Summary of meeting on proposed ANSI standard on passive acoustic monitoring (PAM) using towed array systems: DCLDE workshop, July 17, 2015

Transcription

1 Summary of meeting on proposed ANSI standard on passive acoustic monitoring (PAM) using towed array systems: DCLDE workshop, July 17, 2015 Here is a brief summary of the topics discussed during the first group meeting on the proposed ANSI standard for PAM monitoring, held from 9-10:30 on July 17 at the Scripps Institution of Oceanography. Although two people took notes, the meeting was not recorded in its entirety, so what is listed below is a paraphrased and condensed version of the meeting. Although I do have notes identifying particular speakers, I ve generally chosen to list particular comments without attribution, so people will not hesitate to speak freely in future meetings. The meeting began with the overview presentation (also posted on this website). The key points were that a standard can have required, required to report, and recommended sections. A guiding principle behind the effort is to emphasize the required to report section, to prevent suppression of innovation, but also to permit more transparency to regulators and external reviewers on how a particular PAM operation is conducted. Motivation: Deborah Epperson then provided some background as to why BSSE (Bureau of Safety and Environmental Enforcement) and BOEM (Bureau of Ocean Energy Management) are interested in the effort. As a particular example, in the Gulf of Mexico BOEM now requires a PAM system for seismic operations working in over 100 m of water. On any given day there are vessels working in the region, using PAM systems during times of low visibility. However, standardization of these PAM systems and the operators are currently not laid out. So any so-called PAM system meets current requirements. Hiring a third party and PAM system to meet these requirements is a confusing process to the companies and contractors. Furthermore due to logistical constraints most towed systems will be based off a seismic source vessel. So how do we balance practical ability of towed array work on a source vessel while also creating standards for the regulatory community to feel like they are actually mitigating accurately, and meeting conservation requirements? The National Marine Fisheries Service (NMFS) has some dated guidelines, but they are very nebulous. So really hard for a regulator to look at a proposed system and decide if it s good or useful. Avoiding a worst-case scenario: The discussion then turned to the suggested agenda of the presentation: how to avoid a very common scenario where a PAM system is deployed off a noisy vessel and no marine mammal detections are reported. Were no animals actually present, or did the noisy environment wash out any possibility of detection? The presentation suggested several steps to prevent this situation: (A) The performance of the system needed to be predicted and simulated by the PAM operator before deployment, generally in terms of expected detection range of various species clusters under expected operational conditions. This information would

2 be contained in a planning document. (B) At least one hydrophone in the system needs to be calibrated (i.e. be able to allow power spectral density estimates to be measured in upa 2 /Hz), in order to allow accurate assessment of background noise levels. While uncalibrated hydrophones are fine for detection and localization of sounds, a calibrated hydrophone is needed to determine whether platform noise is exceeding expected levels. (C) Noise statistics need to be provided by the operator after a cruise using a validation document. Examples of what biological signals would look like in this noise environment at required detection range would be either be displayed or simulated (if no marine mammal calls detected). The validation document would also generate updated performance curves for automated detection processes using actual background noise data. Calibration: Discussion began with the calibrated hydrophone recommendation, and what that meant specifically. At least one PAM operator pointed out that a full calibration on a towed system could be very expensive (such as using the TRANSDEC facility in San Diego). While pistonphones can be used to calibrate individual hydrophones in autonomous recorders, they are not designed to be used on hydrophones embedded in a tow cable. It was also pointed out that the entire signal route from hydrophone to recording device needed to be calibrated. I commented that I did not feel that a full independent calibration of the hydrophone system in a specialized facility was necessary. I suggested that it would be sufficient to use the frequency-dependent sensitivity values provided by the manufacture (in terms of db re 1 upa/v) to relate input levels to hydrophone output voltage. The filtering properties of the tow cable can be measured or modeled, and a test calibration signal of known properties (frequency comb, FM sweep, or white noise) can be fed into the point where the array cable connects into the rest of the preprocessing/recording system before the operation starts. Combining all this information should permit a signal amplitude on a digital recording to be converted into a pressure amplitude (or more specifically, to permit a power spectral density generated from such a recording to be converted into units of db re 1uPa 2 /Hz). Some paraphrased comments: I hear about calibrating hydrophones, but that doesn t matter if it s being run through pre-amps, a long cable, etc., and the ENTIRE system, then needs to be calibrated, and I don t know anyone who has done that. And it s not a trivial matter. Every time you change out a part, the system would need to be recalibrated and that gets expensive. But to get actual noise levels, we needs to figure out how to calibrate the whole system in order to interpret the data. Agreed, must at least know the ADC, and many times don t know that. Predicting system performance in a planning document: Is it practical to require PAM performance to be simulated ahead of a cruise? At least at the beginning, many PAM operators would have little information on the expected noise

3 environment, and this lack of information risks turning the simulated performance into an empty paperwork exercise. Furthermore, who would be expected to generate these simulations? The actual PAM field operator has limited time and likely experience to generate these predictions before a cruise. I suggested that while lack of prior information would yield poor predictions during the early stages of reporting under the standard, the use of a validation document might allow a gradual positive feedback cycle of improvement in the quality of the planning document predictions. For example, the standard might encourage a vessel operator to keep copies of a noise fingerprint provided in a validation document after a cruise, to be then used by future PAM operators in future planning documents that would use the same vessel. It was also suggested that PAM software could be adapted in the future to allow performance simulations to be generated quickly without substantial expertise. For example, software could generate synthesized calls at various ranges that could be embedded in either archival or simulated noise signals already displayed on a spectrogram. Some specific comments, paraphrased: Noble goal [predicting performance], and to predict how well these systems work before they are put in the water is a good idea, but I ve never seen a post-hoc analysis of how well the range of PAM services actually worked. We need to come up with good statistics to use for monitoring, and can t just plug into the SONAR equation our first step to look at ad-hoc may not be useful, but the second bit of having a well-calibrated system might be more hopeful. Until we can evolve and actually consider source levels of many species, this is too big of a goal, especially when the operators are not likely specialized acousticians. So our goals need to consider how a relatively untrained person can do this in a couple of days. But if we had a toolbox, with a button to click for each species, that could run through this noise / mitigation goals for a given species one at a time, then an operator could click that species-specific toolbox and not need to be a specialized acoustician [to generate predicted performance]. Recording and reporting background noise levels in validation document There seemed to be general consensus that background noise levels should be reported in a standardized way in the post-operations validation document, but how? One big issue is the time-dependence of the noise background due to shifting environmental and operating conditions. How many simulations are needed to characterize the system performance in a post-cruise validation? The beginning of an answer to this question lies in how often acoustic data need to be sampled or stored. Everyone agreed that each marine mammal detection should be recorded (across all hydrophones), with a large buffer time surrounding each detection. Everyone also agreed that raw acoustic data should be required to be sampled on a regular basis (e.g. every 30 minutes) for a minute or so, so that percentiles of the power spectral density could be generated and reported, thus addressing some of the concerns

4 about how to characterize the time-dependence of the noise background. It was cautioned, however, that there should be a small random component to the sampling times (like a jitter) in order to avoid potentially missing cyclic noise occurrences. There was also vague agreement that noise samples should be collected whenever a major change occurs in operating or environmental conditions. Ideally, a PAM operator should be able to press a button to collect a minute noise sample, in addition to the automated sampling schedule. Should baleen whales be excluded? A question arose as to what stance the standard should take on using towed PAM to monitor low-frequency baleen whales. At present most existing systems seem to have too much flow noise to reliably guarantee detection, although some PAM providers claim the ability. There seemed to be some consensus that a standard should not explicitly prohibit towed PAM used for baleen whales, but should do the following: (1) In the planning document, the PAM operator must list species clusters that are anticipated to be in the area, and must list species clusters that the PAM will monitor for. Thus it would be made explicitly clear to a contractor and regulator whether PAM would attempt to address all anticipated species, or just a subset. (2) If a PAM organization claims that baleen whales could be detected within a certain detection range, then the validation document should contain simulations of baleen whale signals in actual measured noise backgrounds to validate the initial claims, and to show that flow noise levels would not obscure signals (when they are scaled for desired detection range). Can single hydrophones be used to estimate range? There was a bit of discussion about whether detection of a call on a single hydrophone could be used to determine whether a group of animals is within a certain detection range. Apparently a lot of PAM operators use gut reactions to estimate the range of a call. The suggested strawman stance was that such a technique would not be restricted in a standard; however, post-cruise simulations in the validation document would be needed to buttress the claim that a given species cluster call could be detected out to a minimum detection range when embedded in actual noise conditions. That is, this situation would be handled as a required to report item, although a recommendation might be added to discourage the practice for certain species clusters. Preventing misuse of time-delayed synthetic aperture range estimation. A time-honored technique for getting slant range to sperm whales has been to collect a series of bearing estimates from a short-aperture array towed behind a fast-moving ship, and then triangulate the bearings. Some concern was raised that the technique would be inappropriate for arrays towed behind slow-moving (5 kt) seismic vessels, as this situation violates the implicit assumption that the animal s position is fixed during the bearing acquisition time. Rather than restricting the method to certain species or platforms, the current spreadsheet suggests that a PAM operator would need to provide detailed information about the circumstances under which this technique was used, and that every localization

5 obtained by this method also log the ship speed, time window used to acquire the estimate, and left/right ambiguity. These information would be statistically summarized in the planning document, thus rendering transparent whether the assumptions behind time-motion analysis are being violated. Hydrophone sensitivity The draft standard spreadsheet initially suggested that there be a minimum sensitivity required for hydrophones; instead, it was suggested that no particular sensitivity be required, but that the sensitivity would have to be reported.