) elevation range across seven different years (2007, 2010–12, 2015, 2017, 2019).

The distribution of SC_avg with respect to ELEV_avg and month showed that concentrations consistently increased during the season with decreased lake-surface elevation (fig. 5D). The biggest change in distribution with respect to elevation occurred in June, when the median SC_avg increased by 9 μS/cm for the 4,140.5– 4,141.5 ft (1,262.0–1,262.3 m) ELEV_avg range as compared to the 4,141.5– 4,142.5 ft (1,262.3–1,262.6 m) range.

The distribution of deviations of WQ variables from Julian day average values (WT_dev, pH_dev, DO_dev, and SC_dev) with respect to deviations of lake-surface elevation from Julian day average values (ELEV_dev) did not show strong relations (fig. 6). ELEV_dev had a weak, negative relation with SC_dev (ρ = –0.36 [p < .001]) and showed a negligible relation with WT_dev, pH_dev, and DO_dev. For lower-than-expected lake-surface elevations, SC_avg was mostly higher than expected. For higher-than-expected lake-surface elevations, SC_avg was mostly lower than expected, but the distributions were positively skewed with several occurrences of higher than expected SC_avg. DO_avg concentrations that were at least 5 mg/L lower than the Julian day average occurred for every range of ELEV_dev, whereas DO_avg concentrations that were more than 5 mg/L higher than the Julian day average occurred only when lake-surface elevations were in the expected range (within 0.5 ft of the Julian day average) or higher than expected (>0.5 ft higher than the Julian day average). For expected or higher-than-expected lake-surface elevations, the median pH_dev was slightly higher (0.11–0.13 standard units) than expected but negatively skewed with occurrences of pH_avg that were at least 1.5 standard units lower than the Julian day average. In contrast, for lower-than-expected lake-surface elevations, the median pH_dev was slightly lower than expected (0.18 standard units).

Site Comparisons of Water Quality with Respect to Time and Lake-Surface Elevation

At Mid-North, a site with full-pool depth of 4.2 m (table 1), measurements were taken from sondes in the upper (MDNU) and lower (MDNL) water column. For both sondes, extreme water quality in the form of high water temperature, high pH, or low DO concentration mostly occurred from mid-June through mid-September when ELEV_avg ranged from 4,139 to 4,142 ft (1,261.6–1,262.5 m). More days of extremely high water temperature and pH were measured at the upper sonde than the lower sonde, and more days of extremely low DO were measured at the lower sonde than the upper sonde (figs. 7 and 8). Within the 15-year study period, MDNU had 10 days of WT_ext and 11 days of pH_ext, whereas MDNL had 11 days of DO_ext (table 4). In 2006, pH_avg exceeding 9.5 persisted into late September and early October for both sondes. Both sondes showed a similar range in SC_avg measurements.

At Mid-Trench, the deepest core site in Upper Klamath Lake with a full-pool depth of 15 m (table 1), measurements were taken from sondes in the upper (MDTU) and lower (MDTL) water column. The sondes at Mid-Trench had similar timing of extreme water quality for the same lake-surface elevation range as Mid-North, with more days of extremely high water temperature and pH measured at the upper sonde than the lower sonde, and more days of extremely low DO measured at the lower sonde than the upper sonde (figs. 9 and 10). Within the 15-year study record, MDTU had 29 days of WT_ext and 13 days of pH_ext, whereas MDTL had 132 days of DO_ext, which accounted for 85 percent of all days of DO_ext at the core sites (table 4). Days of DO_ext at MDTL occurred as early as July 3 (in 2015) and as late as October 10 (2015) for ELEV_avg ranging from 4,138.0 ft–4,141.9 ft (1,261.3–1,262.5 m; fig. 10). In 2006, like Mid-North, the sondes at Mid-Trench had pH_avg exceeding 9.5 into late September and early October. Both sondes had similar SC_avg measurements to the MDNU and MDNL sites.

The sites at Williamson River outlet (WMR) and Rattlesnake Point (RPT) were more than 10 kilometers (km) apart; however. both sites were set up with a single sonde, had full-pool depths in a similar range (2.5–3.4 m), and had similarities in their spatio-temporal water quality during the study period. For both sondes, extreme water quality in the form of high water temperature and high pH occurred from mid-June through September at several lake-surface elevations from 4,139 to 4,142 ft (1,261.6 to 1,262.5 m), similar to Mid-North and Mid-Trench (figs. 11 and 12). Within the 15-year study period, WMR had 24 days of WT_ext and 61 days of pH_ext, which accounted for 54 percent of all days of pH_ext at the core sites (table 4). RPT had 5 days of WT_ext and 27 days of pH_ext. Both sondes had few days during which DO_avg was less than 5 mg/L and only 1 day of DO_ext occurred at WMR during the study period. The value of pH_avg exceeded 9.5 standard units in late September or early October at WMR or RPT during 2005–08 and 2010–11.

The spatial variability of continuous WQ data was considered by evaluating the variance in daily average values across all core sites for each day of the study period. The spatial variability was smaller for WT_avg and pH_avg, and larger for DO_avg and SC_avg. The range in WT_avg for all sites and days was 5.71–26.1 °C and the daily variance across all sites was less than 0.8 °C (4.1 percent of the range) for 90 percent of days. The range in pH_avg was 7.0–10.4 standard units and the daily variance across sites was less than 0.2 standard units (4.7 percent of the range) for 90 percent of days. The range in DO_avg was 0–18.5 mg/L and the daily variance across sites was less than 0.6 mg/L (3.1 percent of the range) for 25 percent of days and more than 4.7 mg/L for 25 percent of days. Focusing on August, which was the month with the most days of DO_ext (table 3), the range in DO_avg was 0–16.4 mg/L and the daily variance across core sites was less than 3.0 mg/L for 25 percent of days and more than 7.2 mg/L for 25 percent of days. Finally, the range in SC_avg concentration for all sites and days was 86.8 –143 μS/cm and the daily variance across core sites was less than 3.2 μS/cm (5.7 percent of the range) for 50 percent of all days.

The upper sondes at Mid-North (MDNU) and Mid-Trench (MDTU) had the highest estimated concentrations of un-ionized ammonia (NH₃; table 4). For both sondes, extremely high NH₃ mostly occurred in July and August at ELEV_avg ranging from 4,139 to 4,141 ft (1,261.6–1,262.2 m; fig. 13). Within the study period, MDTU had 9 weekly measurement dates when estimated NH₃ exceeded 264 μg/L and those accounted for 60 percent of all days of extremely high NH₃ for the core sites. MDNU also had five weekly measurement dates when estimated NH₃ exceeded 264 μg/L. The highest estimated NH₃ concentration occurred on dates when measured total ammonia concentrations ranged from 312 to 873 μg/L and calculated fractions of un-ionized ammonia ranged from 0.32 to 0.85 owing to the combination of measured pH exceeding 9.0 standard units and measured water temperature exceeding 21.53 °C for all ammonia measurement times. The WMR site had no dates of estimated NH₃ exceeding 264 μg/L despite having the most dates of extremely high pH. The 14 highest total ammonia concentrations measured at WMR ranged from 330 to 740 μg/L but coincided with calculated fractions of un-ionized ammonia (eq. 2) that ranged from 0.01 to 0.43 owing to lower measured pH (< 9.0 standard units) and water temperature (< 22 °C) during high total ammonia measurement times.

Comparisons of Extreme Water-Quality Days with Respect to Time and Lake-Surface Elevation

Water temperatures greater than 24.38 °C (the 99th percentile for all WT_avg) for multiple hours in a day were measured at all sites and in all years of the study (table 5; fig. 14). The most persistent conditions of extreme temperature were measured at the upper sondes of Mid-North (MDNU) and Mid-Trench (MDTU), and at the shallower WMR site. Within the study period, MDTU had 29 days and WMR had 24 days of WT_ext (table 4), and MDTU recorded the highest hourly temperature of 29.46 °C on July 14, 2018. The years with the highest temperatures were 2006, 2009, 2015, and 2017 with totals ranging from 8 to 23 days of WT_ext per year (table 5).

A pH greater than 10.04 standard units for multiple hours in a day was measured in all years of the study, predominantly at the shallower WMR and RPT sites, and the upper sondes at Mid-North and Mid-Trench (fig. 15). WMR had 61 days and RPT had 27 days of pH_ext during the study period (table 4), and WMR recorded the highest hourly pH of 10.64 standard units on June 29, 2018. The years with the highest pH were 2007–08, 2012–16, and 2018, with totals ranging from 7 to 28 days of pH_ext per year and that did not coincide with the years of extreme high water temperature, with the exception of 2015 (table 5).

DO concentrations less than 1.76 mg/L were measured in all years of the study, predominantly at the lower sondes of Mid-North (MDNL) and Mid-Trench (MDTL; fig. 16). MDTL had 132 days of DO_ext during the study period, which accounted for 85 percent of extreme low DO days for the core sites (table 4). The lowest concentrations of less than 0.2 mg/L were measured in:

The years with the most days of DO_ext were 2005, 2008–09, and 2012–19, with totals ranging from 7 to 25 days per year (table 5). A negligible relation was present between annual total days of DO_ext and WT_ext; however, a moderate, positive relation existed between annual total days of DO_ext and pH_ext (ρ = 0.43 [p = .11]) that was not statistically significant. Significant moderate, positive relations existed between annual total days of DO_ext and total days of pH_ext from the previous year (ρ = 0.63 [p = .015]), and also between the combined total days of pH_ext in the current and previous years (ρ = 0.66 [p = .010]).

Distribution of Un-Ionized Ammonia with Respect to Year

High NH₃ concentrations mostly occurred in the latter half of the study period when 86 percent of concentrations greater than 264 μg/L were estimated at the Mid-North and Mid-Trench sites over 2012 to 2019 (table 5; fig. 17). MDTU had 9 days and MDNU had 5 days during which estimated NH₃ exceeded 264 μg/L from 2005 to 2019 (table 4), and MDTU had the highest estimate of 403 μg/L on July 21, 2008. Based on the yearly upper quartiles, the highest estimated NH₃ concentrations occurred in 2005, 2008–09, and 2016–19 (fig. 17); however, concentrations greater than 264 μg/L also were estimated in 2012–15 (table 5).

Seasonal Relations

The seasonality of water temperature, pH, DO, and NH₃ during the summer sampling months was influenced by solar radiation and the life cycle of cyanobacteria in Upper Klamath Lake.

During the study period (2005–19), daily average temperature (WT_avg) and pH (pH_avg):

The highest levels of WT_avg and pH_avg were high but not in the lethal range for 24-hour exposure. The distribution of pH_avg during July–September was skewed left because in some years pH had a mid-season drop. The days with the lowest mid-season DO_avg values occurred with pH_avg ranging from 7.05 to 9.39 standard units.

The central tendency of DO_avg was consistently in the range of 7.5 to 10 mg/L throughout the sampling season, but fluctuated from 0.0 to 18.5 mg/L. Within the study period, DO_avg concentrations below 0.2 mg/L occurred on 13 days in July and August at MDTL and there were 87 days (57 percent of all months) of DO_ext measured in August.

Extreme NH₃ concentrations occurred when high pH_avg coincided with high ammonia concentrations. Within the study period, extremely high NH₃ was measured in August, during which time there were 10 weekly (67-percent) estimated NH₃ concentrations that exceeded 264 μg/L.

SC_avg increased through the sampling season as daily average lake-surface elevation (ELEV_avg) decreased. During the summer sampling season, lake-surface elevation was affected by water loss from evaporation and dam releases, and water gains from snowmelt-derived inflows, groundwater, and irrigation return flows, all of which could have variable SC concentrations and contributions. Higher-than-expected SC_avg (compared to the Julian day average) occurred when ELEV_avg was lower-than-expected (compared to the Julian day average), and lower-than-expected SC_avg occurred when ELEV_avg was higher-than-expected. The relation between SC_avg and DO_avg or pH_avg was negligible when all months of the sampling season were considered; however, in May and June, SC_avg had weak, positive relations with DO_avg (ρ = 0.22 [p < .001]) and pH_avg (ρ = 0.34 [p < .001]), and in September and October SC_avg had moderate, negative relations with DO_avg (ρ = -0.44 [p < .001]) and pH_avg (ρ = -0.56 [p < .001]).

Spatial Relations

The site at Mid-Trench located at the trench along the western shore of Upper Klamath Lake was the deepest of the core sites where depth ranged from 13.2 to 15 meters during the summer sampling period. From 2005 to 2019, the sonde at MDTU had 29 days of WT_ext (40 percent of all sites) and 9 weekly estimated NH₃ concentrations that exceeded 264 μg/L (60-percent of all sites), whereas the sonde at MDTL had 132 days of DO_ext (85 percent of all sites). The high temperatures at MDTU could be due to localized thermal stratification resulting in a “soft” boundary between the upper and lower water column. Although the location of site Mid-Trench is too deep to provide sucker habitat, the extreme WQ there could affect other parts of the lake because of a strong clockwise current observed with prevailing winds. Hydrodynamic studies reported that water flowing through the deepest trench was preferentially routed to the northwestern part of the lake, whereas water nearer the surface was preferentially routed south along the eastern shore.

Overall, the DO_avg time series had similar seasonal patterns between sites each year, although the daily spatial variance in DO_avg across all sites was greater than 4.7 mg/L for 25 percent of all days from 2005 to 2019. The daily spatial variance in August, which was the month with the most days of DO_ext, was greater than 7.2 mg/L for 25 percent of the August days in the study period.

The site at Mid-North was the only core site located in the northwestern part of Upper Klamath Lake and had mid-level depth ranging from 2.4 to 4.2 m through the summer sampling period. Overall, there were fewer days of extreme water quality at Mid-North compared to Mid-Trench, but there were five weekly estimated NH₃ concentrations that exceeded 264 μg/L (33 percent of all sites) at MDNU.

The WMR site, the shallowest core site with a depth ranging from 0.7 to 2.5 m, was located at the outlet of the Williamson River near the restored delta, and the RPT site, with a depth ranging from 1.6 to 3.4 m, was located in the southern part of Upper Klamath Lake, about 10 km south of the WMR site. These two shallowest sites of the core group measured the most days of extreme pH; the WMR site had 61 days of pH_ext (54 percent of all sites) and the RPT site had 27 days of pH_ext (24 percent of all sites). The relatively high number of days of pH_ext at the shallowest sites could be due to the vertical volume within which mixing occurred. At the shallowest sites, the volume was fixed by a solid lower boundary, whereas the deeper sites had variable soft boundaries between the upper and lower water column owing to thermal stratification. Overall, the pH time series had similar seasonal patterns between sites each year, and the daily spatial variance in pH_avg across all sites was less than 0.2 for 90 percent of days in the study period.

Annual Relations

The annual totals of extreme water temperature did not have a trend from 2005 to 2019. The years with the most days of WT_ext were 2005–06, 2009, 2014–15, and 2017–18, when totals ranged from 4 to 23 days. The annual total days of WT_ext had negligible relations with annual total days of pH_ext, DO_ext, or extremely high NH₃ over the 15-year study period.

Most days of pH_ext and DO_ext occurred in the latter half of the study period. The years with the most days of pH_ext were 2007–08, 2012–16, and 2018, when totals ranged from 7 to 28 days, and 85 days of pH_ext (75 percent of total) occurred from 2012 to 2019. The years with the most days of DO_ext were 2005, 2008–09, and 2012–19, when totals ranged from 7 to 25 days, and 119 days of DO_ext (77 percent of total) occurred from 2012 to 2019. The annual total days of DO_ext had a positive relation with annual total days of pH_ext over 2005–2019 (ρ = 0.43 [p = .11]) but had a stronger, positive relation with the previous year’s annual total days of pH_ext, with ρ of 0.63 (p = .015). When the combined total days of pH_ext from the current and previous year were compared to annual total days of DO_ext, the ρ was 0.66 (p = .010). The positive correlation reflected a potential relation between years with greater productivity in the lake, as measured by pH, and oxygen demand the following year. Increased oxygen demand following a high-production year could be due to extensive decay or other processes occurring on or near the lakebed. Future work could evaluate the lagged WQ effects of decay, nutrient release, and oxygen demand, and potential relations to lake circulation and cyanobacteria assemblages, for example.

Like pH and DO, estimated NH₃ had more occurrences of extremely high concentration in the latter half of the study period. The years when at least two weekly estimated NH₃ concentrations exceeded 264 μg/L were 2012, 2016–17, and 2019, and 13 weekly estimates of extremely high NH₃ concentration (87 percent of total) occurred from 2012 to 2019. High NH₃ estimates depended on coincident timing of high total ammonia concentrations and high pH. The days during which estimated NH₃ exceeded 264 μg/L corresponded to total ammonia concentrations that ranged from 312 to 873 μg/L with pH that ranged from 9.0 to 10.0 standard units and estimated fractions of NH₃ greater than 0.3. The increased occurrence of extreme NH₃ concentrations in the latter half of the study period were driven more by increased occurrence of high total ammonia concentrations than increased pH and water temperature. For all sites within the study period, there were more total ammonia measurements higher than 312 μg/L in the second half than the first half of the study period (63 percent from 2012 to 2019), whereas estimated fractions of ammonia greater than 0.3 at the time of sample collection were split more evenly between the first and second halves of the study period (54 percent from 2012 to 2019).

Based on individual WQ variables, different calendar years were considered to have poor water quality. The years that had the most accumulated days of extremely poor WQ (as represented by high temperature, low DO, high pH, or high NH₃) were 2012–15 and 2017. Based on this metric, most days of extremely poor water quality occurred in the latter half of the 15-year study period. The years that had the fewest accumulated days of extremely poor WQ conditions across all variables were 2010 and 2011.