lable at ScienceDirect

Estuarine, Coastal and Shelf Science 168 (2016) 58e70

Contents lists avai

Estuarine, Coastal and Shelf Science

journal homepage: www.elsevier .com/locate/ecss

Drivers of the carbonate system seasonal variations in aMediterranean gulf

Gianmarco Ingrosso a, b, *, Michele Giani a, Cinzia Comici a, Martina Kralj a,Salvatore Piacentino c, Cinzia De Vittor a, Paola Del Negro a

a Istituto Nazionale di Oceanografia e di Geofisica Sperimentale (OGS), Oceanography Section, Via A. Piccard 54, 34151 S. Croce (TS), Italyb University of Trieste, Department of Life Sciences, Via Giorgieri 10, 34127 Trieste, Italyc ENEA-CLIMOSS, Via Catania 2, 90141 Palermo, Italy

a r t i c l e i n f o

Article history:Received 20 November 2014Received in revised form6 August 2015Accepted 2 November 2015Available online 6 November 2015

Keywords:pHCarbonate systemBuffer capacityRiverOcean acidificationGulf of TriesteAdriatic SeaMediterranean Sea

* Corresponding author. Present address: Departmronmental Science and Technology, University of Sale

E-mail address: gianmarco.ingrosso@unisalento.it

http://dx.doi.org/10.1016/j.ecss.2015.11.0010272-7714/© 2015 Elsevier Ltd. All rights reserved.

a b s t r a c t

The effects of different physical and biogeochemical drivers on the carbonate systemwere investigated ina semi-enclosed coastal area characterized by high alkalinity riverine discharge (Gulf of Trieste, NorthernAdriatic Sea, Mediterranean Sea). Our 2-year time-series showed that a large part of the seasonal car-bonate chemistry variation was controlled by the large seasonal change of seawater temperature, thoughair-sea CO2 exchange, biological activity (primary production-respiration), and riverine inputs alsoexerted a significant influence.

With the exception of summer, the Gulf of Trieste was a sink of atmospheric carbon dioxide, showing avery strong CO2 fluxes from atmosphere into the sea (�16.10 mmol m�2 day�1) during high wind speedevent of north easterly Bora wind. The CO2 influx was particularly evident in winter, when the biologicalactivity was at minimum and the low seawater temperature enhanced CO2 solubility. During spring, thedrawdown of CO2 by primary production overwhelmed the CO2 physical pump, driving a significantdecrease of dissolved inorganic carbon (DIC), [CO2], and increase of pHT25 �C. In summer the primaryproduction in surface waters occurred with the same intensity as respiration in the bottom layer, so thenet biological effect on the carbonate system was very low and the further reduction of seawater CO2

concentration observed was mainly due to carbon dioxide degassing induced by high seawater tem-perature. Finally, during autumn the respiration was the predominant process, which determined anoverall increase of DIC, [CO2], and decrease of pHT25 �C. This was particularly evident when the break-down of summer stratification occurred and a large amount of CO2, generated by respiration andsegregated below the pycnocline, was released back to the whole water column. Local rivers alsosignificantly affected the carbonate system by direct input of total alkalinity (AT) coming from thechemical weathering of carbonate rocks, which dominate the river watershed. Our finding clearlydemonstrates a high AT concentration in low salinity surface waters (AT max ¼ 2742.8 mmol kg�1) and anegative AT-salinity correlation. As a result the Gulf of Trieste revealed a low Revelle factor (10.1) and oneof the highest buffer capacities of the Mediterranean Sea (ßDIC ¼ 0.31 mmol kg�1), which allows thesystem to store a significant amount of atmospheric CO2 with a small decrease of seawater pH.

© 2015 Elsevier Ltd. All rights reserved.

1. Introduction

The ocean currently absorbs approximately a quarter of the CO2added to the atmosphere from human activities each year, greatlyreducing the impact of this greenhouse gas on climate (Sabine et al.,

ent of Biological and Envi-nto, 73100 Lecce, Italy.(G. Ingrosso).

2004; Le Qu�er�e et al., 2009, 2010; 2015; McKinley et al., 2011).However, the oceanic uptake of anthropogenic CO2 also leads to agradual acidification of the world's ocean, which is now a wellknown phenomena commonly referred to as “ocean acidification”(Caldeira and Wickett, 2003; Doney et al., 2009). Since the begin-ning of the industrial era, the ocean surface pH has decreased by 0.1pH units and the global pH reduction actually rangesbetween �0.0014 and �0.0024 yr�1 (IPCC, 2013). In the NWMediterranean Sea the ocean acidification process is also occurring

G. Ingrosso et al. / Estuarine, Coastal and Shelf Science 168 (2016) 58e70 59

(De Carlo et al., 2013) and the acidification rate has been estimatedto be 0.003 pH units yr�1 (Geri et al., 2014).

The projections of ocean acidification in the coastal area, whereparadoxically some of the most vulnerable taxa live, are still lackingand they are remarkably difficult to predict due to the greatcomplexity of these ecosystems. Here, the CO2 levels are regulatedby different physical and biological drivers such as air-sea CO2 ex-changes, ecosystem metabolism, net community calcification rate,riverine inputs of carbonate weathering products, as well as hy-drological features that determine the mixing between open oceanand coastal waters (Duarte et al., 2013). These processes can causechanges in pH from seasonal to decadal and longer timescales,enhancing or reducing the ocean acidification. For example, intemperate eutrophic waters with high degree of stratification andreduced ventilation in summer/autumn months (Melzner et al.,2012), the oxidation of organic matter in subsurface waterslowers dissolved oxygen concentrations, adds CO2 to solution, re-duces pH, and, over longer time scales, can exacerbate the oceanacidification process (Byrne et al., 2010; Cai et al., 2011). Conversely,an increase of total alkalinity export from river to the sea couldincrease the seawater buffer capacity and mitigate regionally theeffect of ocean acidification (Raymond and Cole, 2003; Raymondet al., 2008; Watanabe et al., 2009).

The Northern Adriatic Sea (NAd), situated in the northernmostregion of the Mediterranean Sea, receives freshwaters from manyrivers and is subject to wide variations of temperature and salinity.Though oceanographic studies carried out in this area date back tothe 18th century, long time series of chemical oceanography pa-rameters began only in the 1970s and the carbonate chemistryobservations were set up only recently (Turk et al., 2010).

Climatic and anthropogenic driven changes have occurred in thelast decades in the NAd. An oligotrophication trend was observedby different authors (Solidoro et al., 2009; Mozeti�c et al., 2010;Giani et al., 2012), who ascribed this process to a reduction ofannual river discharges and river-borne nutrient inputs. Moreover,Luchetta et al. (2010) have recently detected a pH decrease in theNAd, underlining the importance to follow also in this area theocean acidification process.

The objective of the present work is to understand how physicaland biogeochemical drivers could influence the seasonal cycles ofmarine carbonate system in a shallow Mediterranean gulf. Specif-ically, both the air-sea CO2 exchanges and thermal/not-thermalcontributions to the seasonality of the CO2 system are quantified.We also investigate the buffer capacity of the area, which is affectedby the products of carbonate mineral weathering coming fromfreshwater discharges of the local rivers.

2. Materials and methods

2.1. Study area

The Gulf of Trieste (Fig. 1) is a shallow semi-enclosed basinlocated in the northernmost part of the Adriatic Sea, with amaximum depth of 25 m and a surface area of about 600 km2

(Malej et al., 1995). It is an example of a coastal region of freshwaterinfluence (ROFI, Simpson, 1997), characterized by a high spatial andtemporal variability of its physical, chemical, and biological prop-erties. Four rivers discharge waters into the gulf: Isonzo, Timavo,Ri�zana, and Dragonja. The main freshwater input is through theIsonzo river from the north-west Italian coast (Cozzi et al., 2012),which is characterized by high discharges rate during spring andautumn, while drought periods occur in winter and summer. Thesedischarges, due to the prevailing cyclonic circulation, affect mainlythe north-western part of the gulf and only during the freshets incomitance with southerly winds the plume spread throughout the

gulf (Querin et al., 2007).The water column is subject to a strong variation in tempera-

ture, salinity, and vertical stratification during the year (Mala�ci�c andPetelin, 2001). Seawater temperature shows a seasonal cycle fromwinter minima of 6 �C to summer maxima of >29 �C (Cardin andCelio, 1997; Celio et al., 2006), whereas the salinity ranges from33 to 38 (Mala�ci�c et al., 2006). During spring the freshwater inputand surface heating cause the thermohaline stratification, whichincreases in strength during summer. In autumn and winter,convective and mechanical mixing, induced by water cooling andwind, disrupts the vertical stratification leading to a mostly ho-mogeneous water column. Also in this period, strong Bora windevents, a regional ENE cold wind, cause the formation of densewaters, which are a fundamental component for the generation ofNorth Adriatic Dense Water (Artegiani et al., 1997; Mala�ci�c andPetelin, 2001).

River watersheds along the Gulf of Trieste are composed mainlyof carbonate rocks (e.g., limestone and dolomite) (Pleni�car et al.,2009) and they have some of the highest carbonate-weatheringin the world (Szramek et al., 2007, 2011). The riverine inputs thenrepresent a very important source of inorganic carbon, which cansignificantly affect the marine carbonate system and the whole Ccycle.

2.2. Sampling and analyses

The study was performed at the C1 station in the northeasternpart of the Gulf of Trieste (45�42.050Ne13�42.600E), a coastal timeseries station (bottom depth of 17.5 m) where a long term ecolog-ical monitoring has been in place since 1970 (http://nettuno.ogs.trieste.it/ilter/BIO/). The data presented here refer to the period ofMarch 2011eFebruary 2013. Discrete water samples were collectedmonthly with 5-L Niskin bottles at four depths (0.5, 5, 10, 15 m) inorder to measure total alkalinity (AT), pH, dissolved oxygen, andnutrients.

Depth profiles of salinity (Practical Salinity Scale of 1978) andtemperature (�C) were determined using a Conductivity-Temperature-Depth/Pressure (CTD) recorders (Idronaut 316 andSBE 19 Plus Seacat probe).

For the total alkalinity, samples were pre-filtered on glass-fibrefilters (Whatman GF/F) into a 500 mL narrow-necked borosilicateglass bottle. Filtration was performed to remove phytoplanktoncells and particles of CaCO3, derived from karstic watershed orcalcifying organisms, which respectively can interact with HCltitrant solution and dissolve during measurements due to acidadditions, producing a non-negligible error in the exact estimationof the AT concentration (Gattuso et al., 2010; Hydes et al., 2010;Bockemon and Dickson, 2014). Each bottle was poisoned with100 mL of saturated mercuric chloride (HgCl2) to halt biologicalactivity, sealed with glass stoppers and stored at 4 �C in the darkuntil analysis. Total alkalinity was determined by potentiometrictitration in an open cell (SOP 3b, Dickson et al., 2007) using a non-linear least squares approach. The titration procedure was per-formed with the titration unit Mettler Toledo G20 interfaced with acomputer, using the data-acquisition software LabX. After titrationthe data were processed and AT calculated with a computer pro-gram developed at OGS (Istituto Nazionale di Oceanografia e diGeofisica Sperimentale) similar to that listed in SOP 3 of DOE (1994)and adapted to work in association with the Mettler Toledo LabXsoftware. The HCl titrant solution (0.1 mol kg�1) was prepared inNaCl background, to approximate the ionic strength of the samples,and calibrated against certified reference seawater (CRM, Batch#107, provided by A.G. Dickson, Scripps Institution of Oceanog-raphy, USA). Accuracy and precision of the AT measurements onCRMwere determined to be less than ±2.0 mmol kg�1. Additionally,

Fig. 1. Map of the Gulf of Trieste in the North Adriatic Sea, with the C1 sampling site. Other time series stations of the carbonate system are shown in the map: Paloma in the Gulf ofTrieste, Point B and DYFAMED in the western Mediterranean Sea.

G. Ingrosso et al. / Estuarine, Coastal and Shelf Science 168 (2016) 58e7060

repeated measurements (n � 3) on in-house standards of naturalseawater were undertaken every day prior to sample analysis inorder to monitor the instrument's accuracy and precisionfrequently.

The pH was measured using a double-wavelength spectropho-tometer (Cary 100 Scan UVeVisible) and the indicator dye m-cresolpurple (Merck-105228) following SOP 6b of Dickson et al. (2007). Toavoid CO2 gain or loss, unfiltered seawater samples were collecteddirectly into quartz cuvettes with a 10 cm pathlength, leaving nohead space. pH was always measured within a few hours fromsampling and values are reported on total scale (pHT). The analyt-ical precision was estimated to be ±0.002 pHT units, determined byreplicates from the same Niskin bottle. The accuracy was assessedanalysing the pHT of three different CRM Batch (#107, #124, #125)and comparing it with the pHT reference value calculated fromcertificated AT-DIC and the dissociation constant from Mehrbachet al. (1973) refitted by Dickson and Millero (1987). The pHTmeasured on these reference materials was lower than the theo-retical value by 0.028 ± 0.006 pH units. This difference probablyderived from the use of an unpurified indicator, which coulddetermine an error as large as 0.02 pH units (Yao et al., 2007; Liuet al., 2011). The pHT results at C1 station were therefore cor-rected with the linear fit equation between calculated Batch pHTand measured Batch pHT (pHT Calc. ¼ 1.0111pHT Meas. � 0.0589;R2 ¼ 0.998) in order to reduce the offset found between “true” pHTand measured pHT with unpurified indicator.

For the determination of dissolved oxygen concentration (DO;mmol kg�1) the Winkler method was followed (Grasshoff et al.,1999) using an automated titration system (Mettler Toledo G20)with potentiometric end-point detection (Outdot et al., 1988). Theanalytical precision and accuracy was ±1.5 mmol kg�1.

The samples to measure the dissolved inorganic nutrient con-centration (nitrite, NO2, nitrate, NO3, ammonium, NH4, phosphatePO4, and silicic acid, H4SiO4; mmol kg�1) were prefiltered on 0.7 mmpore size glass-fibre filters (Whatman GF/F) immediately after thesampling and stored at �20 �C in polyethylene vials until analysis.The samples were defrosted and analysed colorimetrically with a

Bran þ Luebbe Autoanalyzer 3, according to Grasshoff et al. (1999).Detection limits for the procedure were 0.003 mM, 0.01 mM,0.04 mM, 0.02 mM, and 0.02 mM for NO2, NO3, NH4, PO4, and H4SiO4,respectively.

CO2 monitoring in air was run on a weekly basis and withcontinuous measurements at ENEA Station for Climate Observa-tions on the island of Lampedusa (Italy). Duplicate air samples werecollected and measured using a Siemens Ultramat 5E non-dispersive infrared analyzer (NDIR). Sampling and measurementprocedures are described elsewhere (Chamard et al., 2003).Determination of CO2 mixing ratio on flask samples was obtainedwith a standard deviation of about 0.04 ppm. To ensure accuratemeasurements, eight CO2 mixtures, certified by NOAA ClimateMonitoring and Diagnostic Laboratory (CMDL), were used as pri-mary standards.

2.3. Calculation of derived parameters

The other carbonate system parameters, including seawaterpartial pressure of CO2 (pCO2; matm), dissolved inorganic carbon(DIC; mmol kg�1), and aragonite saturation state (UAr), werecalculated using the CO2SYS program (Lewis and Wallace, 1998)through in situ AT, pHT (25 �C), temperature, salinity, phosphate,and silicate data. Carbonic acid dissociation constants (i.e., pK1 andpK2) of Millero (2010) were used for the computation, as well as theDickson constant for the ion HSO4

� (Dickson, 1990) and boratedissociation constant of Lee et al. (2010).

The apparent oxygen utilization (AOU; mmol kg�1) was deter-mined as the difference between the saturated oxygen concentra-tion and the observed oxygen concentration, providing anapproximation to the balance between biological processes of pri-mary production and respiration.

The air-sea surface flux of CO2 (FCO2; mmol m�2 day�1) wasestimated from the difference between the partial pressure of CO2at sea surface (pCO2) and the concentration of carbon dioxide in theatmosphere (pCO2 atm) according to the equation:

G. Ingrosso et al. / Estuarine, Coastal and Shelf Science 168 (2016) 58e70 61

FCO2 ¼ k K0ðpCO2 � pCO2 atmÞ

where k is the gas transfer velocity of CO2 and K0 is the solubilitycoefficient of CO2 in seawater at in situ temperature and salinity(Weiss, 1974). We use the gas transfer velocity parameterization asfunction of quadratic wind speed, following the empirical expres-sion of Wanninkhof (1992) for short-term or steady wind speed.The wind speed data (u; m s�1) was recorded at 10 m height in thenearby weather PALOMA e ISMAR CNR station and daily averagevalues were used for the calculation of gas transfer velocity. Thepartial pressure of atmospheric pCO2 was derived from the CO2molar fraction measurements at the remote island of Lampedusa,using daily barometric pressure, sea surface temperature andsalinity measured during experimental activity. Negative FCO2value indicates air-to-sea flux, whereas positive value represents anet CO2 outgassing from the water body to the atmosphere.

To distinguish the effect of seasonal temperature change fromthe effect of the other processes, the method of Takahashi et al.(2002) was applied on surface and bottom pCO2 data. The effectof temperature changes on pCO2 (pCO2thermal) was estimated ac-cording to the equation

pCO2thermal ¼ pCO2mean$exp½0:0423ðTobs � TmeanÞ�

where pCO2mean was the biannual mean pCO2, and (Tobs � Tmean)was the difference between the observed and mean temperatureover two years. The resulting pCO2thermal indicates the values ofpCO2 that would be expected if a parcel of water, having a meanpCO2 of the two years, is subjected to seasonal temperature changesunder isochemical conditions.

To remove the thermal effect from the observed pCO2 (pCO2obs),the values were normalised to a constant seawater temperatureusing the equation

pCO2not�thermal ¼ pCO2obs$exp½0:0423ðTmean � TobsÞ�The not-thermal pCO2 (pCO2not-thermal) then represents the pCO2

variation that can be driven by biological processes (primary pro-duction, respiration, and calcification), advection of water masses,and air-sea CO2 exchange.

2.4. Statistical analysis

The normality of the data was assessed with the ShapiroeWilktest (Shapiro andWilk, 1965). Due to their non-normal distribution,a non parametric approach was adopted. Statistically significantdifferences between seasons for each parameter were tested withthe KruskaleWallis rank sum test (Kruskal and Wallis, 1952),whereas the multiple comparison test (Siegel and Castellan, 1988)was used to determine which pair of seasons were different. Ana-lyses were performed with R program and the package pgirmess.

3. Results

Over the two years of study, the seawater temperature (Fig. 2b)varied from 4.79 �C (February 2012) to 26.16 �C (July 2011),exhibiting a DT of 21.4 �C and of 18.5 �C in the 2011 and 2012respectively. Seawater temperature was significantly different(p < 0.001) among each season (Table 1), with the only exception ofspring and autumn, which instead showed a comparable meanvalue.

Inwinter, the salinity reached amean value of 37.40 (± 0.99) andit was remarkably higher than the surface waters of spring(36.21 ± 0.94) and summer (36.18 ± 0.98). As a result, from May toSeptember of both years a strong vertical halocline gradient was

present (up to DS z 3.5 in June 2012).Dense water masses with density anomaly greater than

29.0 kg m�3 were observed in winter (Fig. 2a). These watersrepresent the Northern Adriatic Dense Water (NAdDW), whichspread as a dense current along thewestern Adriatic shelf, filling upthe middle and south Adriatic depressions. The highest densityanomaly value (30.46 kg m�3) was detected during February 2012as a consequence of an extreme cooling event due to prolongedblowing of the cold ENE wind (Bora), which determined on theAdriatic shelf the extraordinary dense water formation event ofwinter 2012 (Mihanovi�c et al., 2013).

Total alkalinity concentration spanned over a wide range. Thehighest value was detected at the surface in August 2011(2742.8 mmol kg�1), whereas the minimum was reached at thebottom in January 2012 (2657.2 mmol kg�1) (Fig. 2c). These re-sults reflect the AT seasonal variation, which presented higherand comparable average values in spring and summer(2695.5 ± 16.3 and 2696.2 ± 21.3 mmol kg�1, respectively), andlower and comparable concentrations in winter and autumn(2677.7 ± 12.7 mmol kg�1 and 2678.7 ± 11.1 mmol kg�1, respec-tively) (Fig. 3).

In winter the pHT at in situ temperature was elevated and ho-mogeneous along the water column, with an average value of 8.203(SD ¼ 0.03) (Figs. 2d and 3). From spring to summer, the meanseasonal pHT gradually decreased. In these periods, characterizedby a strong thermohaline stratification, the pHT in the bottomwaters was generally lower (up to 0.208 pHT units in late summer)than at the surface. In both years, the lowest pHT values weredetected in the water masses under the pycnocline of September(7.906 in 2011; 7.969 in 2012). After this month, with the break-down of the vertical stratification, the pHT increased and becameagain uniform in the whole water column, with an average value of8.129 ± 0.04.

The dissolved oxygen concentration (Figs. 2e and 3) showedseasonal variations very similar to the in situ pHT. High DO con-centrations (279.6 mmol kg�1, SD ¼ 20.7) characterized the wintermonths, whereas during spring and summer the concentrationswere substantially lower, reaching the minimum values inSeptember at the bottom (137.0 mmol kg�1 in 2011; 155.1 mmol kg�1

in 2012).The seasonal change of dissolved inorganic carbon (Figs. 2f and

3) followed a pattern of lower concentrations during spring andsummer periods (2361.5 ± 27.3 mmol kg�1 and2332.1 ± 42.6 mmol kg�1, respectively) and higher values in winter(2379.2± 9.6 mmol kg�1). Generally, fromMay to late-July, a gradualdecrease at the surface was followed by an increase in the watermasses under the pycnocline, which started in late-spring andreached maximum in the summertime period.

Ammonia varied from 0.23 mmol kg�1 to 3.61 mmol kg�1

(Fig. 4a), with low concentration mainly during spring(0.86 mmol kg�1, SD ¼ 0.54 mmol kg�1) and high values at thebottom waters in summer (3.61 mmol kg�1 in Aug. 2011;3.26 mmol kg�1 in Sept. 2012). PO4 concentrations were low in allseasons, spanning between non-detectable (<0.01 mmol kg�1) to0.15 mmol kg�1, with higher values above the pycnocline duringspringesummer (Fig. 4b).

4. Discussion

In the river-influenced coastal areas, the seasonal variability ofthe carbonate system is the result of several physical and biologicalprocesses such as seasonal changes in seawater temperature, netcommunity production, carbonate mineral dissolution/formation,air-sea CO2 gas exchange, riverine inputs, and mixing betweenwaters with different characteristics. In the following sections we

Fig. 2. Monthly variations at the C1 station of: (a) salinity, density anomaly (st e white isolines), (b) temperature, (c) total alkalinity (AT), (d) in situ pHT, (e) dissolved oxygen (DO),(f) dissolved inorganic carbon (DIC).

G. Ingrosso et al. / Estuarine, Coastal and Shelf Science 168 (2016) 58e7062

analize these factors separately in order to understand how theyaffect the marine carbonate system in the Gulf of Trieste.

4.1. Influence of physical and biological processes on the carbonatesystem

The concentrations of individual species of the carbonate systemchange with temperature, salinity, and pressure (because of varia-tion of the equilibrium constants with T, S, and p) (Zeebe andWolf-Gladrow, 2001). For example, colder seawater has higher [CO2] and½HCO3

�� and lower ½CO32�� at a given DIC and AT, and they present

higher pHT values.To remove the thermal effect on the observed pHT and to eval-

uate how the biological processes of primary production andrespiration can influence seasonal variability of pHT, it is possible toconsider the seasonal variation of pHT measured at constant tem-perature of 25.0 �C (pHT25 �C) with respect to the apparent oxygenutilization (AOU).

The winter season was characterized by a homogeneous watercolumn and showed a very low biological activity (AOU ¼1.6 ± 17.1 mmol kg�1; Figs. 3 and 4c) and pHT25 �C (7.957 ± 0.015;Figs. 3 and 4d). From spring to summer, with the formation of thepycnocline, the water column acted as a two-layer system, withprimary production processes prevailing in the upper layer andrespiration of organic matter in the lower layer. As a consequence,surface waters presented a higher pHT25 �C and were supersatu-rated in oxygen (AOU < 0). In July 2012, for example, pHT25 �C andAOU at 5 m depth were 8.145 and �40.1 mmol kg�1, respectively,whereas the total dissolved inorganic carbon concentration wasvery low (2260 mmol kg�1) compared to other periods (Fig. 4c andd). This strongly negative AOU value reflects a pronounced primaryproductivity, which influences the carbonate system reducing pCO2and DIC, and increasing the pHT25 �C. Otherwise, in spring andsummer bottom waters, the pHT25 �C was generally lower than atthe surface, and in particular three periods with a very low pHT25 �Cwere observed: September 2011, June 2012, and September 2012

Table 1Results of KruskaleWallis rank sum test and multiple comparison test among seasons on different physicalechemical parameters. In bold are reported the pair of seasonssignificantly different.

KruskaleWallis rank sum test Temperature Salinity DO AT pHT in situ

c2 p-value c2 p-value c2 p-value c2 p-value c2 p-value

77.0414 <0.001 15.1352 0.002 61.6022 <0.001 22.3682 <0.001 60.3131 <0.001

Multiple comparison test (p ¼ 0.05) Obs. dif Critical. dif Obs. dif Critical. dif Obs. dif Critical. dif Obs. dif Critical. dif Obs. dif Critical. dif

Autumnespring 6.90 21.22 14.96 21.22 33.44 21.22 27.90 21.00 25.67 21.00Autumnesummer 30.65 21.22 10.48 21.22 8.98 21.00 26.00 21.00 14.13 21.00Autumnewinter 39.58 21.22 13.77 21.22 43.98 21.00 0.47 21.22 42.68 21.22Springesummer 37.54 21.22 4.48 21.22 42.42 21.22 1.90 21.00 39.79 21.00Springewinter 32.69 21.22 28.73 21.22 10.54 21.22 27.43 21.22 17.02 21.22Summerewinter 70.23 21.22 24.25 21.22 52.96 21.00 25.53 21.22 56.81 21.22

KruskaleWallis rank sum test DIC AOU pHT at 25 �C CO2 FCO2

c2 p-value c2 p-value c2 p-value c2 p-value c2 p-value

34.6226 <0.001 33.1994 <0.001 41.4368 <0.001 26.18 <0.001 8.19 0.04231

Multiple comparison test (p ¼ 0.05)a Obs. dif Critical. dif Obs. dif Critical. dif Obs. dif Critical. dif Obs. dif Critical. dif Obs. dif Critical. dif

Autumnespring 7.19 21.00 45.48 21.22 10.04 21.00 16.48 21.00 5.00 9.77Autumnesummer 18.21 21.00 29.88 21.00 23.96 21.00 22.79 21.00 5.33 9.77Autumnewinter 28.54 21.22 23.77 21.00 26.18 21.22 14.47 21.22 4.33 9.77Springesummer 25.40 21.00 15.60 21.22 13.92 21.00 6.31 21.00 10.33 9.77Springewinter 21.35 21.22 21.70 21.22 36.23 21.22 30.95 21.22 0.67 9.77Summerewinter 46.75 21.22 6.10 21.00 50.14 21.22 37.27 21.22 9.67 9.77

a For FCO2 p ¼ 0.1.

Fig. 3. Box-and-whisker plot presenting the seasonal average of physical and chemical parameters measured in the water column at C1 station. Seasons were chosen as 3 monthsperiods: winter (JanuaryeMarch), spring (AprileJune), summer (JulyeSeptember), and autumn (OctobereDecember).

G. Ingrosso et al. / Estuarine, Coastal and Shelf Science 168 (2016) 58e70 63

(Fig. 4d). Concurrently, also a large positive AOU and high con-centration of dissolved inorganic nutrients (e.g., in September 2011,pHT25 �C ¼ 7.805, AOU ¼ 85.2 mmol kg�1, NH4 ¼ 3.26 mmol kg�1,PO4 ¼ 0.15 mmol kg�1) were detected, due to a high organic matterdegradation which consumes oxygen, remineralizes nutrients,produces carbon dioxide, and leads to a higher concentration ofDIC.

During October, the decrease of seawater temperature and thewater mixing, triggered by strong winds, caused the breakdown ofthe pycnocline and the homogenization of the water column. Thus,

a substantial amount of CO2, generated by the heterotrophic ac-tivity, spread from the bottom to the surface, causing a generalincrease of DIC.

The observed seasonal pattern between pHT25 �C and biologicalactivity, and the decoupling of production and respiration due tothe presence of the pycnocline during springesummer seasons, arevery similar to that found by Cantoni et al. (2012) in the same area(Fig. 1), even if they reported a higher acidification of bottom wa-ters. This difference can be ascribed to the greater depth of theirsampling station (25 m) and to the more intense remineralisation

Fig. 4. Monthly variations at the C1 station of: (a) ammonia (NH4), (b) phosphates (PO4), (c) apparent oxygen utilization (AOU), (d) pHT at 25 �C, (e) Revelle factor (RF) and (f)aragonite saturation state (UAr).

G. Ingrosso et al. / Estuarine, Coastal and Shelf Science 168 (2016) 58e7064

processes registered in the bottom waters during August 2008,which led to a hypoxic event, to a rise of pCO2, and to a decrease ofpH.

It has been suggested that hypoxia in coastal habitats willenhance the future ocean acidification (Feely et al., 2010; Cai et al.,2011; Melzner et al., 2012). Strong hypoxic conditions have beenobserved in the Northern Adriatic Sea, as consequence of markedstratification, restricted mixing, prolonged bottomwater residencetime, and organic matter respiration (Ott, 1992; Giani et al., 1992;Degobbis et al., 2000). However, after the year 2000 only fewcases of dissolved oxygen saturation lower than 20% have beenobserved (Socal et al., 2008; Solidoro et al., 2009; Giani et al., 2012)and considering this trend an exacerbation of ocean acidificationprocesses due to hypoxic events is unlikely. The biologicallyinduced acidification of bottom oxygen-depleted water in summerproduces a short-term but very sharp pH drop of approximately 0.3pH units on a seasonal time scale, which lies in the normal range ofseasonal pH variation of coastal habitats (Duarte et al., 2013).

Another relevant process, which influences the global cycle of thecarbonate system, is the exchange of carbon dioxide between atmo-sphere and seawater. Onannual scale, the coastalwaters of theGulf ofTrieste acted as a CO2 sinkat a rate of�2.14±2.55mmolm�2 day�1 in2012 and�3.27±5.80mmolm�2 day�1 in 2013 (Fig. 5). From Januaryto March, in both years, our results showed a clear influx of CO2 to-ward the sea (�2.84±2.24mmolm�2day�1). The averagemagnitudeof the flux during winter was�3.48 ± 3.02 mmol m�2 day�1 in 2012and�2.21± 1.49mmolm�2 day�1 in 2013. This conditionwas drivenby the low seawater temperature, which led undersaturation of sur-face pCO2 with respect to the atmosphere.

During spring, thermally driven pCO2 increase was counter-balanced by the biologically driven pCO2 decrease. The surfacewaters were still undersaturated with respect to the atmosphericCO2 and air-sea surface fluxes persisted toward the sea. Inparticular, a very high absorption of atmospheric CO2 occurred inMay 2012, when the combined effects of primary production andstrong wind speed resulted in an air-sea CO2 influx

Fig. 5. Monthly air-sea CO2 fluxes (FCO2), daily average wind speed at 10 m above sealevel (u), and sea surface temperature (Temp.) at C1 station.

G. Ingrosso et al. / Estuarine, Coastal and Shelf Science 168 (2016) 58e70 65

of �14.22 mmol m�2 day�1.In late summer, the strong seasonal warming drove pCO2 su-

persaturation in surface waters and outgassing of carbon dioxide inthe atmosphere. These CO2 evasions were particularly evidentduring September 2012 (1.70 mmol m�2 day�1) when the CO2concentration throughout the water column increased, likely asconsequence of mixing with deeper layers enriched in CO2 byrespiration. Winds also played an important role in the atmo-spheric carbon dioxide sequestration as showed by the strong Borablowing in November 2012 (up to 12 m s�1), when the absolutemaximum flux (�16.10 mmol m�2 day�1) was observed.

Our annual average of CO2 flux is significantly lower than themagnitude of �6 mmol m�2 day�1 reported by Turk et al. (2010) inthe southern part of the Gulf of Trieste; this dissimilarity could stemfrom differences in the sampling periods and geographical loca-tions of the sampling stations, which are subject to differentoceanographic conditions. However, the high-frequency measure-ments of sea surface pCO2 reported by Turk et al. (2010) could bemore suitable in the coastal zone (Borges and Gypens, 2010) andprobably could explain the difference in annual air-sea CO2 ex-change between these two neighbouring sites. Flux estimatesderived from data collected daily, weekly, or monthly, are less ac-curate than those obtained by (near-)continuous determinations(De Carlo et al., 2013), because in highly dynamic coastal environ-ments there are a number of processes that display short-termvariability in the carbonate system which easily can be missed bylow-frequency sampling.

It is also interesting to compare the CO2 flux measured in theGulf of Trieste and those of other time series in the northwesternMediterranean Sea (Fig. 1): the DYFAMED site (with monthlysampling, B�egovic and Copin-Mont�egut, 2002), and Point B (withweekly sampling, De Carlo et al., 2013). Compared to our work,these studies used a different wind speed parameterization tocalculate the air-sea CO2 exchange. B�egovic and Copin-Mont�egut(2002) used the parameterization of Wanninkhof and McGillis(1999), whereas De Carlo et al. (2013) calculated the fluxfollowing that of Ho et al. (2006). However, the annual averagewind speed in the three cases was relatively modest (5 m s�1 atDYFAMED, 3.9 m s�1 at Point B, 4.17 m s�1 at C1 station), and whenthe wind speed is in the range of 2e5 m s�1 the choice of param-eterization does not have a large effect on the estimation of thefluxes (De Carlo et al., 2013). At the DYFAMED site the authors re-ported an annual average fluxes of�0.87 mmol m�2 day�1 for 1998and �1.86 mmol m�2 day�1 for 1999, whereas at the Point B theyfound an annual CO2 uptake of�0.53mmolm�2 day�1. These fluxes

are lower than those estimated in the present work at the C1 sta-tion and confirm the higher capacity of the Gulf of Trieste to absorbthe atmospheric CO2, due to its higher primary productivity and thevery low seawater temperature during winter dense water forma-tion episodes.

At C1 station, the seasonal air-sea CO2 exchange pattern (Fig. 3)is consistent with the previous works in marginal seas (Chen andBorges, 2009) and also with recent studies in other areas of theGulf of Trieste (Cantoni et al., 2012; Turk et al., 2013), which indi-cated a CO2 sink from autumn to spring, and a source in summer. Inour study, the seasonal mean CO2 fluxes are comparable among allseasons, except between spring and summer, which instead pre-sented a significantly different mean value (Table 1 andFig. 3; �4.80 ± 5.22 mmol m�2 day�1 in spring and 0.15 ± 1.08 insummer).

The magnitude of the flux during the two Bora wind events(�14.22mmol m�2 day�1 in May 2012 and�16.10mmol m�2 day�1

in November 2012) falls within the range of previous estimates(Cantoni et al., 2012; Turk et al., 2013) and confirms the stronginfluence of these events on the air-sea CO2 exchange. However, onannual time scale, the increase of CO2 flux from the atmosphere tothe sea generated by Bora episodes was estimated to be around 5%(Turk et al., 2013) and it could slightly enhance the ocean acidifi-cation process.

In order to estimate which is the predominant driver (photo-synthesis/respiration, carbonate dissolution/formation, and CO2release/invasion) affecting the marine carbonate system in thedifferent seasons, it is possible to consider a graphical approachbased on a propertyeproperty plot of DIC and AT (Fig. 6), where thereaction path can take on variable slopes depending on the ratio ofdifferent processes (Deffeyes, 1965; Suzuki and Kawahata, 2003;Albright et al., 2013; Andersson and Gledhill, 2013; Nomura et al.,2014). DIC and AT were normalized (nDIC and nAT) to a constantsalinity, by multiplying DIC and AT data by the ratio of 37.81 (themean value of winter bottom layer which is representative of theoffshore waters) to the measured in situ salinity, in order to removethe influence of freshwater input, mixing, and evaporation/pre-cipitation. Moreover, the contribution of calcification and dissolu-tion to the carbonate system seasonal variation has been assumedirrelevant with reasonable approximation. This because at thesaturation state level, salinity, and temperature observed in theGulf of Trieste, the natural abiotic precipitation/dissolution ofCaCO3 does not occur (Morse and He, 1993; Morse et al., 2007;Marion et al., 2009). The influence of CaCO3 precipitation couldderive, then, only from calcifying organisms. However, Mozeti�cet al. (1998) reported in the Gulf of Trieste a very low abundanceof coccolithophores, the most important calcifying phytoplanktongroup in temperate sea, and Cantoni et al. (2012) did not found aclear sing of pelagic calcification rate able to alter the carbonatesystem. No information is available on the potential contribution ofthe benthic calcifying organisms, but this could be also smallbecause the study area is characterized mainly by a sandy-muddyseabed with sporadic coralligenous assemblages (Casellato andStefanon, 2008).

In winter, the slope of the regression line between nDIC and nATwas 0.98 (Fig. 6), so the dissolved inorganic carbon increasedslightly compared to the alkalinity. This relation probably wasdetermined by the atmospheric CO2 uptake, which increased DICwithout affecting total alkalinity. As shown in Fig. 3, during winter,when the biological activity was minimal (AOU z 0), the very lowseawater temperature supported a marked invasion of atmosphericCO2, determining a high concentration of DIC and low pHT 25.0values.

In spring, the slope of the nDIC-nAT regression line was 1.67(Fig. 6) and the ANCOVA test for the Homogeneity of Slopes-

Fig. 6. Relationships between the salinity-normalized DIC (nDIC) and salinity-normalized AT (nAT) by seasons. The vectors in the inset indicate theoretical modification of DIC andAT owing to biogeochemical processes of photosynthesis/respiration, carbonate dissolution/formation, and CO2 release/invasion. Isolines represent the pHT 25.0 calculated at thedifferent AT-DIC with the R package seacarb (Gattuso and Lavigne, 2009).

G. Ingrosso et al. / Estuarine, Coastal and Shelf Science 168 (2016) 58e7066

Intercept found that the winter and spring linear regression weresignificantly different (F ¼ 21.06, p < 0.001). This case is a clearindication of the influence of primary production, as evidenced alsoby a negative AOU (�12.8 mmol kg�1, Fig. 3), which caused thedecrease of dissolved inorganic carbon concentration and the in-crease of total alkalinity and pHT 25.0.

During summer, the primary production persisted in the sur-face layer, but it was counterbalanced by the strong respiration oforganic matter in the bottom waters. As a result, the netecosystem production was close to 0 (AOU ¼ 0.7 ± 30.0 mmol kg�1)and the influence of biological processes on carbonate was verylow. Therefore, the nDIC-nAT slope of 1.06 (Fig. 6), significantlydifferent respect to the spring (F ¼ 4.92, p ¼ 0.01), was probablydue to the CO2 release towards the atmosphere, which determinesa reduction of DIC concentration and a rise of pHT 25.0. Finally inautumn, respiration predominated over photosynthesis(AOU ¼ 18.3 ± 9.6 mmol kg�1). The change of nDIC-nAT regression(Fig. 6) reflected this condition (significant difference betweensummer and autumn Y-intercept: F ¼ 15.36, p < 0.001), as well asthe carbonate system: during the transition from summer toautumn the pHT25 �C and total alkalinity decreased, whereas theDIC and dissolved CO2 increases (Fig. 3).

These results confirm that primary production and respirationprocesses are relevant drivers of the large annual pH variation inthe coastal ecosystem (Duarte et al., 2013). When they are inequilibrium, the net CO2 balance is essentially neutral. However, onlonger timescales, if one of these processes prevails over the other,the net community metabolism changes and the pH is affecteddrastically. It has been reported that the eutrophication couldcounteract or enhance the effects of ocean acidification byincreasing pH when primary production CO2 uptake prevails(Borges and Gypens, 2010), or by decreasing pH when respiratoryCO2 release prevails (Cai et al., 2011). Oligotrophication, instead,

could lead to acidification (Nixon, 2009), adding to the ocean pHreduction derived by anthropogenic CO2 uptake.

An oligotrophication trend has been recognised in the Gulf ofTrieste (Solidoro et al., 2009; Mozeti�c et al., 2010; Giani et al., 2012),which can be ascribed to a reduction in outflow and nutrients loadsfrom the Isonzo River, due to increased frequency of drought pe-riods generated by the ongoing climate changes or by a larger hu-man usage of continental waters (Cozzi and Giani, 2011). If thisprocess continues in the future, the primary production coulddecrease and, in this way, the ocean acidification process could beenhanced. Moreover, a decrease in alkalinity discharge wouldreduce the buffer capacity of the system.

4.2. pCO2 seasonal variation

Like other carbonate system parameters also the pCO2 seasonalvariation is affected by physical (i.e. temperature, mixing, and air-sea CO2 exchange) and biological processes (i.e. photosynthesis,respiration, and calcification). To distinguish the effect of seasonaltemperature change from the effect of the other processes, themethod of Takahashi et al. (2002) was applied on surface andbottom pCO2 data. The not-thermal pCO2 (pCO2not-thermal) thenrepresents the pCO2 variation that can be driven by biologicalprocesses (primary production, respiration, calcification), advec-tion of water masses, and air-sea CO2 exchange.

In Fig. 7, the thermal and not-thermal pCO2 variations at C1station are reported for the surface and bottom layers. The resultsare shown as differences (dpCO2) with respect to the February 2012,chosen as reference because it presented the lowest temperature,the minimum biological activity, and the weakest riverine inputs.

The seasonal thermal effect (dpCO2thermal) was predominantboth at surface and in the bottom waters. The highest values werereached at the surface in July 2011 and August 2012 (329 and

Fig. 7. Monthly variations of surface and bottom values of dpCO2, dpCO2thermal, anddpCO2not-thermal at C1 station.

Fig. 8. AT-Salinity and Revelle factor-Salinity plots at C1 station. The lines represent thelinear relationships, black dots and dashed lines fitting refer to the data collected afterhigh riverine discharges.

G. Ingrosso et al. / Estuarine, Coastal and Shelf Science 168 (2016) 58e70 67

323 matm, respectively). This thermodynamic increase of pCO2 wasdamped by the not-thermal effect, probably due to the biologicaldrawdown of CO2 (dpCO2not-thermal < 0), which was shown to beelevated above the pycnocline from winter to late summer. Inparticular, only in March 2012 the not-thermal effect was greaterthan the thermal variation in the whole water column, maybe as aconsequence of the spring phytoplankton bloom influence.

Respiration processes at the bottom (dpCO2not-thermal > 0) werealso evident. In particular, during September 2011, the dpCO2not-

therm was up to 144 matm, and the dpCO2 reached the maximumvalue of 405 matm. This event indicated the importance of theremineralisation processes to generate an excess of carbon dioxidein the deeper layer, which together with the thermal pCO2 increasedetermined a high seawater concentration of CO2 during summer.

To evaluate the relative importance of temperature with respectto mixing, air-sea CO2 exchange and biological processes on theannual pCO2 cycle, it is useful to consider the T/B ratio, whereT ¼ pCO2thermal max � pCO2thermal min and B ¼ pCO2not-thermalmax � pCO2not-thermal min. In marine ecosystem where the not-thermal effect on the pCO2 dynamics exceeds the thermal effect,the T/B ratio varies from 0 to 1, whereas in areas where the influ-ence of temperature changes on pCO2 cycle is predominant, the T/Bratio is greater than 1 (Takahashi et al., 2002). At C1 station the T/Bis 1.01, which indicates a similar importance between thermal andnot-thermal effect. These results differ from those reported byCantoni et al. (2012) who found at PALOMA station, in the center ofthe Gulf of Trieste, a T/B ratio of 1.35, and therefore a strongercontrol of temperature on pCO2 cycle. These two different resultscould be explained by the different locations of the two stations:the PALOMA station is an offshore site, whereas the C1 station isvery close to the coast. As shown previously, the air-sea CO2 ex-changes between these two sites are very similar, so probably the T/B ratio at the C1 station could be a result of a more relevant influ-ence of biological processes and freshwater advection on theannual pCO2 variation.

4.3. Effects of river inputs on the carbonate system

In our study the AT samples collected in low salinity surfacewaters (depth � 5 m) of JuneeJulyeAugust 2011, subsequently to ahigh freshwater input (Fig. 2a and c), showed high concentrations

(AT � 2680.6 mmol kg�1) and strong negative correlation withsalinity (y ¼ �43.63x þ 4258.68 r ¼ �0.90, p ¼ 0.02) (Fig. 8).Considering all AT data, the relationships between AT and salinitywas still evident but with a weaker correlation (Fig. 8,y¼�12.33xþ 3144.13 r¼�0.58, p < 0.001). This probably resultedfrom the influence of different river end-members (Tam�se et al.,2014), as well as the high temporal variability of riverine AT in-puts (Cantoni et al., 2012). Moreover, the total alkalinity concen-tration could be magnified in a highly productive coastal system(Hydes et al., 2010). In particular, the presence of organic carbon,denitrification, biological carbonate mineral precipitation, photo-synthesis and respiration, could all cause perturbation of AT con-centrations without affecting salinity, and then alter the AT-Srelationship.

The AT value extrapolated at 0 salinity was 4258.7 mmol kg�1

during the event of high freshwater inputs, whereas considering alldata it was 3144.1 mmol kg�1. These two results are consistent withthe previous data on the total alkalinity concentration of Isonzo andTimavowatershed (Szramek et al., 2011; Tam�se et al., 2014), the twomost significant sources of freshwater to the Gulf (Cozzi et al.,2012). In the same area, a recent work using stable isotope signa-ture of dissolved inorganic carbon suggested that freshwater inputsduring spring, which is a period of freshets, can represent about 16%of marine DIC (Tam�se et al., 2014).

4.4. Buffer capacity of the system

The Revelle factor is a measure of the seawater's capacity toabsorb CO2 from the atmosphere and it is defined as the ratio of thefractional change in the partial pressure of carbon dioxide in theatmosphere to the fractional increase of the total inorganic carbonin the seawater (Revelle and Suess, 1957). A low Revelle factorimplies that, for a given increase in atmospheric CO2, the concen-tration of dissolved inorganic carbon will be higher than in waterswith a high Revelle factor.

At the C1 station the Revelle factor (RF) was highly variable,ranging from 8.8 to 12.1 (Fig 4e). Due to the strong temperature andalkalinity dependency, the lowest values (between 8.8 and 9.1)were found in spring and summer warm surface waters, when thehighest temperature and total alkalinity concentration weredetected. During autumn and winter, the lowering of seawater

Table 2Statistical summary of several buffer factors in the different seasons at C1 station(March 2011eFebruary 2013). Mean, standard deviation (SD), minimum (Min) andmaximum (Max) are reported.

Buffer factor Seasons Mean SD Min Max

RF Winter 10.7 0.3 10.3 11.5Spring 10.1 0.5 9.1 11.4Summer 9.6 0.8 8.8 12.1Autumn 10.2 0.3 9.7 10.7All data 10.1 0.6 8.8 12.1

gDIC (mmol kg�1) Winter 0.24 0.01 0.23 0.25Spring 0.25 0.01 0.23 0.28Summer 0.27 0.02 0.22 0.28Autumn 0.25 0.01 0.24 0.26All data 0.25 0.01 0.22 0.28

ßDIC (mmol kg�1) Winter 0.29 0.01 0.27 0.30Spring 0.31 0.02 0.27 0.35Summer 0.33 0.03 0.25 0.36Autumn 0.30 0.01 0.29 0.32All data 0.31 0.02 0.25 0.36

uDIC (mmol kg�1) Winter �0.31 0.01 �0.28 �0.33Spring �0.35 0.03 �0.29 �0.41Summer �0.38 0.04 �0.26 �0.43Autumn �0.34 0.02 �0.31 �0.37All data �0.34 0.04 �0.26 �0.43

G. Ingrosso et al. / Estuarine, Coastal and Shelf Science 168 (2016) 58e7068

temperature drove an increase of Revelle factor (10.2 ± 0.3 and10.7 ± 0.3 respectively), but the highest value was found in thebottomwater of September 2011 (max ¼ 12.1), due to the high CO2concentration produced by strong respiration processes.

The two-year average RF value that we found (10.1 ± 0.6) is inagreement with previous studies for the Adriatic Sea (Luchettaet al., 2010; �Alvarez et al., 2014). This confirms that the Gulf ofTrieste has a relatively high uptake capacity for the anthropogenicCO2. However, considering the average RF in the different seasons,two extreme and opposite situations can be detected: the Revellefactor was 9.6 in summer and 10.7 in winter, which representrespectively the lowest and the highest values that we can find inthe open Mediterranean Sea (�Alvarez et al., 2014). Thus, the RF ofthe system is more likely to change seasonally than in openwaters,ranging between a very high and very low capacity to take upcarbon for a given atmospheric CO2 increase.

The influence of freshwater on the carbonate system of coastalwaters has implications with regard to their buffer capacity. Theestuarine buffer capacity is typically lower than the one in the openocean, due to low alkalinity discharge by rivers (Aufdenkampeet al., 2011). However, this is not the case of our sampling area.The riverine input of total alkalinity has a lowering effect on theRevelle factor, i.e. it increases the buffer capacity of the system. Asshown in Fig. 8, during a high river discharge event (black dots) thesystem was well buffered (RF z 9) and the Revelle factor was veryrelated to the river inputs of alkalinity.

However, an explicit determination of carbonate systemresponse as a result of CO2 uptake from the atmosphere cannot bedetermined using the Revelle factor. Therefore, we choose thebuffer factors formulated by Egleston et al. (2010). These factorsprovide a more direct quantification of seawater's ability to bufferchanges in the [CO2] (gDIC), pH (ßDIC), and UAr (uDIC) due to theaccumulation of atmospheric CO2. The buffer factors gDIC, ßDIC, anduDIC express the fractional variation of [CO2], [Hþ], and ½CO3

2��respectively, when DIC changes at constant AT (air-sea CO2 ex-change). The values of gDIC and ßDIC are positive, while uDIC isnegative because the addition of CO2 to seawater increases theconcentration of dissolved carbon dioxide and hydrogen ion, butdecreases the concentration of carbonate. The three buffer factors,thus defined, have dimensions of mmol kg�1 and the results for C1station are shown in Table 2. By examining the buffer factors in thedifferent periods, they presented similar seasonal variations. Thehighest values were observed in summer surface waters, due to theriverine inputs of total alkalinity and high seawater temperaturethat shifts in acid-base dissociation constants. Instead low buffercapacity occurred in cold winter waters and also in the bottomwaters during September as a result of the strong seasonal pHdecrease induced by remineralisation of organic matter.

Comparing the two-year average values of several buffer factorswith those calculated by �Alvarez et al. (2014) in different Mediter-ranean sub-basins, the Gulf of Trieste presented the highest valuesof gDIC and ßDIC, whereas uDIC value was very similar to that foundin the Adriatic Sea. This is a very interesting result, because it showsthat the waters of the gulf are the most resistant of all Mediterra-nean Sea to changes in [CO2] and pH induced by ocean acidificationprocess. They also present a good capacity to buffer the ½CO3

2��decrease and the reduction of aragonite saturation state (Fig. 4e).

The reason for the high buffer capacity of the Gulf of Trieste isthe high total alkalinity concentration resulting from carbonateweathering discharge delivered by local rivers. In general, changesin land use, precipitation, and runoff can alter the alkalinity exportfrom land to coastal areas and their buffer capacity. For example,the Mississippi River in the last 50e100 years exported ~50% morealkalinity in the Gulf of Mexico, due to increasing areas of croplandand increasing precipitation over the watershed, which currently

offsets a portion of regional coastal acidification (Raymond andCole, 2003; Raymond et al., 2008). In the Gulf of Trieste, longtime records of river alkalinity are not present, but recent studiesshowed a marked decrease of river flows (Cozzi et al., 2012) whichprobably could lead to a reduction of river alkalinity export andbuffer capacity of the system.

5. Conclusions

The goal of this study was to understand how physical andbiogeochemical drivers could influence seasonal cycles of marinecarbonate system in a shallow coastal area of the North Adriatic Sea.The experimental results indicate that a large part of seasonal pCO2variation was driven by change in seawater temperature, never-theless also the other processes of air-sea CO2 exchange, biologicalactivity, and mixing play an important role.

The Gulf of Trieste showed a significant CO2 uptake in all seasonswith the exception of summer, when a weak degassing wasdetected. This demonstrates that the Gulf of Trieste is a sink overthe bi-annual observational period, but with marked seasonalvariations. During winter, when the biological activity was verylow, the atmospheric CO2 uptake was set mainly by low seawatertemperature. This thermally driven carbon dioxide sequestrationcontinued in the spring season and it was reinforced by the strongeffect of primary production, which caused a reduction of dissolvedseawater CO2 and a greater flux of atmospheric carbon dioxidetoward the sea. During summer, the primary production in surfacewaters was counterbalanced by respiration of organic matter in thebottom layer, so the net biological effect on the carbonate systemwas very low and the overall reduction of seawater CO2 concen-tration observed was mainly due to CO2 degassing generated fromhigher seawater temperatures. When the water column turnedfrom strongly-stratified in summer to well-mixed in autumn, thehigh amount of CO2, produced by respiration and that was segre-gated below the pycnocline, was released and spread throughoutthe column increasing the seawater CO2 concentration.

Although two years are not enough to detect trends of acidifi-cation in this coastal system, this study showed that the very lowseawater temperature favour the physical pump and absorption ofCO2, especially during dense water formation events, which cantransport anthropogenic CO2 towards the Southern Adriatic Sea

G. Ingrosso et al. / Estuarine, Coastal and Shelf Science 168 (2016) 58e70 69

and the Eastern Mediterranean Deep Waters. By contrast, theriverine inputs of total alkalinity contributes to setting a lowRevellefactor and a great capacity to store atmospheric CO2 with a loweffect on the pH of the system, which, as noted, is one of the mostbuffered areas of the whole Mediterranean Sea.

Our results contribute to a very limited set of observationsregarding the ocean acidification in the coastal Mediterraneanareas and the role of physical and biogeochemical processes oncarbonate system seasonal variation. Additional studies are neededto better characterize the future trend of riverine inflows in the Gulfof Trieste and to understand if the ecosystem goes towards anoligotrophication, because the reduction of river input of alkalinityand nutrients over long time scale could decrease the buffer ca-pacity and primary production, making the ecosystem morevulnerable to the ocean acidification.

Finally, due to our low time resolution of sampling, data re-ported here represent the seasonal variation of the carbonate sys-tem without considering the effects of extreme short-term events(i.e. high stream flow, precipitation, storm, hypoxia), which morefrequently occur in the coastal area. Therefore, continuous in situmeasurements are needed to better characterize these events andtheir influence on the global ocean acidification process.

Acknowledgements

We thank Gianpiero Cossarini and Giorgio Bolzon of OGSOceanography Section for the software development. MassimoCelio and the crew of Folaga vessel of ARPA FVG for the aid insampling and CTD measurements. Meteorological data were kindlysupplied by OSMER ARPA FVG. The authors would like to thankGiovanna Russo for language revision and we are very grateful totwo anonymous reviewers for their critical comments. Thisresearchwas financially supported byMEDSEA FP7 Project Contractn. 265103.

References

Albright, R., Langdon, C., Anthony, K.R.N., 2013. Dynamics of seawater carbonatechemistry, production, and calcification of a coral reef flat, central Great BarrierReef. Biogeosciences 10, 6747e6758. http://dx.doi.org/10.5194/bg-10-6747-2013.

�Alvarez, M., Sanle�on-Bartolom�e, H., Tanhua, T., Mintrop, L., Luchetta, A., Cantoni, C.,Schroeder, K., Civitarese, G., 2014. The CO2 system in the Mediterranean Sea: abasin wide perspective. Ocean Sci. 10, 69e92.

Andersson, A.J., Gledhill, D., 2013. Ocean acidification and coral reefs: effects onbreakdown, dissolution, and net ecosystem calcification. Annu. Rev. Mar. Sci. 5,321e348.

Artegiani, A., Bregant, D., Paschini, E., Pinardi, N., Raicich, F., Russo, A., 1997. TheAdriatic Sea general circulation. Part I: air-sea interactions and water massstructure. J. Phys. Oceanogr. 27, 1492e1514.

Aufdenkampe, A.K., Mayorga, E., Raymond, P.A., Melack, J.M., Doney, S.C., Alin, S.R.,Aalto, R.E., Yoo, K., 2011. Riverine coupling of biogeochemical cycles betweenland, oceans, and atmosphere. Front. Ecol. Environ. 9, 53e60.

B�egovic, M., Copin-Mont�egut, C., 2002. Processes controlling annual variations inthe partial pressure of CO2 in surface waters of the central northwesternMediterranean Sea (Dyfamed site). Deep Sea Res. Part II 49 (11), 2031e2047.

Bockmon, E.E., Dickson, A.G., 2014. A seawater filtration method suitable for totaldissolved inorganic carbon and pH analyses. Limnol. Oceanogr. Methods 12,191e195. http://dx.doi.org/10.4319/lom.2014.12.191.

Borges, A.V., Gypens, N., 2010. Carbonate chemistry in the coastal zone respondsmore strongly to eutrophication than to ocean acidification. Limnol. Oceanogr.55, 346e353.

Byrne, R.H., Mecking, S., Feely, R.A., Liu, X., 2010. Direct observations of basin-wideacidification of the North Pacific Ocean. Geophys. Res. Lett. 37 (2), L02601.http://dx.doi.org/10.1029/2009GL040999.

Cai, W.-J., Hu, X., Huang, W.-J., Murrell, M.C., Lehrter, J.C., Lohrenz, S.E., Chou, W.-C.,Zhai, W., Hollibaugh, J.T., Wang, Y., Zhao, P., Guo, X., Gundersen, K., Dai, M.,Gong, G.-C., 2011. Acidification of subsurface coastal waters enhanced byeutrophication. Nat. Geosci. 4, 766e770.

Caldeira, K., Wickett, M.E., 2003. Oceanography: anthropogenic carbon and oceanpH. Nature 425 (6956), 365.

Cantoni, C., Luchetta, A., Celio, M., Cozzi, S., Raicich, F., Catalano, G., 2012. Carbonatesystem variability in the Gulf of Trieste (North Adriatic Sea). Estuar. Coast. Shelf

Sci. 115, 51e62.Cardin, V., Celio, M., 1997. Cluster analysis as a statistical method for identification

of the water bodies present in the Gulf of Trieste (Northern Adriatic Sea). Boll.Geofis. Teor. Appl. 38, 119e135.

Casellato, S., Stefanon, A., 2008. Coralligenous habitat in the northern Adriatic Sea:an overview. Mar. Ecol. 29, 321e341. http://dx.doi.org/10.1111/j.1439-0485.2008.00236.x.

Celio, M., Mala�ci�c, V., Bussani, A., �Cermelj, B., Comici, C., Petelin, B., 2006. The coastalscale observing system component of ADRICOSM: Gulf of Trieste network. ActaAdriat. 47 (Suppl. l), 65e79.

Chamard, P., Theiry, F., Di Sarra, A., Ciattaglia, L., De Silvestri, L., Grigioni, P.,Monteleone, F., Piacentino, S., 2003. Interannual variability of atmospheric CO2in the Mediterranean: measurements at the island of Lampedusa. Tellus 55B,83e93.

Chen, C.-T.A., Borges, A.V., 2009. Reconciling opposing views on carbon cycling inthe coastal ocean: Continental shelves as sinks and near-shore ecosystems assources of atmospheric CO2. Deep Sea Res. Part II 56, 578e590. http://dx.doi.org/10.1016/j.dsr2.2009.01.001.

Cozzi, S., Giani, M., 2011. River water and nutrient discharges in the NorthernAdriatic Sea: current importance and long term changes. Cont. Shelf Res. 31,1881e1893.

Cozzi, S., Falconi, C., Comici, C., �Cermelj, B., Kovac, N., Turk, V., Giani, M., 2012.Recent evolution of river discharges in the Gulf of Trieste and their potentialresponse to climate changes and anthropogenic pressure. Estuar. Coast. ShelfSci. 115, 1e11.

De Carlo, E.H., Mousseau, L., Passafiume, O., Drupp, P.S., Gattuso, J.-P., 2013. Car-bonate chemistry and airesea CO2 flux in a NW Mediterranean bay over a fouryears period: 2007e2011. Aquat. Geochem. 19 (5e6), 399e442.

Deffeyes, K.S., 1965. Carbonate equilibria: a graphic and algebraic approach. Limnol.Oceanogr. 10, 412e426.

Degobbis, D., Precali, R., Ivancic, I., Smodlaka, N., Fuks, D., Kveder, S., 2000. Long-term changes in the northern Adriatic ecosystem related to anthropogeniceutrophication. Int. J. Environ. Pollut. 13, 495e533.

Dickson, A.G., 1990. Standard potential of the reaction e AgCl(S) þ 1/2h-2(G) ¼Ag(S) þ HCl(Aq) and the standard acidity constant of the ion HSO4

� in syntheticsea water from 273.15 to 318.15 K. J. Chem. Thermodyn. 22, 113e127.

Dickson, A.G., Millero, F.J., 1987. A comparison of the equilibrium constants for thedissociation of carbonic acid in seawater media. Deep Sea Res. Part I 34,1733e1743.

Dickson, A.G., Sabine, C.L., Christian, J.R., 2007. Guide to Best Practices for Ocean CO2Measurements, vol. 3. PICES Special Publication, p. 191.

DOE, 1994. In: Dickson, A.G., Goyet, C. (Eds.), Handbook of Methods for the Analysisof the Various Parameters of the Carbon dioxide System of Sea Water, Version 2.ORNL/CDIAC-74.

Doney, S.C., Fabry, V.J., Feely, R.A., Kleypas, J.A., 2009. Ocean acidification: the otherCO2 problem. Annu. Rev. Mar. Sci. 1, 169e192.

Duarte, C., Hendriks, I., Moore, T., Olsen, Y., Steckbauer, A., Ramajo, L., Carstensen, J.,Trotter, J., McCulloch, M., 2013. Is ocean acidification an open-ocean syndrome?understanding anthropogenic impacts on seawater pH. Estuaries Coasts 36,221e236.

Egleston, E.S., Sabine, C.L., Morel, F.M.M., 2010. Revelle revisited: buffer factors thatquantify the response of ocean chemistry to changes in DIC and alkalinity. Glob.Biogeochem. Cycles 24, GB1002. http://dx.doi.org/10.1029/2008GB003407.

Feely, R.A., Alin, S.R., Newton, J., Sabine, C.L., Warner, M., Devol, A., Krembs, C.,Maloy, C., 2010. The combined effects of ocean acidification, mixing, andrespiration on pH and carbonate saturation in an urbanized estuary. Estuar.Coast. Shelf Sci. 88, 442e449.

Gattuso, J.P., Lavigne, H., 2009. Technical Note: approaches and software tools toinvestigate the impact of ocean acidification. Biogeosciences 6, 2121e2133.

Gattuso, J.-P., Lee, K., Rost, B., Schulz, K., 2010. Approaches and tools to manipulatethe carbonate chemistry. In: Riebesell, U., Fabry, V.J., Hansson, L., Gattuso, J.-P.(Eds.), Guide for Best Practices in Ocean Acidification Research and DataReporting. Office for Official Publications of the European Union, Luxembourg,pp. 41e52.

Geri, P., Yacoubi, El, Goyet, C., 2014. Forecast of sea surface acidifcation in theNorthwestern Mediterraean Sea. J. Comput. Environ. Sci. http://dx.doi.org/10.1155/2014/201819.

Giani, M., Cicero, A.M., Savelli, F., Bruno, M., Donati, G., Farina, A., Veschetti, E.,Volterra, L., 1992. Marine snow in the Adriatic Sea: a multifactorial study. Sci.Total Environ. (Suppl. l), 539e549.

Giani, M., Djakovac, T., Degobbis, D., Cozzi, S., Solidoro, C., Fonda Umani, S., 2012.Recent changes in the marine ecosystems of the northern Adriatic Sea. Estuar.Coast. Shelf Sci. 115, 1e13. http://dx.doi.org/10.1016/j.ecss.2012.08.023.

Grasshoff, K., Kremling, K., Ehrhardt, M., 1999. Methods of Seawater Analysis, thirded. Wiley-VCH, Weinheim, p. 600.

Ho, D.T., Law, C.S., Smith, M.J., Schlosser, P., Harvey, M., Hill, P., 2006. Measurementsof air-sea gas exchange at high wind speeds in the Southern Ocean: implica-tions for global parameterization. Geophys. Res. Lett. 31, L16611.

Hydes, D.J., Loucaides, S., Tyrrell, T., 2010. Report on a Desk Study to Identify LikelySources of Error in the Measurements of Carbonate System Parameters andRelated Calculations, Particularly with Respect to Coastal Waters and OceanAcidification Experiments. National Oceanography Center Southampton,Southampton (UK) (Report No.: 83).

IPCC, 2013. Climate Change 2013: the Physical Science Basis.Kruskal, W.H., Wallis, W.A., 1952. Use of ranks in one-criterion variance analysis.

G. Ingrosso et al. / Estuarine, Coastal and Shelf Science 168 (2016) 58e7070

J. Am. Stat. Assoc. 47 (260), 583e621.Lee, K., Kim, T.W., Byrne, R.H., Millero, F.J., Feely, R.A., Liu, Y.M., 2010. The universal

ratio of boron to chlorinity for the North Pacific and North Atlantic oceans.Geochim. Cosmochim. Acta 74 (6), 1801e1811.

Le Qu�er�e, C., Raupach, M.R., Canadell, J.G., Marland, G., Bopp, L., Ciais, P., Conway, T.J.,Doney, S.C., Feely, R.A., Foster, P., Friedlingstein, P., Gurney, K., Houghton, R.A.,House, J.I., Huntingford, C., Levy, P.E., Lomas, M.R., Majkut, J., Metzl, N.,Ometto, J.P., Peters, G.P., Prentice, I.C., Randerson, J.T., Running, S.W.,Sarmiento, J.L., Schuster, U., Sitch, S., Takahashi, T., Viovy, N., van der Werf, G.R.,Woodward, F.I., 2009. Trends in the sources and sinks of carbon dioxide. Nat.Geosci. 2, 831e836. http://dx.doi.org/10.1038/ngeo689.

Le Qu�er�e, C., Takahashi, T., Buitenhuis, E.T., R€odenbeck, C., Sutherland, S.C., 2010.Impact of climate change and variability on the global oceanic sink of CO2. Glob.Biogeochem. Cycles 24 (4), GB4007.

Le Qu�er�e, C., Moriarty, R., Andrew, R.M., Peters, G.P., Ciais, P., Friedlingstein, P.,Jones, S.D., Sitch, S., Tans, P., Arneth, A., Boden, T.A., Bopp, L., Bozec, Y.,Canadell, J.G., Chini, L.P., Chevallier, F., Cosca, C.E., Harris, I., Hoppema, M.,Houghton, R.A., House, J.I., Jain, A.K., Johannessen, T., Kato, E., Keeling, R.F.,Kitidis, V., Klein Goldewijk, K., Koven, C., Landa, C.S., Landschützer, P., Lenton, A.,Lima, I.D., Marland, G., Mathis, J.T., Metzl, N., Nojiri, Y., Olsen, A., Ono, T., Peng, S.,Peters, W., Pfeil, B., Poulter, B., Raupach, M.R., Regnier, P., R€odenbeck, C., Saito, S.,Salisbury, J.E., Schuster, U., Schwinger, J., S�ef�erian, R., Segschneider, J.,Steinhoff, T., Stocker, B.D., Sutton, A.J., Takahashi, T., Tilbrook, B., van derWerf, G.R., Viovy, N., Wang, Y.P., Wanninkhof, R., Wiltshire, A., Zeng, N., 2015.Global carbon budget 2014. Earth Syst. Sci. Data 7, 47e85. http://dx.doi.org/10.5194/essd-7-47-2015.

Lewis, E., Wallace, D., 1998. Program Developed for CO2 System Calculation. ORNL/CDIAC-105, 38 pp.

Liu, X., Patsavas, M.C., Byrne, R.H., 2011. Purification and characterization of meta-cresol purple for spectrophotometric seawater pH measurements. Environ.Sci. Technol. 45, 4862e4868.

Luchetta, A., Cantoni, C., Catalano, G., 2010. New observations of CO2-inducedacidification in the northern Adriatic Sea over the last quarter century. Chem.Ecol. 26, 1e17.

Mala�ci�c, V., Petelin, B., 2001. Regional studies: Gulf of Trieste. In: Cushman-Roisin, B., Ga�ci�c, M., Poulain, P.-M., Artegiani, A. (Eds.), Physical Oceanography ofthe Adriatic Sea: Past, Present and Future. Kluwer Academic Publishers, Dor-drecht, pp. 167e181.

Mala�ci�c, V., Celio, M., �Cermelj, B., Bussani, A., Comici, C., 2006. Interannual evolutionof seasonal thermohaline properties in the Gulf of Trieste (northern Adriatic)1991-2003. J. Geophys. Res. 111, 1e16.

Malej, A., Mozeti�c, P., Mala�ci�c, V., Terzi�c, S., Ahel, M., 1995. Phytoplankton responsesto freshwater inputs in a small semi-enclosed gulf (Gulf of Trieste, Adriatic Sea).Mar. Ecol. Prog. Ser. 120, 111e121.

Marion, G.M., Millero, F.J., Feistel, R., 2009. Precipitation of solid phase calciumcarbonates and their effect on application of seawater SAeTeP models. OceanSci. 5, 285e291.

McKinley, G.A., Fay, A.R., Takahashi, T., Metzl, N., 2011. Convergence of atmosphericand North Atlantic carbon dioxide trends on multidecadal timescales. Nat.Geosci. 4 (9), 606e610. http://dx.doi.org/10.1038/Ngeo1193.

Mehrbach, C., Culberson, C.H., Hawley, J.E., Pytkowitcz, R.M., 1973. Measurements ofthe apparent dissociation constants of carbonic acid in seawater at atmosphericpressure. Limnol. Oceanogr. 18, 897e907.

Melzner, F., Thomsen, J., Koeve, W., Oschlies, A., Gutowska, M.A., Bange, H.W.,Hansen, H.P., K€ortzinger, A., 2012. Future ocean acidification will be amplifiedby hypoxia in coastal habitats. Mar. Biol. 160, 1e14. http://dx.doi.org/10.1007/s00227-012-1954-1.

Mihanovi�c, H., Vilibi�c, I., Carniel, S., Tudor, M., Russo, A., Bergamasco, A., Bubi�c, N.,Ljube�si�c, Z., Vili�ci�c, D., Boldrin, A., Mala�ci�c, V., Celio, M., Comici, C., Raicich, F.,2013. Exceptional dense water formation on the Adriatic shelf in the winter of2012. Ocean Sci. 9, 561e572.

Millero, F.J., 2010. Carbonate constants for estuarine waters. Mar. Freshw. Res. 61,139e142. http://dx.doi.org/10.1071/MF09254.

Morse, J.W., He, S., 1993. Influences of T, S and PCO2 on the pseudo-homogeneousprecipitation of CaCO3 from seawater: implications for whiting formation. Mar.Chem. 41, 291e297.

Morse, J.W., Arvidson, R.S., Lüttge, A., 2007. Calcium carbonate formation anddissolution. Chem. Rev. 107, 342e381.

Mozeti�c, P., Fonda Umani, S., Cataletto, B., Malej, A., 1998. Seasonal and inter-annualplankton variability in the Gulf of Trieste (northern Adriatic). J. Mar. Sci. 55 (4),711e722.

Mozeti�c, P., Solidoro, C., Cossarini, G., Socal, G., Precali, R., Franc�e, J., Bianchi, F., DeVittor, C., Smodlaka, N., Fonda Umani, S., 2010. Recent trends towards oligo-trophication of the Northern Adriatic: evidence from chlorophyll a time series.Estuaries Coasts 33, 362e375.

Nixon, S.W., 2009. Eutrophication and the macroscope. Hydrobiologia 629 (1),5e19.

Nomura, D., Yoshikawa-Inoue, H., Kobayashi, S., Nakaoka, S., Nakata, K., Hashida, G.,2014. Winter-to-summer evolution of pCO2 in surface water and airesea CO2

flux in the seasonal ice zone of the Southern Ocean. Biogeosciences 11,5749e5761. http://dx.doi.org/10.5194/bg-11-5749-2014.

Ott, J., 1992. The Adriatic benthos: problems and perspectives. In: Colombo, G.,Ferrari, I., Ceccherelli, V.U., Rossi, R. (Eds.), Marine Eutrophication and Popula-tion Dynamics. 25th European Marine Biology Symposium. Olsen & Olsen,Fredensborg, pp. 367e378.

Outdot, C., Gerard, R., Morin, P., 1988. Precise shipboard determination of dissolvedoxygen (Winkler procedure) for productivity studies with commercial system.Limnol. Oceanogr. 33, 146e150.

Pleni�car, M., Ogorelec, B., Novak, M., 2009. The Geology of Slovenia. Geolo�ski zavodSlovenije, Ljubljana, p. 612.

Querin, S., Crise, A., Deponte, D., Solidoro, C., 2007. Numerical study of the role ofwind forcing and freshwater buoyancy input on the circulation in a shallowembayment (Gulf of Trieste, northern Adriatic Sea). J. Geophys. Res. 112 (C3),1e19.

Raymond, P.A., Cole, J.J., 2003. Increase in the export of alkalinity from NorthAmerica's largest river. Science 301, 88e91.

Raymond, P.A., Oh, N.-H., Turner, R.E., Broussard, W., 2008. Anthropogenicallyenhanced fluxes of water and carbon from the Mississippi River. Nature 451,449e452.

Revelle, R., Suess, H.E., 1957. Carbon dioxide exchange between atmosphere andocean and the question of an increase of atmospheric CO2 during the pastdecades. Tellus 9, 18e27.

Sabine, C.L., Felly, R.A., Gruber, N., Key, R.M., Lee, K., Bullister, J.L., Wanninkhof, R.,Wong, C.S., Wallace, D.W.R., Tilbrook, B., Millero, F.J., Peng, T.-H., Kozyr, A.,Ono, T., Rios, A.F., 2004. The oceanic sink for anthropogenic CO2. Science 305,367e371.

Shapiro, S.S., Wilk, M.B., 1965. An analysis of variance test for normality (completesamples). Biometrika 591e611.

Siegel, S., Castellan Jr., N.J., 1988. Nonparametric Statistics for the Behavioral Sci-ences. McGraw-Hill Int., New York, pp. 213e214.

Simpson, J.H., 1997. Physical processes in the ROFI regime. J. Mar. Syst. 12 (1), 3e15.Socal, G., Acri, F., Bastianini, M., Bernardi Aubry, F., Bianchi, F., Cassin, D., Coppola, J.,

De Lazzari, A., Bandelj, V., Cossarini, G., Solidoro, C., 2008. Hydrological andbiogeochemical features of the northern Adriatic Sea in the period 2003-2006.Mar. Ecol. 29, 449e468.

Solidoro, C., Bastianini, M., Bandelj, V., Codermatz, R., Cossarini, G., Melaku Canu, D.,Ravagnan, E., Salon, S., Trevisani, S., 2009. Current state, scales of variability, andtrends of biogeochemical properties in the northern Adriatic Sea. J. Geophys.Res. 114, C07S91. http://dx.doi.org/10.1029/2008JC004838.

Suzuki, A., Kawahata, H., 2003. Carbon budget of coral reef systems: an overview ofobservations in fringing reefs, barrier reefs and atolls in the Indo-Pacific re-gions. Tellus B 55, 428e444.

Szramek, K., McIntosh, J.C., Williams, E.L., Kandu�c, T., Ogrinc, N., Walter, L.M., 2007.Relative weathering intensity of calcite versus dolomite in carbonate-bearingtemperate zone watersheds: carbonate geochemistry and fluxes from catch-ments within the St. Lawrence and Danube river basins. Geochem. Geophys.Geosyst. 8, 1e28.

Szramek, K., Walter, L.M., Kandu�c, T., Ogrinc, N., 2011. Dolomite versus calciteweathering in hydrogeochemically diverse watersheds established on beddedcarbonates (Sava and So�ca rivers, Slovenia). Aquat. Geochem. 17 (4e5),357e396.

Tam�se, S., Ogrinc, N., Walter, L.M., Turk, D., Faganeli, J., 2014. River sources of dis-solved inorganic carbon in the Gulf of Trieste (N Adriatic): stable carbon isotopeevidence. Estuaries Coasts. http://dx.doi.org/10.1007/s12237-014-9812-7.

Takahashi, T., Sutherland, S.C., Sweeney, C., Poisson, A., Metzl, N., Tilbrook, B.,Bates, N., Wanninkhof, R., Feely, R.A., Sabine, C., Olafsson, J., Nojiri, Y., 2002.Global sea-air CO2 flux based on climatological surface ocean pCO2, and sea-sonal biological and temperature effects. Deep Sea Res. Part II 49 (9e10),1601e1622.

Turk, D., Mala�ci�c, V., DeGrandpre, M.D., McGillis, W.R., 2010. Carbon dioxide vari-ability and airesea fluxes in the northern Adriatic Sea. J. Geophys. Res. 115,C10043. http://dx.doi.org/10.1029/2009JC006034.

Turk, D., Book, J.W., McGills, W.R., 2013. pCO2 and CO2 exchange during high borawinds in the Northern Adriatic. J. Mar. Syst. 117e118, 65e71.

Wanninkhof, R., 1992. Relationship between wind speed and gas exchange over theocean. J. Geophys. Res. 97 (C5), 7373e7382.

Wanninkhof, R., McGillis, W.R., 1999. A cubic relationship between air-sea CO2exchange and wind speed. Geophys. Res. Lett. 26 (13), 1889e1892.

Watanabe, Y.W., Nishioka, J., Shigemitsu, M., Mimura, A., Nakatsuka, T., 2009. In-fluence of riverine alkalinity on carbonate species in the Okhotsk Sea. Geophys.Res. Lett. 36, LI5606. http://dx.doi.org/10.1029/2009GL038520.

Weiss, R.F., 1974. Carbon dioxide in water and seawater: the solubility of a non-idealgas. Mar. Chem. 2 (3), 203e215.

Yao, W., Liu, X., Byrne, R.H., 2007. Impurities in indicators used for spectrophoto-metric seawater pH measurements: assessment and remedies. Mar. Chem. 107,167e172.

Zeebe, R.E., Wolf-Gladrow, D., 2001. CO2 in Seawater: Equilibrium, Kinetics, Iso-topes, first ed., vol. 65. Elsevier Science, p. 340.