Nicholas Makris and Kerry Emanuel
Total Page:16
File Type:pdf, Size:1020Kb
XzV-V-03-001 C2 OCEAN ACOUSTIC HURRICANE DETECTION AND CLASSIFICATION Nicholas Makris and Kerry Emanuel MITSG 03-31 ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ OCEAN ACOUSTIC HURRICANE DETECTION AND CLASSIFICATION NicholasMakris and Kerry Emanuel MITSG 03-31 SeaGrant College Program MassachusettsInstitute of Technology Cambridge,Massachusetts 02139 NOAA Grant No.: NA86RG0074 ProjectNo.: RCM-9 Ocean Acoustic Hurricane Detection and Classification Completion Report 2003 Princi le Investi ator: Nicholas Makris MassachusettsInstitute of Technology phone: 617-258-6104 fax: 617-253-2350 email: makris .mi t,edu Co-Princi leInvesti ator: Kerry Emanuel MassachusettsInstitute of Technology phone: 617-253-2462 email: emanuel!texmex.mit.edu Objective: Thepurpose of thisproject is to investigatethe feasibility of usingunderwater acousticsensing techniques to safelyand inexpensively estimate the destructive power of a hurricane. Introduction: We have experimentally demonstratedthat Underwater Acoustic methods may be usedto measurethe wind speedand destructive power of a hurricane,using data from a NOAA hydrophone in the North Atlantic. From this hydrophone data we determined an empiricalrelationship I-V" betweena hurricane'swind speedV and the underwater noiseintensity I it generates.Given this relationshipwe showthat acousticintensity measurementscan be used to accurately estimate hurricane wind speed.To find this relationship, wind speeddata had to be synthesizedfrom satellite and aircraft meteorological measurements.While the results strongly support the use of underwater hydrophone to estimate wind speed,further experiments, with concurrent acoustic and meteorological measurements,are necessarybefore such an acoustic hurricane wind speedmeasurement system could be implemented in practice. This project also includes theoretical modeling of the generation and propagation of hurricane noise through the ocean wave-guide environment. The theoretical modeling shows that underwater hydrophonescould be used to accurately measurethe wind speeds in a hurricane, and from that measurementclassify the destructive power of the hurricane.A singlehydrophone, deployed such that a hurricanepasses over it, can be usedlike an anemometer,measuring the hurricane'swinds asit movesoverhead. Long arrays of hydrophones, in certain deep ~ater ocean environments, can measurethe destructive power of a hurricane at ranges of severalhundred kilometers. Hurricanes are one of the most deadly and costly natural disasters.In 1970 a single storm killed more than 300,000 people in Bangladesh Holland 1993! and in 1900 an unnamedhurricane became the mostdeadly disaster in United Stateshistory killing over 6000 people Hebert 1993!. In 1992 hurricane Andrew becamethe most costly natural disaster in the United Statesat an estimated cost of 2S billion dollars Pielke 1998!.Classification of a hurricane'sdestructive power is critical for hurricaneplanning. For example,inaccurate classification can leadto poor forecastingand unnecessary evacuations Emanuel 1999!, which are costly, or missed evacuations,which can result in loss of life. These fatalities and cost can be reduced if the public is given timely and accurateadvanced warning, but this dependson the ability to accurately classify hurricanes while they are still far from land. Currentlythere are two platformsfor measuringhurricanes while far from land, satellitesand aircraft. While satellitesare useful for detectingand trackinghurricanes they arenot very accuratein determiningthe wind speedand wind power neededto classify a hurricane. The most common method for satellite hurricane classification, the Dvorak method,has been known to yield errorsin wind speedestimates as high as40% Pasch 2000, Franklin 2003, Avila 2001, Stewart 2002!. In the North Atlantic this limitation is overcomethrough the useof reconnaissanceaircraft that are capableof making accuratewind speedmeasurements and hurricane classifications. Unfortunately the expenseof theseaircraft preventstheir useoutside the United States Holland 1993!. In this project we consider how underwater acoustic measurementsof hurricane wind speedand power may provide an additionaltechnique to augmentsatellite and aircraft measurementsand improve our ability to accuratelyclassify hurricanes. At frequenciesabove 10 Hz and wind speedsbelow 35 kts, it hasbeen shown that the oceanambient noise field receivedby a singlehydrophone depends significantly on the local surfacewind speed Piggott 1964,Perrone 1970, Farmer 1984,Hodgkiss 1990, Chapman1993!. This relationshiphas previously been used to estimatethe ambientnoise basedon wind speedmeasurements Cato 1997!,or conversely,to estimatethe wind speedbased on ambientnoise measurements Evans 1984,Katz 1984,Nystuen 1997!. In this project we considerestimating the wind speedsin a hurricaneby measuringthe ambientnoise it generates. Experimental Wor k: In 1999NOAA's Pacific Marine EnvironmentalLaboratory PMEL! deployed severalautonomous hydrophones near the Mid-Atlantic Ridge to continuouslyrecord the low frequency,1 to 50 Hz, ambientnoise in the ocean Fox 2001, Smith2002!. While the purposeof this project wasto study seismicevents, on September15 of 1999, hurricaneGert Lawrence2000! passed@rect! y overone of thesehydrophones fortuitously yielding a clearrecording of the underwatersound generated by the hurricane.Figure 1 showsthe spectrogramof the acousticfield measuredby the hydrophoneas the hurricanepassed overhead. This figure showshigh acousticlevels at around 14:00 as the hurricane was overhead Spectragramat NOAAhydrophone 9 duringhurricane Gert ~ 20 H~~30 LL$5-Sep-1M909.'02:67 is-sep-'N99914:$8.'l7t5-Ssp-f999!934:58 Time dd-mmm-yyyyHH:MM:SS! Figure 1: Spectrogramof the acousticfield at the NOAA hydrophoneduring the time hurricaneGert passedoverhead. The high levelsat around14:00 are due to the high winds of the hurricane. %e also see a decrease.in the acoustic level at 14:18 as the calm eye of the hurricane passedover the sensor, Figure 1 showsthe averagepower spectral level SL of the noise,between 10 and50 Hz andaveraged over 10 min time windows,during the hurricane'spassage. No direct measurementsof the wind speedwere takenat the hydrophonelocation so, to correlate spectrallevel with wind speed,we synthesisethe local wind speed record based on availabledata from US Air Forcehurricane hunting aircraft and NOAA hurricane trackingsatellites. Figure 1 showsthe wind speedprofile of thestorm from aircraft measurementson Sept 16th,a dayafter the acousticmeasurements HurricaneResearch Division1999!; clearly shown is thecharacteristic hurricane structure with the powerful winds,up to 47m/s 92kt!, of theeye wall surrounding the calm center or eye.Satellite locationestimates were provided every six hourswith anerror of +/- 28 km National WeatherService 1999!, effectively yielding an ensemble of possibletracks hurricane Gert could havetaken past the hydrophone.The meansatellite estimated track showsthat hurricaneGert passed 11 km to thesouth of thePMEL hydrophone. Each possible track, coupledwith the aircraft measured wind speed profile, can be used to synthesisethe local hurricanewind speed V at thehydrophone location. Figure3: %ind speedprofile, in mlsat a heightof 10m, of hurricaneGert from US Air Forcereconnaissance aircraft measurements taken on September 16th 1999 Hurricane ResearchDivision 1999!. The low wind speed area at the center is thchurricane's eye surroundedby thehigh winds of theeyc.-wall. The solid linc gives the path of thcPMEL hydrophonerelative to thehurricane based on the likely track showing that hurricane Gertpassed 29 km to thesouth of thePMEL hydrophone. Foreach possible track the power-law hypothesis is tested by calculatingthe hnear regressionbetween the measured spectral level SL andthe log of thelocal wind speed V. Themost likely hurricane track is theonc where the root-mean-square error RMSE!of thelinear regression between the spectral level and the log of thewind speed is minimised,equalling 0.7. For this likely track, the black line, in Fig3 showsthc path of thePMEL hydrophone relative to thehurricane on Sept 15th, In otherwords hurricane Gertpassed 29 km to thesouth of thePMEL hydrophone; this is withinthe +1- 28 km marginor errorof thesateHite estimated track, Note how the powerful eye wall andthe calm eye of the hurricane passedover the hydrophone. From this likely track Fig 2 shows the local hurricanewind speedabove the hydrophonewith the maximumwinds of the eye wall occurring at 13:30 and 15:30 hours and the minimum winds of the eye at 14:30. ln Fig. 1 we see an exceptional correlation between the measurednoise spectral level and the local wind speed. Plotting the noise spectral level versus wind speedfrom hurricane Gert Fig. 4! further illustratestheir closecorrelation. The linearregression of this datais alsoplotted and we find a I f!-V'~' or SL=33.6log V!+37.1relationship between wind speedV and noise intensity I f!. While similar relationships have been measuredpreviously at wind speeds below 15 m/s, this is the first datato suggestsuch a relationshipfor the high windsof a hurricane. This strong relationship between underwater noise and local wind speedsuggests that hydrophonescould be used as 'underwater anemometers' for estimating the surface wind speedsin a hurricane and from the wind speedclassifying the hurricane's destructive power.The powerof a hurricaneis proportionalto the cubeof the hurricane'smaximum wind speed Holland1997! which occursin the eye wall. So a hydrophonedeployed to passunder a hurricane'seye wall