A Pragmatic Multi-Source Remote Sensing Framework for Calcite Whitings and Post-Wildfire Effects in the Gadouras Reservoir

Abstract

The Gadouras Reservoir, Rhodes Island’s primary water source, experiences recurrent whiting events—milky turbidity from calcium carbonate precipitation—that challenge treatment operations, with impacts compounded by a major 2023 wildfire in this fire-prone Mediterranean setting. To elucidate these dynamics, a pragmatic, multi-source monitoring framework integrates archived Sentinel-2 and Landsat imagery with treatment-plant records (2017–mid-2025). Unitless spectral indices (e.g., AreaBGR) for whiting detection and chlorophyll-a proxies are combined with laboratory measurements of turbidity, pH, total organic carbon, manganese, and hydrological metrics, analyzed via spatiotemporal Hovmöller diagrams, Pearson correlations, and interrupted time-series models. Two seasonal whiting regimes are identified: a biogenic summer mode (southern origin; elevated chlorophyll-a; water temperature > 15 °C; pH > 8.5) and a non-biogenic winter mode (northern inflows). Following the wildfire, the system exhibits characteristics that could be related to possible hypolimnetic anoxia, prolonged whiting, a ~50% rise in organic carbon, and a manganese excursion to ~0.4 mg L⁻¹ at the deeper intake. Crucially, the post-fire period shows a decoupling of AreaBGR from turbidity (r ≈ 0.233 versus ≈ 0.859 pre-fire)—a key diagnostic finding that confirms a fundamental shift in the composition and optical properties of suspended particulates. The manganese spike is best explained by the confluence of a wildfire-induced biogeochemical predisposition (anoxia and Mn mobilization) and a consequential operational decision (relocation to a deeper, Mn-rich intake). This framework establishes diagnostic baselines and thresholds for managing fire-impacted reservoirs, supports the use of remote sensing in data-scarce systems, and informs adaptive operations under increasing climate pressures.

1. Introduction

The Gadouras Reservoir in Rhodes serves as the primary water supply source for the city of Rhodes and surrounding areas, covering approximately 80% of the island’s drinking water needs, including coastal settlements and neighboring islands such as Chalki, Symi, and Kastellorizo, with a total storage capacity of approximately 67.5 million cubic meters (effective volume 63 million cubic meters) and designed to supply up to 26.5 million cubic meters annually by 2039. The construction of the reservoir began in 2002, with filling starting in 2005 and gradually completing by 2014, while it became fully operational for water supply on 11 August 2017. As a relatively new impoundment—operational since 2017 and still within the typical 3–11-year trophic upsurge phase following initial filling in 2005—the reservoir has likely experienced during that period transitional biogeochemical shifts, such as initial acidification from the decomposition of flooded organic matter. This may lower pH to acidic levels (e.g., below 7, as documented in similar systems with drops of nearly one unit), before reaching a steady state of increased alkalinity driven by photosynthetic activity; such a maturation period, potentially delaying equilibrium by years, is essential for stabilizing water quality dynamics [1,2,3].

This system has exhibited notable symptoms and endured multiple pressures, rendering it a compelling case study amid broader climate variability challenges in Mediterranean regions [4]. First, it has displayed major whiting events (milky turbidity from calcium carbonate precipitation), particularly evident in 2017 and 2018 [2] (Appendix A.1). These phenomena are attributed in the literature to intense phytoplankton photosynthesis depleting dissolved CO₂, causing a sharp rise in pH and calcite supersaturation, particularly in warm, ion-rich waters [5,6]. While such events can regulate eutrophication by sequestering nutrients, the resultant increase in turbidity presents substantial challenges for water treatment [7,8,9].

The reservoir system is situated in a forested area and was impacted by the 2023 forest fires, which burned approximately 17,000–18,000 hectares on Rhodes [10,11,12]. This disturbance is anticipated to affect reservoir water quality, as supported by relevant literature on wildfire impacts in static systems such as reservoirs and lakes: ash deposition may elevate alkalinity, favoring whiting, while post-fire runoff introduces organic matter and sediments that promote anoxia (oxygen-depleted conditions) and manganese (Mn) release, thereby complicating overall chemistry [13,14,15]. Wildfires degrade water quality through increased erosion, nutrient loading, and heavy metal contamination, impacting aquatic ecosystems and water supplies [16,17]. Most effects subside within five years, though some persist beyond a decade [18]. Recent research emphasizes water quality, climate change interactions, and advanced modeling, with a bibliometric review noting intensified focus on these areas [19]. In Mediterranean regions, integrated watershed-reservoir models and large datasets highlight the burned area ratio as a key driver of reservoir water quality degradation [14,20].

The reservoir meets the needs of Rhodes with significant level fluctuations influencing water volume and dynamics. Such variations are of particular interest, as they affect thermal stratification, mixing regimes, and biogeochemical cycles, potentially impacting water quality [21]. These pressures are amplified by regional climate trends, including reduced tourism demand during COVID-19, which altered water residence times and stratification patterns in summer 2021, and escalating agricultural and tourism demands amid climate change [22]. The system maintains an approximate annual water balance equilibrium, modulated by seasonal demands and operational management. Recently, following a shift in intake depth to greater levels (from the highest intake at +111.5 m to the middle at +104.5 m), elevated manganese concentrations were observed in May 2025. According to the literature, this is linked to anoxic conditions where microorganisms reduce solid Mn(IV) oxides in sediments to soluble Mn(II), exacerbated by thermal stratification and organic matter decomposition [13,23,24].

As evident from the above, the Gadouras Reservoir constitutes a highly complex system subjected to multiple pressures, which has received limited dedicated study in the literature, particularly regarding integrated post-disturbance effects. This work builds upon and extends prior monitoring efforts, such as satellite-based assessments, to encompass recent wildfire dynamics. Current monitoring of water quality parameters in the Gadouras Reservoir is performed by the water facilities operator at the inlet of the water treatment plant. However, these single-point assessments are insufficient for a comprehensive understanding of reservoir-wide dynamics, as they fail to capture the spatial heterogeneity of in-lake processes such as whiting initiation and propagation. To address this, the present study employs satellite remote sensing technology, which has previously contributed to estimating water quality in reservoirs and lakes. Typical examples include detecting phytoplankton blooms, turbidity, and sediment patterns using platforms like Landsat and Sentinel-2 [25,26]. Notably, satellite monitoring has been instrumental in understanding whiting effects [27] by applying spectral indices like AreaBGR to detect and quantify calcite precipitation events in systems such as Lake Geneva, enabling spatiotemporal analysis of event initiation, propagation, and environmental drivers under changing conditions [28].

As a deliberate methodological choice, we rely on radiometrically calibrated surface-reflectance products (Sentinel-2 Level-2A; Landsat 8/9 Level-2). We did not perform a site-specific bio-optical calibration against in situ measurements; therefore, satellite-derived indices (e.g., AreaBGR, MCI) are interpreted relatively (semi-quantitatively) for pattern and timing rather than as absolute concentrations. Nevertheless, the study’s objectives are ambitious, aiming to leverage archived satellite imagery alongside existing single-point measurements from the facility inlet—including daily laboratory analyses, meteorological data, and water levels—over an extensive period from 2017 to July 2025. Accordingly, this study prioritizes identifying trends and relative changes discernible from satellite data, rather than absolute values of the examined parameters.

This strategy seeks the following:

(a)

Elucidate the primary phenomena driving the outcomes that affect facility operations and water quality.

(b)

Develop phenomena estimations accessible for use by other end-users, such as water managers, for early forecasting of events like whiting or Mn spikes and proactive risk management.

(c)

Recommend actions to enable quantitative calibration of future satellite images based on study insights, including optimized multi-depth sampling plans at various intake levels (e.g., +99.0 m, +104.5 m, +111.5 m) to better capture stratified phenomena like anoxia and Mn mobilization.

2. Materials and Methods

2.1. Study Area

2.1.1. Gadouras Reservoir

The Gadouras Reservoir, located in southeastern Rhodes, Greece (Figure 1), is an artificial water body created by damming the Gadouras River [29]. Construction of the dam started in 2001, with reservoir filling initiated in 2005 and full capacity achieved during the winter of 2011 [30]. The overall project concluded in 2014, marking the beginning of operational testing [30]. Efficient operational management began in July 2017 under a third contract (July 2017–July 2022), with water supply to Rhodes city commencing on 11 August 2017 [30]. The dam measures 67 m in height and 585 m in length, yielding a reservoir with a total storage capacity of 67.5 × 10⁶ m³ and an effective usable volume of 63 × 10⁶ m³, draining a catchment area of 148 km² [29]. The Gadouras Reservoir is equipped with three distinct water intakes located at different elevations to support flexible abstraction based on operational needs: Intake No. 1 (lowest) at +99.0 m, Intake No. 2 (middle) at +104.5 m, and Intake No. 3 (highest) at +111.5 m. From 1 July 2017 to 28 August 2024, water was abstracted exclusively from the highest intake (No. 3). Subsequently, from 28 August 2024 to 26 May 2025, abstraction shifted to the middle intake (No. 2). From 26 May 2025 to the present (30 July 2025), water abstraction reverted to the highest intake (No. 3).

The Gadouras Reservoir is the principal potable water source for the island of Rhodes, fulfilling approximately 80% of its demand. The service area encompasses the city of Rhodes, adjacent coastal settlements, and the islands of Chalki, Symi, and Kastellorizo. The associated water treatment plant (WTP) has a current operational capacity of 60,000 m³/day (Phase A), with a planned expansion to 120,000 m³/day (Phase B) to meet a projected annual provision of 26.5 million m³ by 2039. The reservoir is a hardwater, dimictic system characterized by summer hypolimnetic anoxia [2]. Its calcium-rich inflows, originating from a limestone-dominated geology, predispose the system to biogeochemical phenomena, notably calcite precipitation (whiting) events. Operational management is supported by 24-h telemetry and water quality monitoring compliant with EU Directive 98/83/EC and Greek national legislation, under an ISO-certified framework (9001:2015, 14001:2015, 50001:2011). This research was conducted within the broader context of an 18-month collaborative project, initiated in September 2024 between the Thessaloniki Water Supply & Sewerage Company SA (EYATH) and the South Aegean Region [31,32]. The study employs satellite remote sensing to evaluate wildfire impacts on reservoir water quality and to inform sustainable management strategies.

2.1.2. The July 2023 Rhodes Wildfire

The 2023 Rhodes wildfires, ignited on 17 July amid temperatures exceeding 40 °C, scorched approximately 17,625 ha of Mediterranean vegetation within and proximate to the Gadouras Reservoir’s catchment [10,11]. Satellite analysis using PlanetScope imagery revealed varying burn severities: negligible to slight damage (1779 ha), moderate damage (7489 ha), high damage (6004 ha), and destruction (2353 ha) based on differential normalized difference vegetation index [12].

2.2. Study Design and Temporal Framework

This study adopts an exploratory approach to assess the water quality dynamics of the Gadouras Reservoir using satellite remote sensing complemented by available ground-truth data, following similar utility-oriented frameworks for Mediterranean reservoirs [33]. The analysis period spans January 2017 to July 2025, providing seven years of pre-fire baseline conditions and 2 years of post-fire observations. Given the absence of comprehensive concurrent in situ measurements, all satellite-derived parameters are treated as relative indices for spatial and temporal pattern analysis rather than absolute quantification, consistent with data-scarce monitoring strategies [33].

Table 1. Summary of multi-source datasets used for the analysis of the Gadouras Reservoir (January 2017–July 2025).

Data Category	Source/Platform	Parameters	Spatial Resolution	Temporal Frequency
Satellite Remote Sensing	Sentinel-2 MSI (Level-2A) [34]	Surface Reflectance (for MCI, AreaBGR indices)	10–60 m	~5 days
	Landsat 8/9 OLI/TIRS (Level-2) [35]	Surface Reflectance, Water Surface Temperature (WST)	30 m	~8–16 days
In Situ Water Quality	Water Treatment Plant (WTP) Inlet	Raw Water pH, Total Organic Carbon (TOC), Electrical Conductivity (EC), Calcium (Ca), Magnesium (Mg), Manganese (Mn)	Point Sample	Daily
Meteorological Data	Lindos Station (meteo.gr)	Air Temperature, Precipitation	Point Sample	Daily
Hydrological Data	Gadouras Dam SCADA System	Reservoir Water Level	Reservoir-wide	Daily

presents a summary of the datasets used in this study.

2.3. Satellite Remote Sensing—Data Processing and Analysis

We analyzed water-quality dynamics with Sentinel-2 MSI (Level-2A, surface reflectance; 10–60 m) and Landsat 8/9 OLI/TIRS imagery, processed end-to-end in Google Earth Engine (GEE). Scenes were limited to <35% scene-level cloudiness; clouds and low-quality pixels were further masked with the Sentinel-2 scene-classification (SCL) band to minimize artifacts. All geoprocessing, masking, index computation, and figure generation were scripted in GEE [36]. These Level 2 products are already radiometrically and atmospherically corrected; no local, site-specific bio-optical calibration was applied in this study, so indices are used as relative indicators.

Water surface temperature (WST) was derived from Landsat Collection-2 Level-2 Surface Temperature (ST_B10, 30 m). Digital numbers were converted to Kelvin using the USGS scale factor ST = DN × 0.00341802 + 149.0, and then to °C by subtracting 273.15. Reported per-pixel uncertainties for Collection-2 ST are typically on the order of ±1–1.5 °C, reflecting sensor calibration, atmospheric correction, and pixel mixing.

Chlorophyll-a proxy (MCI): Phytoplankton biomass was proxied by the Maximum Chlorophyll Index (MCI), a red-edge “line-height” metric originally developed for MERIS and adapted to Sentinel-2 [37,38]. For MSI, MCI was computed at ~708.75 nm using the B4–B6 baseline and the B5 measurement [37]:

$$MCI=R_{705}-[R_{665}+(R_{740}-R_{665}) \ast {\displaystyle \frac{708.75-665}{740-665}}]$$

where R₆₆₅, R₇₀₅, and R₇₄₀ correspond to Sentinel-2 bands B4, B5, and B6, respectively. Because no site-specific bio-optical calibration is available and bright-water events can affect red-edge reflectance, MCI is interpreted as a relative (semi-quantitative) indicator rather than an absolute Chl-a estimate. Pixels classified as whiting (below) were optionally masked during MCI mapping to reduce false positives. Whitings were identified using the AreaBGR index, defined as the triangular area in RGB spectral space [28]:

AreaBGR = 0.5 × (λ_blue × R_green + λ_green × R_red + λ_red × R_blue − λ_green × R_blue − λ_red × R_green − λ_blue × R_red),

where wavelengths correspond to 490, 560, and 665 nm.

Open water was identified per image using NDWI [39]; all indices were computed on water pixels only. For Sentinel-2, we flagged whitings at AreaBGR > 13,000 (sensitivity tests 8000–20,000). For mapping, we used either a per-image dynamic stretch (water-only percentiles) or a fixed 0–400,000 stretch on the Sentinel-2 scale to enable cross-date comparison [28]. Pixel values were summarised within inlet/central/outlet zones and along a fixed south–north transect of 50 evenly spaced points to produce space–time (Hovmöller-style) plots for AreaBGR, MCI, true color, and WST. All steps ran in GEE [36].

2.4. Ground-Truth Data, Environmental Context

Daily raw-water measurements at the Water Treatment Plant (WTP) inlet were provided by the competent authority. These in situ data refer exclusively to raw water sampled immediately upstream of the WTP intake; the treated (finished) water delivered to consumers complied with EU Directive 98/83/EC as transposed in Greek law (JMD Γ1(δ)/ΓΠ οικ. 67322/2017) and is safe for human consumption. Parameters included pH, total organic carbon (TOC), electrical conductivity (EC), calcium (Ca), magnesium (Mg), and manganese (Mn). Analytical methods adhered to Joint Ministerial Decision Γ1(δ)/ΓΠ οικ. 67322/2017 (transposing EU Council Directive 98/83/EC into Greek legislation). Validation of satellite-derived products relied on temporal correlation analysis, comparison of seasonal patterns, and event-detection cross-checks against WTP measurements. Owing to single-point sampling and spatial-resolution mismatch, this validation constrains conclusions primarily to timing at the intake and does not permit quantitative evaluation of spatial features (e.g., initiation zones or propagation). To provide environmental context, daily meteorological data (air temperature, precipitation, wind speed) from the Lindos station (meteo.gr) and reservoir-water-level data from the Gadouras Dam SCADA system were compiled. Both datasets underwent quality control and gap-filling prior to integration.

2.5. Statistical Analysis

Pairwise linear associations between all numerical variables were quantified using the Pearson correlation coefficient (r). Prior to analysis, all time-series data were programmatically merged, temporally aligned, and aggregated by their mean value to ensure a consistent daily frequency. To investigate temporal shifts, correlation analyses were performed both on the entire dataset and on subsets corresponding to the three identified ecological periods. To formally assess the impact of the July 2023 wildfire, a quasi-experimental Interrupted Time Series (ITS) analysis was conducted. The analysis utilized an Ordinary Least Squares (OLS) model to test for statistically significant changes in the baseline level (immediate impact, β₂) and trend (post-intervention trajectory, β₃). This model was specified using a long pre-fire period (January 2017–June 2023) to establish the baseline and was implemented in Python 3.12.3 using the statsmodels library.

2.6. Study Limitations

Limitations include the absence of a local bio-optical calibration (therefore, MCI is interpreted as a relative proxy), the spatial-scale mismatch between point validation data and satellite pixels, and the possibility of optical confounding by non-algal particulates or surface phenomena (mitigated by AreaBGR-based masking). Despite these constraints, the analysis serves its intended exploratory purpose—pattern recognition and timing—supporting the design of future monitoring campaigns and targeted in situ sampling [33].

3. Results

3.1. Spatiotemporal Distributions

Analysis of satellite imagery reveals two potentially distinct types of calcite whiting events, exemplified by episodes in late-spring 2018 and mid-winter 2020. These events differ fundamentally in their initiation, spatial evolution, and underlying drivers, as evidenced by the combined use of true-color imagery, the Area Under the Red–Green Curve, the Maximum Chlorophyll Index (MCI), and water surface temperature (WST). For example, drawdown, which preconditions the water for supersaturation, leads to precipitation that manifests as a high-AreaBGR whiting event [6,40]. The late-spring whiting event, detailed in Figure 2, exhibited a clear south-to-north progression. Whiting first appeared in the southern basin on 10 May 2018, reached its maximum spatial extent by 15 May 2018, and had largely subsided by 4 June 2018 (Figure 2, True Color and AreaBGR rows). A distinct pre-onset pattern was evident in the MCI data on 5 May 2018. Spatially heterogeneous, nucleus-like patches of elevated MCI were observed across much of the reservoir, indicating widespread phytoplankton activity prior to the whiting. Per-date stretched MCI maxima were approximately 0.007 on 5 May, rose towards the onset, peaked at approximately 0.038 on 10 May, and then decreased to approximately 0.014 by 4 June. Within the main plume at its peak, MCI values were suppressed, consistent with the known depression of the red-edge line-height under intense backscatter from suspended calcite [28,37] (The red-edge line-height is a spectral feature measuring the height of the near-infrared (NIR) reflectance peak relative to a baseline, and it serves as a proxy for phytoplankton biomass (quantified here by the Maximum Chlorophyll Index, MCI). Suspended calcite particles increase backscatter across all wavelengths, which elevates the spectral baseline and consequently reduces the relative height of the NIR peak. This masking effect results in suppressed MCI values within dense whiting plumes, irrespective of the actual phytoplankton concentration).

Concurrently, per-date stretched AreaBGR maxima increased from approximately 9.3 × 10⁴ (index units, Sentinel-2 scale) on 5 May to a peak of approximately (4.3–3.8) × 10⁵ on 10–15 May, before declining to approximately 1.2 × 10⁵ by 4 June. Landsat-derived WST indicated warmer surface water (>15 °C) in the southern basin prior to the onset, with temperatures favorable for CO₂ outgassing and subsequent CaCO₃ precipitation. Taken together, the sequence of (i) widespread, pre-onset MCI nuclei, (ii) southern initiation and northward propagation observed in AreaBGR, and (iii) warm WST is consistent with a biogenic pathway. This pathway involves phytoplankton growth and associated CO₂.

The mid-winter episode, shown in Figure 3, presented a contrasting scenario. The whiting was concentrated in the northern embayment with limited southward propagation. In stark contrast to the spring event, MCI values remained low and spatially uniform throughout the sequence, with per-date maxima ≤0.009. Critically, no pre-onset, nucleus-like patches comparable to those of May 2018 were evident. AreaBGR maxima were considerably lower than in late spring, on the order of (7–14) × 10⁴ index units (Sentinel-2 scale). Throughout the event, WST remained cold, below approximately 10 °C. This combination—cool water, the absence of antecedent MCI nuclei, a northern locus, and moderate AreaBGR intensity—indicates a possible non-biogenic, hydrologically forced mechanism [9]. This is plausibly driven by inflow-driven ionic loading and/or the resuspension of sediments from northern tributaries and shallow shelves, promoting calcite precipitation without a significant antecedent phytoplankton bloom [9,41].

The two events described above delineate distinct operational regimes for whiting formation in the reservoir:

• Biomass-Mediated Regime (Late Spring): Characterized by (i) pre-onset MCI “nuclei” indicating antecedent phytoplankton activity; (ii) initiation in the warm, stratified southern basin with subsequent northward expansion; and (iii) high-intensity whiting, with AreaBGR values peaking at approximately (3–4) × 10⁵ units.
• Hydrologically Driven Regime (Mid-Winter): Characterized by (i) the absence of MCI nuclei; (ii) initiation in the northern embayment, linked to tributary inflows; and (iii) lower-intensity whiting, with AreaBGR values on the order of (5–11) × 10⁴ units in cold (<10 °C) water.

Operationally, these findings demonstrate that while AreaBGR provides specificity for detecting whiting events, MCI supplies critical diagnostic context on antecedent biomass. The presence or absence of pre-onset MCI nuclei, in conjunction with WST and event location, serves as a powerful discriminator between internally driven biogenic whiting and externally forced hydrologic/resuspension events.

3.2. Longitudinal Analysis of Whiting Dynamics Using Hovmöller Diagrams

The case studies presented above highlight the distinct spatiotemporal dynamics of whiting events on specific dates. To place these individual episodes into a broader context and examine long-term patterns, we now analyze the data from a continuous temporal perspective. Figure 4 summarizes in Hovmöller diagrams four variables (true color, WST, MCI, and AreaBGR) along a south–north transect of 50 equidistant points (from dam to northern inflows as shown in the embedded map), from April 2017 to July 2025. This approach allows for the assessment of seasonality, inter-annual variability, and the persistence of the mechanistic regimes identified in the case studies described previously.

All panels use fixed color scales for cross-date comparability (WST: 5–35 °C; MCI: −0.002 to +0.040; AreaBGR: 2 × 10⁵ to 6 × 10⁵); values outside are clipped for display.

(a)

True color. The true-color Hovmöller provides a qualitative cross-check: Whitings appear as pale, milky-turquoise bands whose timing and position generally coincide with AreaBGR peaks. Cloud/haze streaks are identifiable and not interpreted as events.

(b)

Water surface temperature (WST). The annual thermal cycle is clear, with sustained > 15 °C summers and <~10 °C winters that frame the chemical and biological windows for precipitation.

(c)

Whiting intensity (AreaBGR). The AreaBGR panel isolates the calcite signal. In 2017–2019, events are frequent, intensive, and basin-spanning, often initiating in the southern/deeper basin and propagating northward. From early 2020 to mid-2023, events are weaker, winter-biased, and localized to northern segments with limited southward spread. After the July 2023 wildfire, whitings become more frequent, with higher baseline variability across the transect.

(d)

Chlorophyll-a proxy (MCI, relative). During 2017–2019, warm-season whitings show MCI increases that precede AreaBGR maxima by several days; winter episodes show low, flat MCI. After the July 2023 wildfire and especially during summer 2024, MCI also becomes more variable, consistent with episodic external inputs (ash-borne nutrients/particulates) superimposed on seasonal cycles.

Beyond the south first-summer pattern and north first-winter pattern, the diagrams reveal quasi-stationary hotspots: (i) The southern reach near the dam, repeatedly active in warm seasons, and (ii) northern embayments that activate in cool seasons and occasionally persist after peaks. These stationary zones are consistent with morphological and hydrological controls—depth/stratification and residence time in the south; tributary forcing and mixing in the north [42]. While Hovmöller diagrams provide a comprehensive overview, interpreting detailed patterns can be challenging. Therefore, time series are subsequently calculated to quantify key trends more directly.

3.3. Integrated Time-Series Overview and Periods Justification

Figure 5 compiles 7-day means from satellite transect products—whiting intensity (AreaBGR), chlorophyll-a proxy (MCI), and water surface temperature (WST)—together with WTP and hydrologic records (turbidity, pH, electrical conductivity, total organic carbon, Mg/Ca, Mn, rainfall, and reservoir level) for 2017–2025. Pink dashed lines denote summer whitings and light-blue bands denote winter whitings; the broad background shading marks the three analysis regimes. A vertical black dashed line highlights July 2023 (wildfire), and a narrow-shaded band later in the record marks the brief middle-intake interval. Compact heatmaps placed above panels (a)–(b) retain the spatial context of the transects and demonstrate that the time-series features correspond to coherent, basin-scale patterns.

During 2017–2019, AreaBGR (Figure 5a) exhibits recurring late-spring/early-summer peaks that rise in step with turbidity (right axis, log scale). From 2020 to 2022, peaks occur predominantly in winter and are smaller, and the tight co-variation with turbidity observed earlier becomes less consistent. From late-2022 through 2025, AreaBGR displays more frequent excursions; many align with hydrologic pulses and level changes. The strength of the AreaBGR–turbidity coupling decreases across the three regimes, r ≈ 0.86 (Intense), ≈0.42 (Moderate), and ≈0.23 in the post-wildfire span (Appendix A.2).

MCI (Figure 5b) shows a clear annual cycle with maxima in late spring and minima in mid winter. In 2017–2019, MCI commonly precedes AreaBGR by several days to a few weeks before the strongest summer whitings and then weakens at the precipitation peak, consistent with spectral dampening by suspended calcite. The small heatmaps above panel (b) reveal an additional, diagnostic feature in late-spring sequences: spatially patchy MCI anomalies spread across the reservoir before whitening onset (e.g., May 2018), which are less profound in winter sequences. In 2020–2022, MCI remains generally smooth during winter whitings, indicating a pathway not dependent on biomass buildup. After late-2022, MCI becomes more variable in timing and amplitude, with rises and dips tied to pulsed forcing and changing light conditions.

WST and pH are strongly seasonal (Figure 5c). The largest summer whitings occur when WST exceeds ~15 °C and pH is elevated. Short-lived increases in pH variability are repeatedly observed at, or just before, the onset of strong warm-season whitings, consistent with transient CO₂ disequilibria during nucleation/precipitation. In 2020–2022, both WST and pH behave seasonally but with fewer excursions in pH variability; after late-2022, brief pH variability spikes re-appear alongside more frequent optical events.

Reservoir level (Figure 5d) was low in 2017–2018 (drought), recovered in 2019, stayed high and relatively steady during 2020–2022, and then declined through 2023–2025. The unusually steady storage in 2020–2022 likely reflects hydrologic recharge from 2019 coupled with reduced withdrawals during the COVID-19 tourism downturn; similar pandemic-related reductions in water use have been documented in tourism-dependent regions [22]. Because local abstraction records are unavailable here, we treat this as plausible rather than proven. In 2023–2025, rainfall and level pulses align more frequently with AreaBGR and turbidity excursions, consistent with stronger external (catchment) control as storage falls and residence time shortens [42].

TOC (Figure 5e) shows episodic peaks early in the record, plausibly linked to the initial impoundment/first-filling phase, when inundated soils and vegetation leach and decompose, generating pulses of dissolved/total organic carbon and oxygen demand [43]. Afterwards, TOC remains low and stable through 2020–2022, and then shows a step-increase in mean and variance after the wildfires of July 2023. The interrupted time-series analysis (Appendix A.2, Figure A2) indicates no significant immediate step at the wildfire (Δ_level ≈ 0.017, p = 0.824) but a highly significant increase in slope thereafter (Δ_slope ≈ +0.0112 d⁻¹, p < 0.001; model R² ≈ 0.53). In other words, TOC begins to rise persistently after July 2023 relative to the counterfactual, and the variance widens, consistent with sustained post-fire increases in terrestrially derived DOM/TOC delivery modulated primarily by hydrologic pulses rather than a single shock. This interpretation aligns with reports that (i) first-filling disturbs carbon cycling via leaching and decomposition of newly flooded biomass/soils, and (ii) post-fire carbon exports are governed by flow and seasonal wetness [44,45].

Figure 5f plots EC (left axis) and pH (right). EC is lowest during the 2020–2021 high-storage interval and then drifts upward as the reservoir is drawn down after 2022, consistent with evaporative concentration and longer residence times. Within individual whiting episodes, EC shows a reproducible V-shape [9]: it falls during onset, bottoms near the whiting peak, and then rebounds, while pH rises (Figure 5a,c,f). The initial decline reflects removal of Ca²⁺ and bicarbonate into newly formed calcium-carbonate particles and—especially in winter—dilution by low-salinity inflows [46]; the rebound follows cooling/mixing and partial redissolution that returns ions to solution [46]. Winter whitings display this V-shape with low MCI and cold WST, indicating a predominantly abiotic inflow/resuspension-driven process [9], whereas early-summer whitings show pre-onset MCI ‘nuclei’ and stronger pH rises under warm surface waters, consistent with photosynthesis-driven, biologically nucleated precipitation [47].

Figure 5g overlays Mg/Ca (left axis) and Ca (right). During strong whitings (e.g., 2018; again late-2023/2024), the two series move in anti-phase: Ca exhibits sharp troughs, while Mg/Ca spikes. This pattern is expected when calcite forms—Ca and bicarbonate are preferentially removed into CaCO₃ particles, whereas Mg remains comparatively conservative—so the instantaneous Ca decrease drives the Mg/Ca rise; after the event, mixing/cooling and partial redissolution return Ca to solution, and Mg/Ca relaxes. In the high-storage interval of 2020–2021, the ratio is lower and more periodic; winter whitings produce smaller Mg/Ca excursions because low-salinity inflows dilute Ca and Mg together, and biogenic forcing is weak. From late 2022 onward, drawdown and longer residence times increase the amplitude and intermittency of Ca troughs and Mg/Ca peaks, consistent with more active carbonate cycling [8,37,46,47].

Manganese remained near baseline through the middle-intake period; the sole excursion (23–26 May 2025) coincided with the late-May whiting and a synchronous Ca minimum (Figure 5g), plus an EC drop and a pH rise (Figure 5f–h). The brevity and multi-proxy coherence argue against a persistent intake-depth artefact and are more consistent with a short, externally forced redox/inflow pulse that mixed Mn(II)-rich water into an oxygenated, high-pH epilimnion. Rapid oxidation and/or scavenging onto fresh carbonate surfaces likely curtailed the signal within days [18,48,49].

A final reading of Figure 5 makes clear that the time series in Figure 5 show that the examined period can be distinguished into three zones:

• Intense (2017–2019): Frequent, reservoir-wide warm-season whitings with WST > ~15 °C; MCI regularly leads AreaBGR by days–weeks; short-term pH variability spikes at onset; Mg/Ca increases; and AreaBGR tracks turbidity closely.
• Moderate (2020–2022): Higher, steadier water levels; winter-biased whitings with low MCI and cool WST; weaker optical–turbidity coupling and fewer onset signatures in pH variability.
• Heightened (2023–July 2025): Increasing external forcing as levels fall; more frequent optical and geochemical excursions (AreaBGR, turbidity, EC, TOC), with the wildfire introducing a statistically significant upward trend in TOC and greater variability thereafter.

3.4. Daily Average Aggregations and Parameter Interplay

The preceding analyses (Section 3.2 and Section 3.3) identified distinct biogeochemical regimes and multiple phenomena but lack the temporal resolution to resolve the mechanisms driving them. To address this, the following subsections present a detailed analysis of parameter interactions within the discretized spatial regions defined in Section 3.3. The objective is to investigate the coupling of biological productivity, carbonate chemistry, and hydrological drivers, which is not possible at coarser scales. This approach allows for: (1) quantifying the persistence of seasonal patterns relative to disturbance-induced shifts, (2) isolating the effects of wildfires, and (3) identifying thresholds for calcite precipitation and redox transitions. Figure 6, Figure 7 and Figure 8 summarize spatially resolved daily products (left column: AreaBGR, MCI, WST/pH, hydrology) together with laboratory records (right column: TOC, EC/pH, Mg/Ca, Mn) for three operating regimes respectively: High-Intensity (2017–2019), Moderate (2020–2022), and Heightened (2023-July 2025). This layout enables side-by-side comparison of optical, biological, chemical, and hydrologic signals at event scale, while preserving the north–south context visible in the heat-map strips (south-first initiation in late spring; north-first dominance in winter). Turbidity and Mn are on log axes, so small visual steps can represent order-of-magnitude changes. A detailed summary of the parameter interplay for each period is provided in Appendix B.1.

3.4.1. AreaBGR and Turbidity (Panel a)

AreaBGR—an optical proxy for suspended calcite during whitings—tracks in situ turbidity most closely during the High-Intensity period, when late-spring/early-summer whitings are frequent [9,27,28]. The strength of the AreaBGR–turbidity correlation then weakens: Intense r ≈ 0.86 (p < 0.0001, N = 95), Moderate r ≈ 0.42 (p < 0.0001, N = 121), Heightened r ≈ 0.23 (p = 0.0172, N = 104) (Appendix A.2). In the Moderate period, AreaBGR peaks shift toward winter and become more localized to the north, while turbidity baselines are lower. In the Heightened period, excursions become more frequent and variable, with a pronounced warm-season event in May 2025 superimposed on diffuse, basin-scale winter brightening. Collectively, these patterns indicate that (i) the composition of suspended matter varies through time and (ii) external physical forcing (dilution, mixing, resuspension) increasingly modulates the calcite signal [9].

3.4.2. Chlorophyll-a (Chl-a) Index Values (Panel b)

MCI behaves as a relative biomass indicator rather than an absolute Chl-a estimate because incipient calcite can elevate red/green reflectance and subtly inflate the index [45]. In the High-Intensity period, MCI typically leads AreaBGR by days to weeks before the strongest summer whitings, and the transect heat-maps show patchy pre-onset variations (“nuclei”) across the reservoir—textures that are not seen during winter whitings and could be operationally useful as early-warning cues. In the Moderate period, an annual spring rise persists but decouples from winter AreaBGR peaks, supporting an abiotic/inflow mechanism for cool-season whitings [9,46,49]. In the Heightened period, MCI becomes more unpredictable; a mid-winter 2025 increase precedes—but does not immediately coincide with—the May 2025 AreaBGR maximum. A notable cool-season exception occurs on 11 November 2021, when MCI rises modestly during a winter whiting, plausibly reflecting short-lived biomass stimulation by precipitation/mixing.

3.4.3. Water Temperature and pH (Panel c)

Temperature and pH jointly regulate carbonate solubility and CaCO₃ precipitation [50]. Across all periods, large whitings occur when WST > ~15 °C and pH is elevated. Strong events (e.g., May 2018, May 2025) show a brief pH dip at onset followed by rebound, consistent with transient CO₂ disequilibria and alkalinity regeneration during precipitation [5,40,46]. pH variability (7-day s.d.) is modest and seasonally patterned in 2020–2022, then becomes larger and noisier after late-2022, indicating reduced buffering and more variable metabolism/mixing. A weak warming trend and higher winter minima are evident over the record (cf. Figure 5c), in line with regional expectations [4], while pH drifts downwards toward the 8.0–8.4 range by 2025, potentially aided by wildfire-derived acidic inputs [10,11,15,48].

3.4.4. Hydrological Metrics (Panel d)

Rainfall pulses and managed outflows control level and, with them, dilution/evapoconcentration. Low levels in 2017–2018 transition to high, steady levels in 2020–2022, conditions under which whitings are predominantly winter-biased and less intense. After late-2022, levels decline and hydrologic pulses co-occur more often with optical/chemical excursions. The post-wildfire landscape, together with renewed precipitation in early 2025, sets the stage for the strong May 2025 whiting and associated geochemical responses.

3.4.5. Total Organic Carbon (TOC) and Turbidity (Panel e)

TOC shows regime-specific behavior. During 2017–2019, episodic peaks (~3–6 mg L⁻¹) accompany turbidity excursions, plausibly linked to early impoundment effects [43]. TOC is comparatively low and stable in 2020–2022, then rises and becomes more variable after July 2023. Interrupted time-series results indicate no significant instantaneous step at the wildfire but a significant post-fire slope increase (higher trend thereafter), consistent with hydrology-mediated, sustained delivery of terrestrially derived DOC/TOC following fire [44,45].

3.4.6. pH and Electrical Conductivity (Panel f)

EC and pH exhibit seasonality and inverse co-fluctuations tied to dilution and carbonate equilibria: EC tends to rise during evapoconcentration while pH softens, and vice versa during dilution [46]. EC variability is smallest in 2020–2022 (high, steady levels), then becomes more volatile after late-2022 as levels declines and external loading increases. Post-fire deviations suggest additional ionic contributions from fire-derived solutes [14,50,51].

3.4.7. Calcium (Ca) and Magnesium (Mg) Ratio (Panel g) and Manganese (Mn) Concentrations (Panel h)

The Mg/Ca ratio (panel g) rises when Ca is preferentially removed into calcium-carbonate particles while Mg remains comparatively conservative, so it fingerprints whiting episodes [5,51,52]. In Figure 5g, the simultaneous Ca series (right axis) shows the expected anti-phase behaviour: event-scale Ca troughs line up with Mg/Ca peaks during strong whitings [40,46]. The High-Intensity period exhibits the largest excursions (Mg/Ca peak ≈ 5 in mid-2018), consistent with pronounced Ca uptake and abundant particle growth. Under the high, steady water levels of the Moderate period, Mg/Ca settles around ~1.0–2.2 and Ca minima are smaller, in line with winter-dominated, more abiotic events where low-salinity inflows dilute Ca and Mg together and photosynthetic forcing is weak [9,42,49]. From late 2022 (Heightened period), drawdown and longer residence times increase the seasonal amplitude of Ca and the frequency of Ca troughs; Mg/Ca oscillates ~1.3–2.6, and its anti-phase with Ca becomes more evident, indicating stronger carbonate cycling and episodic resuspension/mixing [8,37,47].

Mn (panel h) remains low for most of the record (<0.05 mg L⁻¹ on a log scale), then shows a short-lived spike on 23–26 May 2025 that coincides with a pH rise and an EC minimum (Figure 8f) and a local Mg/Ca minimum (Figure 8g). Intake operations are a necessary context, as abstraction moved from the high (epilimnetic) to the middle intake on 28 Aug 2024 and switched back to the high intake on 26 May 2025. Mechanistically, the co-occurrence of an EC minimum and a Mg/Ca trough with rising pH indicates carbonate precipitation and/or dilution by low-ionic-strength inflow: Removal of Ca–HCO₃⁻ into CaCO₃ particles and/or mixing with dilute water drives EC down, while the elevated pH accelerates Mn(II) oxidation and promotes adsorption/co-precipitation onto freshly formed carbonate and biogenic Mn-oxide surfaces—processes that can produce a brief dissolved-Mn peak if reduced Mn is introduced, followed by rapid decline as it is oxidized [18,41,43,45,47]. Two contributors therefore overlap: (i) A biogeochemical driver—post-wildfire organic loading plus spring stratification likely strengthened hypolimnetic anoxia, enhancing reductive dissolution of Mn oxides and benthic Mn release [13,14,17,18,23,24]; and (ii) an operational exposure—drawing from the deeper intake increased the chance of sampling Mn-rich, reducing water [8,37,44]. The episode is best read as a real but depth-confounded excursion: a short redox/inflow pulse whose observed magnitude was amplified by preferential sampling of a Mn-enriched layer. The rapid fall after the switch back to the high intake implies strong vertical structure and fast near-surface removal under high-pH conditions [8,42,46,52]. With the current data, the fractional contributions of (i) and (ii) cannot be resolved, so reservoir-wide surface concentrations at the peak remain uncertain, and any operational inference should carry this caveat.

4. Discussion

By combining satellite products (AreaBGR, MCI, WST) with in situ and laboratory measurements (turbidity, pH, TOC, EC, Ca, Mg, Mn), three behaviour regimes emerge in the Gadouras system that relate to reservoir maturation, hydrologic variability, and the July 2023 wildfire.

4.1. Pre-Wildfire Dynamics and Reservoir Maturation

From 2017 to mid-2023, the reservoir’s evolution follows the expected maturation of a new impoundment: an early surge and subsequent stabilization [1,38,47]. The Intense period (2017–2019) was highly active, with frequent, basin-wide late-spring/early-summer whitings that initiated in the southern basin and propagated north (Figure 2 and Figure 4d), turbidity spikes of ~10 NTU, and TOC fluctuations of ~3–6 mg L⁻¹. These events aligned with biological indicators—antecedent MCI elevations, WST >15 °C, and intake pH > 8.5. Laboratory data show transient Ca depletions of ~10–20 mg L⁻¹ with relatively stable Mg (~10–15 mg L⁻¹), consistent with selective Ca incorporation into calcite during biogenic precipitation [5,42,46,50]. Together, these observations support a biogenically driven calcite pathway sustained by the elevated productivity typical of a reservoir’s early life [5,6].

The Moderate period (2020–mid-2023) reflects maturation into a steadier state [42]. Whitings became fewer, winter-biased, and localized near the northern inflows at low temperatures (<10 °C), decoupled from biomass build-up (MCI < 0.05 relative units). Background conditions stabilized (EC ~520–620 µS cm⁻¹; Ca ~20–30 mg L⁻¹), and Mn remained low (<0.05 mg L⁻¹). The AreaBGR–turbidity relationship weakened from r = 0.859 to r = 0.417, indicating that, as the system stabilized, winter turbidity reflected a more complex particle mix beyond calcite alone—consistent with hydrologic/resuspension influences in cool-season events [8,9,41,42]. These pre-wildfire phases establish a useful operational baseline with a seasonal duality—biogenic in summer vs. abiotic in winter—modulated by stratification and nutrient cycling [4].

4.2. Wildfire-Induced Shifts in Whiting and Biogeochemical Processes

Two seasonal mechanisms drive whitings (Figure 2, Figure 3 and Figure 4). Summer events follow a biogenic pathway: southern initiation and reservoir-wide expansion, antecedent MCI elevations (≈0.02–0.04 in our per-date stretch), WST > 15 °C, intake pH > 8.5, and selective Ca removal with comparatively steady Mg [6,42,46,50]. Brief pH dips followed by recovery during major events are consistent with transient CO₂ disequilibria and alkalinity restoration via precipitation [5,46,50]. Winter events appear abiotic: localized near northern inflows, occurring at <10 °C with flat/low MCI, plausibly linked to inflow-driven ionic loading and/or resuspension-assisted precipitation largely independent of biology [8,9,41,42]. This seasonal predisposition is expected given the limestone-dominated geology and dimictic stratification [2,42].

The July 2023 wildfire altered these dynamics in three ways (Figure 4 and Figure 5; event documented in [10,11]). First, the AreaBGR Hovmöller shows a persistent, low-amplitude elevation relative to the pre-fire baseline, indicating continuous low-level calcite suspension. Second, the AreaBGR–turbidity correlation declined to r = 0.233. Third, interrupted-time-series analysis shows a significant post-July 2023 increase in the TOC trend (slope change; values clustering ~5–6 mg L⁻¹), not an immediate step. We interpret the decoupling as diagnostic: AreaBGR continues to isolate the spectrally specific calcite signal, while turbidity increasingly reflects a mixed particulate assemblage (calcite plus pyrogenic and eroded material) with different optical properties [27,28,50]. Post-fire ash and mineral soils likely (i) supplied alkalinity and nucleation surfaces sustaining low-level calcite, and (ii) introduced non-calcite particulates that weakened turbidity’s specific association with calcite—consistent with the literature on hydrology-mediated post-fire inputs and carbon export [17,18,44,45,53].

4.3. Manganese Mobilisation Mechanisms

The unprecedented intake Mn spike (~0.4 mg L⁻¹, May 2025) is best explained as a biogeochemical predisposition plus operational exposure. Post-fire organic loading and prolonged stratification favor hypolimnetic oxygen depletion and reductive dissolution of Mn oxides, enhancing benthic Mn release [13,14,17,18,23,24]. Simultaneously, the temporary use of the middle intake during late 2024–May 2025 increased the likelihood of intersecting Mn-rich, reducing water during strong spring stratification [42,49]. The spike co-occurs with an EC minimum, a local Ca (or Mg/Ca) trough, and a pH rise (Figure 5f,g and Figure 8f,g), consistent with carbonate removal/dilution; elevated pH then accelerates Mn(II) oxidation and promotes scavenging onto fresh carbonate and biogenic Mn-oxide surfaces, explaining the rapid decline when abstraction returned to the epilimnetic intake [23,46,48,50,52]. Because depth-resolved DO/Eh were not collected contemporaneously, this remains a well-supported inference that requires targeted profiling and pore-water analyses for confirmation (Appendix A).

4.4. Implications for Monitoring and Management

Two complementary uses of satellite–in situ integration are supported. Event anticipation: AreaBGR provides synoptic detection of whiting; in combination with MCI, WST, and pH it can anticipate biogenic events (e.g., WST > 15 °C, pH > 8.5, rising MCI) [28,37,52]. System diagnostics: A sustained weakening of the AreaBGR–turbidity relation indicates shifts in particulate composition and warrants closer scrutiny of catchment inputs and mixing processes [8,42]. Provisional risk cues (e.g., TOC > 5 mg L⁻¹, together with prolonged stratification as an Mn-risk signal) should be site-calibrated. Priorities include post-fire erosion control, evaluation of hypolimnetic oxygenation/aeration to mitigate anoxia risk and integrating real-time satellite feeds with depth-resolved intake and DO/Eh monitoring to support adaptive operations [16,33,36]. The biogeochemical evolution of the Gadouras Reservoir between 2017 and 2025 can be summarized by three distinct phases, as detailed in

Table 2. Synthesis of biogeochemical regimes, diagnostic thresholds, and management implications in the Gadouras Reservoir (2017–2025).

Parameter	Intense Period (2017–2019)	Moderate Period (2020–2023)	Heightened Period (2023–2025)
Dominant State	Trophic Surge/Maturation	Stabilized/Equilibrium	Post-Wildfire Disturbance
Primary Whiting Mechanism	Summer Biogenic: Southern origin, basin-wide propagation.	Winter Abiotic: Northern origin, localized near inflows.	Amplified Biogenic and Diffuse: Basin-wide, prolonged, plus low-level background whiting.
Key Environmental Drivers	Nutrient release from flooded soil; thermal stratification.	Hydrological stability; reduced water demand (COVID-19).	Wildfire-derived ash and TOC loading; sediment exposure.
BGR-Turbidity Correlation (r)	0.859 (Very Strong)	0.417 (Moderate)	0.233 (Weak)
Interpretation of Correlation	Turbidity is dominated by calcite from biogenic precipitation.	Turbidity reflects a mixed signature of calcite and other inorganic particulates.	Turbidity is a complex mixture dominated by non-calcite pyrogenic particulates (ash, soot, soil), breaking the correlation with the calcite-specific AreaBGR index.
Primary Management Concern	High turbidity from calcite precipitation impacting WTP.	Seasonal, predictable winter turbidity.	Mn toxicity risk; high/complex turbidity; sustained high TOC.

.

4.5. Methodological Framework and Future Directions

We separate direct observations, inferences, and targeted confirmations (Appendix A). This framework, which outlines the study’s conclusions and a roadmap for future validation, is detailed further in Appendix B.2. The main limitations are the semi-quantitative nature of satellite indices without local bio-optical calibration and the reliance on proxies for anoxia. Moving from plausible mechanisms to quantitative certainty requires the following: (i) calibration/validation—multi-depth sampling to calibrate MCI and AreaBGR to chlorophyll-a and suspended CaCO₃; and (ii) mechanistic confirmation—multi-depth profilers (DO, Eh, pH) and sediment core/pore-water chemistry to verify Mn pathways and evaluate wildfire legacy [13,18,19]. Coupled with adaptive intake management, this approach offers a scalable, cost-effective strategy for Gadouras and similar fire-prone Mediterranean reservoirs [33,36,42].

5. Conclusions

This study successfully delineated the controlling mechanisms of water quality in the Gadouras Reservoir from 2017 to 2025, confirming a dual-mechanism hypothesis for whiting events: biogenic, photosynthesis-driven events in summer versus abiotic, inflow-driven events in winter. More significantly, the analysis documents a profound, disturbance-driven regime shift catalyzed by the July 2023 catchment wildfire. The fire forced the reservoir from a predictable, mature state into a “Heightened” biogeochemical regime characterized by altered particulate composition, sustained organic carbon loading, and an elevated risk of hypolimnetic anoxia.

A central finding is that the post-fire statistical decoupling of the satellite-derived AreaBGR whiting index from in situ turbidity (Pearson’s r dropping from 0.859 to 0.233) is not a methodological limitation but a key diagnostic discovery. It confirms a fundamental shift in the composition of suspended matter, as the turbidimeter responded to a new, complex mixture including optically distinct ash and pyrogenic carbon, while the AreaBGR index correctly continued to isolate the calcite signal. This validates the framework’s sensitivity to detect ecosystem-wide changes in water quality drivers.

The unprecedented manganese spike (>0.4 mg/L) in May 2025 resulted from the confluence of a biogeochemical predisposition and a critical operational decision. The wildfire-derived organic matter load created the anoxic conditions necessary for Mn mobilization from sediments, but the magnitude of the recorded spike was a direct consequence of switching to a deeper water intake that abstracted this contaminated, Mn-rich water. This sequence serves as a critical case study, revealing how a system’s biogeochemistry can shift post-disturbance to a point where established operational protocols become a liability. It highlights the imperative of augmenting static protocols with adaptive management tools, particularly real-time, depth-resolved monitoring, to provide operators with the crucial data needed to navigate a dynamically changing environment.

Ultimately, this work provides a transferable framework for monitoring and managing data-scarce Mediterranean reservoirs facing escalating fire and climate pressures. To move from the well-supported inferences presented here to quantitative certainty, future work must prioritize (1) the quantitative calibration of satellite products through targeted in situ sampling, (2) the deployment of high-frequency sensor profilers to directly measure depth-resolved dissolved oxygen and redox potential, and (3) sediment core analysis to conclusively link biogeochemical drivers to water quality outcomes.