Distribution and mixing of intermediate water masses in the Channel of Sicily (Mediterranean Sea)

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 108, NO. C9, 8105, doi: /2002jc001647, 2003 Distribution and mixing of intermediate water masses in the Channel of Sicily (Mediterranean Sea) Daniele Iudicone, 1 Bruno Buongiorno Nardelli, and Rosalia Santoleri Istituto di Scienze dell Atmosfera e del Clima, Consiglio Nazionale dell Ricerche, Rome, Italy Salvatore Marullo Ente per le Nuove tecnologie l Energia e l Ambiente-Centro Ricerche Casaccia, Rome, Italy Received 18 September 2002; revised 27 June 2003; accepted 11 August 2003; published 30 September [1] The Channel of Sicily, which represents the only connection between the western and the eastern Mediterranean Sea, can be described as a three-layer system, where fresher water of Atlantic origin (Modified Atlantic Water) flows eastward in the upper layer, Levantine Intermediate Water (LIW) leaves the eastern basin at around m, and an outflow of Transitional Ionian Deep Water (TIDW) is identified below the LIW. Synoptic Mesoscale Plankton Experiment data, collected during four surveys in the Channel of Sicily, show that LIW and TIDW are subject to major modifications when crossing the eastern sill. A significant Bernoulli aspiration is found there, and the bottom water at the sill is composed of modified deep Ionian waters coming from >800 m. An analysis of the water mass composition proves that LIW mixes essentially with TIDW, and that the TIDW flow through the channel must be larger than the expected 0.5 Sv. The presence of a secondary circulation related to the bottom boundary layer is demonstrated to be the primary factor leading to the dilution and cooling of LIW. INDEX TERMS: 4243 Oceanography: General: Marginal and semienclosed seas; 4283 Oceanography: General: Water masses; 4536 Oceanography: Physical: Hydrography; 4568 Oceanography: Physical: Turbulence, diffusion, and mixing processes; KEYWORDS: hydrography, strait dynamics, mixing processes, Mediterranean Sea, water masses Citation: Iudicone, D., B. Buongiorno Nardelli, R. Santoleri, and S. Marullo, Distribution and mixing of intermediate water masses in the Channel of Sicily (Mediterranean Sea), J. Geophys. Res., 108(C9), 8105, doi: /2002jc001647, Introduction [2] In semienclosed seas, such as the Mediterranean, strait dynamics can play a fundamental role in determining the general circulation of the whole basin. The key role of the Channel of Sicily in the thermohaline circulation of the Mediterranean Sea has been recognized in several previous studies [i.e., Rohling and Bryden, 1992], but it is only with recent studies that the channel dynamics and its influence on the circulation of the basin have begun to be clarified [Astraldi et al., 1999; Rhein et al., 1999; Astraldi et al., 2001]. The channel can be considered an intermediate basin that connects the western and the eastern Mediterranean Sea. It has a trapezoidal shape; its western side (150 km wide) has a mean depth of approximately 200 m (Figure 1). Here, only two narrow sills reach a depth of 430 m (main western sill, Figure 1) and 365 m (secondary western sill, Figure 1), respectively. The eastern end of the channel is much wider, extending for more than 500 km, but its mean depth is approximately 200 m, with the exception of the sill between Medina Bank and the island of Malta (eastern sill, 1 Now at Laboratoire d Océanographie Dynamique et de Climatologie, UPMC/CNRS/IRD, Paris, France. Copyright 2003 by the American Geophysical Union /03/2002JC m, Figure 1) and the wide passage between Medina bank and the Tunisian shelf (southern passage, m, Figure 1). The channel interior is characterized by the presence of two deep trenches, one northwest of Malta (more than 1700 m deep) and the other immediately east of Linosa island (more than 1500 m deep). [3] The flow in the channel is often described as a threelayer system, where fresher water of Atlantic origin (Modified Atlantic Water (MAW)) enters the eastern Mediterranean in the upper layer, Levantine Intermediate Water (LIW) flows westward immediately below the MAW layer, and below the LIW an outflow toward the Tyrrhenian Sea of Transitional Ionian Deep Water (TIDW) is found [Sparnocchia et al., 1999; Astraldi et al., 2001]. MAW and LIW are separated by a transition layer whose thickness varies between 30 m during winter to 100 m in summer [Morel, 1972]. A clear interannual signal has been detected in the LIW layer at the strait of Sicily during the last decade [Astraldi et al., 2002]. During the period from 1989 to 1995, the thermohaline circulation in the eastern Mediterranean Sea experienced a substantial modification. The usual source of deep waters for the eastern basin, namely the Adriatic Sea, was substantially replaced by the Aegean Sea, where a denser and saltier deep water was produced (Cretan Dense Water, CDW) which gradually fed the eastern Mediterranean bottom waters [Roether et al., 1996, Lascaratos et al., 1999]. This phenomenon is known in literature as the Eastern PBE 6-1

2 PBE 6-2 IUDICONE ET AL.: MIXING IN THE CHANNEL OF SICILY Figure 1. Bottom topography of the Channel of Sicily and adjacent western Ionian Sea. Mediterranean Transient. Following this circulation change, the Ionian LIW was gradually diluted with the rising older ADW, and LIW properties inside the channel of Sicily changed, with a weakening of the salinity maximum in the early 1990s [Astraldi et al., 2002]. [4] Although there is some evidence of mixing processes involving intermediate water masses as they enter the Tyrrhenian Sea [Astraldi et al., 2001], very little is known about the driving mechanisms, the variability and the relative role of horizontal versus vertical mixing inside the channel and at its eastern entrance. The LIW in the Ionian Sea has always been found at a depth of m [e.g., Malanotte-Rizzoli et al., 1997]. Consequently, the LIW core is expected to flow westward without major modifications, as in the Ionian Sea it is found at a depth shallower than that of the eastern Sicilian sills (Figure 1). In spite of this, inside the channel of Sicily intermediate waters are fresher than in the Ionian Sea, with salinity values of versus [Warn-Varnas et al., 1999]. [5] Mixing in straits is often associated with strong tidal signals [e.g., Wesson and Gregg, 1994]. In the Sicilian case, moored currentmeter data collected in the eastern sill from April to July 1998 revealed that tidal contributions to the bottom current never exceeded a few cm/s (with maximum values of 5 cm/s), but the mean value recorded, 13.4 cm/s, is indeed important. In addition, significant variations have been found on weekly timescales, leading to extreme cases in which the westward velocity vanished completely, and cases in which a velocity of more than 30 cm/s was measured (G. Budillon, personal communication, 2001). Turbulent mixing associated with high shear zones, which various studies propose to be the dominant mixing process [Warn-Varnas et al., 1999; Lermusiaux and Robinson, 2001], has not been shown to be the only (or ever dominant) mixing process. Given the complex topography, other mixing mechanisms could provide potential, such as the presence of a sloping bottom boundary layer [Garrett et al., 1993; Stahr and Sanford, 1999], a significant secondary circulation [Johnson and Ohlsen, 1994], or the entrainment of the deeper water outflow [Price and Baringer, 1994]. [6] The mixing of intermediate water masses in the Channel of Sicily thus represents an important and almost completely ignored process. The only previous study [Astraldi et al., 2001], focused primarily on the mixing and entrainment of the deep vein of TIDW entering the western Mediterranean basin. Their data were limited almost exclusively to the western sill (Strait of Sicily) and the downstream area in the southern Tyrrhenian Sea. Conversely, CTD data collected during Synoptic Mesoscale Plankton Experiment (SYMPLEX) represents the first hydrographic data set providing extensive coverage of the eastern sills and the adjacent regions. In this paper, a suitable combination of classical and recent methodologies was applied to SYMPLEX data with the objectives of characterizing the patterns and properties of intermediate waters, evidencing areas of active mixing, and discussing the possible mechanisms that drive mixing along the flow. [7] The paper is organized as follows: after a brief description of SYMPLEX surveys in section 2, the water mass properties at intermediate and deep levels inside the Channel of Sicily and in the western Ionian Sea are analyzed in section 3. In section 4 the main modifications of intermediate and deep waters along the channel are described, while section 5 is devoted to the description of the horizontal variability of the general circulation and water mass distribution during SYMPLEX. The mixing processes observed at the eastern entrances of the Sicily channel are discussed in section 6. Finally, a summary and

3 IUDICONE ET AL.: MIXING IN THE CHANNEL OF SICILY PBE 6-3 Figure 2. Locations of CTD measurements during (a) SYMPLEX-1, (b) SYMPLEX-2, (c) SYMPLEX-3, and (d) SYMPLEX-4. Green dots refer to CTD stations classified as Ionian, blue dots mark CTD collected inside the Channel of Sicily, and red dots identify measurements at the eastern sills, while measurements taken at depths shallower than 200 m (black dots) were not used in the q-s diagrams. discussion of the results and observations is presented in the last section. 2. Data Set [8] Four different surveys were conducted by R/V Urania in the framework of the SYMPLEX during 1996, 1997, 1998, and 1999: the SYMPLEX 1 (12 April to 13 May 1996), the SYMPLEX 2 (20 July to 11 August 1997), the SYMPLEX 3 (27 March to 20 April 1998) and the SYMPLEX 4 (21 October to 6 November 1999) and a wide and integrated set of hydrological, optical, chemical and biological data was collected. In this paper, the physical oceanographic data collected during SYMPLEX, consisting of 219, 203, 191, and 212 CTD (Conductivity Temperature Depth) profiles, are analyzed (Figure 2). In 1996 the survey covered the strait of Sicily and the entire channel interior, while only few data were collected east of the eastern sill and in the southern passage. In 1997 and 1998 the channel interior and the areas across the eastern sill were sampled more extensively, extending in the Ionian Sea to 16.5 E, while in 1998 and 1999 it was possible to collect a few stations near the southern passage as well, although no data at all were collected west of 14.4 E in All CTD measurements were acquired using a SBE911 Plus Sea Bird probe calibrated before and after the cruise. The CTD data were corrected basing on precruise and postcruise laboratory calibration coefficients calculated by Sea Bird. 3. Intermediate and Deepwater Properties and Modifications From the Western Ionian Sea to the Channel of Sicily [9] The first step in this analysis will be the classification of the different water masses present in the western Ionian

4 PBE 6-4 IUDICONE ET AL.: MIXING IN THE CHANNEL OF SICILY Figure 3. q-s diagrams of CTD data collected during (a) SYMPLEX-1, (b) SYMPLEX-2, and (c) SYMPLEX-3 and (d) SYMPLEX-4 (d). Colors refer to Figure 2. Sea and inside the channel of Sicily during the SYMPLEX surveys (Figure 2 and Figure 3). Several water masses of different origins have been identified. Starting from the deeper layers, an impressive growth and development of a salinity maximum in the Ionian bottom data (at 2700 m) was observed from 1996 to 1999 (Figure 3). This salinity maximum moved from in 1996 to in 1997 and reached values as high as in 1998, dropping back to in Correspondingly, the temperature maximum increased from C in 1996 to C in 1997 and reached a value of C in 1998, dropping back to C in In 1995 deep water of Aegean origin was invading part of the Ionian Sea at depths below 2000 m, characterized by salinity values between and and with its maximum around 3000 m [Roether et al., 1996]. The increase of salinity observed from 1996 to 1999 during SYMPLEX surveys can be attributed to the westward propagation of this dense CDW, which has now reached the western Ionian Sea. During 1998, the maximum of salinity measured at 2500 m in the western Ionian, at 16 E, was very close to the values observed in This fact supports the hypothesis that since 1998 the CDW replaced the ADW at the western end of the Ionian Sea. [10] On the other hand, inside the channel of Sicily, the deep layers are composed of TIDW, identified by the maximum of density and by salinity values around This water was characterized by a potential density of , during all the surveys, reaching up to in 1997, at a depth between 450 m and 550 m (values were obtained from single profiles, not shown), while in the deepest profiles inside the channel (near the deep trenches, below 700 m), the density was always found to remain below However, the maximum density observed at the sill, at 530 m, ranged from in 1996 to in In the Ionian Sea, the same density is generally found at much greater depths because of the aspiration associated to the acceleration of the LIW at the sill, known as Bernoulli suction. Following Stommel et al. [1973], it is possible to estimate the depth and the characteristics of the water being pushed up, once the velocity of one of the two layers at the sill, and the stratification at the sill and in a region at rest, are known. Assuming either an upper LIW layer velocity v 0 sill of 10 cm/s (this layer was delimited in the calculations by the isopycnal, corresponding to the maximum of salinity at the sill), as suggested by the observations made by Astraldi et al. [2001], or an average bottom current v sill of 13 cm/s, as recorded at the sill during a few months in 1998 (March August) by G. Budillon (personal communication, 2001), and choosing as the deep interior profile that inferred from the CTD station at N, E (which was sampled during all the surveys), substantial interannual variations are evidenced.

5 IUDICONE ET AL.: MIXING IN THE CHANNEL OF SICILY PBE 6-5 Table 1. Potential Density, Salinity, and Temperature for the Reference CTD Cast in the Ionian for SYMPLEX-1 (1996) a Depth, m s Ionian S Ionian T Ionian v 0 sill, cm/s v sill, cm/s a In the two columns on the right, v 0 sill is the LIW velocity at the sill that is required to bring the layer to the sill depth, assuming a 13 cm/s velocity for the bottom layer at the sill, and v sill is the velocity of the layer when reaching the sill bottom, assuming a 10 cm/s LIW velocity at the sill. [11] In 1996, TIDW was found at the sill with a density of The same density is found in the Ionian at 770 m, while Stommel-like calculations indicate that if an upper layer velocity of 10 cm/s is hypothesized, aspiration can push water up from 800 m, resulting in a slow (few cm/s, see Table 1) along channel bottom current at the sill. On the other hand, if we suppose a TIDW current of 13 cm/s at the sill, the aspiration could involve even deeper waters, depending on the effective LIW layer velocity (Table 1). [12] A slightly less dense TIDW was observed at the sill (530 m) in 1997, with a density of The corresponding isopycnal was found in the Ionian interior at 800 m, while a wider range of maximum aspiration depths results possible (Table 2). It is, however, noticeable that values of both TIDW and LIW velocities are coherent with those estimated by assuming that the aspiration involves the Ionian waters down to 1000 m. [13] In 1998, the TIDW was found at the sill (530 m) with a density of , while the same isopycnal was at a depth of 570 m in the Ionian Sea. This actually means that no Bernoulli effects are needed to explain the presence of this TIDW at the sill. [14] An intermediate condition was evidenced in 1999, but with an even lighter TIDW at the sill (s q ). In the Ionian Sea the same potential density is observed at 600 m, while Stommel-like calculations indicate that the suction could reasonably involve waters from 850 m (see Table 3). [15] Moving to the intermediate layers, the signal associated with the LIW is well identified in all the q-s diagrams at m (Figure 3). Its core is characterized in the Table 2. Potential Density, Salinity, and Temperature for the Reference CTD Cast in the Ionian for SYMPLEX-2 (1997) a Depth, m s Ionian S Ionian T Ionian v 0 sill, cm/s v sill, cm/s a In the two columns on the right, v 0 sill is the LIW velocity at the sill that is required to bring the layer to the sill depth, assuming a 13 cm/s velocity for the bottom layer at the sill, and v sill is the velocity of the layer when reaching the sill bottom, assuming a 10 cm/s LIW velocity at the sill. Table 3. Potential Density, Salinity, and Temperature for the Reference CTD Cast in the Ionian for SYMPLEX-4 (1999) a Depth, m s Ionian S Ionian T Ionian v 0 sill, cm/s v sill, cm/s a In the two columns on the right, v 0 sill is the LIW velocity at the sill that is required to bring the layer to the sill depth, assuming a 13 cm/s velocity for the bottom layer at the sill, and v sill is the velocity of the layer when reaching the sill bottom, assuming a 10 cm/s LIW velocity at the sill. Ionian Sea by a temperature of C in 1996, C in 1997 and 1998, and C in 1999 and by salinity values always in excess of Inside the channel the temperature of the LIW core reduces to less than 14.1 C in 1996, gradually decreasing to a value of C in 1999, while the salinity always remains steady at around [16] The layer between intermediate and surface waters presents some interesting interannual differences (Figure 3). In 1996 and 1997 this layer is substantially formed by the transitional waters resulting from the mixing of LIW with surface waters of both Atlantic and Ionian origin. Conversely, in 1998 a different water mass is clearly identified at 150 m, characterized by salinity and temperature values typical of the Adriatic Surface Water (ASW), i.e., S 38.2 and q C [e.g., Malanotte-Rizzoli et al., 1997]. This water is mainly found in the northern profiles in the western Ionian Sea at a depth of 140 m and could have reached that area as a consequence of the observed Ionian anticyclone weakening [International Commission for the Scientific Exploration of the Mediterranean Sea, 2000]. However, only a small part of it is found in the channel interior, probably conveyed by the AIS (Atlantic Ionian Stream) across the Malta plateau [Buongiorno Nardelli et al., 2001]. [17] However, in 1999 a warm and salty vein, with salinity values typical of Ionian LIW and a temperature around 14.4 C, was found at m along the Sicilian- Ionian escarpment, but no traces of this water are identified inside the channel. The circulation at this depth was strongly influenced by the AIS, flowing at that time from the Sicilian shelf directly toward the central Ionian Sea (see section 5, Figure 9). [18] LIW properties in the Ionian Sea were found to differ from those inside the channel of Sicily, indicating that some mixing process must occur along the LIW flow. In addition to this, q-s profiles of the eastern and southern sills (in red) relative to 1996 and 1997 clearly lie between the two characteristics curves for the Ionian Sea and the channel interior. In all the surveys the LIW core substantially freshens and cools while flowing to the western basin. [19] The percentage of each of the original waters from which the LIW found inside the channel of Sicily is formed, i.e., the Ionian deep and intermediate waters (TIDW Ionian and LIW Ionian ), and the transitional/fresh layers between MAW and LIW, as defined by Warn-Varnas et al. [1999] was also computed. The values representative for each water mass

6 PBE 6-6 IUDICONE ET AL.: MIXING IN THE CHANNEL OF SICILY Table 4. Reference Values for the Water Mass Analysis S T S T S T LIW channel LIW Ionian TIDW Ionian Transitional LIW/MAW ASW Ionian during the first 3 surveys are reported in Table 4, while it was not possible to do the same for 1999, as too few data were collected inside the channel. The diagrams in Figure 4 show that in 1996 the LIW inside the channel was composed of approximately 63% LIW Ionian, 28% deep water from the transitional layers in the Ionian (TIDW Ionian ), and less than 10% upper waters. Small differences respect to 1996 were observed in 1997 data, with 72% LIW Ionian, 22% TIDW Ionian Figure 4. The q-s diagrams of the (a) 1996, (b) 1997, and (c) and (d) 1999 cruise considering the ASW and MAW, respectively (see Viúdez et al. [1998] for another application of this method for the Mediterranean). See also Table 4.

7 IUDICONE ET AL.: MIXING IN THE CHANNEL OF SICILY PBE 6-7 Figure 5. Potential temperature, salinity, and potential density of salinity maximum along the channel axis (averaged every 1/3 of longitude) for SYMPLEX-1 (diamonds), SYMPLEX-2 (solid circles), and SYMPLEX-3 (squares). and 6% transitional MAW. In 1998 if we consider the mixing with the transitional MAW/LIW layer, the LIW channel consisted of 61% LIW Ionian, 35% TIDW Ionian, and 8% upper waters, while if we consider ASW as upper layer, the percentages change, with 75% LIW Ionian, 18% TIDW Ionian, and 7% ASW. [20] What is evident from the previous analysis is that LIW Ionian has always primarily mixed with deeper waters and only marginally with the upper layers. Moreover, a minimum TIDW inflow of 0.3 Sv in 1997, rising up to more than 0.5 Sv in 1996, is needed to explain the observed water mass percentages, assuming a LIW transport of 1Svatthe Sicily strait [Astraldi et al., 2001]. These estimates must be taken as a lower limit, as not all the TIDW Ionian entering the channel is diluted in the LIW Channel, while the TIDW inside the channel is generally composed of less than 20% of LIW Ionian. [21] This raises two important questions: where does mixing take place, and what mechanism is responsible for it? A first answer can be found by looking at the mean modification of the LIW core (marked by the maximum of salinity) along the flow. The salinity, potential temperature, and potential density of the salinity maximum, averaged every 0.4 of longitude, have been plotted in Figure 5. The cooling and densification of LIW core from 1996 to 1998 is clearly revealed by this plot, substantially in agreement to what was found by Astraldi et al. [2002]. In Figure 5 the major modification of LIW core characteristics is found around E (S 0.05 and q 0.15 C), i.e., exactly at the location of the eastern sills. At the same time, within the Sicily channel only a slight decrease of the salt content and temperature was measured. Therefore the first conclusion is that the most important mixing of intermediate waters between the eastern Mediterranean Sea and the western basin takes place at the eastern entrances of the channel of Sicily. Figure 5 also shows a regular decrease of the LIW temperature and salinity while it flows westward inside the channel. Therefore mixing activity must also be present inside the channel, and given the observed cooling, the mixing is most likely with TIDW. 4. Bottom Vein Mixing Inside the Channel of Sicily [22] As presented by Astraldi et al. [2001], the TIDW is thought to enter the eastern sill and then to descend into the bottom layers of the channel. This can thus be considered as an overflow, i.e., a bottom vein of dense water under the influence of gravity, friction and Coriolis acceleration, sliding down through a stratified background. The standard approach to the study of overflows is the streamtube model, recently extended by Price and Baringer [1994]. However,

8 PBE 6-8 IUDICONE ET AL.: MIXING IN THE CHANNEL OF SICILY Figure 6. Mean horizontal position of the bottom vein; bottom vein depth, thickness, temperature, salinity, and density along the channel. Please note the reverse axis on the abscissa. The zero corresponds to the eastern sill. because of the lack of velocity measurements along the channel, in the following we will present only an estimate of the main TIDW properties. [23] Assuming the isopycnal s ref = represents the upper boundary for TIDW in the eastern region of the channel [see Girton and Sanford, 2003; Astraldi et al., 2001], we can define for each transect the position of the centre of mass anomaly as X ¼ RR r 0 xdzdx RR r0 dzdx and the center of mass depth as Z ¼ RR r 0 zdzdx RR r0 dzdx where r 0 = r r 0 and r 0 = s ref The averaged thickness H and the tracers properties are computed, instead, as simple means. [24] The mean (X) position of the bottom vein just after the eastern sill (Figure 6a) shows large interannual variability, and it is only in 1996 and 1997 that the vein seems to follow the isobaths. It should be noted that in 1998 the highest density values, i.e., values higher than those at the eastern sill, were observed along the southern flank of the channel. This indicates that this water was coming from the southern entrance. This is the most probable cause for the shift of (X) values toward southwest, even if dynamical effects related to the upper layer velocity could also be taken in account [Astraldi et al., 2001]. Southwest of Malta the deepwater slopes down into the two deepest trenches located inside the channel, without any significant year-to-year variability. [25] The depth (Z) and the mean thickness (H) are presented in Figures 6b 6c. After rising and thinning at

9 IUDICONE ET AL.: MIXING IN THE CHANNEL OF SICILY PBE 6-9 the sill (as a consequence of the Bernoulli suction), the TIDW vein appears to follow the 500 m isobath down to the transect southwest of Malta, while gradually reaching an H value of about 100 m. Afterward, it starts to systematically deepen, and this corresponds to an abrupt change of the bottom slope (Figure 6a). Though experiencing an important change of depth, the layer thickness remains fairly constant in this region. In the following transects, the maximum depth is 1200 m, which is clearly due to the presence of some quiescent water resident in the deep trenches. The corresponding thickness also increases to m, while generally attaining a value of 100 m. [26] These observations suggest that within the channel the main entrainment of upper water occurs when the vein adjusts geostrophically on the slope, a few tens of kilometers downstream of the sill. The successive flow, even if changing depth, is not affected by turbulent mixing. The associated tracer values (Figures 6d, 6e, and 6f ) seem to confirm this hypothesis. [27] Tracer values are scattered at the sill and along the closer transect, but they assume remarkably steady values from year to year within the channel which acts as a reservoir for the TIDW, low-pass filtering the time variability of the inflow at the eastern sill. In addition, absolute values show a very weak spatial variability and the change in depth and in topographic cross-flow slope have no consequences for the tracer distribution. [28] A rough estimate of the entrainment velocity w e between the TIDW and the adjacent water masses inside the channel is given by w e ¼ VH ds s dx ; where s is the difference between the TIDW mean density and the LIW, assumed constant within the channel with values of respectively and 29.12, V is the mean velocity, H the thickness and x is the along channel distance [Girton and Sanford, 2003]. VH can be estimated by considering the transport as = VHL, where L is the current width. Taking s =.042 kg/m 3,ds/dx =10 8 kg/m 4, L = 100 km and = Sv, we obtain w e m/s. It is worth noting that if one assumes H = 100 m, the corresponding value for V (V = /LH) is 3 5 cm/s. [29] Along-flow mean properties have also been computed for an upper layer that includes both the LIW and the transitional LIW/TIDW, identified by the density range < s (Figure 7). Temperature and salinity are again observed to drop at the sill. The layer thickness starts to grow significantly at about 100 km from the sill, where the bottom becomes deeper. Inside the channel, temperature and salinity decrease linearly with distance from the sill. The decrease is smaller than the decrease of the tracer properties associated with the maximum of salinity (Figure 6). The salinity maximum is then eroded by mixing within this isopycnal layer. Nevertheless, assuming a transport of 1 Sv within the layer, we can extend the above analysis to estimate the w e associated to the TIDW entering the LIW layer from below. From the linear decrease of both temperature and salinity one arrives at an estimation of w e 10 5 m/s. [30] Even if errors in the estimates are indeed large, the difference in the w e values estimated from TIDW and LIW modifications (10 6 and 10 5 m/s, respectively) clearly indicate that LIW entrains more water with respect to TIDW, along the flow in the Sicily channel interior. Consequently, it can be hypothesized that the vertical exchange is mostly due to a detrainment of the TIDW layer, while turbulent entrainment of LIW into the TIDW is limited to only the upper part of the TIDW vein. Moreover, the use of a fixed isopycnal as an upper limit for TIDW can produce an underestimate of the entrainment [Dickson and Brown, 1994; Girton and Sanford, 2003]. [31] In the streamtube framework, cross-isobath transport is related to bottom friction, which allows the flow to break the dynamical constraints imposed by geostrophy. A very simple expression relates the descent dz/dx of the overflow to the magnitude of the combined bottom and interfacial stress t: dz dx ¼ t r 0 gh ; where g is the gravitational acceleration [Girton and Sanford, 2003]. When considering H = 100 m, dz/dx = (Figure 6b) and r 0 = 0.04 we obtain t = Pa, i.e., a rather small value, indicating that even with weak friction (i.e., no mixing with resident waters) the deep vein can deepen. One can estimate the interfacial stress as t i = r 0 V shear w e [Girton and Sanford, 2003], where V shear is the velocity difference between the two layers. In the following calculations, we have assumed V shear to be of the order of few (5 10) cm/s and used the two different estimates of w e (associated to the TIDW detrainment and LIW entrainment, respectively). In both cases, the interfacial stress t i remains below 10 3 Pa, indicating that a weak vertical exchange between LIW and TIDW is present and that the total stress driving the TIDW vein descent is given essentially by the bottom friction. 5. Intermediate Water Mass Distribution and Circulation [32] To gain further insight on where mixing occurs, the horizontal variability of the LIW has been inspected by analyzing the distribution of properties on isopycnal surfaces. Some variability is observed in both the values found on the isopycnal, 29.10, corresponding to the LIW core, and in the patterns characterizing the salt distribution on that surface (Figure 8). In particular, even if the analysis of all the maps suggests that the LIW core enters the channel from both the eastern sill and the southern passage, it is evident that in 1996, 1997, as well as in 1999 the LIW core flows inside the channel along the Sicilian continental shelf, while in 1998 the higher values of salinity are found along the Tunisian shelf. [33] Significant differences are found in the slope of the isopycnal between 1998 and the other years (Figure 9). In 1996, at the eastern sill, the isopycnal surface was deeper near Medina bank (up to 280 m) and shallower on the side of the Maltese continental shelf. Conversely, in the southern passage the surface was shallower near the Tunisian shelf. Moving westward, the slope remained

10 PBE 6-10 IUDICONE ET AL.: MIXING IN THE CHANNEL OF SICILY Figure 7. Intermediate layer depth, thickness, temperature, salinity, and density along the channel. Please note the reverse axis on the abscissa. The zero corresponds to the eastern sill. substantially unchanged along the continental shelves, defining a broad maximum of the isopycnal depth in the central part of the basin. In 1997, the same isopycnal slope was found at the eastern sill, but at that time, the surface actually shoaled while moving to the west. A completely different situation was observed in 1998, when the slope of the isopycnal at the sill was reversed with respect to the previous years. According to what was observed in the salinity maps, at that time the core of LIW flowed along Medina bank and not along the Malta and Sicilian shelf. Moreover, the surface exhibited a steeper slope along the whole channel, ranging between 220 m and more than 300 m. Again in 1999, in the same situation was found as in 1996 and 1997, although in 1999 the isopycnal was clearly deeper than during the previous surveys. It can be observed that the core of the LIW was always found at the shallower depths. Moreover, the LIW layer thickness (not shown) was more than 70 m on the Sicilian side and much lower (<40 m) in the southern part of the eastern sill, thus indicating that, despite the isopycnal slope, most of the LIW moves to the right while traveling toward the western basin, adjusting to the Coriolis force. The variability in the position of the main LIW flow could also possibly be influenced by the general circulation in the overlaying layers of the Ionian Sea. A high variability of the surface circulation has been observed during SYMPLEX, as evidenced by the dynamic heights (0/100 dbar) plotted in Figure 10. At the beginning of spring 1996, the Atlantic

11 IUDICONE ET AL.: MIXING IN THE CHANNEL OF SICILY PBE 6-11 Figure 8. Optimally interpolated map of salinity of isopycnal surface for (a) SYMPLEX-1, (b) SYMPLEX-2, (c) SYMPLEX-3, and (d) SYMPLEX-4 surveys. Crosses indicate the CTD measurements included by the mapping algorithm. Ionian Stream (AIS), flowing toward the southeast at the surface, was divided into two branches in the central part of the channel near 36.3 N, 13.6 E. The main branch continued its meandering flow along the Sicilian coast through the Malta channel, not involving the area of the sill, while a secondary branch deviated offshore toward the Tunisian side of the channel. Similarly, in summer 1997, the AIS was found to be more intense than during SYMPLEX 1996, with a broader and stronger main flow confined to the coast of Sicily, which turned to the northeast at the southern tip of Sicily (Capo Passero). In the spring of 1998, the main southeastward flow observed along the island of Sicily deviated to the southwest once it reached the channel of Malta, near Capo Passero, and then entered the Ionian Sea at the latitude of 35.2 N, i.e., just over the eastern sill, directed almost exactly eastward. A secondary and less intense branch was also observed to follow the Sicilian shelf, turning toward the northeast immediately after reaching the eastern tip of Sicily [Buongiorno Nardelli et al., 2001]. No information is available regarding the surface

12 PBE 6-12 IUDICONE ET AL.: MIXING IN THE CHANNEL OF SICILY Figure 9. Optimally interpolated map of the depth on isopycnal surface for (a) SYMPLEX-1, (b) SYMPLEX-2, (c) SYMPLEX-3, and (d) SYMPLEX-4 surveys. Crosses indicate the CTD measurements included by the mapping algorithm. circulation in autumn 1999 inside the channel; however, a very clear signal associated with the AIS was measured in the eastern part of the channel. The AIS detached from Capo Passero, crossing the Sicily Malta shelf and the eastern sill, directed southward, and then gradually turned to the east at the latitude of the Medina bank. 6. Mixing at the Eastern Sill [34] As we have seen, a variety of observations point to the eastern sills as the main site for intermediate water mixing and dilution of LIW flow toward the western Mediterranean basin. In this section we will thus investigate whether and where turbulent mixing was active during SYMPLEX surveys, in order to identify which processes might be responsible for it. Single CTD profiles have been used to estimate the vertical diffusion coefficient K v (intended as an estimate of K r ) through the Thorpe scale, L T, defined as the RMS vertical length scale of overturns. L T is computed by comparing the density profile with the same profile reordered to make it statically stable. By applying a simple algorithm, overturns are identified and, for each overturn, the RMS of the distance between the position of the water parcels in the original and in the

13 IUDICONE ET AL.: MIXING IN THE CHANNEL OF SICILY PBE 6-13 Figure 10. Dynamic heights (0/100 dbar) map computed from (a) SYMPLEX-1, (b) SYMPLEX-2, (c) SYMPLEX-3, and (d) SYMPLEX-4 surveys. Crosses indicate the CTD measurements used for computations. reordered profile is computed. The corresponding N (Brunt- Väisälä frequency) is computed with a linear fit on the reordered profile, using at least ten data, i.e., 10 m. A density difference (r = 0.001) was chosen as a threshold for computing K v. A visual inspection of each single profile ensured that in most cases the overturns where correctly identified and that they were not due to noise. Only in a few cases the overturn length was underestimated by the splitting of a single event into more than one. Thorpe length estimates using CTD data at 1-m resolution have already been computed by various authors and gives reasonable results even if, obviously, it is not possible to resolve lower values of K v. Details about the assumptions and limits of the derivation of K v from CTD measurements are given by Ferron et al. [1998] and Stansfield et al. [2001]. [35] In the following analysis, the bottom mixed layer (BML) was defined as the layer within s q = relative to the density of the bottom. We will consider as a secondary tracer the water transmittance, which is a function of the concentration and size of suspended particles. For the case of flows close to the bottom, these particles are suspended sediments and they are good tracers

14 PBE 6-14 IUDICONE ET AL.: MIXING IN THE CHANNEL OF SICILY Figure 11. the text. Map of the 1998 CTDs that are discussed in of mixing events since an energy source is required to redistribute them throughout the water column. The largest modification of LIW and TIDW properties were observed in the two transects presented here for years 1996 and 1998 (Figures 11 and 15). [36] In Figures 12 and 13, the vertical profiles of potential temperature (T), salinity (S) and transmittance (TR) at the eastern sill during SYMPLEX 1 (1996) are presented. As shown in Figure 12a, the two northernmost casts (126 and 127) are on the northern flank of the strait, along the slope. CTD 126 has a well developed BML, a few tens of meters thick, and high S values indicates that it is formed by the LIW core. CTD 127 (Figure 12b) shows a surprisingly homogeneous layer, well above the bottom, with high S values (LIW core). In addition, the associated TR and S values are similar to values found in the BML of CTD 126, suggesting that the mixing is due to a lateral propagation of an active BML from the northern slope. At the bottom, a thick BML (45 m) is also present, with S values still higher then those characteristic of TIDW. The S maximum at station 128 (Figure 12c), located downstream of a relatively flat area (see Figure 11), has a more standard (or diffusive) shape. An active bottom layer is still present. Layering and associated mixing are also observed in the subsurface MAW layer. CTD 129 (Figure 12d) and 130 (Figure 13a), corresponding to the deepest passage at the sill (Figure 11), display a thicker and weaker S maximum. Active mixing provide a likely explanation for this thickening. A TR minimum at m depth seems to indicate that the observed mixing between LIW and TIDW is related to a lateral transport of the bottom boundary layer on the northern slope of the deep channel. Closer to the bottom, a thin TIDW layer is observed, associated with high levels of turbulence (K v 10 3 ). CTD cast 131 (Figure 13b) is the closest to the southern slope, which is steeper than the northern one, and it is characterized by a 130-m thick homogeneous S layer, constituted by two superimposed extremely homogeneous layers with different T and density, overlying a thin BML. Large mixing activity is represented by two large overturns (more then 20-m thick) whose associated K v reaches 10 2 m 2 /s. [37] These observations show that no mixing was active between MAW and LIW during SYMPLEX-1, and suggest that the lateral and vertical interactions with the bottom topography are the main causes of mixing at the eastern sill in the Channel of Sicily. Consequences of this mixing (and eventually further current-bottom interactions) are expected in the downstream transect (Figures 11 and 14). The first profile along the second transect, CTD 164 (Figure 14a), located along the northern slope and roughly downstream of CTD 127, shows an active mixing in correspondence of the S maximum found in CTD 127. That maximum has been almost completely eroded while two superimposed thick and almost homogeneous layers are visible between 230 m and the bottom. More to the south, a thick weakly stratified layer of modified TIDW is observed (Figure 14b). Here, again, TR indicates the presence of lateral intrusions at about 30 m above the bottom. At station 162 (Figure 14c), on the southern slope and downstream of CTD 131 (Figure 11), a thick layer of modified LIW is observed at the same depth as the one observed in CTD 131. Similarly, a 53 m thick BML is found. [38] The 1998 data are also discussed in detail because they display a very different dynamical situation, with evidence of mixing at the interface between MAW and LIW. In April 1998, the AIS was flowing eastward exactly on top of the eastern sill, forming a steep front around 35.2 N (see section 5). The stratification was reduced and isopycnal surfaces down to outcropped. [39] Starting from the northernmost CTD 007 (Figures 15 and 16a), mixing activity is observed in the deepest 50 m. Here, a weak S maximum is found. At m depth, an intrusion of colder water is visible in correspondence of the maximum of N. Several small overturns are also observed with remarkable continuity, in the other two casts (008 and 009, Figures 16b and 16c) and in stations 011 and 012 (Figure 17), as well as on other southern casts (not shown). These overturns are of the same order of the CTD resolution and the Kv estimates are probably overestimated. At larger depths, a thin S maximum is found at station 008, just above a thick and nearly homogeneous layer (modified LIW) and above a BML with S and T values which lie between characteristic LIW and TIDW values. It is worth noting the strong similarity to what is found for CTD 163 (1996) (Figure 14). In the BML, TR shows a clear minimum. In the southern cast (009), as in 1996, mixing activity Figure 12. (opposite) Vertical profiles of temperature, salinity, and transmittance at the sill in Stations are (a) CTD 126, (b) CTD 127, (c) CTD 128, and (d) CTD 129. Thick vertical segments correspond to overturns (when signal-to-noise ratio is above the threshold, the associated K v value is also given; otherwise they are reported on the vertical axes on the left), thin broken lines are the K v 20-m average, the averaging done by considering only the values not equal zero. The vertical line identifies the isohaline , a proxy for LIW, while the horizontal line identifies the BML, as defined above.

15 IUDICONE ET AL.: MIXING IN THE CHANNEL OF SICILY PBE 6-15

16 PBE 6-16 IUDICONE ET AL.: MIXING IN THE CHANNEL OF SICILY Figure 13. Vertical profiles of temperature, salinity, and transmittance at the sill in Stations are (a) CTD 130 and (b) CTD 131. See Figure 12 for explanations. is associated with a transitional layer at about 400 m where, as in 1996, a minimum in TR is observed. At the bottom a thick BML is present. A thick layer at m, similar to that found in station 009, is present in the two deepest CTD profiles (010 and 011, Figures 16d and 17a) associated with a TR minimum. A thick BML (>70 m) in both stations is characterized by a very large overturn (h > 50 m), underestimated by the automatic detecting algorithm, that has split it in two because of the noise threshold. The large vertical extent of the inversions and the almost coincident depths (the distance between CTDs is about 7 km) suggests that the inversions originate from the same process, an intrusion of a lighter vein, yet well mixed, flowing below a denser one or an extremely large overturn. It must be noticed that, upstream of CTD 010 and 011, in the Ionian Sea an intense mixing activity and a TR minimum were identified at the same depth (not shown), i.e., about 400 m. This seems to suggest that one source of the mixing (and intrusions), is along the slope upstream of the sill (Figure 1). Here bottom currents due to Bernoulli aspiration from deep layers are strong and the convergence of isobaths is probably already accelerating the flow. [40] On the southern slope (CTD 012, Figure 17b), a thick layer characterized by a remarkably homogeneous S and linear T and TR is observed just above the bottom. S is approximately 38.77, corresponding to the value of LIW in the Strait downstream of the sill. The similarity with CTD 131 (1996) suggests that the southern slope is the site of recurrent mixing activity at intermediate depths. [41] Finally, as during 1996, in the downstream transect (Figure 18, here we report only the two deepest stations) the S maximum is eroded and it almost disappears in station 079. An active anomaly (and TR minimum) at 400 m is found within a thick transitional layer composed of both LIW and TIDW. [42] A slightly different situation is evidenced in 1997 and 1999 (not shown). In 1997, LIW appears less actively eroded at the sill. A thick transitional layer is still observed at 400 m but a significant BML (>90 m) is observed only on the northernmost of the two deepest stations, although they are a few kilometers apart. Similarly, in 1999 bottom mixing is present in the stations on the northern slope. A thick transitional layer is observed and a thick BML (80 m) is found only on the northernmost of the two deep stations, while almost no mixing is detected along the water column above. Shear-induced mixing in intermediate and upper layers was probably inhibited because of the higher stratification, as both SYMPLEX-2 and SYMPLEX-4 surveys were conducted during the stratified season (summer/ autumn respectively). It is worth noting that during those

17 IUDICONE ET AL.: MIXING IN THE CHANNEL OF SICILY PBE 6-17 Figure 14. Vertical profiles of temperature, salinity, and transmittance downstream the sill in Stations are (a) CTD 164, (b) CTD 163, (c) CTD 162, and (d) CTD 161. See Figure 12 for explanations.

18 PBE 6-18 IUDICONE ET AL.: MIXING IN THE CHANNEL OF SICILY Figure 15. the text. Map of the 1998 CTDs that are discussed in cruises the main MAW stream, which was flowing across the Sicilian shelf toward the Ionian Sea, was only slightly influencing the sill area (see Figure 10). [43] Summarizing, large modifications of the LIW and TIDW properties are observed during all the cruises. In particular, it has been observed that (see also the sketch in Figure 19): [44] 1) thick BMLs and a strong mixing at m occur on the northern slope, with some lateral intrusions of the mixing layer (e.g., 1996 CTD 127); [45] 2) the profiles in the central part of the strait, where the absolute S maximum is observed, are less perturbed. Here isobaths are too shallow to allow TIDW overflow (e.g., CTD 128 (1996)); [46] 3) active mixing and TR minima are observed at middepth in the deepest passage of the sill. Here a thick transitional layer between 300 and m, also present in the upstream stations, is always found to lie above a very thick BML, composed of largely modified TIDW (e.g., CTD 010 (1998)); [47] 4) on the steep southern flank, in 1996 and in 1998, one or more thick homogeneous layers of largely modified LIW are observed (e.g., CTD 131 (1996)); [48] 5) the LIW erosion is found to continue downstream, especially on the northern flank, while thick BMLs can be observed on the southern flank. 7. Discussion [49] Because of sloping boundaries, the circulation and dynamics in sea straits are quite complex. Cross-isobath Ekman transport in the bottom boundary layer (BBL) and at the interface may vertically redistribute buoyancy and strongly modify the original stratification [Garrett et al., 1993; Johnson and Ohlsen, 1994, Figure 5]. In particular, in the presence of both sloping boundaries, stratification and rotation, the Ekman layer can be arrested, i.e., a near-bottom thermal wind develops and the main flow that results is fundamentally frictionless. Depending on the slope of the lateral boundary with respect to the flow direction, i.e., if a downslope (or upslope) Ekman transport is found, the resulting BML thickness is enhanced (or reduced) [Garrett et al., 1993]. The presence of arrested boundary layers on the slopes can result in a convergence of fluid in the BBL and in the consequent BML detachments on the upwelling favorable flank [Seim et al., 1999]. Moreover, in the presence of shear flow, the resulting secondary circulation can be quite intense and complex. [50] In the upwelling favorable case, the time for the arrest to occur is given approximately by (C d N/f ) 1/2 S 1 f 1, where S = N 2 sin 2 q/f 2 is the Burger number based on the bottom slope angle q and C d is the drag coefficient, while for the downwelling favorable case the arrest time is approximated by 1/2(C d N) 1 S 3/2 [Garrett et al., 1993]. At the sill, the steep southern slope has an approximately uniform value, sinq = , while the northern slope has values of , and for the 400 m to bottom, m, and m layers, respectively. Taking N= s 1 for the southern flank and the northern 300 m to bottom layer, N = s 1 for the m layer and f = s 1, we obtain an arresting timescale of 1 day for the upwelling favorable case, and 9, and 38 days, respectively, for the downwelling favorable case. The upwelling case (southern slope) and the bottom and upper layer of the downwelling case (northern flank of the deep channel and northern flank of the strait) are then likely to get close to a steady state (i.e., to be arrested) within the surroundings of the sill. [51] The presence of both an arrested boundary layer and the observation of thick homogeneous layers on the southern slope appears to be contradictory. However, as discussed by Seim et al. [1999], the presence of an arrested boundary layer does not allow the southward (and upwelling) Ekman flow from the deepest layers to further upwell along the slope (see Figure 19). This convergence can create thick homogeneous layers, eventually detached from the bottom. Another cause for the detachment could be the secondary circulation along the MAW-LIW interface, which is expected to be directed toward the center of the channel [Johnson and Ohlsen, 1994, Figure 5]. [52] On the northern slopes, mixing is observed where the slopes are steep, at 250 and at 400 m, consistent with the hypothesis of a large transport along the downwelling side of the channel. Lateral intrusions are also observed upstream of the sill. Strong bottom currents associated with the Bernoulli suction are probably present there and a thick BML can be expected long before the entrance of the sill. The lateral intrusions (or detachments) of bottom layers could be due to the convergence of the isobaths, and consequently of the flow, as by Pickart [2000], or to the Figure 16. (opposite) Vertical profiles of temperature, salinity, and transmittance at the sill in Stations are (a) CTD 007, (b) CTD 008, (c) CTD 009, and (d) CTD 010. See Figure 12 for explanations. On the right, the vertical distribution of N is shown.