High-frequency in-situ measurements of thermohaline, atmospheric and biogeochemical properties were obtained from two autonomous meteo-marine monitoring systems, positioned ∼ 3.7 km offshore of Fano (43°53^′30.0^′′ N, 13°00^′30.0^′′ E) and Senigallia (43°45^′20.9^′′ N, 13°12^′32.4^′′ E) on the western coast of the northern Adriatic Sea (product ref. 1 and 2; Table 1).

Raw data from these platforms (mentioned as Buoy Fortunae offshore of Fano and Meda Station offshore of Senigallia henceforth) were first converted with the calibration coefficients in the configuration file provided by Sea-Bird Electronics and processed according to the typical SBE data processing sequences. The necessary calculation and analysis of thermodynamic properties was based on the Gibbs Seawater (GSW) Oceanographic Toolbox (IOC et al., 2010), which contains the TEOS-10 subroutines (http://www.TEOS-10.org, last access: 12 February 2025) for the evaluation of seawater properties (McDougall and Barker, 2011). Detailed sensor specifications and processing steps are available in Appendix A1.1–A1.3.

Hourly precipitation data were retrieved from the ERA5 reanalysis dataset, produced by the Copernicus Climate Change Service, which provides gridded atmospheric fields at 0.25° spatial and 1 h temporal resolution (product ref. 3; Table 1). In parallel, 6-hourly river discharge estimates were extracted from the European Flood Awareness System (EFAS) model, based on the LISFLOOD hydrological model forced with gridded meteorological observations (product ref. 4; Table 1). Considering the dominant southeastward flow of the Western Adriatic Current, freshwater input from major (Po, Adige, Reno) and minor regional rivers (e.g., Lamone, Savio, Metauro, and 14 others) was assessed at their upstream mouths along the northwestern Adriatic coast (see Fig. 3 for river mouth locations and characteristic circulation patterns).

To estimate the required freshwater inflow necessary to account for the observed decrease in surface seawater salinity over a given time period, a simplified salt mass conservation approach was applied. This method was based on in-situ salinity data from the Fortunae buoy (product ref. 1; Table 1) and the estimated area of the affected region, derived from satellite SST and ocean-colour observations (product refs. 5 and 6; Table 1) capturing the cold and turbid coastal plume. Assuming that salinity and density are conserved during mixing and proportional to the volumes involved, the following salt mass balance equation was applied: 1ρ1S1V1=ρ2S2V2 where ρ1 and ρ2 are the initial and final density in kg m⁻³; S1 and S2 are the initial and final salinity in g kg⁻¹; V1 is the initial considered volume of seawater in m³; V2 is the final volume after mixing with the added freshwater in m³.

And the freshwater flow (Q) in m³ s⁻¹, required to achieve the observed salinity decrease over the time period Δt is calculated as: 2Q=ΔVΔt A practical example of this calculation, including measured values from the Fortunae buoy and assumptions on the affected volume based on the buoy's position and local bathymetry, is provided in Appendix A2, along with a schematic diagram illustrating the considered water volume (Fig. A1).

To complement the in-situ observations and support the freshwater rate estimation (Sect. 2.3), we incorporated satellite-derived and model-based datasets (products 5–8; Table 1) providing full spatial coverage of the northwestern Adriatic Sea. These included daily fields of sea surface temperature (SST; 0.01° × 0.01°), chlorophyll a concentration (Chl-a, 1 km × 1 km), light attenuation coefficient at 490 nm (KD490; 1 km × 1 km), surface salinity (SSS; 0.042° × 0.042°), and horizontal current components (u and v).

These datasets served three main purposes:

to delineate the surface extent of freshwater-affected waters and define the horizontal area used in freshwater input calculations (see Appendix A2);

Basic statistics were computed following Emery and Thomson (1998). Root Mean Square Difference (RMSD), bias, and cross-correlation were applied to evaluate temporal coherence between thermohaline, biogeochemical and environmental conditions. Detailed equations and assumptions are reported in Appendix A3.

Between May 2023 and November 2024, five distinct surface freshening events were identified in the northern Adriatic Sea from high-frequency salinity observations collected by the Fortunae buoy and the Senigallia Meda Station. These events varied in duration (12 to 53 d), intensity (6.7 to 20.1 g kg⁻¹ salinity drop), and temporal structure, shaped both by prolonged and abrupt freshwater input dynamics (Tables A4 and A5).

Event I (2 May–24 June 2023) was the longest, spanning 53 d. It was characterized by a multi-phase freshening, including an extended period of anomalously low salinity and a gradual recovery. The most pronounced salinity decrease occurred over 13 d (11–24 May), with salinity dropping by 12.77 g kg⁻¹, from 36.19 to 23.42 g kg⁻¹. This prolonged freshening suggests sustained riverine discharge into the northwestern Adriatic. Event II (29 September–16 October 2023, 17 d) showed a more abrupt signature. Over just 5 d, salinity fell by 6.72 g kg⁻¹ (from 37.17 to 30.45 g kg⁻¹), indicating a rapid freshwater input followed by a relatively short recovery. Event III (12 November–2 December 2023, 20 d) included the highest initial salinity observed (37.63 g kg⁻¹), followed by a significant 14 d drop to 26.53 g kg⁻¹, which corresponds to a total decrease of 11.10 g kg⁻¹. Event IV (9–21 April 2024, 12 d) was short but intense, featuring a 6 d salinity decrease of 13.89 g kg⁻¹, from 36.16 to 22.27 g kg⁻¹, representing one of the steepest gradients recorded. Event V (20 October–22 November 2024, 33 d) was the most extreme in terms of salinity reduction. A 10 d drop of 20.12 g kg⁻¹ brought surface salinity from 35.09 to 14.97 g kg⁻¹, making this the most intense freshening event of the record.

These events illustrate the basin's dual sensitivity to sustained meteorological forcing and intense short-term hydrological extremes. Cross-event comparisons emphasize differing recovery behaviors and highlight the coupling between catchment saturation and downstream marine responses (Figs. A2–A6).

Hydrological analysis revealed that major freshening events coincided with intense precipitation and river discharge anomalies in the Emilia-Romagna region, with events driven by both catchment saturation and localized convective storms.

A sequence of five rainfall episodes culminated in catastrophic flooding from 1–24 May 2023 (Fig. 2b). The first rainstorm (1–3 May) brought over 200 mm of precipitation to central-eastern Emilia-Romagna. Initial flooding affected the Reno and Lamone basins. Two subsequent moderate episodes (9–10 and 12–14 May) further saturated soils. The most severe storm, associated with Storm Minerva (16–18 May), delivered over 300 mm in areas such as Forlì, with cumulative precipitation more than 5 times the monthly climatological average.

During this period, the Po River remained well below the climatological average discharge, with values around 2000 m³ s⁻¹, as the most intense rainfall did not affect its main basin, whereas local rivers like the Lamone, Savio, and Reno exceeded historical flood levels (Fig. 3c). Cross-correlation analysis between accumulated precipitation and river discharge (Fig. A5) confirmed statistically significant relationships (p < 0.05), with time lags ranging from -6 to -120 h. Rapid response was observed in local rivers with short catchments, such as Rubicone (r = 0.74, lag = -6 h), while larger basins like the Adige exhibited longer delays (r = 0.71, lag = -36 h).

No widespread heavy precipitation was recorded in correspondence with Event II in late September 2023, but localized convective storms likely impacted the Marche region, just south of Emilia-Romagna, between 25–29 September. An increase in the discharge of the Adige and Po was reported and the freshening signal suggests a short-lived but intense localized input (Fig. 1). In accordance, two prolonged rainfall episodes occurred later during 2–5 November and 30 November–2 December (Event III). The latter was the most severe, with up to 383 mm of rainfall recorded inland, in the Parma Apennines. A third, minor event on 4–5 December followed the main storm.

During spring 2024, an unseasonable storm system between 20–24 April brought 70–80 mm of weak but steady rainfall and rivers with short response times reacted quickly to the precipitation input, leading to the abrupt salinity drop during Event IV (Fig. 1).

Autumn 2024 saw the return of exceptional rains. Between 17–20 September, Storm Boris delivered ∼ 300 mm of rainfall in eastern Emilia-Romagna, causing major river discharges in local rivers like the Savio, Fiumi Uniti and Lamone River (Fig. 3c). Less than a month later (18–19 October), another storm triggered river discharge records for all major rivers, especially for the Po River (record value = 7880 m³ s⁻¹), reflecting a marked hydrological recovery after prolonged drought.

This analysis affirms that both large-scale catchment saturation and localized storm activity contributed significantly to the observed salinity anomalies. Rivers with differing catchment sizes showed distinct response times, with immediate peaks in small rivers and delayed responses in larger ones. Figures 2 and 3, along with Appendix Figs. A2–A6, provide a spatial and temporal context for precipitation and discharge patterns associated with each event.

Each freshening event identified in this study (Fig. 1) triggered a marked biogeochemical response in the northern Adriatic Sea, as captured by the high-frequency in-situ measurements of chlorophyll a, turbidity, and dissolved oxygen collected by the Fortunae Buoy and the Senigallia Meda Station (Fig. 4). These parameters serve as proxies for primary productivity, sediment transport, and oxygen dynamics, offering direct evidence of ecosystem perturbations following extreme hydro-meteorological events.

Following Event I (May 2023), which was preceded by a sequence of five intense rainfall episodes in the Emilia-Romagna region and culminated in catastrophic flooding (Fig. 2b–e), an extreme turbidity increase (∼ 70 NTU average) was recorded for more than a month, along with chlorophyll a concentrations exceeding 10 µg L⁻¹ (Fig. 4). These responses reflect the combined influence of excess sediment transport and nutrient enrichment, especially from local rivers like the Lamone, Reno, and Savio, which exceeded historical discharge levels (Fig. 3c). The sustained rainfall and resulting prolonged runoff seem to have led to a persistent stratified surface layer that confined biological activity and promoted bloom formation.

After the record May flood period, a particularly notable anomaly was observed in June–July 2023, with sustained extreme turbidity (> 200 NTU) and following chlorophyll a values (> 30 µg L⁻¹). Subsequent visual inspections of the buoy infrastructure revealed the presence of mucilage-forming macroaggregates, consistent with recurring gelatinous algal blooms historically observed in the Adriatic under warm, stratified, and nutrient-enriched conditions (Arrighi and Domeneghetti, 2024). These mucilage events have well-documented impacts on marine ecosystems, coastal water quality, and economic sectors such as aquaculture and tourism.

The response to Event II (late September–October 2023) presented a contrasting case. While no widespread heavy precipitation and flooding occurred in Emilia-Romagna, localized convective storms over the Marche region and rising discharges in the Adige and Po rivers triggered a short-lived but intense freshening. While turbidity remained low, the biogeochemical response included a marked increase in chlorophyll a concentration and oxygen supersaturation reaching approximately 150 %, indicating a nutrient rich but low sediment freshwater input. Moreover, dissolved oxygen rose near-synchronously with chlorophyll a concentrations (r = 0.77, lag = 0 h), consistent with well-lit bloom conditions. This event exemplifies how smaller, convective systems can fuel biological productivity without large sediment loads, while favorable light conditions likely allowed rapid phytoplankton growth.

The prolonged rainfall in early and late November 2023 preceded Event III, which produced another strong bloom and variable oxygen patterns. This response suggests that delayed runoff from larger, saturated catchments delivered substantial nutrient loads, with peak phytoplankton activity followed by oxygen variability related to organic matter remineralization and light attenuation during turbid phases.

The corresponding biogeochemical response to event IV followed an unseasonable storm system in April 2024. Rapid response from short catchment rivers like the Rubicone and Uso led to an abrupt salinity drop (Fig. 1) and sharp turbidity and chlorophyll peaks, consistent with the fast land-sea transfer typical of springtime storms over saturated terrain. These dynamics emphasize the system's sensitivity even to moderate precipitation when antecedent soil conditions favor immediate runoff.

Finally, Event V in autumn 2024 was preceded by two major storm systems: Storm Boris (17–20 September) and an intense October storm (18–19 October), the latter driving the Po River to record discharge levels (Fig. 3c). This sequence produced the most intense surface freshening and the highest chlorophyll a concentration observed during the study period, exceeding 60 µg L⁻¹. Interestingly, these extremely high chlorophyll a levels occurred in conjunction with notably low turbidity, suggesting that sediment-laden runoff from the Po River may have settled or dispersed before reaching the offshore monitoring sites at Fano and Senigallia. This decoupling implies that nutrient transport remained effective despite reduced suspended particulate matter, allowing for favorable light conditions and enhanced primary productivity which is consistent with a multi-day negative turbidity–chlorophyll relationship (r = -0.32 at -144 h lag) and near-synchronous, positive chlorophyll–oxygen coupling (r = 0.86 at 0 h lag), together indicating strong photosynthetic control under clear conditions.

Across all events, dissolved oxygen dynamics reflected a combination of photosynthetic production during bloom conditions and oxygen depletion during periods of high turbidity and organic matter degradation. Event-scale lead–lag patterns showed a biphasic turbidity–chlorophyll response with a short-lag negative branch when high turbidity limits light, and a positive branch at multi-day delays when nutrient delivery and stratification promote blooms. The chlorophyll–oxygen coupling is typically positive and near-synchronous, with longer-lag negatives appearing during potential bloom decay and post-bloom respiration.

These findings confirm that hydro-meteorological forcing, whether through widespread rainfall, localized storms, or antecedent soil saturation, directly drive ecosystem-level responses in the northern Adriatic Sea. The magnitude and timing of these responses depend not only on precipitation intensity and discharge volume, but also on the sediment load, nutrient delivery, and seasonal conditions governing stratification and light availability.