Simulated terrestrial runoff shifts the metabolic balance of a coastal Mediterranean plankton community toward heterotrophy

. Climate change is projected to increase the frequency and intensity of extreme rainfall events in the Mediterranean 15 region, increasing runoffs of terrestrial matter into coastal waters. To evaluate the consequences of terrestrial runoff on plankton key processes, an in situ mesocosm experiment was conducted for 18 days in the spring of 2021 in the coastal Mediterranean Thau Lagoon. Terrestrial runoff was simulated in replicate mesocosms by adding soil from an adjacent oak forest that had matured in water from the main river tributary of the lagoon. Automated high-frequency monitoring of dissolved oxygen, chlorophyll-a fluorescence, salinity, light, and temperature was combined with manual sampling of organic and 20 inorganic nutrient pools, pH, carbonate chemistry and maximum quantum yield (Fv:Fm) of photosystem II (PSII). High-frequency data were used to estimate gross oxygen primary production (GPP), respiration (R)


Introduction
Climate change is predicted to increase the frequency and intensity of short extreme rainfall events in the Mediterranean region (Alpert et al., 2002;Sanchez et al., 2004).Consequently, the runoff of terrestrial matter will become more frequent in coastal Mediterranean waters.These runoffs constitute a pulse input of organic and inorganic nutrients into the water column and decrease light penetration (Nunes et al., 2009), substantially impacting marine ecosystems and, notably, plankton communities (Deininger and Frigstad, 2019;Striebel et al., 2023).
T. Soulié et al.: Simulated terrestrial runoff shifts the metabolic balance Plankton is crucial for aquatic ecosystems because it forms the basis of the aquatic food web and plays an important role in multiple biogeochemical cycles, notably that of oxygen (Falkowski et al., 2003;Falkowski, 2012).Indeed, phytoplankton produces oxygen through its gross primary production (GPP), and all planktonic organisms consume it through aerobic community respiration (CR).Hence, assessing GPP and CR provides a community metabolism index (GPP : CR) and determines the capacity of an aquatic ecosystem to serve as a net producer or consumer of oxygen and as a sink or source of atmospheric carbon dioxide (Lopez-Urrutia et al., 2006).This community metabolism index considerably depends on the fate of phytoplankton, which is itself related to phytoplankton growth (µ) and loss (L) rates.Therefore, assessing µ and L provides a trophic index (µ : L) related to the performance of both phytoplankton and its predators (Soulié et al., 2022a).
The consequences of terrestrial runoffs for plankton communities and associated processes remain unclear.The inputs of terrestrial carbon and nutrients have been shown to promote phytoplankton and bacteria in Mediterranean coastal waters (Pecqueur et al., 2011;Liess et al., 2016), possibly leading to higher GPP and CR.However, this positive effect of nutrient enrichment can be mitigated by light attenuation resulting from the runoff, which can depress phytoplankton photosynthesis and therefore GPP, as observed in the North Sea, Baltic Sea, and a North Atlantic bay (Mustaffa et al., 2020;Paczkowska et al., 2020;Soulié et al., 2022b).The contradictory effects of light attenuation and nutrient enrichment induced by terrestrial runoffs on plankton metabolism can change the structure of planktonic communities and, ultimately, their related processes.They can favour bacteria over phytoplankton (Meunier et al., 2017;Andersson et al., 2018;Courboulès et al., 2023) and large phytoplankton species at the expense of smaller cells (Deininger et al., 2016;Mustaffa et al., 2020) and decrease the abundance of protozooplankton (Courboulès et al., 2023).Consequently, these shifts can alter plankton processes because the structure and functions of aquatic communities are closely related (Giller et al., 2004).
Although the consequences of terrestrial runoffs have been well studied in freshwater systems, an important knowledge gap exists regarding the impacts of terrestrial runoffs on coastal marine ecosystems (Blanchet et al., 2022).In this regard, evaluating the consequences of terrestrial runoffs on plankton communities and processes in ecologically and economically important areas such as coastal lagoons (Soria et al., 2022), enclosed systems that are often subject to inputs from the land, is of fundamental concern.In the present study, we conducted an in situ mesocosm experiment in the Mediterranean coastal Thau Lagoon, a shallow productive lagoon which hosts oyster farms and serves as a nursery for several wild fish species (La Jeunesse et al., 2015).Moreover, it is naturally subjected to storm-induced terrestrial runoffs (Pecqueur et al., 2011;Fouilland et al., 2012), notably in fall during the Cévenol events, a meteorologi-cal phenomenon characterized by storms and heavy rainfalls that usually cause flash flooding on the Mediterranean coast (Ducrocq et al., 2008).Six mesocosms were used, with half serving as control mesocosms and with a terrestrial runoff simulated in the other half by adding soil from an adjacent typical Mediterranean oak forest that maturated over 2 weeks in water from the Vène River, the main river tributary of the Thau Lagoon (Plus et al., 2006).The responses of all plankton food web compartments in the present experiment have been detailed by Courboulès et al. (2023).
In the present study, high-frequency data from automated sensors immersed in the mesocosms were used to estimate GPP, CR, µ, and L in each mesocosm and assess how both the metabolic and the trophic indices of the community responded to the simulated runoff.Manual sampling was performed to assess dissolved and particulate materials and photosynthetic efficiency and carbonate system parameters.We hypothesized that (1) the metabolic index would be shifted by the runoff towards heterotrophy through light reduction and inputs of organic matter and that (2) the terrestrial runoff would affect the trophic index by creating imbalance between phytoplankton and its factors of loss.

In situ mesocosm experimental setup
An in situ mesocosm experiment was performed for 18 d in May 2021 in the Thau Lagoon (France) using the facilities of the MEDIterranean platform for Marine Ecosystems Experimental Research (MEDIMEER; 43°24 53 N, 3°41 16 E).The duration of the experiment was set to 18 d in order to monitor the responses of plankton in the medium term (multiple days to weeks), as interesting dynamics were already reported in control treatments during previous experiments in the Thau Lagoon up to almost 3 weeks after the start of the experiment (Courboulès et al., 2021;Soulié et al., 2022a).However, the duration of the experiment was limited by COVID-19 pandemic restrictions, preventing the authors from conducting a longer experiment.The Thau Lagoon is a shallow coastal lagoon of 75 km 2 with a mean depth of 4 m and is located on the French coast of the northwestern Mediterranean Sea (Derolez et al., 2020a).Six mesocosms were deployed in the lagoon.Each mesocosm consisted of a bag, sealed at the bottom, made of nylon-reinforced 200 µm thick vinyl acetate-polyethylene film which was 3 m high and 1.2 m wide, resulting in a total volume of 2200 L (Insinööritoimisto Haikonen Oy, Sipoo, Finland).Each mesocosm was equipped with a sediment trap at the bottom.A schematic representation of the mesocosm setup can be found in Soulié et al. (2021) and in the Supplement.Each mesocosm was covered with a dome of polyvinyl chloride to avoid external inputs.On 3 May (d0), all the mesocosms were filled simultaneously using a pump (SXM2/A SG; Flygt) with 2200 L of subsurface lagoon water preliminarily screened by filtering with a 1 mm mesh to remove large particles and organisms.The water was pooled in a large container before being distributed simultaneously by gravity to the six mesocosms through parallel pipes.In each mesocosm, the water column was continuously homogenized with a pump (Model 360; Rule) immersed at a depth of 1 m, resulting in a turnover rate of approximately 3.5 d −1 .Observations performed with a microscope indicated that organisms (phytoplankton and zooplankton), even fragile organisms such as ciliates, did not seem to be damaged by the mixing procedure.Three mesocosms served as controls, while in three others maturated soil was added on 4 May (d1) to simulate a terrestrial runoff event (these mesocosms are hereafter referred to as the terrestrial runoff treatment).Throughout the experiment, a total of 510 L was sampled from each mesocosm, representing 23 % of the initial volume of the mesocosms.For each treatment, one mesocosm displayed considerable differences in biological, physical, and chemical parameters compared to the two other replicates of the same treatment, most probably because of the malfunctioning of the mixing pumps, and it was therefore removed from the analysis.Data are therefore presented as the mean of the two replicates for each treatment ± the range of observations.Thus, any interpretation of the presented data must take into account the low number of replications and be done cautiously.

Soil extraction, preparation, and maturation
Soil was extracted 2 weeks before the beginning of the mesocosm experiment from the Puéchabon state forest, a fully preserved typical Mediterranean oak forest located approximately 30 km north of the Thau Lagoon (43°44 29 N, 3°35 45 E) (Allard et al., 2008).The soil was then roughly screened by sifting through a 1 cm mesh.On the same day as soil extraction was performed, water was collected from the Vène River, the main tributary of the Thau Lagoon, which is known for its episodic flash floods (Pecqueur et al., 2011).Water was screened by filtering with a 200 µm mesh to remove large particles and organisms.The soil and river water were then mixed to reach a concentration of 416 g soil L −1 , which represents natural flash flood events occurring in the lagoon (Fouilland et al., 2012).This mixture was then left to mature for 2 weeks in transparent Nalgene carboys placed in an outdoor pool continuously supplied with natural water from the Thau Lagoon.During the maturation step, each carboy was homogenized and aerated daily.This maturation was performed to mimic the degradation process of the most labile compounds that naturally occurs during their transportation from the soil to coastal waters during natural runoff events (Müller et al., 2018).After the manual mesocosm sampling on 4 May (d1), 7 L of the soil solution was added to each of the three "terrestrial runoff" mesocosms, representing a final concentration of 1.3 g soil L −1 .Further details regarding the choice and description of the soil addition protocol can be found in Courboulès et al. (2023).

Acquisition, calibration, and correction of the high-frequency sensor data
In each mesocosm, a set of high-frequency sensors was immersed to a depth of 1 m.Campbell Scientific, United States) installed at three different depths (0.5, 1, and 1.5 m).Each sensor recorded measurements every minute during the entire experiment.In the "Results" section, the high-frequency data are presented as daily averages.The fluorometers, oxygen optodes, conductivity sensors, and temperature probes were calibrated before and after the experiment.In addition, Chl a fluorescence and oxygen sensor data were corrected using discrete highperformance liquid chromatography (HPLC) Chl a and Winkler DO measurements, respectively.To do so, three borosilicate bottles (120 mL) were filled with water sampled from each mesocosm using a 5 L Niskin water sampler at a depth of 1 m every other day in the morning.The DO measurement was immediately fixed by adding Winkler reagents (Carrit and Carpenter, 1966).After at least 6 h of fixation during which bottles were kept underwater and in the dark in opaque plastic tanks filled with freshwater at room temperature, the DO concentration in each bottle was measured with an automated Winkler titrator (Metrohm 916 Ti-Touch) using a potentiometric titration method.Similarly, a polycarbonate bottle (2 L) was filled with water that was sampled every morning from each mesocosm using a Niskin water sampler at a depth of 1 m.Samples were then immediately filtered under low-light conditions using a vacuum pump on glass-fibre filters (Whatman GF/F; 0.7 µm pore size).Filters were then stored at −80 °C until analyses with HPLC (Shimadzu) were conducted following the method of Zapata et al. (2000).The HPLC was composed of a pump (DGU-405; Shimadzu), an automatic injector (LC-40D; Shimadzu), a Peltier oven (CTO-40S; Shimadzu), a PDA detector (SPD-M40; Shimadzu), and a fluorimeter detector (RF-20A; Shimadzu).The stationary phase used is a C8 column (SunFire C8; Waters) with a C8 column guard (Waters).Details of the calibration procedure can be found in the Supplement and in Soulié et al. (2023).Each mesocosm was sampled daily using a 5 L Niskin water sampler at a depth of 1 m to monitor dissolved inorganic nutrients (nitrate and nitrite, NO − 2 + and NO − 3 ; ammonium, NH + 4 ; and orthophosphate, PO 3− 4 ), dissolved organic carbon (DOC), and particulate organic carbon (POC) and nitrogen (PON) concentrations; they were sampled every second day to measure pH and total alkalinity (TA).For dissolved inorganic nutrient analyses, 50 mL subsamples of mesocosm water were placed in acid-washed polycarbonate bottles.Directly after, these samples were filtered over 0.45 µm filters (Gelman Sciences, United States) and stored in high-density polyethylene tubes at −20 °C until further analyses that were performed within 48 h.Nitrate, nitrite, and orthophosphate analyses were performed with an automated colourimeter (Skalar Analytical B.V., the Netherlands; Aminot and Kérouel, 2007), and ammonium analyses were performed using the fluorometric method (module 7200-067-W; Turner Design, United States; Aminot et al., 1997;Holmes et al., 1999).For DOC analyses, 30 mL subsamples of mesocosm water were filtered through two pre-combusted (4 h, 450 °C) glass-fibre filters (Whatman GF/F); 90 µL of phosphoric acid (85 % concentration) was then added, and subsamples were then stored at 4 °C in the dark pending analyses, which were performed by high-temperature catalytic oxidation (HTCO) on a total organic carbon analyser (TOC-LCSH; Shimadzu).For POC and PON analyses, subsamples (0.5-1 L) of mesocosm water were filtered over precombusted (4 h; 450 °C) glass-fibre filters (Whatman GF/F).Filters were then placed in a stove at 60 °C for at least 12 h.The POC and PON concentrations were then measured using a CHN analyser (UNICUBE; Elementar).The samples for pH and TA determinations were collected in 300 mL borosilicate glass bottles according to standard sampling methods for carbonate chemistry (Dickson et al., 2007).Samples for TA determination were filtered immediately on glass-fibre filters (Whatman GF/F; 0.45 µm pore size), spiked with 50 µL of HgCl 2 -saturated solution, and stored for later analysis.Samples for pH analysis were spiked with HgCl 2 and were analysed within 36 h.pH was measured spectrophotometrically (LAMBDA 365 UV/Vis; PerkinElmer) on a "total scale" at 25.0 °C (pH 25 ) with m-cresol purple (m-Cp) as an indicator (reproducibility ± 0.002) and thermostated cells (10 cm optical path; 28 mL), according to Clayton and Byrne (1993) and Dickson et al. (2007), with duplicate analysis for control and triplicate for treated mesocosms.TA was measured in the laboratories of the Institute of Marine Sciences of the National Research Council of Italy (CNR-ISMAR) in Trieste (duplicate analysis) using an open-cell potentiometric titration with a derivative determination of the end point, according to Hernández-Ayon et al. (1999) (reproducibility ± 0.1 µmol kg −1 ).Certified reference seawater for carbonate chemistry (provided by Andrew Dickson, Scripps, California) was used for TA analysis and to check the stability of the pH analysis during the experiment.After the experiment, the m-Cp used was checked against a purified m-Cp batch (Liu et al., 2011), showing a difference of < 0.005 pH units.The dissolved inorganic carbon (DIC) concentration, CO 2 partial pressure (pCO 2 ), and pH at in situ temperature (pH) were calculated using the CO2SYS program (Microsoft Excel version 2.5; Lewis and Wallace, 1998;Pierrot et al., 2006) and using the carbonate constants from Lueker et al. (2000), sulfate constants from Dickson (1990), and parameterization of borate from Lee et al. (2010).

Estimation of the daily light integral from the high-frequency PAR sensor data
PAR measurements were used to calculate the daily light integral (DLI).This value corresponds to the average quantity of light available for photosynthesis received by a 1 m 2 surface over a 24 h period (Soulié et al., 2022b).DLI was calculated using Eq. ( 1) as follows: where DLI is expressed in mol m −2 d −1 , mean PAR between sunrise and sunset in µmol m −2 s −1 , and day length in hours.

Estimation of µ and L from the high-frequency Chl a sensor data
The high-frequency Chl a data were used to estimate phytoplankton growth (µ) and loss (L) rates following a method detailed by Soulié et al. (2022a).First, the high-frequency Chl a data were corrected for non-photochemical quenching as detailed in the Supplement.Then, each Chl a cycle was separated into an "increasing period" and a "decreasing period".The increasing period lasted from sunrise until the maximum Chl a fluorescence was reached, generally a few minutes to a few hours after sunset.The decreasing period lasted from this maximum until the next sunrise.For each period, an exponential fit was applied to the Chl a data, and L was estimated from the decreasing period.Then, µ was estimated from the increasing period.The detailed calculations are presented in the Supplement.
2.7 Estimation of GPP and CR from the high-frequency DO sensor data DO data were used to estimate daily GPP and CR during the day (CRdaytime) and the night (CRnight), and daily CR was estimated following the method detailed by Soulié et al. (2021).This method is derived from the free-water diel oxygen technique (Staehr et al., 2010) and was specially developed for mesocosm experiments and to consider variability in both the coupling between day-night and DO cycles and the respiration occurring during the day and at night.Briefly, each DO cycle was separated into a "positive instantaneous net community production period" (during which DO increases) and a "negative instantaneous net community production period" (during which DO decreases).For each period, the DO was smoothed using a five-point sigmoidal model.These smoothed data were then used to estimate oxygen metabolic parameters in two major steps.First, the oxygen exchange term between water and the atmosphere was calculated, taking into consideration its dependence on temperature and salinity.Then, instantaneous and daily metabolic parameters were estimated.A precise description of the method is provided by Soulié et al. (2021) and in the Supplement.

Maximum photosystem II quantum yield measurements
Phytoplankton photosynthetic performance was estimated based on the fluorescence of the photosystem II (PSII).Subsamples of 1.5 mL from the Niskin water sampler were collected daily and analysed using a portable pulse amplitude modulation fluorometer (AquaPen-C AP 110-C; Photon Systems Instruments, Czech Republic).The maximum quantum yield of photosynthesis (F v : F m ) was measured after a 30 min acclimation period in the dark to ensure that all photosystem II reactional centres were open.The measurement was done using the "OJIP" protocol and an excitation wavelength of 450 nm (Strasser et al., 2000).

Heterotrophic bacterial abundance measurements
Heterotrophic bacterial abundance was assessed daily using flow cytometry.For this purpose, 1.5 mL samples were collected from the Niskin water sampler and fixed using glutaraldehyde (Grade I; Sigma-Aldrich; 4 % final dilution) and then frozen into liquid nitrogen before being maintained at −80 °C until further analyses.The samples were stained with SYBR Green I (S7563; Invitrogen; 0.25 % final dilution) (Marie et al., 1997).Analyses were performed using a FAC-SCanto2 flow cytometer (Becton Dickinson; set to low speed for 3 min), and internal cell size standards (cytometry fluorescent beads; Polysciences, Inc.) of 1 and 2 µm diameter were added to each run.Bacterial populations were identified and counted via stained green fluorescence (530 nm) and relative side scatter (Courboulès et al., 2021(Courboulès et al., , 2023)).

Statistical analyses
To test the difference between the control and the terrestrial runoff treatments, we performed repeated-measures analyses of variance (RM-ANOVAs) with the treatment as a fixed factor and time as a random factor (nlme package, R software) over the entire experiment (after the addition of soil, d2-d18) and over shorter periods to assess specific trends.Data from d1 were not included in the statistical analyses as sampling was performed before adding the soil, simulating the terrestrial runoff, in the runoff mesocosms.Statistical significance was set at p<0.05.Before performing the RM-ANOVAs, the assumptions of homoscedasticity and normality were checked using the Levene and Shapiro-Wilk tests, respectively.When these assumptions were not met even after transforming the data (using a log or square-root transformation), a non-parametric Kruskal-Wallis test was performed instead of RM-ANOVA.The non-parametric Spearman's correlation coefficient was used to assess significant (p<0.05)relationships between the logarithms of the response ratio (LRRs) of the variables.All data management and statistical analyses were performed using the R software (version 4.0.1).

Effects of the terrestrial runoff treatment on physical and chemical conditions
In the control treatment, the water temperature varied from 16.68 ± 0.16 to 17.95 ± 0.65 °C (Fig. 1a) and was not significantly different in the terrestrial runoff treatment compared to the control (Table 1).The salinity was on average 38.42 ± 0.11 in the control treatment, increasing almost continuously throughout the experiment (Fig. 1b).In the terrestrial runoff treatment, the salinity was significantly reduced by 0.7 % (Table 1).Similarly, the DLI was, on average, 18.65 ± 1.45 mol m −2 d −1 in the control treatment (Fig. 1c).The terrestrial runoff drastically decreased it by 76 % on d2 and by, on average, 43 % over the entire experiment.This negative effect was stronger during the first half of the experiment (52 % from d2-d11) and was attenuated during the second half of the experiment (27 % from d12-d18) (Table 1).In the control treatment, pH varied between 8.10 ± 0.05 and 8.19 ± 0.01 (Fig. 1d), decreasing from d1 to d10 before stabilization until the end of the experiment.In the runoff treatment, it was significantly reduced by, on average, 0.03 units (8.06 ± 0.01 to 8.19 ± 0.01 in the runoff treatment) (Table 1).In addition, pCO 2 ranged from 292.49 ± 0.45 to 368.27 ± 43.97 µatm in the control treatment (Fig. 1e).In the runoff treatment, it was significantly higher by 9 % compared to the control despite returning to the control level by the end of the experiment (Table 1).DIC concentrations ranged from 2184.04 ± 14.89 to 2230.44 ± 0.76 µmol kg −1 (Fig. 1f).They were significantly higher by 1 % in the runoff treatment than in the control, with the highest difference between treatments on d2 (3 %) (Table 1).DOC concentrations were on average 1.70 ± 0.10 mg L −1 in the control treatment (Fig. 1g).In the terrestrial runoff treatment, DOC concentrations were not immediately enhanced after the addition of soil, reaching higher concentrations than in the control only in the middle and at the end of the experiment.However, no significant https://doi.org/10.5194/bg-21-1887-2024Biogeosciences, 21, 1887-1902, 2024 differences were observed between the treatments (Table 1).POC and PON concentrations displayed similar dynamics over time.POC concentrations ranged from 0.26 ± 0.01 to 0.55 ± 0.09 mg L −1 (Fig. 1h), whereas PON concentrations ranged from 0.04 ± 0.01 to 0.07 ± 0.01 mg L −1 (Fig. 1i).They were both significantly enhanced by 32 %-50 % by the terrestrial runoff at the beginning of the experiment (d2-d12), then decreased to the level of the control (Table 1).The concentrations of dissolved inorganic nutrients exhibited different trends.Nitrate and nitrite concentrations ranged from 0.29 ± 0.03 to 0.50 ± 0.01 µM in the control treatment and were not significantly affected by the terrestrial runoff (Fig. 1j, Table 1).Conversely, while ammonium concentrations remained rather constant in the control treatment during the experiment, ranging from 0.02 ± 0.01 to 0.12 ± 0.07 µM, they increased significantly in the terrestrial runoff treatment, reaching 0.96 ± 0.04 µM on d10 before decreasing to the control level on d16 (Fig. 1k; Table 1).Orthophosphate concentrations ranged from 0.03 ± 0.01 to 0.07 ± 0.01 µM in the control treatment, with peaks at the beginning and the end of the experiment (Fig. 1l).They were significantly higher in the terrestrial runoff treatment but only in the middle of the experiment (63 % from d10-d13) (Table 1).

Effects of the terrestrial runoff treatment on
phytoplankton: Chl a and growth and loss rates In the control treatment, the Chl a concentrations ranged from 0.83 ± 0.30 to 1.91 ± 0.45 µg L −1 (Fig. 3a).They remained rather constant during the first half of the experiment before increasing from d11 to d13 and then decreasing until the end of the experiment.In the terrestrial runoff treatment, they were significantly lower than in the control, particularly during the first part of the experiment (−70 % from d2-d11) (Table 1).However, at the end of the experiment, they in-  creased rapidly from d11-d15, even surpassing the control level.
In the control, µ ranged from 0.06 ± 0.04 to 0.64 ± 0.06 d −1 , peaking on d4, d7, and d12 (Fig. 3b).In the terrestrial runoff treatment, it was significantly lower than in the control by an average of 53 % from d2-d10 (Table 1).However, it increased drastically during the second half of the experiment and was significantly almost 3 times higher than in the control from d12-d17.The value of L varied from 0.19 ± 0.12 to 0.95 ± 0.05 d −1 in the control treatment and was rather constant (Fig. 3c).In the terrestrial runoff treatment, it was significantly lower than in the control by an average of 32 % throughout the experiment and by an average of 60 % from d3-d14 (Table 1).However, it was higher than in the control from d2-d3 and came back to the control level at the end of the experiment.As a consequence of the generally higher L than µ in the control, the µ : L ratio was below 1 on 13 out of the 16 d (Fig. 3d).It ranged from 0.10 ± 0.06 to 2.76 ± 2.26.The terrestrial runoff significantly increased the µ : L ratio by an average of 305 % over the entire experiment (Table 1).The greatest difference between treatments was found on d13, when the ratio was almost 11 times higher in the terrestrial runoff than in the control treatment.

Effects of the terrestrial runoff treatment on primary production, respiration, and photosynthetic efficiency
In the control treatment, GPP ranged from 0.26 ± 0.02 to 0.78 ± 0.03 g O 2 m −3 d −1 (Fig. 4a).After decreasing from d1-d2, it increased until it reached its maximum on d7 and then decreased almost continuously until the end of the experiment.In the terrestrial runoff treatment, it increased significantly by an average of 37 % at the middle of the experiment, from d9-d14 (Table 1).When the GPP was normalized by the daily Chl a concentration, it was significantly higher in the terrestrial runoff treatment than in the control by an average of 312 % throughout the experiment (Fig. 4b).
In the control treatment, CR ranged from 0.18 ± 0.01 to 0.67 ± 0.02 g O 2 m −3 d −1 , and it showed a similar dynamic as GPP (Fig. 4c).It was significantly enhanced in the terrestrial runoff treatment by an average of 46 % over the entire experiment (Table 1).
The GPP : CR ratio ranged from 0.94 ± 0.12 to 1.69 ± 0.19 in the control treatment, and it was higher than 1 on 16 out of 17 d (Fig. 4d).In the terrestrial runoff treatment, it decreased significantly by an average of 32 % during the first half of the experiment (d2-d10) before increasing and reaching the control level during the second half of the experiment (Table 1).Consequently, it was higher than 1 only on 10 out of the 17 d.
In the control, the maximum PSII quantum yield, an indicator of the maximum potential photosynthetic capacity, ranged from 0.24 ± 0.01 to 0.55 ± 0.04 (Fig. 4e).It was not significantly different between the treatments over the entire experiment; however, it increased significantly by 43 % in the terrestrial runoff treatment from d8-d11 (Table 1).
When the CR was normalized by the daily Chl a concentration, it was significantly higher in the terrestrial runoff treatment than in the control by an average of 420 % throughout the experiment (Fig. 4f).
Finally, when CR was normalized by total bacterial abundance (Fig. 4g), it was not significantly different between treatments apart from d9-d14 when it was significantly higher by an average of 154 % in the terrestrial runoff treatment than in the control (Table 1).

Correlation matrix between the responses of phytoplankton processes, community metabolism, and environmental variables
To assess the relationships between the effects of the terrestrial runoff on various variables, Spearman's correlations were calculated between the LRRs of phytoplankton processes, community metabolism, and environmental variables.All significant correlations are shown in the matrix (Fig. 5).GPP was positively correlated with NH + 4 and PO 3− The present study aims to evaluate the effects of a simulated terrestrial runoff on key plankton processes in a coastal Mediterranean lagoon.During the first half of the experiment (d2-d11), the simulated terrestrial runoff strongly decreased available light (−52 %), consequently depressing phytoplankton biomass (−70 %) and growth rate (−53 %), as highlighted by the strong positive correlations among light availability, Chl a, and phytoplankton growth.The phytoplankton community investigated in the present study was typical of the Thau Lagoon in spring (Trombetta et al., 2019), mainly composed of diatoms, cryptophytes, and small nanoand picophytoplankton (Courboulès et al., 2023).The negative effect of light limitation induced by the runoff on phytoplankton biomass is consistent with a mesocosm experiment performed in the Baltic Sea, where terrestrial organic matter addition reduced phytoplankton biomass through light attenuation (Mustaffa et al., 2020), and, generally, with a metaanalysis conducted on 108 studies reporting an average 23 % reduction in photoautotroph biomass in response to experimentally reduced light across various freshwater and coastal ecosystems (Striebel et al., 2023).However, in the Thau Lagoon, a previous experiment reported a positive effect of soil addition, simulating a terrestrial runoff, on phytoplankton (Deininger et al., 2016).Nevertheless, the sinking of the added soil during the experiment performed by Deininger et al. (2016), despite the use of a mixing pump to limit sedimentation, might have rapidly lessened light attenuation, possibly releasing phytoplankton from the negative effect of light limitation.In addition, the experiment was conducted in late spring and early summer, when in light-oversaturated conditions (Trombetta et al., 2019), whereas our experiment was performed in spring, when light could be limiting for phytoplankton metabolism.being farther from natural terrestrial runoffs (Scharnweber et al., 2021).In the present experiment, the maturation step aimed at mimicking processes which naturally occur during the transportation of soil to coastal waters during terrestrial runoffs, such as the degradation of the most labile organic compounds (Müller et al., 2018).The 14 d maturation period can be considered a long residence time in river water, regarding the fact that flash floods in the Mediterranean region are usually faster.Therefore, it can be supposed that the terrestrial matter added in the present study contained lower levels of labile organic compounds than what can be found during flash floods.This emphasizes the need for extreme caution when comparing experimental studies investigating terrestrial runoff effects because protocols often differ from one study to another.
In the present study, the lower phytoplankton biomass and growth rate in the runoff treatment were coupled with an overall decrease in phytoplankton loss rate from d3 until d14 (−60 %).Phytoplankton loss could be caused by multiple factors that occur concomitantly, including grazing by predators, viral lysis, sedimentation, and natural death (Landry and Hassett, 1982;Brussaard, 2004).As the terrestrial runoff in-duced a negative effect on phytoplankton biomass during the first half of the experiment, it may have led to lower prey availability for its predators, resulting in a lower phytoplankton loss rate.This is supported by the negative effect of the simulated runoff on protozooplankton abundances reported in the present experiment (Courboulès et al., 2023), which may be due to both lower phytoplankton abundance and higher grazing pressure from metazooplankton.Finally, the lower phytoplankton loss rate suggests that terrestrial runoffs could have important consequences for the entire plankton food web of coastal Mediterranean waters by disrupting phytoplankton loss processes, including grazing, which is the first link in the herbivorous food web (Legendre and Rassoulzadegan, 1995;Mostajir et al., 2015).
In contrast to phytoplankton biomass and growth, the gross primary production returned quickly to the control level (d4) and was even enhanced by the terrestrial runoff after a few days.This result was unexpected considering that oxygen production strongly depends on light, which was reduced by the runoff.However, we showed that the ratio of primary production to Chl a increased by more than 3 times in the runoff treatment, suggesting a strong enhancement of the phytoplankton photosynthetic efficiency to cope with lower light availability.Supporting this, the maximum PSII quantum yield, an indicator of the maximum potential photosynthetic activity (Strasser et al., 2000), increased significantly in the middle of the experiment in the terrestrial runoff treatment, further suggesting an increase in photosynthetic efficiency under light attenuation induced by the runoff.Moreover, this mismatch between oxygen production and carbon fixation, which has already been reported in a mesocosm experiment in Antarctic coastal waters (Deppeler et al., 2018), might be explained by the fact that photosynthetic carbon fixation is a two-stage process.The first is the conversion of light to energy in the chloroplast, which produces oxygen as a by-product, and the second is the use of the produced energy to convert carbon dioxide into sugars through the Calvin cycle with the RuBisCO enzyme.Under stress conditions, the energy produced can also be used in alternative pathways and not carbon dioxide conversion -that is, mainly respiration and photoacclimation (Behrenfeld et al., 2004;Halsey et al., 2010).Hence, we hypothesized that in the runoff treatment, a significant part of the energy produced by photosynthesis was not converted to growth but was used instead in alternative pathways, which explains the observed mismatch between oxygen production and phytoplankton biomass.An alternative hypothesis is that the high quantity of particulate matter added through the simulated runoff induced a strong sedimentation of a part of the phytoplankton community toward the bottom of the mesocosm enclosures (Kiorboe et al., 1990).This sedimentation could have partly contributed to the mismatch between GPP and Chl a, as sedimented phytoplankton could have continued to produce oxygen, while being undetected by both manual and sensor monitoring of Chl a.Such sedimentation has already been suggested after heavy loadings of terrestrial matter occurred during a natural flash flood event in Thau Lagoon, during which most of the microbial production may have been exported through sedimentation (Fouilland et al., 2012).Nonetheless, it should be noted that the samples of sedimented material in the sediment https://doi.org/10.5194/bg-21-1887-2024Biogeosciences, 21, 1887Biogeosciences, 21, -1902Biogeosciences, 21, , 2024 traps are not fully analysed yet, thus preventing the characterization of the role of sedimentation in the responses of GPP and Chl a with certainty.Simultaneously, community respiration was strongly enhanced (+53 %) by the simulated terrestrial runoff.In marine waters, planktonic bacterial respiration is generally assumed to represent a major part of community respiration (Robinson, 2008).In the present study, bacterial abundance was significantly enhanced by the runoff during the first part of the experiment (d2-d9), which is congruent with the higher respiration at that time.This suggests that higher bacterial abundances are certainly responsible for the higher R reported in the runoff treatment during the first part of the experiment.However, bacterial abundances then significantly decreased during the middle of the experiment (d9-d14) in the runoff treatment, while respiration remained significantly higher than in the control treatment, resulting in a positive response of R normalized by bacterial abundance at this time of the experiment.This suggests that respiration was mostly not sustained by bacteria at that time of the experiment but by other biological compartments instead.Because Chl a was still strongly depressed by the runoff during this period of the experiment, resulting in extremely high CR : Chl a rates, the hypothesis of an increase in phytoplankton respiration is not plausible.An increase in zooplankton respiration might instead explain the positive effect on community respiration, as the abundance of some groups of metazooplankton was significantly enhanced by the runoff treatment (Courboulès et al., 2023), and the concomitant increase in PO 3− 4 suggests a strong phosphorus excretion from zooplankton (Andersen et al., 1986;Vadstein et al., 1995).
As a consequence of the faster and greater increase in respiration compared to that in gross primary production, the terrestrial runoff resulted in a decrease in the productionto-respiration ratio and a shift toward heterotrophy of the metabolic index of the planktonic system during the first half of the experiment, as similarly reported after simulating a terrestrial runoff in a tropical reservoir (Trinh et al., 2016).Concomitantly, pCO 2 was significantly higher in the terrestrial runoff treatment, which was certainly because of the higher respiration as the responses of both variables were strongly correlated.These results are consistent with a study of 15 Swedish lakes that reported higher respiration leading to switches towards a heterotrophic metabolic index and increased pCO 2 in response to increased terrestrial carbon runoffs (Ask et al., 2012).Therefore, the present experiment shows, for the first time, to our knowledge, in Mediterranean coastal lagoons, that terrestrial runoffs could potentially shift coastal Mediterranean lagoons, such as the Thau Lagoon, from being net oxygen producers to net oxygen sinks in spring.Therefore, the respiration-driven gain in CO 2 can temporarily change the magnitude and direction of the airsea CO 2 exchange, potentially switching the ecosystem from a CO 2 sink to a CO 2 source for the atmosphere.

Enhanced nutrient availabilities boosted phytoplankton processes during the second half of the experiment
During the second half of the experiment (d12-d18), the phytoplankton biomass and processes increased in the terrestrial runoff treatment in contrast to what occurred during the first half of the experiment.This might be explained by the higher dissolved inorganic nutrient availability in the runoff treatment, as both NH + 4 and PO 3− 4 concentrations were significantly higher in the terrestrial runoff treatment than in the control in the middle of the experiment before being consumed and returning to the control level.The higher NH + 4 concentrations possibly resulted from bacterial remineralization, as NH + 4 is mostly produced by the bacterial remineralization of organic matter in coastal waters (Nixon, 1981;Glibert, 1982).In contrast, the higher PO 3− 4 availability could be linked to grazing on bacteria, as grazers feeding upon bacteria generally show high phosphorus excretion rates (Andersen et al., 1986).
Enhanced nutrient availability may have fuelled phytoplankton growth to such an extent that the positive effect of nutrient availability surpassed the negative effect of light attenuation.This result suggests a trade-off mechanism between light and nutrient availability, whereby phytoplankton metabolism is enhanced or depressed depending on the extent of nutrient enrichment compared to the light attenuation associated with terrestrial runoffs.This mechanism has already been reported for northern lakes (Klug, 2002;Isles et al., 2021) and even during mesocosm experiments evaluating the addition of dissolved organic matter into coastal waters of various regions (Deininger et al., 2016;Traving et al., 2017;Andersson et al., 2023).The present study provides additional support for this mechanism in Mediterranean coastal waters and highlights the importance of considering it when modelling their response to terrestrial runoffs.
As mentioned earlier, Chl a strongly increased during the second part of the experiment in the runoff treatment.This positive response was mainly due to an increase in the abundance of diatoms, mainly Chaetoceros and Cylindrotheca, Cyanobacteria, and autotrophic dinoflagellates (Courboulès et al., 2023).In addition, the pico-and nanophytoplankton abundances counted with flow cytometry also increased at this time of the experiment (Courboulès et al., 2023).Overall, d very good agreement was found between the response of the Chl a concentration and phytoplankton abundances, measured by both microscopy and flow cytometry, during the entire experiment (Courboulès et al., 2023).The accumulation of phytoplankton biomass during the second part of the experiment in the runoff treatment was related to the strong increase in phytoplankton growth rate from d10, while the phytoplankton loss rate remained low until the end of the experiment.Consequently, the growth-to-loss ratio was significantly enhanced by more than 10 times compared to that of the control.This suggests an uncoupling between phy-toplankton growth and its loss factors, such as zooplankton and/or viruses, at that time in the experiment, possibly because phytoplankton grew too quickly compared to its predators.Nonetheless, this emphasizes the potentially substantial structural impacts of terrestrial runoff on plankton communities and their intricate interactions within aquatic food webs as was recently documented in lakes (Strandberg et al., 2023).
The results of the present experiment suggest that the climate-change-related intensification of terrestrial runoffs could temporarily alter metabolic and trophic indices of the water column of the lagoon during productive seasons (Trombetta et al., 2019), potentially shifting it towards heterotrophy and disrupting its trophic balance.Coupled with terrestrial-runoff-induced shifts in microbenthic net community production towards heterotrophy (Liess et al., 2015), these alterations could interact with ongoing shifts occurring in the lagoon, such as the changes in trophic functioning towards mixotrophy and heterotrophy related to oligotrophication (Derolez et al., 2020b).Such consequences may also be seen in other Mediterranean lagoons, as turbidity and extreme flood events were reported to control phytoplankton abundance and phenology in oligotrophic Mediterranean coastal lagoons in southern France and Corsica (Bec et al., 2011;Ligorini et al., 2022).Even though the results of the present study come from a single mesocosm experiment, implying that their generalization should be implemented with care, they emphasize the importance of considering the effects of terrestrial runoffs on plankton-mediated processes in modelling projections of Mediterranean coastal waters under future climate scenarios.
Author contributions.FrV and BM designed the mesocosm experiment, and FrV, SM, and BM managed it.FrV, JC, MH, SM, FlV, CC, FJ, and BM participated in the daily samplings of the experiment.CC performed the analysis of pH, pCO 2 , and dissolved inorganic carbon, with the help of FlV.TS processed the sensor data, made all related analyses, and wrote the original draft of the paper, with inputs from all authors.All authors read and approved the final version of the paper.
Competing interests.The contact author has declared that none of the authors has any competing interests.
Disclaimer.Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper.While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors.

F 2 Figure 2 .
Figure 2. Daily average bacterial abundances in the control (blue) and terrestrial runoff (gold) treatments.Error bars represent the range of the observations (min and max values).

Figure 3 .
Figure 3. Daily average (a) chlorophyll a (Chl a), (b) phytoplankton growth rate (µ), (c) phytoplankton loss rate (L), and (d) growth : loss ratio (µ : L) in the control (blue) and terrestrial runoff (gold) treatments.Error bars represent the range of the observations (min and max values).Note that µ and L could not be estimated on d18 owing to the lack of a complete fluorescence cycle.

4
concentrations and negatively correlated with bacteria abundance and POC and PON concentrations.CR was positively correlated with pCO 2 and POC and PON concentrations, while being negatively correlated with µ, Chl a, salinity, DLI, and pH.In addition, µ was positively correlated with L, Chl a, salinity, DLI, and NO − 2 + NO − 3 and negatively with bacterial abundances and POC and PON concentrations.Similarly, L was positively correlated with Chl a and salinity and negatively correlated with pCO 2 .In addition, Chl a was positively correlated with salinity, DLI, and NO − 2 + NO − 3 and negatively correlated with POC and PON, while bacterial abundances were positively correlated with DOC and negatively correlated with NO − 2 and NO − 3 , NH + 4 , and PO 3− 4 .Among environmental variables, it should be noted that DLI and POC and PON concentrations were negatively correlated and terrestrial runoff depressed phytoplankton processes and shifted the metabolic balance of the system towards heterotrophy during the first half of the experiment Finally, Deininger et al. (2016)  used a resin in their soil extraction procedure, yielding higher inorganic and organic nutrient concentrations in their extract compared to the protocol performed in the present study but https://doi.org/10.5194/bg-21-1887-2024Biogeosciences,21, 1887Biogeosciences, 21,  -1902Biogeosciences, 21,  , 2024

Figure 4 .
Figure 4. Daily average (a) gross primary production (GPP), (b) GPP normalized by chlorophyll a (GPP : Chl a), (c) community respiration (CR), (d) GPP : CR ratio, (e) maximum quantum yield (F v : F m ) of photosystem II (PSII), (f) CR normalized by chlorophyll a (CR : Chl a), and (g) CR normalized by bacterial abundance (CR : bacteria) in the control (blue) and terrestrial runoff (gold) treatments.Error bars represent the range of the observations (min and max values).Note that GPP and CR could not be estimated on d18 owing to the lack of a complete oxygen cycle.

Figure 5 .
Figure 5. Correlation matrix based on Spearman's correlations between the log response ratio (LRR) of phytoplankton processes, community metabolism, and environmental variables.Only significant (p<0.05)correlations are shown in the matrix.Green illustrates positive correlations and purple negative correlations.(GPP: gross primary production; CR: respiration; µ: growth rate; L: loss rate; Chl a: chlorophyll a; Bacteria: bacterial abundance; DLI: daily light integral; pCO 2 : partial pressure of CO 2 ; DOC: dissolved organic carbon; DIC: dissolved inorganic carbon; POC and PON: particulate organic carbon and nitrogen; NO − 2 and NO − 3 : nitrite and nitrate concentrations; NH + 4 : ammonium concentration; PO 3− 4 : orthophosphate concentration.)

Table 1 .
Summary table of the statistical comparison and the relative change between the terrestrial runoff and the control treatments in percentages.The significance level was set to 0.05 and significant p values and their corresponding relative changes are highlighted in bold.When an RM-ANOVA was performed, its F value is given in brackets, and when a Kruskal-Wallis was performed instead, this is indicated with KW.