• Biggerstaff, M. I., and R. A. Houze, 1991: Kinematic and precipitation structure of the 10–11 June 1985 squall line. Mon. Wea. Rev.,119, 3034–3079.

  • Black, P. G., J. R. Proni, and J. C. Wilkerson, 1995a: Oceanic rainfall detection, classification, and measurement in tropical and sub-tropical mesoscale convective systems using underwater acoustic methods. Fifth Int. Conf. on Precipitation, Elounda, Crete, NOAA, 6020.

  • ——, ——, ——, and C. E. Samsury, 1995b: Classification of convective and stratiform regions in oceanic tropical and sub-tropical mesoscale convective systems using underwater acoustic methods. Preprints, 27th Conf. on Radar Meteorology, Vail, CO, Amer. Meteor. Soc., 237–239.

  • Bom, N., 1969: Effect of rain on underwater noise level. J. Acoust. Soc. Amer.,45, 150–156.

  • Burgess, A. S., and D. J. Kewley, 1983: Wind-generated surface noise source levels in deep water east of Australia. J. Acoust. Soc. Amer.,73, 201–210.

  • Carbone, R. E., and L. D. Nelson, 1978: The evolution of raindrop spectra in warm-based convective storms as observed and numerically modeled. J. Atmos. Sci.,35, 2302–2314.

  • Churchill, D. D., and R. A. Houze Jr., 1984: Development and structure of winter monsoon cloud clusters on 10 December 1978. J. Atmos. Sci.,41, 933–960.

  • Collier, C. G., S. Lovejoy, and G. L. Austin, 1980: Analysis of bright bands from 3-D radar data. Preprints, 19th Conf. on Radar Meteorology, Miami Beach, FL, Amer. Meteor. Soc., 44–47.

  • Czerwinski, N., and W. Pfisterer, 1972: Typen von regentropfenspektren von polaren bis zu tropischen zonen und ihre abhangigkeit von elementen des regens (in German). Meteor. Rundsch.,25, 88–94.

  • Diem, M., and R. Strantz, 1971: Typen der regentropfenspektren II: Abhangigkeit von der regenintensitat (in German). Meteor. Rundsch.,24, 23–26.

  • Dingle, A. N., and K. R. Hardy, 1962: The description of rain by means of sequential raindrop-size distributions. Quart. J. Roy. Meteor. Soc.,88, 301–314.

  • Evans, D. L., and D. R. Watts, 1981: Wind speed and stress at the sea surface from ambient noise measurements. Int. Symp. on Acoustics, Part III, Calgary, AB, Canada, University of Calgary, 69–78.

  • Everest, F. A., R. W. Young, and M. W. Johnson, 1948: Acoustical characteristics of noise produced by snapping shrimp. J. Acoust. Soc. Amer.,20, 137–142.

  • Farmer, D. M., and D. D. Lemon, 1984: The influence of bubbles on ambient noise in the ocean at high wind speeds. J. Phys. Oceanogr.,14, 1762–1778.

  • ——, and S. Vagle, 1988a: On the determination of breaking surface wave distributions using ambient sound. J. Geophys. Res.,93, 3591–3600.

  • ——, and ——, 1988b: Observations of high frequency ambient sound generated by wind. Sea Surface Sound, B. R. Kerman, Ed., Kluwer Academic, 403–415.

  • Franz, G. J., 1959: Splashes as sources of sound in liquids. J. Acoust. Soc. Amer.,31, 1080–1096.

  • Fujiwara, M., 1965: Raindrop-size distribution from individual storms. J. Atmos. Sci.,22, 585–591.

  • Heindsmann, T. E., R. H. Smith, and A. D. Arneson, 1955: Effect of rain upon underwater noise levels. J. Acoust. Soc. Amer.,27, 378–379.

  • Houghton, H. G., 1968: On precipitation mechanisms and their artificial modification. J. Appl. Meteor.,7, 851–859.

  • Houze, R. A., Jr., 1973: A climatological study of vertical transports by cumulus-scale convection. J. Atmos. Sci.,30, 1112–1123.

  • ——, 1993: Cloud Dynamics. Academic Press, 573 pp.

  • Johnson, M. W., F. A. Everest, and R. W. Young, 1947: The role of snapping shrimp (Crangon and Synalpheus) in the production of underwater noise in the sea. Biol. Bull.,93, 122–138.

  • Jorgensen, D. P., and P. T. Willis, 1982: A Z–R relationship for hurricanes. J. Appl. Meteor.,21, 356–366.

  • Joss, J., and A.Waldvogel, 1969: Raindrop size distribution and sampling size errors. J. Atmos. Sci.,26, 566–569.

  • Knudsen, V. O., R. S. Alford, and J. W. Emling, 1948: Underwater ambient noise. J. Mar. Res.,7, 410–429.

  • Lamp, R., 1958: Das tropfenspektrum in niederschlagen und die radar-reflektivitat nach fremden und eigenen messungen (in German). Beitr. Phys. Atmos.,30, 223–245.

  • Laville, F., G. D. Abbott, and M. J. Miller, 1991: Underwater sound generation by rainfall. J. Acoust. Soc. Amer.,89, 715–721.

  • Laws, J. O., and D. A. Parsons, 1943: The relation of raindrop size to intensity. Trans. Amer. Geophys. Union,24, 452–460.

  • Lemon, D. D., D. M. Farmer, and D. R. Watts, 1984: Acoustic measurements of wind speed and precipitation over the continental shelf. J. Geophys. Res.,89, 3462–3472.

  • Lokken, J. E., and N. Bom, 1972: Changes in raindrop size inferred from underwater noise. J. Appl. Meteor.,11, 553–554.

  • Maddox, R. A., 1980: Mesoscale convective complexes. Bull. Amer. Meteor. Soc.,61, 1374–1387.

  • Marshall, J. S., and W. M. Palmer, 1948: The distribution of raindrops with size. J. Meteor.,5, 165–166.

  • McGlothin, C. C., Jr., 1991: Ambient sound in the ocean induced by heavy precipitation and the subsequent predictability of rainfall rate. M.S. thesis, Dept. of Physics, Naval Postgraduate School, 62 pp. [Available from U.S. Naval Postgraduate School, 1 University Circle, Monterey, CA 93947.].

  • Medwin, H., A. Kurgan, and J. A. Nystuen, 1990: Impact and bubble sound from raindrops at normal and oblique incidence. J. Acoust. Soc. Amer.,88, 413–418.

  • ——, J. A. Nystuen, P. A. Jacobus, D. E. Snyder, and L. H. Ostwald, 1992: The anatomy of underwater rain noise. J. Acoust. Soc. Amer.,92, 1613–1623.

  • Nystuen, J. A., 1985: Underwater ambient noise measurements of rainfall. Ph.D dissertation, University of California, San Diego, 86 pp. [Available from Scripps Institution of Oceanography—Archives, University of California, San Diego, 8602 La Jolla Shores Drive, La Jolla, CA 92093.].

  • ——, 1986: Rainfall measurements using underwater ambient noise. J. Acoust. Soc. Amer.,79, 972–982.

  • ——, 1993: An explanation of the sound generated by light rain in the presence of wind. Natural Physical Sources of Underwater Sound, Kluwer Academic, 659–668.

  • ——, 1996: Acoustical rainfall analysis: Rainfall drop size distribution using the underwater sound field. J. Atmos. Oceanic Technol.,13, 74–84.

  • ——, and D. M. Farmer, 1987: The influence of wind on the underwater sound generated by light rain. J. Acoust. Soc. Amer.,82, 270–274.

  • ——, and ——, 1989: Precipitation in the Canadian Atlantic Storms Program: Measurements of the acoustic signature. Atmos.–Ocean,27, 237–257.

  • ——, L. H. Ostwald, and H. Medwin, 1992: The hydroacoustics of a raindrop impact. J. Acoust. Soc. Amer.,92, 1017–1021.

  • ——, C. C. McGlothin, and M. S. Cook, 1993: The underwater sound generated by heavy rainfall. J. Acoust. Soc. Amer.,93,3169–3177.

  • ——, J. R. Proni, C. A. Lauter Jr., J. Bufkin, U. Rivero, M. Boland, and J. C. Wilkerson, 1994: APL disdrometer evaluation. NOAA Tech. Memo. ERL AOML-83, 43 pp. [Available from ERL/NOAA/AOML Library, 4301 Rickenbacker Causeway, Miami, FL 33149.].

  • ——, ——, P. G. Black, and J. C. Wilkerson, 1996: A comparison of automatic rain gauges. J. Atmos. Oceanic Technol.,13, 62–73.

  • Prosperetti, A., L. A. Crum, and H. C. Pumphrey, 1989: The underwater noise of rain. J. Geophys. Res.,94, 3255–3259.

  • Pumphrey, H. C., and L. A. Crum, 1990: Free oscillations of near-surface bubbles as a source of the underwater noise of rain. J. Acoust. Soc. Amer.,87, 142–153.

  • ——, ——, and L. Bjorno, 1989: Underwater sound produced by individual drop impacts and rainfall. J. Acoust. Soc. Amer.,85, 1518–1526.

  • Rosenfeld, D., E. Amitai, and D. B. Wolff, 1995: Classification of rain regimes by the three-dimensional properties of reflectivity fields. J. Appl. Meteor.,34, 198–211.

  • Rowland, J. R., 1976: A comparison of two different raindrop disdrometers. Preprints, 17th Conf. on Radar Meteorology, Seattle, WA, Amer. Meteor. Soc., 398–405.

  • Scrimger, J. A., 1985: Underwater noise caused by precipitation. Nature,318, 647–649.

  • ——, D. J. Evans, G. A. McBean, D. M. Farmer, and B. R. Kerman, 1987: Underwater noise due to rain, hail, and snow. J. Acoust. Soc. Amer.,81, 79–86.

  • ——, ——, and W. Yee, 1989: Underwater noise due to rain: Open ocean measurements. J. Acoust. Soc. Amer.,85, 726–731.

  • Steiner, M., and R. A. Houze Jr., 1993: Three-dimensional validation at TRMM ground truth sites: Some early results from Darwin. Preprints, 26th Int. Conf. on Radar Meteorology, Norman, OK, Amer. Meteor. Soc., 417–420.

  • ——, ——, and S. E. Yuter, 1995: Climatological characterization of three-dimensional storm structure from operational radar and rain gauge data. J. Appl. Meteor.,34, 1978–2007.

  • Takahashi, Y., 1935: Ueber regen und regentropfen (ertse Mitteilung) (in German). J. Meteor. Soc. Japan,13, 67–73.

  • Tokay, A., and D. A. Short, 1996: Evidence from tropical raindrop spectra of the origin of rain from stratiform and convective clouds. J. Appl. Meteor.,35, 355–371.

  • Urick, R. J., 1951: Some directional properties of deep-water ambient noise. Naval Research Laboratory Rep. 3796, 58 pp. [Available from U.S. Naval Postgraduate School, 1 University Circle, Monterey, CA 93947.].

  • ——, 1975: Principles of Underwater Sound. McGraw Hill, 423 pp.

  • Vagle, S., W. G. Large, and D. M. Farmer, 1990: An evaluation of the WOTAN technique of inferring oceanic winds from underwater ambient sound. J. Atmos. Oceanic Technol.,7, 576–595.

  • Waldvogel, A., 1974: The No jump of raindrop spectra. J. Atmos. Sci.,31, 1067–1078.

  • Wenz, G. M., 1962: Acoustic ambient noise in the ocean: Spectra and sources. J. Acoust. Soc. Amer.,34, 1936–1956.

  • Zipser, E. J., 1982: Use of a conceptual model of the life cycle of mesoscaleconvective systems to improve very short range forecasts. Nowcasting, K. Browning, Ed., Academic Press, 191–204.

  • View in gallery

    Major underwater sound-generating mechanisms of raindrops as a function of drop diameter for the drop size range 0.3–5.0 mm (after Medwin et al. 1992).

  • View in gallery

    A composite of sound spectra for the major components of ambient sound in the ocean showing the spectrum shape and level for each source in the frequency range 0.1–50 kHz. For ship traffic and wind components, a range of spectral levels is shown (after Urick 1975). The spectrum for snapping shrimp is a representative worst-case example of biological noise. Rain spectra are measurements recorded by the authors during MCS rainfall on 2 October 1993 in Miami, Florida. The 4–30-kHz frequency band of interest in the analysis of rainfall sound spectra has been highlighted for reference.

  • View in gallery

    Sample rain sound spectra Sr(f, t) measured by other investigators. Those by Bom (1969), Scrimger (1985), and Nystuen (1985) were recorded in lakes. Measurements by Scrimger et al. (1989) were made in the coastal ocean west of Vancouver Island, British Columbia, Canada. Observations by McGlothin (1991) and Nystuen et al. (1993) were taken in the Gulf of Mexico. The Heindsmann et al. (1955) measurement was made in the coastal waters of Long Island Sound, New York.

  • View in gallery

    Schematic vertical cross section of an MCS depicting phases of the system and showing the convective, transition, and stratiform regions (adapted from Biggerstaff and Houze 1991). The early-convective region was added by the authors.

  • View in gallery

    Measurement configurations at the monitoring sites: (a) brackish water pond, Virginia Key, Miami, Florida; (b) Carysfort Reef, Key Largo, Florida; (c) the Army Corps of Engineer’s Research Pier, Duck, North Carolina. Instrument symbols are defined in appendix B, Fig. B1.

  • View in gallery

    (a) Ambient sound levels at the three acoustically different monitoring sites used in this study, measured during the prevailing wind speeds indicated. The coastal sites at Carysfort Reef, Florida, and Duck, North Carolina, where biological activity and surf were major sources of ambient noise, sound levels were as much as 25 dB higher than at the pond where shielding from wind and the absence of marine life resulted in very low levels of background noise. Consequently, the threshold of detection of rain was 1.0 mm h−1 at the coastal sites and an order of magnitude lower at the pond. (b) Ambient sound levels for the coastal sites at Carysfort and Duck during wind speeds of 8 m s−1, and at the pond for a wind speed of 1.5 m s−1, show the wind-noise effect on the ambient background sound level below 10 kHz (see also Fig. 2).

  • View in gallery

    (a) Radar reflectivity from the National Weather Service NEXRAD WSR-88D radar at Miami for a 100 km × 100 km area at 1313 UTC 2 October 1993 when a convective cell,associated with a broad MCS to the southeast, approached the brackish pond site. Time-lapse imagery indicates a cyclonic circulation in the stratiform region centered south of the brackish water pond (denoted by a plus symbol). Curved lines and arrows indicate radar-derived streamlines of the midlevel flow. Low-level motion is also shown. (b) As in (a) except for 1342 UTC when the brackish pond was within stratiform precipitation. Additionally, open arrows indicate direction of upper-level flow. The dBZ scale for (a) and (b) is shown in the lower-right corner of (b). Ground return around the radar site has been removed.

  • View in gallery

    (a) Base-scan horizontal radar reflectivity fields (260 km × 260 km) at 1318 and 1429 UTC 2 October 1993 for the rain event at the brackish water pond. The area centered on the monitoring site has been enlarged to reveal details of the reflectivity measurements at 1 km × 1 km resolution. (b) Results of the convective–stratiform separation analysis for the reflectivity fields in (a) at 2 km × 2 km grid resolution. Enlargements of the area surrounding the pond site (outlined in white) are provided for reference.

  • View in gallery

    A presentation of the brackish pond monitoring site on Virginia Key, Miami. For reference, 1 km × 1 km and 2 km × 2 km squares, representative of the grid resolutions of the radar reflectivity and rain-type classification analysis, have been placed over the pond site. The area of the pond is approximately 0.01 km × 0.01 km.

  • View in gallery

    Time histories of the radar reflectivities and theoretical reflectivities (derived from disdrometer data) for the rain event of 2 October 1993. Rain-rate observations from both disdrometer and weighing rain gauge are also shown. Radar-based rain-type classifications are indicated at 6-min intervals (C—convective, S—stratiform).

  • View in gallery

    (a) Contour plot of the time evolution of drop size distributions from 1300 to 1500 UTC 2 October 1993; (b) corresponding time evolution of the ratioed rain spectra SR(f). Vertical lines indicate the boundaries for the early-convective (1309–1312 UTC), convective (1312–1330 UTC), and stratiform (1336–1450 UTC) phases of the event. The horizontal line in (a) indicating D > 2.2 mm and in (b) indicating 10 kHz are provided for reference. Radar-based rain-type classifications are indicated as in Fig. 10.

  • View in gallery

    (a) Representative drop size distributions and corresponding rainfall sound spectra, Sr(f, t), for the early-convective phase of the event of 2 October 1993 at the brackish water pond; (b) as in (a) but for the convective phase.

  • View in gallery

    (c), (d) Same as Figs. 12a,b but for the transition and stratiform phases.

  • View in gallery

    (a) Time series of the band-averaged sound spectrum levels Sr(f) for the 4–10-kHz and 10–30-kHz frequency bands for the rain event of 2 October 1993; (b) as in (a) but for the ratioed rain sound spectrum SR(f). Radar-based rainfall classifications are indicated as in Fig. 10.

  • View in gallery

    Time series of the discriminant,radar-derived rainfall classifications, and rain rate for the event of 2 October 1993. Rain-type classifications are indicated as in Fig. 10.

  • View in gallery

    (a) Examples of the horizontal radar reflectivity fields (20 km × 20 km) for the rain event at Carysfort Reef on 3 November 1988 during the convective (0927:14 UTC) and stratiform periods (0954:42 UTC). (b) The convective–stratiform analysis of the reflectivity fields in (a). White squares in the center of each image identify the grid location of the acoustic monitoring site. Images are at a cell resolution of 1 km × 1 km.

  • View in gallery

    Time series of radar reflectivity and rain rate for the rain event of 3 November 1988. Radar reflectivities were converted to rain rate using the relation Z = 300R1.35 (Jorgensen and Willis 1982). Shaded areas indicate time periods of missing data. Radar-derived rainfall classifications for available time periods are also shown.

  • View in gallery

    Time evolution of the ratioed rainfall sound spectra over the 0.5–30-kHz frequency band for the rain event of 3 November 1988. Radar-derived rainfall classifications are indicated as in Fig. 16. Vertical lines delineate convective rainfall period.

  • View in gallery

    Time series of the discriminant and radar-based rain-type classifications for the rain event of 3 November 1988 are shown with band-averaged sound levels for the 4–10- and 10–30-kHz bands of the ratioed sound spectrum SR(f). Rainfall classifications are indicated as in Fig. 16.

  • View in gallery

    (a) Horizontal radar reflectivity fields (210 km × 210 km) from the Cape Hatteras WSR-57 radar for the rain event at 0912 UTC 5 November 1992, during stratiform rainfall, and at 0942 UTC, during convective rain; (b) same as in (a) but for the convective–stratiform analyses. Areas outlined in white in the enlarged classification images indicate location of the monitoring site. Image resolution is 1 km × 1 km.

  • View in gallery

    Time series of radar reflectivity, radar-derived rainfall rate, and rain gauge rain rate, for the rain event of 5 November 1992. Radar-based rain-type classifications are indicated at 1-min intervals, except for 0918–0919 UTC where data were unavailable.

  • View in gallery

    Time evolution of the ratioed rainfall sound spectrum over the 0.5–30-kHz frequency band for the rain event of 5 November 1992. Spectra were processed as in Fig. 17. Radar-based classifications are indicated as in Fig. 20. Vertical lines delineate periods of convective rainfall at 0930–0932 and 0934–0949 UTC. A horizontal line at 10 kHz is shown for reference.

  • View in gallery

    Time series of the discriminant, rain rate, and wind speed for the rainfall event of 5 November 1992. Time periods of sudden bursts in wind speed are indicated by dashed lines. Radar-based rainfall classifications are presented as in Fig. 20.

  • View in gallery

    Wind sound spectra measured in the absence of rainfall. A regression fit to the wind spectra ofScrimger et al. (1987) and Vagle et al. (1990) is also shown.

  • View in gallery

    (a) Time-labeled scatterplot of band-averaged sound pressure levels of the 4–10-kHz band, SR(f, t)4–10, vs disdrometer rain rate for the brackish pond event. (b) As in (a) for reflectivity (disdrometer derived).

  • View in gallery

    Fig. A1. One-minute-averaged rain sound spectra for a rain rate of 84.2 mm h−1 in ratioed [SR(f, t)] and nonratioed [Sr(f, t)] form. Also shown is the reference background spectrum [Sb(f, t)]. Data were recorded at the brackish water pond for the case study of 2 October 1993 for the time period 0917–0918 UTC.

  • View in gallery

    Fig. B1. A schematic diagram of the Marine Acoustic Monitoring System as configured during the brackish water pond deployment, 1993–95.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 58 51 3
PDF Downloads 60 51 3

Oceanic Rainfall Detection and Classification in Tropical and Subtropical Mesoscale Convective Systems Using Underwater Acoustic Methods

View More View Less
  • 1 NOAA/Atlantic Oceanographic and Meteorological Laboratory, Hurricane Research Division, Miami, Florida
  • | 2 NOAA/Atlantic Oceanographic and Meteorological Laboratory, Ocean Acoustics Division, Miami, Florida
  • | 3 NOAA/National Environmental Satellite, Data and Information Service, Office of Research and Applications, Washington, D.C.
  • | 4 NOAA/Atlantic Oceanographic and Meteorological Laboratory, Hurricane Research Division, Miami, Florida
© Get Permissions
Full access

Abstract

Measurements of the underwater sound produced by rain were made at three U.S. coastal sites in a study to determine the feasibility and limitations of the acoustic detection and classification of rainfall over water. In the analysis of the rain sound spectra, concurrent radar reflectivity observations were used to identify convective and stratiform regions of the precipitating clouds overhead. It was found that acoustic classifications of rainfall as to type, based on information in the 4–30-kHz frequency band, were in general agreement with radar-derived classifications. The classification technique is based on use of an acoustic discriminant, D R, defined as the difference in average spectral levels between the 10–30- and 4–10-kHz bands. A high correlation was found between sound spectrum levels (in decibels) in the 4–10-kHz frequency band and radar reflectivity, dBZ, suggesting the possible use of the 4–10-kHz band sound spectral level as a classification tool using spatially distributed hydrophones in the same way that radar reflectivity is used in classifying precipitation. The results demonstrate the feasibility of the acoustic method for detecting and classifying rainfall at sea.

Corresponding author address: John R. Proni, NOAA/Atlantic Oceanographic and Meteorological Laboratory, Ocean Acoustics Division, 4301 Rickenbacker Causeway, Miami, FL 33149.

Email: proni@aoml.noaa.gov

Abstract

Measurements of the underwater sound produced by rain were made at three U.S. coastal sites in a study to determine the feasibility and limitations of the acoustic detection and classification of rainfall over water. In the analysis of the rain sound spectra, concurrent radar reflectivity observations were used to identify convective and stratiform regions of the precipitating clouds overhead. It was found that acoustic classifications of rainfall as to type, based on information in the 4–30-kHz frequency band, were in general agreement with radar-derived classifications. The classification technique is based on use of an acoustic discriminant, D R, defined as the difference in average spectral levels between the 10–30- and 4–10-kHz bands. A high correlation was found between sound spectrum levels (in decibels) in the 4–10-kHz frequency band and radar reflectivity, dBZ, suggesting the possible use of the 4–10-kHz band sound spectral level as a classification tool using spatially distributed hydrophones in the same way that radar reflectivity is used in classifying precipitation. The results demonstrate the feasibility of the acoustic method for detecting and classifying rainfall at sea.

Corresponding author address: John R. Proni, NOAA/Atlantic Oceanographic and Meteorological Laboratory, Ocean Acoustics Division, 4301 Rickenbacker Causeway, Miami, FL 33149.

Email: proni@aoml.noaa.gov

1. Introduction

This paper presents new findings on the use of underwater sound for the study and analysis of rainfall at sea. The motivation for this work was the critical need for improved observations of rainfall over the ocean. Although vital to climate modeling and prediction, reliable rainfall estimates for the world’s oceans are largely nonexistent due to the great difficulty of making accurate measurements at sea. During the past decade, underwater sound measurements of rainfall have shown that rain produces a dominant underwater signal that has the potential of yielding useful estimates of rainfall over water. However, to date a thorough investigation in the field has not been attempted. The work reported here is the beginning of a detailed analysis of field data collected at three coastal sites along the Atlantic during the period 1988–95. In this paper, typical rainfall events at each site from mesoscale convective systems (MCS) are presented and discussed in detail. Our goal is to determine the feasibility and limitations of using an acoustic method for detecting and classifying rainfall as convective or stratiform. Many of our results are consistent with the findings of others. In addition, theyshow that the underwater sound of rain is identifiable, with spectral characteristics different from the normally prevailing underwater background noise in the ocean. Based on rainfall classification analysis of concurrent radar observations, our findings indicate that it appears possible to classify rainfall as to type from measurements of the underwater sound spectrum alone.

a. Literature review—The sound production mechanisms of individual drops

Underwater sound produced by rain has been the subject of much study since World War II, when its impact on the operation of military sonars was first recognized (Knudsen et al. 1948). Nonmilitary interest developed nearly 30 years later when it was realized that sound generated by rain might serve as an estimator of rainfall rates at sea for environmental studies (Lemon et al. 1984; Scrimger 1985; Nystuen 1985). In the laboratory, Franz (1959) showed that the main sources of underwater sound in the 0–50-kHz region were the drop impact and the resonance of entrained air bubbles sometimes formed during impact. Other studies elaborated on this finding (Nystuen 1986; Prosperetti et al. 1989; Pumphrey et al. 1989; Pumphrey and Crum 1990; Medwin et al. 1990; Laville et al. 1991; Nystuen et al. 1992). Medwin et al. (1992), in laboratory studies using the full range of drop sizes found in rain, showed that the sound of rain is produced by four acoustically distinct ranges of drop diameters D. These drop size ranges are “minuscule” (D < 0.8 mm), “small” (0.8 < D < 1.1 mm), “midsize” (1.1 < D < 2.2 mm), and “large” (D > 2.2 mm), as illustrated in Fig. 1. Of the four groups, sound produced by the small and large drops were found to be most significant to the sound spectrum, as only drops in these two size groups were observed to produce the entrained microbubbles whose oscillations generate the bulk of underwater sound of rain. The Medwin et al. study also indicated that bubbles generated by small drops radiated sound at frequencies above 10 kHz, typically between 13 and 21 kHz with a peak near 15 kHz. Bubbles from large drops, on the other hand, radiated peak energy levels below 10 kHz, between 2 and 9 kHz. This inherent relationship between drop size distribution and spectral underwater sound levels has utility in the classification of rainfall solely from acoustics.

b. The underwater sound spectrum

Ambient sound in the ocean consists of the normally prevailing underwater sound emanating from natural and man-made sources. In the frequency range 0.2–50 kHz, the major sources are rain, wind, seas (including surf), marine life, and ship traffic (Knudsen et al. 1948). Figure 2 presents a composite of the sound spectra of each of these major components of underwater sound showing the approximate frequency bands occupied by them. Also shown are rain sound spectra measured at one of our sites. The relative importance of any one source to the ambient sound is dependent on time and location. For example, in the open ocean, away from ports, shipping lanes, reefs, beaches, biological activity, and surf, the most important sources of underwater sound are the wind, seas, and rain.

Ship noise is easily distinguishable from the sound of rain (Fig. 2). Ship noise dominates a range offrequencies below 1 kHz and usually occurs in discrete bands associated with operation of the ship’s engines and other equipment on board. In contrast, rain noise is broadband and extends from 0.2 to 50 kHz. For rain monitoring in the open ocean of the Tropics and subtropics, local wind is the principal source of interfering noise below 10 kHz (Urick 1975; Burgess and Kewley 1983). Wind noise is produced by agitation of the surface water and includes bubble noise created by breaking waves and surf. Wind spectra have constant negative slope with log frequency with sound levels decreasing above 500 Hz at a rate of close to 6 dB per octave (Knudsen et al. 1948; Wenz 1962; Urick 1975).

Noise from marine life occurs at frequencies ranging from a few hertz to 100 kHz (Wenz 1962). Sounds of biological origin are extremely diverse, can be intermittent or continuous, and are both narrow and broadband. Biological noise originates from only three groups of marine life: certain kinds of fish, certain kinds of shellfish, and marine mammals such as whales, dolphins, and porpoises (Knudsen et al. 1948; Wenz 1962; Urick 1975). Noise of biological origin is typically encountered in coastal regions of the Tropics and subtropics and is usually intermittent and of low frequency, although noise made by dolphins can go well above 20 kHz. An exception is the sound produced by two types of snapping shrimp. Shrimp noise is the combined sound produced by vigorous claw snapping of individual members of large colonies of shrimp found in coastal regions between 40°N and 40°S where water temperature and bottom conditions are favorable (Knudsen et al. 1948). Because it is both sustained and of high level, this noise completely dominates underwater ambient sound above 2 kHz in the vicinity of large shrimp populations (Everest et al. 1948). Noise levels decrease rapidly, however, with distance and eventually merge with the ambient background (Johnson et al. 1947; Everest et al. 1948). The effect of shrimp noise on the rain sound spectrum can be seen clearly in the example of Fig. 2, which presents a worst-case noise spectrum recorded directly above a shrimp bed. At this range, the impact of shrimp noise on a 1 mm h−1 rain spectrum is significant below 10 kHz. With distance, however, this noise source would not be a factor. It should be noted that in selecting an operational monitoring site, areas with high levels of background noise would be avoided. In summary, the sum of the background noise components present at a monitoring site establishes a noise floor, or detection threshold, for monitoring the sound of rain. Thus, for a given location and time we can define the ambient sound pressure Pa (in units of micropascals) as the sum of the sound pressures. Hence,
PaPrPwPssPmlPstPother
where r, w, ss, ml, and st denote sound pressures at the measurement site arising from rain, wind, sea state, marine life, and ship traffic.
Average spectral values of sound power, in frequency and time, can be expressed in decibels relative to 1 μP2a Hz−1 as
SXf, t10Pxrms2
where subscript X denotes ambient, rain or background; f and t denote frequency and time; and Pxrms denotes root-mean-square underwater pressure.

In this paper, we present spectral sound levels in decibels rather than in units of pressure squared, as decibel units of measure accommodate the many orders of magnitude of sound pressure levels encountered in underwater acoustic measurements. The use of decibels is also convenient when comparing acoustic pressure levels to radar reflectivities, which likewise span many orders of magnitude of signal strength. Therefore, in the acoustic analyses to follow, we express the rainfall sound level spectrum as Sr(f, t). The subscript r denotes that the rain sound level spectrum contains the prevailing background noise as well as the rain noise. We also use the rain sound spectrum in ratioed form SR(f, t), where SR(f, t) is the ratio of the rain spectrum with noise to the noise spectrum (see appendix A). Thus, the quantity SR(f, t) is called the ratioed spectrum and is denoted by the subscript R. The background sound levels used to form SR(f, t) are time-averaged ambient spectral levels existing prior to the onset of the rain and are denoted as 〈Sb(f, t)〉Δt, where the brackets indicate an average over the time interval Δt, typically 30 min.

c. The rain sound spectrum

One of the earliest measurements of the rain sound spectrum, Sr(f, t), was made by Heindsmann et al. (1955) on Long Island Sound, New York. He noted that during the passage of two “heavy” rainstorms the ambient noise spectrum levels rose 25 dB above ambient levels present just prior to the onset of rain. He also reported that “it appears that the spectrum for rain noise can be considered to extend to beyond 40 kc [40 kHz] at approximately constant level.” The high levels of rain noise observed by Heindsmann were consistent with measurements reported earlier by Knudsen et al. (1948), which showed a 19-dB increase in the 0.1–20-kHz band above the background during a “steady but not torrential rain” that occured over the Thames River at New London, Connecticut. These reports were the first to indicate the dominant nature of rain as an underwater sound source in the ocean in the frequency range of 1–50 kHz. Bom (1969) noted that wind effects and contributions from bottom and boundary reflections on measurements of the underwater noise of rain in a small shallow lake were negligible. Later, Scrimger (1985), Nystuen (1985, 1986), Nystuen and Farmer (1987), and Scrimger et al. (1987) showed that light rain over lakes, of the order of a few millimeters per hour or less, produced rain spectra that peaked near 15 kHz. While these observations indicated a correlation between the acoustic spectrum levels and rainfall rates, sufficient observations did not exist to quantify this relationship (Lokken and Bom 1972; Nystuen 1985). Lemon et al. (1984) compared ocean measurements of underwater sound levels produced by wind and rain at 4.3 and at 14 kHz and noted “a clear acoustic signal associated with precipitation.” Farmer andLemon (1984) observed that entrained wind-generated bubbles in the ocean surface layer may have an attenuating effect on ambient noise above 14.5 kHz. Nystuen and Farmer (1989) observed that the ocean surface bubble layer produced by strong winds only partially attenuates the higher-frequency components generated by rain. Nystuen (1993) showed that the spectral signature of sound for very light rainfall (e.g., 0.6 mm h−1) is changed in the presence of wind.

The laboratory model of Medwin et al. (1992) (section 1a) identified the sound-generating mechanisms and energy levels of individual drops, thereby providing a means for calculating the rain sound spectrum given the drop size distribution, and possibly the inverse. Medwin et al. (1992) modeled the sound level spectrum in the 0–20-kHz range for 100 mm h−1 rainfall and compared the results with field measurements by McGlothin (1991) for heavy rain. Nystuen et al. (1993) presented sound spectra obtained in the coastal ocean in the Gulf of Mexico that showed rather flat rain sound spectra over the 4–20-kHz band for rainfall rates of 15.7, 100, and 250 mm h−1. Sample measurements of the rain spectrum Sr(f, t) by these investigators are presented in Fig. 3. Although wind spectra overlap rain spectra below 10 kHz, in most cases depending on rainfall rates and wind speeds, the wind component adds negligible energy to the rain spectrum. This is not to say that wind effects may be ignored in the study of rainfall using underwater sound (see section 3c).

In June 1995, Black et al. (1995a) introduced the concept of an acoustically based rainfall discriminant for classification of rainfall as convective or stratiform and in October (1995b) investigated the relationship between the proposed acoustic discriminant and the drop size distribution gamma-fit parameter μ for the purpose of constructing a parameter space diagram for rainfall classification. Nystuen (1996) attempted an inversion of observed rain sound spectra to determine drop size distributions.

The results of these investigations reveal the dependent nature of the rain sound spectrum on the drop size distribution. Although details of the sound-generating mechanisms for all drops during oceanic rainfall have not yet been fully investigated, the Medwin laboratory model is assumed to closely approximate the acoustic process in the ocean. As noted in the work reported above, short-term changes in the drop size distribution produced significant changes in the rain sound spectrum. Due primarily to the partitioning of bubble sound above and below 10 kHz from the small and large drop groups in the Medwin model, quite different sound spectra are produced when large drops are absent or are present only in small numbers.

d. Tropical and subtropical mesoscale convective systems and rain-type classification

Our study focuses on rainfall from tropical and subtropical MCSs, a term proposed by Zipser (1982) to include a broad range of organized convective cells and their associated rainfall pattern. These formations are a complex of individual thunderstorms, sometimes arranged in lines, that create a broad contiguous area of precipitation 100 km or more in at least one direction and that often contain large areas of stratiform rain (Houze 1993). In general, an MCS in its initial stage is characterized by heavy convective showers with high rain rates from developing and intensifying convective cells within the complex. As thesystem matures, the older convective cells weaken and are advected rearward where they often combine to form a trailing region of stratiform rain with reduced rain rates. The change from convective to stratiform rainfall occurs in a transition zone separating the two regions. When convective cell formation ends, the entire system enters a decay phase where the structure develops into a broad area of slowly dissipating stratiform rain with weak convective cells embedded in it. A conceptual model of a convective line with trailing stratiform after Biggerstaff and Houze (1991) is shown in Fig. 4.

Numerous attempts have been made to define convective and stratiform rainfall regions, especially within the tropical MCS, using observations from drop size distributions, satellites, and ground-based Doppler radars. The link between DSD and rain type has been studied by numerous investigators, all of whom noted marked changes in the drop distribution with changes in rain type or rain intensity (Takahashi 1935; Laws and Parsons 1943; Marshall and Palmer 1948; Lamp 1958; Dingle and Hardy 1962; Fujiwara 1965; Diem and Strantz 1971; Czerwinski and Pfisterer 1972). Waldvogel (1974) observed that the intercept parameter N0 of his Marshall–Palmer exponentially fitted drop distributions did not always remain constant. He noted that abrupt changes in the drop size distribution due to changes in rain type, indicated by radar observations, were easily recognized as “jumps” in N0. When these abrupt changes occurred, Waldvogel observed shifts from large-drop- to small-drop-dominated distributions, suggesting that “convective activity” was replaced by “uniform” rain. Rainfall types in the MCS were originally defined on the basis of thresholds in infrared temperatures measured from satellites (Maddox 1980). More recently, the detailed observations available from radar have been used by various investigators to define precipitation type based upon vertical air motion, radar reflectivity structure, radar reflectivity of the bright band, and variations in reflectivity with rain rate (Houghton 1968; Houze 1973; Collier et al. 1980; Churchill and Houze 1984; Steiner and Houze 1993; Steiner et al. 1995; Rosenfeld et al. 1995; Tokay and Short 1996). In the case studies to follow, we use the radar-based Steiner et al. (1995) method to identify precipitation type.

2. Site descriptions

Our measurements were made at three sites along the Atlantic coast, two in ocean regions, and one at a nearshore brackish water pond. The coastal ocean sites were close to Carysfort Reef, approximately 10.5 km off Key Largo, Florida, and at the Army Corps of Engineers research pier at Duck, North Carolina. The brackish water pond is a small body of water adjacent to our laboratory on Virginia Key in Miami (Fig. 5).

The deployments to the coastal ocean sites at Carysfort Reef in 1988 and the Duck Pier in 1992 were first attempts at collecting field data where the major focus was on engineering problems associated with instrument deployment and recovery and in the development of an acoustic measurement capability. Much of the data recorded there were incomplete. However, someof those measurements clearly support the findings obtained later in the brackish water pond in 1993 where instrumentation and monitoring techniques were developed fully. A discussion of instrumentation and sampling procedures used at each site is given in appendix B. For the present study, a typical event from each location is presented and discussed in detail.

At each of these sites, a submerged, near-bottom, omnidirectional hydrophone was used to record the underwater sound produced by rain. Associated with each hydrophone was an ocean surface monitoring area. The size of the area at each site varied due to the different depths used for deployment of the hydrophones. Calculation of the surface monitoring area is based upon a dipole model of rainfall as a surface sound source (Urick 1951, 1975; Farmer and Vagle 1988b). In this study, surface area radii were estimated using the approximation that the surface area radius equals three times the hydrophone depth. However, since the hydrophones used were omnidirectional, underwater sounds from other sources outside the surface monitoring area such as waves (including surf), marine life, and ships were also received at times.

a. The brackish water pond

The brackish water pond is approximately 70 m by 150 m and extends to a depth of about 2 m. The bottom is composed of mud mixed with decaying vegetation. The pond is sheltered by mangrove trees to a height of 5–7 m. Acoustic monitoring was done near the center of one end of the pond with a single hydrophone submerged to a depth of 1.5 m, which provided a surface monitoring area with a radius of about 4.5 m. Contributions from bottom and boundary reflection effects resulting from the very shallow depth of the pond may be present but are thought to be negligible. Acoustic data were sampled at 100 kHz and processed as FFT (fast Fourier transform) spectra to 50 kHz. Surface wind and rain measurements were made with an anemometer and a rain gauge mounted on a fixed platform in the pond, 6 m from the hydrophone. A disdrometer was located at the pond’s edge, approximately 70 m from the platform. All instrument outputs were fed to equipment housed in a portable van located nearby (Fig. 5a). Base scan radar reflectivity data were recorded at 6-min intervals from the Miami, Florida, National Weather Service (NWS) WSR-88D, Next Generation Weather Radar (NEXRAD) located 29 km to the southwest of the pond. The radar data were used to document the passage of rain events over the brackish pond. Several hundred rain events were recorded at the pond between 1993 and 1995. The brackish pond had very low levels of background noise. Largely devoid of marine life and almost completely sheltered from the wind, the background sound levels there were as much as 25 dB lower than at either coastal site.

b. Carysfort Reef

Carysfort Light (an abandoned lighthouse), located on the reef’s edge 10.5 km offshore from Key Largo, was instrumented to measure surface wind and rain for the acoustic site that was situated 4.5 km northeast of the lighthouse in 45 m of water (Fig. 5b). At this depth, the area of sea surface monitored was approximately 135 m in radius. A Sea Data model 661 Wind Observation Through Ambient Noise (WOTAN) recorder was deployed a few feet above the hard, sandy bottom to record continuous sound spectra from 12 channels in the 0.5–30-kHz frequency band. Further details of the WOTAN system are found in Evans and Watts (1981) and Vagle et al. (1990). Radar coverage was provided by the Miami, Florida, NWS WSR-57 radar located in Coral Gables, Florida, approximately 50 km to the north of Carysfort Reef. The radar data were incomplete for this case study due tofrequent interruptions in recording that produced time gaps in coverage. The radar data that were recorded were available every 20 s. In a 5-week period during October and November 1988, nine rain events were recorded. The ambient background noise at this site, next to the coral reef and adjacent to coastal shipping lanes and recreational boating areas, included the sounds of wind, ship traffic, and marine life. The most significant of these noise sources was biological, from colonies of snapping shrimp.

c. Duck Pier

Measurements at Duck, North Carolina, were made at the seaward end of the army’s research pier, 550 m from shore. A single hydrophone was located 350 m to the south of the pier at a depth of 6 m on a hard, sandy bottom. Redundant measurements were recorded with the WOTAN recorder that was deployed at the same depth several meters away. The surface listening area at this depth was approximately 18 m in radius. Data from both hydrophones were sampled at 100 kHz and processed as sound spectra to 50 kHz in van-mounted equipment located at the end of the pier (Fig. 5c). The WOTAN hydrophone output was also processed within the instrument itself to produce redundant 12-point spectra to 30 kHz. Observations of wind and rain were taken with anemometers and rain gauges mounted on the pier in the vicinity of the van. Radar data from the Cape Hatteras NWS WSR-57 radar 50 km to the south of the pier were available at 1-min intervals. In a 4-week period during October and November 1992, 15 rain events were recorded. Background noise at the Duck Pier site was due mostly to wind, waves, and surf.

Due to the dissimilar types and levels of background noise at each location, the rainfall detection threshold at each of these monitoring sites was different. Background sound levels at each site are shown in Fig. 6. Because of the relatively high background noise levels at the two coastal sites, the rain-rate measurement thresholds were slightly above 1 mm h−1, while at the relatively quiet brackish water pond the measurement level was an order of magnitude lower, 0.1 mm h−1.

3. Case studies

Three case studies, one from each monitoring site, have been selected for presentation to highlight similarities in the rain sound spectra under different conditions of ambient noise. The measurements recorded at the brackish water pond are discussed first because they are the most extensive and permit a more detailed analysis than do the other two studies. The Carysfort Reef study has incomplete radar coverage, and the Duck Pier case lacks concurrent disdrometer measurements. Nonetheless, their data provide strong supporting evidence of the robustness of the acoustic method of rainfall classification.

To evaluate the results of the acoustic classification techniques, we analyzed concurrent radar reflectivity measurements over each site using the objective convective–stratiform rainfall classification method of Steiner et al. (1995). The Steiner et al. algorithm extracts information on rainfall type, structure, and amount from horizontal radar echo patterns. For rainfall classification, we used their version of the rain type separation algorithm tuned for tropical rainfall at Darwin, Australia. While we sought to retune the algorithm for our regions, a proper database for refinement was unavailable. Nevertheless, we believe the Steiner et al. (1995) system can be properly applied to our data.

a. Brackish water pond—2 October 1993

1) Radar observations

On 2 October 1993, a tropical MCS approached Miami from the southeast, arriving over our monitoring site on Virginia Key shortly after 1300 UTC (all time references hereafter are UTC). A base scan from the MiamiNEXRAD radar (Fig. 7a) for the time just prior to the onset of heavy convective rain at the pond shows the horizontal structure of the MCS and the radar-derived environmental flow. The leading convective line is made up of several individual cells that are actively growing at this time and moving toward the northwest at about 8 m s−1. The trailing stratiform anvil region is seen to the southeast of the line. The low-level wind shear was northwesterly, with strong low-level southeasterlies dying off with height. The winds in the midlevels were light. Analysis of the animation of the NEXRAD radar reflectivity data reveals the existence of a midlevel vortex within the stratiform precipitation. Figure 7b provides a view of the system shortly after rainfall ended at the pond.

2) Radar reflectivity data analysis

Base scan radar reflectivity data were analyzed at 6-min intervals on a 400 km × 400 km Cartesian grid to document passage of the storm over the brackish water pond. Figure 8 presents the horizontal reflectivity structure and results of the reflectivity analysis at 1318 during convective rainfall (Fig. 8a) and at 1429 during the stratiform rain (Fig. 8b). An enlargement of the area centered on the pond has been added to show details of the reflectivity pattern and to enhance the lines of separation in the rainfall classification analysis.

The base scan data recorded at 1318 show the MCS in a mature stage with a line of intense reflectivity cells surrounded by a well-formed stratiform region over the area of the pond (outlined in white). Radar observations at 1429, near the end of rainfall at the pond approximately 1 h later, show a reduced stratiform region with convective activity moving toward the northwest. At this time the system is moving away from the pond site.

3) Radar reflectivity analysis for rainfall classification

The convective–stratiform separation algorithm was applied to each base scan reflectivity dataset recorded at 6-min intervals between 1307 and 1452. The grid resolution for the radar is seen in Fig. 9. Time histories of measured radar reflectivity, radar reflectivity calculated from disdrometer measurements, gauge-measured rainfall rate, and disdrometer-derived rainfall rates for the brackish pond site are shown in Fig. 10. Rainfall at 1312 until 1330 was classified as convective. The analysis indicates that at some point after 1330 but before 1336 the transition to stratiform occurred. From 1336 onward, until the end of rainfall at 1447, precipitation was classified as stratiform.

4) Reflectivity–rain rate observations

Letter notations in Fig. 10, indicating the results of the classification analysis, have been added at the appropriate 6-min intervals for reference with rain-rate time histories of both disdrometer and rain gauge observations. These data reveal that convective rainfall occurred over the pond for a period of approximately 20 min, reaching rain rates of 100 mm h−1 before changing to stratiform rain. Note the general similarity of the two reflectivity levels as a function of time, despite the great difference in measurement areas of the two systems. The spatial measurement regions for the disdrometer (an area of about 46 cm2), the acoustical hydrophone (an area about 64 m2), and the radar (a radar cell of about 1 km × 1 km encompassing the pond) are substantially different, so that differences in measurements such as onset time of rainfall of a minute or two are not unexpected.

5) Dropsize distribution and rain sound spectrum analysis

Time evolutions of the drop size distributions and the corresponding rain sound ratio spectra SR(f, t) are presented in the contoured plots of Figs. 11a and 11b. Drop distributions from 1309 to 1312, during the first 3 min of rainfall (Fig. 11a), have low concentrations of drops with near-uniform size distributions from about 0.8 to 5 mm. This time period corresponds to the early-convective phase depicted in the cross-sectional schematic of Fig. 4. Rain rates during this interval are rising rap-idly (Fig. 10) and sound levels are increasing at all frequencies (Fig. 11b). At the end of this 3-min period, convective rainfall begins and continues for an additional 18 min (1312–1330). Note the onset of change in the drop distribution at 1312. This change in drop distribution causes an instantaneous change in the sound spectra. During the convective phase, the rise in drop concentrations of all sizes produces increases in sound levels across all frequencies of the rain spectrum. At the time of transition to stratiform rain, between 1330 and 1336, there is again a significant change in the drop distribution with corresponding change in the sound spectra. As large drop concentrations decrease, there are reductions in sound levels below 10 kHz and an abrupt shift in the energy peak toward 15 kHz. With the appearance of stratiform rain, beginning between 1330 and 1336, and for the duration of the stratiform phase, both drop concentrations and sound levels decrease gradually. Drop distributions now contain few drops greater than about 2 mm in diameter, and sound levels below 10 kHz remain low. At approximately 1412 and again at about 1430 (Fig. 10), short bursts of stratiform precipitation occur. At these times there are slight increases in the concentrations of larger drop sizes in the distributions. Radar analyses show convective cells near the pond at these times that likely contributed to the increases in larger drop concentrations.

Sample drop size distributions and corresponding rain sound spectra representative of the four rainfall phases described in these data are shown in Fig. 12. In the early-convective phase (Fig. 12a), note the low concentration of small drops (0.8–1.1 mm) and the near-equal number of drops of all sizes. The acoustic spectra produced are shown in the figure below. These sound spectra reveal the high levels of acoustic energy produced, particularly below 10 kHz, by large drops (>2.2 mm). More importantly, they show that, during the early-convective phase, the sound spectral levels for the rain rates of 16.5 and 53.7 mm h−1 for the period 1310–1314 exceed the spectral sound levels for the higher rain rates of 56.2, 80.7, and 84.2 mm h−1 that occur between 1317 and 1330 during the convective phase shown in Fig. 12b. The sound produced by large drops dominates during the early-convective period, even for rain rates as low as 1.8 mm h−1 (note the quite different stratiform rain spectrum for the comparable rain rate of 1.4 mm h−1 in Fig. 12d). This suggests that a separate sound-level-to-rain-rate relationship may exist for early-convective rainfall.

Drop distributions during the convective phase (Fig. 12b) are quite different from those in the early-convective phase. The number concentration of small drops has increased greatly. The convective phase distributions also include large numbers of drops greater than 2.2 mm in diameter. These drop sizes areresponsible for rainfall sound over the broad band of frequencies below 10 kHz (section 1a).

The sudden decrease in large drops during the transition phase (Fig. 12c) is clearly reflected in the acoustic spectra. Sound levels below 10 kHz are greatly diminished with reduction of large-drop bubble sound and the loss of broadband acoustic energy from impacts. Absence of the latter allows the bubble sound generated by small drops to emerge. This appears as a spectral peak centered near 15 kHz. Drop distributions during the stratiform phase (Fig. 12d) show decreasing numbers of large drops with corresponding loss of sound below 10 kHz until rainfall ends.

6) Classification of rainfall using an acoustic discriminant

From Fig. 12 it is seen that rainfall generates sound energy in a frequency band extending from 200 Hz to 50 kHz. While in the present study only the band 4–30 kHz was utilized for rainfall analysis, clearly a much larger range of frequencies could have been used. Indeed, future work should examine other frequency bands for investigation.

It has been noted (section 1b) that the dominant rain sound originates from bubble oscillations created by both small and large drops and that these two bubble types resonate in separate frequency bands of the 4–30-kHz spectrum; that is, small-drop bubbles resonate above 10 kHz and large-drop bubbles resonate largely below 10 kHz. We have taken advantage of this frequency dependence of bubble resonance on drop size in the formulation of an acoustic parameter for rain type classification.

The acoustic parameter we introduce, called the discriminant Dr(t), is defined as the difference in band-averaged sound level between the 10–30- and the 4–10-kHz bands. That is,
Drt10–304–10
where the overbar denotes band average.
Figure 13a presents the time histories of the band-averaged sound spectrum levels for Sr(f, t). Note that the convective time period sound levels for the 4–10-kHz band exceed those in the 10–30-kHz band. However, for the stratiform time period the reverse occurs; that is, the 10–30-kHz sound levels generally exceed levels in the 4–10-kHz band. These data show that for convective rainfall periods Dr(t) is negative, while for stratiform rainfall periods Dr(t) is positive. It should be pointed out that the value of Dr(t) at a given measurement site depends in part on the background noise levels present at the site during the measurement period (section 1b; appendix A). Thus, the values of Dr(t) defining rain type will vary from site to site. To reduce this site dependency, a ratioed acoustical spectrum is used. The ratioed spectrum is defined as
SRf, tSrf, tSbf, tΔt
where the background noise spectrum 〈Sb(f, t)〉Δt is measured over a time interval Δt occurring prior to the onset of rainfall; it is, in effect, an estimate of the actual background sound level occurring during the rainfallevent. A plot of 〈Sb(f, t)〉10min for the brackish pond site for 2 October 1993 is shown in Fig. 6. Correspondingly,
i1520-0493-125-9-2014-eq3
4–104–10Sb(f, t)〉Δt4–10
for the 4–10-kHz band and
i1520-0493-125-9-2014-eq4
10–3010–30Sb(f, t)〉Δt10–30
for the 10–30-kHz band. Thus, the discriminant using ratioed acoustic rainfall spectra is
DRt10–304–10

Although both forms of the discriminant work equally well for rainfall classification, we have chosen to use DR(t), as this form [employing SR(f, t)] compensates for background noise levels, thus producing apparently site-independent values in rainfall classification. In order to conclude that DR(t) is in fact site independent, many more case studies are required than are presented in this paper.

Figure 13b shows the ratioed band-averaged rainfall sound spectrum levels corresponding to Fig. 13a. Note that for the time period defined as convective by radar classifications, 10–30 and 4–10 now very closely overlie one another. In other words, for the convective portion of the rainfall event,
i1520-0493-125-9-2014-eq6
10–304–10
Interestingly, this near equality of band-averaged sound spectrum levels for the two frequency bands holds despite a changing rainfall rate during the convective rainfall period of over 60 mm h−1 (Fig. 10). Similarly, in the case of the nonratioed acoustical spectra in Fig. 13a, a near-constant difference in levels of about 3 dB occurs between the two bands during convective rain.
From the rainfall sound spectra behavior illustrated in Fig. 13b, and from other cases observed at the pond and Carysfort sites, it has been generally observed that equality between the 10–30- and 4–10-kHz band-averaged levels holds for convective rainfall periods. It is therefore proposed that for convective rainfall periods
DRt10–304–10
and that for stratiform rainfall periods
DRt
That is to say, use ofSR(f, t) in the calculation of the discriminant produces values of DR(t) that are approximately site independent, suggesting that these classification criteria hold for all ocean areas. [For Dr(t) in Fig. 13a the corresponding values are Dr(t) ≅ −3 dB for convective rain and Dr(t) > −3 dB for stratiform rain.]

Figure 14 presents the time series of the discriminant and the results of the radar-based classifications for the event of 2 October 1993. Rain rates measured by both disdrometer and rain gauge are also shown. Note that DR(t) is in the range interval −4 to slightly less than 0 in the early-convective phase and near zero during the remainder of the convective phase. The discriminant first rises significantly above zero at about 1329, indicating stratiform rainfall, but returns to zero momentarily at 1330 before rising again above zero at 1331 where it remains until rainfall ends approximately 1 h later. According to radar classification, the changeover from convective to stratiform rainfall occurred between 1330 and 1336. The discriminant also indicates a change to stratiform after 1330. However, according to values of the discriminant, the transition at the pond appears to have begun slightly before 1330 (perhaps as early as 1328 with the momentary increase above zero at that time). This is suggested also by the time series of rain rates in this figure as well as the drop distributions in Fig. 12b where point measurements of rain rates derived from disdrometer data dropped precipitously from 19.7 mm h−1 at 1328 to 6.1 mm h−1 at 1329. The discrepancy in time for the onset of the transition phase in this case is not surprising since the classification methods using radar reflectivities and underwater acoustics are based on measurements with orders of magnitude differences in resolution, as discussed in section 3a(4) and as illustrated in Fig. 9.

The observation that DR(t) < 0 for the early-convective period suggests that if the early convective drop size distribution is found to be a consistent or important feature of convective rainfall, then in addition to the propositions that DR(t) ≅ 0 for convective rainfall and DR(t) > 0 for stratiform rainfall, the proposition that DR(t) < 0 for early-convective rainfall may be required.

b. Carysfort Reef

On 3 November 1988, a typical MCS passed over the WOTAN deployment site near Carysfort Reef, producing showers that lasted more than an hour. The MCS was generally moving north while individual cells moved from east to west. Radar coverage from the NWS radar at Coral Gables, Florida, was obtained for a 20 km × 20 km sector centered on the monitoring site. As noted earlier, and in appendix B, radar data were incomplete for this case study. The available observations were analyzed for rain-type classifications using the Steiner et al. method. Figure 15 presents examples of the reflectivity analysis for two rainfall periods, one convective and one stratiform. Unlike the 6-min sampling provided by NEXRAD for the pond case in 1993, radar coverage for this event yielded up to four base scans per minute. Because of this high sampling rate, multiple classifications within a 1-min interval were not always consistent during the transition phase from convective to stratiform (0930–0939).The classification results are presented in Fig. 16 with the time series of reflectivities as well as rain rates derived from them. Where both classifications occurred within the minute interval, both are shown (as for 0929, 0938, and 0939).

The time evolution of the rain sound spectra for this event is shown in the two-dimensional contour plot of Fig. 17. The radar-based classifications are also indicated at 1-min intervals for the time periods available. The rain spectra were derived from continuous 30-s averages from the 12-channel WOTAN recorder. For this figure, an interpolation routine was used to estimate spectral levels between frequencies. Notice that during the convective period indicated by radar (0927–0929) sound levels are high at all frequencies but that they decrease at 0931 when radar classifications indicate a transition to stratiform. The multiple classifications at 0938 and 0939, indicating some convective rain at those times, are also reflected in the sound spectra as slight increases in sound level for frequencies below 10 kHz. The periods of stratiform rainfall between 0930 and 1008 show uniformly low sound levels at all frequencies. These results are consistent with the findings from the pond site in Fig. 11.

Figure 18 displays the time series of the discriminant, the band-averaged spectra for the 4–10- and 10–30-kHz band, and the radar-based rain-type classifications. Comparison of the acoustic-derived and radar-based classifications shows that both indicate a transition from convective to stratiform rainfall at 0930 and that both classify rain as stratiform for the remainder of the comparison period 0944–1008. The discriminant values prior to 0930 are quite variable. They vary approximately 1 dB about zero, significantly more deviation than occurred during the corresponding period in the pond case (Fig. 14). The reason for this variability is uncertain. It may be that the low-resolution 12-channel WOTAN recorder used at Carysfort introduces variability in spectral lines when compared to the high-resolution spectral estimates produced at the pond site (appendix B). Consequently, the WOTAN sound spectra could provide coarser estimates of the discriminant and increased variability.

c. Duck Pier

1) Radar reflectivity analysis

A well-defined MCS rain event occurred at the Duck Pier on 5 November 1992 from 0800 to 1000 UTC. In contrast to the brackish pond and Carysfort case studies, stratiform rainfall preceded convective rainfall. Animation of radar reflectivity data at 1-min intervals revealed that the rainfall was associated with a line of precipitation, organized along a nearly stationary frontal system in the area, that was moving slowly to the southeast. The rainfall moved with the weak west-northwesterly low-level (850 mb) environmental flow, but the more intense rain cells propagated along a line to the east-northeast. Stronger midlevel flow between 700 and 500 mb was from the west-southwest, advecting longer-lived stratiform regions to the east-northeast ahead of the trailing convective precipitation. Upper-level flow, near 200 mb, was from the south-southwest and was advecting higher, newly formed stratiform rain adjacent to the convective line to the north-northeast.

Figure 19 presents examples of the radar reflectivity field and the analysis of this field for two time periods during this event, one stratiform and one convective. Figure 20 shows the time series of radar reflectivity and rain rate with the radar-derived rain-type classifications indicated at 1-min intervals. According to the analysis of radar reflectivities, rainfall was stratiform from0900 to 0934, except for a 2-min convective period (0930–0931) just prior to transition to convective rain at 0935. At 0935 and until 0947 rainfall was convective. At 0947, and until the end of the event at 1000, rainfall was stratiform.

Figure 21 shows the time evolution of the ratioed rain spectra for the time period 0900–1000. The 12-point spectra, recorded by the WOTAN, were smoothed as in Fig. 17. Note the low sound levels during the stratiform period 0900–0930. Note also the abrupt change in the spectra as sound increases to high levels at all frequencies during the 2 min of convective rain at 0930–0931 and after transition to convective rain at 0935. Low sound levels reappear with the return to stratiform rain at 0948 and remain at low levels until the end of the event at 1000. These results are also consistent with those presented for the pond in Fig. 11.

Figure 22 presents the time series of the discriminant, rain rate, and wind speed with radar-derived rain-type classification indicated at 1-min intervals as in Figs. 20 and 21. Notice that during the stratiform period, as determined by radar, the discriminant is generally positive indicating stratiform rainfall. The 2-min burst of convective rain at 0930–0931 is also sensed by the discriminant, which goes slightly negative at that time. During the convective period from 0935 to 0947, both radar and acoustic classifications are in agreement. These results are consistent with the findings from the pond (Fig. 14).

2) Effects of wind on the rain sound spectrum and the discriminant

The arrival of heavy rainfall at the Duck Pier was accompanied by abrupt changes in wind speed at about 0938 and again at 0944 as shown in Fig. 22. These sudden changes occurred in bursts lasting about 5 min each and were separated by a lull of about 1 min. Both bursts occurred during a gradual change in wind direction from 190° alongshore at 0900, to 270° offshore at 0938. The effect of the wind direction change and the sudden increases in wind speed on the acoustic spectra can be seen in the behavior of the discriminant during the wind burst periods. Note that with the onset of convective rain at 0934 (according to radar), DR(t) takes on negative values (indicating convective rainfall), ultimately reaching a value of about −3 dB before the start of the first wind event at 0938:30. As the wind speed increases from about 6 m s−1 to a maximum of 14 m s−1, DR(t) responds by decreasing further, finally reaching a value of about −6 dB at 0941:30. As wind speeds rapidly subside at 0943, DR(t) returns to −3 dB, its value prior to the first wind burst. After about 1 min, the sequence is repeated during the second wind event, 0944–0949. However, the response of DR(t) at this time is less pronounced because the rain rate has decreased substantially, causing 4–10 to become smaller than during the first wind burst; in turn, this causes DR(t) to be less negative during the second wind burst.

Observations made by Farmer and Lemon (1984), Farmer and Vagle (1988a), and Nystuen and Farmer (1989) show that absorption of wind-generated sound occurs at 19.5 and 25 kHz by a layer of near-surface bubbles formed by breaking waves. Theyobserved that such effects are particularly noticeable at wind speeds of 10 m s−1 or greater. Indeed, because of bubble absorption effects, Farmer and Lemon noted that a reduction in sound levels could occur at these frequencies even during increases in wind speed. It is suggested that the principal effect of sudden change in wind speed as occurred at the Duck Pier site is a reduction of the rainfall-generated sound levels in the 10–30-kHz band relative to the 4–10-kHz band.

Further evidence of the wind effect on the 10–30-kHz band is provided in Fig. 23. All spectra shown were obtained at times of no rain; that is, surf and wind sound spectra only. The spectra were recorded at Duck for wind speeds of about 1.3, 7.2, and 13.9 m s−1. In addition, spectra from Scrimger et al. (1987) and Vagle et al. (1990) have been included, with a regression fit to their data. In contrast to the Carysfort site, where biological sound sources were significant, surf-generated noise was dominant at the Duck site from wave breaking on both an offshore sandbar 300 m seaward of the deployment location and along the beach. Surf noise may be responsible for the sound-level floor of about 53 dB above 2 kHz for the 1.3 and 7.2 m s−1 winds. Note that, for the 13.9 m s−1 wind, sound levels above 10 kHz decrease to levels below the noise floor. This reduction in sound level with increasing wind speed occurring at about 15 kHz and higher in frequency is believed to be a consequence of bubble-layer absorption of surf noise as well as noise from other sources. It may well be that, in general, the water column in the surf zone at Duck Pier contains bubbles that serve to reduce the high-frequency content of the surf noise. The overall effect of this surf-zone bubble attenuation is to reduce the value of 10–30 and, in turn, the discriminant DR(t) to values below zero, for example, −2 dB. This is true for both convective- and stratiform-type rainfall periods.

Clearly, the reduction in values of DR(t) due to bubble-layer absorption does not affect the classification of convective rain as this results only in higher negative values. However, the possibility of occasional misclassification of stratiform rain may exist in the presence of strong winds and/or wind speed changes exceeding about 5 m s−1, the threshold for significant white cap development (Wenz 1962). It should be noted, though, that despite substantial variation in wind magnitude during the stratiform period of rainfall in Fig. 22, DR(t) remains on average greater than zero, for example, +1 dB, so that a generally correct acoustic classification of stratiform is obtained. It is also important to realize that because of the limited water column depth available for sound propagation in the surf zone at the Duck Pier site, it is likely that the bubble absorption effect observed here is much more severe than would be observed for a similar change in wind speed in the open ocean, where a spatially extensive water column is available for sound propagation.

4. Interpretation and remarks

Data from the Duck Pier and the brackish water pond show that the 4–10-kHz band-averaged sound level 4–10 is highly correlated with both radar reflectivity and dBZ and rainfall rate R. The correlation coefficients for these data have been calculated to be in excess of 0.9. The data in Figs. 24a and 24b show time-labeled scatterplots for sound level versus rain rate, 4–10R, and for reflectivity versus rain rate, Z–R. Note that when reflectivity exceeds 40 dB, so does the rain sound level. In fact, the sequence of data points in both figures show that both sets march in time much the same way. These time paths are analogous to the parametric cycles described by Carbone and Nelson (1978) for N0 and λ versus rain rate during the growth and dissipation stages of similar storm events. The important feature of the data in Fig. 24 is that they indicate different R and Z–R relationships for the various phases of the MCS.

Note that a single rain rate may be associated with either two sound level values or two reflectivity values. For example, in Fig. 24a a rain rate of 40 mm h−1 corresponds to about 37 dB in the convective region as well as to 53 dB in the early-convective region. Likewise, a 40 mm h−1 rain rate in Fig. 24b corresponds to about 43 dB in the convective region and 53 dB in the early-convective region. The separate Z–R relationships for convective and stratiform rainfall indicated here are consistent with the findings of Tokay and Short (1996) for tropical rainfall at Kapingamarangi Atoll during the Tropical Oceans Global Atmosphere Coupled Ocean–Atmosphere Response Experiment. Notice that the acoustic discriminant can be used to determine the appropriate relationship according to whether DR(t) ≤ 0 or DR(t) > 0. If DR(t) ≤ 0, the convective relationship applies, but if DR(t) > 0, the stratiform relationship is required.

5. Summary and conclusions

An initial investigation of the use of underwater sound for the study of rainfall classification was carried out using data obtained at three acoustically distinct U.S. coastal ocean sites: 1) a brackish pond near Miami, Florida; 2) offshore from Key Largo near the Carysfort Reef; and 3) in the surf zone, at Duck Pier, North Carolina. These sites had varying background noise levels. The pond site was well protected from wind and had low background sound levels. The coastal sites had background noise levels as much as 25 dB higher due to the normally prevailing sound sources of wind, breaking waves (including surf), marine life, and ship traffic. While many rainfall events were recorded and studied at each site, one event from each was selected for detailed analysis.

Underwater sound spectrum levels were measured by bottom-mounted hydrophones over the frequency band 0.5–50 kHz at the pond and Duck Pier sites, and over 0.5–30 kHz at the Carysfort Reef site (and also at the Duck site for redundancy). The sound spectrum levels produced by rainfall in the absence of wind greatly exceed the sound spectrum levels produced by winds for all but the lightest rain intensities, for example, less than 1 mm h−1. Underwater sound spectrum levels in two bands, the 4–10- and 10–30-kHz bands, denoted as 4–10 and 10–30, respectively, were the principal tools used in the analysis of rainfall events.

A discriminant,DR(t), was proposed for rainfall-type classification, defined as the difference between sound spectrum levels, 10–30 and 4–10; that is, DR(t) = 10–304–10 at a given hydrophone measurement site. The selection of band limits of 4 and 10 kHz and 10 and 30 kHz should be further investigated to see if other band limits yield results superior to those described here. The discriminant is a site-local tool for rainfall-type classification in the sense that some estimate of background sound level Sb(f, t) is required in order to estimate SR(f, t). Once SR(f, t) has been determined, reasonably site-independent values for the discriminant are obtained. These values are DR(t) ≅ 0, rainfall is convective; DR(t) > 0, rainfall is stratiform. Whether a site-independent range of values for a discriminate can be determined without recourse to determining Sb(f, t) is a subject for future study.

Fluctuations in rainfall quantities, particularly during transition, are always present so that short-term deviations of ±1 dB or greater, about zero, for DR(t) during this period, can occur. However, for the pond case during early convective rainfall, values of DR(t) less than zero were consistently observed until heavier convective rain began.

Concurrent reflectivity observations from the National Weather Service radars and processes for rain-type classification using the objective analysis method of Steiner et al. (1995) were obtained. In all cases, good agreement was obtained between acoustic-based and radar-based rainfall-type classifications.

The effect of sudden increases in wind during convective rain was observed to decrease the band-averaged sound level in the 10–30-kHz band and not in the 4–10-kHz band due to the selective nature of the absorption of rain sound in the surface bubble layer created by breaking wind waves. The consequence of this absorption on the discriminant was observed to result in negative values, for example, −2 dB, thus indicating that DR(t) ≤ 0 in the presence of a water column rich in bubbles. However, no confusion in classification of convective rains results as DR remains negative. The possibility of occasional misclassification of stratiform rain in the presence of significant wind speed changes was noted however.

A high level of correlation was observed between band-averaged rain sound levels in the 4–10-kHz band, 4–10, and radar dBZ. A similar time evolution of 4–10 and dBZ (disdrometer derived) was observed for the scatterplots of these parameters versus rain rate during the case study rain event at the pond site. Clearly, distinctly different power-law relations between dBZ and rainfall rate, and between 4–10 and rainfall rate, are apparent for each of the rainfall types observed during the pond case study. In these situations the acousticdiscriminant may also be used to determine the appropriate 4–10R versus rain rate relationship for rain-rate estimation. When used appropriately in conjunction with radar, the discriminant may be of use in determining which ZR relationship is appropriate.

The principal limitation of the acoustic methods for rainfall classification appears to be the absorption effect of a wind- or surf-generated surface bubble layer during stratiform rain. While the effect may occur in a shallow surf-zone environment such as at the Duck Pier where our observations were made, the effect of similar wind changes in the deep, open tropical ocean are expected to be far less significant because of the much greater bubble-free water column available for sound propagation. During the convective period, it was observed that DR(t) remained near zero despite a changing rainfall rate. This suggests that the rainfall drop size distribution changes in such a way as to preserve the proportion in energy radiated by its different drop sizes and components.

The results of this work suggest that acoustical methods may be used for the detection and classification of oceanic rainfall in mesoscale and climate studies. The possibility of the use of acoustical methods in rainfall estimation also appears promising. Underwater sound measurement devices deployed from buoys or mounted on the ocean bottom might be used as the equivalent of point-source ocean radars for monitoring coastal and open ocean rainfall. Design of an appropriate scanning strategy in the future might present the possibility of a true acoustic rain mapper, thus more closely approximating the concept of an oceanic rain radar.

Acknowledgments

The authors wish to acknowledge the important contributions made by those who worked on our behalf during this research. Special thanks are due to the following individuals of the Atlantic Oceanographic and Meteorological Laboratory: Charles Lauter, Jeffrey Bufkin, Paul Dammann, and Lt. Mark Pickett for engineering assistance in the design of equipment, hardware construction, and data gathering; and to Jules Craynock for his diving support in the offshore deployments and for his help in establishing the pond site. We are greatly indebted to Ulises Rivero, Mark Boland, Alejandra Lorenzo, Elizabeth Redmond, John Struck, Jack Stamates, and Michael Black for their tireless efforts in software preparation, data reduction, and graphics support, and to Gail Derr and C. C. Stephens for careful typing of the manuscript. David Wolff, Paul Kucera, and Brad Fisher of the Tropical Rainfall Measuring Mission (TRMM Office), NASA/Goddard Space Flight Center gave generously of their time in fulfilling our request for processed NEXRAD radar data. We are most grateful to Paige Bridges of the Office of Research and Applications, NOAA/NESDIS, for his excellent work in preparing the many figures and for producing the final camera-ready graphics; to Paul Chang, NOAA/NESDIS, who gave generously of his time to assist us in many ways during the processing and analysis of radar data; and to Li Li, Caelum Research Corporation, and Steve Chen, Research Data and Systems Corporation, for their programming expertise. Finally, the authors would like to thank Al Caron and Robert Kennedy of the Naval Underwater Systems Center, West Palm Beach, and Matthias Steiner, of the Department of Civil Engineering and Operations Research, Princeton University, for their critical comments and helpful suggestions during the course of this study. We also wish to thank the two anonymous reviewers whose comprehensive and critical comments were vital to the reorganization, revision, and improvement of this manuscript. This work was supported in part by Grant GC95-635 from the Office of Global Programs, NOAA, and by funding from the TRMM Office,NASA.

REFERENCES

  • Biggerstaff, M. I., and R. A. Houze, 1991: Kinematic and precipitation structure of the 10–11 June 1985 squall line. Mon. Wea. Rev.,119, 3034–3079.

  • Black, P. G., J. R. Proni, and J. C. Wilkerson, 1995a: Oceanic rainfall detection, classification, and measurement in tropical and sub-tropical mesoscale convective systems using underwater acoustic methods. Fifth Int. Conf. on Precipitation, Elounda, Crete, NOAA, 6020.

  • ——, ——, ——, and C. E. Samsury, 1995b: Classification of convective and stratiform regions in oceanic tropical and sub-tropical mesoscale convective systems using underwater acoustic methods. Preprints, 27th Conf. on Radar Meteorology, Vail, CO, Amer. Meteor. Soc., 237–239.

  • Bom, N., 1969: Effect of rain on underwater noise level. J. Acoust. Soc. Amer.,45, 150–156.

  • Burgess, A. S., and D. J. Kewley, 1983: Wind-generated surface noise source levels in deep water east of Australia. J. Acoust. Soc. Amer.,73, 201–210.

  • Carbone, R. E., and L. D. Nelson, 1978: The evolution of raindrop spectra in warm-based convective storms as observed and numerically modeled. J. Atmos. Sci.,35, 2302–2314.

  • Churchill, D. D., and R. A. Houze Jr., 1984: Development and structure of winter monsoon cloud clusters on 10 December 1978. J. Atmos. Sci.,41, 933–960.

  • Collier, C. G., S. Lovejoy, and G. L. Austin, 1980: Analysis of bright bands from 3-D radar data. Preprints, 19th Conf. on Radar Meteorology, Miami Beach, FL, Amer. Meteor. Soc., 44–47.

  • Czerwinski, N., and W. Pfisterer, 1972: Typen von regentropfenspektren von polaren bis zu tropischen zonen und ihre abhangigkeit von elementen des regens (in German). Meteor. Rundsch.,25, 88–94.

  • Diem, M., and R. Strantz, 1971: Typen der regentropfenspektren II: Abhangigkeit von der regenintensitat (in German). Meteor. Rundsch.,24, 23–26.

  • Dingle, A. N., and K. R. Hardy, 1962: The description of rain by means of sequential raindrop-size distributions. Quart. J. Roy. Meteor. Soc.,88, 301–314.

  • Evans, D. L., and D. R. Watts, 1981: Wind speed and stress at the sea surface from ambient noise measurements. Int. Symp. on Acoustics, Part III, Calgary, AB, Canada, University of Calgary, 69–78.

  • Everest, F. A., R. W. Young, and M. W. Johnson, 1948: Acoustical characteristics of noise produced by snapping shrimp. J. Acoust. Soc. Amer.,20, 137–142.

  • Farmer, D. M., and D. D. Lemon, 1984: The influence of bubbles on ambient noise in the ocean at high wind speeds. J. Phys. Oceanogr.,14, 1762–1778.

  • ——, and S. Vagle, 1988a: On the determination of breaking surface wave distributions using ambient sound. J. Geophys. Res.,93, 3591–3600.

  • ——, and ——, 1988b: Observations of high frequency ambient sound generated by wind. Sea Surface Sound, B. R. Kerman, Ed., Kluwer Academic, 403–415.

  • Franz, G. J., 1959: Splashes as sources of sound in liquids. J. Acoust. Soc. Amer.,31, 1080–1096.

  • Fujiwara, M., 1965: Raindrop-size distribution from individual storms. J. Atmos. Sci.,22, 585–591.

  • Heindsmann, T. E., R. H. Smith, and A. D. Arneson, 1955: Effect of rain upon underwater noise levels. J. Acoust. Soc. Amer.,27, 378–379.

  • Houghton, H. G., 1968: On precipitation mechanisms and their artificial modification. J. Appl. Meteor.,7, 851–859.

  • Houze, R. A., Jr., 1973: A climatological study of vertical transports by cumulus-scale convection. J. Atmos. Sci.,30, 1112–1123.

  • ——, 1993: Cloud Dynamics. Academic Press, 573 pp.

  • Johnson, M. W., F. A. Everest, and R. W. Young, 1947: The role of snapping shrimp (Crangon and Synalpheus) in the production of underwater noise in the sea. Biol. Bull.,93, 122–138.

  • Jorgensen, D. P., and P. T. Willis, 1982: A Z–R relationship for hurricanes. J. Appl. Meteor.,21, 356–366.

  • Joss, J., and A.Waldvogel, 1969: Raindrop size distribution and sampling size errors. J. Atmos. Sci.,26, 566–569.

  • Knudsen, V. O., R. S. Alford, and J. W. Emling, 1948: Underwater ambient noise. J. Mar. Res.,7, 410–429.

  • Lamp, R., 1958: Das tropfenspektrum in niederschlagen und die radar-reflektivitat nach fremden und eigenen messungen (in German). Beitr. Phys. Atmos.,30, 223–245.

  • Laville, F., G. D. Abbott, and M. J. Miller, 1991: Underwater sound generation by rainfall. J. Acoust. Soc. Amer.,89, 715–721.

  • Laws, J. O., and D. A. Parsons, 1943: The relation of raindrop size to intensity. Trans. Amer. Geophys. Union,24, 452–460.

  • Lemon, D. D., D. M. Farmer, and D. R. Watts, 1984: Acoustic measurements of wind speed and precipitation over the continental shelf. J. Geophys. Res.,89, 3462–3472.

  • Lokken, J. E., and N. Bom, 1972: Changes in raindrop size inferred from underwater noise. J. Appl. Meteor.,11, 553–554.

  • Maddox, R. A., 1980: Mesoscale convective complexes. Bull. Amer. Meteor. Soc.,61, 1374–1387.

  • Marshall, J. S., and W. M. Palmer, 1948: The distribution of raindrops with size. J. Meteor.,5, 165–166.

  • McGlothin, C. C., Jr., 1991: Ambient sound in the ocean induced by heavy precipitation and the subsequent predictability of rainfall rate. M.S. thesis, Dept. of Physics, Naval Postgraduate School, 62 pp. [Available from U.S. Naval Postgraduate School, 1 University Circle, Monterey, CA 93947.].

  • Medwin, H., A. Kurgan, and J. A. Nystuen, 1990: Impact and bubble sound from raindrops at normal and oblique incidence. J. Acoust. Soc. Amer.,88, 413–418.

  • ——, J. A. Nystuen, P. A. Jacobus, D. E. Snyder, and L. H. Ostwald, 1992: The anatomy of underwater rain noise. J. Acoust. Soc. Amer.,92, 1613–1623.

  • Nystuen, J. A., 1985: Underwater ambient noise measurements of rainfall. Ph.D dissertation, University of California, San Diego, 86 pp. [Available from Scripps Institution of Oceanography—Archives, University of California, San Diego, 8602 La Jolla Shores Drive, La Jolla, CA 92093.].

  • ——, 1986: Rainfall measurements using underwater ambient noise. J. Acoust. Soc. Amer.,79, 972–982.

  • ——, 1993: An explanation of the sound generated by light rain in the presence of wind. Natural Physical Sources of Underwater Sound, Kluwer Academic, 659–668.

  • ——, 1996: Acoustical rainfall analysis: Rainfall drop size distribution using the underwater sound field. J. Atmos. Oceanic Technol.,13, 74–84.

  • ——, and D. M. Farmer, 1987: The influence of wind on the underwater sound generated by light rain. J. Acoust. Soc. Amer.,82, 270–274.

  • ——, and ——, 1989: Precipitation in the Canadian Atlantic Storms Program: Measurements of the acoustic signature. Atmos.–Ocean,27, 237–257.

  • ——, L. H. Ostwald, and H. Medwin, 1992: The hydroacoustics of a raindrop impact. J. Acoust. Soc. Amer.,92, 1017–1021.

  • ——, C. C. McGlothin, and M. S. Cook, 1993: The underwater sound generated by heavy rainfall. J. Acoust. Soc. Amer.,93,3169–3177.

  • ——, J. R. Proni, C. A. Lauter Jr., J. Bufkin, U. Rivero, M. Boland, and J. C. Wilkerson, 1994: APL disdrometer evaluation. NOAA Tech. Memo. ERL AOML-83, 43 pp. [Available from ERL/NOAA/AOML Library, 4301 Rickenbacker Causeway, Miami, FL 33149.].

  • ——, ——, P. G. Black, and J. C. Wilkerson, 1996: A comparison of automatic rain gauges. J. Atmos. Oceanic Technol.,13, 62–73.

  • Prosperetti, A., L. A. Crum, and H. C. Pumphrey, 1989: The underwater noise of rain. J. Geophys. Res.,94, 3255–3259.

  • Pumphrey, H. C., and L. A. Crum, 1990: Free oscillations of near-surface bubbles as a source of the underwater noise of rain. J. Acoust. Soc. Amer.,87, 142–153.

  • ——, ——, and L. Bjorno, 1989: Underwater sound produced by individual drop impacts and rainfall. J. Acoust. Soc. Amer.,85, 1518–1526.

  • Rosenfeld, D., E. Amitai, and D. B. Wolff, 1995: Classification of rain regimes by the three-dimensional properties of reflectivity fields. J. Appl. Meteor.,34, 198–211.

  • Rowland, J. R., 1976: A comparison of two different raindrop disdrometers. Preprints, 17th Conf. on Radar Meteorology, Seattle, WA, Amer. Meteor. Soc., 398–405.

  • Scrimger, J. A., 1985: Underwater noise caused by precipitation. Nature,318, 647–649.

  • ——, D. J. Evans, G. A. McBean, D. M. Farmer, and B. R. Kerman, 1987: Underwater noise due to rain, hail, and snow. J. Acoust. Soc. Amer.,81, 79–86.

  • ——, ——, and W. Yee, 1989: Underwater noise due to rain: Open ocean measurements. J. Acoust. Soc. Amer.,85, 726–731.

  • Steiner, M., and R. A. Houze Jr., 1993: Three-dimensional validation at TRMM ground truth sites: Some early results from Darwin. Preprints, 26th Int. Conf. on Radar Meteorology, Norman, OK, Amer. Meteor. Soc., 417–420.

  • ——, ——, and S. E. Yuter, 1995: Climatological characterization of three-dimensional storm structure from operational radar and rain gauge data. J. Appl. Meteor.,34, 1978–2007.

  • Takahashi, Y., 1935: Ueber regen und regentropfen (ertse Mitteilung) (in German). J. Meteor. Soc. Japan,13, 67–73.

  • Tokay, A., and D. A. Short, 1996: Evidence from tropical raindrop spectra of the origin of rain from stratiform and convective clouds. J. Appl. Meteor.,35, 355–371.

  • Urick, R. J., 1951: Some directional properties of deep-water ambient noise. Naval Research Laboratory Rep. 3796, 58 pp. [Available from U.S. Naval Postgraduate School, 1 University Circle, Monterey, CA 93947.].

  • ——, 1975: Principles of Underwater Sound. McGraw Hill, 423 pp.

  • Vagle, S., W. G. Large, and D. M. Farmer, 1990: An evaluation of the WOTAN technique of inferring oceanic winds from underwater ambient sound. J. Atmos. Oceanic Technol.,7, 576–595.

  • Waldvogel, A., 1974: The No jump of raindrop spectra. J. Atmos. Sci.,31, 1067–1078.

  • Wenz, G. M., 1962: Acoustic ambient noise in the ocean: Spectra and sources. J. Acoust. Soc. Amer.,34, 1936–1956.

  • Zipser, E. J., 1982: Use of a conceptual model of the life cycle of mesoscaleconvective systems to improve very short range forecasts. Nowcasting, K. Browning, Ed., Academic Press, 191–204.

APPENDIX A

The Ratioed Rain Sound Spectrum

The rain sound spectrum Sr(f, t), the ratioed sound spectrum SR(f, t), and the reference or background spectrum 〈Sb(f, t)〉Δt are all introduced in section 1b of the text. Here, Sr(f, t) is the rain sound spectrum measured at a given underwater site at a given time and, as such, includes not only rainfall-generated sound, but sound from all of the other sound sources present at the time of measurement [as listed in Eq. (1)]. In an attempt to define an acoustic discriminant DR with values for convective- and stratiform-type rainfall that are site independent, the reference or background noise is subtracted (in decibels) from the rain sound spectrum Sr(f, t) to produce the ratioed rain sound spectrum SR(f, t); for example,
SRf, tSrf, tSbf, tΔt

This is the “compensation” referred to in section 3a(6) of the text. Plots of Sr, SR, and Sb for the brackish water pond are given in Fig. A1. For all but the lightest of rainfall the power contributed by nonrainfall sound sources is very small compared to the power contributed by rain. This may be seen by examination of sound levels in Fig. A1. For example, at 10 kHz the antilog of Sr(f, t) has a value of about 108 and the antilog of Sb(f, t) has a value of 103; thus the contribution by nonrainfall sound sources in the case of the pond (as estimated by the use of Sb) is only 10−5 of the rainfall sound.

Despite this small contribution to the overall sound level, the background sound spectrum is not white and does change the discriminant Dr in going from site to site. We have observed, however, that when using SR(f, t) to calculate the discriminant DR, rather than Dr, a value of zero (or less) during convective rainfall and greater than zero during stratiform rainfall is reasonably consistently observed. In the open tropical ocean it is likely that Sb will be dominated by wind. Wind spectra have been measured extensively and good predictions for their shape are available, so that, although individual site background measurements are needed for the calculation of SR, good estimates for Sb can be made on the basis of knowledge of winds.

The effect of rapid wind changes occurring during rainfall events is to cause a reduction in the accuracy of Sb as an estimate for the background during a rainfall event.

The “ratioed” rain sound spectrum is so titled because it is the spectrum of the ratio of the acoustic power measured during rainfall in the presence of background noise to the background noise measured before the onset of rain. The background noise spectrum measured prior to the onset of rainfall is a reference spectrum. As discussed in the text, the discriminant DR has a value of approximately zero (or less) during convective rainfall. UsingSR to provide a discriminant DR having a value of zero during convective rainfall is equivalent to normalizing band-averaged power ratios. This may easily be seen as follows:
DR
or
S10–30RS4–10R
or
S10–30rS10–30bS4–10rS4–10b
by definition
i1520-0493-125-9-2014-eqa5
Therefore,
i1520-0493-125-9-2014-eqa6
Thus, the reciprocal of the ratio of the 10–30-kHz band-averaged reference spectrum power to the 4–10-kHz band-averaged reference spectrum power serves as a normalization factor for the ratio of the band-averaged power during rainfall in the presence of background noise.

APPENDIX B

Marine Acoustic Rainfall Measurement System (MARS) and Data Sampling Strategies

System integration

Work began in 1988 to assemble an integrated multisensor rainfall measurement system. For the Carysfort Reef deployment in October 1988, three separate datalogging systems were used to record observations: 1) a Sea Data model 661 WOTAN underwater recorder that digitized and stored the acoustic sound spectra at 12 frequencies internally; 2) a Sea Data model 660 high-capacity recorder for logging anemometer winds and rain rates from an optical rain gauge at 5-min intervals; and 3) a datalogging system for recording the hourly average of wind speed and direction, air temperature, water temperature, and tipping-bucket rainfall accumulations.

This system was upgraded in 1992 to a fully operational Marine Acoustical Rainfall System (MARS) as shown in the block diagram of Fig. B1. MARS was developed for the brackish water pond facility in 1992 but was used later that year during the deployment at the Duck Pier site. The system is operated using a 50-MHz 486 personal computer.

Hydrophones

Two acoustical sensor systems are integrated in MARS as shown in the block diagram. They consist of an International Transducer Corporation (ITC) hydrophone and a WOTAN hydrophone system. The WOTAN hydrophone output can be fed to both the MARS central computer as well as to the WOTAN recorder. An important feature of rainfall-produced sound is its wide dynamic range. Acoustical data span a dynamic range on the order of 80 dB, depending on the rainfall rates encountered. The prototype system employed three separate channels to record the acoustic signal levels at three gain settings to assure that the linear receivers were not saturated.

Of some interest is the surface area of sensitivity or surface listening areas of the hydrophones used at the three measurement sites. Discussions of the calculation of this area have been presented by Urick (1951) and by Farmer and Vagle (1988b). Calculations for these deployment sites were based on a “rule of thumb” for estimating shallow-water listening areas; for example, the radius of the listening area is three times the depth of the hydrophone.

For the Duck Pier deployment, the ITC and WOTAN hydrophone outputs to MARS were sampled in a burst mode. A 100-kHz sampling rate was used to record 8192 points per burst for each of the three separate gain settings. This0.08192-s sample was divided into 32 bins of 256 points. The FFT was computed for each bin, and the 32 bins were Bartlett-averaged yielding 128 spectral points from 0 to 50 kHz with a bandwidth of 0.390 kHz. This sample was repeated every 3/4 s, and 12 such spectra were boxcar averaged, yielding a final spectrum every 9 s with 768 degrees of freedom. The internally digitized WOTAN hydrophone output was passed through a group of 12 constant-Q analog filters (−24 dB per octave) centered at 0.5, 1, 2, 4, 8, 10, 12.5, 15, 17.5, 20, 25, and 30 kHz. Six values sampled at 5-s intervals were averaged and digitized every 30 s.

At the brackish pond site, the hydrophone sampling strategy was changed slightly to reduce the data volume. The same 100-kHz sampling rate was used but only for 4096 points and for just the lowest gain setting. This 0.04096-s sample burst was repeated at 1-s intervals. For each burst, an FFT was computed yielding 2048 spectral values and a 0.0244-kHz bandwidth. Averaging was accomplished by boxcar averaging segments of 16 values each yielding a 128-point spectrum with the same 0.390-kHz bandwidth as in the system configuration for the Duck deployment. Prefiltering and tapering yielded 43 degrees of freedom (see Nystuen et al. 1996). Ten 1-s spectra were then averaged to produce one spectra every 10 s with 430 degrees of freedom, about half that for the Duck data. Frequencies were also shifted by one-half bandwidth compared to the configuration at the Duck site.

GPS timing

A GPS receiver was used to provide a timing signal, accurate to better than 1 s, so that the outputs from each of the sensors could be closely compared. In this regard, each instrument or sensor has a characteristic recording time; for example, optical rain gauges have an exponential response effectively producing one data point every 9 s, so that even though the various sensors are sampled at a rate of 1 Hz, temporal resolution is ultimately sensor limited.

Rain gauges

The MARS system for the brackish pond and Duck Pier sites consisted of five rain gauges: Scientific Technology Incorporated models 100 and 700 optical gauges; R.M. Young model 50202 capacitance gauge; a weighing gauge of NASA design from the Calibration Laboratory, Goddard Space Flight Facility; and a Belfort model 382B tipping-bucket gauge. The precision of the optical gauges is 0.05 mm h−1 with a low end noise threshold of 0.6 for the model 700 and and 0.8 mm h−1 for the model 100. At the brackish pond site the model 700 optical gauge was sampled once every 10 s until rain greater than 1 mm h−1 was detected. However, during the deployment at Duck, sampling every 10 s was done until rain greater than 10 mm h−1 was detected. With the onset of rain (as detected by the model 700 optical gauge), all gauges including an anemometer were then sampled once per second. For the Duck Pier deployment, these values were averaged over nine seconds and at the pond site 10.

During the Duck deployment, a 12-bit A/D board was used to sample gauges. This resulted in a rain-rate precision for the weighing gauge of 14.4 mm h−1 for the 9- and 10-s averages, and 2.4 mm h−1 for the 1-min averages. The precision for the capacitance gauge rain rate for the two averaging times was 4.3 and 0.7 mm h−1. Later, for the pond site deployment, a 16-bit board was used that resulted in improved precisions for the weighing and capacitance gauges of 1.84 and 0.5 mm h−1.

The tipping bucket has an inherent quantification error due to the finite quantity of water delivered in each tip and to the act of tipping when a certain amount of raindrops are lost. This error ranges from 2% at 50 mm h−1 to 12% at 200 mm h−1. For a full discussion of the MARS system seeNystuen et al. (1996).

The model 700 optical rain gauge was selected as the instrument for continuous monitoring during the Duck deployment because it had the highest recording precision of all gauges and proved to have an easily correctable linear bias with rain rate. However, the capacitance and weighing gauge proved to be the most accurate. Therefore, their data were regressed on the model 700 gauge data to derive a “corrected” set of model 700 data. Six 10-s averages of this “corrected” dataset were summed to derive a 1-min average for use as the measured rain rate.

Disdrometer

Two types of impact disdrometers were used in the MARS system. They were a Distromet RD-69 disdrometer (Joss and Waldvogel 1969) and an instrument designed by the Applied Physics Laboratory (see Rowland 1976; Nystuen et al. 1994). During the Duck Pier deployment, neither were functional. Both instruments have sensor surface areas of about 50 cm2 and measure drops in the size range 0.3–5.5-mm diameter. The Distromet unit outputs drop-size distributions once per minute. The APL disdrometer output rate is adjustable. Both disdrometer units have separate computers that receive timing signals from the GPS.

Radar

Observations from the National Weather Service WSR-57 and the NEXRAD WSR-88D radars were an additional component of MARS. During the Carysfort and Duck deployments, reflectivity data were recorded by WSR-57 radars at Coral Gables and Cape Hatteras stations, respectively. Each radar was located approximately 50 km from the measurement sites. The WSR-57 10-cm radars have 2° horizontal and vertical beamwidths. The Miami WSR-88D radar (0.95° horizontal and vertical beamwidths) was located 29 km from the brackish water pond site.

To minimize direct supervision of the radar recording equipment during the Carysfort deployment and to limit interference with NWS personnel, reflectivity data from the Miami radar were recorded for only a portion of the full radar scan. An approximately 400-km2 sector of data centered on the Carysfort Reef monitoring site was recorded. The 1-km range-gated data for the 20 azimuths scanned over the target area were averaged and binned to 1-km rectangular coordinates. This sampling method allowed for several days of data to be recorded on a single tape. Unfortunately, upon analysis of the collected data it was found that there were many time intervals when reflectivity data were not recorded due to the frequent updating of header information on the tape. In contrast to the Carysfort radar data sampling technique, the full 360° scans from the Cape Hatteras WSR-57 and Miami NEXRAD radars were recorded for the Duck and brackish water pond deployments, respectively.

Fig. 1.
Fig. 1.

Major underwater sound-generating mechanisms of raindrops as a function of drop diameter for the drop size range 0.3–5.0 mm (after Medwin et al. 1992).

Citation: Monthly Weather Review 125, 9; 10.1175/1520-0493(1997)125<2014:ORDACI>2.0.CO;2

Fig. 2.
Fig. 2.

A composite of sound spectra for the major components of ambient sound in the ocean showing the spectrum shape and level for each source in the frequency range 0.1–50 kHz. For ship traffic and wind components, a range of spectral levels is shown (after Urick 1975). The spectrum for snapping shrimp is a representative worst-case example of biological noise. Rain spectra are measurements recorded by the authors during MCS rainfall on 2 October 1993 in Miami, Florida. The 4–30-kHz frequency band of interest in the analysis of rainfall sound spectra has been highlighted for reference.

Citation: Monthly Weather Review 125, 9; 10.1175/1520-0493(1997)125<2014:ORDACI>2.0.CO;2

Fig. 3.
Fig. 3.

Sample rain sound spectra Sr(f, t) measured by other investigators. Those by Bom (1969), Scrimger (1985), and Nystuen (1985) were recorded in lakes. Measurements by Scrimger et al. (1989) were made in the coastal ocean west of Vancouver Island, British Columbia, Canada. Observations by McGlothin (1991) and Nystuen et al. (1993) were taken in the Gulf of Mexico. The Heindsmann et al. (1955) measurement was made in the coastal waters of Long Island Sound, New York.

Citation: Monthly Weather Review 125, 9; 10.1175/1520-0493(1997)125<2014:ORDACI>2.0.CO;2

Fig. 4.
Fig. 4.

Schematic vertical cross section of an MCS depicting phases of the system and showing the convective, transition, and stratiform regions (adapted from Biggerstaff and Houze 1991). The early-convective region was added by the authors.

Citation: Monthly Weather Review 125, 9; 10.1175/1520-0493(1997)125<2014:ORDACI>2.0.CO;2

Fig. 5.
Fig. 5.

Measurement configurations at the monitoring sites: (a) brackish water pond, Virginia Key, Miami, Florida; (b) Carysfort Reef, Key Largo, Florida; (c) the Army Corps of Engineer’s Research Pier, Duck, North Carolina. Instrument symbols are defined in appendix B, Fig. B1.

Citation: Monthly Weather Review 125, 9; 10.1175/1520-0493(1997)125<2014:ORDACI>2.0.CO;2

Fig. 6.
Fig. 6.

(a) Ambient sound levels at the three acoustically different monitoring sites used in this study, measured during the prevailing wind speeds indicated. The coastal sites at Carysfort Reef, Florida, and Duck, North Carolina, where biological activity and surf were major sources of ambient noise, sound levels were as much as 25 dB higher than at the pond where shielding from wind and the absence of marine life resulted in very low levels of background noise. Consequently, the threshold of detection of rain was 1.0 mm h−1 at the coastal sites and an order of magnitude lower at the pond. (b) Ambient sound levels for the coastal sites at Carysfort and Duck during wind speeds of 8 m s−1, and at the pond for a wind speed of 1.5 m s−1, show the wind-noise effect on the ambient background sound level below 10 kHz (see also Fig. 2).

Citation: Monthly Weather Review 125, 9; 10.1175/1520-0493(1997)125<2014:ORDACI>2.0.CO;2

Fig. 7.
Fig. 7.

(a) Radar reflectivity from the National Weather Service NEXRAD WSR-88D radar at Miami for a 100 km × 100 km area at 1313 UTC 2 October 1993 when a convective cell,associated with a broad MCS to the southeast, approached the brackish pond site. Time-lapse imagery indicates a cyclonic circulation in the stratiform region centered south of the brackish water pond (denoted by a plus symbol). Curved lines and arrows indicate radar-derived streamlines of the midlevel flow. Low-level motion is also shown. (b) As in (a) except for 1342 UTC when the brackish pond was within stratiform precipitation. Additionally, open arrows indicate direction of upper-level flow. The dBZ scale for (a) and (b) is shown in the lower-right corner of (b). Ground return around the radar site has been removed.

Citation: Monthly Weather Review 125, 9; 10.1175/1520-0493(1997)125<2014:ORDACI>2.0.CO;2

Fig. 8.
Fig. 8.

(a) Base-scan horizontal radar reflectivity fields (260 km × 260 km) at 1318 and 1429 UTC 2 October 1993 for the rain event at the brackish water pond. The area centered on the monitoring site has been enlarged to reveal details of the reflectivity measurements at 1 km × 1 km resolution. (b) Results of the convective–stratiform separation analysis for the reflectivity fields in (a) at 2 km × 2 km grid resolution. Enlargements of the area surrounding the pond site (outlined in white) are provided for reference.

Citation: Monthly Weather Review 125, 9; 10.1175/1520-0493(1997)125<2014:ORDACI>2.0.CO;2

Fig. 9.
Fig. 9.

A presentation of the brackish pond monitoring site on Virginia Key, Miami. For reference, 1 km × 1 km and 2 km × 2 km squares, representative of the grid resolutions of the radar reflectivity and rain-type classification analysis, have been placed over the pond site. The area of the pond is approximately 0.01 km × 0.01 km.

Citation: Monthly Weather Review 125, 9; 10.1175/1520-0493(1997)125<2014:ORDACI>2.0.CO;2

Fig. 10.
Fig. 10.

Time histories of the radar reflectivities and theoretical reflectivities (derived from disdrometer data) for the rain event of 2 October 1993. Rain-rate observations from both disdrometer and weighing rain gauge are also shown. Radar-based rain-type classifications are indicated at 6-min intervals (C—convective, S—stratiform).

Citation: Monthly Weather Review 125, 9; 10.1175/1520-0493(1997)125<2014:ORDACI>2.0.CO;2

Fig. 11.
Fig. 11.

(a) Contour plot of the time evolution of drop size distributions from 1300 to 1500 UTC 2 October 1993; (b) corresponding time evolution of the ratioed rain spectra SR(f). Vertical lines indicate the boundaries for the early-convective (1309–1312 UTC), convective (1312–1330 UTC), and stratiform (1336–1450 UTC) phases of the event. The horizontal line in (a) indicating D > 2.2 mm and in (b) indicating 10 kHz are provided for reference. Radar-based rain-type classifications are indicated as in Fig. 10.

Citation: Monthly Weather Review 125, 9; 10.1175/1520-0493(1997)125<2014:ORDACI>2.0.CO;2

Fig. 12.
Fig. 12.

(a) Representative drop size distributions and corresponding rainfall sound spectra, Sr(f, t), for the early-convective phase of the event of 2 October 1993 at the brackish water pond; (b) as in (a) but for the convective phase.

Citation: Monthly Weather Review 125, 9; 10.1175/1520-0493(1997)125<2014:ORDACI>2.0.CO;2

Fig. 12.
Fig. 12.

(c), (d) Same as Figs. 12a,b but for the transition and stratiform phases.

Citation: Monthly Weather Review 125, 9; 10.1175/1520-0493(1997)125<2014:ORDACI>2.0.CO;2

Fig. 13.
Fig. 13.

(a) Time series of the band-averaged sound spectrum levels Sr(f) for the 4–10-kHz and 10–30-kHz frequency bands for the rain event of 2 October 1993; (b) as in (a) but for the ratioed rain sound spectrum SR(f). Radar-based rainfall classifications are indicated as in Fig. 10.

Citation: Monthly Weather Review 125, 9; 10.1175/1520-0493(1997)125<2014:ORDACI>2.0.CO;2

Fig. 14.
Fig. 14.

Time series of the discriminant,radar-derived rainfall classifications, and rain rate for the event of 2 October 1993. Rain-type classifications are indicated as in Fig. 10.

Citation: Monthly Weather Review 125, 9; 10.1175/1520-0493(1997)125<2014:ORDACI>2.0.CO;2

Fig. 15.
Fig. 15.

(a) Examples of the horizontal radar reflectivity fields (20 km × 20 km) for the rain event at Carysfort Reef on 3 November 1988 during the convective (0927:14 UTC) and stratiform periods (0954:42 UTC). (b) The convective–stratiform analysis of the reflectivity fields in (a). White squares in the center of each image identify the grid location of the acoustic monitoring site. Images are at a cell resolution of 1 km × 1 km.

Citation: Monthly Weather Review 125, 9; 10.1175/1520-0493(1997)125<2014:ORDACI>2.0.CO;2

Fig. 16.
Fig. 16.

Time series of radar reflectivity and rain rate for the rain event of 3 November 1988. Radar reflectivities were converted to rain rate using the relation Z = 300R1.35 (Jorgensen and Willis 1982). Shaded areas indicate time periods of missing data. Radar-derived rainfall classifications for available time periods are also shown.

Citation: Monthly Weather Review 125, 9; 10.1175/1520-0493(1997)125<2014:ORDACI>2.0.CO;2

Fig. 17.
Fig. 17.

Time evolution of the ratioed rainfall sound spectra over the 0.5–30-kHz frequency band for the rain event of 3 November 1988. Radar-derived rainfall classifications are indicated as in Fig. 16. Vertical lines delineate convective rainfall period.

Citation: Monthly Weather Review 125, 9; 10.1175/1520-0493(1997)125<2014:ORDACI>2.0.CO;2

Fig. 18.
Fig. 18.

Time series of the discriminant and radar-based rain-type classifications for the rain event of 3 November 1988 are shown with band-averaged sound levels for the 4–10- and 10–30-kHz bands of the ratioed sound spectrum SR(f). Rainfall classifications are indicated as in Fig. 16.

Citation: Monthly Weather Review 125, 9; 10.1175/1520-0493(1997)125<2014:ORDACI>2.0.CO;2

Fig. 19.
Fig. 19.

(a) Horizontal radar reflectivity fields (210 km × 210 km) from the Cape Hatteras WSR-57 radar for the rain event at 0912 UTC 5 November 1992, during stratiform rainfall, and at 0942 UTC, during convective rain; (b) same as in (a) but for the convective–stratiform analyses. Areas outlined in white in the enlarged classification images indicate location of the monitoring site. Image resolution is 1 km × 1 km.

Citation: Monthly Weather Review 125, 9; 10.1175/1520-0493(1997)125<2014:ORDACI>2.0.CO;2

Fig. 20.
Fig. 20.

Time series of radar reflectivity, radar-derived rainfall rate, and rain gauge rain rate, for the rain event of 5 November 1992. Radar-based rain-type classifications are indicated at 1-min intervals, except for 0918–0919 UTC where data were unavailable.

Citation: Monthly Weather Review 125, 9; 10.1175/1520-0493(1997)125<2014:ORDACI>2.0.CO;2

Fig. 21.
Fig. 21.

Time evolution of the ratioed rainfall sound spectrum over the 0.5–30-kHz frequency band for the rain event of 5 November 1992. Spectra were processed as in Fig. 17. Radar-based classifications are indicated as in Fig. 20. Vertical lines delineate periods of convective rainfall at 0930–0932 and 0934–0949 UTC. A horizontal line at 10 kHz is shown for reference.

Citation: Monthly Weather Review 125, 9; 10.1175/1520-0493(1997)125<2014:ORDACI>2.0.CO;2

Fig. 22.
Fig. 22.

Time series of the discriminant, rain rate, and wind speed for the rainfall event of 5 November 1992. Time periods of sudden bursts in wind speed are indicated by dashed lines. Radar-based rainfall classifications are presented as in Fig. 20.

Citation: Monthly Weather Review 125, 9; 10.1175/1520-0493(1997)125<2014:ORDACI>2.0.CO;2

Fig. 23.
Fig. 23.

Wind sound spectra measured in the absence of rainfall. A regression fit to the wind spectra ofScrimger et al. (1987) and Vagle et al. (1990) is also shown.

Citation: Monthly Weather Review 125, 9; 10.1175/1520-0493(1997)125<2014:ORDACI>2.0.CO;2

Fig. 24.
Fig. 24.

(a) Time-labeled scatterplot of band-averaged sound pressure levels of the 4–10-kHz band, SR(f, t)4–10, vs disdrometer rain rate for the brackish pond event. (b) As in (a) for reflectivity (disdrometer derived).

Citation: Monthly Weather Review 125, 9; 10.1175/1520-0493(1997)125<2014:ORDACI>2.0.CO;2

i1520-0493-125-9-2014-fa1

Fig. A1. One-minute-averaged rain sound spectra for a rain rate of 84.2 mm h−1 in ratioed [SR(f, t)] and nonratioed [Sr(f, t)] form. Also shown is the reference background spectrum [Sb(f, t)]. Data were recorded at the brackish water pond for the case study of 2 October 1993 for the time period 0917–0918 UTC.

Citation: Monthly Weather Review 125, 9; 10.1175/1520-0493(1997)125<2014:ORDACI>2.0.CO;2

i1520-0493-125-9-2014-fb1

Fig. B1. A schematic diagram of the Marine Acoustic Monitoring System as configured during the brackish water pond deployment, 1993–95.

Citation: Monthly Weather Review 125, 9; 10.1175/1520-0493(1997)125<2014:ORDACI>2.0.CO;2

Save