• Black, P. G., , J. R. Proni, , J. C. Wilkerson, , and C. E. Samsury, 1997: Oceanic rainfall detection and classification in tropical and subtropical mesoscale convective systems using underwater acoustic methods. Mon. Wea. Rev., 125 , 20142024.

    • Search Google Scholar
    • Export Citation
  • Fairall, C. W., , E. F. Bradley, , D. P. Rogers, , J. B. Edson, , and G. S. Young, 1996: Bulk parameterization of air–sea fluxes for Tropical Ocean Global Atmosphere Coupled Ocean–Atmosphere Response Experiment. J. Geophys. Res., 101 , 37473764.

    • Search Google Scholar
    • Export Citation
  • Freitag, H. P., , M. O’Haleck, , G. C. Thomas, , and M. J. McPhaden, 2001: Calibration procedures and instrumental accuracies for ATLAS wind measurements. NOAA Tech. Memo. OAR PMEL-119, NOAA/Pacific Marine Environmental Laboratory, Seattle, WA, 20 pp.

  • Ma, B. B., , and J. A. Nystuen, 2005: Passive acoustic detection and measurement of rainfall at sea. J. Atmos. Oceanic Technol., 22 , 12251248.

    • Search Google Scholar
    • Export Citation
  • McPhaden, M. J., and Coauthors, 1998: The Tropical Ocean-Global Atmosphere (TOGA) observing system: A decade of progress. J. Geophys. Res., 103 , 1416914240.

    • Search Google Scholar
    • Export Citation
  • 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 , 413418.

    • Search Google Scholar
    • Export Citation
  • Medwin, H., , J. A. Nystuen, , P. W. Jacobus, , D. E. Snyder, , and L. H. Ostwald, 1992: The anatomy of underwater rain noise. J. Acoust. Soc. Amer., 92 , 16131623.

    • Search Google Scholar
    • Export Citation
  • Nystuen, J. A., 1986: Rainfall measurements using underwater ambient noise. J. Acoust. Soc. Amer., 79 , 972982.

  • Nystuen, J. A., 2001: Listening to raindrops from underwater: An acoustic disdrometer. J. Atmos. Oceanic Technol., 18 , 16401657.

  • Nystuen, J. A., , and H. Medwin, 1995: Underwater sound produced by rainfall: Secondary splashes of aerosols. J. Acoust. Soc. Amer., 97 , 16061613.

    • Search Google Scholar
    • Export Citation
  • Nystuen, J. A., , and H. D. Selsor, 1997: Weather classification using passive acoustic drifters. J. Atmos. Oceanic Technol., 14 , 656666.

    • Search Google Scholar
    • Export Citation
  • Nystuen, J. A., , and E. Amitai, 2003: High temporal resolution of extreme rainfall rate variability and the acoustic classification of rainfall. J. Geophys. Res., 108 .8378, doi:10.1029/2001JD001481.

    • Search Google Scholar
    • Export Citation
  • Nystuen, J. A., , C. C. McGlothin, , and M. S. Cook, 1993: The underwater sound generated by heavy rainfall. J. Acoust. Soc. Amer., 93 , 31693177.

    • Search Google Scholar
    • Export Citation
  • Pumphrey, H. C., , L. A. Crum, , and L. Bjorno, 1989: Underwater sound produced by individual drop impacts and rainfall. J. Acoust. Soc. Amer., 85 , 15181526.

    • Search Google Scholar
    • Export Citation
  • Serra, Y., , P. A. Hearn, , H. P. Freitag, , and M. J. McPhaden, 2001: ATLAS self-siphoning rain gauge error estimates. J. Atmos. Oceanic Technol., 18 , 19892002.

    • Search Google Scholar
    • Export Citation
  • Vagle, S., , W. G. Large, , and D. M. Farmer, 1990: An evaluation of the WOTAN technique for inferring oceanic wind from underwater sound. J. Atmos. Oceanic Technol., 7 , 576595.

    • Search Google Scholar
    • Export Citation
  • Weller, B., and Coauthors, 1999: A science and implementation plan for EPIC: An eastern Pacific investigation of climate processes in the coupled ocean–atmosphere system. EPIC Program Doc., 105 pp. [Available online at http://www.atmos.washington.edu/gcg/EPIC/EPIC/v/rev.pdf.].

  • View in gallery

    The acoustic rainfall and wind estimates at 10°N, 95°W from May to December 2001.

  • View in gallery

    The acoustic rainfall and wind estimates at 12°N, 95°W from May to December 2001.

  • View in gallery

    The acoustic rainfall and wind estimates at 0°, 165°E from April to November 2000.

  • View in gallery

    The acoustic rainfall and wind estimates at 20°S, 95°W from March to October 2002.

  • View in gallery

    The acoustic wind estimate vs an R. M. Young anemometer at different time-averaging intervals for 12°N, 95°W. The numbers in the legend box represent how many data points are in each scatterplot.

  • View in gallery

    The monthly rainfall accumulation at 10°N, 95°W and 12°N, 95°W. The RMY accumulation is shown for 10°N (dashed line) but not for 12°N due to RMY failure during the deployment.

  • View in gallery

    The accumulations vs the rainfall rates for 10° and 12°N, 95°W.

  • View in gallery

    The rainfall event definition for the event separation time of 20 min.

  • View in gallery

    The actual rain time vs duration for event separation cases of 10, 15, and 20 min. The dashed lines contain at least 50% of the predictions.

  • View in gallery

    The event accumulation vs the actual rain time for event separations of 10, 15, 20, 30, 60, 120, and 240 min. The dashed lines contain at least 50% of the predictions.

  • View in gallery

    Histogram of rainfall duration with a bin width of 3 mm h−1 and an event separation of 20 min.

  • View in gallery

    Histogram of rainfall event accumulation with a bin width of 3 mm and an event separation of 20 min.

  • View in gallery

    Air–sea temperature differences (solid line) and 1-min acoustic rainfall rates (stars).

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 8 8 2
PDF Downloads 7 7 1

Detection of Rainfall Events Using Underwater Passive Aquatic Sensors and Air–Sea Temperature Changes in the Tropical Pacific Ocean

View More View Less
  • 1 Department of Marine Science, Naval Academy of Taiwan, Kaohsiung, Taiwan*
  • | 2 Applied Physics Laboratory, University of Washington, Seattle, Washington
© Get Permissions
Full access

Abstract

Several years of long-term high temporal resolution ocean ambient noise data from the tropical Pacific Ocean are analyzed to detect oceanic rainfall. Ocean ambient noise generated by rainfall and wind are identified through an acoustic discrimination process. Once the spectra are classified, wind speed and rainfall rates are quantified using the empirical algorithms. Rainfall-rate time series have temporal resolutions of 1 min. These data provide a unique opportunity to study the rainfall events and patterns in two different climate regions, the intertropical convergence zone (ITCZ) of the tropical eastern Pacific (10° and 12°N, 95°W) and the equatorial western Pacific (0°, 165°E). At both locations the rain events have a mean rainfall of 15 mm h−1, but the events are longer in the eastern Pacific. After the rain event is defined, the probability that a rain event can be detected using the change in air–sea temperature often associated with the rainfall is investigated. The result shows that the rain event accompanied by the decrease of air temperature is a general feature, but that using the temperature difference to detect the rainfall has a very high false alarm rate, which makes it unsuitable for rainfall detection.

Corresponding author address: Barry Ma, Dept. of Marine Science, Naval Academy of Taiwan, P.O. Box 90175, Kaohsiung 833, Taiwan, China. Email: binbing@mail.cna.edu.tw

Abstract

Several years of long-term high temporal resolution ocean ambient noise data from the tropical Pacific Ocean are analyzed to detect oceanic rainfall. Ocean ambient noise generated by rainfall and wind are identified through an acoustic discrimination process. Once the spectra are classified, wind speed and rainfall rates are quantified using the empirical algorithms. Rainfall-rate time series have temporal resolutions of 1 min. These data provide a unique opportunity to study the rainfall events and patterns in two different climate regions, the intertropical convergence zone (ITCZ) of the tropical eastern Pacific (10° and 12°N, 95°W) and the equatorial western Pacific (0°, 165°E). At both locations the rain events have a mean rainfall of 15 mm h−1, but the events are longer in the eastern Pacific. After the rain event is defined, the probability that a rain event can be detected using the change in air–sea temperature often associated with the rainfall is investigated. The result shows that the rain event accompanied by the decrease of air temperature is a general feature, but that using the temperature difference to detect the rainfall has a very high false alarm rate, which makes it unsuitable for rainfall detection.

Corresponding author address: Barry Ma, Dept. of Marine Science, Naval Academy of Taiwan, P.O. Box 90175, Kaohsiung 833, Taiwan, China. Email: binbing@mail.cna.edu.tw

1. Introduction

The rainfall climatology of the eastern tropical Pacific Ocean is dominated by the intertropical convergence zone (ITCZ). The narrow ITCZ is marked by heavy and persistent rainfall. The positions and intensities of the ITCZ are highly sensitive to the underlying sea surface temperature distribution (Weller et al. 1999). The East Pacific Investigation of Climate Processes (EPIC) project was a 5-yr process study to improve the description and understanding of the Pacific cold tongue–ITCZ complex and the stratus deck region of the southeastern and eastern Pacific Ocean.

The technique of using underwater acoustic signals generated from raindrops to detect and quantify the rainfall amount has been developed through laboratory studies on individual drop splashes (Pumphrey et al. 1989; Medwin et al. 1990; Medwin et al. 1992; Nystuen and Medwin 1995) and field studies (Nystuen 1986; Nystuen et al. 1993; Nystuen and Selsor 1997; Black et al. 1997; Nystuen 2001). The rainfall rate directly converted from the acoustic sound level has the potential for very high temporal resolution as the measurement is an instantaneous acoustic measure (Nystuen and Amitai 2003). During the EPIC 2001 phase of the study, the underwater acoustic sensors—passive aquatic listeners [PALs; also known as acoustic rain gauges (ARGs)] were deployed at 10° and 12°N, 95°W and were in place for several years. Periods of rain and wind are identified through an acoustic discrimination process (Ma and Nystuen 2005). About 10 000 min of rainfall and 70 000 rain-free data points (at 9-min interval) are identified in these two locations. The time series of acoustically converted rain rate have temporal resolutions of 1 min. This presents an opportunity to define a rainfall event based on the temporal separation of the detected rain spectra. A similar rainfall analysis is also applied to the climate region in the western tropical Pacific Ocean (0°, 165°E). The southeastern Pacific (20°S, 85°W) only had 35 mm of rainfall during a 1-yr span of deployment in 2002. Thus, the rainfall analysis is not performed for this region. The ancillary air temperature data and the sea surface temperature on the deployment moorings are used to study the feasibility of rainfall detection using the decrease in air temperature often associated with rainfall.

2. The instrumentation

The PALs were mounted on the Tropical Atmosphere Ocean (TAO) project (McPhaden et al. 1998) moorings since 1998. The PALs and selected ancillary instruments on the moorings are described in the following subsections.

a. Passive aquatic listeners (PALs)

The PALs consist of an ITC-8263 hydrophone, signal preamplifiers, and a recording computer (Tattletale-8). The nominal sensitivity of these instruments is −160 dB relative to 1 V μPa−1 and the equivalent oceanic background noise level of the preamplifier system is about 28 dB relative to 1 μPa2 Hz−1. A data collection sequence consists of four 1024-point time series collected at 100 kHz (10.24 ms each) separated by 5 s if triggered by rain or drizzle. Each time series is fast Fourier transformed (FFT) to obtain a 512-point (0–50 kHz) power spectrum. These four spectra are averaged together and spectrally compressed to 64 frequency bins, with frequency resolutions of 200 Hz from 100 to 3000 Hz and of 1 kHz from 3 to 50 kHz. These spectra are evaluated individually to detect the acoustic signature of rainfall and then they are recorded internally. The temporal sampling strategy is designed to allow the instrument to record data for up to 1 yr and yet detect the relatively short time intervals associated with rainfall. To achieve this goal, the PAL enters a low-power mode (sleep mode) between each data sample.

b. R. M. Young anemometer

An R. M. Young Company (RMY) anemometer is mounted on the TAO mooring at 4-m height. The data values are 10-min-averaged wind vectors. The error estimate is ±0.3 m s−1 or 3% of the wind speed, whichever is greater (Freitag et al. 2001).

c. R. M. Young rain gauge

Precipitation measurements on the TAO buoys are made using R. M. Young model 50203–34 self-siphoning rain gauges mounted 3.5 m above the ocean surface. The instruments have a 100 cm2 (11.3-cm diameter) catchment cylinder mounted on top of a fill tube. The measuring tube has a maximum capacity of 500 mL, which is equivalent to 50 mm of rainfall accumulation, after which it automatically drains via a siphon. Siphon events take about 30 s, and are typically identified by sharp declines in volume for two consecutive samples (Serra et al. 2001).

3. Data description

a. Ocean ambient sound data

The ocean ambient sound data were collected using PALs deployed on the TAO moorings at different locations with depths from 20 to 98 m (Ma and Nystuen 2005). These depths were chosen so that the measurements would be in the mixed layer, reducing the complexities of sound refraction from the surface (sound source) to the sensor, but deep enough to reduce locally generated noise from the surface component of the mooring itself. The signal contains sound from the desired natural quantity (rain), plus sound from wind and noises from other sources. About 90 buoy months of acoustic data are used from two climate regions, the western Pacific warm pool (WPWP) and the intertropical convergence zone (ITCZ). Although each PAL is focused on rainfall detection, the long-term acoustical spectra from the deployment sites are available for wind speed measurement and various noise budget studies. The data have been classified into sound categories of wind, rain, and noise using the discrimination process described in Ma and Nystuen (2005). Table 1 shows the resulting sound budgets for these deployments.

Once the sound source is identified, wind speed and rainfall rate can be calculated. The surface wind speeds conversion uses the algorithm given in Vagle et al. (1990):
i1520-0493-135-10-3599-e1
where W is the wind speed (m s−1) at 10-m height and SPL8kHz is the sound pressure level at 8 kHz. The rainfall-rate conversion uses the algorithm given in Ma and Nystuen (2005):
i1520-0493-135-10-3599-e2
where R is the rain rate (mm h−1) and SPL5kHz is the sound pressure level at 5 kHz. The acoustically measured wind speeds and rainfall rates are shown in Figs. 1 –4 for 10°N, 95°W; 12°N, 95°W; 0°, 165°E; and 20°S, 85°W, respectively. The acoustic data have the sampling intervals of 9 min for wind, 3 min for drizzle, and 1 min for rain. Each panel in Figs. 1 –4 represents 1 month of data, and only the rainy months are shown. The surface anemometers are also shown for the intercomparison. PALs do not give wind speed estimates during rainfall, since the rain-generated sound is loud and masks the signal from wind when it is present. Although it may be possible to estimate wind speed during periods of drizzle or during the extreme rainfall using changes in the sound signal due to bubble cloud effects, no attempt is made here to estimate wind speed during rain, drizzle, or otherwise noisy conditions. The correlation coefficients between the wind measurements from the PALs and the buoy-mounted R. M. Young anemometers are shown in Table 2. Different time-averaging intervals for the wind measurements, and a comparison using all data points and only data when the wind speeds are above 2.2 m s−1, are considered. The correlation for the wind speeds above (2.2 m s−1) is expected to be slightly higher than the one calculated from all data points because low winds do not generate breaking waves, and thus there is no acoustic signal. Table 2 indicates that a 30-min to hourly average of acoustic wind speeds is sufficient to have a comparable result with the surface anemometer data (correlation more than 0.85). Longer time-averaging intervals improve the correlations, but not significantly. For the data shown in Figs. 1 –4, the acoustic wind speeds are smoothed by taking a three-datapoint (27 min) running mean. The correlation coefficients at 12°N are higher than 10°N. The higher amounts of rainfall and larger percentage of local mooring noises can contribute to this result. The scatterplots for time-averaging intervals of 3 h, 6 h, 12 h, and daily at 12°N, 95°W are shown in Fig. 5. The acoustic wind speeds do not have measurements of less than 2.2 m s−1, due to the physical limitation (no signal from breaking waves), which is reflected in the measurement algorithm [Eq. (1)]. (The algorithm has a minimum output value of 2.2 m s−1.) The annual mean wind speed is 5.2 m s−1 and the maximum wind speed is about 14 m s−1 for this year-long acoustic wind record.

For the rainfall, this part of the ocean has a distinctive wet and dry pattern, each associated with the seasonal movement of the ITCZ. This rainy season is from May to roughly the end of November. The clusters of rainfall from the ITCZ region of the eastern Pacific suggest a separation time of about 4–6 days (Figs. 1 and 2), while the clusters of rainfall from 0°, 165°E are separated by about 8–12 days (Fig. 3), which is roughly twice as long as that in the eastern Pacific. At 20°S, 85°W, rainfall rarely occurs; the annual accumulation was only 30 mm. The resolution of the time series of rainfall rates in Figs. 1 –4 is 1 min, with the maximum rainfall rate of 374 mm h−1 recorded at 10°N, 95°W, but for only 1 min. The maximum 10-min rainfall rate from the surface-mounted R. M. Young rain gauges was 83 mm h−1. This difference is expected as the acoustic rainfall measurement is an instantaneous measurement, whereas the surface rain gauge is an accumulation gauge, necessarily smoothing extreme values. The monthly rainfall accumulation is shown in Fig. 6. At 10°N, 95°W, the month of September 2001 had a rainfall accumulation of 590 mm, which accounted for one-third of the annual precipitation for 2001. In addition, accumulation during two extreme rainy days in September 2001 accounted for one-third of the monthly accumulation of that month. This shows the episodic character of the rainfall climatology in this region. At 12°N, 95°W, the rainfalls are more evenly distributed throughout the wet season with monthly accumulations ranging from 150 to 250 mm. Histograms of acoustic rainfall rates versus the accumulations using a bin width of 2 mm h−1 are shown in Fig. 7. The rainfall rates in the 10 mm h−1 bin have the highest contribution to the overall accumulation of rainfall at these locations. Note that Fig. 7 may be biased because the acoustic detection of lighter rainfall rates (<2 mm h−1) is limited.

b. Ancillary data

Ancillary data collected on the mooring are used for intercomparison of the rainfall and wind speed signal from the acoustic records.

1) Wind speed

Wind measurements are made using an R. M. Young anemometer mounted on the TAO moorings. By using the Coupled Ocean–Atmosphere Response Experiment (COARE) version 2.5 flux algorithm (Fairall et al. 1996), these values are converted to equivalent 10-m-height wind speeds. The 10-m winds are slightly stronger than the 4-m winds values by factors of 1.01 ∼ 1.02. This is smaller than the error estimate of the anemometer.

2) Rainfall

Precipitation measurements are made with an R. M. Young rain gauge on TAO buoys. The 1-min volume samples are stored on board the mooring while at sea and are available for postprocessing after recovery. Once the mooring is recovered, the 1-min accumulations are first flagged for obviously erroneous data. A 16-min Hanning filter is then applied to these data to generate smoothed 10-min accumulations. The estimated instrumental error for 10-min derived rainfall rates is 0.4 mm h−1 when there is rain present, and it is 0.1 mm h−1 when there is no rain (Serra et al. 2001).

3) Air temperature

The air temperature measurements are acquired every 10 min from the mooring’s resistance temperature recorder (Pt-100) with a resolution of 0.04°C and an accuracy of ±0.2°C at 3 m above the sea surface.

4) Sea surface temperature

The sea surface temperatures are acquired from a temperature sensor every 10 min with a resolution of 0.001°C and an accuracy of ±0.03°C at 1-m depth.

4. Rainfall event analysis

Ideally, one would like to define a “rain event” so that a “typical” event for a region can be described, and so that the influence of a typical rain event on the underlying ocean can be described. It is apparent from Figs. 1 –3 that rain events occur in clusters separated by several days, weeks, or even months and that within these clusters, individual events occur, separated by minutes, hours, or days. In fact, rainfall is intermittent even within an organized atmospheric system. This means that defining an event is difficult. The acoustic data provide a very high temporal resolution of rainfall detection; thus, the temporal separation of rainfall detection can be used to define what will be called a separate event. Take the temporal separation of 20 min as an example. A “rainfall event” will be defined as a group of rainfall detections with time intervals in between data points that are all smaller than 20 min. There is no “new” event if the next rainfall detection is less than 20 min away. The schematic diagram of rainfall event definition is given in Fig. 8. Given this definition, there are 216 events at 10°N, 95°W, and 173 events at 12°N, 95°W. The numbers of rainfall events with their correspondent separations are shown in Table 3. Three event quantities are defined: the event accumulation (EACC) is the rainfall accumulated during the event, the event duration (EDU) is the time from the beginning to the end of the event, and the actual rain time (ART) is the time that rain is actually falling during the event. The averages of these three quantities are also given in Table 3. As the event separation time increases, the event number decreases, but the EACC, EDU, and ART increase. Figure 9 shows the scatterplot of EDU versus ART for separation cases of 10, 15, and 20 min. The average EDU and average ART are roughly the same when the event separation is smaller than 20 min. However, the average EDU begins to become larger than the average ART when the separation time is greater than 20 min and continues to grow as the separation time increases. When the ratio of EDU to ART becomes large, it is an indication that there is a gap of nonrain in the event, suggesting that two separate events have been combined together. A separation time of 20 min for the EPIC region, or 20–30 min for the western Pacific site, is chosen (ratio of EDU/ART ≈ 1.10), as this separation time subjectively partitions the data into “events.” The EACC and ART have a fixed ratio even when the separation time is different. Figure 10 is a scatterplot of EACC versus ART using the separation times 10, 15, 20, 30, 60, 120, and 240 min at 10°N, 95°W. The regression is given by
i1520-0493-135-10-3599-e3
where a is the intercept, b is the slope, the ART is in minutes, and EACC is in millimeters. The a and b values for three different locations are given in Table 4. This indicates that the EACC is about one-quarter of the ART regardless of the event separation time, and indicates that the mean rainfall rate during the time when it is actually raining is constant, ∼15 mm h−1.

We can now generate statistics for rain events. Consider the case of 20-min separation, when the average duration is about 33 min and the average accumulation is 7 mm in the EPIC region. In contrast, at 0°, 165°E these numbers are about 19 min and 4.6 mm (Table 3). In other words, the average event in the EPIC region is bigger and longer with higher accumulation than in that in the western Pacific. The histograms of events versus durations (bin width of 3 min) and accumulations (bin width of 3 mm) are shown in Figs. 11 and 12. The mode of the distribution is at 11 min, but with important long-duration outliers. The largest events at 10° and 12°N are 80 and 58 mm, respectively. The longest event duration is about 4 h at 10°N and 3 h at 12°N. The mode of the accumulation histogram is at the smallest event accumulation bin (0–3 mm), but one-half of the annual precipitation at 10°N, 95°W and 12°N, 95°W comes from the events with accumulations larger than 15 mm. At 0°, 165°E, large events only account for about one-third of the total rain. This means that the rainfall climatology includes longer rainfall events in the eastern tropical Pacific Ocean. Yet the mean rainfall rate during rain is the same, roughly 15 mm h−1 (Table 4).

5. Detection of rain events using air–sea temperature difference

Because detection and measurement of rainfall at sea are difficult, other methodologies for identifying rain events have been pursued. One of these is to use the air–sea temperature difference change that occurs as the rainfall begins. S. Yuter et al. (2003, personal communication) attempted to detect drizzle by comparing the time series of the actual drizzle area fraction from a C-band radar and the air–sea temperature difference [ΔT = sea surface temperature (SST) − air temperature (AT); more information available online at http://www.usclivar.org/MTG-PanAmerican-0903.html]. Figure 13 shows the acoustic rainfall rates and air–sea temperature differences for a 20-day period at 12°N, 95°W. This figure suggests that the rainfall events are positively associated with ΔT and demonstrates that there is a cooling of the near-surface atmosphere (AT) by precipitation. The year-long record from 12°N, 95°W shows that the mean SST is 28.11°C with a standard deviation of 2.0, and the mean AT is 26.68°C with a standard deviation of 5.9. The correlation of SST and ΔT is 0.03, while for AT and ΔT it is 0.97. This means that the air–sea temperature differences are caused mainly by the fluctuations of AT.

To assess the potential of rainfall detection using the ΔT thresholds, the acoustic rainfall events (20-min separation) are compared with observed ΔT from the moorings, which are interpolated into 20-min intervals. Different ΔT thresholds are tested for the data from 10°N, 95°W and from 12°N, 95°W. The ΔT values from 10 min prior to the rainfall event and 10 min after the rainfall event are examined for two quantities: average ΔT and maximum ΔT. This ensures that a ΔT can be retrieved even in the shortest rainfall event. Three events from 12°N, 95°W with failed SST records are disqualified. The analysis uses the acoustic rainfall data as ground truth. If the ΔT (maximum or average) in a rainfall event is higher than the given threshold, this event will be defined as “detected.” If a ΔT sample is higher than the threshold but without an associated rainfall event, the sample will be defined as a “false alarm.” The results are given in the Table 5. At 12°N, 95°W, the probabilities of detection are 90% (average ΔT) and 93.5% (maximum ΔT), and the false alarm rate is 17.4% for the threshold setting at 0.8°C. At 10°N, 95°W, the probabilities of detection are 90.2% (average ΔT) and 93.5% (maximum ΔT), and the false alarm rate is 13.3% for the threshold setting at 1.2°C. These results are similar to the study of S. Yuter et al. (2003, personal communication), but the thresholds are lower. Yuter et al. reported that the probability of detection and false alarm rate, for the ΔT threshold at 1.8°C, are 90% and 15%, respectively. However, they studied the heavy drizzle in the stratocumulus regions (18°S, 85°W) where the drizzle is prevalent, and they compared every ΔT within the event, not the average or maximum ΔT. Thus, complete agreement between these two results should not be expected.

A low false alarm rate is desired for rainfall detection given the intermittence of naturally occurring rainfall. Comparing the detection analysis of the PALs and the R. M. Young rain gauges from Ma and Nystuen (2005), a false alarm rate smaller than 0.5% is desirable to make the instrument reliable for year-long rainfall detection at sea. As the temperature threshold setting increases, the false alarm rate decreases, but the probability of detection becomes lower as well (Table 5). For a false alarm rate smaller than 1% (threshold of 3.5°C), the probability of detection will be reduced to 20%–30%. This is much too low to measure rainfall properly. Thus, as a solo indicator, the air–sea temperature difference should not be used in rainfall detection, but it is still an important reference for the occurrences of rainfall, especially for the large accumulation events. Note that with a detection threshold of 3.5°C, the probability of detection is 40.7% using maximum ΔT, yet the accumulation in these detections is 68% of the total accumulation.

6. Conclusions

Long-term rainfall data from the eastern and western Pacific Ocean were acquired using passive underwater sound analysis. A typical rainfall event is defined using the high temporal resolution acoustic record. A separation time of 20 min provides the best regression fit between rain duration and actual rain time among the values tried. The largest events at 10° and 12°N had accumulations of 80 and 58 mm, respectively. The longest event duration is about 4 h at 10°N and 3 h at 12°N. About one-half of the annual precipitation at 10°N, 95°W and 12°N, 95°W comes from the events with accumulations larger than 15 mm. At 0°, 165°E, large events only account for about one-third of the total rain. This means that the rainfall climatology includes longer rainfall events in the eastern tropical Pacific Ocean, yet the mean rainfall rate during rain is the same, roughly 15 mm h−1.

There is a cooling of the near-surface atmosphere by precipitation during most of the rain events. By using the rain event defined by the acoustic method, the probability of detection of rainfall from the difference of the air–sea temperature can be calculated. For a ΔT threshold of about 1°C, the detection of rain events is about 90% with a false alarm rate of about 15%. However, this is a relatively high false alarm rate that makes the change in air–sea temperature difference unsuitable as a solo indicator for rainfall.

Acknowledgments

The PALs in this study were deployed and collected by the Climate Research Division of NOAA/PMEL, headed by Dr. Michael J. McPhaden. We also wish to thank the anonymous reviewers whose comprehensive and critical comments were vital to this manuscript. Funding is from NSF Physical Oceanography, ONR Ocean Acoustics, and National Science Council of Taiwan Project 95-2611-M-012-022.

REFERENCES

  • Black, P. G., , J. R. Proni, , J. C. Wilkerson, , and C. E. Samsury, 1997: Oceanic rainfall detection and classification in tropical and subtropical mesoscale convective systems using underwater acoustic methods. Mon. Wea. Rev., 125 , 20142024.

    • Search Google Scholar
    • Export Citation
  • Fairall, C. W., , E. F. Bradley, , D. P. Rogers, , J. B. Edson, , and G. S. Young, 1996: Bulk parameterization of air–sea fluxes for Tropical Ocean Global Atmosphere Coupled Ocean–Atmosphere Response Experiment. J. Geophys. Res., 101 , 37473764.

    • Search Google Scholar
    • Export Citation
  • Freitag, H. P., , M. O’Haleck, , G. C. Thomas, , and M. J. McPhaden, 2001: Calibration procedures and instrumental accuracies for ATLAS wind measurements. NOAA Tech. Memo. OAR PMEL-119, NOAA/Pacific Marine Environmental Laboratory, Seattle, WA, 20 pp.

  • Ma, B. B., , and J. A. Nystuen, 2005: Passive acoustic detection and measurement of rainfall at sea. J. Atmos. Oceanic Technol., 22 , 12251248.

    • Search Google Scholar
    • Export Citation
  • McPhaden, M. J., and Coauthors, 1998: The Tropical Ocean-Global Atmosphere (TOGA) observing system: A decade of progress. J. Geophys. Res., 103 , 1416914240.

    • Search Google Scholar
    • Export Citation
  • 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 , 413418.

    • Search Google Scholar
    • Export Citation
  • Medwin, H., , J. A. Nystuen, , P. W. Jacobus, , D. E. Snyder, , and L. H. Ostwald, 1992: The anatomy of underwater rain noise. J. Acoust. Soc. Amer., 92 , 16131623.

    • Search Google Scholar
    • Export Citation
  • Nystuen, J. A., 1986: Rainfall measurements using underwater ambient noise. J. Acoust. Soc. Amer., 79 , 972982.

  • Nystuen, J. A., 2001: Listening to raindrops from underwater: An acoustic disdrometer. J. Atmos. Oceanic Technol., 18 , 16401657.

  • Nystuen, J. A., , and H. Medwin, 1995: Underwater sound produced by rainfall: Secondary splashes of aerosols. J. Acoust. Soc. Amer., 97 , 16061613.

    • Search Google Scholar
    • Export Citation
  • Nystuen, J. A., , and H. D. Selsor, 1997: Weather classification using passive acoustic drifters. J. Atmos. Oceanic Technol., 14 , 656666.

    • Search Google Scholar
    • Export Citation
  • Nystuen, J. A., , and E. Amitai, 2003: High temporal resolution of extreme rainfall rate variability and the acoustic classification of rainfall. J. Geophys. Res., 108 .8378, doi:10.1029/2001JD001481.

    • Search Google Scholar
    • Export Citation
  • Nystuen, J. A., , C. C. McGlothin, , and M. S. Cook, 1993: The underwater sound generated by heavy rainfall. J. Acoust. Soc. Amer., 93 , 31693177.

    • Search Google Scholar
    • Export Citation
  • Pumphrey, H. C., , L. A. Crum, , and L. Bjorno, 1989: Underwater sound produced by individual drop impacts and rainfall. J. Acoust. Soc. Amer., 85 , 15181526.

    • Search Google Scholar
    • Export Citation
  • Serra, Y., , P. A. Hearn, , H. P. Freitag, , and M. J. McPhaden, 2001: ATLAS self-siphoning rain gauge error estimates. J. Atmos. Oceanic Technol., 18 , 19892002.

    • Search Google Scholar
    • Export Citation
  • Vagle, S., , W. G. Large, , and D. M. Farmer, 1990: An evaluation of the WOTAN technique for inferring oceanic wind from underwater sound. J. Atmos. Oceanic Technol., 7 , 576595.

    • Search Google Scholar
    • Export Citation
  • Weller, B., and Coauthors, 1999: A science and implementation plan for EPIC: An eastern Pacific investigation of climate processes in the coupled ocean–atmosphere system. EPIC Program Doc., 105 pp. [Available online at http://www.atmos.washington.edu/gcg/EPIC/EPIC/v/rev.pdf.].

Fig. 1.
Fig. 1.

The acoustic rainfall and wind estimates at 10°N, 95°W from May to December 2001.

Citation: Monthly Weather Review 135, 10; 10.1175/MWR3487.1

Fig. 2.
Fig. 2.

The acoustic rainfall and wind estimates at 12°N, 95°W from May to December 2001.

Citation: Monthly Weather Review 135, 10; 10.1175/MWR3487.1

Fig. 3.
Fig. 3.

The acoustic rainfall and wind estimates at 0°, 165°E from April to November 2000.

Citation: Monthly Weather Review 135, 10; 10.1175/MWR3487.1

Fig. 4.
Fig. 4.

The acoustic rainfall and wind estimates at 20°S, 95°W from March to October 2002.

Citation: Monthly Weather Review 135, 10; 10.1175/MWR3487.1

Fig. 5.
Fig. 5.

The acoustic wind estimate vs an R. M. Young anemometer at different time-averaging intervals for 12°N, 95°W. The numbers in the legend box represent how many data points are in each scatterplot.

Citation: Monthly Weather Review 135, 10; 10.1175/MWR3487.1

Fig. 6.
Fig. 6.

The monthly rainfall accumulation at 10°N, 95°W and 12°N, 95°W. The RMY accumulation is shown for 10°N (dashed line) but not for 12°N due to RMY failure during the deployment.

Citation: Monthly Weather Review 135, 10; 10.1175/MWR3487.1

Fig. 7.
Fig. 7.

The accumulations vs the rainfall rates for 10° and 12°N, 95°W.

Citation: Monthly Weather Review 135, 10; 10.1175/MWR3487.1

Fig. 8.
Fig. 8.

The rainfall event definition for the event separation time of 20 min.

Citation: Monthly Weather Review 135, 10; 10.1175/MWR3487.1

Fig. 9.
Fig. 9.

The actual rain time vs duration for event separation cases of 10, 15, and 20 min. The dashed lines contain at least 50% of the predictions.

Citation: Monthly Weather Review 135, 10; 10.1175/MWR3487.1

Fig. 10.
Fig. 10.

The event accumulation vs the actual rain time for event separations of 10, 15, 20, 30, 60, 120, and 240 min. The dashed lines contain at least 50% of the predictions.

Citation: Monthly Weather Review 135, 10; 10.1175/MWR3487.1

Fig. 11.
Fig. 11.

Histogram of rainfall duration with a bin width of 3 mm h−1 and an event separation of 20 min.

Citation: Monthly Weather Review 135, 10; 10.1175/MWR3487.1

Fig. 12.
Fig. 12.

Histogram of rainfall event accumulation with a bin width of 3 mm and an event separation of 20 min.

Citation: Monthly Weather Review 135, 10; 10.1175/MWR3487.1

Fig. 13.
Fig. 13.

Air–sea temperature differences (solid line) and 1-min acoustic rainfall rates (stars).

Citation: Monthly Weather Review 135, 10; 10.1175/MWR3487.1

Table 1.

The percentage of time of the overall acoustic signals and noise.

Table 1.
Table 2.

Correlation coefficients of acoustic wind speed estimates and surface anemometer wind speeds.

Table 2.
Table 3.

Rainfall events at 10°N, 95°W; 12°N, 95°W; and 0°, 165°E. The 20-min separation time is selected for further analysis.

Table 3.
Table 4.

The intercepts and slopes for Eq. (3).

Table 4.
Table 5.

The probabilities of rainfall detection and false alarm rates using the air–sea temperature difference.

Table 5.

* A correction has been made to the online version of this article to remove the word China from this affiliation. A printed corrigenda will be available in the next issue.

Save