Bloom of the potentially toxic cyanobacterium P. rubescens: seasonal distribution and possible drivers of its proliferation in the Vrutci reservoir (Serbia)

The Vrutci reservoir (43°50ʹ34ʹʹN, 19°41ʹ36ʹʹE) was created in 1984 by damming the Đetinja River in Western Serbia (Fig. 1). It is a temperate, mid-altitude reservoir. The normal operational level of the reservoir was set at 621.3 m a.s.l. The reservoir is elongated in shape and is 7 km long, with a maximum depth (located near the dam) of 60 m (Table 1). The geological substrate consists mainly of limestone and serpentines. The average retention time is estimated at 250 days (JCI 1971). Apart from the Đetinja River, there are 28 other permanent and intermittent tributaries. The Vrutci reservoir is the main source of water supply for the city of Užice. It is also used for flood protection, sediment retention, and low-flow management during periods of drought. The total area of the catchment is 160 km². It is sparsely populated and has no industry (JCI 1971). Mixing patterns in the Vrutci reservoir are highly complex and dependent on meteorological conditions and water withdrawals. Monitoring of the reservoir has been insufficient over the last 20 years and carried out by the officials only once a year.

Figure 1

At the end of December 2013, the upstream part of the reservoir turned into a reddish pulp that spread all over the reservoir in just a few days, which was followed by the formation of red surface patches. The study period lasted from January 2014 to December 2015. Water temperature (WT), pH, conductivity (Cond) and chlorophyll a concentration (Chl-a) were measured in situ monthly (along the entire water column), using a multiparameter probe (YSI 6600V2-2), while water transparency was determined using a Secchi disk. Samples for the analysis of nutrients – ammonium (N-NH₄), nitrites (N-NO₂), nitrates (N-NO₃), total nitrogen (TN), orthophosphate (ortho-P) and total phosphorous (TP) – were collected in January, June, September, October and December 2014, as well as from March to December 2015. Samples were collected below the water surface, in the metalimnion (or at the mid-depth during circulation) and near the bottom, and analyzed at the Jaroslav Černi Institute Water Quality Laboratory, in accordance with the standard procedures and methods (Franson et al. 1975; Eaton et al. 2005; SRPS European Norms 12260:2008).

Phytoplankton samples were collected monthly throughout the study period with the exception of November 2014, January and February 2015, due to the lack of funding during these months. The sampling site was located near the dam and water intake and was referred to as Vodozahvat (Fig. 1). Representative samples were collected from stationary depths: 0, 3, 10, 14, 18, 27, 31, 35 and 43 m. Water samples for qualitative analysis of phytoplankton were collected using a plankton net (Ø 22–23 μm). A vertical (beta) Van Dorn bottle of 2.2 l was used to collect samples for quantitative analysis. Samples were put into plastic bottles and preserved with Lugol’s solution according to European Standards 15204 (2006).

The identification of taxa was performed using a light microscope (Karl Zeiss Imager M1), camera (Carl Zeiss AxioCam MRC5) and fluorescence microscope (Olympus BX 5) according to the standard literature (Huber-Pestalozzi et al. 1983; Starmach 1983; Krammer & Lange-Bertalot 1986; 1988; 1991; John et al. 2002; Komárek & Anagnostidis 1998; 2005; Komárek 2013). The Shannon diversity index (Hʹ) was calculated according to the standard formula, based on the number and biomass of species (Shannon & Weaver 1949).

The quantitative analysis of phytoplankton was based on the Utermöhl method (EN 15204 2006), using a Leica inverted microscope. The number of cells was converted into the phytoplankton biomass by geometric approximations, using the standard formulas (Hillebrand et al. 1999; Sun & Liu 2003), while the biomass data were expressed as mm³ l⁻¹.

Meteorological data (air temperature – AT, wind speed – Wind, precipitation – Prec, insolation and cloudiness) were obtained from annual reports of the Republic Hydrometeorological Institute of Serbia (RHMZ), from the nearest weather station (Zlatibor). Data on the water outflow and water level were obtained from the Užice Public Utility Company.

Complex external drivers produced complex-shaped vertical temperature profiles that deviated from those commonly observed. In most cases, the distinction between seasonal and ephemeral thermocline was not obvious. A modified van Genuchten fitting curve (Rimmer et al. 2005; Jennings et al. 2012) was employed to model the observed temperature profiles. The thermocline was then positioned to match the depth at which the second derivative of the temperature plot equals zero.

To answer the question as to which environmental factors shaped the dynamics of the P. rubescens population, redundancy analysis (RDA) was performed using CANOCO for Windows, Version 5.0 (Ter Braak & Šmilauer 2012). The RDA showed the relationship between the biomass of P. rubescens and other phytoplankton taxa as well as preselected environmental (hydrological, meteorological, physical and chemical) parameters – 11 variables included in total. The linear method was used since the gradient on the first axis was 0.9 SD units long. Before the RDA was performed, meteorological and chemical factors were processed separately using the “interactive forward selection” option, where statistical significance of each variable was tested by the Monte Carlo permutation test (p < 0.05). The significance of all other variables was also tested, so that all variables included in the ordination diagram were significant. The explanatory variables included physical (WT, Cond and Secchi depth – Secchi), hydrological (water flushing – Wflush), meteorological (AT, Prec and Wind) and chemical (NH₄, NO₃, ortho-P and TP) parameters. The terms indicating lake stratification (epilimnion – E, metalimnion – M, hypolimnion – H) and water mixing (C) were used as supplementary variables. In addition, response curves that represent the relationship between P. rubescens biomass and two parameters separately (WT and Wflush) were also included for more detailed analysis.

Variation partitioning with three groups of variables was performed. The first group included physical and hydrological parameters (WT, Cond, Secchi, and Wflush), the second group included meteorological parameters (AT, Prec and Wind) and the third group was represented by nutrients or chemical parameters (NH₄, NO₃, ortho-P and TP). The objective of these analyses was to assess how much variability in our data is described by every group of variables, taking in account their simple and conditional effect. The significance of each fraction was also presented.

Species diversity of the phytoplankton in the Vrutci reservoir, expressed as the Shannon diversity index, was generally low (1.18 in 2014 and 1.20 in 2015) and higher in the epilimnion than in the metalimnion. The phytoplankton biomass was characterized by the dominance of the cyanobacterium P. rubescens throughout the study period, except autumn 2014 and spring 2015 when the highest growth of Bacillariophyta was observed (Fig. 2). P. rubescens volume-weighted biomass was higher in 2014 (max 4.40 mm³ l⁻¹; Fig. 2a) than in 2015 (Fig. 2b). During the winter mixing, the P. rubescens population showed a higher biomass in the upper water layers and its gradual decline with increasing water depth (Fig. 3a). After that, at the beginning of the stratification period (April 2014), a sharp decline of P. rubescens biomass was observed. It is important to note that in the period of the highest abundance of P. rubescens (winter 2014), the water level was 5 to 6 m lower than usual (Fig. 4) and red scum appeared on the surface in areas sheltered from the wind. Immediately after the high floods in May 2014, a large amount of inactive P. rubescens filaments formed the purple-red floating patches on the water surface (May and June). After this extreme event and a large outflow of water, P. rubescens reappeared in the metalimnion during the period of stable summer stratification, at a depth of 3 to 7 m (Fig. 3a). The volume-weighted biomass of P. rubescens during the metalimnetic bloom was twice as low as the biomass during the surface bloom in winter (2.19 mm³ l⁻¹). The maximum volume-weighted biomass of P. rubescens in 2015 was recorded in July (3.95 mm³ l⁻¹; Fig. 2b). After that, the biomass of P. rubescens was characterized by a high value in September at a depth of 6 m (8.54 mm³ l⁻¹) and October (0–3 m; Fig. 3a).

Figure 2

Figure 3

Figure 4

Due to the exceptionally mild winter of 2013/2014, the ice cover was not formed, which induced the monomictic mixing behavior of the Vrutci reservoir. High air temperatures in summer induced warming of surface layers and formation of a stable vertical thermal structure. Maximum epilimnetic water temperatures were recorded in August 2014 and July 2015 (Tables 2 and 3). More pronounced differences in mean temperatures between the water layers were observed in 2015.

In both years, the location of the seasonal thermocline was observed from a depth of about 3 m in spring to a depth of about 12 m in autumn (Fig. 3b). As a consequence of the extreme inflow in May 2014, the temperature of hypolimnetic water increased to about 4°C and the thermocline shifted 3 m upward. The pH value was relatively constant, with slightly lower values in deeper strata of metalimnion and hypolimnion compared to the upper strata (Tables 2 and 3).

In May 2014, heavy precipitation was observed when during four consecutive days (from 13 to 16 May) 110 mm of rainfall was recorded at the watershed gauge station. Almost 25% of the reservoir volume was replenished with surface runoff (Fig. 4). The water level varied in a wide range between 615 and 626.7 m a.s.l., and in over 30% of the study period the water level was higher than the usual water level (621.3 m a.s.l.; Fig. 4). Minimum water flushing values were recorded in October 2014 and July 2015, which indicated a higher outflow than inflow (Tables 2 and 3).

N-NH⁴ concentrations were higher in epi- than in meta- and hypolimnion, and the maximum value was recorded in September 2015 (400 μg l⁻¹; Table 3). N-NO₃ mean concentrations were generally high, especially in meta- and hypolimnion (Table 3), but concentrations started to decrease rapidly in June, from the surface to a depth of 15 m (Fig. 5a). Ortho-P concentration was high in the epi- and metalimnion (Table 3) at a depth ranging from 0 to 10 m from July to September (Fig. 5b).

Figure 5

According to data from two years, P. rubescens was characterized by higher biomass during the mixing period and during stratification (metalimnion; Fig. 6). No correlation between the biomass of P. rubescens and other phytoplankton was observed (Fig. 6). P. rubescens biomass was positively correlated with the hydrological parameter Wflush and N H 4 + . $NH_{4}^{+}.$Slightly negative correlation was observed in the case of Wind, NO₃, WT and AT, and negative correlation in the case of physical parameters Cond and Secchi.

Figure 6

The response curves that include the relationship between the P. rubescens biomass and two parameters separately (WT and Wflush) are presented in Figure 7. It appears from Figure 7a that the maximum value of P. rubescens biomass is observed at a temperature close to 16°C. The biomass of P. rubescens was almost stable at Wflush values ranging from 2 to 8, and the highest biomass was observed at the value of about 2 (Fig. 7b).

Figure 7

According to the variation partitioning, physical and hydrological variables (WT, Cond, Secchi and Wflush – fractions a + d + f + g) explained 21.8% of the variability in our data. Irrespective of the others, this group of variables explained 11.8% of the total variation (fraction “a”). Meteorological parameters (Prec, Wind and AT – fractions b + d + g + e) explained a total of 16.2%, and independently – 6.2% of the variability (fraction “b”). Nutrients (chemical parameters) used in the RDA analysis (N H 4 + , $(NH_{4}^{+},$TP, ortho-P and NO₃ – fraction c + e + f + g) explained 18.8% of the total variability. However, this group of variables with no shared effect with other groups explained 9.1% of the variation in our data (fraction “c”). The total explained variation was 40.8%. Comparing these three groups, physical and hydrological variables explained the highest portion of variability in our data and were followed by chemical (nutrients) and then meteorological parameters. The explained variability and the significance of the above-mentioned fractions are presented in Table 4.