COMBINATION OF ACOUSTICS WITH HIGH RESOLUTION OCEANOGRAPHY JÜRGEN SELLSCHOPP Forschungsanstalt der Bundeswehr für Wasserschall und Geophysik Klausdorfer Weg 2-24, 24148 Kiel, Germany E-mail: [email protected] PETER NIELSEN SACLANT Undersea Research Centre, Viale San Bartolomeo 400, 19138 La Spezia, Italy E-mail: [email protected] MARTIN SIDERIUS Science Applications International Corporation, La Jolla, CA 92037,USA E-mail: [email protected] The variability of underwater sound observed on a moving platform in littoral waters is a combination of the effects of temporal and spatial environmental variations, which are intermingled and obscured by the change of the platform position. Sound transmission experiments over a fixed range are an approach to separate different sources of environmental impact on acoustic propagation. Even in a fixed geometry time variability experiment, the most demanding observations are for the spatial distributions of controlling parameters. On a short time scale, the variation of spatial distributions is responsible for acoustic variability, while mean, range averaged conditions may stay unchanged. The instrumentation for monitoring the ocean environment during the ASCOT 01 acoustic trial included moored instruments for the measurement of temperature and current profiles, tidal and surface waves. By means of wave dispersion relations, moored records of ocean variability may be transformed into guessed spatial fine structure. But critical results such as correct slope statistics of sound channel boundaries and iso-velocity surfaces cannot be guaranteed. A 40-sensor CTD chain was continuously towed parallel to the acoustic range. The measured 2-dimensional sound velocity field has 2 m vertical and 4 m horizontal resolution. A primitive model, which propagates rays through measured high resolution sound velocity fields is able to explain observed multipath arrival structures on a vertical hydrophone array. 1 Introduction High quality predictions of the underwater sound channel are desired, since they are the essential input to reliable modeling of the conditions for sonar detection and underwater communication. Environmental parameters of primary concern are the sound velocity profile, the depth and composition of the seabed and the sea state. By these parameters, the ocean volume and the boundaries of the littoral sound channel are covered. Realistically none of these parameters can be prophesied on scales of kilometer or hour orders. The best possible approach, to allow for a realistic environment in acoustic 19 N.G. Pace and F.B. Jensen (eds.), Impact of Littoral Environmental Variability on Acoustic Predictions and Sonar Performance, 19-26. © 2002 Kluwer Academic Publishers. Printed in the Netherlands. 20 J. SELLSCHOPP ET AL. model calculations, is the separation between predictable average conditions and deterministically unpredictable environmental variability, which might however be well described by a statistical parameter such as the surface wave spectrum. Whether small scale environmental variability is treated adequately in acoustic modeling cannot be decided from the comparison of sound measurements with model results, if there are free adjustable parameters in the model, which can be tuned until the results match. If it is impossible to measure all parameters, which have an impact on acoustics, there is always a danger that a parameter with substantial influence is overlooked and a wrong parameter adjusted instead. This becomes especially obvious in reverse modeling, where the restriction on a limited set of tractable parameters forces all deviations of the environment from an assumed predefined state to act into the selected parameter subspace. Consequentially inversely modeled parameters may not be accurate. The range dependence of the results from inverse modeling of bottom parameters [1] is a hint that the treatment of environmental variability in the model is inappropriate. A description of the ambience as accurate as possible, deduced from extensive measurements, is required for a better connection of acoustic variability and uncertainty to ocean variability. Whether or not the variability of a certain environmental parameter can be finally neglected, can only be decided after sensitivity studies have been made with realistic data applied to appropriate models. Small scale fluctuations have a much higher impact on the littoral sound channel than in the open ocean. In a shallow sound channel, multiple interactions with the surface and bottom boundaries influence the energy and direction of the propagated sound. In contrast to stationary bottom scatterers, the rough surface introduces time dependence into acoustic signals. But also in a stratified ocean under summer conditions, when the sound does not interact with the sea surface in motion, time variability of the range dependent sound velocity structure in the water column is able to significantly impact transmissions between fixed locations. Because of the difficulty of measuring the internal variability, its effect on acoustic signals is harder to access than that of surface waves. In order to separate influences of different kinds, it is advantageous to perform experiments with a fixed source and receiver, which would at least keep the range constant and the bottom unchanged. Measurements of acoustic variability should be complemented by measurements of environmental variables, so that the assumptions introduced into acoustic models are minimized. 2 ASCOT 01 experimental setup The acronym ASCOT is nicely translated into “Acoustic Scenario Connected with an Oceanography Trial”, although it originally stands for “Assessment of Skill for Coastal Ocean Transients”. The NATO Research Vessel ALLIANCE spent three weeks in June 2001 for a collaborative effort with Harvard University [2] in Massachusetts Bay and half of the Gulf of Maine. Data from an initial ocean survey were used to initiate the primitive equation forecast model of the Harvard Ocean Prediction System (HOPS). A limited amount of data was acquired for assimilation during the forecast period. The predictive skill experiment was concluded with a verification survey. After the first phase, there was an opportunity to embed ocean acoustics experiments of four days duration, interrupted by one day for logistics. The experimental setup was similar to the ACOUSTICS AND HIGH RESOLUTION OCEANOGRAPHY 21 fixed range experiments on Adventure Bank (Strait of Sicily) two years before [1]. The compulsory second ship was hired from the University of New Hampshire. A site with suitable conditions, which are defined by proximity to the coast, no interference with protected ocean areas and an approximately constant water depth of about 100 m over 10 km range, was identified 30 nautical miles north of Cape Cod, 20 miles east of Cape Ann (Fig. 1). Figure 1. Bathymetry, mooring positions and tow track of the acoustic test site. The bottom was hard and rougher than expected. Soundings were taken every 4 m on a straight line. The bottom contours of Fig. 1 were obtained by swath mapping with 20 m horizontal resolution. At the SE end of the site, a frame was lowered to the ground with a sound source mounted on top. A vertical line array with 64 hydrophones was deployed at distances 5, 2, 10 and 0.78 km from the source tower, each deployment lasting for one day. A suite of instruments was available for environmental monitoring. Three thermistor strings, each equipped with 11 thermistors with 5 m vertical separation, were deployed at the corners of an equilateral triangle with side length 2.35 km. While one standard thermistor string was sampled every 2 minutes, two strings had response times and sample rates of 10 seconds. An upward looking acoustic Doppler current profiler in the center of the triangle was used with 5 minutes averaging interval and sample rate. Sea surface roughness due to wind waves was continuously measured by a wave rider buoy. A meteorological buoy measured wind close to the surface and undistorted by the 22 J. SELLSCHOPP ET AL. ship. A tide gauge, which was combined with the ADCP, was out of range. Qualitatively the measurements compare well with tides in Boston, which they precede by less than 30 minutes. In addition to the moored instruments, a CTD chain [3] was used for the direct measurement of spatial structures. It was deployed from ALLIANCE and handed over to the GULF CHALLENGER for continuous coverage of a 10-km track parallel to the acoustic range. The CTD chain configuration involved 40 packages for the measurement of conductivity, temperature and depth (CTD), distributed over the 70-m aperture between surface float and depressor (Fig. 2). With standard scanning cycles of 2 seconds, the spatial sampling mesh width was 3.3 m horizontally and 1.8 m vertically. CLANT SA ER EN TR E UND SE A C RESEARCH ELECTRODE CONDUCTIVITY TEMPERATURE DEPTH INSULATED WIRE ELECTRODE DEPRESSOR Figure 2. Sketch of the towed CTD chain. Sensors are inductively coupled to the towing cable. 3 Environmental variability and its impact on acoustics In the course of a tidal cycle, the surface level changes by 2 m. The bottom moored thermistor strings feel the rise and fall of the thermocline with the tide. The temperature contours from low-pass filtered time series of the thermistor strings are highly correlated with, but not strictly parallel to surface elevations (Fig. 3). Acoustic energy propagating with low grazing angles would feel the thermocline position rather than the ocean surface. At 10 km range, the first four or five arrivals (see Fig. 4 of Nielsen et al. [4]) are from eigenrays, which do not touch the surface, but are influenced by sub-surface tidal signatures. Tidal currents are approximately aligned with the acoustic track. They have maximum strength 25 cm/s and are nearly independent of depth. The long axis of the tidal ellipse is 2.5 km. Integrated over the 5 days of acoustic experiments, there was a dislocation of water masses to the SE by 10 km. Water mass properties were however sufficiently independent from horizontal coordinates, with the consequence, that flushing did not play a role. Tidal waves alter the sound channel as a whole, but do not change its character in general. Higher components of the wave spectrum introduce range dependence to the sound propagation problem. Original (unfiltered) records from the moored thermistor ACOUSTICS AND HIGH RESOLUTION OCEANOGRAPHY 23 strings contain signatures of internal waves with periods down to a few minutes, which would not be correctly sampled using standard instruments having response times of 2 minutes (Fig. 4). Records at different depths are well correlated. By inspection we find no indication of internal wave modes other than the lowest. From the density profile of the water column, the dispersion relation of internal waves was solved numerically [5]. A first mode oscillation of 1 (10) cycle(s) per hour is related to 0.6 (8.5) wave lengths per km. Phase velocities of these short internal waves are only slightly higher than the background current. The frequency spectrum of internal waves is therefore heavily affected by frequencies of encounter. Figure 3. Tidal surface elevation (upper) and filtered temperature distribution (4°C – 10°C) from the eastern thermistor string. White bars indicate the duration of the acoustic experiments at 5, 2, 10 and 0.78 km range. Figure 4. Temperature contours from original records of the western (left), southern (mid) and eastern (right) thermistor strings. While with 10 s sampling rate, two strings resolve high frequency internal waves, the third with 2 minutes under-samples. For deterministic modeling of acoustic transmissions through a fluctuating ocean, the time and space dependent sound velocity field of the ensonified area must be known. Since it is impossible to obtain three or four dimensional measurements of ocean parameters with sufficient resolution, artificial representations of the sound velocity field are used instead with the correct fluctuation statistics in anticipation. Under favorable conditions and reasonable assumptions such as independence from the azimuth angle, a local frequency spectrum of internal waves may be converted into a horizontal wave number spectrum, which in turn is used for the realization of a spatially 24 J. SELLSCHOPP ET AL. fluctuating sound velocity field. Because of the tidal currents, the moored measurements during ASCOT01 cannot be transformed into wave number spectra. Spatial variability, directly measured by means of the towed CTD chain was analyzed for the wave number spectrum of contour displacements (Fig. 5). Low wave numbers, which would reflect repetitions of the 10 km track were excluded by high pass filtering. This also removes potential contamination by temporal effects such as tides. Figure 5. Wave number spectra of the vertical displacement of high pass filtered temperature contour lines from 6 to 12 °C. Note that the spectral energy increases with depth. The blue curve is for 6°C. Figure 6. Distribution of vertical displacements of the high pass filtered 10°C contour. 6 to 12°C distributions look qualitatively the same. The standard deviation increases with depth (decreasing temperature). A Gaussian standard distribution is shown for comparison. The spatial temperature records of the towed sections show pronounced downward excursions of the contour lines similar to those in Fig. 4. The asymmetry of internal fluctuations is reflected by the distributions of contour line depth. Throughout the thermocline, they deviate significantly from a Gaussian normal distribution (Fig. 6). A primitive ray tracing routine was used for the simulation of acoustic pulses propagating through a real fluctuating sound velocity field. Measured sound velocity (by CTD chain temperature and salinity) was interpolated to a 1 x 1 m grid in the vertical plane spanning between source and receiver. The routine calculates sound velocity gradients, grazing angle, vertical position and time increments at each horizontal grid ACOUSTICS AND HIGH RESOLUTION OCEANOGRAPHY 25 point. Two realizations are displayed in Fig. 7. Bottom topography was taken from echo sounder recordings and the sound velocity field from a parallel tow track. Only the rays of a narrow beam are displayed, which arrive at a receiver 40 m in depth. Figure 7. Multipath propagation through a realistic ocean. A hardly noticed horizontal shift of the sound velocity field causes a significant change of the pattern of eigenrays. The simple relation between the initial ray elevation angle and the number of turning points, which exists in range independent environments with smooth boundaries, gets lost by range dependence. Different rays of the same class can coexist, appear and vanish by small changes in the sound velocity field. Small differences in travel time cause frequency dependent fading. Figure 8 shows pulse arrivals in the water column at the position of the receiving array for a range independent and for a sample range dependent situation. Figure 8. First few arrivals from multipath propagation over the 2 km test range. Left: Range independent. Right: Realistic range dependent example. Bottom roughness amplifies the effect of fine structure ocean variability. Slightly modified refraction in the water column may shift the point of contact with the bottom to a facet with different slope, so that differential adjustment of the eigenray path will not bring the reflected ray back to the desired direction. With bottom roughness single 26 J. SELLSCHOPP ET AL. eigenrays die and come to life much easier than with a smooth bottom (see Fig. 7). Because multipath arrival patterns become less stable, matched field processing is a challenging task in this environment. 4 Conclusion Nielsen et al. [4] have shown that irrespective of a similar experimental setup the results of the ASCOT01 acoustic experiments differ largely from the ADVENT99 results. The matched field correlation of an early received signal with subsequent signals drops off very fast, even for low frequencies. While source localization by common matched field processing failed in the ASCOT01 environment, Siderius et al. [6] successfully used a more robust processor, which correlates signal envelopes separately for each hydrophone. The unexpected resistance of the ASCOT01 environment against standard data processing, as far as can be judged at present, is due to high bottom reflectivity, bottom roughness and higher sound speed variability than during the ADVENT99 experiments [1]. The large amount and quality of environmental data, which were collected together with acoustics offers the opportunity to study the processes in detail, which are responsible for coherence loss and signal degradation in a relatively benign littoral ocean area. Acknowledgements The experiment was carried out as part of the SACLANTCEN Programme Of Work. We thank crews and cruise participants for their outstanding dedication to the task. FWG sponsored a considerable part of data processing including the development of processing tools. References 1. M. Siderius, P.L. Nielsen, J. Sellschopp, M. Snellen and D. Simons, Experimental study of geo-acoustic inversion uncertainty due to ocean sound-speed fluctuations, J. Acoust. Soc. Am. 110(2), 769–781 (2001). 2. A.R. Robinson et al., http://www.deas.harvard.edu/~leslie/ASCOT01/ (2001). 3. J. Sellschopp, A towed CTD chain for two dimensional high resolution hydrography, Deep Sea Research 44(1), 145–163 (1997). 4. P.L. Nielsen, M. Siderius and J. Sellschopp, Broadband acoustic signal variability in two “typical” shallow-water regions. In Impact of Littoral Environmental Variability on Acoustic Predictions and Sonar Performance, edited by N.G. Pace and F.B. Jensen (Kluwer, The Netherlands, 2002) pp. 237–244. 5. R. Evans, Program WAVE, http://oalib.saic.com/Other/wave/wave.zip (2001). 6. M. Siderius, P. Nielsen and J. Sellschopp, Source localization in a highly variable shallow water environment: Results from ASCOT-01. In Impact of Littoral Environmental Variability on Acoustic Predictions and Sonar Performance, edited by N.G. Pace and F.B. Jensen (Kluwer, The Netherlands, 2002) pp. 425–432.
© Copyright 2025 Paperzz