Inorganic carbon distribution and CO2 fluxes in a large European estuary (Tagus, Portugal)

Ten field cruises were carried out in Tagus estuary from 1999 to 2007 to study the dynamics of the inorganic carbon system. Dissolved inorganic carbon (DIC) and total alkalinity (TA) increased with salinity. DIC and TA were generally conservative in the estuarine mixing zone (salinity > 10), while a complex distribution pattern was observed at the upper estuary. DIC values peaked 1786.9 ± 155.8 µmol kg−1 at that segment. Estimated annual mean fluxes of DIC were 0.27 Tg C yr−1 from the river to the estuary, and 0.37 Tg C yr−1 from here to the coastal area. The Tagus estuary was always CO2 supersaturated, with partial pressure of CO2 (pCO2) reaching 9160 µatm in the upper estuary. An average emission of 0.11 Tg C yr−1 was estimated from the estuary to the atmosphere, corresponding to 23% of exported DIC. Only 8% of the riverine DIC was ventilated. The non-conservative behaviour of CO2 parameters in the estuary segment under freshwater influence was attributed to alternations in the relevance of riverine/terrestrial runoff, photosynthesis, aerobic respiration, organic matter mineralization and CaCO3 precipitation/dissolution.

Variations of inorganic carbon system parameters. pH values in the marine section were rather constant (amplitude 0.21; Fig. 4A), while the riverine end-member pH varied significantly (amplitude 1.10; Fig. 4A). Seasonally, pH showed scattered patterns and variations were not the same for all sampling periods (Fig. 5A): in autumn/winter more acidic features were noticed upstream (7.13-7.68) and increased downstream (7.83-8.03); in spring/summer, most of the time, more basic values were present upstream than downstream.
TA and DIC values were always lower in the riverine section than in the marine area, ranging between 977 µmol kg −1 in winter and 2587 µmol kg −1 in summer. These parameters displayed less irregular values in the marine influenced area and varied, respectively, from 1469 µmol kg −1 in winter to 2885 µmol kg −1 in summer (Fig. 4B,C). A negative correlation was found between riverine TA and average freshwater discharges (r 2 = 0.482, p < 0.05). TA and DIC increased from winter to spring/summer (Fig. 3E,F), with values varying from 1401 to 3535 µmol kg −1 during the productive period and from 977 to 2884 µmol kg −1 during the non-productive period ( Table 1).
pCO 2 values were higher in the riverine section, reaching as high as 9160 µatm in February 2004 ( Fig. 4D; Table 1), a value ~24 times higher than pCO 2 of atmospheric equilibrium (383.9 µatm). Seasonally average values of pCO 2 remain rather stabilized from winter to spring/summer and increased in autumn (Fig. 3G). Values of pCO 2 dropped rapidly in the upper estuarine zone (Fig. 5D) at salinities below 10. pCO 2 values as high as 9160 µatm (Table 1) were accompanied by low pH (7.13) at salinities below 5. Downstream, pCO 2 in general dropped to values of 620 µatm at the estuary mouth (Fig. 5D). CO 2 evasion pattern. Table 2 shows CO 2 gas transfer velocities at the water-air interface, proposed by different authors and calculated for Tagus estuary, as well as wind speed and tidal current for each season. Daily wind velocity was random and variable, oscillating between 1.7 m s −1 and 3.9 m s −1 . The maximum tidal current was lower in summer (58 cm s −1 ) and reached 122 cm s −1 in autumn depending on the tidal regime, river discharges and bathymetry of the estuary. Two of the algorithms used to estimate gas transfer velocity (k), namely k B04 7 and k A09 8 , gave the highest values when compared with other algorithms. This was an expected outcome since both approaches are based on the same technique (floating dome) and took into account the water current velocity. The k OD58 9 algorithm led to minimum k values, since it considers water current alone. However, maximum values of k were obtained in spring for all the parameterizations associated with maximum wind speed conditions (Table 2).
When applying a sensitive analysis to the fluxes calculated for each of the various parameterizations, differences emerged varying from 8 to 30%. Hence, in order to minimize any substantial errors in the fluxes estimated due to the use of a single generic relationship, CO 2 fluxes were averaged for the proposed parameterizations (k C96 , k RC01 , k B04 and k A09 ). A seasonal pattern of CO 2 fluxes to the atmosphere was revealed in Tagus estuary, values increasing from 80.4 ± 87.7 mmol C m −2 d −1 in spring to 133.9 ± 89.3 mmol C m −2 d −1 in autumn (Table 2). Concerning spatial variability, the amplitude of fluxes variability was ~250 mmol C m −2 d −1 (calculated as the

Season
Sampling dates Q a (m 3 s −1 ) S T (°C) SPM (mg l −1 ) Chl a (mg m −1 ) DO (%) pH TA (µmol kg −1 ) DIC (µmol kg −1 ) pCO 2 (µatm)   Inorganic carbon balance. The following inorganic carbon fluxes were considered to establish the inorganic carbon balance in Tagus estuary: the riverine DIC input to the estuary, the estuarine DIC output to the adjacent coastal waters and the water-air CO 2 flux. Mixing curves of DIC for each sampling period (exception of September 1999 and May 2007) were well reproduced by polynomial equations (Table 3). DIC fluxes and advective export of DIC added during estuarine transport (Internal DIC Flux; Table 3) were estimated 10 , and the magnitudes and temporal variation of inorganic carbon fluxes examined.
Freshwater end-member DIC concentration (C 0 ; Table 3) did not vary much during all survey periods (1786.9 ± 155.8 µmol kg −1 ), except during an exceptionally intense river discharge in March 2001, and in June 2002, with low river flow and relatively high photosynthetic activity (Table 1). However, the estimates show that DIC riverine exports to the estuary varied from 16.6 × 10 6 mol C d −1 (summer 2001) to 179.2 × 10 6 mol C d −1 (winter 2001) (Table 3), with a mean annual flux of 0.27 Tg C yr −1 .
It was observed that during the estuarine transport, generated DIC (C S − C 0 ) was higher in winter 2001 and lower (sometimes zero) in the other seasons. These internal fluxes ranged from 0 to 167.9 × 10 6 mol C d −1 , being estimated a mean annual internal flux of 0.10 Tg C yr −1 (Table 3). Fluxes from the estuary to the adjacent coastal waters varied from 16.6 × 10 6 mol C d −1 (in summer 2001) to 347.1 × 10 6 mol C d −1 (in winter 2001), corresponding to an annual flux of 0.37 Tg C yr −1 (Table 3). Thus, 27% of DIC was generated within the estuary, while the exported DIC to the atmosphere was 0.11 Tg C yr −1 , which accounted for approx. 23% of the total DIC exported from the estuary.

Discussion
Tagus waters environmental properties were strongly affected by the river discharge and showed a seasonal variability typical of a mid-latitude system, with most parameter values increasing from winter to spring/summer. Generally, river water carried higher contents of particulate material (Fig. 2D), nutrients (Fig. 7C) and phytoplankton (pers. comm.), than estuarine water. A tendency of a decrease ocean-ward was observed in winter, although high levels of SPM were present during the productive period in the 20-30 salinity range. This predominance of estuarine suspensions in the central/lower estuary can be interpreted as the result of salt marsh particles mixing and/or water circulation in the southern estuary, rather than the occurrence of biogeochemical processes. SPM values are comparable to values recorded for the Scheldt and the Thames estuaries and much lower (~14 times) than the ones in the highly turbid Gironde 11 . pH values in the marine section were rather constant due to the seawater buffering capacity, while the riverine end-member pH varied largely being such variability attributed to runoff coupled to changes in biological activity. pH seasonal trends can be attributed to more runoff in winter and more intense primary productivity, in particular in spring/summer, upstream. Overall, pH values were higher in spring/summer and lower in autumn/ winter (Fig. 3D).
TA and DIC riverine values were similar to those reported for the uppermost sections of other estuaries [12][13][14][15] . To a certain extent, the seasonal variability of TA and DIC was related to the river discharge. Highest values occurred during the low flow period when higher salinity water dominates the estuary and the lowest during intense river discharges. TA and DIC plots against salinity (Fig. 5B,C) show values increasing downstream. These trends have also been recorded in other estuaries 13,14,[16][17][18][19] . The nearly-conservative behaviour can be explained by the counterbalance between sources and consumption of TA and DIC along the salinity gradient. Nevertheless  SCIEnTIfIC RepORts | 7: 7376 | DOI:10.1038/s41598-017-06758-z salinities below 5, and a range of 1000 to >6000 µatm at low-salinity areas in the Satilla and Altamaha Rivers 17 . As suggested by some authors 19,21,[23][24][25] , high pCO 2 values at riverine waters might be due to organic carbon mineralization in soils, river waters and sediments. Differences in inorganic carbon variability patterns in the two end-members indicate, beside the dilution effect, the complexity of interactions between input sources and processes acting along the estuary. Spatial trends and range of values for this estuary are also similar to those reported for other estuaries 4,5,14,18,19 .
Mixing curves are a commonly used approach for the interpretation of source/sink dynamics of estuarine constituents 10,13,26 . Assuming salinity as a satisfactory mixing indicator, the profiles of water properties as functions of salinity should be linear, if only mixing processes occur in an estuary. Thus, analysis of correlations between different parameters was used for identification of other processes in Tagus estuary.
As mentioned above, CO 2 parameters behaved conservatively in most of the sampling periods, with the system dominated by mixing processes. But comparing the TA/DIC salinity profiles with theoretical mixing lines drawn between the two end-members, TA/DIC anomalies came out in the upper estuary at salinities below 10 Actually, a non-conservative behaviour of TA and DIC in upper estuaries, mostly at salinities below 5, has been referred by some authors 27 .
In March 2001, an exceptional river discharge (1861 m 3 s −1 ; Table 1) occurred, probably not allowing enough time for biogeochemical reactions to occur in the estuary 28 . However, a mechanism involving increase of pCO 2 and decrease of pH and DO (Fig. 7A) has been hypothesized to be the aerobic respiration. Such mechanism would lead, nevertheless, to DIC increase and would have practically no effect on TA. Still, the simultaneous increase in both parameters (TA and DIC) was observed, pointing to a distinct process responsible for TA production, possibly calcium carbonate (CaCO 3 ) dissolution. To assess this possibility, CaCO 3 saturation state was calculated for calcite and aragonite (respectively Ω c and Ω a ), using the thermodynamic solubility products 29 . Calculations have shown that for this high flow period, the estuary was undersaturated with values of Ω c and Ω a varying from 0.01 to 1.02, eventually indicating CaCO 3 dissolution. This process has referred to as the generator of alkalinity in Loire and Godavari estuaries 27,30 .
In June 2002, a TA/DIC deviation from linearity was noticed in the central part of the estuary, at salinities approximately 20-30 (Fig. 5F,J). A production of TA of ca. 540 and of ca. 565 µmol kg −1 of DIC was estimated by comparing the theoretical mixing line to in situ values. The anthropogenic influence of Trancão River (Fig. 1) in that zone, with high loads of organic carbon and associated mineralization, may explain such increase in TA. Additionally, the concomitant decrease of pH and DO and increases of pCO 2 (Fig. 7C), as well as of SPM and nitrate, decrease at that respective range of salinities (Fig. 7B), supports the occurrence of organic carbon mineralization. The considerable amount of suspended matter attained in that region (Fig. 7C) may have favoured some carbonate dissolution as well. In fact, it has been reported the occurrence of dissolution of CaCO 3 and consequent generation of alkalinity in other estuaries, in zones of maximum turbidity, namely in Loire 27 and in the highly turbid Gironde 31 and Ems 32 .
A strong decrease of TA/DIC (430 and 640 µmol kg −1 , respectively for TA and DIC) was observed during February 2004 (Fig. 5G,K) in the extremely low range of salinity (from 0 to 1), accompanied by a relatively conservative (r 2 = 0.765, p < 0.05, n = 12) behaviour downstream. Simultaneously, pCO 2 attained extremely high values (9160 µatm), contrasting with the values obtained in other sampling surveys (hardly attaining 4500 µatm; Table 1). This might be explained by the efflux of CO 2 to the atmosphere in the very low salinities region, leading  ), water current contribution, water-air CO 2 fluxes and CO 2 emission. a calculated as the ratio k OD58 /(k OD58 + k C96 ) b averaged CO 2 water-air fluxes for the four parameterizations proposed (k C96 , k RC01 , k B04 and k A09 ) c surface area of 265 km 2 computed as the definite integral of bidimensional masks defined from latitude, longitude, and coastline information, using the ferret software.
to the consumption of TA/DIC. Moreover, Chl a values (just up to 2.6 mg m −3 ; Table 1) reflecting winter conditions, support the above conclusion, since photosynthetic carbon fixation very unlikely would justify TA/DIC consumption. At salinity 20 another decrease of TA/DIC (~270 µmol kg −1 ) occurred, which also might correspond to CO 2 degassing reflected in the decrease of pCO 2 from 920 µatm to 670 µatm.
In May 2006, TA/DIC increase at low salinities seemed to be due to a combination of processes. Even if considerable primary productivity was underway in that region, reflected in elevated Chl a values (up to 73.4 mg m −3 ; Table 1), low pCO 2 (487-650 µatm) and as well as relatively high pH (8.25; Table 1), consumption of TA/DIC did not occur, as expected for high phytoplankton biomass development. As no clear evidence of particulate organic matter mineralization was revealed, a possible explanation for production of DIC in that range of salinity could be mineralization of labile organic material (mainly glicids) produced by phytoplankton exudation and/or lysis. Thus, it means that photosynthetic activity might indirectly contribute to DIC increase, counteracting the expected decrease under conditions of relatively high productivity. Besides, another mechanism likely acting and leading to generation of alkalinity may have been dissolution of CaCO 3 , as by this period only slightly supersaturated conditions were present in the estuary.
It should be taken into account that the patterns of TA and DIC could also be related to other processes. As about one third of Tagus estuary surface is composed by intertidal areas, sediments are sites of organic matter degradation 33 . Moreover, salt marshes sediments store carbon and their pore waters enriched in DIC are transferred to the estuary waters by tidal pumping, also producing DIC. This mechanism has been reported to act in mangrove creek 34,35 as well. In addition, denitrification occurs in Tagus 36 , and it is known that such processes were responsible for TA increases in several estuaries 17,31,34,37 .
The predominance of CO 2 supersaturation conditions in Tagus estuary suggests that the system is dominated by heterotrophy 6 , which is backed by the O 2 data (pers. comm.). The highest pCO 2 values were noticed at the salinity ranges 0-5 and also 20-30 (Fig. 5D). Upstream pCO 2 was most likely originated from riverine waters probably having no time to degas. pCO 2 increase observed at higher salinities might be attributed to lateral transport of CO 2 from salt marshes mainly located at the southern margin of Tagus in the north margin near the polluted Trancão River.
Studies carried out in Tagus marshes in 2001/2002 revealed CO 2 supersaturation, and pCO 2 values varying from 877 to 3950 µatm (Oliveira, unpublished data). Besides, other studies suggest that in tidally flooded salt marshes of some estuaries of Georgia/U.S.A, the CO 2 supersaturation was controlled by inputs from organic carbon respiration 17,38 . Also, CO 2 supersaturation of three Georgia estuaries were due to CO 2 inputs from the intertidal marshes and the rivers 39 . Thus, within Tagus estuary, mechanisms such as mixing of supersaturated freshwater with seawater, CO 2 efflux to the atmosphere and carbonate dissolution seem to emerge as the major regulators of pCO 2 spatial distribution.
As mentioned in another study 6 , higher CO 2 emissions were generally related to more intense wind speed, although no statistical significance was observed (p > 0.05). Tagus estuary water current was found to largely affect the gas transfer velocity over it (~48% on average) ( Table 2), an effect estimated by applying the ratio k OD58 / (k OD58 + k C96 ).
The value is rather high when compared with estimates for other European estuaries, such as the Scheldt (25%) 40 and Guadalquivir (30%) 41 . Assuming that water current and wind speed have an additive effect on k 7 , the water current expression proposed by O'Connor and Dobbins 9 was combined with the algorithms from Carini et al. 42 and Raymond and Cole 43 . The outcome of the respective additive effect resulted in k' C96 = k C96 + k OD58 and k' RC01 = k RC01 + k OD58 . Also, the possibility of a fetch effect on k has been investigated. Since the suspended matter levels present in the estuary were low to moderate (<165 mg l −1 ; Table 1) it was assumed that turbidity might have a rather limited effect on k.
Tagus estuary was always CO 2 supersaturated inducing effluxes to the atmosphere at an average flux of 33.6 ± 29.7 mol C m −2 yr −1 leading to an average CO 2 emission of 0.11 Tg C yr −1 . Tagus CO 2 fluxes are similar to the ones estimated for some inner estuaries (32.1 mol C m −2 yr −1 ) and analogous to fluxes of other European estuaries like Sado/Portugal (31.1 mol C m −2 yr −1 ), Gironde/France (30.8 mol C m −2 yr −1 ) and Guadalquivir/ Spain (31.3 mol C m −2 yr −1 ) 4, 6 .
The nearly-conservative DIC distribution observed most of the time along the estuary implies that inputs and outputs are in equilibrium, meaning that the CO 2 flux to the atmosphere must be balanced by net CO 2 production in the water column or sediment. In fact, several researchers 13,20,38,44 referred the dynamic coupling between CO 2 generation and evasion being extremely fast in estuaries.  Table 3. Dissolved Inorganic Carbon (DIC) distributions and fluxes in Tagus estuary. Equations are polynomial equations used to fit the data from DIC versus salinity for each sampling transects. All equations have p < 0.05. C 0 and C S are, respectively, freshwater and seawater DIC concentrations. The flux of freshwater DIC is defined as (Q · C 0 ), whereas the flux of internal DIC is (Q · (C S − C 0 )) and the estuarine flux is (Q · C S ). Q represents the Tagus River flow. a DIC concentration where the polynomial equation intersects the y-intercept (or the concentration at zero salinity) b DIC concentration where the tangent at the marine end-member crosses the y-intercept Freshwater end-member DIC concentration did not vary much during all survey periods. Such near stable values suggest that the concentration of the inorganic carbon entering the estuary is, to a certain extent, independent of the season of the year and very likely independent of the Tagus discharges.
The relative contribution of riverine CO 2 to the overall CO 2 emission by the estuary (estimated after 45 ) was ~6% in spring and ~10% in autumn. In fact, these values are very close to those reported for the very productive Scheldt estuary (10%) 45 and for 11 European estuaries (median value 10%) 46 . Still, that proportion can be highly variable from one estuary to the other, for example for the Randers Fjord the riverine contribution reached a value of 50% 46 , for Guadalquivir estuary 30% 41 and for Rhine estuary ~300% 46 . Hence, most of the emission of CO 2 from Tagus estuary (~90%) was attributed to heterotrophic activity. Our study suggests that changes in the Tagus estuary carbonate parameters seemed to be consequence of rather more complex feature than just the result of simple mixing of freshwater and seawater. Riverine/terrestrial runoff and occurrence of specific biogeochemical processes as photosynthesis, aerobic respiration, organic matter mineralization (degradation) and CaCO 3 precipitation/dissolution might be responsible for carbonate variability. Probably, the pore waters rich in DIC transferred to the estuary waters by tidal pumping may also contribute to the occurred changes. The inorganic carbon budget revealed that 0.27 Tg C enter into Tagus estuary annually, while 0.37 Tg C are exported to the adjacent coastal waters. Overall, 23% of the total DIC exported from Tagus estuary is emitted to the atmosphere, and about 8% of riverine DIC is ventilated. Most of the estuarine emissions (average flux of 33.6 ± 29.7 mol C C m −2 yr −1 ) were attributed to heterotrophy.

Methods
Geographic coverage. Tagus Estuary (Fig. 1) is located at the southwest Portugal (38°36′-39°N, 08°54′-09°24′W) and supports important human communities and natural resources. It is an inundated valley with a submerged area of 320 km 2 where ~40% are intertidal areas (~20 km 2 salt marsh vegetation and ~80 km 2 mudflats). This mesotidal estuary (tidal range 1-4 m) has a narrow and fault-controlled inlet channel separating two distinct regions: an outer wave-dominated area from an inner broad and tide-dominated part. Winds predominate from south and southwest during winter, rotating progressively to the northwest and north directions during spring, and maintaining these directions throughout the summer months. There is a strong seasonal hydrodynamic and biogeochemical variability due to seasonal fluctuations in meteorological conditions and river discharges (mean annual flow of 350 m 3 s −1 ). There is also a strong horizontal gradient inside the estuary as a result of the hydrodynamic conditions, mostly tidally-controlled, with a dominant semidiurnal period and maximum amplitude of 4.8 m in spring tide. Middle estuarine areas, and to a lesser extent upper areas, have more stable and homogenous conditions, displaying a higher residence time. Lower estuarine areas, influenced by the tidal regime, are characterized by high variability. The mean residence time of water in the estuary varies between 26 and 8 days 47 . The system is vertically well-mixed and has a mean tidal prism of 600 × 10 6 m 3 , about a third of the mean volume.
Even with recent reduction in organic loadings due to the implementation of considerable sewage treatment plants, the Tagus estuary is under intense anthropogenic disturbance, with a major population centre within its catchment basin (~2.3 million of inhabitants).
Sampling, analysis and calculation methods. The sampling period was from 1999 to 2007, covering an area extending from the estuary mouth to the point where freshwater was encountered. Sampling locations were selected to provide a full coverage of the salinity gradient (from 0 to 35). Surface seawater samples were collected at ebb conditions with Niskin bottles, for a total of 18 sites along a ~50 km stretch of the estuary (Fig. 1). Temperature (T) and salinity (S) parameters were determined in situ with a CTD (Conductivity -Temperature -Depth) Aanderaa probe. Salinity was calibrated with an AutoSal salinometer using IAPSO standard seawater, with a variation coefficient of 0.003%.
Dissolved oxygen (DO) was analysed following the Winkler method 48 using a whole-bottle manual titration. The coefficient of variation associated with the method ranged from 0.08 to 0.25%. pH was measured immediately after sample collection at 25 °C, using a Metrohm 704 pH-meter and a combination electrode (Metrohm) standardised against 2-amino-2-hydroxymethyl-1,3-propanediol seawater buffer (ionic strength of 0.7 M), at a precision of 0.005 pH units 49 . Total alkalinity (TA) samples were filtered through Whatman GF/F (0.7 μm) filters, fixed with HgCl 2 and stored (refrigerated not frozen) until use. Samples were then titrated automatically with HCl (~0.25 M HCl in a solution of 0.45 M NaCl) past the endpoint of 4.5 49 , with an accuracy of ±2 μmol kg −1 . The respectively accuracy was controlled against certified reference material supplied by A.G. Dickson (Scripps Institution of Oceanography, San Diego, USA).
Chlorophyll a (Chl a) was determined by filtering triplicate aliquots of ~150 ml water through Whatman GF/F filters (0.7 µm) under a 0.2 atm vacuum, and immediately frozen and later extracted in 90% acetone for analysis in a fluorometer Hitachi F-7000, calibrated with commercial solutions of Chl a (Sigma Chemical Co.). The variation coefficient was 1.8%. For suspended particulate matter (SPM) determinations, six aliquots of 100-1000 ml water samples were filtered through pre-combusted (2 h at 450 °C) Wathman GF/F filters and determined gravimetrically (drying at 70 °C). The respective filters were subsequently used for particulate organic (POC) and inorganic (PIC) carbon determinations using a CHN Fissons NA 1500 Analyser, with acetanilide as the calibration standard. System blanks were obtained by running several empty ashed tin capsules. The analyser provides a measure of total carbon, so the inorganic fraction was removed by drying the filters at 450 °C. The method precision was of 0.47%.
Meteorological data. Wind speed and direction were measured in situ with a Vaisala ® meteorological station (Datalogger Campbell Scientific CR510) coupled with a MetOne 034 A anemometer. Continuous measurements were acquired with 1-minute intervals at 11 m height. Wind speed was referenced to a height of 10 m (u 10 ) 50 . We assume one standard deviation of ± 2 m s −1 as wind speed error. Atmospheric CO 2 data was obtained from the Terceira Island's reference station (Azores, Portugal, 38.77°N 27.38°W), from the network of the National Oceanic and Atmospheric Administration (NOAA)/Climate Monitoring and Diagnostics Laboratory (CMDL)/Carbon Cycle Greenhouse Gases Group (CCGG) 51 . Some algorithms 49 were used to convert observed atmospheric CO 2 content in mole fraction (in dry air) to wet air values. Direct measurements of atmospheric CO 2 partial pressure made on-board was only available for some sampling periods. Significant correlations were found out between Terceira data and shipboard data (r 2 = 0.910, p < 0.05, n = 45) and the discrepancies lied just between 3 and 13 µatm. The impact of using Terceira data on this study was considered negligible.
SCIEnTIfIC RepORts | 7: 7376 | DOI:10.1038/s41598-017-06758-z Estimated parameters. pH values corrected to in situ temperature were calculated from total alkalinity (TA) and in situ pH and temperature 52 . For these calculations the carbon dioxide constants of Millero et al. 53 were applied.
The partial pressure of CO 2 in seawater (pCO 2 ) and the dissolved inorganic carbon (DIC) were calculated from the in situ temperature, TA and corrected pH, using the carbonic acid dissociation constants given by Millero et al. 53 and the CO 2 solubility coefficient of Weiss 54 . Errors associated with pCO 2 and DIC calculations were estimated to be ± 10 μatm and ± 5 μmol kg −1 , respectively (accumulated errors on TA and pH).
The water-air CO 2 fluxes (CO 2 Flux) were computed according to: where k is the CO 2 gas transfer velocity, K 0 the solubility coefficient of CO 2 and ΔpCO 2 the water-air gradient of pCO 2 . Positive CO 2 flux indicates emission of CO 2 from water to the atmosphere and negative flux the opposite direction. Some authors 7, 55, 56 suggest that k-wind speed relationships are site-specific. More recently, it was argued 8 that wind, water current, surface area and turbidity, all significantly affect k in estuaries. Various measurement techniques on the gas transfer velocity in estuaries have been implemented, using methods such as dual tracer addition 42 , natural gas tracer 57 or floating dome technique 7 . The tracer methods involve long term measurement of k over the entire estuary, while the floating dome technique is a short term measurement affected by the system heterogeneity that is typical in estuaries. In any case, the selection of a particular value for k will affect the overall representation of the net ecosystem metabolism. Since k was not determined in situ in Tagus estuary, we based our calculations to bracket the most likely value for k on parameterizations used on similar studies: (1) from a SF6 release experiment in the Parker River and estuary 42 (hereafter referred to as C96), (2) from a compilation of published k values in various rivers and estuaries, using different methodologies 43 (hereafter referred to as RC01), (3) considering the contribution of the water current 7, 9 (hereafter referred to as B04), and (4) based on a generic equation that gives k as a function of water current velocity, wind speed, estuarine surface area and suspended matter content 8 (hereafter referred to as A09). The choice of these formulations was motivated by the relative similarity between the Tagus estuary and the systems studied by the mentioned studies, in their physical characteristics (e.g., shallow, well-mixed, influenced by tides). Water current and tidal height data at Tagus estuary were obtained with hindcast simulations using the MOHID Modelling System (www.mohid.com), based on real forcing for river discharge, tide and wind 58-60 . Estimated DIC fluxes. The internal flux of dissolved constituent was estimated based on mixing curves 10 , by quantifying how much DIC was added by net heterotrophy during estuarine transport through DIC versus salinity plots. Whenever the distribution of a dissolved constituent is continuous and predictable using simple polynomial equations, C 0 is where the polynomial equation defining DIC concentrations intersects the y-intercept (or the concentration at zero salinity), and C S is the concentration of the constituent where the tangent at the seawater end-member crosses the y-intercept. Still, total exported flux from the estuary is given by (Q · C S ), where Q is the freshwater flow, (Q · (C S − C 0 )) the internal flux, and (Q · C 0 ) the flux from the freshwater end-member.
Statistical analysis. Exploratory analysis and statistical procedures were implemented using the statistical software Statistica 6.0 ® (Statsoft Inc., 2001). Differences between sampling periods in the measured/calculated physical-chemical and biological parameters, were assessed using an analysis of variance (ANOVA), and differences between means have been considered statistically significant for p < 0.05. The dominant processes influencing surface water chemistry were identified using linear correlations between the system parameters.