Factors contributing to cucumber odor in a northern USA reservoir

Abstract

Episodic occurrences of cucumber odor caused by the alga Synura petersenii in Meander Creek Reservoir (MCR), northeastern Ohio, USA, are partly attributed to increased water transparency resulting from decreased phosphorus and suspended solids loading from the watershed. The first documented occurrence of nuisance odor levels was in 1984, 52 years after the reservoir was filled. This indicates that previous environmental factors constraining the growth of S. petersenii have been relaxed, probably from changes in the physical–chemical environment in the reservoir caused by changes in land use in the catchment. Reduction in farming since 1950, and diversion of sewage around the reservoir in 1977, reduced suspended solids and total phosphorus loading into the reservoir during the time that the cucumber odors occurred. These observations support the hypothesis that increased transparency of the reservoir, resulting from decreased sediment loading and reduced productivity, has permitted the occasional occurrence of nuisance densities of S. petersenii. Based on available data, pH, iron, and silica do not appear to be key factors regulating growth of S. petersenii in MCR. The transition to lower turbidity and total phosphorus concentrations from catchment restoration actions may increases the risk of S. petersenii blooms and cucumber odor episodes. While the overall benefits of cleaner raw water for water supply may outweigh this risk, it is desirable to understand the factors that promote nuisance growths and take actions to control them.

Key words:

• diatom inferred phosphorus
• land use
• nutrient export
• odor
• reservoir
• sediment core
• Synura petersenii
• taste

Objectionable odors and tastes in finished drinking water, although usually benign, are the primary criteria used by consumers to judge water safety (McQuire 1995). Aside from chlorination, most odor/taste problems are of biological origin, caused by algae, bacteria or fungi (Mallevaille and Suffet 1987, Suffet et al. 1999). Most odors and tastes have more than one source; for example, musty or moldy odors are produced by many species of bacteria, cyanobacteria, fungi, and probably algae (Suffet et al. 1999). However, cucumber odor, a property of the aldehyde E2,Z6 nonadienal (Hayes and Burch 1989), is caused by only a few species of algae. Algae implicated as possible sources of cucumber odor include Synura petersenii (Wee et al. 1994), Uroglenopsis (Mallevaille and Suffet 1987), Scenedesmus subspicatus (Cotsaris et al. 1995) Peridinium (Burlingame 1992) and Uroglena (Palmer 1977). Synura petersenii is the most common cause of cucumber odor in domestic water supplies (Watson et al. 2001).

Nuisance levels of cucumber odor first occurred in water from Meander Creek Reservoir (MCR) in northeastern Ohio in December 1984. Subsequently, nuisance levels of the odor have occurred five times, all during winter, with the last episode in February 2009 (Table 1). During the 2002 odor episode, the causative organism was identified as Synura petersenii based on scanning electron microscope (SEM) discerned scale morphology (L.A. Schroeder, S.C. Martin, Youngstown State Univ., Dept. Civil Engineering, 2002, unpubl.). The abundance of S. petersenii during the 2002 episode correlated with the concentration of E2,Z6 nonadienal in the reservoir water, providing substantive evidence that S. petersenii was the cause of the episodic occurrence of cucumber odor in MCR (L.A. Schroeder, S.C. Martin, Younstown State Univ., Dept. Civil Engineering, 2002, unpubl.). Synura petersenii also was identified during the 2009 episode based on colony and scale morphology with light microscopy. There is no direct evidence that the other episodes of cucumber odor are from S. petersenii; however, all episodes occurred during winter, and scales from S. petersenii were observed to a depth of about 20 cm in the sediments, corresponding to ca. 1995.

Table 1 Occurrence of cucumber odor episodes in Meander Creek Reservoir (M. Kielbasa, Mahoning Valley Sanitary District, 2003, pers. comm.).

Onset of occurrence of cucumber odor
December 1984
8 January 1993
1 December 1996
27 January 1999
23 January 2002
17 February 2009

S. petersenii is a flagellated, planktonic, colonial alga belonging to the class Synurophyceae, whose members are often abundant in the metalimnion of oligotrophic to mesotrophic lakes and reservoirs (Siver 2003). S. petersenii is a strict photoautotroph (Holen and Boraas 1995, Jansson et al. 1996) that has a wide tolerance for many environmental variables (Kristiansen 1975, Nicholls and Gerrath 1985, Siver 1987, 1995).

Despite the wide range of conditions suitable for S. petersenii, this alga was not present in MCR at nuisance densities until 1984, suggesting that changing conditions in the reservoir released the algae from prior limiting factors. Since 1984, nonoptimum conditions have generally prevailed, keeping the population of S. petersenii below nuisance densities. Occasionally, at 3–5 yr intervals, the constraining factor or factors were relaxed and the alga attained nuisance densities, releasing precursors that form E2,Z6 nonadienal. We investigated the changes in reservoir catchment use, historical data for chemico-physical properties of MCR water, and diatom-based inferred total phosphorus (TP) levels for factors that may explain the recent and sporadic occurrence of nuisance cucumber odor in MCR.

Study site

Meander Creek Reservoir, located in the Eastern-Ontario Lake Plain (EOLP) ecoregion, is the primary potable water source for 300,000 people in the Youngstown, Ohio, metropolitan area. The reservoir and the water treatment plant are managed by the Mahoning Valley Sanitary District (MVSD), a public utility. The reservoir was completed in 1931 and first filled in January 1932. An inflatable dam extension was installed in 1995, adding 41 cm to the reservoir depth. The reservoir was 11.2 km long, 1.6 km wide, and had a maximum depth 15.7 m, mean depth 5.1 m and a 221 km² catchment area. Since 1995, the filled reservoir (Fig. 1) had a surface area of 8.77 km² and an impoundment capacity of 42.5 × 10⁶ m³, with water residence time of 188 days. The reservoir was dimictic with a thermocline forming usually 3–4 m below the surface. The land immediately around the reservoir (1600 ha) was planted to red pine (Pinus resinosa). The reservoir and surrounding land is a wildlife refuge and is closed to the public.

Figure 1 Meander Creek Reservoir and catchment. An “x” marks the location of the core samples.

Methods

Sediment cores

Two sediment cores were collected during June 2005 from the deepest part of the reservoir (Fig. 1) using a 5-cm dia piston corer. One core was analyzed for sediment particle size distribution and the other was sectioned for analysis of diatom remains, wet sediment specific gravity, dry matter content, and loss on ignition (LOI). Particle size of the sieved material was determined using a hygrometer method (Das 2001). Specific gravity was determined as the mass of a measured volume of sediment compacted by centrifugation in a calibrated centrifuge tube. Dry matter was determined as the mass remaining after drying at 105 C for 24 hr. Loss on ignition was determined as mass loss after heating to 550 C for 4 hr.

The 144-cm long diatom core was uniformly divided into 1.12 cm sections. Sediment chronology was determined by Daniel Engstrom (St. Croix Watershed Research Station, Science Museum of Minnesota, Marine on St. Croix, MN) from ¹³⁷Cs activity on core sections that bracketed the maximum ¹³⁷Cs activity and extended to a depth of no activity. Diatom sediment subsamples were taken from each section with a #2 cork borer (0.1–0.2 g wet weight), cleaned by concentrated nitric acid, washed with DI water, settled onto 22 × 22 mm glass cover slips and mounted in ZERAX®. Diatom valves were counted with a 100× objective using an Olympus AO 70A microscope. Consecutive 100-μm square quadrats were scored along a transect of the slide until at least 400 valves were counted. Diatom identification was based primarily on the works of Krammer and Lange-Bertalot (1986–1991) and Patrick and Reimer (1966, 1975). Diatom nomenclature primarily follows Stoermer et al. (1999) then Krammer and Lange-Bertalot (1986–1991, (see Supplement for additional details).

Determination of inferred relative total phosphorus

Diatom inferred relative total phosphorus (IRTP) was determined using the weighted averaging method (McCune and Medford 1999). No regional training set includes reservoirs and/or lakes in the EOLP ecoregion of Ohio; therefore, a surrogate training set was constructed from appropriate published training sets. Five training sets were use: (1) Reavie set, 64 alkaline lakes in southeastern Ontario (Reavie and Smol 2001); (2) Bennion set, 32 eutrophic ponds in southeast England (Bennion 1994); (3) Fritz set, 42 lakes in northern and western Michigan (Fritz et al. 1993); (4) Dixit set, 257 lakes and reservoirs in northeastern USA (Dixit et al. 2006); and (5) Ramstack set, 55 lakes in Minnesota (Ramstack et al. 2003). Because of the high variability in the optimum TP among training sets, the data were normalized to the maximum value in each training set. The surrogate training set was then calculated as the mean of the normalized TP optima for each taxon. Therefore, the normalized optimum TP range for each training set and also the surrogate set was between 0 and 1. The resulting inferred TP is a relative value and indicative of magnitude of changes in TP with time but not absolute TP concentrations (see Supplement for additional details).

Estimation of total phosphorus (TP) and suspended solids (SS) loading

Total phosphorus and SS loadings into MCR were estimated from land use and mean export coefficients derived from published data (see Supplement for details).

Results

The core section thickness (1.12 cm) is less than the annual sediment deposition rate of 3.2 cm/yr for upper sections and 1.6 cm/yr for the deepest part of the core. Thus, seasonal differences in deposition rates may increase the variability in data derived from core section analysis: LOI, density, dry matter and diatoms abundance. To compensate for the seasonal source of variability in the core sections, the data were smoothed using the T3524 function in SPSS® (Norusis 1993).

Sediment dry matter and density

The sediment dry matter content increased with sediment depth from 22% near the surface to a maximum of 50% in the deepest sediment, (Fig. 2a). Sediment wet density follows a pattern similar to dry matter content (Fig. 2b). The upper sediment has relatively low density (mean = 1.3 ± 0.03 g/cc, sections 1–57) with peaks at sections 22 and 40 corresponding to maxima in the dry matter content (Fig. 2a and Fig. 2b). The density then increases with sediment depth through the remainder of the core to a maximum of 1.45 g/cc in the deepest layers (Fig. 2b). The dry matter content of each core section increased from 0.25 g/cm²/section at the surface to 0.92 g/cm²/section at the bottom of the core (Fig. 2c).

Figure 2 Changes in sediment physical properties and ¹³⁷Cs activity with sediment depth.

Chronology

Four time markers establish the chronology of core deposition. The top of the core was deposited in 2005 and the bottom in 1932. Maximum ¹³⁷Cs activity occurred in section 83, representing 1963, and last detectable activity in section 98, representing ca. 1950 (Fig. 2d). Aerial annual deposition rates (AAD) were calculated for three of the core regions delimited by the ¹³⁷Cs chronology: sections 1–83, 84–102 and 103–139 (Table 2). The AAD increased with core depth from 0.93 g/cm²/yr for the upper region (2005–1965) to 0.98 g/cm²/yr for the mid region (1964–1950) and 1.2 g/cm²/yr in the deepest region (1951–1932). The time span represented by each core section was estimated from the linear regression of AAD on core section and the measured dry weight content of the section (see Supplement for additional details). Although the change in deposition rate is likely nonlinear, the estimates of sediment age based on AAD and measured dry matter content of each section are likely more accurate than those based on a constant deposition rate.

Table 2 Mean dry matter deposited during time intervals delineated by the ¹³⁷Cs chronology (n = core section number from top of core).

Interval core section (n)	Mid-point section (n)	Time range (yr)	Interval (yr)	Dry Wt (g/cm^2)	Mean (g/cm^2/yr)
1–83	42	2005–1964	42	39.2	0.932
84–102	93	1963–1950	13	12.8	0.985
103–129	116	1949–1932	17	20.3	1.196

Loss of weight on ignition (LOI)

Organic content of the sediments (LOI) decreased steadily with sediment age, from 7.5% of dry mass in 2005 to 5.5% in 1955, then formed a minimum between 1950 and 1942 and increased to 6% from 1946 through 1932 (Fig. 3a). There was no evidence of organic matter from terrestrial sources in the core (e.g., tree leaves and twigs); thus, the organic content of the sediment is attributed predominantly to reservoir productivity. The pattern of change in organic matter depends on both the productivity of the reservoir and the rate of dilution by inorganic material from the catchment. The pattern is consistent with the decreased rate of inorganic deposition with time (see suspended solids export, below); the lowest organic content occurred in the deepest sediment, which also had the greatest rate of inorganic deposition.

Figure 3 Change in organic content, sediment particle size and diatom abundance with time of deposition. Occurrence of odor episodes are marked with an “o.”

Sediment particle size

Coarse silt and sand (particle size 8–74 μm) dominated the sediment from 1932 to ca. 1957. Between 1957 and ca. 1980 sediment particle size changed to predominantly fine silt and clay (size < 8 μm) and remained clay–silt through 2005 (Fig. 3b). All odor episodes occurred after the transition from coarse to fine sediment size (Fig. 3b).

Diatom abundance and composition

Diatom valve production remained relatively low from 1932 to ca. 1973 at about 6 × 10⁶ valves/cm²/yr, then increased to a maximum of about 23 × 10⁶ valves/cm²/yr by 1987 and remained relatively high to the core top (2005) during the period when odor episodes occurred (Fig. 3c).

The relative abundance (RA, the ratio of number of valves for a taxon to the total valves counted) of the diatom assemblage changed from predominantly Asterionella formosa, (RA = 0.15) and Stephanodiscus hantzschii (RA = 0.4; Fig. 4a, Fig. 4d) prior to 1985 to predominantly Cyclotella ocellata (RA = 0.4) and Aulacoseira sp1 (RA = 0.2) (Fig. 4b and Fig. 4c) after 1985. The decrease in A. formosa, Stephanodiscus hantzschii (Fig. 4a and Fig. 4d) and increases in C. ocellata and A. sp1 (Fig. 4b and Fig. 4c) roughly correspond to the onset of cucumber odor in MCR. Stephanodiscus hantzschii has a high, and A. formosa moderately high, relative TP optima (0.62 and 0.35, respectively). Cyclotella ocellata and A. sp1 (A. cf alpigena) have relatively low optimal TP requirements (0.23 and 0.21, respectively), which is consistent with increased RA as TP decreases. The TP optimum for A. sp1 is not available, but a similar species, Aulacoseira alpigena, has an optimum relative TP value of 0.2l. Changes in relative abundance of diatom taxa with sediment depth imply changes in water chemical and/or physical characteristics. Patterns of changes in diatom community composition and abundance correspond with the occurrence of odor episodes in MCR. Both patterns are consistent with decreased nutrient (TP) levels during the period that odor episodes were observed.

Figure 4 Change in relative abundance of diatom valves with time of sediment deposition: Occurrence of odor episodes are marked with an “o.”

Water chemistry

The MVSD has maintained water chemistry records from the beginning of water treatment plant operations in 1932. Although some of the records, particularly between 1984 and 1995, are missing, the available data are useful in assessing the trends in reservoir conditions. The pH of MCR has remained relatively constant (geometric mean 7.67 ± 0.23; Fig. 5a). The pH decreased during the first 5 years after construction to the lowest value (7.1) then increased steadily to a maximum of 8.05 ca. 1972, then decreased to 7.7 at present (Fig. 5a). Free carbon dioxide was not monitored after 1983. The highest CO₂ values of 5 ppm were recorded shortly after the reservoir was filled and declined to 2 ppm by 1980 (Fig. 5b). Total alkalinity trended higher from 45 ppm shortly after reservoir filling to 65 ppm (CaCO₃ equivalents) by 1995 (Fig. 5c). Total iron was recorded between 1950 and 1982. Iron content was relatively high, 40 ppb, until 1949 then decreased to 20 ppb by 1983 (Fig. 5d). Turbidity was high following reservoir filling, > 10 NTU, then declined to 4 NTU by 1980 and remained relatively low for the remainder of the study period (Fig. 5e).

Figure 5 Meander Creek Reservoir water analysis data: pH, free carbon dioxide, total alkalinity, iron, turbidity and turbidity. Data are from Mahoning Valley Sanitary District, blank space = no available data.

There are no early direct measures of transparency in MCR; however, reservoir turbidity (tb) is correlated with Secchi disk transparency (sd in m; Equation 1). The regression equation for transparency as a function of turbidity in MCR, calculated from measurements during three winter seasons, 2000–2003 (L.A. Schroeder, S.C. Martin, Youngstown State Univ., Dept. Civil Engineering, 2003, unpubl), is:

Secchi disk transparency calculated from the MVSD turbidity data using Equation 1, was relatively low (annual mean = 0.97 ± 0.3 m) prior to 1950 then increased to a mean of 2.1 ± 0.8 m after 1970 and occasionally > 3 m (Fig. 7e).

Water chemistry including TP, silica, pH, conductivity, turbidity, suspended solids (SS) and transparency were monitored during the winter seasons 2000–2003 (Table 3, data from L.A. Schroeder, S.C. Martin, Youngstown State Univ., Dept. Civil Engineering, 2003, unpubl.).

Table 3 Summary of chemical/physical data for Meander Creek Reservoir for the winter seasons of 2000–2003.

ANALYTE	N	Mean	SD
Total phosphorus (ppb)	48	27	17
Silica (ppm)	36	2.4	1
pH	48	7.8	0.3
Conductivity (μS)	48	540	40
Turbidity (NTU)	48	4.1	2
Suspended solids (ppm)	48	4	3
Transparency (Secchi disk, m)	48	2.0	0.7
N = number of determinations; Mean = mean of all data for the site nearest where the sediment core was taken; SD = standard deviation.

Land use patterns

Three major changes in the 221 km² MCR catchment have occurred since reservoir construction: (1) total farmland decreased but the proportion planted to soybeans increased; (2) the human population increased with the concomitant increase in residential area; and (3) the sewage effluent from the City of Canfield was diverted around the reservoir in 1977. Land use in the catchment in 1932 when MCR was completed was: water 3.5%; residential 1.4%; woods and shrub 33%; and farmland 62%. By 2005 land use was: water 4%; residential 16%; shrub and woods 53%; and farmland 27% (NASS 2006 and extrapolated from Christou 2002). There is little industry, commercial or dense urban development in the catchment.

Residential:

Population changes in the catchment were estimated from the U.S. 10-year township census data weighted for the proportion of each township included in the catchment (Fig. 6). The catchment population grew exponentially, with a doubling time of about 18 yr between 1930 and 1970, then more slowly during the 1970s, and even declining during the 1980s. The population growth resumed during the 1990s (Fig. 6). Residential area of the catchment increased from 10.3 km² in 1965 to 30.0 km² in 1994 (measurements from aerial photographs, Christou 2002). Extrapolating from these data, the residential area increased from 3.7 km² in 1950 to 14 km² by 1970 and 45 km² in 2005 (Fig. 7a). Most of the residential area in the catchment is in and around the City of Canfield (Fig. 1). Prior to 1977, all sewage from the city was treated at the Canfield waste water treatment plant (WWTP) and the effluent discharged into Sawmill Creek, a tributary that flows into the southern end of MCR (Fig. 1). The Canfield WWTP was closed in 1977, and the raw sewage from Canfield was diverted around MCR to a county WWTP downstream.

Figure 6 Change in population in the Meander Creek Reservoir catchment.

Figure 7 Factors affecting the growth of Synura petersenii in Meander Creek Reservoir (MCR): (a) change in farmland and residential area on MCR catchment; (b) estimated total phosphorus export from MCR catchment; (c) estimated suspended solids export from MCR catchment; (d) relative inferred total phosphorus (IRTP) and (e) Secchi disk transparency (sd) calculated from measured turbidity. Odor episodes are marked by an “o.”

Agriculture:

Farmland area (NASS 2006) of the catchment steadily decreased after ca. 1945 (Fig. 7a). The most rapid change in farmland occurred between 1945 and 1965 when farmland decreased by nearly 50% (Fig. 7a). Changes in catchment land use are reflected in the loading of nutrients and suspended materials entering the reservoir.

The SS loading is dominated by agricultural land use (Fig. 7c). Suspended solids export was high until 1950 then decreased rapidly, in concert with the decrease in farmland, to relatively low values by 1970. The decline in farmland corresponds to the change in particle size of the sediment from predominately coarse silt to fine silt and clay (Fig. 3b). This is in contrast to the TP export, where the rapid decrease occurred after 1977 in response to completion of the sewage bypass. After 1990, the TP export from increasing residential area became substantial. All of the recorded incidents of cucumber odor occurred after the closing of the WWTP, and the marked decrease in TP, and during a period with relatively low SS loading (Fig. 7c). The pattern of change in SS export (Fig. 7c) with time corresponds to changes in measured turbidity in the reservoir. Turbidity was high during the period of high farming activity (1932–1950) then decreased through about 1980, remaining low until 2005 (Fig. 5e).

Inferred relative total phosphorus (IRTP)

Inferred relative total phosphorus increased steadily from 0.34 after reservoir filling in 1932 to a maximum of 0.47 in 1977 then decreased to a minimum of 0.32 ca. 1995 (Fig. 7d). The decrease in IRTP beginning in 1977 corresponds to completion of the sewage bypass by the City of Canfield. At the time of the first odor episode in 1984, IRTP was 0.37; during the subsequent episodes between 1995 and 2002, the IRTP was at a minimum of about 0.32 (Fig. 7d). Changes in IRTP reflect the changes in TP export, especially from 1932 to about 1990. After ca. 1990, increases in TP export do not correspond to changes in IRTP (Fig. 7b and Fig. 7d), possibly due to over-estimation of phosphorus export from the developing urban areas. The mean TP measured during winter seasons 2000–2003 was 27 ppb, N = 47 (Table 3). If TP remains proportional to IRTP then the IRTP cardinal points correspond to TP of: 29 ppb following reservoir filling, 1932; 40 ppb at maximum TP, 1977; 31 ppb for first odor episode, 1984; and about 27 ppb during the remaining odor episodes.

The decrease in SS (Fig. 7c) and TP (Fig. 7b and Fig. 7d) is manifested in increased transparency (Fig. 7e). Transparency, calculated from turbidity data (Fig. 5e), increased slowly from 1932 to ca. 1977 then increased rapidly and remained relatively high until 2005 (Fig. 7e). The change in transparency is consistent with changes in TP and SS export and with the changes in IRTP. All odor episodes occurred during the period of relatively high transparency.

Discussion

Six episodes of cucumber odor in water from MCR have occurred between December 1984 and 17 February 2009. The synurophyte, S. petersenii has been identified as causing the cucumber odor of the 2002 and the 2009 episodes and presumably is the causative agent for the previous episodes (L.A. Schroeder, S.C. Martin, Youngstown State Univ., Dept. Civil Engineering, 2002, unpubl.). The relatively recent (1984) first odor episode implies that conditions in the reservoir changed in a manner that released S. petersenii from prior limiting factors. The sporadic nature of the nuisance odor implies that constraining factors generally prevail but occasionally are relaxed, permitting the alga to attain nuisance densities. Thus, an explanation for the occurrence of cucumber odor in MCR must include a general change in one or more factors that permit the growth of S. petersenii to nuisance densities and recognize that these factors are only occasionally effective.

Several environmental factors have been implicated as affecting, or at least corresponding with, growth of S. petersenii populations, including: pH (Siver and Hammer 1989); phosphorus (Siver and Hammer 1989, Rashash et al. 1995); conductivity (Dixit et al. 2000); light (Rashash et al. 1995, Paterson et al. 2004); silica (Sandgren et al. 1996); free carbon dioxide (Saxby-Rouen et al. 1998); and iron (Guseva 1935, cited in Nicholls 1995).

Scaled chrysophytes in general and the synurophyte S. petersenii in particular have recently increased in abundance in many boreal lakes in North America (Paterson et al. 2004). The recent increases were attributed to “broadly applicable anthropogenic effects” but not to changes in pH (acid precipitation), TP, or catchment disturbances (Paterson et al. 2004). They proposed that climatic changes or synergism between climatic changes and acidification cause an increase in lake transparency resulting in increased productivity in deeper layers of the lakes. Because many colonial scaled chrysophytes and synurophytes, including S. petersenii, are most prevalent in the metalimnion (Paterson et al. 2004 and references therein), they would benefit from greater lake transparency.

The silica requirement for S. petersenii is low, less than for most diatoms (Klaveness and Guillard 1975, Sandgren et al. 1996). Growth rate of S. petersenii was reduced in media containing < 1.0 μM silica (Klaveness and Guilliard 1975). The mean silica concentrations of MCR measured during winter 2002 (2.4 ppm = 40 μm/l; Table 3) is 40 times the threshold for reduced growth of S. petersenii (Klaveness and Guilliard 1975); therefore, silica is not likely a limiting factor for the growth of S. petersenii in MCR.

The use of cytochrome c as an electron donor by S. petersenii results in relatively high requirements for iron (Raven 1995). Synura petersenii growth was enhanced by the addition of iron to Hall Lake water (Munch 1972 cited in Sandgren 1988). Guseva (1935 cited in Nicholls 1995) found that iron was limiting to growth of S. petersenii below about 1.2–1.4 ppm. The measured iron content in MCR (mean = 0.3 ppm between 1951 and 1983; data from MVSD) is considerably less than the reported limiting concentration (Guseva 1935 cited in Nicholls 1995); however, the iron content of MCR was decreasing rather than increasing prior to the first occurrence of cucumber odor (Fig. 5d). The trend in iron concentration is consistent with expectations from changes in watershed use (i.e., decreased SS and TP loading). Thus the episodic increases in S. petersenii are not likely due to relief from iron limitation.

Synura petersenii has been described as pH indifferent (Kristiansen 1975, Takahashi 1978) and is found over a wide range of pH conditions (Roijackers and Kessels 1986). The optimum pH reported for S. petersenii (i.e., the mean pH weighted for abundance) ranges from 6.4 for 105 Sudbury Ontario lakes (Dixit et al. 2002) to 7.2 for 146 northeastern USA lakes (Dixit et al. 1999). Synura petersenii abundance and distribution was not correlated with pH among 30 Sudbury Ontario lakes (Dixit et al. 1988), and pH was not implicated in the recent increase in S. petersenii abundance in boreal lakes of Canada (Paterson et al. 2004). Changes in abundance of S. petersenii in five Connecticut lakes were not attributed to pH, but to changes in land use that affected conductivity and alkalinity (Marsicano and Siver 1993). Laboratory cultures of S. petersenii showed no significant difference in growth rates over a pH range of 5.5 to 8.5 (Wee et al. 1991); however, pH and conductance were the dominant variables for axis 1 in a Principal Component Analysis (PCA) of environmental factors affecting the distribution of S. petersenii in northeastern U.S. lakes (Siver and Hammer 1989). Although the decrease in pH in MCR between 1972 and 2005 includes the period when cucumber odor episodes occurred and is consistent with more favorable pH conditions for the growth of S. petersenii, it is unlikely the causative factor for the episodic occurrence of cucumber odor in MCR water. The pH of MCR (7.8; Table 3) remains well within the range for occurrence of S. petersenii, and prior to the odor episodes there were periods when the pH was considerably lower (pH = 7.2, 1933–1940; Fig. 5a) with no recorded reports of cucumber odor episodes.

The inability of S. petersenii to utilize bicarbonate as a carbon source for photosynthesis (Saxby-Rouen et al. 1998) limits their distribution to lakes, or regions within lakes, with sufficient concentrations of free CO₂ to support growth. Because free CO₂ is inversely related to both alkalinity and pH, as described by the Henderson-Hasselbalch equation (see Hutchinson 1957), S. petersenii is indirectly sensitive to alkalinity and pH. This is consistent with the optimum pH for S. petersenii of 6.7 and the observation that all 11 incidents of cucumber odor problems in Ontario, noted by Nicholls and Gerrath (1985), were in Canadian Shield lakes of low alkalinity (< 45 mg CaCO₃ /L).

In soft water lakes with low alkalinity, pH must be low enough to provide sufficient free CO₂ for S. petersenii growth; however, in lakes with moderately high hardness and bicarbonate alkalinity, sufficient free CO₂ can be available even at relatively high pH. The geometric mean pH in MCR is 7.8, but, because of the high total inorganic carbon, the free CO₂ is relatively high at about 2 ppm. Free CO₂ in MCR does not vary in a manner consistent with the recent sporadic winter occurrence of cucumber odor; therefore, we do not expect that free CO₂ is limiting S. petersenii growth in MCR.

Cucumber odor episodes in MCR (1984–1999) occurred during a period of inferred TP minima, implicating TP as a factor related to the abundance of S. petersenii. Synura petersenii were more prevalent in Experimental Lake 227 prior to experimental nutrient enrichment with nitrogen and phosphorus (Zeeb et al. 1994). Synura petersenii occurred in substantial numbers after TP was lowered by 90% by dredging and other restoration procedures in Lake Trummen, Sweden (Cronberg 1982). Recent increases in S. petersenii in two oligotrophic lakes, Russell and Willard Ponds, NH, were attributed to watershed disturbances but could not be directly linked to changes in pH, chloride or phosphorus (Dixit et al. 2001).

By contrast, anthropogenic effects including eutrophication have been proffered as a cause for recent increases in S. petersenii in Joe's Pond, VT (TP = 7.9 ppb) and Kenoza Lake, MA (TP = 13 ppb), while an increase in S. petersenii in the third lake, French Pond, NH (TP = 14 ppb) was attributed to increases in electrolytes and TP (Dixit et al. 2000). Marsicano and Siver (1993) attributed recent increases in S. petersenii in sediments of lakes from northeastern U.S. to increased TP. Principle Component Analysis of the abundance of S. petersenii and environmental factors in 62 lakes (TP = < 10 to > 40 ppb) from New York and New Hampshire revealed a high correspondence of TP (second PC axis) to the abundance of S. petersenii (Siver and Hammer 1989).

The inconsistent response of S. petersenii populations to changes in nutrient levels in fresh water is also characteristic of other chrysophytes populations (Nicholls 1995). Recent widespread increase in S. petersenii in oligotrophic and mesotrophic boreal lakes of North America could not be attributed to TP enrichment (Paterson et al. 2004). The requirement for phosphorus, as determined in laboratory cultures, is lower than for most other algae (Sandgren 1988, Sandgren et al. 1996). The TP in MCR, measured during winter seasons 2000–2003, was 27 ppb (Table 3), which is considerably greater than the weighted average TP of 17 ppb from lakes in northeastern U.S. where S. petersenii were present (Siver and Hammer 1989).

There is no evidence for inhibition of S. petersenii by high concentrations of TP (Rashash et al. 1995). High TP concentrations may indirectly suppress S. petersenii populations by stimulating growth of competing algae that limit light availability in the deeper water where S. petersenii and other colonial scaled synurophytes and chrysophytes are commonly found (Fee et al. 1977, Sandgren 1988). The recent occurrence of S. petersenii in MCR is consistent with this hypothesis.

As farming activities on MCR catchment declined (Fig. 7a), the SS load to the reservoir decreased (Fig. 7c). The decrease in farming along with completion of a sewage effluent bypass in 1977 reduced nutrient loading to the reservoir (Fig. 7b), confirmed by the pattern of IRTP (Fig. 7d). The reduced TP and SS increased reservoir transparency during the time that odor episodes were observed (Fig. 7e), consistent with the hypothesis that S. petersenii were consistently light limited prior to ca. 1980 and periodically light limited after ca. 1980. Superimposed on the increased light penetration are physical vagaries affecting light penetration on a shorter time scale that may be caused by snow covered ice, irregular runoff events that temporarily increase turbidity or prolonged cloudy periods. These short-term vagaries in light penetration provide a mechanism for the sporadic occurrence of nuisance levels of S. petersenii in MCR.

The onset and duration of ice cover, and depth and character of snow cover in particular, may affect the sporadic growth of nuisance levels of S. petersenii. The onset and duration of ice cover on MCR is quite variable, ranging from early December to mid-January with duration of a few weeks to several months (L.A. Schroeder, S.C. Martin, Youngstown State Univ., Dept. Civil Engineering, 2000, unpubl.). Ice cover eliminates wind mixing, resulting in reduced SS with a concomitant increase in transparency. Snow cover on the ice reduces light penetration, perhaps favoring growth of low-light tolerant species (e.g., S. petersenii). When ice melt occurs after prolonged ice cover, especially when not accompanied by a major runoff, the nascent S. petersenii populations may be released to produce nuisance density populations. This is consistent with the correspondence between January snow melt and odor episodes in MCR (L.A. Schroeder, S.C. Martin, Youngstown State Univ., Dept. Civil Engineering, 2000, unpubl.).

The phenomenon of nuisance growth of S. petersenii triggered by increasing reservoir transparency would most likely occur in dimictic, meso- to eutrophic reservoirs. As transparency increases, S. petersenii are released from the limitation of low light intensity and may form nuisance level populations, if other conditions are favorable. In reservoirs with moderate bicarbonate, such as MCR, subsequent phytoplankton growth in the epilimnion would reduce free CO₂ (Wetzel 1983), probably limiting the duration of light-induced S. petersenii growth. In oligotrophic reservoirs with greater free CO₂ and high transparency (e.g., Cannon Reservoir, on a branch of the Delaware River in New York), the duration of S. petersenii problems may persist for longer periods (Burlingame et al. 1992).

Synura petersenii in MCR present a paradox where improving water quality, as assessed by increased water transparency, was accompanied by instances of excessive growth of a nuisance alga, S. petersenii. This is, of course, not a valid argument for reducing watershed management efforts. Water treatment plant managers reported that only 7% of taste/odor problems were attributed to S. petersenii, while over 56% were caused by cyanobacteria (Knappe et al. 2004). Cyanobacteria and other problem algae are most common in eutrophic conditions and could be alleviated by catchment and reservoir management that reduces nutrient loading. The increased cost of water treatment to remove excess algae and odors formed in eutrophic reservoirs likely exceeds the additional cost of removing the occasional cucumber odor that may result from S. petersenii growth in reservoirs of improving water quality (i.e., increased transparency).

The threshold concentration for detection of cucumber odor is low, from 2 ng/L (Burlingame et al. 1992) to 60 ng/L of E2,Z6 nonadienal (Young et al. 1996), corresponding to a S. petersenii cell density of about 20 cells/ml (Rashash et al. 1995, Watson et al. 2001). Peak density of S. petersenii observed in MCR was 300 colonies/ml (21,000 cells/ml). The corresponding concentration of E2,Z6 nonadienal was 64 ng/L in suspended particulate matter on 22 January 2002 (L.A. Schroeder, S.C. Martin, Youngstown State Univ., Dept. Civil Engineering, 2002, unpubl.). The aldehyde E2,Z6 nonadienal is found only at very low concentrations in living cells (Wee et al. 1994, Rashash et al. 1995); cucumber odor forms primarily after cell death and lysis.

Cucumber odor can be readily removed from water supplies by activated carbon adsorption (Burlingame et al. 1992) or chlorine (Burlingame et al. 1992, Watson et al. 2001). In situ treatment with algaecides is more problematic. During summer thermal stratification, S. petersenii are most often found below the thermocline and therefore are less susceptible to surface application of algaecides (Sung 1990). During thermal turnover when algae are vulnerable, the entire reservoir water column must be treated, requiring larger dosages of algaecides (Sung 1990). Often S. petersenii produce odor problems during winter and early spring when the reservoir is wholly or partially ice covered, making application of algaecides difficult. If algaecide treatment is successful, the death of S. petersenii can be expected to result in a temporary increase in cucumber odor.

Acknowledgments

Dr. Eduardo A. Morales, Curador, Herbario Criptogámico, Universidad Católica Boliviana San Pablo, Cochabamba, Bolivia and research scientist at the Philadelphia Academy of Sciences, Philadelphia, PA, helped with identification of diatoms. Dr. Gary Walker, Department of Biological Sciences, Youngstown State University (YSU), gave valuable advice on microscopy and microphotography and permitted use of his dark room and dark room facilities. Carl Leet III, YSU Media Center, provided resources and help digitizing film negatives and advice on microphotography. John Bralich, Center for Urban and Regional Studies, YSU, provided the census data and the map of MCR catchment. Ellen Wakefield-Banks at the YSU library was exceptionally helpful in tracking literature sources. Ed McCormack, Joe Paris and Martin Kielbasa of the MVSD provided access to MVSD data and logistical support for field work. We are especially appreciative to the editor of LRM and to the several reviewers for their insightful and constructive suggestions. Funding was provided by the MVSD and YSU Research Council, grant number 2004–2005 #1.