Assessing the Trophic Condition of a Reservoir: A Combined Analysis of Watershed, Inter-Lake Connections and Internal Nutrient Loads

Abstract

Eutrophication is a pervasive issue in Mediterranean reservoirs, where external nutrient inputs and internal sediment releases interact to impair water quality and ecological stability. This study assessed the trophic condition of the artificial lake Cuga in Sardinia (Italy), mainly used for irrigation and providing potable water, by integrating watershed nutrient load estimates, inter-lake transfers, and internal phosphorus release. Field campaigns between July 2022 and May 2023 provided bi-monthly measurements of physical, chemical, and biological parameters, complemented by GIS-based land cover analysis and export coefficient modeling to quantify spatial nutrient sources. Additional phosphorus inputs from water transfers with a nearby reservoir were calculated, while internal sediment release was estimated using a calibrated mass balance model. Results revealed high nutrient concentrations, with mean total phosphorus of 128 mg P m⁻³, chlorophyll a averaging 9.9 mg m⁻³, and Secchi depth below 1 m, classifying the reservoir as eutrophic to hypertrophic under OECD and Carlson indices. Spatial loads were dominated by agricultural areas, while inter-lake transfers and internal sediment release contributed substantially to the overall phosphorus budget. The predictive Vollenweider model closely matched the observed conditions, confirming the robustness of the combined approach. Maintaining good ecological status in Mediterranean reservoirs is essential for safeguarding human well-being, as eutrophication degrades drinking-water quality, increases treatment costs, and can promote toxin-producing algal blooms with direct implications for public health. These findings highlight the need for integrated management strategies addressing both external and internal nutrient sources to mitigate eutrophication in Mediterranean reservoirs, which affects the ecosystem functioning and the related human needs and well-being.

1. Introduction

Eutrophication is a widespread process in freshwater ecosystems, characterized by excessive nutrient enrichment that can create self-sustaining feedback loops, leading to severe ecological degradation and serious economic and social consequences [1,2,3].

Beyond ecological impairment, eutrophication directly affects human health and well-being by compromising the safety of drinking-water supplies, increasing the risk of cyanotoxin exposure, and reducing the reliability of water resources for agriculture and domestic use. Healthy lake ecosystems provide key regulating services—such as natural water purification and buffering against climatic extremes—that are fundamental for public health and socio-economic stability [4].

Effective mitigation of eutrophication requires the control of both phosphorus (P) and nitrogen (N) inputs from anthropogenic sources, including sewage, industrial effluents, agricultural fertilizers, and livestock waste [5,6]. Vollenweider [7] first quantified the relationship between nutrient inputs, hydraulic retention, and water body morphometry, highlighting the susceptibility of lakes and reservoirs to eutrophication based on watershed nutrient loads and water residence time. The methodological framework and ecosystem-based approach proposed by Vollenweider [7] remain highly relevant today, as they continue to form the foundation of contemporary models assessing nutrient loading and trophic status in lakes and reservoirs [8]. Due to its broad scope and ability to provide fast-response, streamlined, yet scientifically robust assessments, this approach has maintained widespread application for decades, making it the gold standard for an immediate diagnosis of trophic status [9].

However, for an integrated management strategy that links lake response to specific territorial dynamics, recent literature suggests that deterministic process-based tools—such as the Soil and Water Assessment Tool (SWAT) or EROSION-3D—could be proposed for subsequent, in-depth studies [10,11,12].

These process-based simulations have demonstrated that nutrient export is often dominated by ‘Critical Source Areas’ (CSAs), where a small fraction of the watershed contributes the majority of the phosphorus (P) load [13]. Specifically, studies using tools like EROSION-3D and SWAT have revealed that sediment-associated P can account for over 60–90% of total P exports during high-flow events, highlighting the critical role of ‘legacy P’ stored in surface soils from decades of intensive land use [11,13]. Furthermore, modeling has quantified how hydraulic residence time and reservoir morphometry act as biogeochemical filters; these systems can retain up to 70–90% of the incoming nutrient load through sedimentation, which effectively creates a long-term internal nutrient reservoir that delays ecosystem recovery even after external loads are reduced [14,15].

These models enable detailed simulation of erosion and runoff, allowing for the precise mapping of source areas and the differentiation between current inputs and ‘sediment legacy’ [15,16]. This is particularly relevant in Mediterranean catchments, where intermittent hydrological regimes make nutrient transport a critical factor in eutrophication.

In Mediterranean catchments, soil erosion plays a disproportionate role in eutrophication due to the extreme intra-annual hydrological variability. During prolonged summer droughts, phosphorus accumulates in the surface soil and within dry stream beds; subsequent intense autumn rainfall events trigger massive erosive pulses that transport this material as particulate phosphorus (PP). For instance, in Mediterranean watersheds, it has been observed that a very small number of extreme events can account for up to 50–90% of the total annual sediment and nutrient yield [17,18]. This ‘flush effect’ makes erosion-driven nutrient transport a far more significant driver of trophic degradation than constant point sources, as it delivers high-concentration nutrient pulses that exceed the immediate assimilative capacity of the receiving reservoir.

Furthermore, such detailed modeling is especially complex in reservoirs, where sedimentation and retention dynamics differ significantly from natural lakes [14].

Despite these advantages, the implementation of such models requires extremely demanding data characterization and significant economic and computational investment. Therefore, they serve as support tools for more specific and localized technical interventions, logically following the global ecosystem assessment provided by mass balance models. This ensures a balance between analytical efficacy and the sustainability of resources allocated for monitoring.

Internal loading is an additional driver of degradation that exacerbates eutrophication, prolonging algal blooms and worsening water quality. This is particularly evident in stratified lakes and reservoirs, where nutrient accumulation in the hypolimnion under reducing conditions triggers the release of phosphorus from sediments—specifically iron-bound phosphorus, which remains stable under oxic conditions [19].

Building on this foundation, contemporary studies emphasize the integration of both external and internal nutrient sources to accurately assess trophic conditions [20].

External nutrient loading assessments can be conducted through direct experimental methods, which are often costly and time-intensive, or indirectly using empirically derived nutrient release coefficients from similar systems [21,22]. Nutrient sources are commonly classified as diffuse, such as agricultural and urban runoff and sediment release, or point sources, including municipal sewage and industrial discharges. Phosphorus is typically prioritized in modeling studies because of its role as a primary limiting nutrient in freshwater systems and its relative ease of management [23].

To specifically evaluate internal nutrient loading from lake sediments, several methodological approaches are employed. Direct measurements include sediment core incubations under controlled oxic and anoxic conditions, which allow quantification of phosphorus fluxes across the sediment–water interface [24]. These experiments are often complemented by in situ benthic chambers that capture nutrient release dynamics under natural stratification [25]. Indirect approaches rely on mass balance models, where differences between observed and expected nutrient concentrations during stratification are attributed to internal loading [26]. Empirical models, such as the Grøterud–Haaland approach [27], estimate internal fluxes by calibrating changes in hypolimnetic phosphorus concentrations over time. More advanced techniques integrate diagenetic modeling of sediment chemistry, combining redox-sensitive phosphorus fractions with hydrodynamic simulations to predict release rates [28]. Together, these methods provide complementary insights, balancing experimental precision with applicability across diverse lake systems.

Where implemented, inter-basin transfers and the hydraulic interconnection of reservoirs constitute a critical driver of eutrophication processes in receiving water bodies. When nutrient-rich waters are transferred into downstream reservoirs, the trophic status of the recipient system can deteriorate rapidly [29]. This occurs because external nutrient inputs from connected reservoirs often exceed the assimilative capacity of the receiving ecosystem, thereby intensifying phytoplankton growth and promoting harmful algal blooms (HABs). From a biogeochemical perspective, the transfer of hypereutrophic waters accelerates the accumulation of bioavailable nutrients in the epilimnion, while operational practices such as rapid flushing or altered residence times modify the balance between nutrient retention and export. Short hydraulic residence times can reduce natural attenuation processes, limiting denitrification and sedimentation, whereas prolonged retention favors internal recycling and sediment release of phosphorus under anoxic conditions [30]. Operationally, interconnections are often necessary to ensure water supply security, hydropower generation, and irrigation in regions with high seasonal variability in precipitation. In Mediterranean climates, for example, transfers between reservoirs are essential to balance spatial and temporal mismatches between water availability and demand [31]. However, these transfers inadvertently propagate eutrophic conditions downstream, creating a cascade effect where multiple reservoirs within a network exhibit degraded water quality.

Remediation strategies for eutrophication can be broadly divided into preventive and symptomatic measures. Preventive measures target pollution at the source and aim for long-term, systemic improvements, whereas symptomatic measures address specific ecological symptoms in the short term, often with limited efficacy [24]. An integrated understanding of both approaches is essential to designing effective management interventions.

In Sardinia (Italy, Western Mediterranean), reservoirs play a critical role in water supply, yet eutrophication affects nearly 80% of these water bodies [32]. Frequent algal blooms, often producing toxins, are driven by excessive nutrient loads and exacerbated by warm monomictic conditions typical of the region. Stratification during warmer months results in hypolimnetic oxygen depletion, promoting anoxia and the release of nutrients and harmful substances from sediments, thereby complicating recovery [33]. Within this context, hydrological connectivity among lakes represents a key component of water management but may also act as a mechanism for propagating environmental pressures when transfers involve nutrient-enriched waters. Regular and seasonally concentrated inter-lake transfers can intensify downstream eutrophication by redistributing external loads and enhancing nutrient accumulation and internal loading in receiving basins, which progressively function as sinks within interconnected reservoir networks. This dynamic poses a central management challenge: limiting internal nutrient recycling while maintaining cascade-based redistribution strategies. When supported by robust methods and reliable estimates, studies addressing these processes offer substantial potential for upscaling, enabling the identification of areas where hydrological connectivity shifts from a management benefit to a driver of environmental risk.

This study adopts a combined ecosystem and watershed approach to assess the trophic state of Lake Cuga, one of the 38 artificial lakes of Sardinia used mainly for drinking and irrigation purposes. Lake Cuga is a highly eutrophic reservoir characterized by substantial water transfers resulting from its interconnection with another reservoir. We integrate experimental analysis of the lake’s chemical and physical characteristics with a spatial assessment of nutrient sources. The objectives are to (i) characterize the lake’s trophic status, (ii) identify key nutrient sources within the watershed, (iii) construct nutrient mass balances, and (iv) model the potential effects of nutrient reduction scenarios. By linking internal and external nutrient dynamics, this dual approach provides a comprehensive framework to understand eutrophication processes and support effective management strategies for reservoirs. This is especially important in the Mediterranean context, where future climate change is expected to be a serious threat to the water resources of this region, with severe consequences for human needs and well-being [34,35].

2. Materials and Methods

2.1. Study Area

Lake Cuga is situated in the northwestern part of Sardinia, in the municipality of Uri (Figure 1). The urban area within the catchment basin includes part of the municipality of Ittiri. The lake was created by damming the river Cuga in 1965, but it was not completely filled until 1975. It is located at an altitude of 114 m above sea level and covers an area of about 58 × 10⁶ m², with a maximum depth of 45 m, an average depth of 11 m, and a volume of about 34 × 10⁶ m³ [36]. The water demand, which is higher than the average annual outflow, is met by transferring water to Lake Cuga from Lake Alto Temo through a pipeline. The water is used mainly for irrigation and drinking purposes. There are no special environmental protection statuses at the site. The hydrographic basin has an area of about 52 km² and is characterized by sparse woods and considerable agricultural and pastoral activity. The municipality of Uri, with about 2950 inhabitants, is not far from the lake. Since 1979, surveys of the reservoir have shown a state of hypertrophy, with significant oscillations from year to year: the annual averages of total phosphorus have always been higher than 100 mg P m⁻³, with maximum levels higher than 300 mg P m⁻³; the annual average of chlorophyll a, as an indicator of algal productivity, has varied from minimums of about 10 mg m⁻³ to maximums of over 30 mg m⁻³, with peaks of 70–80 mg m⁻³ [37]. Thanks to the long-term series of ecological data, Lake Cuga is one of the stations of site n° 10, Lake Ecosystem of Sardinia, belonging to the Italian Long Term Ecological Research Network (LTER-Italy; www.lteritalia.it).

2.2. Experimental Design

The investigations were carried out in the Cuga reservoir bi-monthly from July 2022 to May 2023. Sampling was conducted at the control platform of the reservoir management authority, Ente Acque della Sardegna (ENAS), located at the deepest point of the lake system and near the main intake tower.

Vertical water profiles were sampled using a Niskin bottle at predetermined depths: 0.5, 1, 2.5, 5, 7.5, 10, 15, and 20 m, with subsequent samples taken every 10 m to the lake bottom. In situ measurements of temperature (TEM), pH, conductivity (CON), and dissolved oxygen saturation (DOS) were taken using a Hydrolab HL7 multiparameter probe (Loveland, CO, USA). Water transparency was determined using a Secchi disk (SDT).

Collected water samples were stored in darkness and at low temperature, with laboratory analysis conducted within 24 h. Nutrient concentrations for total nitrogen (TN), dissolved inorganic nitrogen (DIN, calculated as the sum of nitrate + nitrite + ammonia), total phosphorus (TP), reactive phosphorus (RP) and reactive soluble silica (RSS) were determined according to APHA [38]. Chlorophyll a (CHL) was determined according to Golterman et al. [39].

2.3. Estimation of Spatial Nutrient Loads from Watershed

The calculation of nutrient loads was estimated by means of a theoretical statistical methodology following the export coefficient modeling approach [40]. The reported nutrient loads are strictly associated with surface runoff processes. The estimation, based on land-use specific export coefficients, considers the nutrients (TN and TP) transported from the soil surface to the drainage network. This includes both the dissolved fraction carried by the water flow and the particulate fraction bound to eroded soil particles (sediment-associated transport). Data on point and non-point sources have been converted by nutrient export coefficients into the same standardized unit of measurement (tons) of phosphorus and nitrogen (Table 1) to derive a theoretical potential load from the watershed.

Nutrient loads = ∑ source × export coefficient

The watershed area was identified through a precise definition of the natural boundaries of the drainage basin underlying Lake Cuga, through Geographic Information Systems analysis (Figure 1).

Table 1. Export coefficients used for the calculation of the potential nutrient load of phosphorus and nitrogen for the sources present in the watershed of Lake Cuga. (LA = Livestock Activity; UA = Urban Areas; NsNA = Natural and semi-Natural Areas; AA = Agricultural Areas).

Type	Source	Sub Category	Units	P Export	N Export	Reference
Point	LA	ovine/caprine	Kg × unit y−1	0.8	0.2482	[41,42]
		bovine	Kg × unit y−1	7.4	2.74115
		equine	Kg × unit y−1	8.7	3.09885
		swine	Kg × unit y−1	3.8	0.5694
		poultry	Kg × unit y−1	0.17	0.02555
Non-point	UA	All areas identified in the preceding subdivision 1 (artificial surfaces) of the 4th-level CORINE LC	kg × ha y−1	0.4	8	[43,44]
	NsNA	All areas identified in category 3 (forests and semi-natural areas), except the category 3315 (beds of streams wider than 25 m) + category 244 (agroforestry areas) of the 4th-level CORINE LC	kg × ha y−1	0.30	2.0002
	AA	All areas identified in category 2 (used agricultural surfaces), except category 244 (agroforestry areas) of the 4th level CORINE LC	kg × ha y−1	0.60	15.99795

Table 1. Export coefficients used for the calculation of the potential nutrient load of phosphorus and nitrogen for the sources present in the watershed of Lake Cuga. (LA = Livestock Activity; UA = Urban Areas; NsNA = Natural and semi-Natural Areas; AA = Agricultural Areas).

Type	Source	Sub Category	Units	P Export	N Export	Reference
Point	LA	ovine/caprine	Kg × unit y−1	0.8	0.2482	[41,42]
		bovine	Kg × unit y−1	7.4	2.74115
		equine	Kg × unit y−1	8.7	3.09885
		swine	Kg × unit y−1	3.8	0.5694
		poultry	Kg × unit y−1	0.17	0.02555
Non-point	UA	All areas identified in the preceding subdivision 1 (artificial surfaces) of the 4th-level CORINE LC	kg × ha y−1	0.4	8	[43,44]
	NsNA	All areas identified in category 3 (forests and semi-natural areas), except the category 3315 (beds of streams wider than 25 m) + category 244 (agroforestry areas) of the 4th-level CORINE LC	kg × ha y−1	0.30	2.0002
	AA	All areas identified in category 2 (used agricultural surfaces), except category 244 (agroforestry areas) of the 4th level CORINE LC	kg × ha y−1	0.60	15.99795

2.3.1. Point Sources

The point loads were calculated exclusively for zootechnical sources, as contributions from residential and fluctuating populations (UW) and from industrial activities (IA) were absent in the watershed. Data for Livestock Activities (LA) were obtained from the Banca Dati Nazionale dell’Anagrafe Zootecnica (BDN) managed by the Italian Ministry of Health [45], with reference to the various animal species (bovine, ovine/caprine, swine, equine, and poultry), and relating to a standard median condition for the period considered (January 2023).

2.3.2. Non-Point Sources

The contribution of non-point sources to the extent of Agricultural Areas (AA), Urban Areas (UA) and Natural or semi-Natural Areas (NsNA) was estimated through a GIS land cover analysis. The analysis focused on evaluating land cover using object-oriented classification techniques. These techniques are based on polygons generated from the segmentation and automatic classification of medium spatial resolution digital images. The analyses were performed on the vector layer of CORINE Land Cover, version 2020_20 u1, released in May 2019 [46].

The analysis covered an area of approximately 58.42 km². Each segmentation analysis was calibrated to obtain objects consistent with the CLC 2000 nomenclature system (minimum mappable unit: 25 ha). The thematic classes were subdivided according to this hierarchical structure, ensuring that lower-order polygons (3rd hierarchical level CLC; EEA, 2000) retain information from the higher hierarchical level. This approach provides all necessary information for subsequent and more homogeneous aggregation of areas/classes for specific use in calculating released nutritional loads. The analysis of the study area resulted in the classification of 11 land use/cover classes (Table 1). Homogeneous polygons/areas were identified, and for the purpose of calculating, the elements identified at the third level were aggregated into 4 macro-classes, maintaining the representativeness of the first level as much as possible.

2.4. Estimation of Inter-Lake Connection Nutrient Loads from Lake Alto Temo

The input of nutrient load from water transferred from Lake Alto Temo was calculated from the total volumetric supplies multiplied by the average hypolimnetic water phosphorus concentration, where the waters are taken from. The data relating to the volumes and the total phosphorus concentrations of water transferred from Lake Alto Temo were kindly provided by ENAS.

2.5. Estimation of Internal Nutrient Loads from Lake Cuga Sediments

The internal load, i.e., the release of phosphorus accumulated in the sediment, was estimated using the model proposed by Grøterud and Haaland [27]. The model assumes that phosphorus released from the sediment is produced and occurs during the hypolimnetic stratification period, becoming available throughout the entire water column at the onset of the circulation phase. The model was calibrated by assuming that the difference between the measured and simulated phosphorus concentrations in a water body at two successive times, i.e., at the beginning and end of summer stratification, is due to the internal load. This method is applicable and robust in shallow lakes and is able to provide an approximate estimate, especially when direct measurements of the internal phosphorus load are lacking. If the lake behaves as a completely mixed reactor, a first-order differential equation is used to estimate the change in phosphorus concentration with time due to a change in the input value. The equation applied to the case of Lake Cuga is shown below:

$${\displaystyle \raisebox{1ex}{$\mathrm{d}\mathrm{P}$}\!\left/ \!\raisebox{-1ex}{$\mathrm{d}\mathrm{t}$}\right.}=\mathrm{q}\ast {\displaystyle \raisebox{1ex}{${\mathrm{P}}_{\mathrm{i}}$}\!\left/ \!\raisebox{-1ex}{$\mathrm{V}$}\right.}-\mathrm{q}\ast {\displaystyle \raisebox{1ex}{$\mathrm{P}$}\!\left/ \!\raisebox{-1ex}{$\mathrm{V}$}\right.}$$

or in its integrated form:

$${\mathrm{P}}_{\mathrm{t}}={\mathrm{P}}_{\mathrm{i}}+\left({\mathrm{P}}_0-{\mathrm{P}}_{\mathrm{i}}\right)\ast \mathrm{e}-\mathrm{q}\ast {\displaystyle \raisebox{1ex}{$\mathrm{t}$}\!\left/ \!\raisebox{-1ex}{$\mathrm{V}$}\right.}$$

Internal loading can be calculated as:

$${\mathrm{P}}_{\mathrm{i}\mathrm{n}\mathrm{t}}={\mathrm{P}}_{\mathrm{t}\mathrm{m}\mathrm{e}\mathrm{a}}-{\mathrm{P}}_{\mathrm{t}}$$

where

P_i = Phosphorus concentration in the inflow;

P₀ = Phosphorus concentration in the lake at the beginning of stratification;

P_t = Phosphorus concentration at time t at the end of circulation;

P_{t mea} = Phosphorus measured concentration after time t;

q = discharge from the lake;

t = days of stratification;

V = volume of the lake.

2.6. Total Mass Balance and Predictive Modeling

The mass balance was limited to phosphorus, a nutrient for which modeling tools are available and reasonably used, being a key nutrient in controlling the primary production in Sardinian reservoirs [32,47,48]. The Vollenweider model [21] in the variant for shallow water bodies and artificial reservoirs was adopted, being already proficiently tested in various regional contexts of Sardinian lakes [42,49,50,51,52]. This model synthesizes the standard OECD equations for the relations between the external phosphorus loads (P_ext) and the expected phosphorus lake concentration (P_exp) and chlorophyll a (CHL_exp) as a function of the average lake water residence time (t_w), as calculated by two equations:

$${\mathrm{P}}_{\mathrm{e}\mathrm{x}\mathrm{p}}=1.22{[{\displaystyle \raisebox{1ex}{${\mathrm{P}}_{\mathrm{e}\mathrm{x}\mathrm{t}}$}\!\left/ \!\raisebox{-1ex}{$(1+{\mathrm{t}}_{\mathrm{w}})$}\right.}]}^{0.87}$$

$${\mathrm{C}\mathrm{H}\mathrm{L}}_{\mathrm{e}\mathrm{x}\mathrm{p}}=0.18{{\mathrm{P}}_{\mathrm{e}\mathrm{x}\mathrm{p}}}^{1.09}$$

The P_ext concentration was determined considering the sum of the external phosphorus loads (spatial and inter-lake connection input), calculated as in Section 2.3 and Section 2.4, while water residence time was derived from the ratio between the lake volume (LV) and the water inflow from the watershed (WI) over the year of investigation. Specifically, the lake volume was officially estimated by the Watershed Authority of the Sardinia Autonomous Region [53]. To calculate the water inflow (WI), the monthly cumulative precipitation (CP), measured at a station representative of the watershed (Cantoniera Rudas–Alghero; lon. 8.38111 lat. 40.60989), was multiplied by the average runoff coefficient (0.28) for the watershed area [54].

To the expected phosphorus concentration, derived from external loads, the phosphorus concentration resulting from internal releases from the sediment was subsequently added:

$${\mathrm{P}}_{\mathrm{t}\mathrm{o}\mathrm{t}}={\mathrm{P}}_{\mathrm{e}\mathrm{x}\mathrm{p}}+{\mathrm{P}}_{\mathrm{i}\mathrm{n}\mathrm{t}}$$

To evaluate the model’s consistency, the total phosphorus concentration of the lake (P_tot) was compared with the experimentally measured phosphorus lake concentration (TP). Following this verification, we simulated the effects of reducing external (P_ext) and/or internal (P_int) phosphorus loads on the lake’s trophic state. The inherent uncertainties of this empirical approach were addressed by considering the variability of input data, such as export coefficients and hydrological fluctuations. According to Harmel et al. [55], uncertainty in nutrient load estimations for small watersheds typically ranges between 10% and 30%. This variability is particularly relevant in Mediterranean systems due to episodic ‘flush effects’. Accordingly, the results were interpreted as best-estimates, and the resulting trophic shifts were assessed using the OECD probabilistic diagrams [21], which naturally account for the uncertainty in trophic classification.

2.7. Trophic State Evaluation

To assess the trophic status of Lake Cuga, the percentage probability of classification within a given trophic category was determined using three complementary approaches: the OECD fixed-interval criteria, the OECD probability distribution model [21], and Carlson’s Trophic State Index (TSI) [56]. The OECD procedure estimates the probability distribution across five trophic classes (ultra-oligotrophic, oligotrophic, mesotrophic, eutrophic, and hypertrophic) based on the following parameters: annual mean TP and CHL concentration, maximum CHL concentration, and annual mean SDT.

The TSI, in contrast, is derived from monthly measurements of TP, CHL, and SDT, incorporating values from each sampling event. The index ranges from 0 to 100, with thresholds defining the same trophic categories (0–20: ultra-oligotrophic, 20–40: oligotrophic, 40–50: mesotrophic, 50–70: eutrophic, 70: hypertrophic).

3. Results

3.1. Lake Water Conditions

During the study period, the maximum water temperature (27.0 °C) was recorded at the surface in September 2022, while the minimum (9.4 °C) occurred at 30 m depth in January 2023 (Figure 2a). Monthly water column averages (

Table 2. Monthly values of lake volume (LV), cumulative precipitation (CP), Watershed inflow (WI), temperature (TEM), pH, dissolved oxygen saturation (DOS), dissolved inorganic nitrogen (DIN), total nitrogen (TN), reactive phosphorus (RP), total phosphorus (TP), reactive soluble silica (RSS) and chlorophyll a (CHL).

Month	LV Volume Volume	CP	WI	TEM	pH	DOS	TP	RP	DIN	TN	RSS	CHL
	106 m3	mm	103 m3	°C		%	mg P m−3	mg P m−3	mg N m−3	mg N m−3	mg Si l−1	mg m−3
Jun 22	8.67	1.2	19.6
Jul 22	7.24	0	0.0	20.8	8.35	53.3	100	54	257	1310	3.80	9.7
Aug 22	8.3	67.2	1099.2
Sep 22	9.31	78.8	1288.9	22.4	8.68	48.6	140	45	219	1323	4.74	9.8
Oct 22	7.77	13	212.6
Nov 22	10.1	126.8	2074.1	14.8	7.26	63.1	128	36	423	1833	3.70	5.3
Dec 22	11.27	43.4	709.9
Jan 23	14.65	89.39	1462.2	10.2	7.64	74.1	171	63	585	2241	7.01	2.3
Feb 23	13.47	31.48	514.9
Mar 23	13.36	18.54	303.3	13.0	7.86	67.9	110	58	547	1644	7.03	3.0
Apr 23	10.75	26.15	427.7
May 23	7.51	6.35	103.9	18.4	8.13	65.4	118	45	360	2054	1.89	3.1

) ranged from 20.8 °C in July 2022 to 10.2 °C in January 2023, reflecting a clear seasonal pattern. Strong stratification was evident during the summer months, contrasting with the full winter circulation typical of a warm monomictic lake. This thermal structure drives the variability observed across all studied variables, particularly throughout the critical stratification phase.

pH (Figure 2b) exhibited marked vertical and temporal fluctuations typical of a eutrophic system. During stratification, both the maximum value (10.2) at the surface in September 2022 and the minimum (7.1) at 20 m depth in July 2022 were recorded. Monthly averages for the water column ranged from 7.3 in November 2022 to 8.7 in September 2022 (Table 2), reflecting the transition from low- to high-productivity periods. Similarly, during stratification, DOS (Figure 2c) reached peak values (>140%) in surface waters between July and September 2023, while the hypolimnion exhibited clear signs of anoxia. Accordingly, monthly water column averages ranged from 48.6% in September 2022 to 74.1% in January 2023 (Table 2). CON remained relatively stable over time, with monthly water column averages ranging from 323 µS cm⁻¹ in January 2023 to 403 µS cm⁻¹ in May 2023. SDT was generally low, varying between 0.5 m in November 2022 and 1.5 m in January 2023.

Nutrient concentrations also exhibited considerable variation according to stratification and circulation phases. During stratification, phosphorus compounds (Figure 2d,e) reached their maxima in September 2022 (with RP and TP at 280 mg P m⁻³ and 471 mg P m^−3, respectively, in the hypolimnion at 20 m depth) and their minima in July 2022 (2 mg P m⁻³ for RP and 42 mg P m⁻³ for TP at the surface). Regarding monthly averages for the water column (Table 2), values ranged from 36 mg P m⁻³ in November 2022 to 63 mg P m⁻³ in January 2023 for RP, and from 100 mg P m⁻³ in July 2022 to 171 mg P m⁻³ in January 2023 for TP.

Throughout stratification, nitrogen compounds DIN and TN (Figure 2f,g) exhibited different temporal maxima in the hypolimnion (at 20 m depth): in September 2022 for DIN (1091 mg N m⁻³) and in May 2023 for TN (2755 mg N m⁻³). Conversely, minima were recorded in the epilimnion in September 2022 for both DIN (42 mg N m⁻³) and TN (1013 mg N m⁻³). Regarding the monthly averages for the water column (Table 2), values ranged from 219 mg N m⁻³ in September 2022 to 585 mg N m⁻³ in January 2023 for DIN, and from 1310 mg N m⁻³ in July 2022 to 2241 mg N m⁻³ in January 2023 for TN.

RSS (Figure 2h) peaked at 8.1 mg Si l⁻¹ at 30 m depth during the winter (January–March 2023), dropping to a minimum of 0.4 mg Si l⁻¹ in the upper layers (0–5 m) in May 2023. Monthly averages (Table 2) ranged from 1.9 mg Si l⁻¹ (May 2023) to over 7.0 mg Si l⁻¹ (January–March 2023).

CHL (Figure 2i) concentrations reached their maximum at the surface in July 2022 (31 mg m⁻³) and a minimum of 1 mg m⁻³ at 20 m in March 2023. Monthly averages across the water column (Table 2) ranged from 2.3 mg m⁻³ (Jan 2023) to 9.8 mg m⁻³ (Sep 2022).

The analysis of the hydrological regime between June 2022 and May 2023 highlights a pronounced seasonality typical of Mediterranean catchments. Precipitation was almost absent during the summer months, reaching a minimum value of 0 mm in July 2022. Rainfall increased markedly from autumn onward, with peak values recorded in November 2022 (126.8 mm) and January 2023 (89.39 mm). Stream discharge closely followed the precipitation pattern: the absence of flow in July 2022 was followed by a rapid increase in hydrological inputs, with maximum discharge volumes observed in November 2022 (2074 × 10³ m³) and January 2023 (1462 × 10³ m³).

Consequently, the reservoir storage also exhibited substantial fluctuations. The minimum storage volume was recorded in July 2022 (7.24 × 10⁶ m³), whereas the maximum storage capacity was reached in January 2023, with a volume of 14.65 × 10⁶ m³. This corresponds to approximately a twofold increase in stored water compared to summer conditions.

Trophic State

Average values of TP, CHL, and SDT measured during the study period were used to classify the trophic status of Lake Cuga (

Table 3. Trophic state classification using OECD probability distribution [21] and TSI [56].

Method	Trophic State	SDT	TP	CHLmean	CHLmax	Mean
OECD %	Ultra-oligotrophy	0	0	0	0	0
	Oligotrophy	0	0	2	0	1
	Mesotrophy	0	8	38	32	20
	Eutrphy	11	59	53	50	43
	Iper-eutrophy	89	33	7	18	37
TSI	score	63	74	53	-	63

).

TP mean concentration (128.0 mg P m⁻³) corresponds to eutrophic conditions under OECD fixed-interval classification. OECD Probability distribution indicates 59% likelihood of eutrophy, 33% of hypertrophy, and 8% of mesotrophy, yielding a 92% probability of conditions beyond eutrophy. The TSI-TP value (74.1) places in the hypertrophic range.

Mean CHL photic-zone concentration (9.9 mg m⁻³) indicates eutrophy for OECD fixed-interval classification. OECD Probability distribution shows 53% eutrophy, 38% mesotrophy, and 7% hypertrophy, with a 60% chance of conditions exceeding eutrophy. The corresponding TSI-CHL value (60.7) indicates high eutrophy. CHL peak (31.0 mg m⁻³) suggests 50% probability of eutrophy, 32% mesotrophy, and 18% hypertrophy, with 68% likelihood of conditions beyond eutrophy.

SDT mean value (0.84 m) classifies the lake as hypertrophic under OECD fixed-interval criteria. OECD Probability distribution confirms hypertrophy (88%) with 12% eutrophy. The TSI-SDT value (62.5) instead places in the eutrophic range.

3.2. Spatial Nutrient Loads from Watershed

The total spatial nutrient loads from watershed account for 3.152 t P a⁻¹ for phosphorus and 76.414 t N a⁻¹, originating from the areas and activities described below.

3.2.1. Point Sources

Nutrient loads are attributed solely to LAs, although wastewater from UAs within the Lake Cuga watershed is discharged into the adjacent Rio Mannu di Porto Torres watershed. Data (

Table 4. Values of point and non-point (grouped by CORINE Land Cover Level 2 classes) sources and estimation of phosphorus and nitrogen nutrient loads from the Lake Cuga watershed, (LA = Livestock Activity; UA = Urban Areas; NsNA = Natural and semi-Natural Areas; AA = Agricultural Areas). Extent is expressed as number of heads for point sources and hectares for non-point sources.

Type	Source	Sub-Category	Extent	Pload (t ha−1 y−1)	Nload (t ha−1 y−1)	Pload (%)	Nload (%)
Point	LA	ovine/caprine	156	0.135	0.827	4.3	1.1
		bovine	13,459	0.014	0.107	0.4	0.1
		equine	59	0.006	0.046	0.2	0.1
		swine	269	0.013	0.038	0.4	0.0
		poultry	-	-	-	-	-
Non-point	UA	Discontinuous urban fabric	75.3	0.158	0.527	5.0	0.7
	NsNA	Forests	107.5	0.011	0.215	0.3	0.3
		Shrub and/or grasslands	937.0	0.092	1.874	2.9	2.5
	AA	Arable land	1180.2	0.706	18.880	22.4	24.7
		Permanent crops	2796.8	1.674	44.744	53.1	58.6
		Pastures	131.9	0.079	2.109	2.5	2.8
		Heterogeneous agricultural areas	440.5	0.264	7.046	8.4	9.2
Total			-	3.152	76.414	100	100

) indicate a significant ovine/caprine population, estimated at 13,494 units within the basin. This is followed by swine (269 units), bovine (156 units), and equine (59 units). Poultry units are estimated to be negligible.

In the overall total, the LA expresses a nutrient load of 0.168 t P a⁻¹ for phosphorus and 1.018 t N a⁻¹ for nitrogen. The most impactful activity is ovine, representing 4.3% of spatial phosphorus loads and 1.1% of spatial nitrogen loads.

3.2.2. Non-Point Sources

The largest quota of land use (Table 4) is for AA (4549.3 ha), followed by NsNA (1044.6 ha). The extent of UA is extremely negligible and is attributable to a limited number of urban areas (75.3 ha).

The largest share of loads is attributable to AA, amounting to 2.723 t P a⁻¹ (86.4% of spatial loads) and 72.780 t N a⁻¹ (95.2% of spatial loads); in this category, the main contribution comes from permanent crops (53.1% for P and 58.6% for N), followed by arable lands (representing 22.4% and 24.7% of the total spatial loads for P and N, respectively). Loads from UA are of lower magnitude, estimated at only 0.158 t P a⁻¹ (5% of total spatial loads) and 0.527 t N a⁻¹ (0.7% of spatial loads). Likewise, NsNA involve an estimated release of 0.103 t P a⁻¹ and 2.089 t N a⁻¹ (accounting for 3.2% and 2.8% of P and N spatial loads, respectively), almost entirely originating from shrub and/or grasslands.

3.3. Nutrient Loads from Lake Alto Temo Inter-Connection

The phosphorus load from pipeline water input from Lake Alto Temo was estimated by analyzing the water volumes transferred to Lake Cuga between July and October 2022. While the average annual monthly transfer is approximately 5.9 × 10⁶ m³, the total volume transferred during the study period was 23.7 × 10⁶ m³. The phosphorus concentration of the transferred water, representing the hypolimnetic average during this period, was 118 mg P m⁻³.

Based on the values reported above, an additional phosphorus load to Lake Cuga of approximately 2.78 t P a⁻¹ was therefore estimated, which should be added to the loads estimated through the spatial analysis (Section 3.2).

3.4. Internal Nutrient Loads

Based on the Input values and calculations performed (

Table 5. Model calibration parameters for the estimation of internal loads in Lake Cuga.

	Parameters	Units	Values
Input	Lake surface area	m2	1.450.000
	Lake outflow	m3	31.855.570
	Daily lake outflow (q)	m3 d−1	87.276
	Concentration at the beginning of stratification (P0)	mg P m3	124
	measured concentration after time t (Pt mea)	mg P m3	216
	Concentration in the inflow (Pi)	mg P m3	186
	Lake volume (V)	m3	10.200.000
	Start of stratification period	dd/mm/yy	15/05/2022
	End of stratification period	dd/mm/yy	15/10/2022
	Stratification days (t)	no. of days	153
Result	Internal load (Pint)	mg P m3	46.5

), the internal phosphorus sediment release has been estimated at 46.5 mg P m⁻³. This value appears quite high, but it is presumably accurate in relation to the high phosphorus availability identified during stratification in the deep hypolimnetic portion of the water column.

3.5. Total Mass Balance

As described in the materials and methods, the mass balance was limited to phosphorus, a nutrient for which modeling tools are available that can reasonably be used. The OECD Vollenweider predictive model [21] was adopted in the variant for shallow water bodies and artificial reservoirs, already used and tested for the artificial lakes of Sardinia. The model result (

Table 6. Modeling simulation of the expected concentration of phosphorus starting from the generated external land load of phosphorus and from the inputs of the Lake Alto Temo, from the internal load of sediments and from the hydraulic turnover time, according with OECD Vollenweider predictive model [21].

Parameters	Units	Values
Lake volume	m−3	10.200.000
Volume inflow	m−3	31.855.570
Water residence time	y	0.320
Concentration from spatial and transfers from Lake Alto Temo	mg P m−3	78.0
Concentration from internal load	mg P m−3	46.5
Final expected total phosphorus concentration	mg P m−3	124.5

), starting from a spatial phosphorus load of 3.15 t P a⁻¹ and the related additional load from water transfers from Lake Alto Temo of 2.78 t P a⁻¹, with an estimated hydraulic residence time of 0.320 years (equivalent to 117 days), for the 2022–2023 study year, would determine an annual average total phosphorus concentration (Pexp) in the lake of approximately 78 mg P m⁻³. As previously described, Lake Cuga is significantly affected by a substantial release of phosphorus from sediments, which must necessarily be added to the load described above. The model-estimated contribution of the internal load released (Pint) from sediments was 46.5 mg P m⁻³. The final expected total phosphorus concentration (Ptot) in the lake rises to 124.5 mg P m⁻³.

To verify that the coupling of the predictive models is effective in describing real conditions, this theoretically defined phosphorus concentration must be compared with that measured experimentally during the lake monitoring. The mean experimentally measured phosphorus concentration (TP) was 128 mg P m⁻³, a value practically coinciding with the model result, with an overestimation of only 3.5 mg P m⁻³. We can therefore state, obviously with all due caution given the different nature of the data derived from experimentally collected data and theoretical simulation, that the estimations carried out substantially fit. This means that the model calibration is statistically capable of predicting the effect of estimated loads on the lake trophic state. The model can therefore reasonably be used to simulate scenarios in which selective control (i.e., reduction) of the polluting sources that determine the external and internal phosphorus load is implemented.

4. Discussion

The results of this study confirm that Lake Cuga is subject to severe eutrophication, with phosphorus concentrations and chlorophyll a values consistently exceeding thresholds for eutrophic to hypertrophic conditions.

In this context, the observed thermal regime confirmed the warm monomictic nature of the lake, with pronounced seasonal stratification characterized by warmer surface waters in summer and cooler conditions at depth during winter. Physico-chemical parameters showed marked temporal variability, particularly for pH and dissolved oxygen saturation, which reached high values in surface waters during the summer period of high primary productivity, reflecting intense photosynthetic activity. Water transparency remained generally low, suggesting persistent phytoplankton biomass and suspended matter. Nutrient concentrations indicated high availability of both phosphorus and nitrogen in the water column, with occasional peaks at intermediate depths, possibly linked to internal processes such as sediment release or vertical mixing. Dissolved silica showed seasonal depletion in surface layers during spring, consistent with biological uptake by diatoms. Overall, chlorophyll-a values and nutrient concentrations highlight a highly productive system characterized by strong seasonal dynamics in biological activity and nutrient cycling.

The analysis of hydro-meteorological data confirms that soil erosion and sediment transport in the studied basin are largely driven by impulsive events, a characteristic feature of Mediterranean hydrological regimes. During the monitoring year, the main runoff events were concentrated between November 2022 (2.07 × 10⁶ m³) and January 2023 (1.46 × 10⁶ m³), triggered by significant precipitation peaks (e.g., 126.8 mm in November). Such events are essential for sediment delivery, as flash floods in Mediterranean rivers can transport most of the annual sediment load in just a few days [57].

Lake water quality parameters showed a rapid and pronounced response to these hydrological inputs. Concurrent with the January runoff peak, TP reached its maximum value of 171 mg P m⁻³, compared with 100 mg P m⁻³ during the summer period (July). This 71% increase highlights the critical role of catchment runoff in transporting sediment-bound particulate phosphorus to the lake. During the summer period (July–September), dissolved oxygen levels declined to 49–53%. This pattern, combined with the minimum lake volume observed in July (7.24 × 10⁶ m³), indicates the strong influence of thermal stratification. By isolating the hypolimnion from atmospheric exchange, stratification promotes hypoxic conditions and may enhance the internal release of nutrients from bottom sediments. Winter mixing is reflected in the increase in oxygen saturation (74% in January) and in the simultaneous redistribution of nutrients throughout the water column, as evidenced by the sharp increase in TN (from 1310 to 2241 mg N m⁻³). Overall, the observed dynamics indicate that the system is characterized by periods of intense external loading (November–January), followed by phases of relative stability and summer stratification.

The integration of watershed nutrient loads, inter-lake transfers, and internal sediment release provides a comprehensive picture of the drivers sustaining the high trophic status of Lake Cuga.

Results indicate that external loads generated within the watershed amount to 3.15 t P yr⁻¹, while water transfers from Lake Alto Temo contribute an additional 2.78 t P yr⁻¹, representing nearly the same magnitude as the locally generated load. When these inputs are translated into lake concentration through the Vollenweider model, they correspond to an expected external phosphorus concentration of 78 mg P m⁻³, before accounting for internal recycling.

Among the external sources, the watershed runoff represents a critical pathway for nutrient delivery. Land-use analysis showed that agricultural areas occupy more than 77% of the basin surface and generate 86.4% of the total spatial phosphorus load, corresponding to 2.72 t P y⁻¹. These results indicate that diffuse agricultural runoff is the dominant driver of nutrient inputs from the catchment. In Mediterranean basins characterized by episodic rainfall and intense runoff events, phosphorus transport from fertilized soils and cultivated land can occur in short but highly effective pulses, leading to elevated concentrations in inflowing waters [18]. The relevance of runoff-derived phosphorus is further supported by the high phosphorus concentrations measured in the lake water column during the study period (mean 128 mg P m⁻³), which largely exceed the concentration predicted by external loads alone when internal recycling is excluded. This suggests that the combined effect of agricultural runoff and sediment recycling creates a feedback mechanism that maintains high phosphorus availability throughout the year. The dominance of agricultural land use in the watershed aligns with previous findings that diffuse sources, particularly cropland runoff, are the primary contributors to phosphorus enrichment in Mediterranean reservoirs [58,59]. To provide insight into the seasonal dynamics of nutrient loading, the annual loads were disaggregated based on the monthly cumulative water inflow. Although this approach assumes a constant concentration (thus not accounting for specific event-based flushing effects), it effectively highlights peak loading periods. The highest phosphorus delivery coincides with the autumn and winter months, representing a critical window for the accumulation of the nutrient pool that supports primary production during the subsequent growing season. The highest values were recorded in September 2022 (0.93 t P mo⁻¹), November 2022 (1.50 t P mo⁻¹) and January 2023 (1.06 t P mo⁻¹). In contrast, the other months showed significantly lower values, all below 0.5 t P mo⁻¹. This seasonal partitioning, albeit indirect, offers a more nuanced view of the risk periods for algal blooms compared to a simple annual average.

In addition to local watershed contributions, the inter-lake transfer from Lake Alto Temo represents a major external input that directly introduces nutrient-rich water into the system. The estimated load of 2.78 t P yr⁻¹ corresponds to nearly 47% of the total external load, highlighting the importance of considering hydrological connectivity in amplifying nutrient pressures and reservoir management. This system is characterized by regular hydraulic pulses that introduce water already compromised in terms of quality, as it is itself markedly eutrophic. This volumetric input is substantial and occurs during the summer period to compensate for withdrawals from Lake Cuga to an extensive irrigation district. Over time, Lake Cuga has progressively assumed the functional role of a settling and redistribution basin within the interconnected reservoir network.

Internal phosphorus release from sediments was found to be substantial, contributing to an additional 46.5 mg P m⁻³, representing approximately 37% of the final predicted concentration (124.5 mg P m⁻³) and corroborating evidence that hypolimnetic anoxia and sediment recycling can perpetuate eutrophication even when external inputs are reduced [19,28]. The water body underwent a significant volumetric disturbance in 2016–2017, which plausibly led to the concentration of nutrients in a very low volume of water, facilitating their mineralization and sedimentation. This event has certainly facilitated the release of nutrients from the sediment in the years to come, continuously suspending the nutrients with each emptying and refilling of water by the dam. For a lake such as the Cuga, of a certain size and ancient age (if compared to other younger systems), the internal input can have a considerable importance in the total calculation of loads, also in relation to recent episodes of poor water storage that increase the mineralization of biomasses [60]. The close agreement between observed phosphorus concentrations and those predicted by the Vollenweider-type model underscores the robustness of this framework in Mediterranean contexts, while also demonstrating the necessity of including internal loading to avoid underestimation of trophic state.

The application of the Vollenweider model in this study showed a remarkably low discrepancy between the predicted P concentration (125 mg P m⁻³) and the observed mean (128 mg P m⁻³). While this high level of agreement supports the model’s robustness for the specific hydrological cycle of 2022–2023, we acknowledge that such a narrow gap may partly reflect the stability of the environmental conditions during the monitoring period rather than a universal predictive certainty.

A potential limitation of this study is the lack of multi-year validation. However, the reliability of our results is supported by long-term ecological data from Lake Cuga; specifically, Total Phosphorus (TP) concentrations have consistently exceeded 100 mg P m⁻³ over the last five years (e.g., 121 mg P m⁻³ in 2021–2022; 131 mg P m⁻³ in 2023–2024; 115 mg P m⁻³ in 2024–2025). These data confirm that our ‘snapshot’ year is representative of the lake’s chronic hypertrophic state. Furthermore, the model’s accuracy is reinforced by the integration of three independent data sources: land-use-based export coefficients, measured inter-lake transfers, and experimental sediment release rates. Although uncertainty in literature-based export coefficients can typically range between 20% and 30%, the convergence of these independent lines of evidence provides a solid diagnostic framework for the reservoir’s phosphorus mass balance, even in the absence of a decadal time series.

Similar studies in stratified reservoirs have shown that internal nutrient cycling can shift nutrient limitation patterns and prolong algal blooms, complicating management interventions [61].

Overall, the results demonstrate that eutrophication in Lake Cuga cannot be attributed to a single source but rather to the interaction of diffuse agricultural runoff, nutrient transfers from connected reservoirs, and substantial internal sediment recycling. Effective mitigation strategies must therefore address all three pathways simultaneously, as reducing only one component would likely produce limited improvements in the trophic condition of the reservoir.

By focusing on the most accessible interventions, such as managing inputs from the Lake Alto Temo transfer, we can immediately assess the potential trophic effects on the Lake Cuga receptor system. Complete elimination would indeed remove a significant load contribution but would also eliminate hydraulic input. This would affect the potential for dilution of lake concentrations due to a reduction in hydraulic turnover time, which would increase from the current 117 days to 453 days. Ultimately, this condition could lead to an expected lake concentration even higher than the current level, reaching 130 mg P m⁻³. From a hydraulic perspective, where the incoming water volume is greater than the lake’s volumetric size, nutrients and organisms can be removed before they exceed critical levels that trigger algal blooms, thus decreasing the hydraulic retention time [62]. Furthermore, halving the inputs from Lake Alto Temo would not yield significant results in the simulation, leading to only a slight reduction in the expected concentration to 120 mg P m⁻³. In this context, a simple regulation of inter-lake transfers should be calibrated not only on water supply needs but also by considering the quality of the transferred water. Adopting adaptive management criteria—such as limiting or blocking the inflow of hypertrophic waters during critical periods (e.g., summer), shifting transfers to winter, and accumulating as much volume as possible for summer use only during the mixing period when phosphorus concentrations are lower—could help mitigate the worsening of trophic conditions.

Limiting internal loading is critical if the current water management plan—transferring volumes from Lake Alto Temo to the Nurra agricultural district through Lake Cuga—is to remain viable. The release from sediments, by itself, would contribute about 37% of the total load. Ultimately, only sediment inertization techniques could provide a substantial contribution. With a potential reduction of 46.5 mg P m⁻³, the concentration value would decrease to 78 mg P m⁻³ (with an 80% probability of expressing a higher eutrophic/hyper-eutrophic condition and a 20% probability of mesotrophy). Within this framework, potential remediation strategies oriented to limit internal nutrient load are strongly constrained by cost–benefit assessments of the proposed interventions. Reliance on simplified measures is not recommended, as such approaches generally provide only palliative and short-term effects. Therefore, effective recovery requires the implementation of structural interventions capable of producing immediate and quantifiable outcomes. In this perspective, two potential strategies can be considered:

(a)

sediment capping to limit internal phosphorus release. This technique involves covering phosphorus-rich sediments with inert materials (such as sand or bentonite), thereby limiting resuspension and phosphorus release into the water. This requires careful selection of materials, as unsuitable ones may fail to provide the necessary barrier and could even release harmful substances. Furthermore, constant monitoring is necessary to ensure its effectiveness over time, requiring additional resources and expertise, which, combined with periodic material maintenance, can further increase costs and complications [46]. Use of chemical additives (phosphorus precipitants), such as compost or clays, can bind phosphorus in the sediments, reducing its availability for phytoplankton and consequently limiting algal blooms and associated problems. However, the effectiveness of this approach depends on various factors, including the choice of additive materials. Some composts, for example, may contain additional nutrients that could contribute to further system enrichment. This solution, therefore, proves to be extremely difficult to implement, also due to the basin’s nature, being shallow and thus susceptible to sediment resuspension caused by wind, currents, and waves, but especially due to the long periods of strong anoxia at the hypolimnetic sediment-water interface.

(b)

hypolimnetic oxygenation to mitigate anoxia and reduce nutrient mobilization from sediments. Under anoxic conditions, as observed in our case study, redox dynamics play a crucial role: the shift from oxidizing to reducing environments can release phosphorus from insoluble sediment compounds, making it bioavailable. Phosphorus is mainly found in deep waters as phosphate ions (PO₄³⁻), which, under oxidizing conditions (documented from late October to March), bind to iron, forming insoluble compounds. These compounds are vulnerable to background bacterial activity (from May to early October), causing the reduction of iron from Fe³⁺ to Fe²⁺ and making phosphorus compounds bound to ferric oxides unstable, releasing phosphates (HPO₄²⁻) into the water and making them bioavailable again, especially for phytoplankton that naturally prefer this ionic form. An oxygen injection system in the hypolimnetic waters during stratification could help limit the bioavailability of phosphorus. Several technologies are available to reduce anoxic conditions, such as micro-diffuser plates [63], full-air lift systems [64], siphoning [65], and pipeline aeration along the lakeshore [66]. The key with these techniques is to minimize turbulence in order to avoid resuspending sediments or, even worse, disrupting stratification [67]. These approaches, of course, cannot be applied extensively across the entire lake surface, but they can be optimized when implemented near the lake’s water intake points.

As a subsequent step, spatial planning should focus on limiting watershed sources. However, regulatory policies related to land use and management (e.g., agricultural practices) are often challenging to enforce and monitor; therefore, they should be accompanied by in-lake structural interventions to ensure long-term effectiveness [68]. The adoption of sustainable farming practices—such as rational fertilizer management, the use of cover crops, and the establishment of vegetated buffer strips along watercourses—can significantly reduce phosphorus and nitrogen inputs. At the same time, stricter control of livestock activities, through integrated waste management and reduced stocking densities, would help contain point-source loads. By primarily limiting point sources from the LA, phosphorus concentrations could reach 76.1 mg P m⁻³, without significant changes in the trophic state. Conversely, the more challenging reduction of non-point sources—i.e., loads from AA and UA runoff—would provide a substantially greater improvement, with concentrations estimated at 41.6 mg P m⁻³ (with a 56% probability of achieving a meso/oligotrophic condition and a 45% probability of remaining in a eutrophic/hyper-eutrophic state).

As can be inferred, these scenarios would significantly improve the trophic state of Lake Cuga. However, these projections are not able to retrieve the system at optimal condition, with concentrations that would still remain above the threshold level of 30–35 mg P m⁻³, characteristic of a mesotrophic to incipient eutrophic state. Anyway, the value is very close, leaving open the possibility for further interventions to support and assist the previously hypothesized remediation strategy.

In summary, even considering systemic and integrated approaches, this path is experimentally the least feasible in the short to medium term, as it requires careful planning and in-depth analysis of the chemical and environmental dynamics involved, as well as significant economic expenditure [69]. From a management perspective, the findings emphasize that effective mitigation must address both external and internal nutrient sources. The integration of watershed management with in-lake restoration strategies has been shown to be more effective than isolated approaches [59]. Future research should also explore adaptive management frameworks that account for hydrological transfers between reservoirs, as these connections can redistribute nutrient pressures across systems.

Despite the robustness of the integrated methodological framework adopted in this study, several sources of uncertainty should be acknowledged. Export coefficient modelling, although widely applied for watershed-scale nutrient assessments, relies on generalized coefficients derived from literature or comparable catchments. These coefficients may not fully capture local variability in land management practices, soil characteristics, hydrological connectivity, or seasonal runoff dynamics, potentially leading to uncertainties in the estimation of diffuse nutrient loads [40,70]. In Mediterranean catchments, these uncertainties may be further amplified by the highly episodic nature of rainfall and runoff events, which strongly control nutrient transport from agricultural and natural areas to receiving water bodies. As demonstrated by Harmel et al. [55], the cumulative error associated with load estimations can range between 20% and 30%.

Regarding the application of the Vollenweider framework to Lake Cuga, it is important to consider the specific hydraulic and thermal dynamics of this reservoir. Although water transfers from Lake Alto Temo introduce temporal variability in both phosphorus loads and lake volume, potentially challenging the steady-state assumptions of classical mass balance formulations, the model proved sufficiently robust to incorporate these external inputs and provide a reliable first-order estimate of trophic conditions. Vollenweider-type models have long been used as practical tools for predicting lake responses to nutrient loading; however, several studies have highlighted that simplified empirical approaches may not fully represent the spatial and temporal heterogeneity of nutrient cycling in stratified lakes. For example, Brett and Benjamin [9] showed that empirical phosphorus loading models generally reproduce long-term average concentrations but may overlook spatial variability in sediment phosphorus release and hydrodynamic transport processes.

Similarly, other studies [24,26,71] have demonstrated that internal phosphorus loading can represent a substantial fraction of total nutrient inputs in stratified eutrophic lakes and that its magnitude is strongly influenced by the duration of hypolimnetic anoxia and by sediment characteristics.

More detailed process-based models have been developed to address these limitations by explicitly simulating hydrodynamic circulation, sediment diagenesis, and nutrient cycling. Mechanistic eutrophication models, such as those described by Chapra et al. [72], allow a more refined representation of sediment–water interactions and the temporal dynamics of internal phosphorus release during stratification and mixing events.

Moreover, the role of climate variability, especially prolonged stratification under warming scenarios, warrants further investigation given its potential to exacerbate internal nutrient release [58].

However, these approaches require extensive monitoring datasets for calibration and validation, including high-frequency measurements of temperature, oxygen profiles, sediment fluxes, and inflow characteristics. Such detailed datasets are often unavailable for many Mediterranean reservoirs, where monitoring programs are typically limited in temporal and spatial resolution.

For this reason, the present study integrated the Vollenweider framework with a simplified mass-balance approach to estimate internal phosphorus loading based on hypolimnetic phosphorus accumulation during stratification. Although empirical estimates of internal loading cannot fully reproduce the complex interactions among redox conditions, temperature, sediment composition, and biological activity, they provide a reasonable approximation of the magnitude of sediment-derived nutrient inputs when direct measurements are unavailable. The synergy with field monitoring data, as performed in this study, helps constrain uncertainties and provides a robust estimate of nutrient inputs. Therefore, the values obtained in the present study should be interpreted as integrated and conservative estimates, suitable for ecosystem diagnosis and for defining management priorities and intervention scenarios.

Lake Cuga thus represents a system inherently susceptible to eutrophication. Indeed, the influence exerted by the upstream watershed on the water body remains considerable: it can be highlighted by the watershed/lake surface ratio. The watershed extent determines a ratio of 41:1, which is quite high according to Politi et al. [73], indicating eutrophication risk. However, other factors, such as slope, shape, and soil permeability, must also be considered when assessing nutrient dynamics. The degraded water quality observed in Lake Cuga has clear implications for human well-being, given the reservoir’s role in potable and irrigation supply. Elevated nutrient concentrations and the potential for harmful algal blooms can undermine water safety and increase treatment demands, whereas restoring ecological integrity would enhance the ecosystem’s capacity to support safe water provision and reduce health-related risks.

This study identifies the critical sources of nutrients deriving from land use and sediments. This is crucial to successfully predict eutrophication and select the correct decision-making in its prevention or mitigation. Lake Cuga exemplifies the challenges of managing eutrophication in Mediterranean reservoirs, where external agricultural inputs, inter-lake transfers, and internal sediment release interact to sustain high trophic states. The combined methodological approach adopted here provides a valuable template for similar systems, but further refinement through site-specific measurements and climate-sensitive modelling is essential to guide sustainable management strategies.

5. Conclusions

Eutrophication has multidimensional consequences linked to water quality, ecosystem and human health, as well as economic activities.

The integrated analysis conducted on the Lake Cuga basin highlighted that the trophic condition of the reservoir results from the combined effect of external inputs and internal nutrient recycling processes. Experimental data collected between 2022 and 2023 confirmed total phosphorus and chlorophyll a concentrations consistent with a eutrophic–hypertrophic state, in line with OECD and TSI classifications. The applied modelling demonstrated the robustness of the combined approach, with the Vollenweider model faithfully reproducing the observed conditions and quantifying the relative impact of the different nutrient sources. Spatial loads derived from agricultural surfaces represent the predominant share of nutrients, while water transfers from contiguous basins and internal sediment release contribute substantially to the overall phosphorus budget. The presence of monomictic stratification and hypoxic conditions in the deeper layers further promotes the mobilization of sedimentary phosphorus, amplifying the persistence of eutrophic phenomena. These findings emphasize the need for integrated management strategies that simultaneously address diffuse and point sources within the watershed, regulate inter-lake transfers, and mitigate internal release processes. Targeted interventions aimed at reducing agricultural loads, combined with water and sediment management practices, appear essential to contain eutrophication and ensure the ecological sustainability of Mediterranean reservoirs.