Introduction

   Small-scale turbulence is a random phenomenon, and theoretical relationships about turbulent processes are often only crude approximations. There are relatively few accurate statements that can be made about a turbulent flow without recourse to experimental evidence from flow itself (Tennekes and Lumley 1972). In the atmosphere, turbulent flows are relatively easy to observe. In the ocean, however, it is very difficult to directly visualize small-scale turbulence. Satellite remote sensing images allow visualizations of large scale turbulent motions, such as eddies and river plumes entering the ocean. Below the water's surface however, it is difficult and expensive to obtain information about the details of fluid motion. Wind speed at the water's surface has been used to estimate mixed-layer turbulence. This is because the mixing layer developed by wind stress is known to be a principle source of turbulent mixing in the waters above the thermocline. It is usually assumed that the energy available for mixing, and the dissipation, are both proportional to the cube of the wind speed (Dillon and Caldwell, 1980).

   Often low resolution (on the order of one sample per meter) measurements of the vertical gradients of properties, such as temperature, salinity and density, are used to estimate the extent of the mixed layer. Brainerd and Gregg (1995) calculated mixing layers and mixed layers on data sets for which microstructure data were available. This allowed the depths estimated from the density difference and the density gradient methods to be compared with appropriate microstructure data. It appears the best way to find mixing layers is from measurements that resolve the turbulent overturns within the mixing layer (Brainerd and Gregg 1995). Moum et al. (1989) measured diurnal turbulent kinetic energy dissipation rate fluctuations well below the depth of the mixed layer, as defined by density gradient criteria, near the equator.

   Many investigators need to know more than how deep the mixing is occurring. They need to calculate vertical heat, momentum, chemical and biological fluxes through the water column. Turbulence is the dominant contributing process for diapycnal transport of these water properties. Turbulent mixing events in the ocean are extremely intermittent, and basin-scale averages of transport are dominated by rare, energetic events. Samples of mixing must be intensively collected over large space and time scales to obtain a meaningful average.
 

Cruise Narrative, R/V Endeavor Cruise E9608, August 14 to September 1 1996

This was the first of two physical oceanography cruises conducted by the co-PIs Jack Barth and Mike Kosro as part of the ONR-sponsored Coastal Mixing and Optics (CMO) Accelerated Research Initiative. The objective was to rapidly survey a region around 40.5^oN, 70.5^oW where a set of moorings and a stationary vessel conducting profiling operations were located. The first cruise, took place during a period of strong summer stratification and will be described elsewhere. The second cruise, was conducted in the spring (25 April to 15 May 1997) as water over the shelf restratified after being mixed by winter storms. The water column was sampled by towing the undulating vehicle SeaSoar from the surface to within 5-7 m of the bottom. The vehicle was equipped with a Seabird 911+ conductivity-temperature-depth (CTD), a nine-wavelength light absorption and attenuation meter (WETLabs ac-9) and a new microstructure instrument (MicroSoar) which measured conductivity and temperature using robust, fast-response probes. Maps of hydrographic properties, velocity from a shipboard ADCP, and optical properties, were obtained over the continental shelf and slope. The SeaSoar tows were concentrated in two patterns: 1) a small box (SB) roughly 25 by 30 km centered on 40.5^oN, 70.5^oW (in 70 m of water on the mid-shelf) and with north-south lines separated by about 5 km; 2) a big box (BB) roughly 70 by 80 km which included the small box region but extended out over the continental slope and with north-south lines separated by about 10 km. Each of these boxes was sampled repeatedly during the 21-day cruise. Between SeaSoar tows, CTD/rosette casts were made and underway surface temperature and salinity, and meteorological measurements were made continuously.

The primary activities during the R/V Endeavor CMO cruises were SeaSoar profiling (details below) and the measurement of subsurface velocities using a shipboard acoustic Doppler current profiler (ADCP). To achieve higher vertical resolution, Endeavor's standard 150-kHz RDI ADCP transducer was replaced with a 300-kHz transducer from Oregon State University.

The SeaSoar vehicle was equipped with a Seabird (SBE) 911+ CTD with its pressure case mounted inside the vehicle and dual temperature/conductivity (T/C) sensors mounted pointing forward through SeaSoar's nose (Figure 1). Dual SBE pumps mounted inside the vehicle ensured a steady flow past the T/C sensors (Figure 1). A nine-wavelength light absorption and attenuation instrument, ac-9 (WETLabs, Inc., Philomath, Oregon), was mounted on top of SeaSoar in a rigid saddle and with a streamlined nose cone to minimize drag (Figure 1). For more details of the ac-9 installation, operation and data processing see Barth and Bogucki (1998) and for a data report see Barth et al. (1998).

On the bottom of SeaSoar, where a streamlined lead weight or "bomb" normally is mounted, was a new microstructure instrument (MicroSoar) which measured conductivity and temperature using robust, fast-response probes (Figure 1). The MicroSoar pressure case and lead weights, the latter in the form of a streamlined nose cone, were configured to approximate the weight of the "bomb". MicroSoar is capable of either sending its entire data stream (~1 MByte per minute or 16 kBytes per second) topside, or storing all the data internally on hard disks and sending a subset of the data topside to allow data quality to be monitored.

During the CMO SeaSoar cruises, SeaSoar was towed using a bare (i.e., no streamlined fairing attached as required for deep tow profiling; see Barth et al., 1996), 5/16" armored, seven-conductor (plus ground) cable from a trawl winch aboard Endeavor. Flight characteristics were similar to previous experiments using bare-cable towing (e.g. Barth et al., 1996) with maximum depths reached of around 105 m. The vehicle profiled from the surface to 105(55) m and back in 4(1.5) minutes at the deep and shallow ends of the north-south survey lines, respectively. The presence of the external instruments - ac-9 and FlashPak on top, MicroSoar on bottom - did not adversely impact the flight performance of the SeaSoar system in this bare cable configuration.

To supply power to each of the instruments onboard SeaSoar and to return a merged data stream, a prototype power supply and signal multiplexor unit was used during the August 1996 CMO SeaSoar/ADCP cruise. The Modular Ocean Data and Power System Plus (MODAPS+) was manufactured by WETLabs, Inc., Philomath, Oregon motivated by the need of Oregon State University scientists for a system capable of supplying more power and returning more data than possible with WETLabs' existing MODAPS (WETLabs, 1994). The MODAPS+ was installed inside the SeaSoar vehicle (Figure 1) and operated using 3 wires plus ground in the conducting tow cable. A topside power supply sent 300 volts down the cable where the subunit converted and parceled out power to the CTD, ac-9, ac-9 pump and MicroSoar. The data from each of these instruments was multiplexed and sent topside for storage as raw binary files on a PC. The signals were also split out by the MODAPS+ deckunit and sent to each instrument's data display computer. The CTD signal was passed through a WETLabs SBE deckunit emulator (a 286-based processor card) whose purpose was to turn the CTD signal communicated by MODAPS+ into that produced by a standard SBE deck unit. The CTD data stream was then fed into a data display, acquisition and flight control system (details below). The ac-9 data was sent to a display and monitor package running on a Unix workstation (Barth and Bogucki, 1998) and the MicroSoar data to a PC-based, LabWindows/CVI display system (see section "The A/D Convertion and Data Acquisition Software" of this report).

Because MODAPS+ could not maintain the bandwidth required to bring the full MicroSoar data stream to the surface, data stored on its hard drives were transferred to topside computers via a direct connection while MicroSoar was on deck between tows. It was decided to remove the MODAPS+ power and communications module and to replace it with a solution based on a WETLabs MODAPS (WETLabs, 1994) which was brought along as a spare.

Since MODAPS was not capable of powering and communicating with MicroSoar, the latter was reconfigured to accept 300 volts directly from topside by installing a power converter inside the MicroSoar pressure case to supply 12 volts to the instrument. A new RS-232 communications channel was also installed in the MicroSoar subunit to allow it to communicate via MODAPS (previous is MicroSoar-to-MODAPS+ communication was via Ethernet). The CTD and ac-9 continued to run via MODAPS, which requires three of the seven conducting wires plus ground. SeaSoar control signals still require two wires. A capacitor was installed in MicroSoar to isolate it from the MODAPS RS-485 communication lines.

R/V Endeavor sailed at 1400 on 14 August 1996 (all times UTC) with the science party from Oregon State University, a technician from WETLabs Inc. (Philomath, Oregon), and a marine technician and a graduate student from the University of Rhode Island (URI) aboard (Table 1). This was Endeavor cruise EN-287, but we shall refer to it as E9608 to conform to our traditional way of naming cruises using the first letter of the ship's name, followed by the year and month. The cruise was split into a 5-day leg, during which three individuals (Chang, Hankins and Holt) assisted in setting up and testing equipment to assure quality data collection, followed by an approximately two-week long second leg.

At 2000 on 14 August, a CTD/rosette cast was conducted at 40.9^oN, 70.5^oW at a station previously sampled during the CMO project. Bill Fanning, URI Marine Technician, trained the science party in the use of the CTD/rosette system. The Endeavor then proceeded south along 70.5^oW, checking for fishing gear, primarily fixed individual lobster pots marked with surface buoys and strings of lobster pots marked on both ends by surface buoys and sometimes with radar reflectors. Fortunately, the 70.5^oW line was not too heavily populated with lobster gear. We proceeded to south of the shelfbreak (~40^oN), where there was a local concentration of fishing gear, checking the southern end of the "big box" survey region. We then occupied 6 CTD stations from 39^o 54' to 40^o 19' N. The next activity was to visually survey for, and electronically chart, fishing gear on the six north-south lines of the "small box" survey grid. During the time it took to conduct the visual survey (approximately 0850-2300 15 August) we conducted a dip test of the SeaSoar vehicle to check for proper instrument performance and data acquisition.

The 15 August dip test was successful and the visual survey of the small box was completed by the end of that day. At 2314, SeaSoar was launched to begin the first survey of the small box grid (SB1). During the first tow, MODAPS+ intermittently (approximately every 30 s) dropped data scans from the CTD or sent bad data packets. This impacted both the flight control software, which is expecting every 24-Hz scan, and the CTD data processing system which is expecting 24 values to construct a 1-s averaged value. The former of these is the potentially more serious problem as the flight control software must fly "blind" during the dropouts. The flight control software was also designed to issue full "wings up" in the event of a break in the CTD data stream. This is not desirable should the MODAPS+ data dropouts come in mid-water column. Both software systems were modified by OSU personnel to handle the data dropouts from the MODAPS+ CTD data stream. The MODAPS+ SBE deckunit emulator also was not processing the second T and C channels from the dual sensors mounted on SeaSoar. The dual sensor capability has proven useful during SeaSoar operations in productive coastal waters to limit data loss due to (usually temporary) fouling of one T/C sensor pair by biological material.

At 0840 on 16 August, SeaSoar was recovered after high tensions (around 3000 lbs) were recorded presumably due to snagging lobster fishing gear. High tensions did not persist, as the vehicle presumably slipped off the moored fishing gear. No damage was done to the SeaSoar vehicle or to the onboard sensors. A connector in the FlashPak water supply had broken due to strong flow past the exposed fitting. It was decided to remove the FlashPak from the optical water supply and to fit it with a forward-pointing elbow connector so that the FlashPak would flush itself in the oncoming flow.

SeaSoar was redeployed at 0930 on 16 August, towed through completion of SB1, then recovered at 1535. The optical flow tubes and windows on the ac-9 were cleaned and reinstalled, as was done between every SeaSoar tow throughout the remainder of the cruise. Since MODAPS+ could not maintain the bandwidth required to bring the full MicroSoar data stream to the surface, data stored on its hard drives were transferred to topside computers via a direct connection while MicroSoar was on deck between tows. Meanwhile, a visual survey and electronic charting of fishing gear on the "big box" sampling grid was carried out on the north-south lines to the east of the SB grid.

At 0111 on 17 August, SeaSoar was deployed at 40^oN on line F (70^o 3.5' W) for the first big box survey (BB1). BB1 was completed at 0900 on 18 August and after a short transit, SB2 was begun. SB2 was completed at 0130 on 19 August and after an hour and half to test some new flight software, SeaSoar was recovered and the transit to the URI Narragansett pier was begun.

We arrived dockside at URI (Narragansett, RI) at 1200 on 19 August and three scientists (Chang, Hankins and Holt) disembarked. A new set of EPROMS was received from WETLabs containing MODAPS+ SBE emulator code to process the secondary T/C data correctly. The EPROMS were installed and tested successfully.

At 1700 on 19 August, R/V Endeavor sailed for the second leg of the CMO SeaSoar/ADCP survey with a reduced science party of 12 (Table 1). SeaSoar was deployed at 0130 on 20 August at the northeast corner of the small box survey grid for the start of SB3. SB3 was completed at 1445 on 20 August when SeaSoar was recovered and then readied for the next tow (optics cleaned, MicroSoar data transferred, etc.). At 1700 on 20 August, SeaSoar was deployed on the north end of line A for the start of BB2. At the south end of line C, SeaSoar was recovered at 1130 on 21 August to realign the absorption flow tube on the ac-9 which had come ajar. In the middle of line C, ADCP data stopped being acquired due to failure of a board in the ADCP deck unit. A replacement board was available on board, installed and tested successfully.

SeaSoar was deployed at the south end of line C at 1300 on 21 August and then towed north, repeating line C to collect good light absorption and ADCP data. Beginning at approximately 1530, more frequent data dropouts and eventual loss of signal from the MODAPS+ SBE emulator data stream occurred. Cycling the power to the SBE emulator card sometimes successfully restarted the CTD data stream. This did not always work, so cycling power to the entire MODAPS+ system was needed to recover the full data stream. After data loss became more frequent, SeaSoar was recovered at 2230 on 21 August.

From late on 21 August through 22 August, a series of tests (on deck, in the lab, and eventually with the MODAPS+ subunit pressure case opened up) were conducted in an effort to fix the MODAPS+ data communication problem. Meanwhile, a time series of 15 CTD casts to 85 m was made at 40^o 18'N, 70^o 21.4'W. No obvious failed or unseated components were found inside the MODAPS+ subunit, so it was reinstalled in SeaSoar for further towing. At 2057 on 22 August, SeaSoar was deployed, but the MODAPS+ system only worked for one and a half undulations before failing. At 2217 it was decided to recover SeaSoar and remove the MODAPS+ power and communications module and to replace it with a solution based on a WETLabs MODAPS (WETLabs, 1994) which was brought along as a spare. During this changeover on 23 August, CTD stations were performed every 5 nautical miles along the big box lines D and E in an effort to complete BB2.

Since MODAPS was not capable of powering and communicating with MicroSoar, the latter was reconfigured to accept 300~volts directly from topside by installing a power converter inside the MicroSoar pressure case to supply 15~volts to the instrument. A new RS-232 communications channel was also installed in the MicroSoar subunit to allow it to communicate via MODAPS (previous MicroSoar-MODAPS+ communication was via Ethernet). The CTD and ac-9 continued to run via MODAPS, which requires three of the seven conducting wires plus ground. SeaSoar control signals still require two wires. Since the remaining conducting wires were being used to power the MicroSoar, the power and data lines to the onboard echosounder (altimeter) were lost. This was deemed acceptable since SeaSoar was being flown using depth information from the ship's echosounder.

At 2311 on 23 August during the 47th CTD cast of the cruise, the pressure signal on Endeavor's SBE 911+ failed, thus ending the series of CTD stations at the end of BB2. Bill Fanning, URI Marine Technician, was unable to rectify the problem at sea after pursuing the problem with SBE technicians.

At 1240 on 24 August, SeaSoar with the new configuration based on the old MODAPS was deployed. It was found that stable CTD and ac-9 data were only obtainable with the MicroSoar turned off, presumably due to interference between the MODAPS communication lines and the MicroSoar power lines. At 1500, after diagnosing a possible short in the conducting cable, SeaSoar was recovered. One of the conductors was shorted, so it and the three wires carrying the MODAPS signal and power were reterminated. At 1949, SeaSoar was redeployed and towed three times from east to west on the small box grid (SB4, SB5, SB6). During this tow, MicroSoar was not powered up since stable CTD and ac-9 data were not possible with it running. At 1700 on 26 August, SeaSoar was recovered and readied for the next deployment (optics cleaned). While SeaSoar was on deck and MicroSoar detached and moved to the lab, a capacitor was installed in MicroSoar to isolate it from the MODAPS RS-485 communication lines and a successful deck test with stable CTD, ac-9 and MicroSoar data communication was performed.

While SeaSoar was on deck, a visual survey and electronic charting of fishing gear on the "butterfly" sampling pattern was done. At 2350 on 26 August, SeaSoar was deployed and reliable data from all onboard instruments was being acquired. Unfortunately, a short developed on the SeaSoar control conductors and the vehicle was recovered at 0100 on 27 August. After reterminating all conductors, SeaSoar was deployed at 0400 for the start of sampling on the butterfly (BF) pattern. Three complete cycles of the butterfly pattern were completed (BF1-BF3) before recovering SeaSoar at 2130 on 27 August. After cleaning the optics and transferring MicroSoar data, SeaSoar was redeployed at 2344 to begin an approximately 24-hour period of sampling aimed at capturing internal solitary wave packets (solitons) as they propagated through the CMO region. During previous tows (e.g., the north-south line on BF3), we had noticed evidence for packets of ISWs in both the CTD and optical data. Packets were located by rapidly undulating SeaSoar, making an estimate of their propagation direction by assuming they were formed at the shelfbreak to the southeast of our study region, and then attempting to recross the packets in a direction orthogonal to the wave crests. This was repeated several times, and included flying SeaSoar at a number of fixed (± 1m) levels across the soliton packets. At 0109 on 29 August, SeaSoar was recovered and the optics cleaned and MicroSoar data transferred. At 0348, SeaSoar was deployed and towed for SB7 and SB8 before being recovered at 1216 on 30 August. Upon inspection, the ac-9 attenuation flow tube had sea slime stuck in it which contributed to degraded data during the previous tow. The optics were cleaned for the next deployment.

At 1453 on 30 August, SeaSoar was deployed and towed on SB9 followed by sampling on BB3. Around 0600 on 31 August it was noticed that the ac-9 absorption data looked fouled, so SeaSoar was recovered at 0810, the ac-9 optics cleaned, followed by SeaSoar being redeployed at 0856. Sampling was continued on BB3 until 1109 on 1 September when SeaSoar was recovered and the Endeavor began a transit to Newport, RI to wait out the passage of Hurrican Edouard which had been steadily moving north along approximately 70^oW.

At 1700 on 1 September, Endeavor was dockside in Newport, RI ending the science portion of E9608. Hurricane Edouard turned east and passed over Cape Cod. On 3 September from 1230 to 1400, Endeavor transited from Newport to the URI Narragansett pier, thus ending cruise E9608 (EN-287).

In summary, despite a number of instrumental challenges a total of 11 days of SeaSoar towing were conducted yielding high-quality CTD, optical and microstructure data. Nine occupations of the small box grid, three of the big box grid, three repeats of the butterfly pattern and a day of soliton chasing were completed. The total number of water column profiles produced by SeaSoar was approximately 17,400. In addition, 46 CTD/rosette stations were occupied. Overall, this was a very successful cruise and operation of SeaSoar in this region of considerable shipping and fishing activity could not have been accomplished without the superb efforts of the captain, mates and crew of the R/V Endeavor. In particular, the electronic charting of lobster fishing gear and the around-the-clock vigilance of the captain and mates made it possible to slalom along the survey grids.

Cruise Narrative, R/V Endeavor Cruise E9704, April 25 to May 15 1997

Cruise E9704 was the second of two physical oceanography cruises conducted by the co-PIs Jack Barth and Mike Kosro as part of the ONR-sponsored Coastal Mixing and Optics Accelerated Research Initiative. The SeaSoar vehicle was equipped as in the August 1996 CMO cruise (E9608): SBE 911+ CTD; WETLabs ac-9; MicroSoar and a next-generation prototype single-channel fluorometer (WETLabs FlashPak). A major difference from E9608 was that the WETLabs MODAPS+ power and data communications module did not work when installed in SeaSoar and connected to the seven-conductor tow cable while dockside. With that result, the SeaSoar vehicle was loaded with the old WETLabs MODAPS as done during the second half of E9608.

After waiting out a storm on the original sailing day of 24 April, the R/V Endeavor sailed from Narraganset, Rhode Island at 1300 UTC on 25 April 1997 (all times UTC) with the science party from Oregon State University aboard (Table 2).
 

This was Endeavor cruise EN-299, but we shall refer to it as E9704 to conform to our traditional way of naming cruises using the first letter of the ship's name, followed by the year and month. A CTD/rosette cast was conducted at 40.9^oN, 70.5^oW at a station sampled previously during the CMO project. Bill Fanning, URI Marine Technician, trained the scientists in use of the CTD/rosette system. After the CTD, a transit to the south along 70.5^oW was conducted to visually survey for fishing gear so that the ship could avoid obstacles while towing SeaSoar along this NS line. Locations of visible gear were marked on the electronic chart on the bridge of the Endeavor. At 0100 on 26 April, we arrived at 39.9^oN, 70.5^oW and made a CTD/rosette cast to 1000 m. Endeavor turned north and began a CTD section of seven more casts along 70.5^oW finishing near 40.5^oN at 1000. For the remainder of that day, a visual survey was conducted to indicate fishing gear on the Small Box (SB) sampling grid, working from west to east. At 0038 27-April, SeaSoar was launched to begin the SB1 survey. At 1240, SeaSoar was recovered after the flight pattern degraded. Inspection showed a faulty hydraulic unit, which provides power to change the wing angle. The hydraulic unit was replaced, while making use of the time to conduct a visual survey for fishing gear on the Big Box (BB) survey grid. At 2220, SeaSoar was deployed and towed to the north to begin SB2. At 1450 on 28 April, SeaSoar was recovered at the captain's request as a storm built.

Early on 29 April, SeaSoar was redeployed and then towed on the SB sampling grid for SB3 and SB4. At 0150 on 30 April, SeaSoar was recovered after communication with the vehicle was lost. Inspection showed the need to replace two shorted pigtail leads near the SeaSoar bridle which had worn through. As a result of the short, a communication channel within the SeaSoar data telemetry unit (MODAPS) was made inoperative. After attempts to fix the unit, OSU scientists aboard the nearby R/V Knorr (located at 40.5^oN, 70.5^oW and conducting vertical profiling operations as part of the CMO experiment) offered to loan us their spare MODAPS. It was brought over by small boat and installed in SeaSoar, but the communication channel still did not function properly. During testing of the spare communications unit, a shallow CTD cast was conducted to test Endeavor's CTD winch #1. The winch drum was rubbing on the clutch plate and stripping off metal. The captain suggested limiting the number and depth of any future CTDs. Meanwhile, OSU scientists onboard R/V Knorr repaired the formerly defective communications unit and delivered it back to Endeavor via a floating transfer since the high winds and seas made a small boat transfer unsafe. The repaired unit was intstalled in SeaSoar, but the communication unit was still unable to perform up to specifications. A decision was made to hook the Seabird CTD onboard SeaSoar directly to its deck unit, rather than going through the defective communications unit. The ac-9 instrument and data continued to be run from the MODAPS, but now the RS-485 MODAPS communications on the sea cable was being run in parallel with the SBE data telemetry, and thus there was potential for crosstalk. However, with both systems powered up and running data was successfully acquired from both the CTD and the ac-9 (as well as the MicroSoar data sent via MODAPS to monitor the performance of the microstructure sensors). At 0200 on 2 May, SeaSoar was ready for deployment, but a 10 hour delay was necessitated by high winds and seas. At 1200 on 2 May, SeaSoar was deployed and towed on SB5 and SB6. At 0200 on 4 May, SeaSoar was recovered as the weather worsened.

At 1230 on 4 May, SeaSoar was deployed and towed on the BB grid, completing BB1 at around 1800 on 6 May. This was followed immediately by SeaSoar surveys SB7 and SB8. During SB8, the counterweight on SeaSoar's rudder, which acts in concert with the rudder to keep the vehicle flying level and right-side-up, fell off. Upon recovery of the vehicle at 1945 on 7 May during rough seas, the MicroSoar microstructure probes were damaged, necessitating replacement. A CTD cast was performed to test repaired CTD winch #1. The winch was deemed to be working properly again, and the go-ahead was given for future deep CTD casts. SeaSoar was deployed again at 0530 on 8 May and towed on the remainder of SB8. At 1320, SeaSoar failed to respond to wing angle changes and was recovered. Upon inspection, the 1/2" stainless steel push rod in the hydraulic unit had broken, presumably by excessive use over the last 10 days as SeaSoar completed a full undulation in this shallow-water experiment every 1-4 minutes (compared with deep-water, 300+ m undulations which take about 10 minutes). A new hydraulic unit was installed, and SeaSoar was back in the water at 1900. SB8 was completed, followed by surveys SB9 and SB10 and recovery of SeaSoar at 1705 on 10 May.

On the morning of 11 May, five CTD casts along 70.5^oW from 39.83 to 40.15^oN were completed. Deployed SeaSoar was deployed at 1100, and completed surveys SB11 and SB12 by 1455 on 12 May. Surveying began immediately on the BB grid for the second time (BB2). On 13 May at 0955, the ship's #1 generator failed and ship speed fell below 5 knots, the minimum required for SeaSoar to fly. The vehicle began sinking with 230 m of cable out in 87 m of water. The engineers reacted quickly and brought the #2 generator on line and the SeaSoar cable was immediately begun to be brought in at full speed. From the CTD onboard SeaSoar, the closest approach to the bottom was determined to be 3 m. The remainder of BB2 was completed by 2300 on 14 May; SeaSoar was recovered and the transit to Narragansett was begun. ENDEAVOR arrived dockside at 1215 (0815 local) on 15 May 1997.

Overall, E9704 was a very successful cruise, with in excess of 13 days continuous towing of the SeaSoar vehicle, including 12 occupations of the small box centered around the CMO central site, and 2 occupations of the larger box, which included sampling the shelfbreak frontal region out over the continental slope. The total number of water column profiles produced by SeaSoar was approximately 17,500. As during E9608, operation of SeaSoar in this region of considerable shipping and fishing activity could not have been accomplished without the superb efforts of the captain, mates and crew of the R/V Endeavor.

SeaSoar towed profiling platform

    MicroSoar to date has been used as a passenger instrument aboard the Oregon State University (OSU) SeaSoar towed profiling platform. MicroSoar mounts underneath, replacing the SeaSoar ballast weight, (Figure 2). The SeaSoar "flies" through the water behind the towing vessel. It was most commonly used to make a series of depth profiles of the water column (e.g. 80 meters depth for 200 meters horizontal). When it approaches the surface, the wings are commanded to tilt, so that it dives rapidly. When it approaches the bottom, the wings are commanded to tilt, so that it climbs rapidly to the surface. The OSU SeaSoar is equipped with a SeaBird 9/11+CTD that samples dual sensors 24 times per second. During a series of successful oceanic cruises, the SeaSoar mounted CTD data quality has been continuously improved. The SeaBird CTD onboard SeaSoar serves as constant calibration reference against which the low-frequency accuracy of the MicroSoar sensors can be calibrated.

   The CTD temperature and conductivity sensors are mounted pointing forward through a hole in SeaSoar's nose, considerably improving performance over a previous sideways-plumbed T/C configuration (Barth et al. 1996).
 

MicroSoar functional description

    MicroSoar is a general-purpose high-frequency computer-controlled measurement system. It uses a high performance PC/AT-compatible PC/104 100 MHz 486DX4 CPU CoreModule^TM (CM/486, AMPRO Inc.^TM ). Within just 14 square inches of space, the CM/486 includes the equivalent functions of a PC/AT motherboard and meets the demands of embedded systems, through its extreme compactness, low power consumption, +5V-only operation, wide operation temperature range and high reliability. The motherboard includes 8 Mbytes on board DRAM memory, a 16-bit expansion bus, and dual-serial and parallel controllers. It also has an enhanced embedded BIOS, Watchdog Timer and Socket for a bootable "Solid State Disk". The CM/486 is plugged into the application circuit board described in May (1997). Besides the motherboard, the Computer Electronic System (CES) includes PC/AT compatible disk drive interfaces (floppy controller and IDE hardware interface) supporting two hard disks, high resolution display controller, and an Ethernet NE2000 compatible LAN communication adapter. Two 16-bit, 8-channels 100 kHz fully differential analog to digital (A/D) converters (AIM-16, Analogic Inc.^TM ), sampling at 2048 Hz per channel (higher or lower frequencies can be used, depending on the specifics of the experiment). Multiple modules in CES can be stacked together, or they can be mounted separately on the application circuit board. The CES also contains two Western Digital IDE hard drives with 2 Gbytes each. The operating systems used were DOS-6.22 and Windows 3.11 for Work Groups^TM . The Intelligent Data Acquisition Software (IDAS) was developed at OSU to be executed in a DOS/Windows environment. The IDAS runs automatically after the CES has powered up and computer reboots. A typical operating mode for MicroSoar includes delivering digital data signals to the surface (at least one second averaged data for the visual analysis in real time), as well as recording all data on MicroSoar's hard disks. Windows is usually used for high speed transferring of data to the deck host computer, using the TCP/IP protocol and Ethernet interface. As a rule, it takes 2-3 days to fill up the hard disks, depending on what sample rate is used. After that, SeaSoar and MicroSoar are pulled on deck, and MicroSoar is directly connected to the deck computer via an Ethernet twisted pair (10BaseT). The next step is downloading raw data to another storage medium, and then clearing MicroSoar's hard disks. The data transfer rate is approximately 27 Mbytes per second, and it takes usually 3 hours to move all data from the MicroSoar Submerged Computer to the Deck Unit hard disks. After that, MicroSoar is ready for next measurements.
 

MicroSoar sensors

     MicroSoar's Analog Electronic System (AES) provides power conditioning, sensor excitation voltages, offset and gain calibration adjustments, buffering, filtering and signal routing of all analog signals. AES was configured to have 11 analog differential channels. Conductivity, conductivity derivative, and a reference ground signal, were sampled at 2048 per second. Temperature, temperature derivative, pressure, pressure derivative, three single axis accelerometers (axis X, axis Y and axis Z), and reference ground signal, are sampled 256 times per second by the second ADC board. The complete sample rate of all 11 channels is 8192 samples per second, corresponding to 131,072 bits per second. MicroSoar carries a capillary microconductivity probe (Paka et al. 1998; May 1997); a temperature probe, built at OSU, utilizes a ThermometricsÔ FastTip FP07 glass-coated bead thermistor. A stainless steel protective sleeve protects the thermistor probe tip during handling and deployment. The pressure sensor is an Endevcoä model 8510B, 500 psig, piezoresistive pressure transducer with a sensitivity of .5 mv/psi. MicroSoar uses three IC Sensorsä model 3140-002 accelerometers, featuring approximately 1 volt/g output, with flat response up to 500 Hz, with a maximum of two gravities acceleration. MicroSoar's design details, and detailed information about the microconductivity sensor and analog and digit electronics, can be found in May (1997).
 

Structure of raw data

    The length of one record of MicroSoar raw data is 16384 bytes (8192 samples) and is storred in a RAM double-buffer once each second. Raw sata records are copied from a memory buffer to a hard disk file during the data acquisition process. Each record is supplemented by a header of 32 bytes, containing the start record label, PC time and date, record number, etc. (see Figure 3). Each record in a raw data file consist of 256 scans and each scan contains 3_channels x 8_times = 24 samples of the fast channels, and 8 samples of the slow channels, with the sampling frequencies 2048Hz and 256Hz respectively.

 The sequence of MicroSoar channels in a scan is shown below.

The A/D Conversion and Data Acquisition Software

    The AIM16/12-1/104 A/D converters have 8 differential input channels with software programmable gains for each channel. The channel list consists of a starting channel code and an ending channel code. Data is stored in a 256x16 bit FIFO buffer and transferred to the host via programmed I/O or DMA. The converter sets a status bit "true" when the FIFO is half full.

   The MicroSoar data acquisition software is based on an interrupt driven I/O routine. There are 3 modes of operation for reading data from an A/D converter. We used a procedure when the mode of operation which initiates a conversion each time that a trigger signal occurs, and a burst of conversions is taken. There are 3 functional regimes of the trigger, corresponding to an external timer, an internal timer, and a programmed register bit. The internal trigger timer counter is used in MicroSoar. In this mode, a 24 bit counter is pre-scaled with a 24 bit code and the counter is clocked by the on-board 10MHz clock. When a conversion trigger occurs, a burst of conversions takes place. A channel range is programmed in the A/D setup register and the trigger counter begins as soon as the GO bit is set true. A conversion burst begins at the start channel, and ends with the end channel Within such burst, conversions occur at a rate of 10⁵ conversions per second.

    MicroSoar can acquire data using up to 16 channels in differential mode, that is, 8 channels per each A/D converter. One of the A/D converters is used as a "master" board. It is programmed using an internal trigger timer counter. The second one is a "slave" board, and is programmed using a software trigger. The software trigger is set via a programmed I/O instruction; when triggered a burst of conversions occurs in the "slave" A/D converter. For example, if it is necessary to collect data from 11 channels (3 channels at the sample rate 2048Hz, and 8 channels at the slower rate 256Hz), we can use the master A/D converter for the faster channels, and the slave A/D converter for the slower channels. Schematically, the acquisition process is shown below:
 

   All data collected from both A/D boards are stored in a large memory FIFO buffer. When this buffer is half full, a "save to a hard disk" operation invoked by a background program. Every 5 minutes, the current raw data file is closed and new file is opened. This procedure was designed to insure that only small portion of raw data will be lost if some incidents, such as unplanned loss of power, occurs. If, for example, the MicroSoar computer "locks up", less than 5 minutes of data will be lost. This procedure also keeps raw data in sequentially stored 5 minutes data files, which is very convenient for calibrating MicroSoar's sensors.

   The background software also includes the following procedures: 1) checking how many bytes are available on the hard disks; 2) one second data averaging; 3) sending the data to the Deck Unit computer via a standard RS232 interface for visual monitoring in real time. During measurements, the Deck Unit computer is used for a real time visualization of MicroSoar data. One second averaged data for 11 channels are shown on the screen simultaneously, in a strip chart display and are continuously scrolled in their windows from the right to the left (Figure 4). The following information is displayed on the screen: 1) Start time of measurements; 2) current time (GMT-time); 3) how many Mbytes have been already acquired by the MicroSoar submerged computer (MSC); 4) how many Mbytes are available on MSC. There is also an option to send complete sets of raw data to the Deck Unit Computer in real time. We used the PC/TCP OnNet Developer's Toolkit 3.1 software to develop DOS functions for transferring data using the TCP/IP communication protocol.
 

Comparison with calibrated SeaBird CTD

   As shown in Figure 5, SeaSoar is fitted with a SeaBird 9/11 CTD, whose redundant sensors are factory calibrated. MicroSoar data can be compared at any time to these calibrated reference standards. Actual cruise data can be used to evaluate MicroSoar's performance compared to the SeaBird CTDs. The SeaBird temperature and conductivity sensors are designed to drift as little as possible over time. This stability is achieved at the expense of sampling speed and sensitivity, which is the goals of the microconductivity sensor design. To determine the amount of drift over time in the MicroSoar system, nine 5 minutes segments of data were selected over a 4 hour period. The CTD and MicroSoar data sets were averaged down to one second. Linear regression analysis were then performed for pressure, temperature and conductivity for each of nine 5 minute segments of data. All data sets have very good correlation. The results of linear fitting for MicroSoar and SeaBird data is shown in Table 3.
 
 

    These data were measured during a scientific cruise off the coast of Oregon in May 1996. The coefficients in the summary represent averaged calibration coefficients over period of almost 4 hours of measurements. Standard root mean square errors for the coefficients are also included. Figure 5 shows one of the 5 minutes segments of data used in the analysis above. Note that there are no significant drifts or offsets in the MicroSoar data when graphed using the calibration coefficients in Table 3.

    Figure 6 shows a time series analysis of calibration coefficients for the MicroSoar microconductivity sensor for about 50 hours duration. Each point in the graph represents one pair of slope and intercept calculated for over sequential 5 minute time interval, and divided by their averaged values. Calibration of MicroSoar raw data is strait-forward, and is an automatic procedure. A time series analysis of calibration coefficients has been used for "outlier" detection for the MicroSoar microconductivity sensor. Considering the huge amount of MicroSoar raw data, this analysis is very helpful for detecting segments of raw data (if slope and/or intercept are outside of ± %5 interval) which are candidates for hand editing and calibration. Usually there are not many points outside of ± %5 interval, and mostly they are related to some intermittent problems with the MicroSoar analog electronics.
 

Analysis of instrument vibration using three-axis accelerometer data

     It is important to insure that MicroSoar does not affect the deployment or flying characteristics of SeaSoar, and that vehicle motion is sufficiently small. "Sufficiently small" means that measurements of scalar properties, such as temperature or conductivity, are unaffected by SeaSoar vibrations. A three-axis accelerometer was used to estimate the displacement of MicroSoar in all three directions: Z-axis (straight forward), Y-axis (up and down) and X-axis (in horizontal plane).

     An acceleration can be expressed as a sum of harmonics. A value of acceleration for each harmonic at a frequency w can be estimated by the formula , where A is the amplitude, and . Displacement for the harmonic with the amplitude A and frequency w can be defined as

We estimated displacement at dominant frequency, _peak, of an acceleration power spectrum. Maximum displacement in the straight forward direction was Z_max= 0.06mm at the _peak=88Hz, maximum displacement in the up and down direction Y_max= 3mm at the _peak=22Hz, maximum displacement in the left and right direction X_max=0.7mm at the _peak=20Hz. An estimation of displacement also can be calculated for a frequency range, giving an integrated estimation of displacements. If  is the FFT transformation of an acceleration a, then the variance of a is estimated as

If the spectrum of acceleration is known, we can estimate the velocity variance as

and the displacement variance as

Data from MicroSoar's three-axis accelerometer, measured during CMO cruise in August 1996 (E9608), were used to examine the effect of SeaSoar vibration. MicroSoar was towed at 6 knots vessel speed in the depth range of 0-45 meters. The values of a root mean square displacement, ,  and , are shown on the Figure 7. The acceleration components were measured at a 256Hz sample rate, and each point on the figure is calculated using 256 raw data points (i.e., one second intervals). The rms of displacements on the Figure 7a, 7b, 7c and 7d were calculated for the frequency ranges 1 to 128Hz, 1 to 50Hz, 10 to 128Hz and 50 to 128Hz respectively. (± 2 ) for 1 second interval are shown in the Table 4.
 

frequency range, (Hz)	1 to 128		1 to 50		10 to 128		50 to 128
depth range, (m)	0 - 5	5 -45	0 - 5	5 -45	0 - 5	5 -45	0 - 5	5 -45
2x , (mm) - lateral	21.6	7.4	21.6	7.4	0.12	0.16	0.007	0.03
2y , (mm) - vertical	19.6	5.2	19.6	5.2	0.16	0.14	0.006	0.02
2z , (mm) - forvard	42.0	24.0	41.4	23.0	0.12	0.1	0.004	0.014

The largest values of the displacements typically appear in the upper 5 meters depth range, and smaller values appear in 5-45 meters depth range. It is important that displacements, as well as vibration, have significant values only in the lower 1-10Hz frequency range, and negligible values in the 10-128Hz frequency range. Typical rms of displacements in the X, Y and Z directions have been calculated for the 10-128Hz frequency range. They are practically the same at every depth. The 2 values in the range of 0.1-0.2mm are negligible and cannot be considered as a source of significant errors for MicroSoar small scale measurements.
 
 

Temperature variance dissipation rate calculation.

   It is not easy to directly measure the flux of water properties across stratified boundaries in the ocean. In principal, if the velocity fluctuations and concentration gradient of the water property (e.g., heat, chemical composition, or density) can be precisely measured at the same point and time, the flux can be calculated. For instance, the heat flux F_H is given by F_H= C_p<u'_zT'>. Here , u'_z is vertical velocity fluctuations, <u'_zT'> is the correlation of u'_z and T',  is density and C_p is the specific heat. A problem with directly measuring a set of correlated data for u'_z and T' arises because the correlations are quite small. These measurements may include errors introduced by non-turbulent internal waves, imprecisely matched sensor response time, and movements of the measurement platform. It has proven more effective to infer fluxes from measurements of the turbulent kinetic energy dissipation rate,  , or from the temperature variance dissipation rate, _T, both of which rely on the measurement of only one fluctuating quantity.

Cox Number, C_x. The Cox number, C_x, is a ratio of the varience in the temperature gradient fluctuations (a function of stirring due to turbulent mixing) to the mean temperature gradient, against which the turbulence was established. Higher Cox numbers indicate that turbulence is playing an increasingly important role in mixing the water column. The Cox number can be determined from vertical microstructure temperature measurements (Dillon and Coldwell 1980):

Heat Flux F_H can be defind as

where, typically,  = 1025 kg/m³, C_p = 4000 J/(kg ^oC), D = 1.4 x 10^-7 m²s^-1, and I_so = 3. Numerically, in SI units,

The Temperature Variance Dissipation Rate,_T, is

Existing thermistor temperature sensors can resolve _T in the range of 1-20Hz, at best. A way to increase the frequency range is to use a micro conductivity sensor, which can resolve fluctuations up to 1000Hz or more. It is important to note that vertical profiles of small-scale fluctuations are not necessary, and that one-dimensional measurements in any arbitrary (say, ) direction provide as much information, as long as the mean gradient is known. The temperature variance dissipation rate is given by

where u is the instrument velocity.

   Because temperature is well-known function of conductivity and salinity, we can write

We also assume that the local measured T-S relation, given by

holds from meter scales to the smallest temperature fluctuation scales, and we can write

where b_C is the sensor sensitivity, and ¶V/¶t is the time derivative of voltage. The final result for the temperature variance dissipation rate is:

 The value of  of the variance of conductivity derivative, in voltege units, is estimated by integration of the power spectrum in a time/space domain. If
S_x() is the frequency domain spectrum, then

and

where  = 2  f. Thus,

 The advantage of this method is that we can make a correction to the variance by excluding the noise spectral component from the conductivity derivative signal. Figure 8 shows the noise spectra of conductivity derivative as a dashed line. This model of the noise spectra is used for correcting . Instrumental velocity u, in formula (15) is calculated from approximation

 If X is latitude and Y is longitude, ship velocity u_ship  is defined as

Figure 9 shows a typical example of calculated instrument velocity and all components used for calculation.

    The temperature variance dissipation rate cannot be calculated if _TS ST = 1, and _T becomes less precise, whenever _TS ST is near unity. A "figure of merit" for the calculated _T can be defined as
 

                                                                                        (21)

If M is very different from unity, we cannot trust our T-S approximation, because the spectrum may be dominated by salinity rather than by temperature fluctuations. Calculations of M for entire E9704 cruise give us next results:
 

Condition	% of Occurance
0.1  M  10	93%
0.3  M  3	83%
0.5  M  2	66%

MicroSoar data processing steps

Source: MicroSoar raw data:

C_v - conductivity ("v" mean volts), C'_v - conductivity derivative (electronically differentiated from C). Sampling frequency 2048 Hz
P_v - pressure, T_v - temperature, A_x, A_y, A_z - three axes accelerometer. Sampling frequency 256 Hz:

Primary data processing (for each one second time interval):

• Remove spikes from all signals;
• Calculate D C_v = ( C_vⁱ⁺¹ - C_vⁱ ) /  t;
• Detrend signals and remove mean values before calculating spectra;
• Create one second averaged data set (in volts). Table 5 below shows the structure of one second averaged MicrosSoar data.

<T>	<C>	<P>	<T' >	<C' >	<C/t >
<P' >	<Accel-X>	<Accel-Y>	<Accel-Z>	<Shorted-1>	<Shorted-2>
T	C	etc.	...	...	...
...	...	...	...	...	...
Skewness T	Skewness C	etc.	...	...	...
...	...	...	...	...	...
Kurtosis T	Kurtosis C	etc.	...	...	...
...	...	...	...	...	...
Fast channels frequency array, Nfast points
Power spectrum of C/ t, Nfast points
Power spectrum of C ' , Nfast points
Coherence spectrum of C/ t and C ' , Nfast points
Phase spectrum of C/ t and C ' , Nfast points
In-Phase power spectrum of C/ t and C ' , Nfast points
Slow channels frequency array, Nslow points
Power spectrum of T ' , Nslow points
Power spectrum of C ' , Nslow points
Coherence spectrum of T ' and C ' , Nslow points
Phase spectrum of T ' and C ' , Nslow points
In-Phase power spectrum of T ' and C ' , Nslow points

One second averaged MicroSoar data, in volts, includes these characteristics:

• Four first moments for all 11 channels, calculated over one second time interval (mean, standard deviation, skewness, kurtosis);
• Band averaged power spectra for the of C/t, C' and T ' coherence spectra between C/t and C' , and between C' and T ' ; Phase spectra between DC/D t and C' and between C'  and T ' , In-Phase power spectrum between C/t and C'  and and between C' and T ' .
• Match MicroSoar and SeaBird CTD data by pressure, correct MicroSoar time base and incorporate MicroSoar one second averaged data with navigational data.
• Calibration of MicroSoar data using SeaBird CTD measurements. finding calibration coefficients for each 5 minutes of MicroSoar one second averaged data sets:
• C_ctd = a_C  + b_C C_v
• T_ctd = a_T  + b_T T_v
• P_ctd = a_P  + b_P P_v

 
•  Calculate microstructure parameters:
• Cox number;
• Heat Flux;
• Temperature variance dissipation rate;

• Create final MicroSoar data sets in terms of tows and lines.

Data presentation

   The final 1-Hz data files contain unfiltered GPS latitude and longitude; pressure; temperature, salinity and sigma-t from the preferred sensor pair of SeaBird instrument; date and time (in both decimal day-of-year and integer year, month, day, hour, minute, second); SeaSoar velocity, Temperature Variance Dissipation rate, Cox Number and Heat Flux, an integer flag: 0 - means original CTD data are used, 1 - means there was not CTD data and MicroSoar pressure temperature, salinity and sigma-t are used.

    For the SeaSoar observations, we split the tow data into the small box and big box surveys. Sections which connect one box to another were used in the maps for both boxes. Maps of Temperature Variance Dissipation rate, Cox Number and Heat Flux are shown for every ten meters between 5 and 75 meters depth for the small box surveys; the big box surveys continue that down to 105 meters. Data used in the maps were obtained by first binning the data into 2-db bins in the vertical, and 1.25 km bins in the horizontal. Then, the depth of interest was extracted from the appropriate sections for the maps. Contour maps were then created by gridding these data using zgrid (Crain, 1968, unpublished). The small box grid used a spacing of 0.025 degrees in E-W spacing, and 0.0125 degrees in N-S spacing, while the big box grids used twice that (0.05 degrees E-W and 0.025 N-S spacing). Any grid point more that two grid spaces away from a data was set to be undefined.

    Vertical sections of Temperature Variance Dissipation rate, Cox Number and Heat Flux are shown for each of the SeaSoar lines.
 

Acknowledgements

We are indebted to Robert O'Malley, who supplied the SeaBird CTD data for MicroSoar sensors calibrations. We would like to thank Marine Technicians Linda Fayler and Marc Willis, they were respossible for the highly successful SeaSoar operations. We also thank Andy Dale who was in charge of asembling/disassembling and mounting MicroSoar instrument to the SeaSoar vehicle. This work was funded by the Office of Naval Research Grant N0014-95-1-0382.
 

References