Inverted Echo Sounder Data Report Ulleung Basin of Japan/East Sea June 1999 to July 2001 by Douglas A. Mitchell Yongsheng Xu Karen L. Tracey D.Randolph Watts Mark Wimbush William Teague GSO Technical Report 2004-0 GRADUATE SCHOOL OF OCEANOGRAPHY UNIVERSITY OF RHODE ISLAND NARRAGANSETT, RHODE ISLAND October 2004 ABSTRACT Observatons were conducted from June 1999 to July 2001 to study the shallow and deep current variability in the southwest Japan/East Sea. Data were collected during the field experimentwith a two-dimensional array of pressure-gauge equipped inverted echo sounders (PIES) and deep ecording current meters (RCM). The collection, processing and calibration of the PIES are documented in this report. Descriptions of the processes used to identify and remove jumps (offsets) in the pressure and travel time recors, that resulted when the instruments were jostled or moved deep crab-fishing activities in the Japan/East Sea, are give. Time series plots of travel time, bottom pressure, and mperature are presented for the 24 recovered instruments. Basi statistics of the hourly data are tabulated. Setting and Experiment Design Introduction Ths report focuses on data collected from an array of 25 pressure-gauge-equipped inverted echo sounders (PIES) deployed in the Ulleng Basin of the southwestern Japan/East Sea (JES) during June 1999 to July 2001. In this report the collection, processing, and calibration of the PIES data are documented. The measurements were made under the support of the Office of Naval Research. Other data collected as part of the experiment included expendable bathythermograph (XBT) surveys, conductivity-temperature-depth (CTD) surveys, and velocity measurements obtained with moored current meters. The locations of all the moored instruments is shown in Figure 1. The PIES were deployed in an approximately 5 by 5 two-dimensional array with instruments separated by about 55 km in both the North/South and East/West directions. Interspersed within the PIES array were 18 current meter moorings, each consisting of one current meter approximately 23 m above the bottom. The current meter records are documented in Xu et al. [2003]. PIES Description The PIES array during the JES experiment was a collaborative effort between the University of Rhode island (URI) and the Naval Research Laboratory (NRL). Three different models of inverted echo sounders were used in this experiment. The type designated SD-NRL in Table 1 was Sea Data Model 1665 with standard options built for NRL. Type SD-URI was also Sea Data Model 1665, but the components were assembled at URI. Type WHISL in Table 1 was built by Woods Hole Instrument Systems Ltd. for NRL. All three instrument types ping at a frequency of 10.24 kHz. The Model 1665 instruments record the data to internal cassette tapes and the WHISL models record to solid state memory. In addition to the travel time detector, these instruments were also equipped to measure pressure and temperature. Chaplin and Watts (1984) provide a description of the IES circuitry and mooring configuration and Watts and Kontoyiannis (1990) describe instruments with pressure and temperature sensors included. The Paros Digiquartz pressure sensors (Table 1 ) had one of two full-scale ranges, either 6000 psi (approximately 4000 dbar) or 10000 psi (approximately 6800 dbar). The 6000 psi sensors were owned by URI (Paros serial numbers primarily ranging between 20000 and 40000) and had been previously deployed for many months. Because of this pre-pressurization, they performed with minimal drift, typically <0.2 dbar during the two year deployment. The WHISL instruments (owned by NRL) had never been deployed, and their new 10000 psi sensors (Paros serial numbers >50000) xhibited slightly larger drifts (0.1-0.7 dbar). The internal reference oscillators in these WHISL instruments may also have drifted, but this could not be confirmed. Five 10000 psi sensors (Paros serial numbers 18860-19934) owned by NRL were installed in non-WHISL model instruments. Although these gauges had been previously deployed, the exhibited very large (>2 dbar), atypical drifts. The PIES instruments were moored one meter above the ocean floor and sampled at hourly intervals throughout the deployment. At the deignated sampling time, a burst of 24 acoustic pulses (10.24 KHz) were transmitted at 10 second intervals, and the time eh ping took to travel the round trip distance to the sea surface was measured. Typical deployment depths of 1000-3000 m produce full travel times between 1-4 seconds; thus with a resolution of 0.05 ms, each measurement would require a storage space of 18 bits. However, the variability of the travel time signal in the JES region is about 0.01 seconds making it unnecessary to record the full travel time measurement. By recording just the 13 least significant bits, variability of up to 0.4 seconds can be resolved, with only a constant integer multiple of 0.4 seconds excluded. The constant can be determined a priori by knowing the bottom depth at the instrument site to within 300 m; it may be added back into the recorded travel times after the instrument is recovered. Table 1. PIES Positions, Serial Numbers, Instrument type, Pressure Sensor Serial Number and Rating, and Deployment Depth. P4-1 was not recovered. P3-2 was found later (see text). Site PIES instr. Paros Paros Latitude Longitude depth S/N type S/N rating-psi (N) (E) (m) P1-1 88 SD-URI 31162 6000 38 01.50 129 12.00 1175 P1-2 121 WHISL 51911 10000 37 53.70 129 45.00 1682 P1-3 118 WHISL 50496 10000 37 57.02 130 25.00 1881 P1-4 119 WHISL 50526 10000 37 49.72 130 58.38 1386 P1-5 117 WHISL 50532 10000 37 50.00 132 00.02 2644 P1-6 66 SD-URI 33822 6000 37 53.69 132 42.02 2536 P2-1 23 SD-NRL 18862 10000 37 33.20 129 45.00 1047 P2-2 120 WHISL 50543 10000 37 33.20 130 21.00 1650 P2-3 71 SD-URI 31724 6000 37 33.20 131 14.57 2207 P2-4 68 SD-URI 40275 6000 37 33.20 131 49.97 2382 P2-5 116 WHISL 50530 10000 37 33.20 132 30.00 1696 P3-1 17 SD-NRL 36873 6000 37 04.00 129 56.40 1023 P3-2 70 SD-URI 17849 6000 37 03.40 130 18.72 2214 P3-3 69 SD-URI 33816 6000 37 03.40 130 56.40 2221 P3-4 72 SD-URI 19539 10000 37 03.40 131 40.98 2165 P4-1 107 WHISL 47151 10000 36 30.30 130 03.03 1375 P4-2 77 SD-URI 28197 6000 36 30.30 130 37.32 2038 P4-3 74 SD-URI 19552 10000 36 30.30 131 13.98 2058 P4-4 112 WHISL 53566 10000 36 35.32 131 49.97 1809 P4-5 106 WHISL 50525 10000 36 40.30 132 30.00 1194 P5-1 24 SD-NRL 18860 10000 35 50.27 129 51.51 1036 P5-2 111 WHISL 53565 10000 35 52.00 130 34.02 1402 P5-3 15 SD-NRL 19934 10000 35 57.50 131 15.00 1257 P5-4 86 SD-URI 36884 6000 36 03.00 131 55.98 1165 P5-5 87 SD-URI 36883 6000 36 12.00 132 27.60 1055 Data Recovery The deployment locations of the PIES are listed in Table 1 and plotted in Figure 1. The instruments were deployed on a cruise aboard the R/V Revelle (HAHNARO-06) during June 6 - June 15, 1999. They were recovered during June 21 - July 4, 2001 aboard the R/V Melville (Cook Leg 09). Initially, 23 PIES were recovered and two instruments were not. Acoustic communication was established with P3-2, but it would not release from the bottom. Acoustic communication was not established with P4-1, and it is believed to have been lost to crab fishing activity. Nearl three years after the recovery cruise, in March 2004, the P3-2 instrument was found by a Korean fisherman near Ulleung Island and its data tape was returned to URI. Complete travel time, pressure, and temperature records were obtained for 22 of the 24 recovered instruments. For the two remaining instruments, the travel time record for P1-5 failed after three months and the temperature record for P5-4 failed completely (Figure 2). Time Base The PIES instruments sampled and recorded nominally at one-hour intervals. However, the instruments exhibited varying amounts of clock drift during the two-year deployment. If the clock drifted less than 2 minutes between launch and recovery, the timing associated with each sample was shifted by half the total drift and the one-hour sampling interval was maintained. If the clock rift exceeded 2 minutes, the sampling interval was adjusted to be slightly longer or shorter than one hour, so that the observed times at launch and recovery were assigned to the first and last data records, respectively. All times are reported as decimal day since January 1, 1999 0000 UTC. Thus, January 1, 1999 at 1200 UTC would b given as 0.5. Data Processing and Calibration Data Processing The basic steps in the PIES data processing included recovery of the data from the cassette tapes and conversion of th recorded counts into scientific units. The data processing and low-pass filtering was accomplished by a series of MATLAroutines specifically developed for the PIES. The steps are described below and schematically illustrated in Figures 3 and 6. Sections 2.2-2.7 provide a complete description of the processing schemes applied to the PIES data. New difficulties were encountered during the processing of the PIES data, specifically, pressure (P) and travel time (tau) jumps (offsets) at 11 sites, and complicated sensor drifts at 10 sites (Figures 12-14). The temperature records were not affected. The jumps apparently resulted from the instruments being displaced to different depths by the deep crab fishing activity in the region. The P and tau jumps always occurred simultaneously at an individual site and had relative displacements consistent with vertical movement of the instrument ranging from tens of meters to centimeters, either upward or downward. The most complicated drifts all occurred on pressure sensors with similar serial numbers, suggesting the possibility of a bad batch. These sensors have since been retired. Four of the records had the added complication of a very strong drift at the beginning of the record that trended in opposition to the remainder of the record. In order to produce coherent maps of P and tau, these jumps and drifts needed to be estimated and removed from the records. The dynamically important pressure signal was on the order of 1-3 cm, nearly an order of magnitude smaller than either the dominant basin average signal (~50 cm) or the tidal signal (~15 cm). This dynamic pressure signal was also considerably less than many of the jumps and drifts, which ranged up to 34 dbar and 3-4 dbar, respectively. The smallness of the dynamic pressure signal relative to the basin average, tides, jumps, and drifts required that these other signals be calculated with a high order of accuracy. Temperature Temperature is measured by the PIES in order to account for the temperature sensitivity of pressure sensors. It is measured inside the glass sphere and therefore does not provide an accurate measurement of the instantaneous in situ water temperature. The measurement is taken during a 50-60 sec time window at the end of the hourly sampling period. The time assigned to each temperature measurement is the midpoint of this sampling window. The temperature records from 23 of the 24 PIES required minimal processing to remove spikes (large outliers). The PIES at P5-4 required special processing because the temperature measurement failed. Although the temperature counter of P5-4 functioned normally while the instrument was in the laboratory, after the instrument was deployed only zero values were recorded. In order to accurately determine pressure from the recorded frequencies, it was necessary to generate a temperature record for this instrument. We decided to combine the temperature records from neighboring sites P5-3 and P5-5 to produce the temperature record for site P5-4. We synthesized temperatures by averaging time-shifted and low-pass filtered temperatures from these adjacent sites. We shifted the temperatures from P5-3 and P5-5 in opposite directions, to account for signal propagation, for time lags ranging between 0 and 50 days. We also low-pass filtered the records using a suite of cutoff periods ranging from 0 to 75 days. Since the dynamically-important pressure signals were small, it was important to minimize the noise introduced by the choice of temperature signal. Initially we attempted to minimize the RMS of the temperature record because it did not require us to recalculate the pressures. However, we decided it was best to minimize the RMS of the pressure record itself, because the pressure record was the data we were interested in studying. Thus, a pressure record was generated for each temperature record synthesized. The smallest RMS (0.008782 dbar) was obtained using the temperatures synthesized by averaging the records from P5-3 and P5-5 that were time shifted by 24 days and low-pass filtered using a cutoff period of 65 days (Figure 83). The hourly temperature records are shown in Figures 78-83 and basic statistics are given in Table 7. Bottom Pressure All pressure measurements were corrected for the temperature sensitivity of the sensor using calibration coefficients supplied by the manufacturer. Since the temperature record failed at site P5-4, the pressure record was corrected using a proxy temperature record, which is described in the above section. The raw pressure records are shown in Figures 12-14. Processing the Jump-Free Pressure Records A simple scheme designed to increase the signal-to-noise ratio for each step, shown schematically in Figure 3, was used to process the pressure records that had no jumps (P1-3, P1-6, P2-2, P23, P2-5, P3-2, P3-3, P3-4, P4-3, P4-5, P5-3, P5-4, and P5-5). The scheme was broken into five steps: 1. Remove large spikes from the raw data, where large was defined as spikes greater in magnitude than 10 dbar. These isolated large spikes were present in most of the records and occurred from internal electronic sources or tape reading errors. These records were referred to as P_DS. 2. Calculate and remove the tidal signal (Detide) from P_DS using Tidal Response Analysis (TRA, described below), where this preliminary tidal signal is called TS_0. 3. Remove the Basin Average (described below), referred to as BA. 4. Calculate and remove the drift curve (Dedrift) from P_DS-TS_0-BA, where the drift curve is referred to as DC. 5. Detide the original despiked record with TRA after removal of the A and the drift curve, i.e. detide P_DS-BA-DC. This generates the greatest signal-to-noise ratio of the tidal signals possible. The improved tidal signal is called simply TS. Detiding Tidal response analysis (TRA) (Munk and Cartwright, 1966) was used to determine the tidal constituents from the pressure records of each instrument. In orer to enhance the signal-to-noise ratio of the tidal signal, which was less than a millimeter for some constituents, the BA and drift curves (DC) were removed from the pressures prior o running the TRA for the final time. The amplitude and phase of the constituents for each instrument are given in Table. To illustrate the tidal variation between sites, a 65-day segment of the tidal signals for all 24 sites are plotted in Figure 15. Table 2. Amplitudes (H) and phases(G) of the eight major tidal constituents. Amplitudes are given in millibars and Greenwich Epoch phases are given in degrees. Site O1 K1 Q1 P1 M2 K2 N2 S2 P1-1 H (mbar) 4.670 4.895 1.013 1.675 6.537 0.455 1.511 1.796 G (deg) 193.514 226.517 173.788 224.712 183.798 201.626 166.643 202.079 P1-2 H (mbar) 4.629 4.801 1.009 1.644 5.986 0.384 1.406 1.535 G (deg) 192.174 224.452 172.831 222.724 182.390 198.889 164.323 199.893 P1-3 H (mbar) 4.707 4.851 1.039 1.661 5.692 0.359 1.368 1.431 G (deg) 189.925 221.175 170.961 219.626 178.634 192.540 160.745 194.168 P1-4 H (mbar) 4.754 4.847 1.020 1.665 5.248 0.316 1.298 1.257 G (deg) 188.003 218.600 169.541 216.899 172.992 181.783 156.237 184.374 P1-5 H (mbar) 4.927 5.042 1.085 1.724 5.444 0.346 1.348 1.368 G (deg) 186.499 215.765 169.064 214.309 169.841 180.094 152.359 182.477 P1-6 H (mbar) 5.008 5.140 1.095 1.758 5.655 0.381 1.405 1.488 G (deg) 185.714 214.501 168.687 213.029 167.786 180.245 150.855 182.073 P2-1 H (mbar) 4.524 4.590 0.991 1.576 5.567 0.329 1.329 1.334 G (deg) 191.779 224.252 172.487 222.493 182.946 199.009 163.882 200.491 P2-2 H (mbar) 4.590 4.683 1.015 1.606 5.182 0.289 1.279 1.173 G (deg) 189.405 220.512 170.017 219.021 177.795 187.292 158.563 190.465 P2-3 H (mbar) 4.750 4.831 1.051 1.654 5.000 0.282 1.255 1.131 G (deg) 186.827 216.787 168.781 215.326 172.517 178.607 154.437 182.262 P2-4 H (mbar) 4.883 4.974 1.077 1.702 5.057 0.308 1.285 1.214 G (deg) 185.722 214.903 168.206 213.473 167.940 172.859 150.611 176.395 P2-5 H (mbar) 4.950 5.077 1.094 1.733 5.415 0.353 1.375 1.377 G (deg) 185.721 214.132 169.169 212.723 165.584 173.207 148.670 176.019 P3-1 H (mbar) 4.362 4.336 0.954 1.495 4.474 0.176 1.099 0.778 G (deg) 190.042 221.911 170.215 220.252 182.882 190.361 161.588 195.573 P3-2 H (mbar) 4.217 4.144 0.930 1.430 3.928 0.144 0.997 0.628 G (deg) 186.835 216.692 167.433 215.320 174.741 176.605 157.792 182.537 P3-3 H (mbar) 4.499 4.514 1.004 1.548 4.199 0.194 1.106 0.780 G (deg) 184.684 214.623 166.347 213.229 171.010 159.821 151.461 168.493 P3-4 H (mbar) 4.876 4.894 1.085 1.677 4.481 0.256 1.196 0.987 G (deg) 182.925 211.819 165.246 210.484 162.997 153.267 145.541 159.788 P4-2 H (mbar) 3.948 3.804 0.900 1.309 3.021 0.116 0.865 0.366 G (deg) 184.764 213.270 166.609 212.162 172.648 104.588 149.919 124.359 P4-3 H (mbar) 4.362 4.435 0.971 1.516 3.454 0.209 0.992 0.729 G (deg) 176.322 205.066 159.054 203.726 158.483 117.823 141.051 127.624 P4-4 H (mbar) 5.159 5.083 1.163 1.745 4.092 0.270 1.137 0.986 G (deg) 178.134 207.705 159.948 206.392 153.752 131.099 137.969 138.100 P4-5 H (mbar) 5.408 5.213 1.251 1.787 4.636 0.317 1.248 1.181 G (deg) 188.803 212.758 172.295 212.191 152.884 140.691 138.641 145.824 P5-1 H (mbar) 3.654 3.062 0.848 1.078 1.529 0.431 0.213 1.434 G (deg) 199.824 237.624 178.765 235.352 268.839 12.257 195.783 6.611 P5-2 H (mbar) 2.708 2.177 0.695 0.756 1.246 0.369 0.605 1.185 G (deg) 184.540 199.715 166.280 201.486 127.720 56.874 121.421 56.051 P5-3 H (mbar) 3.282 4.065 0.727 1.347 2.703 0.331 0.868 1.093 G (deg) 162.584 180.406 152.212 179.775 135.899 86.002 124.749 89.663 P5-4 H (mbar) 6.639 6.086 1.479 2.124 3.964 0.357 1.164 1.242 G (deg) 165.870 202.897 144.187 200.634 139.404 111.060 127.365 115.861 P5-5 H (mbar) 6.426 5.635 1.468 1.973 4.502 0.358 1.253 1.290 G (deg) 180.493 210.394 160.714 209.186 143.917 125.136 131.393 129.925 Basin Average The bottom pressure records of all 24 PIES in the Ulleung Basin exhibited a remarkably similar, coherent signal. This coherent signal was determined to be in response to atmospheric forcing (Park and Watts, 2003). The basin-wide average signal was constructed from three pressure records (P2-3, P3-3, P5-5) that displayed little sensor drift, good signal-to-noise ratio, and no jumps. Prior to averaging, the three records were demeaned, detided with TRA, and dedrifted with anexponential/linear scheme described in Watts and Kontoyiannis (1990). The mean of the three resultant records was calculated to determine the basin-wide signal (BA) plotted in Figure 4. Dedrifting The determination of best-fit drift curves followed a hierarchal order of complexity from a single linear function to double exponential linear functions. In each case, drift curves were calculated in an ascending order of complexity until a suitable fit was found. By doing this we ensured that the simplest function of time was used to calculate the drift curves. Three of the jump-free records (P1-6, P3-4, P4-3) and seven of the records with jumps (P1-1, P1-2, P1-4, P2-1, P4-2, P4-4, and P5-1) required the drift curves be calculated piece-meal due to the complexity of the drifts involved. In all cases, te rate of drift decayed as a function time and was approximated by the following functions, Drift=Et+F Drift=C exp^{-Dt}+Et+F Drift=Aexp^{-Bt}+C exp^{-Dt}+Et+F, depending on the complexity of the drift. The overdetermined sets of equations were solved for A, B, C, D, E, and F with the MATLAB function fmincon, which uses a constrined nonlinear optimization method. Drifts were fitted to residual pressure records that had the record mean, basin average, and tidal signals removed ( P_{DS}-TS_0-BA ). The fitted dift curves for each instrument are shown in Figures 16-18 superimposed on the P_{DS}-TS_0-BA data. P1-1, P1-2, P1-4, P6, P2-1, P3-4, P4-2, P4-3, P4-4, and P5-1 required fitting distinct drift curves to different portions of the records. We did this because the nature of the drifts, for undetermined reasons, changed character and could not be fitted with a single function. Figure 5 shows the drift curve calculated for site P1-6. The drift changes character near hour 1000. A double exponential/linear function was used prior to hour 1000, and a single exponential/linear was used after hour 1000. To be sure that the drift curves matched at hour 1000, a linear function between hours 850 to 1000 was used to ensure continuity between the two fits. Similar techniques were used on the remaining sites. Processing the Pressure Records with Jumps The pressure records from the remaining 11 sites (P1-1, P1-2, P1-4, P1-5, P2-1, P2-2, P3-1, P4-2, P4-4, P5-1, P5-2) contained jumps (Figures 12-14). A combined total of 32 jumps ranging up to 34 dbar in magnitude (tabulated in Table 3) were identified in these records. Special care was required during the processing to identify and remove these jumps correctly. The processing scheme of the records with jumps, outlined in Figure 6, was very similar to the processing described above for records without jumps. The processing sequence was as follows: 1. Remove large spikes from the raw data, where large is defined as outliers greater in magnitude than any jumps present. These spikes were not identified and removed by the standard despiking routine, instead new code was written. This made certain that the jumps and their adjacent data points would not be mistakenly identified as spikes. Records with the large spikes removed are referred to as P_{DS}. 2. Perform Iteration 1 of the dejump routine, referred to as P_{DJ1} , described below. 3. Calculate the preliminary tidal signal, TS_0, using TRA. 4. Remove BA and TS_0 from the first iteration dejumped record, P_{DJ1}-TS_0-BA. 5. Perform Iteration 2 of the dejump routine, called P_{DJ}, described below. 6. Calculate the drift curve DC detailed above and remove it, as P_{DJ}-BA-DC. 7. After adding TS_0 back in, detide again with TRA to find the improved tidal signal, TS. Jump Identification and Removal Each jump occurred within a relatively short time period of four hours or less. Over this time interval, the tidal signals dominated, allowing us to use the periodicity of the tides as a means of identifying the timing and size of the jumps. Once the initial jump estimates were removed, the preliminary tidal signal ( TS_0 ) was calculated and removed from the pressure records. Additionally, the BA was removed to further reduce true ocean signal variability. The resulting residual pressure records were expected to vary negligibly over the short jump time interval, allowing us to set all points within the jump equal to the sample just prior to the jump and adjust the remainder of the record accordingly. The process used to remove the jumps will be described next using the pressure record from site P3-1. The jump used in this illustration had a magnitude of 2.1844 dbar (Figure 7, top panel). All other jumps in this record as well as at the other sites were removed with this same scheme. Iteration 1. The purpose of this step is to eliminate jumps from the pressure records reasonably well, so that the tidal signals can be determined with the response analysis program. The hourly measurements affected by the jumps are identified in the P_{DS} data record. These are depicted by two open circles in the top two panels of Figure 7 and will be referred to here as J1 and J2, respectively. Measurements that occurred 24 hours prior to J1 and J2 are also found. These points,referred to as P1 and P2, are marked by two open diamonds in the middle panel of Figure 7. Because the tides are periodic, P1 and P2 should be in approximately at the same phase of the tide as J1 and J2. Thus, P1 and P2 can be used to estimate the amount of variability due to the tides that would be expected at J1 and J2. The size of the tidal signal at measurement J1 is estimated as follows. The pressure difference between P1 and the measurement immediately prior to it (P0) is calculated as Delta p_1 = P1 - P0 . This difference is assumed to be caused solely by the tides. The expected value at J1 is estimated as DJ1=J0+Delta p_1 , where J0 is the measurement taken just before J1. The same procedure is applied to J2: estimate Delta p_2 = P2 - P1 and apply it to give DJ2=DJ1+Delta p_2 . The final step is to calculate Delta_{jump}=J2-DJ2 and remove from all points in the record after point DJ2. This procedure is followed each jump identified in the record. The resulting jump-free record is referred to as P_{DJ1} . This tidal signal TS_0 was determined with TRA from the P{DJ1} records. Iteration 2. The purpose of this step is to produce pressure records with the largest ocean signals removed so that jumps would be eliminated with more confidence. he TS_0 and BA signals were removed from the despiked record P_{DS} to produce a record with small residuals, P_{DSTS_0-BA . This residual record contains jumps, the drift, and dynamically important pressures, as well as the errors from the tidal and basin average calculations. We make the assumption that the dynamic signals were static over the course of a jump (1-4 hours). This permitted us to set the measurements during the jump ( J1 and J2 ) to be equal to the last value prior to the jump ( J0 ). The magnitude of the jump was then calculated as the value of the last measurement involved in the jump prior to correction minus its corrected value ( Delta = J2 - J0 ). All subsequent measurements in the record were then adjusted by the value of Delta . This process is depicted by asterisks in the bottom panel of Figure 7. Travel Time Processing and Calibration Raw Time Series and Biological Interference During each hourly sample burst, the PIES emitted a set of 24 pings and the times of the returning echoes were recorded internally. Figures 46-48 display all the echo returns (tau) as individual dots. The desired surface reflections appear as a continuous line in each panel. Jumps, caused when the instruments were jostled by fishing activity, are clearly visible as abrupt offsets in these traces. In addition to the surface returns, many of the records exhibited a large amount of scatter. Most of the scatter was attributable to echoes that returned early. An enlargement of the scatter at site P2-2 (Figure 8) shows a diurnal pattern to the echo returns. The earliest echo returns consistently occurred during the daytime hours (note that local time leads UT by 9 hours). The echoes became progressively longer (closer to the sea surface) during the evening hours. This pattern is consistent with vertical migration of zooplankton. It is conjectured that the pings emitted from the PIES reflected off the high concentrations of fish or squid which followed the food source. The records in Figures 46-48 show more activity at sites nearer the coasts than in the center of the basin. They also show more activity during the first year of the deployment period. The 24 travel times taken each sampling period are first windowed to remove the largest scatter. From the remaining values, a single representative tau was chosen by using an average of several points near the first-quartile value of a Rayleigh distribution. After jump removal and low-pass filtering, the final calibration step was to project the measured travel times onto a common pressure level. Removal of Travel Time Jumps The 11 sites with pressure record jumps described above also have corresponding travel time jumps that must be removed. The travel time jump is directly proportional to the pressure jump and calculated as Delta tau=2(Delta p/1.01) c^{-1} , where c is the speed of sound at the bottom and Delta p/1.01 expresses the identified pressure jump as a depth change. We use the sound speed at the bottom, because this is where the extra distance traveled by the sound pulse occurs. The process is illustrated in Figure 9 for site P1-2. The four circled points represent a tau jump, and are referred to as J1 through J4. First, Delta tau is subtracted from J4 and all subsequent measurements. Points J1-J3 are replaced with values linearly interpolated between J0 and DJ4, where DJ4 is the corrected J4 value and J0 is the measurement just prior to J1. This procedure is repeated for all jumps. Calibration to tau_500 It is convenient to have all the travel time data referenced to a common pressure level for subsequent scientific analyses. The calibration steps employed here primarily follow the methods developed by Meinen and Watts (1998). Because baroclinicity in the JES below 500 dbar is weak, and because most CTD casts extend only that deep, 500 dbar was selected as the common pressure level. To determine the projections, we used hydrographic data from several sources: Japan Oceanographic Data Center, Korea Oceanographic Data Center, Northwestern Pacific Hydrobase (MacDonald et al., 2001), Lynne Talley's hydrographic surveys (http: //sam.ucsd.edu/onr_jes/) and casts collected on our recovery cruise. Each hydrocast was interpolated to a common pressure vector with 10 dbar spacing. The casts were then integrated from a suite of deep pressure levels to the surface to simulate the round trip travel times ( tau_p ) that would be attained from instruments deployed at those depths. For this study, tau_p was simulated for bottom pressures between 600-2700 dbar. Linear relationships were found for the tau_p versus tau_{500} comparisons for all depths (Figure 10). The slopes A and intercepts B of these relationships were determined. However, since the variations in tau_p are on the orders of milliseconds while the magnitudes are on the order of several seconds, large computational errors can occur when determining these parameters. To minimize the errors, A and B were obtained after removal of a large constant from the travel times. This constant, tau_{con}, was defined as the round trip travel time that would be measured if the sound speed were fixed at 1500 ms^{-1}: tau_{con}(p)=\frac{2(\frac{p}{1.01})}{1500} where the factor 1.01 dbar m^{-1}converts bottom pressure in decibars to depth in meters. Different slopes and intercepts were obtained for the relationships between tau_{500} and tau_p for 600 <= p <= 2700 dbar. Figure 11 shows the values obtained for A, b, and B versus pressure. The parameter b is the intercept value with the large constant removed, while {\bf B} is the full intercept value. No physical justification could be determined for the observed structure in A for p > 2300 dbar. Instead this structure was attributed to the limited number of hydrocasts extending to deeper levels (i.e. sampling error). Superimposed on the top two panels are the best fit curves: A(p)= a_1p^4+a_2p^3+a_3p^2+a_4p+a_5 for p <= 2300 dbar 9.7547 * 10^{-1} for p>2300 dbar b= b_1 * p + b_2 where a_1=1.6944* 10^{-14}, a_2=-1.1256* 10^{-10}, a_3=2.6970* 10^{-07}, a_4=-2.7879* 10^{-04}, a_5=1.08353, b_1=-3.8869*10^{-6}, b_2=-6.5050*10^{-03} . The value of A for pressures below 2300 db are forced to equal that obtained for P=2300 dbar to avoid introducing nondynamical variability. The intercept B is defined as B(p)= -A(p) * p/750 + 500/750 +b where 500/750 is tau_{con}(p) for p=500m (removed from tau_{500} ). All the PIES deployed in the JES were equipped with pressure sensors, and the mean pressure measured at each site was used in the above formulas. The mean values were calculated from the records that had been dejumped, detided and dedrifted. Before applying the formulas, additional adjustments to the pressure and travel time records were required. First, the mean regional atmospheric pressure of 10.14 dbar was removed, since the pressure sensors measured absolute (not gauge) pressure. Second, 0.6 dbar was subtracted to account for the physical separation of the pressure sensor and transducer on the PIES. Third, 3.0 msec were subtracted from the measured travel times to account for an internal response delay of the echo detector. The travel time measured by each instrument ( tau_{PIES}) was projected onto tau_{500} using tau_{500}=A(p)*tau_{PIES}+{\bf B}(p) where P is the mean measured pressure, as determined above. Due to the pressure jumps and sensor drifts, the accuracy of the mean pressures is questionable and may introduce errors into tau_{500} So further calibration of tau_{500} was necessary. This was accomplished through the use of CTD casts taken at each PIES site approximately 24 hours prior to their recovery and from CTD casts taken at some sites by the National Fisheries and Research Development Institute (NFRDI) and the Japanese Oceanographic Data Center (JODC). There were a total of nine CTD casts taken in August, November, and December 1999, February, April, June, August, October, and December 2000 over 10 sites (P1-2, P1-3, P1-4, P2-2, P2-3, P3-1, P3-2, P3-3, P4-2, and P4-3) by NFRDI. There were two CTD casts taken in June 2000 and June 2001 over two sites (P4-4, P5-5) by JODC. For each CTD cast, tau_{500} was calculated from the temperature and salinity profiles. The calibration coefficient was then calculated as C_{CTD} = \overline{tau_{500_{PIES}}-tau_{500_{CTD}}} , where the overbar means the average of all available CTD measurements at the site (except for P1-5 which required special calibration, described below). The calibration coefficients are shown in Table 4. The best estimate of tau_{500} is then, tau_{500}=A(p)*tau_{PIES}+ B(p)+C_{CTD}. Table 4. tau_{500} calibration coefficient C_{CTD} calculated from CTDs taken at the PIES sites. Site C_{CTD} Site C_{CTD} P1-1 2.1815 ms P3-2 -0.0028 ms P1-2 1.5096 ms P3-3 -0.3165 ms P1-3 2.9111 ms P3-4 10.3334 ms P1-4 1.6807 ms P4-2 9.8386 ms P1-5 2.2495 ms P4-3 6.6115 ms P1-6 -0.6929 ms P4-4 6.4957 ms P2-1 16.4823 ms P4-5 0.6740 ms P2-2 3.6025 ms P5-1 5.4287 ms P2-3 -1.2967 ms P5-2 2.9832 ms P2-4 1.3002 ms P5-3 12.9066 ms P2-5 3.8464 ms P5-4 2.7802 ms P3-1 -1.2226 ms P5-5 2.2836 ms Site P1-5 required a different method of calibration because the tau measurement failed after three months. This rendered the CTD collected 24 hours prior to its recovery useless for calibration. We chose to calculate C_{CTD} for P1-5 by using the calibrated data from the adjacent sites. First we calculated the mean tau_{500} at P1-5 over the 90-day period it operated. We also calculated the mean tau_{500} (calibrated using C_{CTD} ) of sites P1-4 and P1-6 over the same time period. The calibration coefficient C_{CTD} for P1-5 was calculated by linearly interpolating these means as: C_{CTD,P1-5}=\overline{tau_{500}(P1-4, 90 d)}\frac{Delta x_{P1-5}}{Delta x_{P1-6}}+ \overline{tau_{500}(P1-6, 90 d)}\frac{Delta x_{P1-6}-Delta x_{P1-5}}{Delta x_{P1-6}} - tau_{500}(P1-5, 90 d) where Delta x is the distance of sites P1-5 and P1-6 from P1-4. Expected errors in tau_{500} arise from uncertainties in the tau and pressure measurements and the processing methods. The accuracy of the travel time measurement, which is 1 ms (Chaplin and Watts, 1984), contributes an uncertainty in tau_{500} after 120 hour low-pass filtering, of epsilon_{tau}=1 ms / sqrt{120}=0.09 ms. An additional error arises from the projection to 500 dbar from the actual bottom pressure. This error is due to the scatter about the A and B versus pressure curves shown in Figure 11. The combined error epsilon_{AB} was determined by quantifying the scatter of the tau_{500} versus tau_{p} relationships shown in Figure 10. The scatter ranged from 0.1 to 0.4 ms, thus, we set epsilon_{AB} =0.4 ms. The accuracy of the pressure sensor is given as +/- 0.01 % of the sensor's full scale. Sensors with two different full scales were used for this experiment: 6000 psi (~4000 dbar) and 10000 psi (~ 6800 dbar), giving pressure sensor accuracies of 0.40 and 0.68 dbar. Assuming the average sound speed is 1500 m s ^{-1} , these yield travel time uncertainties epsilon_{ps} of 2(0.4)/1500 and 2(0.68)/1500 , or about .5 ms and 1 ms. Additional sources of error in the pressure records due to the removal of pressure jumps and drifts must also be determined. The minimum accuracy of the method used to remove the pressure jumps is determined by the scatter in the pressure record immediately before and after the jumps, which is typically 0.01 dbar. This gives a travel time uncertainty due to pressure jumps epsilon_{pj}=0.013 ms. The uncertainty arising from removing the drift is difficult to quantify, however, a reasonable estimate is possible due to the dominance of the basin average signal (BA). The final pressure records, which have been detided, dejumped (if necessary), dedrifted, and low pass filtered to remove the error associated with epsilon_{ps} , should only contain the basin average and the remaining ocean signal. Since we do not know the ocean signal a priori, we assume its long term mean is zero and take the mean of the final pressure record minus the basin average as the uncertainty of the drift removal, that is \overline{p_{fin}-BA}. This yields a typical uncertainty for the drift removals of 0.0012 dbar or epsilon_{pd}=10^{-6} ms. Thus, the total uncertainty of tau_{500} is epsilon_{total}=(epsilon_{tau}^2+epsilon_{AB}^2+epsilon_{ps}^2+epsilon_{pj} ^2+ epsilon_{pd}^2)^{\frac{1}{2}} = 0.65 ms for the 6000 psi sensor and 1.1 ms for the 10000 psi sensor. Bias errors may exceed these random errors: Echo detector times may have varied between these old model instruments by 2 ms. Pressure sensors may have drifted in calibration by a few decibars, equivalent to a few milliseconds. Some instruments changed depth. These latter errors account for much of the C_{CTD} offsets in Table 4. Leveling and Low-Pass Filtering The hourly pressure records were also leveled to a common geopotential using the mean currents measured by the current meter moorings following the procedures described in Watts et al. (2001). Leveled records are obtained by subtracting the mean pressure and adding the leveling constants listed in Table 5. The leveled pressure and tau_{500} records were low-pass filtered using a fourth-order Butterworth filter with a cutoff period of 120 hours. The filter was passed forward and backward in time to avoid introducing phase shifts. Approximately 40 hours of data at each end of the filtered records were discarded to avoid startup transients. After filtering, the time series were sampled at 12-hour intervals centered at 0000 and 1200 UT. The low-pass filtered pressures are shown in Figures 22-45 together with the hourly leveled pressures. The low-pass filtered tau_{500} records are in Figures 72-77. Acknowledgments We gratefully acknowledge the efforts of the crews of the R/V Revelle and the R/V Melville for their efforts during the deployment and recovery cruises. The successful deployment and recovery of the inverted echo sounders is due to the instrumentation development and careful preparation done by Gerard Chaplin and Michael Mulroney. Special thanks goes to Angela Adams, a SURFO student at URI during Summer 2001, who helped with the initial data processing. We appreciate the current meter data supplied by Dr. Moon-Sik Suk and Dr. K.-I. Chang (KORDI), and Dr. J-H Yoon (RIAM) which were used to level the bottom pressure measurements. The CTD data sets were kindly provided by Dr. Lynne Talley (SIO), Dr. Alison MacDonald (WHIO), the Korea Oceanographic Data Center, the National Fisheries and Research Development Institute (Korea), and the Japan Oceanographic Data Center. The work of D. R. Watts and M. Wimbush was supported by the Office of Naval Research Japan/East Sea DRI through contract N00014-98-10246. The work of W. Teague was supported through the Naval Research Laboratory's Linkages of Asian Marginal Seas system (O601153N). References Chaplin, G. F., and D. R. Watts. 1984. Inverted ehco sounder development. IEEE Ocean '84 Conf. Record, Washington, DC, IEEE, 249-253. MacDonald, A. M., T. Suga, and R. G. Curry. 2001. An isopycnally averaged North Pacific climatology. J. Atmos. Oceanic Technol., 18, 394-420. Meinen, C. S. and D. R. Watts. 1998. Calibrating inverted echo sounders equipped with pressure senors. J. Atmos. Oceanic Technol., 15, 1339-1345. Munk, W.H. and D.E. Cartwright. 1966. Tidal spectroscopy and prediction. Philos. Trans. R. Soc. London, 259, 533-581. Park, J.H., and D. R.Watts. 2004. Response of the southwestern Japan/East Sea to the atmospheric pressure. Deep-Sea Res., [accepted]. Watts, D. R., and H. Kontoyiannis. 1990. Deep-ocean bottom pressure measurement: drift removal and performance. J. Atmos. Oceanic Technol., 7, 296-306. Watts, D. R., X. Qian, and K. L. Tracey. 2001. Mapping abyssal current and pressure fields under the meandering Gulf Stream. J. Atmos. Oceanic Technol., 18, 1052-1067. Xu, Y., K. L. Tracey,D. R. Watts, M. Wimbush, W. Teague, and J. Book. Current meter data report Ulleung Basin of Japan/East Sea June 1999 to July 2001. University of Rhode Island. GSO Technical Report 2003-02. 118 pp.