• Churnside, J. H., , Tatarskii V. V. , , and Wilson J. W. , 1997: Lidar profiles of fish schools. Appl. Opt., 36 , 60116020.

  • Churnside, J. H., , Tatarskii V. V. , , and Wilson J. W. , 1998: Oceanographic lidar attenuation coefficients and signal fluctuations measured from a ship in the Southern California Bight. Appl. Opt., 37 , 31053112.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Feldman, G., , and McClain C. R. , cited 2004: Ocean Color Web, SeaWiFS Reprocessing Level 2. NASA Goddard Space Flight Center. [Available online at http://oceancolor.gsfc.nasa.gov/.].

  • Franz, H., , Gehlhaar U. , , Günther K. P. , , Klein A. , , Luther J. , , Reuter R. , , and Weidemmann H. , 1982: Airborne fluorescence LIDAR monitoring of trace dye patches—A comparison with shipboard measurements. Deep-Sea Res., 29 , 893901.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Gordon, H. R., 1982: Interpretation of airborne oceanic lidar: Effects of multiple scattering. Appl. Opt., 21 , 29963001.

  • Günther, G. C., , Thomas R. W. L. , , and LaRoque P. E. , 1996: Design considerations for achieving high accuracy with the shoals bathymetric LIDAR system. Laser Remote Sensing of Natural Waters: From Theory to Practice, V. Feigels and Y. Kopilevich, Eds., SPIE, 54–71.

    • Crossref
    • Export Citation
  • Gyory, J., , Rowe E. , , Mariano A. J. , , and Ryan E. H. , cited 2005: The Florida current. [Available online at http://oceancurrents.rsmas.miami.edu/atlantic/florida.html.].

  • Hoge, F. E., , and Swift R. N. , 1981: Absolute tracer dye concentration using airborne laser-induced water raman backscatter. Appl. Opt., 20 , 11911202.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hoge, F. E., , and Swift R. N. , 1983: Airborne dual laser excitation and mapping of phytoplankton photopigments in a Gulf Stream warm core ring. Appl. Opt., 22 , 22722281.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Irish, J. L., , and Lillycrop J. W. , 1999: Scanning laser mapping of the coastal zone; the shoals system. ISPRS J. Photogramm. Remote Sens., 54 , 123129.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Klett, J. D., 1981: Stable analytical inversion solution for processing LIDAR returns. Appl. Opt., 20 , 211220.

  • Klett, J. D., 1985: Lidar inversion with variable backscatter/extinction ratios. Appl. Opt., 24 , 16381643.

  • Kullenberg, G., 1971: Vertical diffusion in shallow waters. Tellus, 23 , 129135.

  • Lee, T., , and Williams E. , 1988: Wind-forced transport fluctuations of the Florida current. J. Phys. Oceanogr., 18 , 937946.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Mack, S. A., , Vasholz D. P. , , Larson E. C. , , Scheerer D. J. , , Calman J. , , and Schoeberlein H. C. , 1997: Estimation of diapycnal diffusivity from a dye tracer study in the upper seasonal thermocline. Eos, Trans. Amer. Geophys. Union, 78 .(Fall Meeting Suppl.), Abstract F374.

    • Search Google Scholar
    • Export Citation
  • Morel, A., , and Maritorena S. , 2001: Bio-optical properties of oceanic waters: A reappraisal. J. Geophys. Res., 106 , 71637180.

  • Okubo, A., 1971: Oceanic diffusion diagrams. Deep-Sea Res., 18 , 789802.

  • Phillips, D. M., , and Koerber B. W. , 1984: A theoretical study of an airborne laser technique for determining sea water turbidity. Aust. J. Phys., 37 , 7590.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Squire, J. L., , and Krumboltz H. , 1981: Profiling pelagic fish schools using airborne optical lasers and other remotes sensing techniques. Mar. Technol. Soc. J., 15 , 2731.

    • Search Google Scholar
    • Export Citation
  • Terray, E. A., and Coauthors, 2005: Airborne fluorescence imaging of the ocean mixed layer. Proc. Eighth Working Conf. on Current Measurement Technology, Southampton, UK, IEEE/OES, 76–82.

  • Vasholz, D. P., , and Crawford L. J. , 1985: Dye dispersion in the seasonal thermocline. J. Phys. Oceanogr., 15 , 695711.

  • Yoder, J. A., , Aiken J. , , Swift R. N. , , Hoge F. E. , , and Stegman P. M. , 1993: Spatial variability in near-surface chlorophyll-a fluorescence measured by the Airborne Oceanographic Lidar (AOL). Deep-Sea Res., 40 , 3753.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    Chlorophyll concentration inferred from satellite sea surface color over the Florida Straits (Feldman and McClain 2004). Inset to the figure shows the ship track during the 3 June dye surveys—red, green, and magenta curves indicate the first transect, the zigzag survey, and the final transect, respectively. Bold lines indicate the ship’s location at times corresponding to the lidar overflights.

  • View in gallery

    Injection time series showing the staircase of different depths where the dye was injected. The injection was performed while the ship was headed southward, hence, early times correspond to the northernmost extent of the patch, while later times are at the southern extent.

  • View in gallery

    Raw backscatter for 2500 profiles outside the dye patch plotted on a logarithmic scale showing approximately constant attenuation with depth. A mean regression line corresponding to an attenuation of 0.085 m−1 is shown as a black line.

  • View in gallery

    (a) Power spectra of normalized lidar flat target response (thick solid), raw backscatter (solid), and deconvolved backscatter (dashed) for a single waveform within the dye patch. (b) Same as in (a), but for corresponding fluorescence channel waveform.

  • View in gallery

    (left) Temperature, (middle) salinity, and (right) potential density profiles during the 3 June dye release experiment.

  • View in gallery

    Mean velocity profiles during the 3 June dye release experiment: eastward (solid) and northward (dashed). Dotted lines indicate one standard deviation about the means.

  • View in gallery

    In situ dye concentration [log10(C/ppb)] as measured during the first transect along the major axis of the dye streak showing the three near-surface (centered at about 2.5-m depth) segments of the dye patch, interspersed with deeper (5 m) segments. Two gaps where no dye was found (approximately 25.944° and 25.95°N) correspond to the portions of the streak that were injected at 10 m, which, due to a strong vertical shear, had already been advected westward of the surface patch. Black triangles indicate locations of the downcast portions of the surveys.

  • View in gallery

    Dye concentration observed during the 3 June zigzag survey showing the vertical and cross-streak structure of the dye patch 1–1.5 h after injection. (a)–(d) Four transects consist of the (top) in situ, (middle) green channel, and (bottom) fluorescent channel with plan view of ship track and corresponding lidar overflight on the rhs. Color scale is log10(C/ppb). Note, in situ concentration estimates near the sea surface (uppermost 1–2 m) may be artificially elevated due to the sampling sled breaching the surface.

  • View in gallery

    (left) Raw IR, (middle) green, and (right) fluorescence backscatter lidar signals observed (top) during a pass over the near-surface segment of the dye patch and (bottom) for a series of profiles outside the dye patch. Each panel contains 100 profiles.

  • View in gallery

    Dye concentration [log10(C/ppb)] inferred from airborne lidar: (top) green channel and (bottom) fluorescence channel. Vertical slices are approximately every 2 m starting from the surface down to 12 m.

  • View in gallery

    Dye concentration [log10(C/ppb)] as measured by lidar during an overflight approximately 10 min into the first line transect along the major axis of the dye streak based on the green channel and fluorescent channel inversions, respectively. Data shown correspond to averages of profiles taken within 10 m of the ship-based transect positions. An offset in the longitudinal position has been applied to the lidar positions, corresponding to approximately 0.5 m s−1 to account for the westward advection of the patch between the in situ and lidar data sampling times.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 4 4 1
PDF Downloads 2 2 0

Three-Dimensional Mapping of Fluorescent Dye Using a Scanning, Depth-Resolving Airborne Lidar

View More View Less
  • 1 School for Marine Science and Technology, University of Massachusetts, Dartmouth, New Bedford, Massachusetts
  • 2 Department of Applied Ocean Physics and Engineering, Woods Hole Oceanographic Institution, Woods Hole, Massachusetts
  • 3 Optech Incorporated, Vaughan, Toronto, Ontario, Canada
  • 4 Joint Airborne LIDAR Bathymetry Technical Center of Expertise, U.S. Army Corps of Engineers, Mobile, Alabama
© Get Permissions
Full access

Abstract

Results are presented from a pilot study using a fluorescent dye tracer imaged by airborne lidar in the ocean surface layer on spatial scales of meters to kilometers and temporal scales of minutes to hours. The lidar used here employs a scanning, frequency-doubled Nd:YAG laser to emit an infrared (1064 nm) and green (532 nm) pulse 6 ns in duration at a rate of 1 kHz. The received signal is split to infrared, green, and fluorescent (nominally 580–600 nm) channels, the latter two of which are used to compute absolute dye concentration as a function of depth and horizontal position. Comparison of dye concentrations inferred from the lidar with in situ fluorometry measurements made by ship shows good agreement both qualitatively and quantitatively for absolute dye concentrations ranging from 1 to >10 ppb. Uncertainties associated with horizontal variations in the natural seawater attenuation are approximately 1 ppb. The results demonstrate the ability of airborne lidar to capture high-resolution three-dimensional “snapshots” of the distribution of the tracer as it evolves over very short time and space scales. Such measurements offer a powerful observational tool for studies of transport and mixing on these scales.

Corresponding author address: M. A. Sundermeyer, School for Marine Science and Technology, University of Massachusetts, Dartmouth, 706 Rodney French Blvd., New Bedford, MA 02744-1221. Email: msundermeyer@umassd.edu

Abstract

Results are presented from a pilot study using a fluorescent dye tracer imaged by airborne lidar in the ocean surface layer on spatial scales of meters to kilometers and temporal scales of minutes to hours. The lidar used here employs a scanning, frequency-doubled Nd:YAG laser to emit an infrared (1064 nm) and green (532 nm) pulse 6 ns in duration at a rate of 1 kHz. The received signal is split to infrared, green, and fluorescent (nominally 580–600 nm) channels, the latter two of which are used to compute absolute dye concentration as a function of depth and horizontal position. Comparison of dye concentrations inferred from the lidar with in situ fluorometry measurements made by ship shows good agreement both qualitatively and quantitatively for absolute dye concentrations ranging from 1 to >10 ppb. Uncertainties associated with horizontal variations in the natural seawater attenuation are approximately 1 ppb. The results demonstrate the ability of airborne lidar to capture high-resolution three-dimensional “snapshots” of the distribution of the tracer as it evolves over very short time and space scales. Such measurements offer a powerful observational tool for studies of transport and mixing on these scales.

Corresponding author address: M. A. Sundermeyer, School for Marine Science and Technology, University of Massachusetts, Dartmouth, 706 Rodney French Blvd., New Bedford, MA 02744-1221. Email: msundermeyer@umassd.edu

1. Introduction

a. Overview and background

For decades, dye release experiments in the upper ocean and pycnocline have been used to study the dispersal and transport of passive tracers in natural waters (e.g., Kullenberg 1971; Okubo 1971; Vasholz and Crawford 1985). The naturally integrating properties of Lagrangian tracers in both space and time allow slow but persistent motions to be detected on a variety of time and space scales. One of the difficulties of using Lagrangian tracers, however, is that quantitative estimates of dye concentrations have traditionally required in situ measurements of the dye patch. Yet conventional shipboard measurements are time consuming and often lack the spatial and/or temporal coverage required to understand the underlying physics. This is particularly true for small scales, that is, spatial scales of meters to kilometers and temporal scales of minutes to hours.

Of particular interest here are both vertical and horizontal mixing processes in the ocean surface layer, and along and across its base. These processes are central to understanding and modeling how the mixed layer and the deeper ocean communicate, and are central issues in a number of fundamental questions in oceanography, such as the contribution of atmospheric forcing to mixing and circulation in the ocean interior, the transport of nutrients across the base of the mixed layer, and the observed patchiness of biological organisms in the photic zone. Despite the importance of mixing and lateral dispersion on these scales, however, little is known from direct observations, mainly due to the inability to make observations with the requisite spatiotemporal resolution and coverage.

One approach that offers considerable promise in this regard is the coupling of fluorescent dye releases with the mapping capability of airborne lidar operating at low altitude (on the order of several hundred meters). In combination, these two approaches offer a powerful tool for studying mixing and stirring processes in the upper ocean. Particularly because of the relatively short temporal and spatial scales characteristic of these processes, the ability of an airborne lidar to nonintrusively capture high-resolution three-dimensional “snapshots” of the distribution of the tracer patch as it evolves over time represents a unique observational capability.

Specialized lidar systems have been developed for oceanographic use by a number of organizations, including the National Aeronautics and Space Administration (NASA), the Unites States Navy, and the National Oceanographic and Atmospheric Administration (NOAA). These systems have been shown to be very effective for their intended uses: bathymetric surveys (e.g., Irish and Lillycrop 1999), chlorophyll mapping (e.g., Hoge and Swift 1983; Yoder et al. 1993), and fish stock assessment (e.g., Squire and Krumboltz 1981; Churnside et al. 1997). Despite efforts in these other areas, however, there have been few studies using lidar in conjunction with the fluorescent dye releases to study small-scale mixing in the ocean. In fact, we are aware of only three other studies in the published literature in which airborne lidar was used to obtain quantitative estimates of dye concentration in the upper ocean. These studies differ from the present work, however, in that they reported only depth-integrated concentrations of dye, whereas here we estimate dye concentration as a function of depth.

One of the prior studies, reported by Hoge and Swift (1981), used NASA’s Airborne Oceanographic lidar (AOL). The AOL employs a conically scanning laser, with a multispectral time-gated receiving capability. In that study, the observed ratio of fluorescence to water Raman backscatter was calibrated with similar measurements in the laboratory and used to infer absolute dye concentrations in the field. A major advantage to this approach is that it enabled absolute dye concentrations to be inferred from airborne measurements without the need for in situ water column sampling. Despite the time gating capability of AOL, however, Hoge and Swift (1981) did not report on depth dependence of the dye patch. Still, their study was significant in that they tracked the dye for approximately 2 h after release, with horizontal resolutions of better than 10 m (absolute positioning accuracy of the airborne measurements was not reported) and a minimum detectable single-pulse integrated water column dye concentration of approximately 2 ppb.

A second published study, by Franz et al. (1982), mapped the distribution of Rhodamine B over a period of 2 days using an airborne lidar flown in a helicopter. In that study, the lidar system used was a nonscanning flashlamp-pumped laser operated by the University of Oldenburg, Germany. There the authors relied on simultaneous in situ measurements of dye concentration and attenuation depth to determine dye concentration from the lidar signal. They also used a comparatively high-powered laser (1 J compared to the AOL’s 1 mJ), which presumably enabled them to detect dye deeper in the water column. They reported being able to map dye concentrations ranging from less than 1 ppb up to 30 ppb along individual flight tracks of the aircraft, with an estimated positioning accuracy of about 100 m (based on the Decca radio-location system). However, their detector did not have time-resolving capability, so that like Hoge and Swift (1981), although for different reasons, they also did not resolve depth-dependent dye concentrations.

Finally, a third study was conducted by a group from the Johns Hopkins University Applied Physics Laboratory (Mack et al. 1997). They used a lidar operated by the Naval Air Warfare Command to track the horizontal and vertical spreading of dye streaks injected in the Sargasso Sea. Their lidar measurements did provide depth resolution of their dye patch. However, they ultimately used the towed fluorometer chain to infer the diapycnal diffusivity of the dye. They did not report estimates of the horizontal diffusivity.

b. Scope and outline

Here we revisit the idea of using airborne lidar for surveying dye release experiments in the upper ocean and pycnocline. Advances in navigational systems, improvements in airborne lidar technology, and the advent of new airborne and in situ observational capabilities (e.g., shipboard ADCP) all lend themselves to better observing of the physics of ocean mixing on short time and space scales. The most notable advance of the present study compared with previous experiments of this type (with the possible exception of Mack et al. 1997) is the ability to obtain depth-resolved dye concentrations with a vertical resolution on the order of 1 m and an absolute horizontal position accuracy of 2.5 m. We also present a different method for converting measured intensity in the laser and dye wave bands to absolute dye concentration. This is done using both the backscatter and dye fluorescence. Detailed in situ observations are also used to compare the airborne measurements with more traditional shipboard surveys using towed conductivity–temperature–depth (CTD) and fluorometer observations.

The present paper is organized as follows: in section 2 we describe the field experiments and the methodology used to inject and sample the dye patch, as well as the methods used to invert the measured backscatter and fluorescence intensity to dye concentration. Results are described in section 3, including both the in situ and airborne measurements. Similarities and differences between the in situ and airborne measurements are described in section 4, along with possible sources of errors and uncertainties and considerations for future studies. Section 5 summarizes and concludes.

2. Description of field experiments

Two dye releases were conducted in the near-surface waters approximately 5 km due east of Ft. Lauderdale, Florida, one on 2 June and the other on 3 June 2004 (Fig. 1). The study site was chosen in part because of the favorable optical qualities of the water in this region and in part to coincide with scheduled activities of the U.S. Army Corps of Engineers (USACE) Joint Airborne Lidar Bathymetry Technical Center of Expertise (JALBTCX) group. The dye injections and in situ sampling were carried out using the Florida Atlantic University research vessel, R/V Stephan. Both experiments were performed in approximately 200-m water depth, far enough offshore to be outside the nearshore band of higher chlorophyll and suspended particulates, but not so far as to be subjected to significant currents associated with the Gulf Stream. On each of the 2 days, 5 kg of Rhodamine WT were injected in a single streak and subsequently mapped over a period of 1–2 h using a fluorometer/CTD system towed from the ship, as well as airborne lidar. For reasons described below, we focus here on results from the second of these two releases.

a. Dye injection and in situ sampling

The 3 June dye release was conducted at approximately 0900 local time. A 20% solution of 5 kg of Rhodamine WT was first mixed with isopropyl alcohol to match the density of the dye mixture with that of the near-surface waters at the injection site to within 0.001 kg L−1. The resulting solution was then diluted with surface seawater to obtain a total volume of dye solution of approximately 200 L. The dye mixture was pumped at a rate of approximately 0.2 L s−1 down a garden hose to the injection sled, where it was released through a T-shaped pvc diffuser. The injection sled, which was also used for subsequent sampling of the dye, consisted of the diffuser, an Ocean Sensors OS200 CTD, and two Chelsea fluorometers, one for dye and one for chlorophyll.

The injection was performed at a ship speed of approximately 3.6 kt. The total injection time for the experiment was approximately 15 min. This resulted in an injection streak length of approximately 1600 m, with an average initial dye concentration based on estimates of the size of the patch on the order of 40 ppb. The dye was injected in a stairstep profile, starting at a depth of 2.5 m for approximately 1 min, followed by a deeper segment at 5 m, followed by an even deeper segment at 10 m, then returning to the surface and repeating (Fig. 2). This resulted in a series of surface segments of the patch, interspersed with progressively deeper segments at discrete depths. The pumping rate of the dye was held constant throughout the injection, which allowed quantitative assessment of the depth-resolving capabilities of the airborne lidar using similar dye concentrations at various depths. The surface segments of the patch provided the additional advantage of enabling the aircraft to visually align with the axis of the tracer streak during successive lidar passes.

Ship-based sampling of the dye patch commenced immediately after injection and continued for 1–2 h thereafter. Surveys of the patch consisted of a single line transect along the major axis of the dye streak, followed by a zigzag survey, followed by another line transect (see inset to Fig. 1). Ship speed during the surveys varied between 2 and 4 kt, while the depth of the tow sled was cycled between the surface and 10–20 m with the winch. This resulted in an oblique up–down profile approximately every 1 min, or 50–100 m, depending on the ship speed at the time. The CTD and fluorometers all sampled at a rate of 6 Hz, which, for a winch speed of approximately 0.5 m s−1, resulted in a vertical sample resolution of approximately 10 cm. In addition to the tow sled, the ship was equipped with a 300-kHz RD Instruments Workhorse acoustic Doppler current profiler. The ADCP was set to sample at a rate of 1 Hz, with a vertical bin size of 2 m. Data from the ADCP were logged nearly continuously during the injection and sampling.

b. Measurements from airborne lidar

Overflights of the dye patches were conducted using an airborne lidar. We used a SHOALS-1000T lidar manufactured by Optech Incorporated and operated by JALBTCX (Irish and Lillycrop 1999; Günther et al. 1996). The SHOALS-1000T system uses a frequency-doubled Nd:YAG laser to produce 1064- and 532-nm pulses 6 ns in duration. The laser pulses at a rate of 1 kHz, and scans approximately transverse to the direction of flight at a rate of 10–15 Hz through a conical arc with a constant incidence angle of 20°. In its bathymetric configuration, the system has four receive channels—one at 1064 nm [infrared (IR)], two at 532 nm (green), and one at 650 nm to measure Raman backscatter. The transmit pulses at both 532 and 1064 nm are approximately linearly polarized, while the receivers are essentially unpolarized. The infrared and Raman channels are used to obtain redundant estimates of the location of the sea surface, while the two green channels (one for shallow returns and one for deep) are normally used to detect the bottom. Following each pulse, the received waveforms of all four channels are digitized at a rate of 1 GHz and recorded for postflight processing. Measurements are georeferenced using GPS, using inertial sensors to measure the aircraft’s attitude. At an aircraft altitude of 200 m, horizontal resolution is on the order of 3–4 m over a swath width of 100 m. The range resolution within the water column, which is set by the pulse length, is approximately 1 m, although higher resolutions are possible for mapping bottom bathymetry.

To measure dye fluorescence, we used a slight modification of the SHOALS-1000T bathymetric configuration. Namely, we replaced the receiver optics of the existing Raman channel with a narrowband filter centered around the peak emission wavelength of the dye (i.e., centered at 580 nm with cutoffs at 560 and 600 nm). This receive channel was then routed to the high gain electronics of the shallow green channel. The latter spanned a wider range of depths than the Raman channel, typically down to about 20 m, compared to only 5–8 m in the Raman. The reconfiguration took about 20 min on the ground. In dye sampling mode, we recorded the fluorescence (nominally 580 nm), the infrared, and the deep green channels. The latter used a photomultiplier (PMT) detector and was the most sensitive of the three channels. The fluorescence channel used an avalanche photo detector (APD).

c. Method of inversion for dye concentration

We use the measured profiles of backscatter at 532 nm and fluorescence at 580 nm to estimate dye concentration as a function of depth. The inversion approach is described briefly here. Additional details are given in Terray et al. (2005).

We make the simplifying assumption that the only unknowns are the dye concentration C(z) and the light intensity in the water just below the surface. To invert the measured intensity profiles directly, one approach would be to condition the problem further by estimating C(z) on a coarser grid than the depth resolution of the measurements. Alternatively, in the interest of achieving a fast inversion, we can reduce the problem to just two dimensions by using an exact inversion of our fluorescence model (Klett 1981, 1985). The latter allows us to parameterize the concentration profile in terms of an unknown constant of proportionality P, which is related to the incident intensity just below the surface, and the dye concentration CrC(zr) at some reference depth zr. These parameters can then be determined by a joint least squares fit to the fluorescence intensity profile and to the backscatter-derived concentration profile over a limited region where the signal-to-noise ratio is high.

Before discussing the general inversion, we review the use of the backscatter profile at 532 nm to infer dye concentration. This approach is based on the fact that the dye absorption/fluorescence process removes photons from the incident 532-nm beam. The measured profiles of backscatter outside the patch in dye-free water are approximately exponential and give a mean lidar attenuation coefficient of k = 0.085 ± 0.003 m−1. Meanwhile, the absorption coefficient of Rhodamine WT at the laser wavelength of 532 nm is approximately αD532 = 0.03 m−1 ppb−1, so that a dye concentration of a few ppb will lead to an O(1) contribution to the total (dye plus water) attenuation. The received backscattered intensity I532 from the range gate at depth z can thus be modeled as
i1520-0426-24-6-1050-e1
where IS532 is the attenuation profile associated with the ambient seawater, and the exponential accounts for the additional absorption by the dye. The dye concentration C(z) (ppb) can then be estimated as
i1520-0426-24-6-1050-e2
Note that if the “background” backscatter due to everything but the dye is stationary and homogeneous over the time and space scales of our experiment, then IS532 can be estimated from measurements outside the dye patch. For our experiments, we have checked that this is a good approximation by comparing mean profiles of I−1532dI532/dz taken in dye-free water to the north and south of the dye streaks (the locations were separated by roughly 1 km, and the aircraft flew at a constant altitude throughout the experiment). The profiles are essentially indistinguishable over the upper 30 m of the water column. Thus, (2) provides a local estimate of C(z) based entirely (apart from the specific dye attenuation coefficient αD532) on measured backscatter profiles in the 532-nm “green” channel.
Our model for the received signal in the 580-nm “fluorescence” channel is based on the backscatter models of Gordon (1982) and Phillips and Koerber (1984), extended to include fluorescence and attenuation by the dye. We write it as
i1520-0426-24-6-1050-e3
where αD is the sum of the specific dye absorption coefficients in the 532- and 580-nm bands, and kS(z) (m−1) denotes the sum of the lidar attenuation coefficients in those two bands due to scattering and absorption by everything but the dye. Subsumed in P are the incident intensity of the laser light and the transmissivity of the surface to ingoing and outgoing radiation, as well as a number of system-dependent constants such as the receiver aperture, aircraft altitude, and pulse length. The P also contains a factor (nh + |z|)−2 describing the diminution of the intensity due to geometrical spreading, where n is the index of refraction of seawater and h is the aircraft altitude. However, in this experiment, h ∼ 200 m and zmax ∼ 10 m, and therefore this factor varies by less than 10%. Consequently, while we permit P to vary from profile to profile, we assume here that it is independent of depth.
Following Klett (1981), (3) can be used to relate the concentration at any depth to the concentration at a deeper depth. Differentiating (3), we obtain the nonlinear differential equation
i1520-0426-24-6-1050-e4
where g(z) = I−1580 dI580/dz + kS(z) is given by observation. Equation (4) can be integrated to give the concentration at any depth in terms of the concentration at another “reference depth” zr. The solution can be written as
i1520-0426-24-6-1050-e5
where Cr = C(zr) and I580(z) is the measured intensity profile. As noted by Klett (1981), the integration is stable if z < zr, although in practice it is possible to integrate to depths somewhat deeper than zr. Substituting the solution for C(z) back into (3), we obtain a model for I(z) that depends on the two unknown values Cr and P, and functionally on the observed I580 profile.
Combining the above approaches for both the backscatter and fluorescence channels, we estimate Cr and P by fitting the measured I580 profile and part of the concentration profile C532 estimated from the excess attenuation in the 532-nm channel [Eq. (2)]. Here we use the cost function
i1520-0426-24-6-1050-e6
This function has been shown in simulations to provide a reasonably shaped error surface (Terray et al. 2005). The weight w is adjusted to balance the contributions from the two error terms, and only the part of the C532 profile that is well above the noise floor is used.

In this procedure, both αD and the lidar attenuation coefficient kS, which may be depth-dependent, must be known. From laboratory measurements, we take αD = αD532 + αD580 = 0.035 m−1 ppb−1. Similarly, kS can be written as the sum kS = k532 + k580. We estimate k532 from the measured backscatter profiles at that frequency (Churnside et al. 1998). A representative sample of 2500 profiles outside the dye patch are shown in Fig. 3. Each profile was fit to a single exponential, and the resulting slope estimates were averaged to give k532 = 0.085 ± 0.003. We note in passing that this value is close to lidar attenuation coefficients reported by Churnside et al. (1998) in relatively clear water in the Southern California Bight.

Estimating k580 is more problematic. As discussed by Gordon (1982), Phillips and Koerber (1984), and Churnside et al. (1998), the attenuation coefficient in (3) is phenomenological, and its relation to such properties as the beam attenuation coefficient c or the diffuse attenuation coefficient Kd depends on the amount of multiple scattering. Unfortunately, we do not have the requisite hydro-optical measurements to estimate this reliably. Gordon (1982) has argued that k lies in the range Kd < k < c, with its exact value depending on both the single scattering albedo, defined as the ratio of the scattering to beam attenuation coefficients, and the product cR, where R is the radius of the receiver field of view projected on the surface. For large cR, then kKd, whereas kc as cR → 0. In our case R ≃ 4.5 m, so that for the 532-nm backscatter, cR > kR ∼ 0.45, and we likely are in the transitional regime, although the presumption is that we are closer to Kd. In the absence of a model relating k580 to its value at 532 nm, we have taken k580 = [Kd(580)/Kd(532)]k532, where the ratio of diffuse attenuations was estimated using the bio-optical model of Morel and Maritorena (2001). From our towed fluorometer measurements, we estimate that the chlorophyll concentration in the upper 20 m was roughly 0.1 mg m−3, so that Kd(580)/Kd(532) = 1.78 and k580 = 0.151 m−1. This gives the estimate kSk532 + k580 = 0.24 m−1, which we have used in our inversion.

Note that a 50% error in k580 is a 33% error in kS. We have checked that the inversion is not strongly sensitive to errors in kS of this magnitude. Specifically, we have found that for a typical overflight, increasing or decreasing kS by 33% leads to an approximately 0.5-ppb rms change in inferred dye concentration in the upper 5 m, increasing to approximately 3.0 ppb at a depth of 15 m. We attribute the general increase in rms error with depth to the lack of stability of our inversion deeper in the water column. Meanwhile, the lack of a strong sensitivity to kS, particularly higher in the water column, is due to the fact that our least squares fit is constrained by the dye concentration profile inferred from the attenuation measurements at 532 nm, which is independent of any assumptions concerning kS.

In all of the above, we have not dealt explicitly with the lidar system response. Specifically, the measured lidar backscatter in both the 532- and 580-nm receive channels is the convolution of 1) the water column backscatter/fluorescence and 2) the laser pulse, receiver detector, amplifier, and digitizer bandwidth. For both channels, the latter was estimated on the ground by firing the system at a flat target from a distance of 80 m. For these measurements, attenuation filters were used on the outgoing laser beam to reduce the backscatter signals so that they would not saturate the various detectors in the system. In addition, a 532-nm filter was used in place of the 580-nm narrowband fluorescence filter to permit a backscatter measurement in the fluorescence channel. Regarding the latter, we note that the response time of the APD used in the fluorescence channel did not depend strongly on wavelength, so that using 532 nm rather than 580 nm for our flat-target tests of the APD resulted in only a small difference in the rise and fall response times, less than 10% of the total width of the flat-target response.

The pulse widths of the flat-target responses for both the backscatter and fluorescence channels were on the order of 10 ns, or about 1 m in depth (accounting for round-trip travel), with the backscatter channel showing a slightly wider peak than the fluorescence channel (not shown). Relevant here is that the vertical distribution of our dye patch was of comparable or greater scale, on the order of 2–5 m (e.g., see both the in situ and the lidar inversion results in Figs. 7, 8 and 11). Comparing power spectra of the flat-target responses with typical spectra from the raw received waveforms, we thus find that the bandwidths of both flat-target responses tend to be greater than those of their respective channel’s raw waveforms (Fig. 4). We also note that the raw waveform spectra for both channels drop to near or below the noise floor of our measurements before their concomitant flat-target spectra roll-off (e.g., frequencies greater than approximately 0.5–0.7 cycles per meter in Fig. 4). These two factors combined imply that deconvolving the flat-target response from the total received waveforms would amplify the noise but would not provide much signal gain (e.g., see dashed lines in Fig. 4).

In the above inversion procedure, even without performing the deconvolution, it was necessary to smooth the raw waveform data with an approximately 2-m smoothing filter to help reduce the noise. To address the increased noise that would accompany a deconvolution, in principle we could apply an additional low-pass filter to the deconvolved water column response. The precise shape of such a filter would likely differ somewhat from that of the flat-target response (e.g., it might be flatter). However, its effect would still be similar to that of the flat-target response we just deconvolved. Important here is that the amplitude of the lidar response is immaterial since our backscatter inversion is insensitive to amplitude, and our fluorescence inversion fits for amplitude. Also, given the approximately Gaussian shape of the flat-target response, based on theoretical considerations and tests using actual deconvolved waveforms, the slope of the deconvolved profiles and hence our estimate of the lidar attenuation coefficient ks is not sensitive to whether we use the convolved or deconvolved signals. With these considerations in mind, in the interest of simplicity and speed of our inversion, we thus choose not to perform such a deconvolution, but rather simply note that the resolution of our inverted dye profiles, whether it be due to the bandwidth of the lidar system response or due to the inherent bandwidth of the water column response, is limited to approximately 2 m.

3. Results

a. Experimental setting

The study site was located along the narrow portion of the East Florida Shelf, along the inshore edge of the region known as Miami Terrace. As with much of the East Florida Shelf, this area is strongly influenced by the Florida Current, which meanders back and forth across the region and varies in total transport, with periods ranging from 2 to 20 days, to seasonal and annual time scales (e.g., Lee and Williams 1988). At the location of our study site, east of Miami and Ft. Lauderdale, the inner edge of the Florida Current is typically about 10 miles from the coast (e.g., Gyory et al. 2005). At times, northward alongshore currents can exceed 2 m s−l within a few miles of the coast, while at other times, when the current is farther offshore, the nearshore flow can be significantly reduced or even reversed (southward). Tidal currents in the region are generally small compared to the Florida Current and furthermore would not have contributed significant variability over the time scale of our experiments (less than 2 h).

Weather for the week prior as well as during the experiments was generally fair, except for occasional thunderstorms during the afternoon and/or evening hours, although no thunderstorms passed over our field site during either of our experiments. Meteorological data were not available from the ship; however, data collected at the nearby Ft. Lauderdale airport were generally consistent with conditions observed at the study site, namely, scattered to partly cloudy with winds generally less than 10 kt out of the south-southeast. Waves at the study site throughout the experiment were generally small (i.e., on the order of 0.5 m or less).

CTD data collected during the dye surveys showed a strongly stratified surface layer (N ≈ 10 cph) extending to about 10 m, overlying a less stratified region down to 35 m (N ≈ 3.5 cph), followed by an abrupt density jump (approximately 0.5 kg m−3) at 35 m (Fig. 5). The associated temperature stratification was consistent with a remnant mixed layer, capped by surface heating. Specifically, the elevated heat content of the upper 10 m of the water column compared with deeper waters was consistent with 3–4 days of net surface warming by solar insolation, combined with relatively light winds. Currents consisted of a mean northward flow near the surface on the order of 10 cm s−1, overlying a 25 cm s−1 northwestward subsurface maximum centered at about 10 m, and a return to northward flow below 20 m of 15–20 cm s−1 (Fig. 6).

b. Results from the 3 June experiment

We present here results from the 3 June experiment, namely, the initial line transect and the zigzag survey, since these provided the densest sampling of the dye patch and since concurrent lidar data were also available at the same times (the final line transect took place after the aircraft had already left the study region). Dye concentrations along the major axis of the streak, as measured during the initial transect, are shown in Fig. 7. The transect shows three distinct patches of dye. From right to left (note, the x axis in Fig. 7 is reversed compared to Fig. 2, since the dye was injected from north to south), the northernmost (rightmost) segment of the dye patch starts near the surface, then deepens to approximately 5 m; the middle segment starts at 5 m, shoals to the surface, then deepens again to 5 m; and the southernmost (leftmost) segment starts at 5 m, and shoals to the surface. This pattern of surface and deep segments agrees well with the injection segments shown in Fig. 1, with the exception of the two 10-m sections of the patch, which by the time of the transect had already been advected to the west of the major axis of the surface patches and hence were not sampled by the line transect (recall that the subsurface shear would have advected the deeper dye westward compared to the surface patch; see Fig. 6). That dye concentrations measured at the southern end of the patch were somewhat less than those at the northern end is consistent with the fact that the injection was performed while steaming southward, while the initial sampling transect was done while heading northward. Thus, the time between injection and sampling was larger for the northern portion of the dye streak.

Cross-streak sections of the dye patch taken during the zigzag survey are shown in Fig. 8, starting from the northern end of the patch and progressing southward. (Note, for compactness of presentation, lidar-derived sections are also shown in Fig. 8, although we defer comparison between the lidar and in situ results until section 3d. Also, since corresponding airborne lidar data were only available for the early part of the survey, we show here only sections occupied during the first half of the survey.) In all sections, particularly those passing through the surface segments of the patch, evidence of an east–west shear in the upper portion of the water column is again visible in the dye data. Assuming an elapsed time of 1–1.5 h (the time between the injection and the zigzag survey) and a westward velocity of 0.1 m s−1 at 5-m depth (see Fig. 6), we would expect a total differential displacement on the order of 350–550 m between the surface and 5-m portions of the patch. This is roughly consistent with the observed displacements, most notably the two transects taken through the surface segments of the patch (i.e., Figs. 8a,d). Transects taken across the deeper segments of the patch, well away from the near-surface segments (Figs. 8b,c), show relatively little dye near the surface, with a weak maximum at 5-m depth toward the western edge of the survey. These displacements are again consistent with a stronger westward current at depth.

c. Results from airborne lidar

Overflights of the dye patch using airborne lidar commenced at the start of the dye injection and continued for 1.5 h (i.e., about halfway through the zigzag survey). A total of 21 flight lines over the dye patch were carried out: 12 in the north–south direction (i.e., along the axis of the dye streak) and 9 in the east–west direction (across the axis of the dye streak). The average time between successive overflights was about 4.5 min.

A subset of the raw lidar waveforms (intensity of backscatter versus time/depth) both inside and outside the dye patch for the infrared, green, and fluorescence channels is shown in Fig. 9. A peak in backscatter from the ocean surface is clearly visible in the IR and green channels. The IR signal is confined to the near surface, while the green channel shows a more gradual exponential decay with depth. Peaks in the fluorescent channel are also evident in profiles through the dye patch, distinct from profiles outside the patch. These correspond to profiles in the green channel in which the attenuation rate was significantly enhanced compared to ones where no dye was found.

The inversion procedures for both the green and fluorescence channels were applied for each of the lidar overflights per (2) and (5). In the case of the green channel, data collected well to the north and south of the dye patch were used to compute a mean I−1532 dI532/dz outside of the patch. Results of the two inversion steps for an overflight made approximately 0.5 h after the injection are shown in Fig. 10. The results show a clear signal in both channels. Noteworthy, however, are the differences between C computed from (2) and that from (5). Specifically, the noise level in the green channel is more evident than the noise in the fluorescence channel. Meanwhile, though the fluorescence channel appears cleaner, this is at the expense of a higher detectability threshold. These differences stem from differences in the two inversion procedures. Specifically, in the case of the green channel, attenuation profiles within the dye patch are compared to the mean profile outside the patch, per (2). As such, the noise in the inverted solution is the result of small spatial variations in the ambient attenuation profile, combined with instrument noise at low signal-to-noise levels. Both of these can result in spurious dye peaks, as well as zero or even negative dye concentrations. By contrast, the noise in the dye channel is primarily instrumental noise. To identify peaks in the dye channel that are clearly above the noise floor, we have chosen a threshold intensity, below which we set the inverted dye signal equal to zero. This results in an apparently cleaner signal with fewer spurious peaks but also with a higher detectability threshold. Despite these differences, however, the surface segments of the dye streak are clearly visible in both the green and fluorescence channel inversions. Also visible in the green channel inversion are the two southernmost 5-m segments of the dye patch. The northernmost 5-m segment and the two 10-m segments do not appear, possibly because they had already been advected westward out of the field of view of the lidar.

d. Comparison with in situ results

The results of the lidar surveys can be compared directly to those from the towed fluorometer data. The lidar data corresponding to the first transect along the major axis of the dye streak are shown in Fig. 11. Results from both channels show good qualitative agreement with the in situ data (see also Fig. 7) in terms of the locations and mean depths of the dye patch, as well as the amplitude of the dye concentrations. Again, however, significant noise is visible in the green channel C, particularly below regions where dye was found. By contrast, C from the fluorescence channel shows less noise at depth, but also fewer dye peaks overall, particularly for lower peak concentrations.

Similar comparisons can be made between the zigzag transects of the airborne versus the in situ observations (Fig. 8). Again, the lidar results show good agreement with the in situ data, both qualitatively and quantitatively, although the same differences between the lidar results and the in situ data, as well as between the two lidar channels, also exist.

A final check of the lidar inversions can be made by comparing the total mass of dye inferred from each of the inversions to the total amount that was actually injected. Taking an average over the 9 north–south overflights following the injection (3 of the 12 north–south overnights were done during the injection, before it was complete), the total dye mass accounted for in the lidar signal was 1.7 kg for the green channel and 1.4 kg for the fluorescent channel. These can be compared to the 5 kg of dye that were actually injected. There are three major reasons for these underestimates. The first reason, which would lead to both channels underestimating the total dye, particularly during later overflights, is that because of vertical shear in the water column, the deeper segments of the patch were advected westward, out of the field of view of the lidar. Because the lidar overflights were aligned with the surface segments of the dye patch, they generally did not capture these deeper dye segments. Taking this into consideration, and noting that about 60% of the total dye was injected in the near-surface segments of the streak, we would expect that approximately 3 kg of the dye would be contained in the near-surface segments. As the deeper portions of the surface dye segments were likely also advected westward, due to the strong vertical shears (e.g., Fig. 8), the actual amount of dye sampled by the lidar may have been even less than this. A second reason, particularly relevant to the earlier surveys when dye concentrations were much higher (up to 40 ppb peak concentration), was that the attenuation of the incident laser light due to absorption by dye, plus that due to the water itself, would have led to a more than 100-fold reduction in the intensity of the incoming pulse over the first (upper) 2 m of the dye patch [e.g., see Eq. (1)]. Such strong attenuations would have rapidly reduced the lidar signal to the noise floor of our measurement, and hence led to our inability to detect dye deeper in the water column. Although our inversion approach generally takes both the dye and water column absorption into account, it cannot overcome the extreme situation in which the laser pulse is attenuated away before it has fully traversed the dye patch. As a result, absorption by the upper portion of the dye patch could have led to the “shadowing” of dye at deeper depths, and hence an underestimate of the overall dye mass. Relevant to the fluorescence channel inversion, a third reason contributing to the underestimate of the dye concentration is that whenever the signal for an entire profile fell below a prespecified level, the signal was deemed indistinguishable from noise and the entire dye profile was assumed to be zero. As a result, profiles that had a very weak signal (i.e., near or below the noise level) underestimated the dye. This also explains why less of the dye was accounted for by the fluorescence channel inversion than the green channel, since the signal to noise was generally less in the fluorescence channel.

4. Discussion

a. Comparison of green and dye channel results

Comparing the lidar data to the in situ data, we find that the green channel inversion is better able to detect lower concentrations of dye but that it also has a greater amount of noise, including spurious but coherent peaks that are not in the in situ data. This is particularly so below regions where significant dye concentrations (on the order of 10 ppb or greater) were found. Meanwhile, the fluorescence channel shows only the higher concentration peaks in the dye, but with less noise otherwise. As noted previously, the latter is the result of constraints imposed by the inversion itself.

The results of the lidar inversions suggest that for both the green and fluorescence channels, dye concentrations on the order of 1–2 ppb are readily resolved, with the possibility of detections as low as 0.5 ppb in the green channel. Particularly noteworthy is that the dye concentrations inferred from the lidar agree as well as they do with the in situ data, despite the fact that the two sets of measurements were made independently of one another. This suggests that our inversion procedures capture at least the major attenuation, scattering, and absorption characteristics so as to yield robust results. A more thorough treatment of the inversion will likely lead to further improvements in the dye concentration estimates.

One of the difficulties of comparing the lidar data with in situ data is the problem of space–time aliasing. Specifically, the lidar measurements are nearly instantaneous compared to the in situ measurements (e.g., 20 s to complete an overflight versus 20 min for a single line transect). We have addressed this to some extent in our comparisons in section 3, inasmuch as we compare measurements from the two platforms for times and locations that are as close as is practically possible. Nevertheless, the fact that the dye patch is constantly evolving because of both advection and diffusion means that any mismatch in time inevitably translates to a mismatch in space as well. Still, the results shown in Figs. 7, 8 offer a reasonable comparison.

b. Errors and uncertainties

Neglecting possible errors and uncertainties associated with the model itself, the dominant source of error in the green channel inversion is natural variation in the background attenuation. The error associated with such variation can be estimated directly. Specifically, taking the two standard deviation limit of I−1532dI532/dz outside the dye patch, the uncertainty in the attenuation in the upper 10 m of the water column is about 0.07 m−1, increasing to 0.14 m−1 at 15 m. Dividing by 2αD532, this translates to an uncertainty in the dye concentration of 1 ppb from 0- to 10-m depth, increasing to 2 ppb at 15-m depth. This increase in uncertainty with depth is a direct result of the decreasing signal-to-noise ratio.

Uncertainties in the fluorescence channel dye estimates are more difficult to quantify, in part because of our use of an a priori noise threshold in the inversion but also because of the least squares fitting method used to constrain the integration constants, plus other assumptions such as that of a constant lidar attenuation coefficient. In the present analysis, we have set the noise threshold at 1.0 × 10−4 in the raw fluorescence intensity (see Fig. 9). This translates to a noise floor in absolute dye concentration again of about 1 ppb between 0- and 10-m depth, increasing to about 2 ppb at 15-m depth. Meanwhile, if variations in background attenuation as a function of depth were similar in the two channels, the uncertainty due to such variations in the fluorescence channel would also have been on the order of 1 ppb between 0 and 10 m, increasing to 2 ppb at 15-m depth.

One possible approach to attaining a greater dynamic range in the observations in future studies would be to increase the total amount of dye released in each experiment, thus increasing the overall signal to noise. The problem with this approach, however, is the self-shadowing of the dye described in section 3d, which for a given level of incoming radiation (e.g., a given laser pulse power) sets an upper limit on the total amount of dye the laser can penetrate. Based on the present inversion results, we estimate this limit to be approximately equivalent to a 3–4-m-thick layer of dye with concentration on the order of 10 ppb, or about 3.5 ppb of dye averaged over the upper 10 m of the water column.

Putting the above uncertainties into perspective, the minimum detectable level for Rhodamine WT in clear water using a commercial fluorometer is approximately 0.01 ppb. Given that the uncertainties in the lidar inversions are set in part by variations in the background attenuation, a more refined inversion approach will thus be required to come close to the in situ detection level. Nevertheless, the very large number of observations made by the lidar (on the order of 30 000 profiles during a single overflight compared to 200 in situ profiles for the entire experiment) provides a wealth of information about the overall dye distribution, which would otherwise not be obtainable from in situ measurements alone. The two types of observations are complementary in this regard.

5. Summary and conclusions

The results here are from a pilot study using airborne lidar to survey dye release experiments in the upper ocean on spatial scales of meters to kilometers and temporal scales of minutes to days. Dye concentrations as a function of horizontal position and depth from two inversion approaches are compared to ship-based observations using an in situ fluorometer. Results show qualitative as well as quantitative agreement between dye distributions inferred from the two lidar channels versus the in situ measurements. Dye concentrations detected by the airborne lidar ranged from 1 ppb to greater than 10 ppb, with the lower limit set by variations in the background (ambient) attenuation coefficient and the upper limit set by self-shadowing of the dye.

The present results represent a significant advance in the use of airborne lidar for surveying dye release experiments insofar as this is the first time, to our knowledge, that depth-dependent dye concentrations have been reported in the peer-reviewed literature. A second novel aspect of this work is the method used to invert the raw lidar signal for absolute dye concentration. Specifically, the method used here requires no a priori knowledge of in situ dye concentration, although we used such measurements to provide validation of the technique.

While the present study shows the basic capabilities of airborne lidar for obtaining three-dimensional maps of dye release experiments in the upper ocean, there is still considerable room for improvement in terms of signal to noise, dye detection limits, and depth penetration. Given the relative success of the simple inversion approach applied here, however, it is likely that a more sophisticated inversion procedure that takes advantage of known characteristics of the background attenuation, instrument and environmental noise, and the dye signal could lead to significant improvement in the accuracy of the dye concentration estimates. On the instrumentation side, it should be noted that the system used here was a slight modification of a bathymetric system. A system designed specifically with the above issues in mind, including possibly the use of more sensitive detectors, would also greatly enhance the capability of airborne lidar for surveying dye release experiments. Given the very high resolution, both temporally and spatially, provided by airborne lidar, we believe measurements such as those presented here hold great promise for significant advances in our understanding of small-scale dispersion in the upper ocean.

Acknowledgments

Support was provided by the Cecil H. and Ida M. Green Technology Innovation Fund under Grant 27001545, the Office of Naval Research Grant N00014-01-1-0984, and the Woods Hole Oceanographic Institution Coastal Ocean Institute. The authors thank E. Culpepper and C. Wiggins of the U.S. Army Corps of Engineers Joint Airborne LIDAR Bathymetry Technical Center of Expertise for their assistance with the airborne measurements. We are grateful to Captains R. Franks and M. Andrews and Florida Atlantic University for use of the R/V Stephan, and T. Donoghue and S. Bohra for assistance with staging and deployment related to the ship-based portion of the field work. We also thank F. Hoge, J. Churnside, J. Buck, and H. Sosik for helpful discussions.

REFERENCES

  • Churnside, J. H., , Tatarskii V. V. , , and Wilson J. W. , 1997: Lidar profiles of fish schools. Appl. Opt., 36 , 60116020.

  • Churnside, J. H., , Tatarskii V. V. , , and Wilson J. W. , 1998: Oceanographic lidar attenuation coefficients and signal fluctuations measured from a ship in the Southern California Bight. Appl. Opt., 37 , 31053112.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Feldman, G., , and McClain C. R. , cited 2004: Ocean Color Web, SeaWiFS Reprocessing Level 2. NASA Goddard Space Flight Center. [Available online at http://oceancolor.gsfc.nasa.gov/.].

  • Franz, H., , Gehlhaar U. , , Günther K. P. , , Klein A. , , Luther J. , , Reuter R. , , and Weidemmann H. , 1982: Airborne fluorescence LIDAR monitoring of trace dye patches—A comparison with shipboard measurements. Deep-Sea Res., 29 , 893901.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Gordon, H. R., 1982: Interpretation of airborne oceanic lidar: Effects of multiple scattering. Appl. Opt., 21 , 29963001.

  • Günther, G. C., , Thomas R. W. L. , , and LaRoque P. E. , 1996: Design considerations for achieving high accuracy with the shoals bathymetric LIDAR system. Laser Remote Sensing of Natural Waters: From Theory to Practice, V. Feigels and Y. Kopilevich, Eds., SPIE, 54–71.

    • Crossref
    • Export Citation
  • Gyory, J., , Rowe E. , , Mariano A. J. , , and Ryan E. H. , cited 2005: The Florida current. [Available online at http://oceancurrents.rsmas.miami.edu/atlantic/florida.html.].

  • Hoge, F. E., , and Swift R. N. , 1981: Absolute tracer dye concentration using airborne laser-induced water raman backscatter. Appl. Opt., 20 , 11911202.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hoge, F. E., , and Swift R. N. , 1983: Airborne dual laser excitation and mapping of phytoplankton photopigments in a Gulf Stream warm core ring. Appl. Opt., 22 , 22722281.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Irish, J. L., , and Lillycrop J. W. , 1999: Scanning laser mapping of the coastal zone; the shoals system. ISPRS J. Photogramm. Remote Sens., 54 , 123129.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Klett, J. D., 1981: Stable analytical inversion solution for processing LIDAR returns. Appl. Opt., 20 , 211220.

  • Klett, J. D., 1985: Lidar inversion with variable backscatter/extinction ratios. Appl. Opt., 24 , 16381643.

  • Kullenberg, G., 1971: Vertical diffusion in shallow waters. Tellus, 23 , 129135.

  • Lee, T., , and Williams E. , 1988: Wind-forced transport fluctuations of the Florida current. J. Phys. Oceanogr., 18 , 937946.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Mack, S. A., , Vasholz D. P. , , Larson E. C. , , Scheerer D. J. , , Calman J. , , and Schoeberlein H. C. , 1997: Estimation of diapycnal diffusivity from a dye tracer study in the upper seasonal thermocline. Eos, Trans. Amer. Geophys. Union, 78 .(Fall Meeting Suppl.), Abstract F374.

    • Search Google Scholar
    • Export Citation
  • Morel, A., , and Maritorena S. , 2001: Bio-optical properties of oceanic waters: A reappraisal. J. Geophys. Res., 106 , 71637180.

  • Okubo, A., 1971: Oceanic diffusion diagrams. Deep-Sea Res., 18 , 789802.

  • Phillips, D. M., , and Koerber B. W. , 1984: A theoretical study of an airborne laser technique for determining sea water turbidity. Aust. J. Phys., 37 , 7590.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Squire, J. L., , and Krumboltz H. , 1981: Profiling pelagic fish schools using airborne optical lasers and other remotes sensing techniques. Mar. Technol. Soc. J., 15 , 2731.

    • Search Google Scholar
    • Export Citation
  • Terray, E. A., and Coauthors, 2005: Airborne fluorescence imaging of the ocean mixed layer. Proc. Eighth Working Conf. on Current Measurement Technology, Southampton, UK, IEEE/OES, 76–82.

  • Vasholz, D. P., , and Crawford L. J. , 1985: Dye dispersion in the seasonal thermocline. J. Phys. Oceanogr., 15 , 695711.

  • Yoder, J. A., , Aiken J. , , Swift R. N. , , Hoge F. E. , , and Stegman P. M. , 1993: Spatial variability in near-surface chlorophyll-a fluorescence measured by the Airborne Oceanographic Lidar (AOL). Deep-Sea Res., 40 , 3753.

    • Search Google Scholar
    • Export Citation
Fig. 1.
Fig. 1.

Chlorophyll concentration inferred from satellite sea surface color over the Florida Straits (Feldman and McClain 2004). Inset to the figure shows the ship track during the 3 June dye surveys—red, green, and magenta curves indicate the first transect, the zigzag survey, and the final transect, respectively. Bold lines indicate the ship’s location at times corresponding to the lidar overflights.

Citation: Journal of Atmospheric and Oceanic Technology 24, 6; 10.1175/JTECH2027.1

Fig. 2.
Fig. 2.

Injection time series showing the staircase of different depths where the dye was injected. The injection was performed while the ship was headed southward, hence, early times correspond to the northernmost extent of the patch, while later times are at the southern extent.

Citation: Journal of Atmospheric and Oceanic Technology 24, 6; 10.1175/JTECH2027.1

Fig. 3.
Fig. 3.

Raw backscatter for 2500 profiles outside the dye patch plotted on a logarithmic scale showing approximately constant attenuation with depth. A mean regression line corresponding to an attenuation of 0.085 m−1 is shown as a black line.

Citation: Journal of Atmospheric and Oceanic Technology 24, 6; 10.1175/JTECH2027.1

Fig. 4.
Fig. 4.

(a) Power spectra of normalized lidar flat target response (thick solid), raw backscatter (solid), and deconvolved backscatter (dashed) for a single waveform within the dye patch. (b) Same as in (a), but for corresponding fluorescence channel waveform.

Citation: Journal of Atmospheric and Oceanic Technology 24, 6; 10.1175/JTECH2027.1

Fig. 5.
Fig. 5.

(left) Temperature, (middle) salinity, and (right) potential density profiles during the 3 June dye release experiment.

Citation: Journal of Atmospheric and Oceanic Technology 24, 6; 10.1175/JTECH2027.1

Fig. 6.
Fig. 6.

Mean velocity profiles during the 3 June dye release experiment: eastward (solid) and northward (dashed). Dotted lines indicate one standard deviation about the means.

Citation: Journal of Atmospheric and Oceanic Technology 24, 6; 10.1175/JTECH2027.1

Fig. 7.
Fig. 7.

In situ dye concentration [log10(C/ppb)] as measured during the first transect along the major axis of the dye streak showing the three near-surface (centered at about 2.5-m depth) segments of the dye patch, interspersed with deeper (5 m) segments. Two gaps where no dye was found (approximately 25.944° and 25.95°N) correspond to the portions of the streak that were injected at 10 m, which, due to a strong vertical shear, had already been advected westward of the surface patch. Black triangles indicate locations of the downcast portions of the surveys.

Citation: Journal of Atmospheric and Oceanic Technology 24, 6; 10.1175/JTECH2027.1

Fig. 8.
Fig. 8.

Dye concentration observed during the 3 June zigzag survey showing the vertical and cross-streak structure of the dye patch 1–1.5 h after injection. (a)–(d) Four transects consist of the (top) in situ, (middle) green channel, and (bottom) fluorescent channel with plan view of ship track and corresponding lidar overflight on the rhs. Color scale is log10(C/ppb). Note, in situ concentration estimates near the sea surface (uppermost 1–2 m) may be artificially elevated due to the sampling sled breaching the surface.

Citation: Journal of Atmospheric and Oceanic Technology 24, 6; 10.1175/JTECH2027.1

Fig. 9.
Fig. 9.

(left) Raw IR, (middle) green, and (right) fluorescence backscatter lidar signals observed (top) during a pass over the near-surface segment of the dye patch and (bottom) for a series of profiles outside the dye patch. Each panel contains 100 profiles.

Citation: Journal of Atmospheric and Oceanic Technology 24, 6; 10.1175/JTECH2027.1

Fig. 10.
Fig. 10.

Dye concentration [log10(C/ppb)] inferred from airborne lidar: (top) green channel and (bottom) fluorescence channel. Vertical slices are approximately every 2 m starting from the surface down to 12 m.

Citation: Journal of Atmospheric and Oceanic Technology 24, 6; 10.1175/JTECH2027.1

Fig. 11.
Fig. 11.

Dye concentration [log10(C/ppb)] as measured by lidar during an overflight approximately 10 min into the first line transect along the major axis of the dye streak based on the green channel and fluorescent channel inversions, respectively. Data shown correspond to averages of profiles taken within 10 m of the ship-based transect positions. An offset in the longitudinal position has been applied to the lidar positions, corresponding to approximately 0.5 m s−1 to account for the westward advection of the patch between the in situ and lidar data sampling times.

Citation: Journal of Atmospheric and Oceanic Technology 24, 6; 10.1175/JTECH2027.1

Save