The Adriatic Sea can be divided into three distinct sub-basins on the basis of their oceanographic characteristics and bathymetric features (Fig. 1) . The northern part is a shallow shelf area with a gently sloping bottom down to the 100 m isobath at its southern limit. The central part encounters the Middle Adriatic Pit; a circular 270 m deep area delimited by the 170 m deep Palagruza Sill to the south. Finally, the South Adriatic Pit is roughly of a circular shape and the deepest portion of the Adriatic Sea (maximum depth about 1200 m). It is delimited by the 750 m deep Sill of Otranto in the south and by narrow smooth shelves along the coasts.
The freshwater discharge in the Adriatic is rather high (up to 5700 m³/sec), and more than 50% of this discharge is concentrated in the northern shallow part.
The circulation in the Adriatic is forced by the longitudinal pressure gradient caused by the freshwater contributions to the northern shallow part as well as by the differential cooling/heating of the water column [1] during winter/summer (the heat losses/gains in the northern part of the Adriatic are larger than in the Central and South Adriatic). During the winter, dense water is formed in both the northern shelf area and in the South Adriatic Pit. The Northern Adriatic shelf area is a source of two water masses, one relatively fresh in the surface layer due to the influence of the riverine inflow, and the other cold and dense formed in the winter, which occupies the bottom layer. The southward spreading of these two water masses, occurs in the form of a narrow swift surface coastal current along the Italian shore with a width of the order of 10 kilometers, and in the form of a bottom density driven current, respectively. The much larger compensating inflow occurs along the eastern coast.

Figure 1. Bathymetric map of the Adriatic Sea.
Depths are given in meters. CTD and biogeochemical sampling transects are marked by lines, while the mooring positions are marked by open circles.

This transversal asymmetry is possible since the width of the Adriatic Sea is about 200 km, i.e. an order of magnitude larger than the maximum value of the internal Rossby radius of deformation [2, 3].
One of the objectives of the project PRISMA1 was to assess longitudinal fluxes of water, and suspended and dissolved matter across four transects delimiting the three Adriatic sub-basins and in the Strait of Otranto - the communicating inlet between the Adriatic and Ionian Seas. In this paper, the estimated fluxes of water, and suspended and dissolved matter based on the measurements carried out during the period May 1995 - February 1996 for the summer (stratified water column) and winter (homogeneous water column) will be presented and discussed. Furthermore, the aim of the present research is to assess the importance of the Northern Adriatic eutrophic area in the functioning of the entire Adriatic as a unique physical and biogeochemical system. Since the coverage of the selected transects by measurements was not complete (the measurements were carried out up to the Croatian territorial waters), only estimates of the southward transport along the western shore could have been given. The paper is organized as follows: in section 2, a brief description of the experimental design is given; section 3 contains a short discussion of the vertical distributions of thermohaline properties, and of some chemical parameters like dissolved oxygen and nutrients, as well as chlorophyll a; in section 4, the vertical distributions of the longitudinal current components and their seasonal variability are described; section 5 gives the methodology of flux calculations and includes a discussion on the estimates of dissolved and suspended matter fluxes; finally, section 6 contains conclusions.
 

Four seasonal cruises with the R/V Urania were carried out in May, August, and October 1995 and February 1996. The spatial distribution of the sampling locations is given in the Fig. 1. The following parameters were sampled: temperature and salinity with a SeaBird CTD probe and dissolved oxygen, dissolved inorganic nitrogen (DIN), phosphates, silicates, chlorophyll a and total suspended matter (TSM) determined by standard chemical methods [4, 5]. Current measurements were carried out using moored current meters at selected locations (see Fig. 1 for instrument locations). Independently, but simultaneously, an intensive current measurement and hydrological sampling program was undertaken in the Strait of Otranto [6, 7]. In addition, a basin-wide survey with a ship-borne ADCP was carried out on a continuous basis from May 1995 until February 1996 with the R/V Thetis.
 
 

The phenomenological description of the spatio-temporal variability of the thermohaline and biogeochemical parameters will be focused on the comparison of the functioning of the shallow northern area influenced heavily by the freshwater discharge and vertical convection, with that of the deeper Adriatic areas which is mainly influenced by the horizontal advection of open-sea waters. In this context, attention will be mainly paid to the northernmost transect B located in an area 50 m deep and to transect H (see Fig. 1 for transect locations).

Figure 2a.Vertical distributions of temperature (C) for transect B.

Vertical distributions of the water temperature reveal a complete vertical homogenization of the water column along transect B only in winter (Fig. 2a, b and c). In the salinity field, the vertical homogenization is evident in all seasons excepting summer. A coastal layer is prominent in the salinity and density fields in all seasons. Its width, however varies in time reaching a minimum in winter when it does not exceed 10 km, due to the smaller horizontal length scales in this season (yearly minimum of the baroclinic radius of deformation).

Figure 2b. Vertical distributions of salinity for transect B.

Figure 2c. Vertical distributions of density (kg/m³) for transect B.

The dissolved oxygen content is minimum in autumn, while in winter the whole water column is well oxygenated due to vertical mixing (Fig. 3a). The oxygen content is rather high in spring as well, but this is probably a remnant of the winter situation when oxygen was high due to the strong vertical mixing and weak vertical stratification. Another interesting feature of the vertical oxygen distribution in this transect is the occurrence of hypoxic bubbles near the bottom that are evident in summer and autumn. These features are generated by remineralization processes associated with fast sedimentation of suspended biogenic particles, and their occurrence is also helped by a stable vertical stratification that prevents mixing and ventilation of the sub-pycnocline layers. Vertical distributions of DIN, phosphates and silicates (Fig. 3b, c and d) show that these hypoxic bubbles correspond with maximums of dissolved nutrient concentration that confirm the occurrence of the enhanced degradation of organic matter in these areas. The coastal layer shows a clear signal in the nutrient distributions only in the season when the biological activity is low and consequently when the surface layer does not appear depleted in nutrients, i.e. in winter. This layer is also evident in the nitrate distribution in autumn, when it is probably associated with a more intense freshwater input and the stronger terrestrial influence.
 
Figure 3a. Vertical distributions of dissolved oxygen (µmoles/dm³) for transect B.

Figure 3b. Vertical distributions of DIN (µmoles/dm³) for transect B.

Figure 3c. Vertical distributions of phosphate (µmoles/dm³) for transect B.

Figure 3d. Vertical distributions of silicate (µmoles/dm³) for transect B.

Transect E, with a maximum depth of about 120 meters, shows somewhat different behavior. Again, only in winter the temperature distribution is vertically homogeneous, while the prominent halocline is present only in summer (Fig. 4a and b). The winter density distribution (Fig. 4c) shows a column of a very high density water (> 29.3) situated on the continental slope that is probably associated with a dense water formation and spreading. The boundary layer of low salinity water evident in the density field as well, occurs in all seasons as on transect B, and it is again the narrowest in the winter season. The oxygen distribution (Fig. 5) is appreciably different along transect E when compared to that of transect B. In all seasons excepting winter, there appears an intermediate oxygen maximum associated with maximum of primary production. In winter, the overall oxygen content is maximum and vertically homogeneous. The low-oxygen bottom layer is much less pronounced along this transect, although it is still present and coincides again with the local maximums of nutrients. The coastal boundary layer is much less evident from the nutrient distribution (not shown here).
 

Figure 4a. Vertical distributions of temperature (C) for transect E.

Figure 4b. Vertical distributions of salinity for transect E.

Figure 4c. Vertical distributions of density (kg/m³) for transect E.

Figure 5. Vertical distributions of dissolved oxygen (µmoles/dm³) for transect E.

Deeper water areas (transect H with a maximum depth of about 230 m) display important influences of the horizontal advection processes in sub-pycnocline layers. Temperature and salinity transects (Fig. 6a and b) reveal a signal associated with the cold (T<12.5⁰ C) and relatively fresh (S<38.4) vein at the continental shelf break at a depth of about 100 m.

Figure 6a. Vertical distributions of temperature (C) for transect H.

Figure 6b. Vertical distributions of salinity for transect H.

Figure 6c. Vertical distributions of density (kg/m³) for transect H.:w

The vertical dissolved oxygen distribution (Fig. 7a) displays a distinct sub-surface maximum at a depth of about 25 m in all seasons excepting winter, that is associated with a Deep Chlorophyll Maximum (DCM) layer (Fig. 7e) situated slightly below the oxygen maximum layer (about 50 m). The DCM coincides with the nutricline which is the most prominent in autumn and summer.

Figure 7a. Vertical distributions of dissolved oxygen (µmoles/dm³) for transect H.

Figure 7b. Vertical distributions of DIN (µmoles/dm³) for transect H.

The sub-pycnocline pattern in the dissolved oxygen distribution is similar to that of the distributions of DIN and phosphate (Fig. 7b and c), but is different from the silicate distribution (Fig. 7d). The maximum concentrations of silicate in the deeper layers of transect H are the result of the advection of the diatom-dominated waters from the Northern Adriatic.

Figure 7c. Vertical distributions of phosphate (µmoles/dm³) for transect H.

Figure 7d. Vertical distributions of silicate (µmoles/dm³) for transect H.

Figure 7e. Vertical distributions of chlorophyll a (µg/dm³) for transect H.

Along the Otranto transect in the vertical distributions of thermohaline properties and chemical parameters, the sub-pycnocline layer shows stronger evidence of the influence of the inflow of Levantine Intermediate Water (LIW), than that of the outflow of Adriatic waters, [6, 8]. The LIW core appears as the local temperature maximum of about 14 C (not shown here) and the maximum of salinity (S>38.75) at a depth of about 300 m.


Figure 8. Vertical distributions of salinity for transect O (Strait of Otranto).

The tongue of LIW is pressed against the eastern coast but reaches almost the western continental slope. The seasonal thermocline is absent only in winter while the halocline is present throughout the year. The coastal boundary layer is evident in the salinity field only in autumn and winter. In addition, in winter the coastal boundary layer is characterized by the presence of the less saline water that is at the same time colder than the adjacent off-shore waters. The surface nutrient-depleted layer is present in all seasons and the intermediate layer of the maximum nutrient concentrations and the minimum dissolved oxygen content coincides in this area with the LIW core. The exception are silicates for which the maximum layer occurs below the LIW depth [8].

The entire period of the ADCP measurements was divided into two time-segments; the stratified season (May - September, 1995), designated as "summer", and the season characterized by a vertically homogeneous water column (October, 1995 - February 1996) denoted henceforth as "winter", which were then analyzed separately (the winter and summer spatial coverage of the ADCP measurements are shown in Fig. 9).
The tidal signal was eliminated with the method developed by Candela et al. [9]. Afterwards, the detided current field was expressed as a sum of the steady component and a time-varying one. The latter one will be called the "residual" current field in the rest of the paper and contains the time variability on all time scales from the tidal periods up to the entire record length. Some details on the data analysis are presented in a separate paper [10]. Vertical transects of a longshore current (current component perpendicular to the transect) were reconstructed extrapolating towards the surface and the bottom, to account for the portions of the water column not covered by the ADCP measurements. In the surface layer, the extrapolation was performed assuming a constant current speed from the first measured data up to the sea surface. In the bottom layer, the vertical profile of the velocity was assumed logarithmic and no-slip conditions were imposed. Some intercomparison between the ADCP data and Eulerian current measurements from moored current meters were also carried out.
Figure 9. Ship-borne ADCP measurements for summer and winter. For the definitions of two seasons, see text.

Steady longshore current components for winter and summer for locations where moored current meter data were available were compared with the averaged low-pass moored currentmeter data. The results of the intercomparison were found to be satisfactory [11] considering the fact that the two measurement methods are completely independent, and also considering that the data processing and methods of calculations of the steady current field are very different.
The residual (time-dependent) current field reveals a strong small-scale variability in both time and space (Fig. 10). Typical horizontal scales are of the order of ten kilometers, while in the vertical during the stratified season, the typical scales are of the order of ten meters. During the winter season, these meso-scale structures extend from the surface to the bottom, i.e. the vertical shear is very small. The signal associated with the coastal boundary layer is clearly present in the steady current field in both summer and winter seasons but very rarely in the residual current field. In fact, the southward coastal current was noted in the residual current field only on one occasion subsequent to a strong Po River discharge pulse (Fig. 10a).


Figure 10. Vertical distributions of the residual current component (cm/s) perpendicular to the transect B for 22 September 1995 (a) and 20 October 1995 (b). The current is defined negative if directed out of the paper. The corresponding Po river discharge rate is depicted below (c). Vertical dashed line denotes the ADCP measurement data.

 

Comparison of summer and winter steady currents (Fig. 11a, b, c and d) reveal for all transects a more pronounced horizontal shear (cyclonic shear) in the upper thermocline layer during the winter than in summer. This is mainly due to the strengthening of the coastal boundary current during the winter. From the Strait of Otranto where there is a complete coverage by current measurements of the entire transect, an increase in both inflowing and outflowing surface and intermediate currents in the winter season is evident. On the other hand, the Adriatic Deep Water (ADW) outflow in the Strait of Otranto shows weaker variability from one season to the other. Typically, the entire water column along transects is characterized by the cyclonic shear decreasing from the surface to the bottom. However, along the transect E to the north of the Middle Adriatic Pit, the shear changes sign from the cyclonic in the surface layer to the anticyclonic in the bottom layer. This may be associated with the remnants of the anticyclonic gyre whose presence was evidenced over the Middle Adriatic Pit during one part of the year [12], that is in spring. It is also interesting to note that an appreciable part of the transect area is occupied by the inflowing current only along the transect B. Along transects E and H, major portions of the area, if not all of it, are occupied by the southward current. Since, with the ADCP surveys, only the minor eastern part of these transects remained uncovered, it means that along both these transects the recirculation is rather strong and can be associated with the bathymetric constraints at the Palagruza Sill and the strong bottom slope to the north of the Middle Adriatic Pit. This recirculation is noticeable especially well in the IR satellite imagery when the thermal contrast between the inflowing and outflowing water is very prominent.

Figure 11a.Vertical distributions of the steady current component (cm/s) perpendicular to transect B. The sign of the current component is the same as in Fig. 10.

Figure 11b.Vertical distributions of the steady current component (cm/s) perpendicular to transect E. The sign of the current component is the same as in Fig. 10.

Figure 11c.Vertical distributions of the steady current component (cm/s) perpendicular to transect H. The sign of the current component is the same as in Fig. 10.

Figure 11d.Vertical distributions of the steady current component (cm/s) perpendicular to transect O (Otranto Strait). The sign of the current component is the same as in Fig. 10.

Steady southward fluxes of water, and dissolved and particulate matter for the winter and summer seasons were estimated using the steady longshore current component and the average distributions of biogeochemical parameters. The mean "summer" distribution of biogeochemical parameters was defined as an average of the spring (May, 1995) and summer (August, 1995) cruise data, while the "winter" distribution was defined as an average of the autumn (October, 1995) and winter (February, 1996) cruise data. First, for each transect and for each data set, an interpolation on a regular grid was performed using the kriging method. A grid of 200 X 200 points was reconstructed. The interpolation procedure was carried out using the PC software SURFER 6.0. Subsequently, by averaging, the summer and winter interpolated fields were obtained. These average fields were then used for the calculations of water, dissolved and suspended matter fluxes. For what concerns the Strait of Otranto, fluxes were obtained re-calculating on a six-months time scale the results from Civitarese et al. [8].
 

  Analyzing these results (Tab. 1), it is important to remember that all of the transects except the one at Otranto are only partially covered with current measurements. Since the measurements were carried out starting from the western (Italian) coast, only the estimates of the southward branch of the basin-wide circulation can be obtained. Also, since the Adriatic Sea is a semiclosed basin, one has to remember that the net water transport at each individual transect has to be close to zero (or equal to the freshwater gain/loss through the free surface and coastal boundaries). Consequently, the estimates of the northward flow across individual transect should be smaller or at most equal to the southward transport. Indeed, this is true for transects E and H, but not for the transect B where the inflowing transport is almost three times larger than the outflow when assuming the constant current speed in the surface layer. This is probably due to the underestimated surface transports since from the thermal wind relationship in areas with maximum horizontal density gradients, we obtained a surface velocity of up to 15 cm/s while the constant velocity assumption resulted in surface currents of up to 5 cm/s. In addition, the relative importance of this surface layer increases in shallow water since it becomes an important fraction of the total transect area. Thus at the transect B, the assumption of the vertically constant current speed can introduce an appreciable error in estimates of the southward transport. Also, the missing part of the transect between the westernmost measurement point and the coast can appreciably contribute to the error in the transport estimates. Therefore, we have reconstructed the surface layer velocities at the transect B calculating the geostrophic shear from the horizontal density distribution as obtained from the hydrographic campaigns. Then the surface current was allowed to go to zero logarithmically from the westernmost measurement location towards the coast. After that, at each location the vertical logarithmic profile with no-slip conditions was assumed. The obtained water fluxes then are still unbalanced but the difference between the two is now much smaller than it was before with the assumption of the constant current speed in the surface layer. The obtained values of the fluxes reveal stronger longitudinal flows during winter than during summer. This difference is especially prominent along transect E since the re-circulation rate to the north of the Mid-Adriatic Pit in winter is almost one order of magnitude larger than in summer. On average, during both summer and winter, about 25% of the volume that enters the Adriatic through the Strait Of Otranto passes over the Palagruza Sill to the Middle and Northern Adriatic. This means that 75% of the water entering the Adriatic remains trapped in the South Adriatic Pit. As said earlier, the varying influence of the bottom topography under different stratification conditions on the re-circulation is most evident along transect E, whereas along transect H the re-circulation rate changes by only small amount from summer to winter. If we consider the volume of water passing through transect B as a contribution of the Northern Adriatic, it follows that it adds to the South Adriatic Pit a volume of water that is very small: during winter, it represents only 4% of the total water volume exchanged through the Strait of Otranto, while during summer its contribution is only 3% of the water exchanged with the Ionian Sea.
In Table 2, the estimates of biogeochemical southward fluxes through four investigated transects are summarized. Either in summer or in winter the fluxes increase from north to south, owing not only to an obvious increase in the water transport, but also to the intrinsic biogeochemical characteristics of the Adriatic sub-basins. The relative increase in the amount of the exchanged inorganic nutrients inside the Adriatic, that is excluding the Strait of Otranto transect, is maximum passing from the transect E to the transect H in summer.
 

This increase can be explained in terms of the relation of the total transect depth and the surface nutrient depleted layer thickness. The two are almost the same at transect B while at E and H the water depth is larger than the thickness of the surface nutrient depleted layer. In winter, the relative increase in the nutrient exchange rate is maximum passing from the transect B to transect E (two to five times), while the flux increase from the transects H to E is reduced. On the contrary, in summer the Chlorophyll a and TSM southward fluxes, after an increase from transect B to transect E, decrease by about 50% from E to H. This decrease is due to the presence of the South Adriatic water that recirculates across the transect H, and which is characterized by a phytoplankton biomass lower than in the Middle and Northern Adriatic areas. In winter, the southward increase in Chlorophyll a and TSM fluxes is reduced since the longitudinal phytoplankton biomass variations are not so pronounced as in summer. In any case, transect H in correspondence to the abrupt change of the bottom depth (from 170 to 1200 m) appears as a discontinuity zone separating the two areas of the Adriatic: the northern part, relatively shallow and productive, and the oligotrophic deep water areas of the South Adriatic Pit.

  Estimates of the southward water fluxes have indicated that the Northern Adriatic adds to the South Adriatic Pit only 3-4% of the total water volume exchanged through the Strait of Otranto. Also, it was shown that at the Palagruza Sill, the northern border of the South Adriatic Pit, about 75% of the water entering through the Strait of Otranto recirculates. Thus, the Palagruza Sill represents a discontinuity area separating the relatively shallow and productive northern shelf from the oligotrophic South Adriatic Pit.
Water and biogeochemical fluxes estimates at the Strait of Otranto, on the other hand, made it possible to assess the relative importance of the northern productive area in determining the overall biogeochemical physionomy of the Adriatic Sea as a unique water body. These calculations, in fact, show that the Adriatic Sea through the Strait of Otranto exports inorganic nutrients and imports particulate organic carbon and nitrogen. This means that the Adriatic Sea acts as a mineralization site in which the oligotrophy is more marked than in the basin with which exchanges the biogeochemicals. Furthermore, this suggests that the Northern Adriatic, even though being an organic matter source, does not have an important influence on the whole Adriatic biogeochemical physionomy.
In interpreting these results one should however bear in mind that they are obtained on the basis of a single year realization (1995/96), and on the other hand, it is well known that the year-to-year variations of the oceanographic conditions in the Adriatic are rather large [12, 13]. Consequently, the longitudinal fluxes, as well as the exchange with the Ionian Sea are subject to strong interannual variations. Thus, one should assess whether these conclusions are obtained for extreme conditions of the longitudinal exchange or they are representative of the average conditions. As mentioned earlier, the most important driving force of the longitudinal fluxes is the pressure gradient which is to a large extent dependent on the buoyancy input and, during the winter, on the magnitude of the surface heat losses and subsequent dense water formation. Thus, climatic conditions of the specific year have to be compared to the long-term climate variations.

Acknowledgments. This research was supported by the Italian Ministry of University, Scientific Research and Technology, the National Council of Research and the Central Institute for Applied Marine Research within the project PRISMA1 (Programma di RIcerca e Sperimentazione per il Mare Adriatico). The data set originated from:

The dataset was kindly put at our disposal by the Scientific Coordinator of the PRISMA1 project, R. Pagnotta. We express our thanks to S. Ghergo for the technical assistance in data recovery. Useful comments by T. Hopkins to an early draft of this paper were greatly appreciated. Finally, it is almost impossible to name all the technicians, oceanographic vessel crew members and scientists which give a great contribution in collecting and analysing this data set.