The Bidighinzu Lake is located in the northern part of Sardinia (Italy, 40°33'22"N 8°39'41"E) at an altitude of 334 m a.s.l. (Figure 1). The reservoir was constructed in 1956, it has a surface area of 1.7 × 10⁶ m², a theoretical maximum volume of 12.2 × 10⁶ m³, a maximum depth of 39.5 m and a mean depth of 7.3 m. The watershed has a surface area of 52.18 km². The main use of Bidighinzu Lake is to store drinking water. Problems affecting the potability of Bidighinzu Lake’s water have occurred since the lake was built, particularly hypolimnetic deoxygenation and the excessive presence of algae in the epilimnion (Lugliè et al. 2001). To address these problems, an aeration system was installed adjacent to the water intake tower in 1966. A by-pass was built in 1987 to divert urban and industrial wastes from the town of Thiesi to downstream of the reservoir (Marchetti et al. 1992, Lugliè et al. 2001). An experimental water aeration system with micro-air bubbles was activated in the reservoir from spring 2014 to autumn 2015 (Mariani et al. 2016).

Figure 1.

Location of the study site and sampling station.

Water sampling was conducted in the Bidighinzu Lake at monthly frequency from March 1988 to April 1989, from March 1994 to November 1997, from July to October 2003 and from March 2006 to October 2015 at a single station close to the deepest part of the reservoir (Figure 1). As part of the results of the WISER project (Water bodies in Europe: Integrative Systems to assess Ecological status and Recovery), the data analyses of phytoplankton and environmental variables collected at different stations within the reservoir demonstrated that the single sampling station, routinely considered, is representative of the entire Bidighinzu Lake (Thackeray et al. 2013). The sampling was more frequent in the summers of 2006, 2009, 2014 and 2015 (4 samples in each summer of these years), during the present study. Samples were collected at fixed depths using 2.5 l Niskin bottles at 0, 1, 2.5, 5, 7.5, 10, 15 m and subsequently at intervals of 10 m until 1 m from the bottom.

Figure 2.

Multiannual (2006–2015) variation of mixing zone and euphotic zone ratio (Z_mix/Z_eu) (upper panel) and of total phosphorous (TP) (lower panel) in Bidighinzu Lake in summer. Each black dot represents a single sample and each column represents the mean value.

The water transparency was measured with a Secchi disc and the water temperature was recorded using a multi-parametric probe (Hydrolab Datasonde 5 and YSI 6600V2). Water samples (1.5 l) from the Niskin bottles were stored in cold (4 °C) and dark conditions prior to laboratory analysis for ammonium (N-NH₄), nitrite (N-NO₂), nitrate (N-NO₃), total nitrogen (TN), reactive silica (Si-SiO₄), orthophosphate (P-PO₄) and total phosphorus (TP) according to Strickland and Parsons (1972). Total dissolved inorganic nitrogen (DIN) was calculated as the sum of ammonium, nitrate and nitrite. Total organic nitrogen (TON) was calculated by subtracting DIN from TN.

The Consiglio per la Ricerca in Agricoltura (CREA) provided daily meteorological data of rainfall and air temperature from a meteorological station nearby the Lake (about 15 km) from January 2006 to December 2015. Daily temperature values were monthly averaged and daily rainfall values were monthly cumulated. Monthly data on the climatic index WEMO (Western Mediterranean Oscillation) were provided by the Climatology Group of the University of Barcelona (Spain). This index measures the difference between the standardised atmospheric pressure recorded at Padua in northern Italy and San Fernando in south-western Spain (Martín-Vide and Lopez-Bustins 2006). Compared to the Mediterranean Oscillation index (Conte et al. 1989), which covers the atmospheric dynamics of the whole Mediterranean basin, WEMO is defined only within the synoptic framework of the western Mediterranean basin and its vicinities and shows a greater independence from the external dynamics of the Mediterranean basin (Martín-Vide and Lopez-Bustins 2006). WEMO is able to explain the pluviometric variability in the eastern fringe of the Iberian Peninsula, an area weakly or not related to the NAO (North Atlantic Oscillation index) pattern (Rodó et al. 1997, Martín-Vide and Lopez-Bustins 2006). Monthly data of the WEMO index were available continuously (i.e. without gaps) from 1988 to 2015.

The phytoplankton samples (100 ml), taken from the Niskin bottles, were immediately fixed after collection in 2% acid Lugol’s solution. Sample aliquots of 5 to 10 ml (depending on cell density) were analysed to estimate the cell density using the Utermöhl method (1958) with an inverted microscope (Zeiss, Axiovert 10). Cell counts were made at magnifications of 200× and 400× on at least 10% of the total bottom area of the sedimentation chamber. Additional non-fixed samples were observed immediately after collection to facilitate the identification of certain species. The species were identified following the taxonomic guides listed in Mariani et al. (2015). Results are shown at class level (and at order level only for the most abundant classes) in the present work.

Mean cell volume of each species was obtained by geometrical approximations from measurement of at least 30 cells in each sample according to Sun and Liu (2003). The mean volumes of all the species observed were added to obtain the total cell volume in each sample. The total cell volume was divided by the total cell number to obtain the mean cell volume of the whole phytoplankton community in each sample (abundance-weighted mean cell volume according to Sommer et al. 2017a). The biovolume in each sample was converted to biomass based on the assumption that 1 mm³ = 1 mg of fresh-weight biomass (Weyhenmeyer et al. 2013). The cell biomass of each species was determined in each sample by multiplying its mean cell volume by its cell density.

Water samples (1.5 l) for measuring chlorophyll a concentrations were stored in cold (4 °C) and dark conditions before the laboratory analysis (within 24 hours), which were conducted with a spectrophotometer (50 SCAM, Varian), according to Goltermann et al. (1978).

Environmental data were depth-averaged for the entire water column, whereas phytoplankton data were depth-averaged only for the euphotic zone assuming that these organisms live and grow mainly in this water layer.

The depth of the euphotic layer was calculated as Z_eu = 2.5 times the Secchi disc depth (Poikane 2009). The depth of the mixed layer (Z_mix) was calculated as the zone from the surface to the depth where the temperature gradient exceeded 1 °C x m^-1 (Ruttner 1963), except in the case of absence of stratification, when mixing depth was taken equal to the maximum depth of the reservoir. The mixing depth and euphotic depth ratio (Z_mix/Z_eu) was used as a good proxy for underwater light climate since it indicates the time spent by phytoplankton in good or poor light conditions when they are in the mixed water column: the highest is the Z_mix/Z_eu value in the lake, the lowest is the lifetime spent in suitable light conditions (Naselli Flores and Barone 2007).

Seasons were considered as: summer = July–September, autumn = October–December, winter = January–March and spring = April–June.

For each separate season, the non-parametric Mann-Kendall test (Gilbert 1987) was applied to detect significant monotonic trends in environmental and phytoplankton variables. Since the Mann-Kendall test results indicated the most significant trends for the summer season and for the phytoplankton cell density and cell volume datasets, all the subsequent statistical analyses were performed only on these phytoplankton descriptors in summer. Redundancy analysis (RDA, Ter Braak and Šmilauer 1998) was used to assess which environmental variable (amongst WEMO index, air temperature, rainfall, Z_mix/Z_eu, water temperature, Si-SiO₄, P-PO₄, TP, N-NO₃, N-NO₂, N-NH₄, DIN, TN and TON) significantly affected the multiannual patterns of the cell density and cell volume of all the phytoplankton classes observed consistently throughout the analysed decade (Bacillariophyceae, Chlorophyceae, Chrysophyceae, Conjugatophyceae, Cryptophyceae, Cyanophyceae, Dinophyceae, Euglenophyceae). Before analysis, phytoplankton data were log₁₀(×+1) transformed to stabilise variance and reduce the influence of dominant taxa on the ordination.

All statistical analyses were performed using R 3.4.3 software (R Core Team 2017). An a priori level of significance was established at p < 0.05, i.e. 95% confidence interval.

The non-parametric Mann-Kendall test was applied to detect significant monotonic trends in the summer WEMO index over all the historical datasets.

One-way analysis of variance (ANOVA) was performed to assess significant differences in environmental (WEMO index, air temperature, rainfall, Z_mix/Z_eu, water temperature, Si-SiO₄, P-PO₄, TP, N-NO₃, N-NO₂, N-NH₄, DIN, TN and TON) and phytoplankton (cell density and cell volume of each phytoplankton class observed) variables amongst the four continuous time cycles of samplings available (1988–1989, 1994–1997, 2003, 2006–2015). Prior to ANOVA, all data were log₁₀(x+1) transformed to meet ANOVA assumptions: normal distribution (Kolmogorov-Smirnov test) and homogeneity of variance (Bartlett test). When significant differences in the dependent variables based on these factors were observed, the post hoc Tukey’s pairwise comparisons test was performed.

All the analyses were performed using R 3.4.3 software (R Core Team 2017). An a priori level of significance was established at p < 0.05, i.e. 95% confidence interval.

To explore the land cover and land use changes of the Bidighinzu Lake watershed (52.18 km²) over time, a spatial data processing analysis with the support of a geographic information system (GIS) was performed (Padedda et al. 2015). The analysis was focused on the evaluation of land cover by using techniques of classification generating polygons by clipping the shape of the three time-series CORINE land cover maps: 1990 (EEA 2018a), 2000 (EEA 2018b) and 2012 (EEA 2018c). The automatic segmentation was carried out on the whole shape layer of the watershed of the reservoir. Each segmentation analysis was performed in order to obtain objects in accordance with the CLC system 2000 nomenclature (EEA 2000). For the purposes of this paper, the identified objects were aggregated into three homogeneous macro-classes (Urban Areas, Agricultural Areas and Natural and semi-Natural Areas) in accordance with the CLC classification at the 4^th hierarchical level.

The Mann-Kendall test revealed no significant multiannual trends in the meteo-climatic variables considered nor in the water temperature and transparency during the study decade (Table 3). Instead, a very significant increasing trend of Z_mix/Z_eu ratio was observed in summer (Table 3). This increment was confirmed, although not significant, when the year 2015 (breaking of stratification due to the activation of the aeration system) was excluded from the analysis (Mann Kendall test: S = 12, p = 0.828, n = 35). The multiannual patterns of Z_mix/Z_eu summer mean values showed fluctuations over the years, from 0.80 in 2012 up to 11.0 in 2015 (Figure 2). Strong decreasing trends of TP, N-NH₄ and DIN were detected in summer (Table 3). TP values decreased from a summer mean value of 390.8 mg m^-3 in 2006 to 160.3 mg m^-3 in 2015 (Figure 2). DIN values decreased strongly from a summer mean value of 1351.4 mg m^-3 in 2006 to 391.6 mg m^-3 in 2015, with N-NH₄ always representing more than 80% of total DIN (Figure 3). Si-SiO₄ decreased in summer too, whereas TON increased, although their trends were close to the limit of statistical significance (Table 3). N-NH₄ decreased significantly in winter and DIN even in spring. TON and N-NO₂ increased significantly in autumn and TON even in spring (Table 3).

Table 3.Download as CSV 

Results of the Mann-Kendall test for detection of long-term trends (2006–2015) in the environmental parameters (WEMOi, Western Mediterranean Oscillation index; Z_mix/Z_eu, mixing zone and euphotic zone ratio; Si-SiO₄, silicate; P-PO₄, orthophosphate; TP, total phosphorous; N-NO₃, nitrate; N-NO₂, nitrite; N-NH₄, ammonia; DIN, dissolved inorganic nitrogen; TN, total nitrogen; TON, total organic nitrogen) in Bidighinzu Lake. Significant trends are in bold (S = Kendall score, it indicates the trend direction; p = significance; n = number of observations).

	Winter			Spring			Summer			Autumn
	S	p	n	S	p	n	S	p	n	S	p	n
WEMOi	33	0.568	30	40	0.486	30	-34	0.556	30	15	0.802	30
Air Temperature	27	0.566	26	24	0.649	28	-6	0.912	26	20	0.657	25
Rainfall	38	0.440	27	20	0.721	29	53	0.303	28	20	0.707	28
Water Temperature	-17	0.544	24	2	0.986	30	52	0.469	35	21	0.620	24
Water Transparency	-21	0.444	24	62	0.274	30	-19	0.797	35	45	0.271	24
Zmix/Zeu	9	0.715	24	2	0.985	30	181	0.007	35	16	0.671	24
Si-SiO4	-50	0.063	24	100	0.077	30	-141	0.047	35	-65	0.112	24
P-PO4	-10	0.773	24	17	0.775	30	-62	0.386	35	-16	0.710	24
TP	-14	0.621	24	-27	0.643	30	-175	0.013	35	-39	0.346	24
N-NO2	-23	0.404	24	-50	0.382	30	132	0.063	35	83	0.042	24
N-NO3	12	0.677	24	-53	0.353	30	73	0.306	35	27	0.519	24
N-NH4	-56	0.037	24	-67	0.239	30	-159	0.025	35	-33	0.427	24
DIN	10	0.733	24	-157	0.005	30	-161	0.023	35	35	0.399	24
TN	-6	0.850	24	103	0.069	30	113	0.112	35	63	0.124	24
TON	32	0.163	24	128	0.027	30	221	0.040	35	113	0.031	24

Figure 3.

Multiannual (2006–2015) variation of dissolved inorganic nitrogen (DIN) (upper panel) and of the contribution of nitrate (N-NO₃), nitrite (N-NO₂) and ammonium (N-NH₄) to DIN (lower panel) in Bidighinzu Lake in summer. Each black dot represents a single sample and each column represents the mean value.

A significant increasing trend resulted for total phytoplankton cell density in summer during the study decade (Table 4). This increment was confirmed also when the year 2010 (exceptional peak of the total phytoplankton density) was excluded from the analysis (Mann Kendall test: S = 190, p = 0.002, n = 35). Summer total phytoplankton cell density varied from about 20 × 10⁶ cells l^-1 in 2006 to about 126 × 10⁶ cells l^-1 in 2015, with an exceptional event (about 1 × 10⁹ cells l^-1) in 2010, when a bloom of Cyanophyceae was observed (Figure 4). Cyanophyceae contributed most to the total phytoplankton cell density (> 60%), followed by Chlorophyceae (1–25%) and Bacillariophyceae (1–16%) (Figure 5). Significant increasing trends were observed for Bacillariophyceae and Cyanophyceae cell density in summer over the study period (Table 4). Bacillariophyceae cell density increased from a summer mean value of about 1 × 10⁶ cells l^-1 in 2006 to about 7 × 10⁶ cells l^-1 in 2015 (Figure 6A). Cyanophyceae cell density increased from a summer mean value of about 13 × 10⁶ cells l^-1 in 2006 to about 1 × 10⁹ cells l^-1 in 2010, decreasing to about 107 × 10⁶ cells l^-1 in 2015 (Figure 6C). Instead, no significant multiannual trend was observed for Chlorophyceae cell density (Table 4).

Figure 4.

Multiannual (2006–2015) variation of total cell density (upper panel) and mean cell volume (lower panel) of the whole phytoplankton community in Bidighinzu Lake in summer. Each black dot represents a single sample and each column represents the mean value.

Figure 5.

Multiannual (2006–2015) variation of the percentage contribution of all phytoplankton classes to the total cell density (upper panel) and mean cell volume (lower panel) of the whole phytoplankton community in Bidighinzu Lake in summer. BAC = Bacillariophyceae, CHL = Chlorophyceae, CHR = Chrysophyceae, CON = Conjugatophyceae, CRY = Cryptophyceae, CYA = Cyanophyceae, DIN = Dinophyceae, EUG = Euglenophyceae.

Figure 6.

Multiannual (2006–2015) variation of total cell density (upper panels) and mean cell volume (lower panels) of Bacillariophyceae (A and B) and Cyanophyceae (C and D) in Bidighinzu Lake in summer. Each black dot represents a sample and each column represents the mean value.

Table 4.Download as CSV 

Results of the Mann-Kendall test for detection of long-term trends (2006–2015) in the phytoplankton variables in Bidighinzu Lake. Significant trends are in bold (S = Kendall score, it indicates the trend direction; p = significance; n = number of observations).

	Winter			Spring			Summer			Autumn
	S	p	n	S	p	n	S	p	n	S	p	n
Total cell density	24	0.300	24	106	0.065	30	175	0.009	34	43	0.236	24
Bacillariophyceae	30	0.192	24	-142	0.008	30	249	0.000	34	81	0.024	24
Chlorophyceae	4	0.892	24	12	0.836	30	-85	0.213	34	-65	0.071	24
Chrysophyceae	22	0.344	24	23	0.677	30	201	0.003	34	51	0.058	24
Cryptophyceae	-2	0.964	24	62	0.252	30	47	0.495	34	-49	0.176	24
Conjugatophyceae	-1	1	24	-140	0.007	30	-212	0.002	34	-99	0.006	24
Cyanophyceae	60	0.008	24	130	0.015	30	167	0.014	34	65	0.071	24
Dinophyceae	38	0.074	24	76	0.139	30	-68	0.318	34	12	0.744	24
Euglenophyceae	24	0.300	24	39	0.476	30	-10	0.894	34	-29	0.430	24
Mean cell volume	-18	0.444	24	-105	0.063	30	-129	0.030	34	-9	0.821	24
Bacillariophyceae	-18	0.444	24	168	0.002	30	-197	0.004	34	-29	0.430	24
Chlorophyceae	8	0.753	24	14	0.807	30	51	0.458	34	43	0.236	24
Chrysophyceae	7	0.766	24	5	0.910	30	-128	0.049	34	22	0.526	24
Cryptophyceae	17	0.428	24	34	0.536	30	-47	0.406	34	87	0.015	24
Conjugatophyceae	25	0.235	24	46	0.064	30	112	0.072	34	129	0.000	24
Cyanophyceae	-29	0.166	24	-142	0.005	30	-195	0.004	34	-71	0.048	24
Dinophyceae	-8	0.465	24	26	0.260	30	37	0.453	34	18	0.244	24
Euglenophyceae	-3	0.921	24	-41	0.378	30	-7	0.915	34	22	0.526	24
Total cell biomass	16	0.599	24	23	0.695	30	89	0.192	34	17	0.691	24
Bacillariophyceae	24	0.420	24	-141	0.012	30	141	0.038	34	89	0.029	24
Chlorophyceae	-6	0.861	24	21	0.721	30	-85	0.213	34	-53	0.197	24
Chrysophyceae	6	0.861	24	30	0.602	30	121	0.075	34	111	0.006	24
Cryptophyceae	-18	0.551	24	107	0.059	30	49	0.477	34	-25	0.551	24
Conjugatophyceae	2	0.972	24	-143	0.009	30	-166	0.014	34	-53	0.197	24
Cyanophyceae	20	0.505	24	67	0.239	30	13	0.859	34	25	0.551	24
Dinophyceae	28	0.323	24	89	0.101	30	-64	0.348	34	28	0.486	24
Euglenophyceae	16	0.599	24	-8	0.900	30	-19	0.790	34	-7	0.882	24
Chlorophyll a	4	0.892	24	15	0.803	30	9	0.909	34	-139	0.100	24

Considering the mean cell volume of the whole summer phytoplankton community, a multiannual decreasing trend was detected in the total assemblage, with a strong decrement in Bacillariophyceae and Cyanophyceae (Table 4). Summer Chrysophyceae cell volume also decreased, but to a lesser extent. The mean cell volume of the whole phytoplankton community decreased from a summer mean value of about 11 µm³ in 2006 to about 1 µm³ in 2015, apart from a high summer mean value of 18 µm³ in 2012 (Figure 4). Summer Bacillariophyceae cell volume increased from a mean value of 45 µm³ in 2006 to 58 µm³ in 2009, then decreased to 4 µm³ in 2015 (Figure 6B). Summer Cyanophyceae cell volume decreased from a mean value of 0.081 µm³ in 2006 to 0.005 µm³ in 2015 (Figure 6D). Dinophyceae was the most important class contributing to the mean cell volume of the whole phytoplankton community (19–97%), followed by Conjugatophyceae (1–67%) and Euglenophyceae (1–38%) (Figure 5).

No significant multiannual trends were observed for total phytoplankton biomass and chlorophyll a (Table 4).

Opposite significant trends were observed for the Cyanophyceae Chroococcales and Bacillariophyceae Centrales: cell density increased (Mann-Kendall test: S = 197, p = 0.004, n = 34 and S = 253, p < 0.001, n = 34, respectively) while their cell volume decreased (Mann-Kendall test: S = -179, p = 0.008, n = 34 and S = -205, p = 0.002, n = 34, respectively). In addition, a multiannual decrement in summer was detected for the Cyanophyceae Nostocales only in cell density (Mann-Kendall test: S = -102, p = 0.040, n = 34; Suppl. materials 1, 2).

Chroococcales were mainly represented by Aphanocapsa sp., Merismopedia sp. and Chroococcus sp. (mean cell volume of 0.76 µm³, 1.12 µm³ and 77.15 µm³, respectively) at the beginning of the study decade and by Aphanothece sp., Aphanocapsa spp. and Merismopedia tenuissima Lemmermann (mean cell volume of 1.30 µm³, 0.44 µm³ and 0.54 µm³, respectively) at the end. Amongst Nostocales, Dolichospermum flos-aquae (Brébisson ex Bornet & Flahault) P. Wacklin, L. Hoffmann & J. Komárek (mean cell volume of 73 µm³) and Dolichospermum spiroides (Klebhan) Wacklin, L. Hoffmann & Komárek (mean cell volume of 519 µm³) determined the highest cell density values at the beginning and at the end of the considered period, respectively. Centrales were mainly represented by Cyclotella spp. and Aulacoseira granulata (Ehrenberg) Simonsen (mean cell volume of 1035 µm³ and 982 µm³, respectively) at the beginning of the study and by A. granulata and Aulacoseira granulata var. angustissima (Otto Müller) Simonsen (mean cell volume of 608 µm³ and 180 µm³, respectively) during the last years.

The results of the RDAs showed that, in summer, all environmental variables accounted for 57.8% and 53.8% of the variation in the 2006–2015 phytoplankton total cell density and volume, respectively. These results were confirmed when the years 2010 (exceptional peak of the total phytoplankton density) and 2015 (breaking of stratification due to the activation of the aeration system) were excluded from the analysis.

DIN (F = 4.21, p = 0.004), N-NH₄ (F = 3.97, p = 0.002) and TP (F = 3.18, p = 0.007) were significant environmental variables, which provided a greater explanation for the variability in summer total phytoplankton cell densitiy, followed by Z_mix/Z_eu (F = 2.62, p = 0.029) (Figure 7A). Total phytoplankton cell density and Cyanophyceae, Bacillariophyceae, Chrysophyceae and Euglenophyceae cell densities were negatively correlated with these nutrients, but Dinophyceae was positively correlated (Figure 7A). Instead, the cell density of total phytoplankton and of the phytoplankton classes mentioned above, with the exception of Dinophyceae, were positively correlated with Z_mix/Z_eu.

Figure 7.

Results of Redundancy Analysis on the relationships between environmental explanatory variables (vectors) and phytoplankton variables (responses), considering (A) the total phytoplankton cell density and (B) the mean phytoplankton cell volume data during the decade 2006–2015 in Bidighinzu Lake in summer. Eigenvalues of the first two axes are indicated by λ₁ and λ₂. Asterisks indicate statistical significance (*p < 0.05; **p < 0.01; ***p < 0.001) of environmental variables. WEMOi, Western Mediterranean Oscillation index; AirTem, air temperature; Rain, rainfall; Z_mix/Z_eu, mixing zone and euphotic zone ratio; WatTem, water temperature; Si-SiO₄, silicate; P-PO₄, orthophosphate; TP, total phosphorous; N-NO₃, nitrate; N-NO₂, nitrite; N-NH₄, ammonia; DIN, dissolved inorganic nitrogen; TN, total nitrogen; TON, total organic nitrogen; Bac, Bacillariophyceae; Chl, Chlorophyceae; Chr, Chrysophyceae; Con, Conjugatophyceae; Cry, Cryptophyceae; Cya, Cyanophyceae; Dino, Dinophyceae; Eug, Euglenophyceae; TotDens, total phytoplankton density; MeanVol, mean cell volume of the whole phytoplankton community.

TP (F = 3.34, p = 0.004), DIN (F = 2.82, p = 0.007) and N-NH₄ (F = 3.18, p = 0.009), followed by water temperature (F = 3.52, p = 0.027), provided a greater explanation for the variability in summer mean phytoplankton cell volume (Figure 7B). Mean cell volumes of the whole phytoplankton community and of Bacillariophyceae, Cyanophyceae, Dinophyceae and Conjugatophyceae were positively correlated with TP and negatively correlated with water temperature (Figure 7B). Euglenophyceae cell volume was negatively correlated with TP Chrysophyceae and Chlorophyceae cell volumes were related positively and negatively to DIN and N-NH_4, respectively.

A significant decreasing trend (Mann-Kendall test: S = -949, p < 0.001, n = 84) was observed for the WEMO index in summer during the period 1988–2015. One-way ANOVA highlighted significant differences in the WEMO index, TP and P–PO₄ amongst the four time-cycles of samplings (i.e. 1988–1989, 1994–1997, 2003, 2006–2015; Table 5). Mean values of the summer WEMO index increased from -0.1 in the first cycle (1988–1989, A) to 0.8 in the third cycle (2003, C), then decreased strongly to -1.0 in the last period (2006–2015, D) (Tukey’s test: B-D, p = 0.011; C-D, p < 0.001) (Figure 8). Mean values of summer TP decreased from 540.5 mg m^-3 at the beginning of the time series (1988–1989, A) to 246.9 mg m^-3 at the end (2006–2015, D) (Tukey’s test: A-D, p < 0.001) (Figure 8). The analysis did not reveal significant differences in total phytoplankton cell density amongst the different time cycles, even if significant differences were detected in total cell density of Bacillariophyceae, Chlorophyceae, Chrysophyceae, Cryptophyceae, Conjugatophyceae, Dinophyceae and Euglenophyceae, but not Cyanophyceae (Table 5). Instead, significant differences in mean cell volume of the whole phytoplankton community and in Bacillariophyceae, Chlorophyceae, Cryptophyceae, Cyanophyceae, Dinophyceae and Euglenophyceae cell volumes were assessed (Table 5). Summer mean cell volumes of total phytoplankton and Cyanophyceae increased at the beginning of the time series, from 0.004 µm³ and 0.001 µm³ to 2.48 µm³ and 0.02 µm³, respectively, from the first (1988–1989, A) to the second sampling cycle (1994–1997, B). Subsequently, summer mean values decreased strongly to 1.45 µm³ and to a value very close to 0 µm³ in the last sampling cycle (2006–2015, D) (Tukey’s test: B-D, p < 0.001 for both) (Figure 8).

Table 5.Download as CSV 

Results of the one-way ANOVA (F-test and p-value) to assess significant differences in selected environmental and phytoplankton variables amongst sampling year cycles (1988–1989, 1994–1997, 2003, 2006–2015) in summer in Bidighinzu Lake (WEMOi, Western Mediterranean Oscillation index; Z_mix/Z_eu, mixing zone-euphotic zone ratio; Si-SiO₄, silicate; P-PO₄, orthophosphate; TP, total phosphorous; N-NO₃, nitrate; N-NO₂, nitrite; N-NH₄, ammonia; DIN, dissolved inorganic nitrogen; TN, total nitrogen; TON, total organic nitrogen, ns = not significant).

	F	p		F	p		F	p
WEMOi	7.83	***	Total cell density	1.76	ns	Mean cell volume	3.60	*
Water Temperature	1.03	ns	Bacillariophyceae	10.92	***	Bacillariophyceae	77.56	***
Water Transparency	1.32	ns	Chlorophyceae	11.27	***	Chlorophyceae	406	***
Zmix/Zeu	1.24	ns	Chrysophyceae	13.35	***	Chrysophyceae	0.15	ns
Si-SiO4	2.26	ns	Cryptophyceae	20.45	***	Cryptophyceae	46.96	***
P-PO4	9.44	***	Conjugatophyceae	9.44	***	Conjugatophyceae	0.89	ns
TP	7.33	***	Cyanophyceae	2.01	ns	Cyanophyceae	123.3	***
N-NO2	0.55	ns	Dinophyceae	9.47	***	Dinophyceae	9.43	***
N-NO3	1.45	ns	Euglenophyceae	6.02	**	Euglenophyceae	3.91	*
N-NH4	0.25	ns
DIN	0.33	ns
TN	2.21	ns
TON	1.17	ns

*p < 0.05, **p < 0.01, ***p < 0.001

Figure 8.

Variation of the Western Mediterranean Oscillation index (WEMOi), total phosphorous (TP), mean cell volume of the whole phytoplankton community (Total phytoplankton) and mean Cyanophyceae volume during the four time cycles of samplings (A = 1988–1989, B = 1994–1997, C = 2003, D = 2006–2015) in summer in Bidighinzu Lake. The lowest, second lowest, middle, second highest and highest lines in the box plots represent the 10^th percentile, 25^th percentile, median, 75^th percentile and 90^th percentile, respectively. Means are represented by black dots.

Significant variations in land cover and land use have been assessed in the Bidighinzu Lake’s watershed during the last 20 years. The GIS analyses of CORINE maps for the time series 1990, 2000 and 2012, highlighted a decrement in Agricultural Area of about 7.2% from 1990 (32.59 km²) to 2000 (30.25 km²) and of a further 16.8% from 2000 to 2012 (25.17 km²), with a total decrement of about 23% from 1990 to 2012. On the other hand, Natural and semi-Natural Areas increased by about 14.6% from 1990 (16.63 km²) to 2000 (19.06 km²) and by a further 25.3% from 2000 to 2012 (23.89 km²), with a total increment of about 44% from 1990 to 2012. A noteworthy variation in Urban Area was also assessed with an increase of about 2.5% from 1990 (1.61 km²) to 2000 (1.65 km²) and of a further 4.8% from 2000 to 2012 (1.73 km²), with a total increment of about 7.5% from 1990 to 2012.

We observed strong decreases in summer concentrations of various nutrients, i.e. Si-SiO₄, TP, DIN and N-NH₄ and a significant but weaker increment of TON values during the decade 2006–2015 in Bidighizu Lake. It has been previously documented that the phosphorous reduction in summer has been on-going since 1978, immediately after the application of recovery actions for the reservoir and it was most likely favoured by the diversion of urban and industrial wastes being initiated in 1987 (Marchetti et al. 1992, Lugliè et al. 2001, Mariani et al. 2015). The ANOVA results of the present study suggest that the decrease in summer nitrogen and silicate and the increase in TON instead may have been more recent and they had not been reported so far (Mariani et al. 2015).

Explaining how the dissolved nitrogen forms vary in lakes is not straightforward as they depend upon variations in natural and pollutant sources and a variety of physical, biological and metabolic features of the lake (Quirós 2003). The strong decrement of DIN values in Bidighinzu Lake could be attributed to the concomitant decrease of N-NO₃ and N-NH₄ in the reservoir. In eutrophic lakes, such as the Bidighinzu Lake, nitrification may stop in conjunction with summer thermal stratification associated with hypolimnetic relatively low dissolved oxygen and low pH. As a consequence, denitrification may increase and lead to a complete natural depletion of nitrate. In hypoxic and anoxic conditions, N-NH₄ is mainly generated as a product of decomposition of the organic matter instead of N-NO₃ (Wetzel 2001). Consequently, N-NH₄ concentrations also depend on the quantities of settled organic matter and oxygen concentrations in the waters. Further, in hypoxic and anoxic conditions, denitrification causes the formation of volatile nitrogen compounds that are released from aquatic ecosystems, contributing to reduced concentrations of inorganic nitrogen compounds (Wetzel 2001). Unfortunately, the rate of denitrification was not measured in Bidighinzu Lake; however, based on oxygen concentrations over time (data not shown), it was likely to be relevant in summer in the majority of the considered annual cycles. N₂-fixation due to Cyanobacteria was also not measured. Finally, as summer is also the dry season in the Mediterranean climate, the reduction of external loadings in that season causes the loss of an important input of inorganic nitrogen carried by the drainage from river basin waters (Coppens et al. 2016, Padedda et al. 2010, 2017). In the present work, the GIS analysis of the Bidighinzu watershed land cover and land use variations over the last 20 years highlighted the significant decrease in agricultural areas and the opposing relevant increase in natural and semi-natural areas accompanied by a small increase in urban areas. These changes may have contributed significantly to a shift in material inputs, supporting the significant trends observed. In particular, considering the different forms of nitrogen, the significant increase of TON in the reservoir may be linked to the partially re-naturalisation of the watershed (Volungevicius et al. 2015). This latter, in turn, may also be responsible, at least in part, for the significant detriment of inorganic nitrogen compounds (DIN and N-NH₄) due to reduction in the use of fertilisers (Jomaa et al. 2016). This is a relevant outcome compared to other reservoirs in Sardinia, where dissolved inorganic nitrogen is generally the main component of TN. For example, Padedda et al. (2015) found a contribution of DIN around 60% of TN in Lake Cedrino, an ecosystem similar to Lake Bidighinzu.

Relevant changes in summer multiannual phytoplankton dynamics were observed simultaneously to the summer nutrient reduction in Bidighinze Lake during the decade 2006–2015. The total phytoplankton cell density increased significantly, whereas the mean cell volume of the whole phytoplankton community decreased significantly. Consequently, it is not surprising that the total phytoplankton biomass and chlorophyll a concentration did not show any significant trend in the same period. The phytoplankton classes that contributed more to total density in the analysed decade remained mainly Cyanophyceae, followed by Chlorophyceae and Bacillariophyceae as observed in the previous years (Mariani et al. 2015). While significant multiannual trends were not observed for Chlorophyceae, the Cyanophyceae and Bacillariohyceae cell density increased significantly from 2006 to 2015 and their mean cell volume decreased strongly, determining the same multiannual trends for total cell density and mean cell volume of the whole phytoplankton community. According to the ANOVA results, the increment in total phytoplankton cell density was recent (i.e. concerning the period 2006–2015), whereas the reduction in mean cell volume of the whole phytoplankton community and of Cyanophyceae was found to have begun earlier, from the late 1990s. The strong decrease in mean phytoplankton cell volume can be explained by changes in the taxa composition of phytoplankton, not observed at classes’ level, but present at order level and consisting of an increasing density of smaller taxa during the decade under study. In fact, the density of Cyanophyceae Chroococcales increased strongly, replacing the larger Cyanophyceae Nostocales at the end of the considered period. In addition, changes in Chroococcales species composition, with an increasing importance of smaller Chroococcales species (e.g. Aphanothece sp. and Merismopedia tenuissima instead of Aphanocapsa sp. and Chroococcus sp.) caused a strong decrement of the Chrooccocales cell volume during the years. Similarly, the density of Bacillariophyceae Centrales, smaller than the Bacillariophyceae Pennales, increased significantly in the reservoir. Amongst Centrales, smaller species (e.g. Aulacoseira granulata var. angustissima) replaced the larger ones (e.g. Cyclotella spp.) in the last years of the study, inducing the reduction of mean Centrales cell volume during the decade. The increasing density of smaller Cyanophyceae Chroococcales and Bacillariophyceae Centrales during the long-term decrement in nutrient concentrations was also observed in Temo Lake, a reservoir located in the same geographical area of Bidighinzu Lake in Sardinia (Pulina et al. 2016), as well as in the natural subalpine Maggiore Lake (Kamenir and Morabito 2009).

Statistical analysis revealed a strong relationship between the interannual patterns of algal nutrients and of total cell density and mean cell volume of the whole phytoplankton community at the Bidighinzu Lake in summer. Specifically, the summer TP and DIN (mainly N-NH₄) reduction affected, significantly and negatively, the summer total cell density. On the other hand, summer TP and DIN reduction significantly influenced the summer mean cell volume in a positive way. The decrement in phytoplankton size at lower nutrient concentrations was observed elsewhere, such as in another eutrophic Sardinian reservoir (Temo Lake, Pulina et al. 2016) and in different lake typology worldwide (e.g. Lehman 1991, Masson et al. 2000, Kamenir and Morabito 2009). The positive relationship between phytoplankton cell volume and nutrient concentrations was also detected experimentally in a highly eutrophic urban lake in Austria (Dokulil et al. 2007). Small-cell sized phytoplankton are expected to be favoured under lower nutrient concentration because of the high surface area to volume ratio and smaller diffusion boundary layer that enables rapid nutrient exchange through the cell surface (Harris 1994). In addition, the different strategies of nutrient utilisation can explain the distribution patterns of phytoplankton cell size classes along nutrient concentration gradients (Litchman and Klausmeier 2008). According to recent sudies on phytoplankton’s nutrient uptake-related traits, low nutrient environments should favour smaller-celled species while high nutrient environments should favour larger-celled species since maximum nutrient uptake velocity (V_max) and half-saturation constant for nutrient uptake (K) are positively correlated with cell size (Litchman et al. 2007).

In the hypertrophic Lake Arancio (Sicily, Italy), as well as in many other Sicilian reservoirs, nutrients have never been observed to play an important role in determining the structure of the phytoplankton community (Naselli-Flores and Barone 2007). The morphological changes observed in phytoplankton of this lake were instead strongly related to water mixing and underwater light availability (Naselli-Flores and Barone 2007). Specifically, Z_mix/Z_eu was detected as one of the most important factors in determining phytoplankton structure in a study of 21 Sicilian reservoirs (Naselli-Flores 2000). In our study case, a significant increasing trend was detected for Z_mix/Z_eu values during the present study decade 2006–2015. This change might have been partially linked to the activation of the water aeration system in the reservoir only in the period 2014 and 2015. This experimental activity prevented the summer thermal stratification of Bidighinzu Lake in 2015 but not in 2014 (Mariani et al. 2016). Our data analysis showed that the summer Z_mix/Z_eu value was a significant positive driver of summer dynamics of total phytoplankton density in Bidighinzu Lake, although its role was statistically weaker than that of nutrients. Higher Z_mix/Z_eu values in Bidighinzu Lake enhanced an increment in total phytoplankton density and especially in smaller species density, though significant relationships with total phytoplankton cell volume were not observed. These findings agreed with the evidence that smaller cells have a lower sinking rate compared to the larger ones (Litchman and Klausmeier 2008) and may be more efficient at utilising low light because of the smaller packaging effect, with self-shading of light-capturing pigments (Kirk 1994, Litchman et al. 2010).

The strong size reduction of marine phytoplankton under increasing ocean warming has also been well documented, although without univocal evidence (Gardner et al. 2011). It remains controversial whether the cell size reduction under increasing temperatures is i) a direct temperature effect or ii) an indirect effect mediated by changes in size selective grazing or by enhanced nutrient limitation that alter the selection on phytoplankton’s nutrient uptake-related traits, favouring smaller species over larger cells (Peter and Sommer 2013, Lewandowska et al. 2014). In the present work, no significant multiannual trends of meteo-climatic parameters were detected at the Bidighinzu Lake during the decade 2006–2015. However, summer water temperature significantly affected summer phytoplankton cell volume in the reservoir, though in a weaker way compared to the nutrients. The opposite trends observed for summer phytoplankton cell volume (decreasing) and summer water temperature (increasing), though not significant, seem to agree with the Atkinson’s ecological rule, according to which the average size of individuals is inversely related to temperature (Atkinson et al. 2003), also supporting similar findings proposed by several study cases in literature (Sommer et al. 2017b and reference therein).