1. Introduction
Acoustic Doppler velocity profilers (ADCPs) mounted on CTD rosettes—so-called lowered ADCP (LADCP) systems—are routinely used to collect velocity profiles in the ocean. LADCP data have been processed for horizontal velocity for over two decades (Fischer and Visbeck 1993). More recently, a method has been developed to obtain vertical ocean velocity as well (Thurnherr 2011). LADCP-derived velocities can be used directly, for example, for circulation studies (e.g., Thurnherr et al. 2011; St. Laurent et al. 2012). Importantly, LADCP velocities can also be used to estimate turbulence and mixing levels using so-called finestructure parameterization methods (Gregg 1989; Polzin et al. 2014; Thurnherr et al. 2015).
Lowered ADCP work puts high demands on the instruments. For horizontal velocity, the ADCP measurements must be sufficiently accurate so that the errors in vertical shear, integrated over the full profile depths, do not exceed a few centimeters per second (cm s−1; Firing and Gordon 1990). For vertical velocity, the ADCP measurements must be sufficiently accurate to yield signals of a few millimeters per second (mm s−1) from a platform moving up to
Here, we analyze two LADCP datasets that were collected with additional external accelerometer/magnetometer measurements (section 2). Large instrument tilts due to very strong upper-ocean currents adversely affect many of the profiles in the first set, collected in the northeastern Gulf of Mexico (section 3). When processed with the external attitude measurements, the differences between the corresponding LADCP and shipboard ADCP (SADCP) velocities decrease by
2. Methods
a. Magnetometer and accelerometer measurements
A simple instrument called the Independent Measurement Package (IMP; Fig. 1) was built using a datalogger connected to inexpensive magnetometer/accelerometer chips that are readily available as robotics components [so-called inertial measurement unit (IMU) breakout boards]. In its present configuration, the IMP collects magnetometer/accelerometer data from two microchips manufactured by STMicroelectronics: the LSM303DLHC and the similar, but somewhat more recent, LSM303D. The IMP records 100-Hz time series of all three components of acceleration and the magnetic field strength in a coordinate system that is aligned with the sensor chips. In a first step, the data are despiked with a five-wide (1/20 s) median filter and bin averaged to 5 Hz, primarily to reduce file size. Next, the data are low passed with a simple frequency-domain filter with a 2-s cutoff, because high-frequency motion is highly damped underwater.
Schematic of the IMP. CPU and data storage are provided by a Raspberry Pi microcontroller running the Arch Linux operating system and public domain firmware that is available on request. Several peripherals are attached to the CPU via a simple two-wire interintegrated circuit (I2C) bus: 1) A real-time clock (Macetech ChronoDot), 2) Robotics breakout boards based on the LSM303DLHC and LSM303D accelerometer/magnetometer chips (labeled IMU1 and IMU2, respectively), and 3) A 128-byte EEPROM; microchip 24AA02E48) for sensor configuration and usage logging. In the most recent incarnation of the IMP, the components on the gray background are housed in a separate small pressure case that can be mounted away from any magentic disturbances.
Citation: Journal of Atmospheric and Oceanic Technology 34, 8; 10.1175/JTECH-D-16-0258.1





b. Compass calibration







c. Merging ADCP data with external attitude measurements
To calculate the replacement values for the ADCP attitude data from external measurements, the relative alignment of the external sensors with respect to the ADCP transducer must be known. Here, the offset angles are calculated from in situ profile data with the following simple algorithm:
For both instruments (ADCP and IMP), subtract the mean instrument tilts from the measured data, that is, replace pitch and roll with their temporal anomalies.
Using the pitch/roll time series from both instruments, calculate the corresponding time series of the instrument tilt angle (from vertical) and azimuth (heading).
Use temporal lag correlation to determine the clock difference between the corresponding time series of tilt magnitude, which are independent of the heading offset between the instruments.
Determine the heading offset between the IMP and the ADCP pitch/roll sensors from the differences between the two corresponding tilt–azimuth estimates.
Use this heading offset to rotate the external pitch/roll measurements into the coordinate frame of the ADCP. (The differences between the rotated mean tilts give the pitch/roll offsets of the external accelerometers with respect to the ADCP, but these are not required.)
Construct replacement time series for ADCP pitch and roll by adding the rotated external pitch/roll anomalies to the corresponding ADCP means determined in step 1; construct a replacement time series for ADCP heading by adding the heading offset determined in step 4 to the external heading time series.
To avoid having to modify multiple LADCP processing software packages to work with the external attitude data, “patched” binary ADCP data files are created by replacing the pitch, roll, and heading data with the corresponding replacement time series from the IMP. Ensembles without valid external heading measurements are effectively removed from the data files by marking the corresponding velocity measurements as invalid. For LADCP data collected in Earth coordinates, before processing the velocities also have to be transformed back to beam coordinates, which is accomplished by inverting the rotation matrices of the beam-to-instrument and instrument-to-Earth transformations (RD Instruments 1998).
d. LADCP data processing and quality control
The LADCP data are processed for horizontal velocity with the LDEO_IX implementation of the velocity inversion method (Visbeck 2002). The profiles are processed without the SADCP referencing constraint, that is, using only the ship drift (GPS) and bottom tracking (where available) to constrain the barotropic velocities. With this processing, rms differences between the LADCP profiles and simultaneous on-station SADCP velocities from the upper ocean can be used to quantify the uncertainty of the LADCP measurements (Thurnherr 2010). High-quality LADCP datasets typically have LADCP–SADCP velocity differences between 2 and
3. ECOGIG EN586 data
In July 2016, 42 LADCP/CTD/IMP profiles were collected in the northeastern Gulf of Mexico during the R/V Endeavor EN586 cruise of the Gulf of Mexico Research Initiative (GoMRI)-funded Ecosystem Impacts of Oil and Gas Inputs to the Gulf (ECOGIG-2) program. Two Teledyne RD Instruments (TRDI) 300-kHz Workhorse ADCPs, recording beam-coordinate velocities in 6-m bins without blanking, were installed on the CTD rosette together with an IMP. A TRDI 75-kHz Ocean Surveyer SADCP measured the velocity field in the upper
Differences between uplooker and downlooker measurements of (top) heading, (middle) pitch, and (bottom) roll from an example profile from the ECOGIG EN586 cruise showing downcast data (red samples) and upcast data (blue samples). (left) Processed with ADCP attitude data. (right) Processed with IMP attitude data.
Citation: Journal of Atmospheric and Oceanic Technology 34, 8; 10.1175/JTECH-D-16-0258.1
Compass calibration was carried out by subtracting visually determined
Postcalibration horizontal magnetic field data from a yo-yo cast (three profiles) that includes the profile shown in Fig. 2. Shown are the calibration circle (blue dots) and horizontal field strengths within 20% of the calibration circle (red samples); green samples have field strengths with greater deviations. (left) From (older) the LSM303DLHC chip, and (right) from the LSM303D chip.
Citation: Journal of Atmospheric and Oceanic Technology 34, 8; 10.1175/JTECH-D-16-0258.1
A positive consequence of the large tilt angles in the EN586 profiles is that the relative instrument alignment is very tightly constrained by pitch and roll data. The profiles from each of the three magnetometer calibrations (between battery changes) were combined to determine the mean heading offsets between the instruments for each installation of the IMP. For the LSM303D chip, the corresponding standard errors lie between
Using the instrument alignment offsets to construct replacement heading time series for the ADCPs yields estimates for the heading-dependent compass errors of the two instruments (Fig. 4). These estimates indicate that the heading differences shown in Fig. 2 are dominated by errors in the uplooker compass.
Heading-averaged ADCP compass errors and standard deviations in the yo-yo profile shown in Fig. 3; headings collected at instrument tilts
Citation: Journal of Atmospheric and Oceanic Technology 34, 8; 10.1175/JTECH-D-16-0258.1
Significantly improved consistency between the data from the two ADCPs is readily apparent when processing the EN586 profiles with external attitude measurements. In particular, there are no longer any heading-dependent compass offsets, and the pitch and roll differences show reduced scatter (Fig. 2). In many profiles, there is less spatial structure in the inversion residuals from horizontal-velocity processing (Visbeck 2002) when external attitude data are used, indicating that the measurement errors are more random (not shown).
More importantly, the LADCP velocities processed with external attitude data agree more closely with the corresponding SADCP velocities than the original profiles (Fig. 5). Averaged over the entire dataset, external attitude measurements improve the rms differences between the LADCP and SADCP velocities by
Histograms of the profile-averaged rms differences between the corresponding upper-ocean LADCP and SADCP velocities in the ECOGIG data.
Citation: Journal of Atmospheric and Oceanic Technology 34, 8; 10.1175/JTECH-D-16-0258.1
The LADCP-derived vertical velocities in the ECOGIG dataset also improve when processed with external attitude measurements, noting that only pitch and roll matter in this case because vertical velocity does not require any heading data (Thurnherr 2011). In case of the ECOGIG profiles, the vertical-velocity differences between the two instruments decreases by
4. NABOS 2015 data
In September 2015, 70 LADCP/CTD/IMP profiles were collected in the Arctic Ocean along the Russian margin of the Nansen and Amundsen basins (
Heading time series from NABOS profiles (left) 12 and (right) 27. ADCP-derived headings (red) and IMP-derived headings (blue). Horizontal geomagnetic field strength is printed above each panel; the corresponding field inclinations are
Citation: Journal of Atmospheric and Oceanic Technology 34, 8; 10.1175/JTECH-D-16-0258.1
For the NABOS profiles, magnetometer calibration was again carried out by visually determining
Magnetometer calibration for NABOS profiles (left) 12 and (right) 27. Shown are the calibration circle (blue dots) and horizontal field strengths within 20% of the calibration circle (red samples); green samples have field strengths with greater deviations.
Citation: Journal of Atmospheric and Oceanic Technology 34, 8; 10.1175/JTECH-D-16-0258.1
During collection of the NABOS profiles, neither of the instruments was moved onto the CTD rosette. Excluding two shallow profiles, the mean heading offset between the ADCP transducer and the IMP accelerometer is
Many of the original NABOS LADCP profiles, when processed without external attitude data, show differences exceeding
LADCP vs SADCP velocities in the NABOS data. (left) Velocities of profile 27. Because of magnetic contamination from the ship, there are no valid velocity samples in the top 60 m of the water column; the processing software extrapolates the uppermost valid velocity sample to the surface. (right) Histograms of the rms LADCP vs SADCP velocity differences from all profiles.
Citation: Journal of Atmospheric and Oceanic Technology 34, 8; 10.1175/JTECH-D-16-0258.1
5. Discussion and conclusions
The results presented above indicate that LADCP velocity profiles can be improved significantly by processing with 3D magnetometer and accelerometer measurements made with common microelectromechanical systems (MEMS) sensors. We have used an external self-contained datalogger to record these ancillary measurements, but the same methodology can be applied to data collected with ADCPs that also record 3D magnetometer and accelerometer data. In the case of the ECOGIG profiles, the external attitude measurements reveal large compass errors in the uplooker ADCP as the main reason for the heading differences in the original ADCP data files. When processed with the external attitude measurements, the discrepancies between SADCP and LADCP velocities reduce by 10%–20%. Based on observed compass differences from thousands of additional available dual-headed profiles, we expect similar improvements in other LADCP datasets.
While the improvements in the ECOGIG LADCP data quality is certainly welcome, it is important to note that the improvements are relatively modest, indicating that even without the external attitude data most profiles are of high quality. The main reason why compass errors do not contaminate regular LADCP profiles more fatally is that compass errors are heading dependent and average to zero; package rotation during the casts ensures that the same velocity is sampled at different instrument headings, thus averaging out the compass errors to some degree. Averaging the heading data from the two instruments further mitigates the problem for dual-headed LADCP systems. We conclude that for regular LADCP work external attitude measurements are optional.
The main benefit of using external attitude measurements is that they allow processing of LADCP profiles with bad attitudes. In the case of NABOS, the ADCP heading measurements are invalid because of a combination of a weak horizontal geomagnetic field and heavy sea state (there are many similar unprocessable profiles from the Southern Ocean in the LDEO LADCP data archive). When processed with external attitude data, most of the resulting velocity profiles are of high quality. It is expected that even better profiles are possible with dual-headed LADCP systems.
We envision several further improvements to the instrument and methodology described here. Support for additional magnetometer and accelerometer chips can be easily added to the IMP firmware. For future deployments we plan to add gyroscopes to distinguish instrument tilt from horizontal acceleration, with the eventual goal of replacing the magnetometer with fiber-optic gyroscopes to remove the effects of magnetic disturbances, especially near the sea surface, and to allow collection of LADCP data arbitrarily close to the magnetic poles.
Acknowledgments
Part of this research was made possible by a grant from the Gulf of Mexico Research Initiative to support the Ecosystem Impacts of Oil and Gas Inputs to the Gulf (ECOGIG-2) research consortium. Funding for acquisition of the 2015 Arctic data was provided by NSF (1203473 and 1249133) and NOAA (NA15OAR4310155) under the NABOS-II program. Development of the prototype external magnetometer/accelerometer package (IMP) and methodology was carried out without external funding; extensive testing was carried out during cruises of the NSF-funded DIMES project (OCE-1232962). Participation of Ilona Goszczko on the NABOS cruise was made possible by the Polish National Science Center MIXAR project (2012/05/N/ST10/03643) and by funding from the Leading National Research Centre (KNOW) to the Centre for Polar Studies for the period 2014–18. Overall responsibility for CTD data acquisition and processing by Joe Montoya and Igor Polyakov for the Gulf of Mexico and Arctic CTD data, respectively, is gratefully acknowledged, as is the support by Piotr Wieczorek (IOPAN) for adapting an old ADCP pressure case and power supply for use with the IMP. The ECOGIG EN586 LADCP data are publicly available through the Gulf of Mexico Research Initiative Information and Data Cooperative (GRIIDC; https://data.gulfresearchinitiative.org; doi:10.7266/N7K072BN). The 2015 NABOS LADCP data are available on request from the authors.
REFERENCES
Firing, E., and R. Gordon, 1990: Deep ocean acoustic Doppler current profiling. Proceedings of the IEEE Fourth Working Conference on Current Measurement, IEEE, 192–201, doi:10.1109/CURM.1990.110905.
Fischer, J., and M. Visbeck, 1993: Deep velocity profiling with self-contained ADCPs. J. Atmos. Oceanic Technol., 10, 764–773, doi:10.1175/1520-0426(1993)010<0764:DVPWSC>2.0.CO;2.
Gregg, M. C., 1989: Scaling turbulent dissipation in the thermocline. J. Geophys. Res., 94, 9686–9698, doi:10.1029/JC094iC07p09686.
Hamilton, J. M., 2001: Accurate ocean current direction measurements near the magnetic poles. The Proceedings of the Eleventh (2001) International Offshore and Polar Engineering Conference, J. S. Chung et al., Eds., Vol. 1, ISOPE, 656–660.
Polzin, K. L., A. C. Naveira Garabato, T. N. Huussen, B. M. Sloyan, and S. Waterman, 2014: Finescale parameterizations of turbulent dissipation. J. Geophys. Res. Oceans, 119, 1383–1419, doi:10.1002/2013JC008979.
RD Instruments, 1998: ADCP coordinate transformation: Formulas and calculations. RDI Manual P/N 951-6079-00, 29 pp.
St. Laurent, L., A. C. Naveira Garabato, J. R. Ledwell, A. M. Thurnherr, J. M. Toole, and A. J. Watson, 2012: Turbulence and diapycnal mixing in Drake Passage. J. Phys. Oceanogr., 42, 2143–2152, doi:10.1175/JPO-D-12-027.1.
Thurnherr, A. M., 2010: A practical assessment of the errors associated with full-depth LADCP profiles obtained using Teledyne RDI Workhorse acoustic Doppler current profilers. J. Atmos. Oceanic Technol., 27, 1215–1227, doi:10.1175/2010JTECHO708.1.
Thurnherr, A. M., 2011: Vertical velocity from LADCP data. 2011 IEEE/OES Tenth Current, Waves and Turbulence Measurements (CWTM), IEEE, 198–204, doi:10.1109/CWTM.2011.5759552.
Thurnherr, A. M., J. R. Ledwell, J. W. Lavelle, and L. S. Mullineaux, 2011: Hydrography and circulation near the crest of the East Pacific Rise between 9° and 10°N. Deep-Sea Res. I, 58, 365–376, doi:10.1016/j.dsr.2011.01.009.
Thurnherr, A. M., L. St. Laurent, K. J. Richards, J. M. Toole, E. Kunze, and A. Ruíz Angulo, 2015: Vertical kinetic energy and turbulent dissipation in the ocean. Geophys. Res. Lett., 42, 7639–7647, doi:10.1002/2015GL065043.
Visbeck, M., 2002: Deep velocity profiling using lowered acoustic Doppler current profilers: Bottom track and inverse solutions. J. Atmos. Oceanic Technol., 19, 794–807, doi:10.1175/1520-0426(2002)019<0794:DVPULA>2.0.CO;2.
This assertion was tested with an Xsens IMU that distinguishes between instrument tilt and horizontal acceleration, and that was deployed on a CTD rosette during two casts in rough seas in the Southern Ocean during the 2010 DIMES second U.K. (UK2) cruise (rms horizontal acceleration