AN AUTONOMOUS AND ADAPTIVE SAMPLING LOW-DUTY CYCLE AMBIENT SOUND SENSOR FOR MARINE OBSERVATIONS: DESIGN AND APPLICATIONS Marios N. Anagnostou, Roberto Bonzano, Sara Pienseri, Emmanouil N. Anagnostou, and Jeffrey A. Nystuen Acknowledgments: Dimitris Kassis, Paris Pagonis from HCMR and the PERSEUS project for the contribution to the new PAL data collection
Motivation Sound budgets Policy: Monitoring of the marine environment is driven by data needs with the scope of an efficient protection from anthropogenic impacts, coupled with the sustainable and effective exploitation of marine resources. Ambient noise: Ambient noise pollution is one of the essential physical variables affecting the marine environment and ecosystem mainly caused from excess human activities. Climate data – Rainfall and wind speed are two critical climatic variables that define atmosphere-ocean interactions. Marine mammals–Monitoring the marine mammal movements, communications and feeding process. Monitoring oceanic conditions (winds, precipitation, sea roughness) for real-time operations
Purpose and Goal What we need: Develop of an in-situ autonomous long-term underwater ambient noise sensor To do: Monitor the marine mammal movements, communications and feeding process Monitor precipitation rate and type and wind speed in oceans Monitor geophysical deformations (underwater landslides, ice cracking, etc.) Monitor anthropogenic activities (ships, sonars, etc.) Monitor noise pollution and create baseline sound source budgets Process and transfer data in real-time all from ONE sensor … not just recorded sound levels
How do we interpret the ambient sound? Assume that different sound sources have unique features, spectral or temporal, that allow the source to be identified Classification is CRITICAL! Quantify (wind speed, rainfall rate, bubbles, whale identification, … Validate (anemometer, weather radar, sound clips, …)
Brief History 1940s – Knudsen – underwater noise 1962 – Wenz curves – rain noise identified 1984 – acoustic wind measurements (Evans et al.) 1990 – acoustic wind speed algorithm (Vagle et al.) 1992 – lab studies of sound from raindrops (Medwin et al.) (acoustic physics of drop splashes) 1999 – field measurements on TOGA-TAO array (Nystuen et al.) (1st generation development: Acoustic Rain Gauge - ARG) 2004 – Ionian Sea Rainfall Experiment (Anagnostou et al. 2008) (2nd generation development: Passive Aquatic Listener – PAL for deep ocean deployment) 2008 – POSEIDON weather observation system (Anagnostou) (First operational use; data transfer through buoy) – 2nd generation PAL deployed at Liguria Sea (Anagnostou) – New generation PAL deployed at EU PERSEAS project (Anagnostou)
Technological Overview of PAL Improved power management -> longer deployment FPGA technology -> increased real-time processing of acoustic data. New smart and more advanced algorithms for wind speed and rainfall detection and quantification. Housing minimization. Full Scale input range (to Analog stage) = 1Vpp Analog output dynamic range = 10Vpp System resolution = 15.2uV Max measurable sound level = 19,8 Pa Min measurable sound level= 0,00059 Pa Ein,ADC = 171.5uVrms (total noise input to ADC) SNRin,ADC = 86,32dB (input signal to ADC) Total continuous time acquisition (one time series) = about 10 minutes (with 256MB DDR memory, increased linearly when DDR is increased) Total spectra PSD files = about 2 million (with 4GB card, increased linearly when flash card is increased)
Technological Overview of PAL Deployment Modes Surface moorings Autonomous (sub-merged moorings) Real time data transfer for operational control
Two data collection modes 1) high sample rate (>120 kHz) 2) low duty cycle (1%) 3) long deployment (low power) 4) adaptive sampling strategy PALIII is far more than a simple acoustic recorder with the following characteristics: Phase 1: it is collecting data for a time interval (Window) specified by the user. The role of the operating software is to control the sampling strategy of the UPAL. - adaptive sampling strategy - variable time step depending on sound source Phase 2: The automated sound analysis software, which is the processing of the data into the “flag” file (where in turn can also optional create and transmit a message). - used to identify sound source: - classify/quantify, send the summary message
Overview
Classification depends on 1) spectral characteristic 2) temporal patterns
Wind Sound generated by resonating bubbles at leading edge of breaking waves No signal for U < 3 m/s Distribution of bubble sizes controls the shape of the spectrum At very high wind speed small bubbles are stirred down and absorb new surface generated sound Uniform spectral slope from 2-10 kHz at moderate wind speeds
Rain Sound generated by resonant bubbles created during the splashes of individual raindrops Different distribution of bubbles sizes than for breaking waves 1 mm raindrops generate 15-22 kHz bubbles (drizzle) 2-3 mm raindrops generate larger bubbles (2-10 kHz) (heavier rainfall) 4-5 mm generate bubbles sound to 1-2 kHz (convective rain) Spectral slope flatter than wind from 2-10 kHz
Shipping Sound from engines, propellers, hull, sonars, etc. Generally more low frequency sound than wind or rain Close ships loud at all frequencies Distant ships (shipping) relative loud at lower frequencies Spectral slope from 2-8 is steeper than wind or rain
Compare spectra
R/V Dallaporta maintaning the W1M3A observatory Underwater ambient sound and meteo obs. Convective rain (30-Aug-2016 20-23 UTC) Rainfall starts at 19:40 UTC, lasts for 32 minutes and the cumulative is11.31 mm (10.59 mm in only 14 minutes). R/V Dallaporta maintaning the W1M3A observatory
Underwater ambient sound W1M3A MSC Fantasia 2 Km Striped dolphin
Marine mammal verification Aegean sea dolphin
Aegean Sea Dolphin Detection at the Athos POSEIDON buoy Nystuen & Anagnostou et al. 2013, JTECH
Validation of Shipping Temporal record of ship passage
Wind speed verification Nystuen & Anagnostou et al. 2013, JTECH Sound levels highly correlated with wind speed
Pensieri et al. 2013 OCEANS2013
Definitions for ambient noise Trends in the annual average of the squared sound pressure associated with ambient noise in each of two third octave bands, one centred at 63 Hz and the other at 125 Hz, expressed as a level in decibels, in units of dB re 1 μPa, either measured directly at observation stations, or inferred from a model used to interpolate between or extrapolate from measurements at observation stations [Van der Graaf et al., 2012].
Sound Budgets Percentage of time present 20 S 85 W 10 N 95 W Bering Sea Ionian Sea Carr Inlet Haro Strait Wind 93% 86% 90% 74% 80 % 21 % Rain - 8 % 3 % 5 % Ships 0.5 % 1.5 % 1 % 20 % 2 % 59 % Whale* 1.8 % 0.6 % Other 4 % 6 % 10 % 15 % Haro Strait is easily the noisiest environment studied. Shipping dominates most of time. There is also an "other" noise associated with the mooring, possibly flow noise. The "whale" signal is a 30-kHz click. This is consistent with a beaked whale echo location click. It is present at the deep water locations, but not the shallow water locations. Other types of biological sounds are present at the shallow water locations, but are not identified. *30 kHz click detected – no visual confirmation
Sound Summary at 2 kHz Sound Summary at 20 kHz Wind is the dominant sound source by frequency, but 88% of the loudest 1% of the spectra are from boating. Wind Rain Boating Other All 76% 10% 3% 11% Loudest 5% 39% 8% 47% 6% Loudest 1% 9% 87% 1% Sound Summary at 20 kHz Wind Rain Boating Other All 76% 10% 3% 11% Loudest 5% 0% 96% 1% Loudest 1% 94% Nystuen & Anagnostou et al. 2014, JTECH
PAL is a unique sensor that uses underwater sound to monitor the marine environment Increase the availability of long-term climatic data in oceanic locations. Support implementation of Maritime Policies particularly those relating to underwater sound pollution. Promote new discoveries leading to better understanding of the sea –in situ time-series of data to support a more detailed understanding of several key marine aspects: 1. impact of human-induced noise on the behavior of marine mammals 2. the communication and feeding habits of mammals 3. the interaction mechanism of the atmosphere with the sea-surface (high wind effects) Real-time monitoring for marine infrastructures (off-shore wind farms, oil platforms, etc.) Cost efficiency and security relative to surface buoys