Phytoplankton Composition and Functional Groups in Cascade Hydropower Reservoirs of the Drina River (Bosnia and Herzegovina): Trophic Status and Ecological Potential Assessment

Abstract

Cascade reservoirs on the Drina River (Bosnia and Herzegovina) are heavily modified water bodies that require reliable biological tools for assessing trophic status and ecological potential. Under the Water Framework Directive (WFD), assessments of surface water ecological status and potential rely on biological quality elements, since aquatic communities integrate and respond to prevailing environmental conditions and thus serve as reliable indicators of water quality. This study aims to (i) describe phytoplankton diversity, biomass, and functional-group composition along the Drina reservoir cascade, (ii) examine monthly changes across the studied reservoirs, (iii) determine trophic status and ecological potential, and (iv) provide a preliminary estimate of total phosphorus thresholds that may support future setting of ecological potential boundaries. Phytoplankton composition and functional groups were analysed in three longitudinally connected reservoirs of the Drina River during four monthly surveys in 2024. A total of 80 phytoplankton taxa were recorded, with diatoms dominating most of the study period. The highest biomasses were recorded for Fragilaria crotonensis, Dinobryon divergens, Acanthoceras zachariasii and Sphaerocystis sp., while the dominant functional groups were P, E, A, and F. Phytoplankton assemblage structure showed moderate spatial differentiation among the reservoirs. Mean chlorophyll a and Carlson’s Trophic State Index indicated eutrophic conditions in the Višegrad Reservoir and mesotrophic conditions in the Perućac and Zvornik reservoirs, while biomass showed a pronounced summer maximum, particularly in Perućac. Ecological potential was generally classified as good or better, except for a moderate classification in the Zvornik Reservoir in late summer. The good/moderate TP boundary was estimated at 39 µg L⁻¹, linking EQR-based ecological assessment with the onset of eutrophic conditions. Overall, this study represents the first application of the phytoplankton functional group approach in cascade reservoirs in Bosnia and Herzegovina and may provide a valuable basis for the development of a phytoplankton-based monitoring framework in lakes and reservoirs, which is currently lacking.

1. Introduction

Water reservoirs support integrated water resources management through flow regulation, flood control, hydropower generation, drought mitigation, and a range of economic activities, including tourism, recreation, aquaculture, and fisheries [1,2]. In contrast, dams fragment rivers, leading to substantial habitat alteration and changes in flow and sediment regimes [3], thereby posing a major threat to freshwater biodiversity [4]. Approximately 90% of global river volume is already affected by fragmentation [5].

Despite their apparent similarity to natural lakes, reservoirs differ fundamentally because they integrate riverine conditions in inflow zones with lacustrine features near the dam [6]. Water quality in reservoirs is shaped by interacting environmental and anthropogenic drivers, and excessive nutrient and pollutant inputs from the catchment can trigger progressive degradation. In combination with reduced flow velocity, prolonged water residence time, and increased exposure to solar radiation, these conditions make reservoirs particularly susceptible to eutrophication. Although eutrophication may occur naturally, it is often accelerated by human activities [7]. Its consequences include enhanced primary production, proliferation of algae and cyanobacteria, reduced transparency, oxygen depletion (anoxia), toxic blooms, and gradual biodiversity loss [8].

Water quality of the reservoir exhibits pronounced spatial and seasonal variability, highlighting the need for systematic ecological assessment and continuous monitoring. In this context, phytoplankton is a key biological quality element and a sensitive indicator of trophic conditions in lentic ecosystems. As primary producers in the water column, phytoplankton communities respond rapidly to changes in nutrient availability and physical and chemical conditions [9]. Accordingly, shifts in species composition, abundance, biomass, and functional structure can provide early signals of eutrophication pressures, including within assessment frameworks aligned with the EU Water Framework Directive [10].

Modern trophic assessment systems increasingly use functional classifications that link phytoplankton community structure to environmental constraints and eutrophication pressure [11,12]. Based on this concept, several phytoplankton indices have been developed, notably the Q phytoplankton index [13] and the Hungarian Lake Phytoplankton Index (HLPI) [14], which were successfully applied in both natural and anthropogenic lentic ecosystems. These indices integrate functional composition with biomass-related metrics and can support robust assessment in systems with strong seasonal dynamics [15]. Integrative phytoplankton metrics have been successfully applied in artificial reservoirs, capturing spatial and temporal variability in the community structure and often aligning with trophic state assessments [16,17] and water column stability in cascade reservoirs [18].

Specifically, shifts in functional groups due to stratification mixing dynamics underscore the enhanced diagnostic benefits of functional approaches compared to solely using biomass metrics for reservoir evaluation [17,19]. While phytoplankton-based ecological assessment is not equivalent to trophic state classification, the two can be interpreted in a comparable way when supporting physical and chemical elements and type-specific thresholds are clearly defined. Although phytoplankton functional groups have been applied in individual artificial reservoirs in Europe [16], comparable studies on large, longitudinally connected cascade reservoirs remain much less common and are documented mainly outside Europe, particularly in China [17,18].

Phytoplankton in the freshwaters of Bosnia and Herzegovina remains poorly investigated. Published phytoplankton studies of Blidinje Lake [20] have mainly relied on chlorophyll a concentration, total cell counts, and saprobity index calculations, whereas biomass-based assessments of trophic status have been largely unexplored. Phytoplankton is formally recognised as a biological quality element in the entities’ water-management framework, but its assessment is still not fully operationalised in accordance with European Water Framework Directive principles [21]. For lakes, currently available criteria rely mainly on chlorophyll a and supporting physico-chemical parameters, whereas values for heavily modified and artificial water bodies are still to be defined after their formal designation [22].

We selected three reservoirs along the Drina River in Bosnia and Herzegovina constructed for cascade hydropower exploitation to investigate phytoplankton community structure, algal biomass, functional groups and to demonstrate the applicability of certain metrics for trophic status evaluation and ecological potential assessment in Bosnia and Herzegovina. We proposed a total phosphorus threshold as an additional pilot contribution intended to support future boundary-setting efforts.

The objectives of this study are to (i) assess species diversity and perform qualitative and quantitative analyses of algae and cyanobacteria within phytoplankton functional groups in the three selected reservoirs, (ii) explore monthly phytoplankton dynamics across the cascade of reservoirs, (iii) determine the trophic status and ecological potential of reservoirs, and (iv) provide a preliminary estimate of total phosphorus thresholds that may support future setting of ecological potential boundaries.

2. Materials and Methods

2.1. Study Area

The Drina River is a 346 km long system, with a catchment area of 20,319.9 km² [23] and a mean discharge at its confluence of approximately 400 m³ s⁻¹ [24]. It is formed by the confluence of the Tara and Piva Rivers near Šćepan Polje at 432 m a.s.l., and it is the largest tributary of the Sava River (Black Sea basin), joining the Sava near Sremska Rača at 75.4 m a.s.l. The mean annual precipitation in the Drina River basin is 1100 mm, while the mean annual air temperature ranges from 10.5 to 11.1 °C in the northern part of the basin and from 4 to 5 °C in the southern part [25].

In the present study, three large reservoirs longitudinally arranged along the Drina River were investigated: Višegrad Reservoir (V), Perućac Reservoir (P), and Zvornik Reservoir (Z) (Figure 1,

Table 1. Sampling locations and basic hydrological parameters of the studied reservoirs on the Drina River. The data for volume, Qav, dam height, dam type, installed capacity, and mean annual electricity generation were sourced from [26].

Reservoirs	Višegrad (V)	Perućac (P)	Zvornik (Z)
Longitude of sampling sites (WGS84)	19°16′22.66″ E	19°21′31.77″ E	19° 6′43.40″ E
Latitude of sampling sites (WGS84)	43°45′19.49″ N	43°58′25.92″ N	44°21′39.78″ N
Elevation (a.s.l.) Elevation (m a.s.l.)	341	290	153
Ecoregion/Subecoregion	Dinaric/Continental	Dinaric/Continental	Dinaric/ Peripannonian
Surface area (km2)	8.5	12.5	8.1
Catchment area (km2)	13,934	15,308	17,423
Reservoir volume total/active (106 m3)	161/105	340/218	47.43/21.32
Maximal depth (m)	43	60	39
Length (km)	40	55	25
Average discharge Qav (m3 s−1)	342	349	369
Year of dam construction	1989	1966	1955
Dam height (m)	79.5	90	42
Dam length (m)	280	460	166.5
Dam type	Concrete	Concrete	Concrete
Installed capacity (MW)	333	96	368
Mean annual electricity generation (GWh)	1010	500	1650

). The reservoirs have existed for 37, 60, and 71 years, respectively. Višegrad Reservoir is an artificial storage reservoir (surface area 8.1 km², volume 161 × 10⁶ m³, length 40 km, maximum depth 43 m) created by the construction of the Višegrad Hydropower Plant. Perućac Reservoir (surface area 12.5 km², volume 340 × 10⁶ m³, length 55 km, maximum depth 60 m) was impounded to support the Bajina Bašta run-of-river hydropower scheme. Most of the reservoir lies within the Drina River canyon and is bordered by steep, rocky cliffs. Zvornik Reservoir (surface area 8.1 km², volume 47.43 × 10⁶ m³, length 25 km, maximum depth 39 m) was formed by damming for the Zvornik Hydropower Plant. Since impoundment, approximately 50% of its basin has been infilled by erosional deposits delivered by torrential tributaries and small streams discharging into the reservoir. The reservoirs primarily serve hydropower generation but are also used for flood-wave attenuation, tourism, and recreation.

2.2. Sampling and Laboratory Analysis

Samples were collected at the same site in the pelagic (limnetic) zone of each reservoir during four surveys at monthly intervals in 2024: June (Va, Pa, Za), July (Vb, Pb, Zb), August (Vc, Pc, Zc), and October (Vd, Pd, Zd).

Water temperature (T), pH, dissolved oxygen (O₂), and conductivity (EC) were measured in situ using a multiparameter WTW MultiLine 3410 (WTW, Weilheim, Germany). Water transparency (Z_SD) was determined using a Secchi disk. To determine the thermal stratification depth, temperature, and dissolved oxygen were measured at 2 m intervals from the surface to 20 m depth in the Višegrad and Perućac Reservoirs and at 1 m intervals in the Zvornik Reservoir. A 1.2 L horizontal water sampler (LaMotte, Chestertown, MD, USA) was used to obtain a depth-integrated composite sample every 2 m up to a depth of 20 m in Višegrad and Perućac reservoirs, while in the Zvornik Reservoir, samples were collected to the maximum available depth of 11 m since this is the shallowest reservoir in the cascade. Subsamples for chemical analyses (1 L), chlorophyll a determination (1 L), and phytoplankton analysis (250 mL) were subsequently taken from this composite sample. Water samples (250 mL) for the quantitative analysis of phytoplankton were preserved using Lugol’s solution. Phytoplankton samples from Višegrad and Perućac Reservoirs were collected using a 25 μm mesh and vertically hauled from 20 m to the surface, while from the Zvornik Reservoir, the vertical sample spanned from 11 m to the surface. The samples were preserved with a 2% formaldehyde solution.

Laboratory water chemistry analyses included the quantification of total phosphorus (TP), total nitrogen (TN), nitrates (NO₃⁻), nitrites (NO₂⁻), ammonium (NH₄⁺), total Kjeldahl nitrogen (TKN), total organic carbon (TOC), chemical oxygen demand (COD), total hardness (GH), chloride (Cl⁻), and sulfate (SO₄²⁻). Chlorophyll a concentration was determined spectrophotometrically after filtration and ethanol extraction [27].

Phytoplankton analysis was conducted following the Utermöhl method [28] using an inverted microscope (Axiovert 5, Carl Zeiss^®, Göttingen, Germany). Species determination was performed using standard taxonomic literature. Counting and cell measurements were performed according to CEN EN 15204 [29] and Lund et al. [30]. Algal biovolume was calculated using standard geometric approximations [31,32,33,34]. Phytoplankton biomass (mg L⁻¹) was derived by assuming an equivalence of 1 mg biomass per 1 cm³ biovolume [35,36]. Diatoms were identified from permanent slides prepared after chemical sample treatment and mounting in Naphrax [37]. Species nomenclature and classification followed AlgaeBase [38].

2.3. Assessment of Trophic Status and Ecological Potential

The Carlson Trophic State Index [39] was applied for the assessment of trophic status, calculated from Secchi depth (TSI_SD; Z_SD, m), chlorophyll a (TSI_Chl-a; Chl-a, µg L⁻¹), and total phosphorus (TSI_TP; TP, µg L⁻¹). In addition, trophic status was evaluated using threshold values for phytoplankton biomass [40], Secchi depth, chlorophyll a, and nutrients (TP), following boundaries given by OECD [41] and the Official Gazette (42/01) [22] (Appendix A 

Table A1. Class boundaries of the trophic status for the Z_SD—Secchi depth, TP—total phosphorus, Chl-a—Chlorophyll-a, [41,65], Biomass [40] and TSI—Carlson Trophic Index [39].

Trophic Status	ZSD (m)	TP (µg L−1)	Chl-a (µg L−1)	Biomass (mg L−1)	TSI
Ultra-oligotrophic	>12	<4	<1	<0.12	0–30
Oligotrophic	12–6	4–10	1–2.5	0.12–0.4	30–40
Oligo-mesotrophic				0.4–0.6
Mesotrophic	6–3	10–35	2.5–8	0.6–1.5	40–50
Eutrophic	3–1.5	35–100	8–25	1.5–2.5	50–60
Polyeutrophic				2.5–5
Hypereutrophic	<1.5	>100	>25	>5	>60

).

The ecological potential of the Višegrad and Perućac reservoirs was assessed using the Hungarian Lake Phytoplankton Index [14]. HLPI integrates the normalized Ecological Quality Ratios (EQR) of two metrics: chlorophyll a concentration (EQR_Chl-a) and the Q_k index (EQR_Q). Q_k is a compositional metric based on phytoplankton functional groups [11,12,13]. Normalized EQR_Chl-a values were derived using a polynomial equation adopted from the intercalibrated approach developed for deep Croatian lakes [42]. The Q_k index was standardized (Qk_stand) by dividing the calculated Qk value by the maximum Q value (Q_kmax = 9). Qk_{_stand} was subsequently transformed to EQR_Qk (HLPI) using a third-order polynomial regression equation, referenced to the hydromorphologically comparable deep Lake Peruća in Croatia [43].

The Hungarian River Phytoplankton Index (HRPI) [44] was applied for Zvornik Reservoir, given its location in the Peripannonian subregion, short water residence time, and absence of thermal stratification. This index is also based on chlorophyll a concentration and a compositional metric derived from phytoplankton functional groups (Q_r). Qr_stand was gained by dividing the calculated Q_r value by a maximum value of 5 and subsequently normalised to EQR_Qr (HRPI) using a third-order polynomial regression equation, following the intercalibrated equations developed for reservoirs on large rivers with short residence times in the Pannonian ecoregion [45].

Class boundaries for ecological potential (good or better/moderate, moderate/poor, and poor/bad) were set as an equidistant division of the EQR gradient at 0.6, 0.4, and 0.2 [21] (Appendix A 

Table A2. Class boundaries of the ecological potential for the studied reservoirs. HLPI—Hungarian Lake Phytoplankton Index [14], HRPI—Hungarian River Phytoplankton Index [44], EQR—ecological quality ratio.

).

Total phosphorus (TP) thresholds were estimated in relation to EQR derived from the HLPI and HRPI indices using the Supporting Elements Toolkit [46]. In this approach, TP was treated as the pressure variable and EQR as the biological response, allowing threshold values to be defined statistically rather than arbitrarily. The analysis focused on class boundaries, and particularly the good/moderate transition was defined by the 75th percentile of the good class.

2.4. Data Analysis

The Shannon diversity index (H′, ln), Simpson index (1 − Λ), and Pielou’s evenness (J′) were calculated based on phytoplankton cell counts, whereas phytoplankton biomass was used in all subsequent statistical analyses. Community composition and characteristic taxa were explored using SIMPER and non-metric multidimensional scaling (nMDS) based on a Bray–Curtis similarity matrix, with overlaid vectors for the dominant species (based on the Pearson correlation of r > 0.5). Shade plots were used for visualisation of compositional data. Prior to analysis, biomass data were log-transformed using log(x + 1). Principal component analysis (PCA) based on Euclidean distance was applied to describe sample patterns regarding the physico-chemical variables. Differences in assemblage structure among the three reservoirs were tested using PERMANOVA. Homogeneity of multivariate dispersions among the reservoirs was assessed using PERMDISP, and significant differences were further explored using pair-wise PERMANOVA tests [47]. All of the aforementioned analyses were performed in PRIMER v6 [48]. Spearman’s rank correlations were computed among physico-chemical parameters, the HLPI index, chlorophyll a, phytoplankton biomass, and TSI values. ANOVA was applied to test for differences in total biomass, chlorophyll a, TSI_avg, and EQR values among groups defined by reservoir and by sampling months. These two analyses were performed in IBM SPSS Statistics, version 25.0 [49].

3. Results

3.1. Physico-Chemical Parameters

Results of the physico-chemical water analyses are presented in Appendix A 

Table A3. Physical and chemical parameters of studied reservoirs V (Višegrad), P (Perućac) and Zvornik (Z) in different months of sampling: a—June, b—July, c—August/September, d—October in 2024. T—temperature, O₂—oxygen concentration, % O₂—oxygen saturation, EC—electrical conductivity, Z_SD—Secchi depth, TN—total nitrogen, NO₃⁻—nitrate, NO₂⁻—nitrite, NH₄⁺—ammonium, TKN—total Kjeldahl nitrogen, TP—total phosphorus, TOC—total organic carbon, COD—chemical oxygen demand, Cl⁻—chlorides, SO₄²⁻, GH—total hardness.

	T (°C)	O2 mg/L	%O2	pH	EC µS/cm	ZSD m	TN mg/L	NO3− mg/L	NO2− mg/L	NH4+ mg/L	TKN mg/L	TP µg/L	TOC mg/L	COD mgO2/L	GH °dGH	Cl− mg/L	SO42− mg/L
			mg/L
Višegrad Reservoir
Va	25.9	10.12	130	8.35	276	1.8	2.87	2.02	0.003	0.074	0.84	20	2.75	4.8	8.12	0.09	1.63
Vb	26	10.72	137	8	281	2.5	2.19	1.63	0.003	0.028	0.56	24	3.53	4.8	8.4	0.18	0.002
Vc	23.6	11.04	135	8.03	279	1.6	3.39	1.71	0.0008	0.028	1.68	44	3.48	6.72	5.32	0.09	0.012
Vd	17.9	9.41	103	8.18	292	3.6	5.82	4.7	0.004	0.131	1.12	18	3.26	13.44	8.12	0.01	10.3
$\overline{\mathrm{x}}$	23.4	10.32	126	8.14	282	2.4	3.57	2.52	0.0027	0.065	1.05	26.5	3.26	7.44	7.49	0.09	2.99
SD	3.8	0.7	15.8	0.2	7.0	0.9	1.6	1.5	0.0014	0.049	0.48	11.9	0.4	4.1	1.5	0.1	4.9
Perućac Reservoir
Pa	26.2	9.46	121	8.24	286	4.3	2.34	1.53	0.0058	0.028	0.81	15	3.14	4.3	8.79	0.04	1.44
Pb	25.3	9.1	114	8	283	4.3	2.28	1.44	0.003	0.028	0.84	38	2.28	3.84	11.2	0.09	0.002
Pc	24.5	9.5	117	8	276	4.5	2.48	1.36	0.0002	0.028	1.12	45	3.53	2.82	5.32	0.09	0.037
Pd	17.9	8.81	95	7.99	287	5.9	3.32	2.47	0.012	0.076	0.84	25	2.81	2.88	7.67	0.01	10.71
$\overline{\mathrm{x}}$	23.48	9.22	112	8.06	283	4.8	2.61	1.70	0.0053	0.040	0.90	30.8	2.94	3.46	8.25	0.06	3.05
SD	3.78	0.33	11.53	0.12	4.97	0.77	0.48	0.52	0.0050	0.024	0.15	13.38	0.53	0.73	2.44	0.04	5.15
Zvornik Reservoir
Za	19.7	9.06	101	8.3	294	2.1	2.14	1.57	0.01	0.028	0.56	33	3.11	4.8	8.62	0.05	1.05
Zb	24.7	8.48	104	8.3	294	2.3	2.45	2.17	0.0015	0.028	0.28	40	2.46	10.56	11.2	0.09	0.004
Zc	22.7	8.7	102	8.25	281	2	5.7	4.29	0.0005	0.028	1.4	53	2.13	1.92	5.88	0.09	0.011
Zd	16.5	9.5	100	8.24	300	1.2	5.96	4.56	0.003	0.071	1.4	15	3.04	5.76	8.17	0.01	24.72
$\overline{\mathrm{x}}$	20.9	8.94	102	8.27	292	1.9	4.06	3.15	0.0038	0.0388	0.91	35.3	2.69	5.76	8.47	0.06	6.45
SD	3.58	0.45	1.71	0.03	8.02	0.48	2.05	1.50	0.0043	0.0215	0.58	15.8	0.47	3.59	2.18	0.04	12.19

. Water temperature ranged from 16.5 °C (Zd) to 26.0 °C (Pa). Thermal stratification down to 20 m depth was recorded in the Višegrad Reservoir (monthly temperature range: 5.4–9.0 °C) and the Perućac Reservoir (1.9–8.3 °C). Profiles Va, Vb, and Vc indicated clear thermal stratification, with the most pronounced thermoclines occurring at 2–4 m, 3–5 m, and 6–12 m depth, respectively. In contrast, profile Vd in October showed a weak thermal gradient without a sharply defined thermocline. Similarly, the temperature profiles in Perućac indicated clear thermal stratification in Pa, Pb, and Pc, with the most pronounced thermocline occurring at 4–8 m in Pa and at 6–8 m in Pb and Pc, whereas Pd showed a weak gradual temperature decline and no clearly developed thermocline. In contrast, temperatures in the Zvornik Reservoir profiles were largely homogeneous with a narrow vertical temperature range (0.4–1.5 °C) from the surface to the maximum sampling depth of 11 m.

Secchi depth varied between 1.2 m (Zd) and 5.91 m (Pd). Dissolved oxygen concentrations ranged from 8.48 mg L⁻¹ (Zb) to 11.04 mg L⁻¹ (Vb). Across all reservoirs, pH remained slightly alkaline, ranging from 8.00 to 8.35. Electrical conductivity was low and showed little spatio-temporal variation (276–300 µS cm⁻¹).

Total nitrogen (TN) reached its lowest and highest values in spring and autumn (2.14 and 5.96 mg L⁻¹, respectively) in the Zvornik Reservoir. The highest concentration of nitrates (NO₃⁻) occurred in the Višegrad Reservoir in autumn (4.7 mg L⁻¹), whereas the Perućac Reservoir consistently displayed the lowest concentration of nitrates (1.36–2.47 mg L⁻¹). Values of nitrites were generally very low at all sampling sites (≤0.01 mg L⁻¹), with a slight increase observed in autumn at Perućac (0.012 mg L⁻¹). The maximum measured concentration of ammonium was 1.31 mg L⁻¹ (Vd).

Total phosphorus (TP) values ranged from 15 to 53 µg L⁻¹ (Zd and Zc, respectively). Total organic carbon (TOC) peaked at 3.53 mg L⁻¹ (Vb, Pc) and reached its lowest value at 2.13 mg L⁻¹ (Zc). Chemical oxygen demand (COD; mg O₂ L⁻¹) varied widely, from 1.92 (Zc) to 13.44 (Vd). Total hardness ranged from 5.32 °dGH (Vc, Pc) to 11.2 °dGH (Pb, Zb). Sulfates and chlorides were uniformly low at all sampling sites.

Statistically significant positive correlations, assessed using Spearman’s rank correlation coefficient (p < 0.05; r > 0.67), were observed between temperature and oxygen saturation, temperature and chloride concentration, dissolved oxygen and total organic carbon, total nitrogen and nitrate concentrations, and ammonium and sulfate concentrations. Statistically significant negative correlations (p < 0.05; r < −0.67) were found between electrical conductivity and oxygen saturation, nitrite and total phosphorus, total Kjeldahl nitrogen and water hardness, and chloride and sulfate concentrations.

Principal component analysis (PCA) of the 14 environmental variables indicated that the first 2 principal component axes explained 52.4% of the total variance (Appendix A 

Table A4. Relative variance explained and factor coordinates of the variables for the first two principal components (PC1 and PC2) of the Principal Component Analysis (PCA): T—temperature, O₂—oxygen concentration, %—oxygen saturation, EC—electrical conductivity, Z_SD—Secchi depth, TN—total nitrogen, NO₃⁻—nitrate, NO₂⁻—nitrite, NH₄⁺—ammonium, TP—total phosphorus, TOC—total organic carbon, COD—chemical oxygen demand, GH—hardness.

Variable	PC1	PC2
Variation (%)	33.8	18.6
Cumulative variation (%)	33.8	52.4
Eigenvalues	4.73	2.61
T	−0.382	−0.097
O2	−0.258	0.455
% O2	−0.378	0.273
pH	0.198	−0.016
EC	0.371	−0.083
ZSD	−0.024	−0.304
TN	0.324	0.307
NO3−	0.373	0.241
NO2−	0.155	−0.268
NH4+	0.313	0.25
TP	−0.171	−0.136
TOC	−0.129	0.37
COD	0.209	0.214
GH	0.103	−0.349

). Axis 1 was primarily associated with temperature, oxygen saturation, and nitrates (intra-set correlations: −0.382, −0.378, and 0.373, respectively). Axis 2 was driven mainly by dissolved oxygen, total organic carbon, and total hardness (intra-set correlations: 0.455, 0.370, and −0.349, respectively). The ordination (Figure 2) identified three main clusters (Euclidean distance of 4.8), while two samples were positioned outside these groups. The first group comprised all samples collected in June and July (except Vb), the second cluster included sample Vb together with two samples from August marked by the highest values of O₂ and TOC, while the third group consisted of samples from October defined by the highest values of TN. Two samples, Zc (in Zvornik in August, with the highest value of TP) and Pd (in Perućac in October, with the highest nitrite), did not cluster with the others.

3.2. Description of Phytoplankton Community

In total, 80 phytoplankton taxa were identified (52 in the Višegrad Reservoir, 37 in the Perućac Reservoir, and 49 in the Zvornik Reservoir). The taxa were assigned to six major taxonomic groups: Charophyta (6), Chlorophyta (27), Cryptista (5), Cyanobacteria (8), Dinoflagellata (2), and Heterokontophyta (32). Within Heterokontophyta, Bacillariophyceae comprised 16 taxa, Mediophyceae 8 taxa, Coscinodiscophyceae 2 taxa, and Chrysophyceae 6 taxa. Within Charophyta, Zygnematophyceae was the most species-rich class, with 5 taxa (

Table A5. Phytoplankton functional groups (FG) and identified species within taxonomic groups (TG) in the studied reservoirs.

FG	Taxa	TG	FG	Taxa	TG
A	Acanthoceras zachariasii (Brun) Simonsen	Medi	LM	Microcystis aeruginosa (Kützing) Kützing	Cyan
	Cyclostephanos invisitatus (M.H.Hohn & Hellerman) E.C.Theriot, Stoermer & Håkanasson	Medi	LO	Chroococcus minutus (Kützing) Nägeli	Cyan
B	Discostella pseudostelligera (Hustedt) Houk & Klee	Medi		Snowella lacustris (Chodat) Komárek & Hindák	Cyan
	Pantocsekiella costei (J.C.Druart & F.Straub) K.T.Kiss & E.Ács	Medi		Synechocystis sp.	Cyan
	Pantocsekiella ocellata (Pantocsek) K.T.Kiss & Ács	Medi		Ceratium hirundinella (O.F.Müller) Dujardin	Dino
	Stephanodiscus parvus Stoermer & Håkansson	Medi		Glenodinium sp.	Dino
C	Asterionella formosa Hassall	Baci	MP	Achnanthidium minutissimum (Kützing) Czarnecki	Baci
	Asterionella formosa var. acaroides Lemmermann	Baci		Cocconeis placentula Ehrenberg	Baci
	Discostella stelligera (Cleve & Grunow) Houk & Klee	Medi		Nitzschia dissipata (Kützing) Rabenhorst	Baci
	Lindavia radiosa (Grunow) De Toni & Forti	Medi		Nitzschia palea (Kützing) W.Smith	Baci
D	Ulnaria acus (Kützing) Aboal	Baci		Nitzschia recta Hantzsch	Baci
	Ulnaria ulna (Nitzsch) Compère	Baci		Gomphonema parvulum (Kützing) Kützing	Baci
E	Dinobryon bavaricum Imhof	Chry		Gomphonema pumilum (Grunow) E.Reichardt & Lange-Bertalot	Baci
	Dinobryon crenulatum West & G.S.West	Chry		Fragilaria vaucheriae (Kützing) J.B.Petersen	Baci
	Dinobryon divergens O.E.Imhof	Chry		Hippodonta capitata (Ehrenberg) Lange-Bertalot, Metzeltin & Witkowski	Baci
	Dinobryon sertularia Ehrenberg	Chry		Melosira varians C.Agardh	Cosc
	Elakatothrix gelatinosa Wille	Kleb	N	Cosmarium subcostatum Nordstedt	Zygn
	Sphaerocystis sp.	Chlo		Cosmarium tinctum Ralfs	Zygn
	Raphidocelis danubiana (Hindák) Marvan, Komárek & Comas	Chlo	P	Closterium acutum Brébisson	Zygn
	Oocystis parva West & G.S.West	Treb		Closterium limneticum Lemmermann	Zygn
	Willea apiculata (Lemmermann) D.M.John, M.J.Wynne & P.M.Tsarenko	Treb		Closterium venus Kützing ex Ralfs	Zygn
J	Crucigenia quadrata Morren	Chlo		Fragilaria crotonensis Kitton	Baci
	Crucigenia sp.	Chlo		Fragilaria saxoplanctonica Lange-Bertalot & S.Ulrich	Baci
	Desmodesmus abundans (Kirchner) E.H.Hegewald	Chlo		Fragilaria sp.	Baci
	Desmodesmus bicellularis (Chodat) S.S.An, T.Friedl & E.Hegewald	Chlo		Aulacoseira granulata (Ehrenberg) Simonsen	Cosc
	Golenkinia radiata Chodat	Chlo	S1	Pseudanabaena catenata Lauterborn	Cyan
	Scenedesmus ecornis (Ehrenberg) Chodat	Chlo		Pseudanabaena limnetica (Lemmermann) Komárek	Cyan
	Scenedesmus intermedius var. acaudatus Hortobagyi	Chlo	T	Planctonema lauterbornii Schmidle	Treb
	Scenedesmus quadricauda (Turpin) Brébisson	Chlo	X1	Monoraphidium contortum (Thuret) Komárková-Legnerová	Chlo
	Stauridium tetras (Ehrenberg) E.Hegewald	Chlo		Monoraphidium minutum (Nägeli) Komárková-Legnerová	Chlo
	Tetradesmus dimorphus (Turpin) M.J.Wynne	Chlo		Pseudodidymocystis planctonica (Korshikov) E.Hegewald & Deason	Chlo
	Tetradesmus lagerheimii M.J.Wynne & Guiry	Chlo	X2	Chlamydomonas sp.	Chlo
	Tetraëdron minimum (A.Braun) Hansgirg	Chlo		Plagioselmis lacustris (Pascher & Ruttner) Javornicky	Cryp
	Tetraedron tumidulum (Reinsch) Hansgirg	Chlo		Plagioselmis nannoplanctica (Skuja) G.Novarino, I.A.N.Lucas & Morrall	Cryp
	Tetrastrum glabrum (Y.V.Roll) Ahlstrom & Tiffany	Chlo		Kephyrion rubri-claustri Conrad	Chry
	Verrucodesmus verrucosus (Y.V.Roll) E.Hegewald	Chlo	X3	Pseudodidymocystis inconspicua (Korshikov) Hindák	Chlo
	Lagerheimia genevensis (Chodat) Chodat	Treb		Chrysococcus rufescens Klebs	Chry
	Lemmermannia tetrapedia (Kirchner) Lemmermann	Treb	Y	Cryptomonas erosa Ehrenberg	Cryp
K	Aphanocapsa delicatissima West & G.S.West	Cyan		Cryptomonas marssonii Skuja	Cryp
	Aphanocapsa holsatica (Lemmermann) G.Cronberg & Komárek	Cyan		Cryptomonas ovata Ehrenberg	Cryp

).

Heterokontophyta, dominated by diatoms (Figure 3), prevailed in biomass across all reservoirs and sampling months (43.5–95.4%), except in August in the Višegrad Reservoir, when Chlorophyta was the dominant group (49%), driven primarily by Sphaerocystis sp. (sample Vc). The most frequently recorded diatoms included Fragilaria crotonensis Kitton and Asterionella formosa Hassall, present in all reservoirs, Fragilaria saxoplanctonica Lange-Bertalot & S. Ulrich (in Vb, Vc, Vd), Ulnaria ulna (Nitzsch) Compère (in Va, Pa), Acanthoceras zachariasii (Brun) Simonsen (in Zb, Pb, Zd), and Pantocsekiella ocellata (Pantocsek) K.T. Kiss & Ács (in Pc, Pd).

Within Heterokontophyta, chrysophytes, particularly Dinobryon divergens O.E. Imhof, also contributed substantially to the total phytoplankton biomass in the Perućac Reservoir during June and August (samples Pa and Pc with 32.3% and 44.7%, respectively), in the Zvornik Reservoir during August (sample Zc with 26.6%), and in the Višegrad Reservoir during July and October (samples Vb and Vd with 37.1% and 43.1%, respectively). The Zvornik and Višegrad reservoirs were further characterized by the presence of Cryptista (including Cryptomonas erosa Ehrenberg, Cryptomonas ovata Ehrenberg, and Plagioselmis nannoplanctica (Skuja) G. Novarino, I.A.N. Lucas & Morrall), reaching from 18% to 11% of the relative biomass contribution in July (sample Vb) and August (sample Zc).

The highest phytoplankton biomass was recorded in the Perućac Reservoir during late August (Pc), where Dinobryon divergens and Fragilaria crotonensis reached 1.27 and 1.18 mg L⁻¹, respectively (Figure 4). Both taxa were also abundant in the Višegrad Reservoir, but their contribution declined toward the Zvornik Reservoir. Acanthoceras zachariasii was another important contributor, codominating in Perućac (Pb: 1.21 mg L⁻¹) and remaining abundant in Zvornik (Zb: 0.99 mg L⁻¹). In the Perućac Reservoir, Asterionella formosa and Pantocsekiella ocellata were also present as subdominant taxa.

The seasonal variation in species composition among the reservoirs is shown as follows. In Višegrad, dominance shifted from Ulnaria ulna (Nitzsch) Compèrein June to D. divergens in July and October, while Sphaerocystis sp. prevailed in late August (Vc: 0.81 mg L⁻¹). In Perućac, June biomass was mainly shared by D. divergens, F. crotonensis, and A. formosa, followed by the dominance of A. zachariasii in July and renewed codominance of D. divergens and F. crotonensis in late August. In October, F. crotonensis remained an important species together with P. ocellata. The Zvornik Reservoir was characterized by dominance of Aulacoseira granulata (Ehrenberg) Simonsen in June, the July peak of A. zachariasii, codominance of Pantocsekiella costei (J.C.Druart & F.Straub) K.T.Kiss & E.Ács, Pantocsekiella ocellata (Pantocsek) K.T.Kiss & Ács, D. divergens, and F. crotonensis in August, and a peak of Cocconeis placentula Ehrenberg, A. granulata, P. ocellata in October.

Analysing species diversity, the highest number of taxa was recorded in the Višegrad Reservoir sample collected in late summer (Vc: S = 31), whereas the lowest richness was observed in the Perućac Reservoir sample collected in June (Pa: S = 16). The lowest cell density (543,760 cells L⁻¹) occurred in sample Zd, which also exhibited the highest values of the Shannon diversity index (H′ log_e = 2.67), Simpson’s diversity (1 − Λ = 0.91), and Pielou’s evenness (J′ = 0.89). In contrast, the highest cell density (12,863,038 cells L⁻¹) was counted in sample Va, where the lowest diversity and evenness values were detected (H′ log_e = 0.56; 1 − Λ = 0.20; J′ = 0.18) (

Table 2. Values of total species richness (S), cell abundance, Pielou’s evenness index (J′), Shannon diversity index H′ (log_e), and Simpson’s diversity index (1 − Λ) per studied samples.

Samples	S	N	J′	H′ (loge)	1 − Λ
Va	21	12,863,038	0.18	0.56	0.20
Vb	25	8,889,602	0.54	1.72	0.67
Vc	31	10,434,610	0.50	1.71	0.66
Vd	28	4,763,141	0.69	2.29	0.85
Pa	16	2,494,861	0.75	2.07	0.84
Pb	21	6,280,137	0.69	2.10	0.78
Pc	15	3,061,460	0.62	1.67	0.74
Pd	16	1,720,008	0.56	1.55	0.68
Za	20	997,832	0.75	2.23	0.82
Zb	28	1,685,608	0.78	2.59	0.87
Zc	22	1,192,259	0.81	2.50	0.89
Zd	20	543,760	0.89	2.67	0.91

).

Non-metric multidimensional scaling (nMDS) of phytoplankton species composition at a 20% similarity level grouped samples into three clusters, with sample Va remaining unassigned (Figure 5). The first cluster comprised samples from the Zvornik Reservoir (Za, Zc, Zd), characterized by the occurrence of Cyclostephanos invisitatus (M.H. Hohn & Hellerman) E.C. Theriot, Stoermer & Håkanasson, Cocconeis placentula, and Achnanthidium minutissimum (Kützing) Czarnecki. The second cluster included samples from the Višegrad and Perućac reservoirs, typified by D. divergens, F. crotonensis, P. ocellata, Sphaerocystis sp., P. nannoplanctica, C. ovata, and F. saxoplanctonica. The third cluster consisted of samples Zb and Pb, distinguished by A. zachariasii. Sample Va was characterized by Discostella pseudostelligera (Hustedt) Houk & Klee.

PERMANOVA analysis based on a Bray–Curtis similarity matrix revealed a significant effect of reservoir on assemblage structure (Pseudo-F = 2.151, p = 0.024). PERMDISP showed no significant differences in multivariate dispersion among the locations (F = 1.3318, p = 0.482), indicating that the spatial pattern detected by PERMANOVA was not driven by differences in within-group variability. Pairwise PERMANOVA tests did not detect significant differences between individual reservoir pairs, although the comparison between Perućac and Zvornik approached marginal significance (t = 2.0185, p = 0.054), suggesting that this pair contributed strongly to the observed spatial differentiation. Overall, the results indicate a moderate, marginally supported spatial differentiation in assemblage structure among the reservoirs.

SIMPER analysis identified samples from the Perućac Reservoir as the most similar to each other (35.88%), followed by those from the Višegrad Reservoir (25.77%), while samples from the Zvornik Reservoir showed the lowest within-group similarity in terms of monthly dynamics (19.62%) (

Table 3. Results of the SIMPER analysis showing taxon contributions (%) by grouping factor reservoirs (V: Višegrad P: Perućac, Z: Zvornik) and sampling months.

Species	V	P	Z	June	July	August	October
Average similarity (%)	25.77	35.88	19.62	18.71	30.16	10.61	S: 8.02
Species	Contribution (%)
Fragilaria crotonensis Kitton	9.51	56.81	16.74	21.3	24.20	37.45
Dinobryon divergens O.E.Imhof	29.05	17.72	9.94	38.02		19.84	7.12
Fragilaria saxoplanctonica Lange-Bertalot & S.Ulrich	30.90
Pantocsekiella costei (J.C.Druart & F.Straub) K.T.Kiss & E.Ács	8.68		18.32
Achnanthidium minutissimum (Kützing) Czarnecki			16.5
Acanthoceras zachariasii (Brun) Simonsen					56.52
Pantocsekiella ocellata (Pantocsek) K.T.Kiss & Ács				9.93		9.93	36.44
Cryptomonas ovata Ehrenberg							19.66
Ulnaria ulna (Nitzsch) Compère				16.28
Plagioselmis nannoplanctica (Skuja) G.Novarino, I.A.N.Lucas & Morrall				6.66		6.66
Asterionella formosa Hassall				9.98
Plagioselmis lacustris (Pascher & Ruttner) Javornicky							7.42
Tetraëdron minimum (A.Braun) Hansgirg							7.28

). The highest species contributions within each reservoir group were from F. saxoplanctonica for Višegrad (30.9%), F. crotonensis for Perućac (58.81%), and P. costei for Zvornik (18.32%). SIMPER analysis based on the sampling month showed the greatest similarity among July samples (30.61%), largely driven by the biomass contribution of A. zachariasii. In contrast, the lowest similarity was recorded in autumn (October; 8.02%), with notable contributions from P. ocellata and C. ovata. Across both grouping schemes (reservoirs and months), D. divergens (except in July) and F. crotonensis (except in October) consistently contributed to group similarities.

3.3. Description of Phytoplankton Functional Groups

A total of 19 functional groups were identified (Appendix A Table A5). The highest biomass contribution was recorded for functional group E in the Perućac Reservoir in August (Pc: 1.31 mg L⁻¹), and it was also present in the Višegrad Reservoir. Functional group A exhibited high biomass in the Perućac Reservoir in July (Pb: 1.21 mg L⁻¹) and slightly lower values in the Zvornik Reservoir in July (Zb: 0.99 mg L⁻¹). Functional group P reached its maximum biomass in the Perućac Reservoir (Pc: 1.18 mg L⁻¹) and was continuously represented across all three reservoirs, with varying biomass contributions (Figure 6).

The relative contribution of functional groups varied among reservoirs and sampling months (Figure 7). In the Višegrad Reservoir, functional group D dominated in June (Va: 55%), whereas groups B and E were subdominant (19% and 9%, respectively). In July (Vb), functional groups P and E accounted for most of the biomass (41% and 37%), with a notable contribution from groups Y (8.6%) and J (7.4%). In August (Vc), group F was dominant (44.5%), while P remained substantial (28%) and group C increased to 7%. In October (Vd), functional group E (43%) dominated together with P (32.5%), while in the Perućac Reservoir, functional groups E, C, and P contributed most to biomass in June (Pa; 32%, 26%, and 23%, respectively). In July (Pb), group A dominated (56%), whereas P remained an important contributor (21%). In August (Pc), codominance of groups E and P was observed (45% and 40%), with group B contributing 13%. During autumn (Pd), group P again became dominant (47%), while group E was replaced by group B (6% and 40%, respectively). In the Zvornik Reservoir, the most abundant groups in June (Za) were P, C, Y, MP, and E (29%, 23%, 18%, 14%, and 12%, respectively). In July (Zb), group A showed an overwhelming dominance (80%), whereas group B contributed 8% of the relative phytoplankton biomass. In August (Zc), group A was not recorded in the biomass. Instead, groups B, E, MP, and P co-dominated, accounting for 22%, 20%, 15%, and 16%, respectively. In October (Zd), group A reappeared but with a much smaller share (5%), while groups C, MP, and P contributed 18%, 36%, and 21% of the relative phytoplankton biomass, respectively.

Non-metric multidimensional scaling (nMDS) based on phytoplankton functional groups produced the same sample-grouping pattern as the species-composition approach with a higher similarity threshold (30%), yielding three clusters and one unassigned sample (Figure 8). Applying Pearson correlation vectors (r > 0.5) for functional groups to the ordination, the cluster of samples from the Zvornik Reservoir (Zc, Za, Zd) was primarily associated with functional groups MP and Y. The second cluster, including samples from the Višegrad and Perućac reservoirs, was primarily associated with functional groups E and P. The July samples from the Perućac and Zvornik reservoirs (Pb and Zb) formed a distinct cluster driven by functional group A, whereas the unassigned sample was characterised by functional group D.

3.4. Trophic Status and Ecological Potential of the Cascade Reservoirs

The results of the trophic status and ecological potential analyses are presented in

Table 4. Trophic status and ecological potential parameters of the Višegrad Reservoir (Va-Vd), Perućac Reservoir (Pa-Pd), and Zvornik Reservoir (Za-Zd) (Chl-a—Chlorophyll a; Z_SD—Secchi depth; TP—Total phosphorus; TN—Total nitrogen; TSI_Chl-a—Trophic state index for chlorophyll a; TSI_TP—Trophic state index for total phosphorus; TSI_SD—Trophic state index for Secchi depth; TSI_avg—mean TSI), Q_k—Compositional metric based on phytoplankton functional groups for lakes, Q_r—Compositional metric based on phytoplankton functional groups for rivers, HLPI—Hungarian Lake Phytoplankton Index, HRPI—Hungarian River Phytoplankton Index, EQR—Ecological quality ratio.

and

Table A6. Trophic and ecological status of the cascade reservoir using different metrics. HYPER- hypertrophic, POLY—polytrophic, EU—eutrophic, MESO—mesotrophic, OLIGO-MESO—oligo-mesotrophic, OLIGO—oligotrophic, ULTRA-OLIGO—ultra-oligotrophic. Chl-a—Chlorophyll-a, Z_SD—Secchi depth, TP—total phosphorus, TSI_avg—mean TSI, EQR—ecological quality ratio based on HLPI or HRPI. The mean trophic categories by reservoir are shown in bold.

. The highest measured chlorophyll a concentration was 13.98 µg L⁻¹ (Va), whereas the lowest was 1.97 µg L⁻¹ (Pd). Applying the OECD criteria [41] and Carlson’s Trophic State Index (TSI) based on mean chlorophyll a values, the Višegrad Reservoir was classified as eutrophic, while the other two reservoirs were mesotrophic. Based on mean total phosphorus (TP) concentrations, all investigated reservoirs were mesotrophic, with a tendency toward eutrophic status during the summer period (August). In contrast, mean TSI_TP values indicated a lower-class assessment, classifying all reservoirs as eutrophic, with a tendency ttowardhypertrophy in summer. Mean water transparency measured with a Secchi disc suggested mesotrophic conditions in the Perućac Reservoir and eutrophic conditions in the other two reservoirs. According to TSI_SD, the Zvornik and Višegrad Reservoirs ranged from mesotrophic to eutrophic, whereas the Perućac Reservoir was assessed as oligotrophic. Overall, the mean TSI (TSI_avg) indicated a eutrophic status for the Višegrad Reservoir and a mesotrophic status for the Perućac and Zvornik Reservoirs. Total plankton biomass ranged from 0.237 mg L⁻¹ to 2.946 mg L⁻¹ (Zc and Pc, respectively). Following the guidance of Brettum [40], mean monthly biomass values indicated mesotrophic conditions in the Višegrad Reservoir, eutrophic conditions in the Perućac Reservoir, and oligo-mesotrophic conditions in the Zvornik Reservoir. All reservoirs showed a pronounced increase in biomass during summer, followed by a decline toward autumn.

EQR showed statistically significant negative correlations (p < 0.05) with temperature, oxygen saturation, (r = −0.582 and r = −0.660, respectively), and a positive correlation with electrical conductivity (r = 0.698). Chlorophyll a correlated positively with oxygen saturation (r = 0.664) and negatively with electrical conductivity (r = −0.741) and EQR (r = −0.919) at a higher significance level (p < 0.01). Total phosphorus and TSI_avg were negatively correlated with nitrate concentrations (r = −0.676 and r = −0.790, respectively) (Appendix A 

Table A7. Summary of the Spearman’s rank correlation coefficient (r, N = 12) between the environmental variables, Chl-a—Chlorophyll-a, TSI_avg—Average Trophic State Index by Carlson, Algal Biomass, EQR—Ecological Quality Ratio. Correlations in bold indicate statistical significance. ** Correlation is significant at the 0.01 level (2-tailed). * Correlation is significant at the 0.05 level (2-tailed). T—temperature, % O₂—oxygen saturation, EC—electrical conductivity, Z_SD—Secchi depth, TN—total nitrogen, NO₃⁻—nitrate, NO₂⁻—nitrite, TP—total phosphorus, TP—total phosphorus. Statistically significant values of the correlation coefficient are shown in bold. ** Correlation is significant at the 0.01 level, * Correlation is significant at the 0.05 level.

	T (°C)	% O2	EC (µS cm−1)	ZSD (m)	TN (mg L−1)	NO3−(mg L−1)	NO2−(mg L−1)	TP (mg L−1)	Chl-a (µg L−1)	Biomass (mg L−1)	TSIavg	EQR
T (°C)	1.000	0.806 **	−0.514	0.182	−0.634 *	−0.641 *	−0.157	−0.007	0.438	0.481	0.224	−0.582 *
% O2	0.806 **	1.000	−0.692 *	−0.077	−0.343	−0.497	−0.374	0.053	0.664 *	0.546	0.462	−0.660 *
EC (µS cm−1)	−0.514	−0.692 *	1.000	−0.084	0.067	0.432	0.522	−0.412	−0.714 **	−0.403	−0.552	0.698 *
ZSD (m)	0.182	−0.077	−0.084	1.000	−0.357	−0.406	0.281	0.063	−0.322	0.606 *	−0.559	0.232
TN (mg L−1)	−0.634 *	−0.343	0.067	−0.357	1.000	0.769 **	−0.228	−0.084	0.245	−0.315	0.042	−0.028
NO3−(mg L−1)	−0.641 *	−0.497	0.432	−0.406	0.769 **	1.000	0.117	−0.252	−0.035	−0.655 *	−0.077	0.119
NO2− (mg L−1)	−0.157	−0.374	0.522	0.281	−0.228	0.117	1.000	−0.676 *	−0.470	−0.166	−0.790 **	0.423
TP (mg L−1)	−0.007	0.053	−0.412	0.063	−0.084	−0.252	−0.676 *	1.000	0.112	0.128	0.666 *	−0.144
Chl-a (µg L−1)	0.438	0.664 *	−0.714 **	−0.322	0.245	−0.035	−0.470	0.112	1.000	0.207	0.566	−0.919 **
Biomass (mg L−1)	0.481	0.546	−0.403	0.606 *	−0.315	−0.655 *	−0.166	0.128	0.207	1.000	−0.109	−0.120
TSIavg	0.224	0.462	−0.552	−0.559	0.042	−0.077	−0.790 **	0.666 *	0.566	−0.109	1.000	−0.533
EQR	−0.582 *	−0.660 *	0.698 *	0.232	−0.028	0.119	0.423	−0.144	−0.919 **	−0.120	−0.533	1.000

). According to the HLPI and HRPI indices, the reservoirs generally exhibited good to better ecological potential, with the exception of Zvornik Reservoir in August, when ecological potential was classified as moderate. The ANOVA did not reveal significant differences (p > 0.05) in total biomass, chlorophyll a, TSIavg, or HLPI when sampling months and reservoir location were used as factors, except for a statistically significant difference in biomass between reservoirs, specifically Perućac and Zvornik (F = 4.921, p = 0.036).

The boundary value of 39 µg L⁻¹ for total phosphorus (TP) was derived using the categorical (boxplot) approach in the Supporting Elements Toolkit, in accordance with the toolkit’s methodology for establishing preliminary supporting thresholds. The class boundaries were extrapolated equidistantly from the established good/moderate threshold (0.039 mg L⁻¹ TP), using the same class width as between the high/good and good/moderate boundaries (0.010 mg L⁻¹). This resulted in moderate/poor and poor/bad boundaries of 0.049 and 0.060 mg L⁻¹ TP, respectively (

Table 5. Proposed class boundaries of the ecological potential status in the studied reservoirs of the Drina River using total phosphorus (TP).

Ecological Potential	Total Phosphorus (mg L−1)
Good and better	≤0.039
Moderate	0.040–0.049
Poor	0.050–0.060
Bad	>0.060

).

4. Discussion

Reservoirs are human-made ecosystems that balance hydropower and water management with sensitivity to multiple stressors. They are typically affected by eutrophication, chemical and organic pollution, and hydromorphological alterations (loss of connectivity, water-level fluctuations), with additional influences from invasive species and land-use changes in the catchment [2]. The present study provides new evidence on phytoplankton community structure and functional groups from a large cascade hydropower system in the Western Balkans. Within the Water Framework Directive (WFD), phytoplankton is a key biological quality element for assessing the ecological potential in heavily modified water bodies [21]. Since the biological metrics for freshwater ecosystems have not yet been intercalibrated in Bosnia and Herzegovina [50], this study relied on regionally developed approaches for phytoplankton to test their applicability in the Drina cascade reservoirs.

The physico-chemical characteristics of the Drina reservoirs indicate moderate nutrient and organic loading, with seasonal and local variability. Total phosphorus generally suggests mesotrophic to eutrophic conditions, while TOC remains within a narrow range. Despite its fluctuations, COD does not exceed the thresholds for very good and good ecological status [22]. Although oxygen solubility typically decreases with increasing temperature, warm conditions in productive systems can enhance photosynthetic activity and lead to higher daytime oxygen saturation [51]. Similarly, the positive relationship between dissolved O₂ and TOC may indicate a contribution of autochthonous organic carbon associated with in-lake primary production [52,53]. Nitrate represents a significant fraction of total nitrogen, and the negative nitrite-TP correlation suggests that nitrogen and phosphorus dynamics are not tightly linked because nitrite is a transient and unstable intermediate in surface waters [54].

An interannual comparison was only possible for the Višegrad Reservoir, as the Perućac and Zvornik Reservoirs are not included in the national monitoring program. Compared to the 2023 data [55], phosphorus concentrations in 2024 were similar and mostly indicated mesotrophic conditions. Chlorophyll a also followed a comparable seasonal pattern in both years, while Secchi depth remained consistently below 2.5 m. In contrast, total nitrogen and nitrate were substantially higher in 2024 than in 2023, when both parameters remained below 1 mg L⁻¹. This points to increased nitrogen-related pressure during the study year.

Multivariate analyses support the interpretation that reservoir functioning was structured by both monthly succession and spatial gradients. PCA indicates the main axis of variation as seasonal, driven primarily by temperature, oxygenation, and nitrogen forms, whereas outlying samples reflect local extremes in phosphorus availability or transparency. Specifically, the separation of the late-summer Zvornik sample (Zc) corresponds to maximum TP values, suggesting that episodic phosphorus loading contributes to shifts in community composition and reduced ecological potential. In contrast, the October Perućac sample (Pd) is associated with the highest Secchi depth, indicating that high transparency and reduced resuspension constrain phytoplankton biomass.

Across the Drina cascade, phytoplankton assemblages showed certain monthly turnover and partially reservoir-specific structuring. Diatoms dominated the biomass during most of the study period, favoured by mixed conditions and efficient resource exploitation, while other algal groups peaked only during stable summer conditions. The late summer predominance of chlorophyte Sphaerocystis sp. in the Višegrad Reservoir, together with the high biomass contribution of Dinobryon divergens across all reservoirs, indicates reduced grazing control and selective resistance to zooplankton consumption. Spherical colonies of Sphaerocystis can reach up to 500 µm and are encased in mucilage with robust cell walls, rendering them resistant to zooplankton ingestion [56]. Furthermore, Dinobryon divergens uses a mixotrophic nutritional strategy that combines photosynthesis with bacterivory, allowing it to thrive under moderate temperatures and reduced light conditions in stratified water columns, particularly within the metalimnion. Simultaneously, Dinobryon divergens is considered low-quality food for zooplankton due to its grazing resistance and limited nutritional value, thus further supporting its seasonal dominance in freshwater ecosystems [57]. At the taxonomic level, several species reflected the interaction of trophic conditions, light, and hydrodynamics. Fragilaria saxoplanctonica reached its highest abundance in the Višegrad Reservoir in October, aligning with high nitrate values and nutrient-enriched conditions typical of eutrophic waters [58]. The dominance of Fragilaria crotonensis, especially in Perućac, reflects conditions with sufficient nutrient supplies, slight thermal stratification with adequate light availability, combined with wind velocities sufficient to keep the species within the surface mixed layer [59]. In addition, the co-occurrence of F. crotonensis and Asterionella formosa points to elevated nitrogen availability. While this relationship was originally described for oligotrophic alpine lakes and should therefore be interpreted cautiously in reservoir systems [60]. The phytoplankton assemblage of the Zvornik Reservoir strongly differed from those of the two upstream reservoirs, likely due to its shallower water, lack of stable thermal stratification, and stronger mixing regime. Consistent with these conditions, SIMPER analysis identified the benthic diatom Achnanthidium minutissimum and the mixing-tolerant species Pantocsekiella costei as the principal taxa contributing to this differentiation. The phytoplankton assemblages of the Drina cascade reservoirs revealed certain monthly patterns and particular reservoir-specific features. Višegrad and Perućac Reservoirs showed higher similarity in species composition and functional structure, with assemblages largely characterized by diatoms and functional groups associated with mixed, moderately enriched conditions, whereas the Zvornik Reservoir was more distinct due to its shallow depth, stronger mixing, and higher contribution of benthic and mixing-tolerant taxa.

The functional group approach provided a useful ecological synthesis of these species-level patterns. According to the Reynolds framework, phytoplankton taxa are grouped based on shared adaptive traits and habitat templates rather than taxonomy alone [11,12]. In the Višegrad Reservoir, the co-occurrence of functional groups B, D, and E suggests mixed conditions, diminished light availability, and variable nutrient supply, although these groups should not be treated as ecologically identical. The low transparency in the Višegrad Reservoir is consistent with the shade tolerance of groups B and P, whereas the occurrence of group E is better explained through the mixotrophic strategy of Dinobryon taxa than by its traditional shallow-lake habitat template, especially considering the hydromorphology of the Višegrad Reservoir [11,61,62]. In July, group P codominated with group E in the Višegrad Reservoir. The prevalence of group P, which includes taxa such as Fragilaria crotonensis, adapted to moderately light-limited, eutrophic epilimnetic conditions, further substantiates the eutrophic signal captured by the Carlson Trophic State Index [9,11]. In early summer, group A (predominantly Acanthoceras zachariasii) characterized the Perućac and Zvornik Reservoirs. While its preference for clear, mixed, and base-poor lakes is consistent with the higher transparency of the Perućac Reservoir [63,64], its July dominance under low-transparency conditions in the shallower Zvornik Reservoir likely reflects downstream advection from Perućac rather than exclusively local environmental selection. Furthermore, the co-occurrence of groups MP, P, C, and Y in the Zvornik Reservoir reflects their adaptation to a shallower, optically constrained, and frequently mixed ecosystem. Specifically, group MP is characteristic of frequently stirred, inorganically turbid shallow waters, whereas group Y consists of highly tolerant cryptophyte assemblages capable of thriving across diverse lentic habitats under low grazing pressure [12,65]. The findings primarily reflect late-spring to autumn phytoplankton succession and provide insight into seasonal patterns during this part of the year, while the full annual cycle of trophic dynamics and community turnover remains to be explored in future research.

The Shannon diversity, Simpson diversity, and Pielou’s evenness indices are highly sensitive to pronounced taxonomic dominance. Consequently, conditions promoting high phytoplankton abundance generally reduce overall diversity and evenness, whereas more moderate conditions support a more balanced distribution of abundances among coexisting taxa [66,67,68]. In the present study, the highest cell densities in the Višegrad Reservoir coincided with the lowest diversity and evenness scores, while samples from the Perućac Reservoir exhibited lower cell numbers coupled with higher diversity. The diversity indices further support the interpretation that high biomass events were accompanied by reduced community evenness.

Multivariate analysis revealed that phytoplankton community structure was jointly shaped by reservoir location and partially by monthly succession. nMDS showed that the Zvornik Reservoir formed a distinct cluster based on functional groups MP and Y, whereas the Višegrad and Perućac Reservoirs exhibited high assemblage overlap, characterized by associations E and P. SIMPER further identified a separate July cluster, primarily driven by the proliferation of Acanthoceras zachariasii (group A) across the entire system. The high degree of congruence between species-based and functional group-based ordinations indicates that the Reynolds framework successfully retains key ecological information while simplifying the complex dataset for monitoring purposes [69,70].

Studies on cascade reservoirs demonstrate that phytoplankton succession is shaped by a complex interplay of water column stability, hydrodynamics, and local environmental filtering rather than by nutrient availability alone [18]. While physical stability was not directly measured in the Drina system, the observed functional signatures serve as effective proxies for hydrodynamic conditions. The co-occurrence of groups P, E, and B suggests a shared tolerance to turbulent mixing, a pattern also reported in the Wujiang cascade [18] and Bulgarian reservoirs [16]. Such patterns suggest that these functional associations are frequently observed in temperate reservoir environments, representing a common community response to the specific environmental filters prevalent in these systems. Despite the lack of lateral connectivity typical of floodplain systems in the Danube basin [71], the hydrological impulses generated by reservoir operations and inflow variability in the Drina cascade appear to play a comparable structuring role. These disturbances favor mixing-tolerant taxa (groups P and B), while more specialized groups (A, MP, and D) respond to local seasonal and spatial heterogeneity, mirroring patterns observed in other pulse-dominated systems [44,72]. Furthermore, longitudinal connectivity likely facilitates downstream transport, contributing to a partial homogenization of assemblages [18,73]. However, local environmental sorting driven by variations in transparency, stratification, depth, and residence time maintains distinct differences among the reservoirs. This aligns with the consensus that phytoplankton biogeography is shaped by the interplay between dispersal and local habitat filters [74].

The trophic status of the Drina reservoirs revealed a functional divergence among metrics, reflecting the distinct ecological roles of each parameter: total phosphorus (TP) as a proxy for algal growth, chlorophyll a (Chl-a) as realized biomass, and transparency as a variable influenced by both biotic and abiotic factors [39,41]. While Chl-a levels indicated eutrophic conditions in the Višegrad Reservoir and predominantly mesotrophic states elsewhere, TP and TSI_TP suggested eutrophy with a shift toward hypertrophy during summer. In contrast, Secchi-based metrics pointed toward more oligotrophic conditions in the Perućac Reservoir. Such variations in trophic classification are characteristic of reservoir ecosystems, where light climate, stratification, and mixing regimes shape the relationship between nutrient availability and algal biomass [75,76]. The mismatch between TP and biomass in some samples, particularly under reduced transparency, suggests that light limitation may constrain algal production even when phosphorus is abundant, as observed in regional reservoir studies [77]. The metrics applied provided partially consistent and partially contrasting assessments. TSI, chlorophyll a, and total biomass described the magnitude of trophic expression, whereas functional groups and HLPI/HRPI reflected the ecological organisation of the phytoplankton assemblage. The trophic indices consistently identified August as the period of strongest trophic expression and indicated generally mesotrophic to eutrophic conditions across the reservoir cascade. Although biomass-based metrics inclined toward stronger trophic enrichment, HLPI/HRPI mostly indicated good or better ecological potential, showing that increased algal quantity was not always matched by an equally pronounced structural deterioration of the phytoplankton assemblage.

Compared to other systems in the Balkan region, such as the Gruža Reservoir [78] or the Grlište Reservoir [77], the Drina reservoirs exhibit similar variations in trophic classification, where phosphorus-based indices may overestimate the actual realized biomass. However, the Drina cascade shows a more pronounced spatial variability in biomass, likely due to higher flow-through rates and the specific connectivity of the reservoir chain. While these regional reservoirs often maintain a more stable eutrophic state, our results indicate that the Drina reservoirs fluctuate between mesotrophy and hypertrophy, driven by pulses in the hydrological regime rather than steady nutrient enrichment alone.

The observed relationship between the EQR-based assessment and the recommended good/moderate threshold of 39 µg L⁻¹ TP reflects a shift in the phytoplankton community toward species typical for moderate status. This value is closely aligned with the threshold for eutrophic status (>35 µg L⁻¹ TP) defined by the Republic of Srpska regulations [22], indicating that the decline to moderate ecological status corresponds to a shift toward eutrophy. While a higher threshold of 50 µg L⁻¹ is established for reservoirs of similar hydromorphology in the Adriatic River Basin District [79], the EQR-based classification for the Drina cascade reflects a direct biological response rather than just nutrient-based potential. The proposed values, therefore, provide a basis for the proposed class boundaries in these reservoirs, even when physical factors such as light and mixing disrupt the link between phosphorus and biomass.

From the WFD perspective, good ecological potential is the appropriate management target for heavily modified water bodies [21]. In this context, the integrative HLPI/HRPI metric combines biomass and community composition. Overall, the Drina reservoirs reached “good or better” ecological potential, although the metric also captured a decline to moderate potential in the late-summer Zvornik sample.

Results show that an integrative phytoplankton assessment is capable of detecting both the general ecological status and short-term seasonal degradation, making it a reliable tool for management in systems where single trophic metrics give conflicting signals. Nevertheless, the similarity between species-based and functional-group analyses indicates that functional groups represent a practical, interpretable, and ecologically responsive tool for assessing the ecological potential of reservoirs in Bosnia and Herzegovina.

Given that the present study primarily focuses on the phytoplankton dynamics from late spring to autumn, broader seasonal coverage and repeated observations over multiple years would help to further refine ecological classification. This would also allow for a more complete interpretation of phytoplankton dynamics in relation to key hydrological descriptors, including inflow, residence time, and stratification.