Observations and modeling of a pulsating density current

GEOPHYSICAL RESEARCH LETTERS, VOL. 35, L14603, doi:10.1029/2008GL034123, 2008

Observations and modeling of a pulsating density current E. Biton,1 J. Silverman,1 and H. Gildor1 Received 26 March 2008; revised 27 May 2008; accepted 18 June 2008; published 22 July 2008.

[1] We investigate density current formation and flow in the northern Gulf of Eilat, Red Sea, using in situ observations and a high resolution, nonhydrostatic general circulation model. These density currents occur in relatively warm water 20°C and are probably the warmest in the world ocean. We demonstrate the intrinsic nonlinearity of density currents by comparing two high-resolution simulations forced by different heat flux. This nonlinearity, which is poorly represented in global models, is shown here to affect the properties of simulated density currents. Such a bias in simulating density currents in global models may lead to significant biases on the larger scale global ocean circulation and water mass distribution. Citation: Biton, E., J. Silverman, and H. Gildor (2008), Observations and modeling of a pulsating density current, Geophys. Res. Lett., 35, L14603, doi:10.1029/ 2008GL034123.

1. Introduction [2] Density Currents (DCs) are considered to be an important component of the oceanic thermohaline circulation, affecting ocean climate and maintaining a stably stratified world ocean [Killworth, 1983; Hughes and Griffiths, 2006; Wahlin and Cenedese, 2006]. Near-boundary buoyancy loss from the ocean’s surface is one of the ways in which DCs form. Because of the larger surface area to depth ratio, water over the shallow continental shelf becomes denser relative to the adjacent deeper water and then starts flowing under the force of gravity down the continental slope [e.g., Symonds and Gardiner-Garden, 1994; Killworth, 1983; Baines and Condie, 1998; Ivanov et al., 2004; Gordon et al., 1993; Wells and Sherman, 2001]. At high latitudes these flows are associated with deep-water formation over continental shelves, such as the formation of the Antarctic Bottom Water, the most widespread water mass in the world oceans [Baines and Condie, 1998]. [3] At present the dynamics of density currents is not fully understood because observing them at high spatial and temporal resolution is very difficult. The ability of most oceanic general circulation models to realistically simulate density current formation and flow is also limited by spatial resolution (which is constrained by present computing power). We use both observations and a high-resolution numerical general circulation model to study the formation and dynamics of DCs that form over the shelf in the northern tip of the Gulf of Eilat, the northern Red Sea

1 Department of Environmental Sciences and Energy Research, Weizmann Institute of Science, Rehovot, Israel.

Copyright 2008 by the American Geophysical Union. 0094-8276/08/2008GL034123

(Figure S1).1 This is an ideal location to study DCs because it is a relatively small and simple oceanic sub-system that enables us to conduct detailed observations of DC as well as run a high resolution model capable of simulating small scale processes associated with them. Using our highresolution model results, we demonstrate the nonlinearity of these DCs by comparing two experiments with and without a diurnal cycle of surface heat flux. [4] These DCs are also of interest because their formation occurs in relatively warm (20°C) and salty (40.7 psu) seawater. Under these conditions seawater density is particularly sensitive to changes in temperature, where for example a 0.5°C temperature change results in a 0.13 kg
m 3 density change, which is three times greater than the density change near 1°C for an equivalent temperature change (Figure S2). Our study may thus shed light on the evolution of DCs during warmer periods in Earth’s past, when deep water ventilation may have been stronger [Boer et al., 2007]. Our results show that dense water forms over the shallow shelf at the northern tip of the Gulf during winter and cascades down-slope in daily pulses following the buoyancy loss. Similar pulses have been observed in a small canyon descending from the eastern shore of the Gulf of Eilat [Niemann et al., 2004] and also in Lake Geneva [Fer et al., 2001]. Somewhat similar thermally driven shallow water flows were also observed by Monismith et al. [2006].

2. Region and Methods [5] The Gulf of Eilat (Aqaba) is a long (177 km), narrow and deep terminal basin, at the northern tip of the Red Sea. The Gulf is connected to the Red Sea via the Tiran Strait, which is a short, narrow, and relatively shallow sill with a maximum depth of 252 m [Murray et al., 1984]. The arid climate and the high evaporation rate drive a reversed estuarine thermohaline circulation, which allows only warm surface water from the northern Red Sea to enter the Gulf [e.g., Genin, 2008]. Hence, deep water forms in situ during the winter and is relatively warm, 20– 21°C [Plahn et al., 2002], leading to a weakly stratified basin, with less than 0.5°C temperature difference between the upper and deep basin water during winter, and not more than 6°C during the summer. The weak stratification prevailing during winter allows dense water anomalies produced over the shallow shelves to descend to a great depth in the form of DCs. [6] Field studies were conducted during winter 2007 near the northern shore of the Gulf. This region is characterized by a relatively gentler slope (0.1 m/m) of the continental shelf in comparison to the much steeper shelves along the western and eastern shores of the northern Gulf (0.2 – 0.4 m/m). During these field campaigns, vertical profiles of tempera1 Auxiliary materials are available in the HTML. doi:10.1029/ 2008GL034123.

L14603

1 of 5

L14603

BITON ET AL.: PULSATING DENSITY CURRENTS

Figure 1. Meridional velocity component profile (top) and temperature at constant depths of 3, 8, 13, 18, 27, and 42 m above the bottom (bottom) measured in the north shore region of the Gulf of Eilat during 6 – 8 March 2007. The arrows indicate two succeeding cold water pulses (separated by one day), characterized by high down-slope velocities (up to 15 cm
s 1 (negative values)) in the bottom layer (up 15 m above the bottom). ture and salinity were measured with a Sea Bird Electronics SBE-19 CTD profiler. Vertical profile stations were situated along a transect extending out from the northern shore (bearing 220°) at 10, 20, 30, 50, 75,100, 150 and 200 m bottom depth. An upward looking 600 kHz RDI Workhorse Sentinel Acoustic Doppler Current Profiler (ADCP) was deployed at 45 m bottom depth (for the exact location and time of the CTD casts and ADCP, see Figure S1). Finally, six Vemco Minilog 12 bit temperature data loggers were distributed along the ADCP mooring line between the bottom and the surface. Four of these loggers were deployed in the lower 18 m above the ADCP in order to successfully capture the cold water anomalies associated with the DCs signal. [7] We use the Massachusetts Institute of Technology Ocean General Circulation Model (MITgcm) [Marshall et al., 1997a, 1997b], which solves the Navier Stokes equations including the non-hydrostatic terms. The model domain includes the entire Gulf of Eilat and part of the northern Red Sea with a bathymetry that is adapted from Hall and Ben-Avraham [1973], while partial cell method was used to give smooth bottom topography [Adcroft et al., 1997]. At the southern end of the domain, we have a ‘sponge-layer’ in which temperature and salinity are relaxed at all levels to the monthly mean hydrographic climatology from present-day Red Sea model simulations of Biton et al. [2008]. The horizontal resolution is 75 m and the water column is divided into 38 levels in the vertical. The time step for all simulations is 60 sec. At the surface we used climatological wind stress and thermohaline fluxes to drive the model (see the auxiliary information for more details).

3. Results [8] The ADCP and temperature data show pulses of anomalous high density water confined to the bottom with

L14603

a diurnal frequency. The meridional velocity component profile during the 7 – 8 of March 2007 is characterized by high southward velocities up to 15 cm
s 1 in a layer 15 m above the bottom, which begins to develop at 0500 and disappears at 1500 (Figure 1 (top)). During the same time period, the temperature at 3, 8 and 13 m above the bottom begins to drop, while temperature in the upper layer increases due to daytime heating (Figure 1 (bottom)). Finally, the density profiles indicate the increasingly deepening penetration over time of high density water along the slope from the north shore (Figure 2). These observations clearly indicate that the northern shore of the Gulf is the source of cold bottom water anomalies. During the night, maximal buoyancy flux loss from the surface leads to the formation of relatively denser water over the shallow shelves at the northern end of the Gulf, when compared to the adjacent deeper open water. The dense plume then cascades down-slope and gains buoyancy by entraining the lighter ambient water, until reaching a depth of 200 m (the maximal measured depth), approximately 2.5 km from its formation site. Similar pulses were observed during all winter months of 2007 (not shown) suggesting that these pulsating DCs are common. [9] The observed DCs were compared to two 5-day, high resolution simulations, which differed only in their surface heat flux forcing. The first run was forced by a complete diurnal cycle of heat flux (EXP1), while the second was forced by a time independent surface heat flux equal to the time average of the heat flux used in EXP1 (EXP2) (heat flux forcing for both experiments is shown in Figure S3). The results of EXP1 (Figure 3 (left)), which show a formation of dense water over the shelf during the nighttime and cascading down-slope during the daytime, nicely capture the main characteristics of the measured DC signal (Figures 1 –2). These reproduced characteristics include: 1) arrival and departure times of the pulse to and from the 50 m depth mark, 2) downward velocities up to 15 cm
s 1, 3) plume thickness of 15 m. Both the EXP1 results and the observations have, 4) completely mixed open water at least

Figure 2. Down-slope vertical contour plots of sq (kg
m 3) calculated from the salinity and temperature profiles measured at stations along the eastern section between 20 and 200 m bottom depth (stations indicated by the black lines) at six different times during 7 –8 March 2007.

2 of 5

L14603

BITON ET AL.: PULSATING DENSITY CURRENTS

L14603

Figure 3. Down-slope vertical contour plots of simulated sq (kg
m 3) (left) with a complete diurnal heat flux cycle (EXP1) and (right) constant average heat flux (EXP2) at different times on the 5th day of the model run, in a section close to the measured profiles (Figure 2). down to 200 m in the background density and, 5) dilution of plume water by entrainment of ambient water. Although the model densities are comparable with observations, there are some small differences between them. The observed plume and ambient water densities are slightly higher than the model densities, by 0.05 kg
m 3. This small difference is caused by the initialization of the model by climatological density profiles for February, while the observations were made in March. Indeed, CTD measurements from February 2007 indicate that density varied between 28.810 at the surface and 28.840 kg
m 3 at 400 m depth, which is very similar to the model results. In addition, the simulations were made with climatological heat flux while in reality heat loss from the surface can change dramatically from one night to the next, causing large differences in shallow water density. [10] The dense plumes that form along the north shore are channeled into two canyons that descend southwards leaving the area in between these two canyons almost free from DC flux (Figure 4 (top)). During its decent the modeled dense plume increases its flux (calculated according to equation 2 in the auxiliary material) due to entrainment and convergence. At a depth of 170 m, the dense plume veers westward in response to the Coriolis force [e.g., Griffiths, 1986; Nof, 1983; Wahlin, 2002] and the along slope fluxes start to increase (Figure 4 (bottom)). The model DC flux converges on the western side of the Gulf, while descending down to 300– 400 m depth. At this depth the DC continues to flow southward isobathically as a narrow stream in almost perfect geostrophic balance (balancing the buoyancy and Coriolis forces). Based on the model results, almost all of the DC signal originates at the north shore of the Gulf, where the coast has a relatively moderate slope, while along the much steeper eastern and western side boundaries, density pulses hardly exist, suggesting that

narrow and steep shelves are not able to support the formation of DCs. [11] Our observations and model results (i.e. EXP1) suggest that the pulses are the results of the diurnal heat flux cycle. Since many General Circulation Models neglect the diurnal cycle, we compared the results of the model driven by diurnal heat flux to a simulation driven by daily averaged flux (EXP2) in order to test the sensitivity of the simulated DC to the heat flux term. In both experiments, a high density signal is formed in the shallow water, which flows mainly down-slope to 175 m depth. Most of the entrainment occurs at this stage and causes a sharp increase in the downward flux and an exponential decrease in density. As might be expected, the main difference between the experiments is the diurnal formation of DCs in EXP1, while in EXP2 the DC is nearly time-independent and the dense water feeds the downward stream constantly (Figures 3 – 4). Most important, the model also shows much higher time average fluxes in EXP2, up to 40% more than in EXP1 (see Figure S5), demonstrating the intrinsic nonlinearity of the DCs.

4. Summary [12] We used observations made during the winter of 2007 in conjunction with high-resolution numerical simulations to investigate the formation of DCs over the shallow shelves at the northern tip of the Gulf of Eilat. Our results show that pulsating DCs are a common occurrence during the winter months in the Gulf of Eilat and that the diurnal air-sea buoyancy flux cycle is their main driving force, while their spatial distribution is influenced by bathymetry and rotation. The flow associated with these DCs is not continuous but occur in pulses. The dense pulses are formed during late night and early morning over the shallow shelves when buoyancy loss is greatest, and flow down-

3 of 5

L14603

BITON ET AL.: PULSATING DENSITY CURRENTS

L14603

Figure 4. Modeled down-slope (top) and along-slope (bottom) flux distribution associated with the high density signal in the northern Gulf of Eilat (m3
s 1) at different times during the diurnal cycle for EXP1 (left) and EXP2 (right). The black contours indicate the 0, 100, 200, 300, and 400 isobaths. Both experiments indicate that the northern shore is a major source for the deeply penetrating DC signal and that the flux distribution is highly dependent on topography and rotation. The dense plume is channeled into two canyons that descend from the north beach. Six hours after it forms, the dense plume veers westward in response to the Coriolis force and converges on the western side of the Gulf at a depth of 300– 400 m. At this depth the dense current continues to flow southward isobathically as a narrow stream in almost perfect quasigeostrophic balance. Summing the flux over depth shows that the total flux is 40% greater in EXP2 compared to EXP1 (Figure S5). slope during the daytime afterwards. These DCs occur under unique conditions of temperature (20oC) and salinity (41 psu) at which the density of seawater is very sensitive to temperature changes unlike seawater at a temperature of a few degrees in which DC usually form. [13] The ability of most oceanic general circulation models to realistically simulate density currents is limited by spatial resolution. Important small-scales processes such as turbulent entrainment - detrainment, have to be parameterized in those models whose typical grid size is a few tens of kilometers [Hughes and Griffiths, 2006; Legg et al., 2006; Wahlin and Cenedese, 2006]. The unique characteristics of the Gulf of Eilat make it a natural laboratory for studying DCs and allows us to both observe and simulate density currents at exceptionally high-resolution. We use this observational and modeling capability to demonstrate the critical importance of the intrinsic nonlinearity of density currents which is impossible to properly account for in global climate models. This nonlinearity, which is poorly represented in global models, is shown here to affect the volume flux associated with the simulated density currents. Such a bias in simulating density currents in global models may lead to significant biases on the larger scale global ocean circulation and water mass distribution. In future work we plan to use the high resolution model results in order to evaluate various entrainment schemes that may be used in coarser resolution climate models.

[14] Acknowledgments. This research was supported by The Israel Science Foundation (781/04) and by NATO (SFP982220). HG is the Incumbent of the Rowland and Sylvia Schaefer Career Development Chair. We thank the management and the staff of the Inter-University Institute for Marine Sciences of Eilat (IUI) for their cooperation and help. We are grateful to Martin Loch for helping us with the set up of the numerical model.

References Adcroft, A. J., C. N. Hill, and J. Marshall (1997), Representation of topography by shaved cells in a height coordinate ocean model, Mon. Weather Rev., 125, 2293 – 2315. Baines, P. G., and S. Condie (1998), Observations and modelling of Antarctic downslope flow: A review, in Ocean, Ice, and Atmosphere: Interactions at the Antarctic Continental Margin, Antarct. Res. Ser., vol. 75, edited by S. S. Jacobs and R. F. Weiss, pp. 29 – 49, AGU, Washington, D.C. Biton, E., H. Gildor, and W. R. Peltier (2008), The Red Sea during the Last Glacial Maximum: Implications for sea level reconstruction, Paleoceanography, 23, PA1214, doi:10.1029/2007PA001431. Boer, A. M., D. M. Sigman, and J. L. Russell (2007), Effect of global temperature change on deep ocean ventilation, Paleoceanography, 22, PA2210, doi:10.1029/2005PA001242. Fer, I., U. Lemmin, and S. A. Thorpe (2001), Cascading of water down the sloping sides of deep lake in winter, Geophys. Res. Lett., 28, 2093 – 2096. Genin, A. (2008), The physical settings of the Gulf of Aqaba: An explanation for the occurrence of tropical communities in the subtropics, in Aqaba-Eilat, the Improbable Gulf. Environmental, Biodiversity and Preservation, chap. Gordon, A. L., B. A. Huber, H. H. Hellmer, and A. Ffield (1993), Deep and bottom water of the Weddell Sea’s and Western Rim, Science, 262, 95 – 97. Griffiths, R. W. (1986), Gravity currents in rotating systems, Annu. Rev. Fluid Mech., 18, 59 – 89. Hall, J., and Z. Ben-Avraham (1973), Bathymetric chart of the Gulf of Eilat, Geol. Surv. of Israel, Minist. of Infrastruct., Jerusalem.

4 of 5

L14603

BITON ET AL.: PULSATING DENSITY CURRENTS

Hughes, G. O., and R. W. Griffiths (2006), A simple convective model of global overturning circulation, including effects of entrainment into sinking region, Ocean Modell., 12, 46 – 79. Ivanov, V. V., G. I. Shapiro, J. M. Huthnance, and D. L. Aleynik (2004), Cascades of dense water around the world ocean, Prog. Oceanogr., 60, 47 – 98. Killworth, P. D. (1983), Deep convection in the world ocean, Rev. Geophys., 21, 1 – 26. Legg, S., R. W. Hallberg, and J. B. Girton (2006), Comparison of entrainment in overflow simulated by z-coordinate, isopycnal and non-hydrostatic models, Ocean Modell., 11, 69 – 97. Marshall, J., C. Hill, L. Perelman, and A. Adcroft (1997a), Hydrostatic, quasi-hydrostatic, and nonhydrostatic ocean modeling, J. Geophys. Res., 102, 5733 – 5752. Marshall, J., C. Hill, L. Perelman, and A. Adcroft (1997b), A finite-volume, incompressible Navier Stokes model for studies of the ocean on parallel computers, J. Geophys. Res., 102, 5753 – 5766. Monismith, S. G., A. Genin, M. A. Reidenbach, G. Yahel, and J. R. Koseff (2006), Thermal driven exchanges between a coral reef and the adjoining ocean, J. Phys. Oceanogr., 36, 1332 – 1346. Murray, S., A. Hecht, and A. Babcock (1984), On the mean flow in the Tiran Strait in winter, J. Mar. Res., 42, 265 – 284. Niemann, H., C. Richter, H. M. Jonkers, and M. I. Badran (2004), Red Sea gravity currents cascade near-reef phytoplankton to the twilight zone, Mar. Ecol. Prog. Ser., 269, 91 – 99.

L14603

Nof, D. (1983), The translation of isolated cold eddies along a sloping bottom, Deep Sea Res., 30, 171 – 182. Plahn, O., B. Baschek, T. H. Badewien, M. Walter, and M. Rhein (2002), Importance of the Gulf of Aqaba for the formation of bottom water in the Red Sea, J. Geophys. Res., 107(C8),3108, doi:10.1029/2000JC000342. Symonds, G., and R. Gardiner-Garden (1984), Coastal density currents forced by cooling events, Cont. Shelf Res., 14, 143 – 157. Wahlin, A. (2002), Topographic steering of dense currents with application to submarine canyons, Deep Sea Res., Part I, 49, 305 – 320. Wahlin, A. K., and C. Cenedese (2006), How entraining density currents influence the stratification in a one-dimensional ocean basin, Deep Sea Res., Part II, 53, 172 – 193. Wells, M., and B. Sherman (2001), Stratification produced by surface cooling in lakes with significant shallow regions, Limnol. Oceanogr., 46, 1747 – 1759. E. Biton, H. Gildor, and J. Silverman, Department of Environmental Sciences and Energy Research, Weizmann Institute of Science, Rehovot 76100, Israel. ([email protected])

5 of 5