Computers & Electronics

Erratum to: Characterizing the Relative Contributions of Large Vessels to Total Ocean Noise Fields: A Case Study Using the Gerry E. Studds Stellwagen Bank National Marine Sanctuary

Description
Erratum to: Characterizing the Relative Contributions of Large Vessels to Total Ocean Noise Fields: A Case Study Using the Gerry E. Studds Stellwagen Bank National Marine Sanctuary
Published
of 18
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
Share
Transcript
  Characterizing the Relative Contributions of Large Vesselsto Total Ocean Noise Fields: A Case Study Using theGerry E. Studds Stellwagen Bank National Marine Sanctuary Leila Hatch   Christopher Clark   Richard Merrick   Sofie Van Parijs   Dimitri Ponirakis   Kurt Schwehr   Michael Thompson   David Wiley Received: 20 December 2007/Accepted: 24 May 2008   Springer Science+Business Media, LLC 2008 Abstract  In 2006, we used the U.S. Coast Guard’sAutomatic Identification System (AIS) to describe patternsof large commercial ship traffic within a U.S. NationalMarine Sanctuary located off the coast of Massachusetts.We found that 541 large commercial vessels transited thegreater sanctuary 3413 times during the year. Cargo ships,tankers, and tug/tows constituted 78% of the vessels and82% of the total transits. Cargo ships, tankers, and cruiseships predominantly used the designated Boston TrafficSeparation Scheme, while tug/tow traffic was concentratedin the western and northern portions of the sanctuary. Wecombined AIS data with low-frequency acoustic data froman array of nine autonomous recording units analyzed for2 months in 2006. Analysis of received sound levels(10–1000 Hz, root-mean-square pressure re 1  l Pa  ±  SE)averaged 119.5  ±  0.3 dB at high-traffic locations. High-traffic locations experienced double the acoustic power of less trafficked locations for the majority of the time periodanalyzed. Average source level estimates (71–141 Hz, root-mean-square pressure re 1  l Pa  ±  SE) for individual vesselsranged from 158  ±  2 dB (research vessel) to 186  ±  2 dB(oil tanker). Tankers were estimated to contribute 2 timesmore acoustic power to the region than cargo ships, andmore than 100 times more than research vessels. Our resultsindicate that noise produced by large commercial vesselswas at levels and within frequencies that warrant concernamong managers regarding the ability of endangered whalestomaintainacousticcontactwithingreatersanctuarywaters. Keywords  Shipping    Underwater noise    Whales   Automatic Identification System    U.S. NationalMarine Sanctuary    Marine protected area Introduction The anthropogenic components of underwater noise andtheir potential impacts on marine resources are topics of substantial interest and concern among scientists and thepublic (NRC 1994, 2000, 2003, 2005). In the past 5 years, many national and international policy forums haveaddressed ocean noise concerns (ACCOBAMS 2003;ASCOBANS 2003; World Conservation Union [IUCN]2004; IWC 2005; U.K. IACMST 2006; U.S. MMC 2007). These concerns have mainly focused on injury and/orbehavioral disturbance of whales exposed to impulsive(e.g., short-duration) sounds, such as sonars utilized fornaval defense and seismic sources used for oil and gasexploration (NRC 2000, 2003). However, possible impacts Electronic supplementary material  The online version of thisarticle (doi:10.1007/s00267-008-9169-4) contains supplementarymaterial, which is available to authorized users.L. Hatch ( & )    M. Thompson    D. WileyGerry E. Studds Stellwagen Bank National Marine Sanctuary,U.S. National Oceanic and Atmospheric Administration, 175Edward Foster Road, Scituate, MA 02066, USAe-mail: leila.hatch@noaa.govC. Clark     D. PonirakisBioacoustics Research Program, Cornell University Laboratoryof Ornithology, 159 Sapsucker Woods Road, Ithaca, NY 14850,USAR. Merrick     S. Van ParijsNortheast Fisheries Science Center, U.S. National Oceanic andAtmospheric Administration, 166 Water Street, Woods Hole,MA 02543, USAK. SchwehrCenter for Coastal and Ocean Mapping Joint HydrographicCenter, University of New Hampshire, 24 Colovos Road,Durham, NH 03824, USA  1 3 Environmental ManagementDOI 10.1007/s00267-008-9169-4  to marine animals exposed to continuous sources, such ascommercial shipping, have recently begun to garner moreattention (Southall 2005, 2007). Evidence of increasing levels of underwater noise asso-ciated with shipping has heightened concerns regarding the‘‘masking’’ of marine animal signals, particularly low-fre-quency vocalizations, with possible negative effectsincluding diminished abilities to find mates, maintain socialstructure, forage, navigate, and/or evade predation (Payneand Webb 1971; Erbe and Farmer 1998, 2000; Southall and others2000;Erbe2002;Morisakaandothers2005;Nowacek  andothers2007).Duetotheconcentrationofacousticenergyfrom large commercial vessels within low-frequency band-widths, and the efficiency of the propagation of lowfrequencies underwater, distant commercial shipping domi-nates low-frequency, omnipresent background or ‘‘ambient’’noiseinmanypartsoftheworld’soceans(Wenz1962,1969; Gray and Greeley 1980; Ross 1993; Greene and others 1995). The relative contribution of vessel noise to ambientocean noise varies with the distribution of vessel traffic, withareas such as shipping lanes (Andrew and others 2002;McDonald and others 2006) and the northern hemisphere ingeneral (Cato 1976) showing higher noise levels. In suchareas, increasing commercial maritime transport over thepast 30 years is correlated with 10 decibel (dB) re 1 micro-pascal ( l Pa) increases in low-frequency noise levels(Andrew and others 2002; Cato and McCauley 2002; McDonald and others 2006).A key recommendation of reports dealing with anthro-pogenic noise is the need to establish ‘‘noise budgets,’’defined as the sum of the relative contributions made byidentified sound sources to total noise fields, for areas of the ocean (NRC 2003). The NRC specifically identified theneed to define the sound contributions of different vesseltypes within the major category of shipping and to char-acterize the temporal (e.g., annual, seasonal, monthly, anddaily) and spatial variation of noise production and soundfields (NRC 2003). This information can then be used tounderstand the potential impact of anthropogenic noise onlocal marine animals.Our study site was the Stellwagen Bank National MarineSanctuary (SBNMS or sanctuary). The sanctuary is a fed-erally designated marineprotectedarea locatedtothe eastof Boston, Massachusetts, USA, and in close proximity to adensely populated coastal zone (Fig. 1). Because of this,substantial commercial shipping transits the sanctuary toand from the port of Boston, and the sanctuary hosts aUnitedNations’InternationalMaritimeOrganization(IMO)recommended route for commercial vessels (the BostonTraffic Separation Scheme; BTSS). The sanctuary is also animportant feeding ground for endangered marine mammalssuch as the North Atlantic right whale ( Eubalaena glacial-is ), humpback whale (  Megaptera novaeangliae ), and finwhale (  Balaenoptera physalus ), which are vulnerable tocollisions with vessels and persistent exposure to shipping-generated noise. As a result, the sanctuary makes an ideal Fig. 1  The upper-left inset shows the location of the Gerry E. StuddsStellwagen Bank National Marine Sanctuary (SBNMS) in Massa-chusetts Bay, off the northeast coast of the United States. Bothregional maps show the boundaries of the SBNMS and the locationsof the nine Autonomous Recording Units (ARUs) deployed betweenApril 7 and May 24, 2006. All vessel traffic tracked using the U.S.Coast Guard’s Automatic Identification System during the months of April and May 2006 using four receivers located on Fisher Island nearGroton, CT, on Cape Cod near Provincetown, MA, in downtownBoston, MA, and at SBNMS headquarters in Scituate, Massachusettshas been plotted on the left. April’s traffic is plotted in black andMay’s traffic is in gray. The map at the right allows the reader toreference the ARU numbers referred to in the text, the location of the2006 Boston Traffic Separation Scheme (dashed line), and thebathymetry of the study areaEnvironmental Management  1 3  studysitefortheinvestigationofanthropogenicnoiseanditspotential impact on endangered large whales.For this investigation, we used the U.S. Coast Guard’sAutomatic Identification System (AIS) to track individualvessels transiting the SBNMS throughout 2006. For Apriland May 2006, we combined ship transit data with acousticdata collected from autonomous recording units (ARUs)placed throughout the sanctuary. Our goal was to establish anoise budget for the sanctuary by (1) quantifying the tem-poral and spatial vessel traffic patterns in the sanctuary,including variation in vessel size and speed, (2) quantifyingthe noise produced by specific vessel classes, and (3)extrapolating those data to the entire sanctuary area for a 1-year period. We also used data on historic distribution of marine mammals in the sanctuary to quantify the acousticconsequences of the sanctuary’s recent shifting of the BTTSfrom areas of high whale density to low-density areas. Methods Acquisition of Automatic Identification System DataUnder the IMO’s current mandates, all ocean-going com-mercial traffic [ 300 gross tons or carrying more than 165passengers, as well as all tug/tows, are required to carryAIS transmitters (Federal Register 2003; IALA 2004). The AIS is a VHF ‘‘line-of-sight’’ transmitter that broadcasts avessel’s position, identity, and various characteristics(including but not limited to length, beam, draught, cargotype, destination, and speed) as often as every 2 s. FourAIS receivers, located near Provincetown, Boston, andScituate, Massachusetts, USA, as well as on Fishers Island,New York, USA, allowed for the tracking of all vesselscarrying AIS transmitters as they transited the study areaand beyond (Fig. 1). Through collaboration with the U.S.Coast Guard, vessel tracking data from these receiverswere continuously archived for the entire year of 2006 on aserver at the SBNMS and available for real-time viewing aswell as post hoc analysis.Analysis of AIS DataAn area defined by the boundaries of the SBNMS extendedby 5 nautical miles (nmi) was chosen for analysis. Thisspatial extent was chosen as a course estimate of the areawithin which ships with source levels (SLs)  C  180 dB re 1 l Pa would ensonify the sanctuary at levels [ 120 dB re 1 l Pa. Source levels for the majority of large commercialvessels range from 170 to 190 dB (Wenz 1962; Gray andGreeley 1980; Greene and others 1995). For a SL of  180 dB, a course estimate of transmission loss (TL) of 60 dB (bringing received levels [RL] to 120 dB), usingTL  =  15 log10 (distance), would occur approximately 5.4nmi (10 km) from the source. As discussed further below,this simplified TL equation was chosen to represent theaverage sound field during our study, while retainingdependence on the principle features of the environment.Archived AIS data collected in 2006 were extracted,reformatted, filtered, and quality controlled. Reformattingwas necessary to provide a continuous data stream ratherthan daily log files collected for the U.S. Coast Guard’spurposes, and filtering was necessary to extract only thevessel information that was necessary for this study.Extraction and filtering were completed using the U.S.Coast Guard’s newly updated software (U.S. Coast GuardResearch and Development 2007) and custom softwarewritten in Python v2.5.1 (Python Software Foundation2007) added to the noaadata package (Schwehr 2007).Ship positions were loaded in PostgreSQL/PostGIS(Santilli and Leslie 2007) for transit analysis. Using thedatabase, each vessel’s position reports were grouped intotransits. A vessel’s transit was defined as a set of chrono-logically ordered position reports preceded and followedby not receiving a position report from the vessel within thestudy area for 1 h. Noaadata then generated a report of alltransits. Following extraction from AIS Miner (U.S. CoastGuard Research and Development 2007) and ArcGIS(ESRI 2006), AIS data were manipulated and/or graphed inMicrosoft Excel (Microsoft Office 2003) and/or MicrosoftAccess (Microsoft Office 2003), depending on the file size.Finally, because some AIS data fields (e.g., ship iden-tification numbers, ship name, ship type, cargo type, anddimensions) rely on manual input from shipboard users,they are more likely to contain errors (Harati-Mokhtariaand others 2007). Thus, information on each of the 541vessels was reviewed by hand, with additional information(including gross tonnage, flag of ship, and year built) andany errors identified by cross-checking information pro-vided with that available from Web-based ship registries(e.g., Equasis and individual company websites). Ship typecategories were taken directly from the InternationalAssociation of Marine Aids to Navigation and LighthouseAuthorities’ Guidelines for AIS (IALA 2004), with furtherspecificity provided through Web-based research. Onceships were classified, the separate groups of ships weregridded using noaadata. The study area was divided into agrid with 1  9  1 nmi (1.85  9  1.85 km) cells. Each transitwas linearly interpolated across the gridded region basedon all of the received ship position reports. A cell wasincremented by one for each time the ship entered the cell.If a ship left and re-entered a cell, the cell was againincremented. The amount of time a ship resided within acell was not considered for the cell counts. The total countswere then written to an ArcASCII grid and imported intoArcGIS for display. Environmental Management  1 3  AIS data were visualized and spatially analyzed eitherusing custom queries added to the noaadata software orArcGIS. Minimum great circle distances (meters) betweenvessel locations and ARU locations were calculated todetermine each vessel’s closest point of approach (CPA) tothe array. Each transit of the sampled area per vessel wasthen further documented by recording the times and dates of entries and exits. Vessels’ speeds over ground were ana-lyzed, and results are presented by Wiley and others (2008).As discussed by Wiley and others (2008), calculations of average speedover groundinaspecifiedareausingAISdatamust account for covariance between the number of datapoints received and the speed of the transmitting vessel. Forthis paper, two summary statistics that are robust to non-uniform sampling rates were calculated for each vesseltransit: minimum speed over ground and maximum speedover ground (as knots or nmi/h and as km/h). Number of hours spent and number of nautical miles/kilometers trans-ited within the sampling region were calculated per transit.The numbers of transits approaching within 5-nmi (9.26-km) radii of each of the ARUs were calculated. Finally,each-nmi 2 (3.42-km 2 ) block of the SBNMS was coded perday according to the presence/absence of vessels falling intofive broad categories (e.g., cargo, tanker, services andresearch, passenger, and tug/tows). These daily presence/ absence grids were then summarized as density plots toshow the distribution of vessel types within the SBNMS.Acquisition of Acoustic DataAcoustic data (10–1000 Hz) were collected using an arrayof 9 or 10 ARUs (ARUs) deployed in the Stellwagensanctuary from January 2006 to January 2007 (Fig. 1). Allunits were synchronized just prior to deployment and justafter recovery. These ARUs, or ‘‘pop-ups,’’ were developedby Cornell University’s Bioacoustics Research Programand are comprised of an external hydrophone attached to aglass sphere containing a battery, computer electronics, andmemory which is temporarily anchored to the ocean floorwith sandbags (Calupca and others 2000). The ARU hy-drophones were calibrated at U.S. Naval facilities in NewLondon, Connecticut, USA, and the operational ARUswere calibrated at a U.S. Naval facility on Seneca Lake,New York, USA. Operational ARUs had flat frequencyresponses ( ± 1-dB variation) in the 55- to 1000-Hz range.To retrieve data from the ARUs, the release of theiranchorage was acoustically triggered using a shipboardtransponder allowing surface retrieval. ARUs wereretrieved and redeployed every 2–3 months in order todownload data and replenish batteries. The units recordedcontinuously at a 2000-Hz sampling rate; raw binaryacoustic data files, as well as finalized multichannel AIFFfiles, were archived onto 160- to 320-GB hard drives.Analysis of Acoustic DataThe months of April and May 2006 were chosen forintegration of acoustic and AIS data because many non-commercial shipping activities that are presumed to impactthe SBNMS’s acoustic environment (e.g., fishing vessels,fishing activities, and small recreational craft) were eitherprohibited (most fishing activity) or reduced (recreationalactivity) during that period. This time period was slightlyrestricted by late deployment and earlier retrieval of a fewARUs in the array. Thus, the analysis period, during whichthefullARUarray wasdeployed,was April 7–May 24,2006(48 days). In addition, one ARU (no. 6) was found to havemalfunctionedonMay4,2006.Thus,analysisatthislocationwas only possible for 27 days.Archived raw acoustic data from all nine ARUs wereprocessed to create synchronized, nine-channel, time-aligned files. Synchronized acoustic data were then visu-alized using an open-source extensible sound analysisapplication for developing sound analysis tools written inMatlab (The Mathworks Inc. 2006) called XBAT (Millsand Figueroa 2005; Figueroa 2007). Daily, weekly, and monthly RLs for 3 broad-frequency bandwidths (10–1000,10–400, and 71–141 Hz) and 17 third-octave bandwidths(center frequencies at 10, 12.5, 16, 20, 25, 31.5, 40, 50, 63,80, 100, 125, 160, 200, 250, 315, and 400 Hz) were cal-culated. Total bandwidth RLs were also broken down bypercentages of the time period analyzed, with RL thresh-olds for 5, 25, 50, 75, and 95% of the sampling periodreported. These calculations were performed using a Mat-lab program called LTspec (Cortopassi 2007). LTspec usesa Fourier transform and time aggregation to generate along-term root-mean-square (RMS) spectrogram whichcontains RMS magnitude values for each frequency bandof width  D  f   over the total spectrogram bin interval  D t  . RMSspectrograms were created using a sampling rate of 2000 Hz, a FFT size of 2048,  D t   =  1.024 s (aggregatedover hours, days, weeks, and months),  D  f   =  0.98 Hz, aHanning window function, and a calibration reference levelof 85.5 dB re 1  l Pa with a reference bit depth of 12.Spectrograms and power curves for each ARU locationwere generated to display variation in received frequenciesand acoustic intensity (dB re 1  l Pa) over days, weeks, andeach of the 2 months sampled.Integration of AIS and Acoustic DataAIS data were integrated with acoustic data in two ways:(1) RLs were examined relative to the number of vesselstransiting each of the ARUs over multiple temporal scales;and (2) the acoustic footprints of 17 vessels representing allvessel types in our 2-month sample were characterized attheir points of closest approach to ARUs in the array. Environmental Management  1 3  Standard errors were calculated to accompany all aver-ages reported in this paper. For the first analysis, low-frequencyacousticeventsidentifiedwithinthedaily,weekly,andmonthlyacousticrecordsforeachARUwerematchedintime with close approaches by AIS-tracked vessels tracked.The total and average numbers of each type of AIS-trackedvesselpassingwithina 5-nmi(9.26-km)radius ofeachoftheARUs were calculated. The relationship between the num-bers ofcloselyapproachingvesselsandthevariationinRMSRLs recorded by each of the ARUs during the months of April and May 2006 was evaluated statistically in a numberof ways. First, the average number of vessels within 5 nmi(dependent variable) was assessed relative to averagereceived levels in the 10- to 400-Hz band (independentvariable) for each ARU in April and May 2006 using linearregression. Analyses were also completed using daily RMSRLs and the daily close approaches by each vessel type, todetermine whether some vessel types were stronger deter-minantsofreceivedlevelsthanothers.Allstatisticalanalyseswere completed in JMP v5.0.1a (SAS Institute Inc. 2002).Seventeen vessels which closely approached the locationofanARUinthe arrayandrepresentedthediversityofvesseltypes in the sample were chosen for the second analysis. Thelocations and times of their CPAs to the ARUs in the arraywere used to match the vessel tracks to the acoustic records.Oncematched,threeestimatesofRLs(1-speakRL,10-speak RL, and 5-min RMS RL) were calculated at the time of theCPA to each ARU inthe array, and in a frequency bandwidth(71–141 Hz) important for vocalizing North Atlantic rightwhales in the SBNMS (Urazghildiiev and Clark  2006). Theaverage of these values was used as the empirical basis forinterpolating the levels of sound inside the boundaries of thearray. Kriging (a group of geostatistical techniques used tointerpolate the value of a random field at an unobservedlocation from observations of its value at nearby observedlocations [see Cressie 1993]) was used to compute the bestlinear unbiased estimator of the RLs based on a stochasticmodel of the spatial dependence quantified by a variogram of the empirically measured levels. Parameter settings used tokrige the acoustic data were as follows: Xmin = –70.4568,Xmax = - 70.1362, Ymin =  42.1204, Ymax =  42.6455,dx  =  0.0168737, and dy =  0.0276368. Variogram settingsincluded the selection of a spherical model type with a rangeof 1.4142, length of 1.4142, resolution of 0.025, power of 3.3973, sill of 2.2495, and nugget of 0. The intensity of thekriged sound field in the 71- to 141-Hz frequency band wasrepresented on a gray scale from black (high) to white (low),withagridsizeof13  9  18points.Thefinalimagewasrefinedusing a smoothing interpolation function in Matlab.Interpolating RLs necessitated identifying a simplifiedestimate of TL (decreasing sound energy over distancefrom source) that could be used to characterize the averagesound field in our study area during the months analyzedhere. To estimate TL, we referred to equations devised byMarsh and Schulkin (1962) which represent average soundfields in areas where conditions gradually transitionbetween spherical spreading in the near field to cylindricalspreading in the far field. Based on the range of bottomdepths and thermocline depths (referenced from Valentineand others 1999) in our study area during April and May,our conditions were characterized as ‘‘intermediate’’and are described by the simplified equation TL  = 15 log(distance)  +  5 log(  H  )  -  K  , where  H   represents theskip distance of a propagating wave and  K   represents thecontribution by bottom and surface reflections (Marsh andSchulkin 1962). H (as kiloyards) is further defined to equal[1/8(depth to bottom in feet  +  depth to thermocline infeet)] 1/2 (Marsh and Schulkin 1962). By calculating TLover the appropriate range of empirical and reference-based bottom and thermocline depths, we determined thataccounting for the  H   term could increase our TL estimates(and thus raise our SL estimates) by no more than 4 dB.Further accounting for  K   would likely offset this theoreticalmaximum to some unknown degree. Thus, in the absenceof the data necessary to accurately calculate  H  , we reliedon the main term 15 log(distance), noting that our resultsare conservative by as much as, but likely far less than,4 dB.Using this equation for TL, the SL of a vessel wascalculated as its RL plus TL over the distance between thesource and the receiver (Ross 1976; Urick  1983). Three estimates of the SLs (associated with the three estimates of RL) for each vessel at its CPA and three estimates of distances from CPA locations to isopleths (contours thatmap the spatial extent of a sound within specified intensitybounds) of interest were calculated.To estimate the total relative contributions of differentsampled vessel types to the low-frequency (71- to 141-Hz)band of the sampled region during 2006, average SLs forvessels of each type were converted from decibels toabsolute intensity (watts). Absolute intensity per vesseltype was then multiplied by the total number of hours thateach type spent in the sampling region in 2006. Theseestimates of total absolute intensity per year per vessel typewere then converted back to decibels and scaled relative tothe lowest contributing vessel type. Results Automatic Identification System Data  Number and Types of Vessels Commercial vessels accounted for 78% of AIS-trackedvessels transiting the study area in 2006, with the remaining Environmental Management  1 3
Search
Similar documents
View more...
Tags
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks