Pressure, temperature, and salinity data collected during the winter of 2011/2012 by an Argo profiling float over the Adriatic shelf were used to document the dense water formation and subsequent bottom density current (BDC) normally occurring along the shelf slope. The float was advected to the Jabuka Pit and neighboring shallow area (<275 m) after October 2010. The parking depth was set to approximately 150 m, enabling the float to mostly follow the firstOnlineseabed between December 2011 and July 2012. The profiler measured strong spatial-temporal changes in the BDC thickness (from a few to about 50 m) and the bottom density (between 29.46 and 29.88 kg/m3). These observations show that an Argo float has the capability to observe a bottom density current and suggest that it would be possible to systematically use such floats to investigate these processes on coastal shelves.
 Originally developed for monitoring global processes [Roemmich and Argo Steering Team, 2009; Le Traon et al., 2009], the potential of the Argo program for use in regional studies was early recognized. The launching of the MedArgo program in 2004 [Poulain et al., 2007] with different parameters (e.g., setting the parking pressure to 350 dbar and using cycles of 5 days) introduced new opportunities to study a variety of basin-scale processes, including the tracking of intermediate [Emelianov et al., 2006] and deep water [Smith et al., 2008] generation and dynamics, and it significantly improved basin-scale forecasts [Nilsson et al., 2010]. The growth of the MedArgo program was apparent in the following years, reaching more than 30 active floats in 2012 in the Mediterranean and Black Seas. This growth largely increased the possibility to capture shelf and shelf-break processes by Argo measurements in a limited basin with complex topography, such as the Mediterranean.
 Indeed, an Argo profiling float drifting in the middle Adriatic (Figure 1) allowed for a quantification of shelf processes that followed the severely cold January and February of 2012. The float was drifting near the bottom during a dense water formation (DWF) episode and an associated bottom density current (BDC), in the area where the BDC normally occurs. The shelf DWF site, the source of the North Adriatic Dense Water (NAdDW), is located in the northernmost part of the wide Adriatic shelf, having depths less than 50 m, where strong bora cools the entire water column [Bergamasco et al., 1999]. After generation, the NAdDW flows as a BDC over the western part of the shelf and cascades toward the deep Adriatic and eastern Mediterranean areas [Robinson et al., 2003; Vilibić and Supić, 2005]. The dense water outflow is compensated by the advection of saltier Levantine Intermediate Water, which affects thermohaline properties and circulation in the middle and north Adriatic.
 Proper capturing of the BDC requires measurements in the first few meters above the bottom, and it is quite difficult to obtain measurements that close to the seabed. However, one Argo float was advected at the “right” place and at the “right” time to capture the BDC and associated processes in the middle Adriatic, thus introducing the potential of observing and quantifying BDC by intentionally repeating measurements with a number of floats. The analysis of the Argo data collected over the Adriatic shelf in 2011 and 2012 is presented in section 2. The use of Argo profiling floats to observe and quantify BDC for use over any shelf is discussed in section 3.
2 The Argo Float Trajectory and the Data
 We used the trajectory data and profiles collected by Argo float 1900848 (Arvor Iridium float model) from its crossing of the 200 m isobath and arrival at the Palagruža Sill in October 2010 (ascending profile 50) to the beginning of July 2012 (ascending profile 176), focusing on measurements obtained after November 2011. Figure 1 shows the float's trajectory, where the red line denotes the position of the float during the preconditioning and generation phase of the NAdDW until its detection by the profiler (from ascending profile 132, 1 December 2011, to ascending profile 150, 29 February 2012). The NAdDW preconditioning and generation during the winter of 2011/2012 are documented by Mihanović et al. . The yellow line corresponds to the profiler position once embedded and flowing within the BDC (from ascending profile 155, 25 March 2012, to ascending profile 176, 8 July 2012).
 The default parking pressure of the float was originally set to 350 dbar. However, after being pushed across the 200 m isobath by the east Adriatic current [Orlić et al., 2006] toward the northwest, the Argo operators changed the parking pressure to 150 dbar from cycle 63 (the cycle started on 16 December 2010 and ended with an ascending profile on 21 December 2010). This change enabled the float to cross over the deepest part of the Palagruža Sill (between cycles 63 and 68) and to drift and profile in the middle Adriatic afterward. Only data that passed the initial quality check were used in the analyses; however, all the used data were visually inspected for eventual errors. Potential temperatures (PTs) and potential density anomalies (PDAs) were computed from pressure, temperature, and practical salinity data following TEOS-10 algorithms and using scripts available at http://www.teos-10.org.
 The Argo float regularly profiled every 5 days, except for a few individual profiles, which were either completely missing or had poor quality flags for a certain parameter. The only significant data gap occurred between 29 February (ascending profile 150) and 25 March 2012 (ascending profile 155), when the float was drifting in the relatively shallow northern Adriatic shelf to the northwest of the Jabuka Pit. The trajectory data were sampled every 3 h during the float's subsurface drift at the parking depth or along the bottom, when the bottom depth was shallower than the parking depth. It can be seen that the float was transported faster by deep currents during certain periods, while at other times, it was much slower, following the bathymetry. For example, the along-slope speed of the float reached 2 cm/s between ascending profiles 161 and 165, while the corresponding theoretical speed, computed from the reduced gravity g’, the bottom slope S, and the Coriolis parameter f as g’S/f [Nof, 1983], was between 4 and 6 cm/s. This indicates that the float was confined to the very bottom along the shelf, scratching the seabed and presumably measuring just a portion of the bottom current speed due to friction with the bottom. Information about the float touching the ground for a certain cycle is contained in the float's trajectory files, and it occurred frequently when it drifted in the shallow area.
 When the float flows in line with the seabed, it is reasonable to assume that temperature and salinity sensor problems can occur due to the effects of sediments and eventual mechanical failures. Unfortunately, only a few conductivity-temperature-depth (CTD) measurements were conducted in that part of the Adriatic in the first half of 2012. A CTD profile taken at the Jabuka Pit on 23 February 2012 found potential temperatures between 11.44 and 13.25 °C and a salinity of approximately 38.71, while the temperatures and salinities measured by the Argo float located 60 km to the west-northwest (ascending profile 148, 19 February 2012) were within the ranges of 10.96–12.79 °C and 38.68–38.73, respectively (Figure 2, see insets).
 However, interesting temperature-salinity (T-S) values were observed close to the bottom during cycle 148. The PT measured during the float's subsurface drift decreased suddenly during the cycle, a day before the related ascending profile. The temperature drop was from 12.6 to 10.1 °C, with an associated salinity decrease of approximately 0.1 and a PDA increase of approximately 0.4 kg/m3, reaching PDA values close to 29.8 kg/m3 (Figure 3, red arrows). The float was confined to the bottom at that time, capturing the BDC for the first time. The ascending profile on 19 February 2012 (the end of cycle 148) shows that the PT and salinity sharply dropped in the last 2–3 m, to less than 11 °C and 38.67, respectively (Figure 2, see corresponding insets). Such a sudden near-bottom drop was not observed during previous ascending profiles, and it represents the first measurement of the BDC vertical structure, as the density at the bottom surpassed the density above the current by approximately 0.32 kg/m3. Indeed, the January–February period in 2012 was very harsh and cold in the Adriatic, and the energy loss during the severe bora episode that lasted from 25 January to 14 February 2012 was modeled to be more than 0.5 GJ/m2 over most of the north and middle Adriatic [Mihanović et al., 2012]. This event resulted in the generation of the NAdDW and the associated BDC captured by CTD measurements at the Palagruža Sill and other transects in the middle and north Adriatic [Mihanović et al., 2012]. The cooling intensity is apparent in the PT temporal changes (Figure 2), where the temperature decreased throughout the water column from 14.6 °C (ascending profile 143, 25 January 2012) to 12.5 °C (ascending profile 147, 14 February 2012). Fortunately, the Argo float moved toward and along the western Adriatic shore in the following months (March to June 2012), exactly where the BDC normally has its maximum (Figure 1) [Artegiani and Salusti, 1987; Vilibić et al., 2004].
 A week after cycle 148, the PT restored its values to approximately 12 °C, while the float was advected less than 5 km to the northwest, still being near the seabed. The float apparently moved out of the BDC, which was presumably very shallow and narrow at the time of its first arrival. Later, in early April 2012 (cycle 157), the profiling float again entered the BDC, measuring the minimum bottom PT and maximum PDA of 9.54 °C and 29.88 kg/m3, respectively (Figure 3). Afterward, the float was positioned in the BDC (Figure 1), following the isobaths around the western perimeter of the Jabuka Pit. However, the float gradually sank to greater depths during its along-slope advance, suggesting the existence of a cross-slope BDC component. During most of this period, the float was confined to the seabed itself (documented by the float's trajectory files, which contain subsurface pressure-temperature-salinity (p-T-S) data and a flag indicating whether or not the float touched the ground during a particular cycle). However, that was not the case during vertical profiling, during data transmission at the surface, and in June and early July 2012, when the float was advected above the western Jabuka Pit depression (depths between 160 and 180 m).
 The large temporal variability of the NAdDW density current over the small area surveyed by the profiling float in March–May 2012 is apparent (Figure 2). As stated, the first detection of the cold and dense water by the float on 18–19 February 2012 indicated that the measured BDC was just a few meters thick (cycle 148 and corresponding ascending profile), while the maximum measured BDC thickness of approximately 50 m coincided with the maximum measured PDA and can be found in early April 2012 (cycle 157 and corresponding ascending profile). Unfortunately, no data whatsoever were available between ascending profiles 150 and 155, when a strong BDC may have been expected. Nevertheless, the measured BDC was again confined to the bottom in mid April 2012 (ascending profile 159) and grew in early May 2012 (ascending profile 164), indicating strong spatial and temporal changes of the BDC strength and thickness. A “conflict” of the much denser NAdDW and lighter ambient waters probably resulted in complex deep dynamics, with bottom eddies, filaments, and turbulence. These spatial-temporal changes could also be due to unsteadiness of dense water production and/or intermittent southward transport of the pool of dense waters trapped on the northern shelf. Afterward, in June and in early July 2012, the float profiled in deeper water around the western perimeter of the Jabuka Pit (Figure 1, ascending profiles 170–176), where a BDC with lower densities was measured (Figure 2, maximum PDA was approximately 29.6 kg/m3).
3 The Use of Profiling Floats in Bottom Dense Water Observation
 The Argo 1900848 profiling float was active and successful in measuring the BDC and the outflow of the dense water in the Adriatic. This success introduces the possibility of the systematic use of profiling floats according to the Argo philosophy of observing density currents and the cascading of dense water. This cascading occurs in a large number of places in the world ocean and in marginal and coastal seas [Ivanov et al., 2004], and it is quite important for the ventilation, oxygenation, and biogeochemistry of deep waters. No CTD measurements normally take place near the seabed, and only bottom-mounted instrumentation would be able to capture such a data set. However, the moorings are normally planned in advance, and it is not easy to deploy moored instruments rapidly in a desired area. However, a profiling float such as the Argo is a much cheaper and efficient solution, is easy to deploy and recover, and measures both vertical profiles and bottom water properties (if the parking pressure/depth is set up properly).
 Several other oceanographic platforms can be also used for measuring BDC very near the bottom using a Lagrangian approach. For example, gliders have been used to make seasonal evaluations of the DWF of southwestern Australia [Pattiaratchi et al., 2011]. Gliders could also be forced to measure close to the bottom; however, this would be highly risky because the cost of a glider is an order of magnitude higher than that of an Argo profiler. RAFOS profiling floats [Rossby et al., 1986; Gould, 2005], neutrally buoyant floats that follow isopycnal planes, could also be used for observations near the bottom, which is much closer to a true Lagrangian approach with respect to Argo isobaric floats (whose Lagrangian drift depends on interaction with the bottom). Additionally, the RAFOS floats could measure vertical movements of a density current in a Lagrangian manner, which may be important over some topographic corrugations (underwater canyons and bulges) with strong vertical movements of a BDC.
 In general, RAFOS floats are tracked acoustically, whereas Argo floats use satellite tracking [Gould, 2005]. Even for multi-cycle RAFOS floats (that surface occasionally to transmit the signal arrival times, e.g., MARVOR), the data are not available immediately, in contrast to profiling floats tracked by satellites. However, the acoustic tracking provides high-resolution trajectories, thus resolving small-scale processes (e.g., eddy motions and narrow boundary currents), while satellite-tracked floats give only the mean velocity during each subsurface drift. On the other hand, multi-cycle Argo floats measure T-S profiles on each cycle, providing valuable information on the BDC vertical structure (Figure 2).
 It has to be emphasized that Arvor Iridium floats (developed by the French Research Institute for Exploitation of the Sea) use improved float technology and Iridium satellite transmission (http://www.euro-argo.eu). They were developed specifically for marginal seas, and three profilers were deployed in the Mediterranean and the Adriatic in 2009–2010. These floats have significantly shorter surface time (approximately 30 min, enabling even 1 day cycles), and their mission parameters can be remotely modified using downlink capability of the satellite transmission. Shorter surfacing intervals mitigate negative effects of having standard Argo floats (6–12 h surface time) that surface too frequently [Getzlaff et al., 2006]. The downlink feature could even facilitate their use as isopycnal floats, since their parking density can be periodically calculated from CTD sensor and their buoyancy could be adjusted to the desired potential density [Shimizu et al., 2004].
 The BDC observed during the Adriatic winter of 2011/2012 was occasionally very thin, changing rapidly vertically over a few meters and horizontally over a few kilometers; this somewhat complicates the use of RAFOS floats for BDC tracking in marginal seas. Namely, to observe the BDC, a RAFOS float should be programmed to follow an isopycnal level positioned near the bottom, which would increase the probability of the float hitting the seabed due to spatial-temporal changes of the BDC. The Argo float 1900848 documented significant variations in the BDC, and the plume thickness varied substantially between successive ascending profiles (e.g., from 5 m in ascending profile 163 to 30 m in ascending profile 164, see Figure 2). The bottom density changed strongly as well (up to 0.4 kg/m3 over a few days, see PDA time series in Figure 3). However, RAFOS floats can be intentionally programmed at high-density levels and therefore used similarly to isobaric Argo float profilers. Still, both of them will measure only a fraction of the real BDC current speed, as we found that the speed estimated from profiling float positions was at least two to three times lower than the theoretical BDC speed computed by the model of Nof . The main reason for that is probably the friction between the profiler and the seabed.
 Therefore, a modification of profiling floats to the operating conditions is envisaged. A float may be placed inside a protecting frame similarly to sub-ice floaters [Toole et al., 2011], or mechanical filters could be placed around the sensors to prevent problems that can occur when the sensors contact the seabed. Also, a solution for minimizing the friction between the instrument and the seabed would be helpful if the real speed of a BDC is to be monitored [e.g., Prater and Rossby, 2005]. More rapid and more frequent vertical profiling may be set up, e.g., with a 1 day resolution or even higher, similarly to SEPTR moorings [Grandi et al., 2005], to capture the changes in the dense water flow and to assess the turbulence and mixing at the edge between the dense water plume and adjacent waters. High-resolution vertical sampling near the seabed is also envisaged. The inclusion of additional sensors, e.g., for oxygen, turbidity, or biogeochemistry, would allow for the proper assessment of different aspects of BDCs, including sedimentary transport, oxygenation and transport of nutrients, and inorganic and organic matter. A proper solution for adopting these sensors on profiling floats could follow the recently introduced bio-Argo approach [Johnson et al., 2009].
 The systematic use of floats in the observation of BDC and dense water cascading should include the following: (1) the monitoring of dense water formation in DWF areas through detection of the atmospheric conditions and preconditioning favorable for the DWF, which can be achieved by operational atmospheric and ocean forecasting modeling systems; (2) the deployment of a number of floats at critical points of the expected BDC before its arrival; and (3) setting of the parking depth or buoyant density to greater than the sea depth or maximum measured density at any location in the investigated region where the BDC is intended to be measured.
 Several operational met-ocean systems are available for the Adriatic and may be used for DWF and BDC monitoring, e.g., operational COSMO/ROMS one-way coupled system [Mihanović et al., 2012] or the Adriatic Sea Forecasting System (http://gnoo.bo.ingv.it/afs). Also, the shelf Adriatic DWF site and shelf DWF sites in marginal seas are generally close to the coast, which makes feasible an execution of rapid in situ campaigns just after presumed DWF events, to quantify the strength of a DWF. The latter was done for the Adriatic DWF observed in the winter of 2011/2012, with record-breaking densities (above 30.50 kg/m3) documented in the northernmost part of the Adriatic, just a few days after the event [Mihanović et al., 2012]. Moreover, the DWF itself was observed by operational moorings positioned in the area. Once the DFW event is detected and quantified, several floats could be deployed at the DWF site and at a few locations along the presumed BDC track. As the BDC occurs for up to 2 months after the DWF along the western Adriatic shelf [Vilibić et al., 2004], these locations may span from the western Jabuka Pit toward the northern Adriatic. Also, the deployment should be carried out immediately after the DWF event, which would allow for the detection of the first arrival of the BDC in a specific area. In our study, the BDC was observed for the first time about 2 weeks after the DWF (cf. ascending profile 148), about 150 km to the southeast of the major northern Adriatic DWF site. However, the best selection of deployment locations and optimal number of profilers should be investigated by numerical models applied to a DWF and BDC event, with incorporated synthetic profiling floats.
 The data from the ARGO profiling float 1900848 were collected and made freely available by the International Argo Program and the national programs that contribute to it (http://www.coriolis.eu.org; http://www.argo.ucsd.edu). The Argo Program is part of the Global Ocean Observing System. The Adriatic bathymetric chart was kindly provided by the Hydrographic Institute of the Republic of Croatia. We are grateful to Rade Garić (Institute for Marine and Coastal Research, Dubrovnik, Croatia) for sharing his knowledge and enthusiasm related to the Adriatic Argo floats. Anonymous reviewers gave precious comments which significantly improved the manuscript. The support for the study was received through the Ministry of Science, Education and Sports of the Republic of Croatia (grants 001-0013077-1122 and 119-1193086-3085).