Open Access
Add to BibSonomyAbstract
Two single-waveband low-light radiometers were developed to characterize properties of the underwater light field relevant to biological camouflage at mesopelagic ocean depths. Phenomena of interest were vertical changes in downward irradiance of ambient light at wavelengths near 470 nm and 560 nm, and flashes from bioluminescent organisms. Depth profiles were acquired at multiple deep stations in different geographic regions. Results indicate significant irradiance magnitudes at 560 nm, providing direct evidence of energy transfer as described by Raman scattering. Analysis of a night profile yielded multiple examples of bioluminescent flashes. The selection of high-sensitivity, high-speed silicon photomultipliers as detectors enabled measurement of spectrally-resolved irradiance to greater than 400 m depth.
© 2014 Optical Society of America
1. Introduction
Biological camouflage in the open ocean is a complex problem. There is no physical structure to hide within or behind, and potential predators can approach from any position in a three-dimensional space. The ambient light field is constantly changing due to sun position, sky conditions, sea state, inherent optical properties of the water, depth in the water column, and so on. A specific framework for examining the visual ecology of marine pelagic animals, i.e., predator-prey relationships, can be depicted as shown in Fig. 1. The solar input at the sea surface, modified by the aforementioned factors, contributes to the background radiance field. For an organism attempting crypsis to hide from a predator, the background radiance is important, as it results in an incident irradiance on the surface of the animal. The incident irradiance is scattered by the prey animal, as quantified by the bidirectional reflectance distribution function (BRDF) [1], which results in a reflected radiance. It is also possible for a predator to use self-generated light sources, i.e., bioluminescence, to add to the incident irradiance on the prey, which increases the reflected radiance. A potential predator can then detect both the background and reflected radiances, so that perfect camouflage is attained when these quantities are equal over all wavelengths accessible by the predator visual system. Appearance of the prey, as perceived by the predator, is therefore dependent on all of the above components.
Fig. 1 Schematic of predator-prey visual ecology problem described in text. Solar radiance Ls incident on the sea surface results in a transmitted radiance Lt, which contributes to the underwater background radiance Lb. Background radiance Lb is scattered by the prey animal, resulting in a reflected radiance Lr. A potential predator may detect both radiance quantities, such that perfect camouflage occurs for Lr = Lb.
Early work on understanding the connections between marine animal appearance and behavioral ecology focused primarily on the interaction of light with fish scales [2] and silvery layers in and around the eyes of various fish and cephalopods [3]. Qualitative observations were made that the color of scattered light varied with incident and viewing angles. Based on underwater optical measurements, mathematical models were created to determine the optimal shape and reflective properties that a fish would need in order to succeed in various situations. For example, the type of camouflage required to avoid predators in different viewing locations was considered [4–6]. However, the conclusions were limited, since radiometric data for the ambient light field and animal BRDF was not available for comparison. More recent studies have attempted to quantify the observed appearances, through development of physical scattering models from structural information [7] as well as direct measurement of the BRDF of mirrored animal specimens [8].
Recent comprehensive radiative transfer simulations examined the ambient light field from the surface to 1000 m depth [9]. The results suggest that the generally accepted assumption of mesopelagic (200 m to 1000 m depth) light that is increasingly centered near 475 nm with diminishing angular extent, as it travels through the water column [10], is too limiting. In particular, it was shown that the light field does not approach the near-asymptotic regime until ∼400–500 m depth in clear water. In addition, secondary spectral maxima emerge primarily as a result of Raman (inelastic) scattering processes, near 565 nm and 695 nm, with biologically-significant magnitudes. Existing spectrally-resolved deep-sea light data are too sparse to confirm these predictions, in the sense that the measurements were broadband [11] or limited to wavelengths less than 540 nm [12, 13]. This clearly indicates the need for new instrumentation and field experiments focused specifically on obtaining spectrally-resolved light measurements, including those near the secondary maxima wavelengths resulting from inelastic scattering effects, down to at least 500 m depth.
At night, or below mesopelagic depths, ambient light may not be sufficient for visual detection of prey [14]. Successful predation may then require the use of bioluminescent sources to yield a detectable magnitude of reflected radiance from the target animal. In some cases, such as for giant squid, bioluminescence stimulated by potential predators is observed and used as an aversion technique [15]. A proper examination of visual ecology scenarios thus requires characterization of bioluminescent flashes. Most of the available information on flash kinetics, e.g., maximum response and flash duration, was acquired through laboratory measurements of electrically-stimulated bioluminescence [16, 17]. Field measurements of bioluminescence have been conducted, most at blue wavelengths, including an example measurement of stimulated bioluminescence [18], characterization of distribution patterns using the spatial plankton analysis technique (SPLAT) [19], and data from a system designed to measure fine-scale coastal bioluminescence [20]. However, a deep-sea capable instrument that can measure in situ flash kinetics at high temporal rates at multiple wavelengths would be useful.
Based on the present limited ability to acquire measurements of vertical changes in the deep-sea ambient light field as well as bioluminescent flash kinetics, two low-light radiometers have been developed that are uniquely suited to measuring irradiance due to both phenomena of interest. In the current configuration, the instruments are depth-capable to 1000 m, and acquire irradiance at two wavelengths (472 nm and 562 nm), with detector gate times as short as 100 μs. The instrument design, initial field testing, and statistical data analysis are described in the following sections.
2. Instrument for measuring mesopelagic light
2.1. Design considerations
For examination of ambient light resulting from solar radiance at the sea surface, the physical quantity of interest is spectral downward plane irradiance Ed [21], which can be written as
with spectral radiance L, spatial x and temporal t parameters, zenith θ and azimuthal ϕ angles, and wavelength of light in vacuum λ. The specified integration limits correspond to downward light incident on a planar measurement surface. Note that spectral radiance L is of fundamental importance in ocean optics, and spectral irradiance E represents the sum of this quantity over a specified angular range.Neglecting spatial and temporal parameters, the optical power Φ incident on the focal plane of a detector is written as
with system throughput Tλ = T (λ), spectral radiance Lλ = L (λ), entrance pupil area Ap, detector solid angle Ωp, and spectral bandwidth Δλ. Note that the approximation results from the assumption that irradiance does not change significantly over the measured wavelength band. The validity of this approximation is directly related to the resultant estimation error.The present quantity of interest is irradiance Eλ = E (λ), such that the radiance is given by Lλ = ρEλ/π, where ρ is the diffuser transmission, since the incident light is scattered into a flat plane by the Lambertian diffuser. The entrance pupil area Ap can be further expanded as , where Dp is the entrance pupil diameter. For small angles, the solid angle is expressed as Ωp ≈ θ2 = (d/f)2, where θ is detector field of view (FOV), d is detector size, and f is focal length. For a small detector size d, the FOV will be very limited, so the small-angle approximation is reasonable. The incident power is then written as
Recognizing that f/Dp is the f -number (f/#), the incident optical power becomes Note that the optical power incident on the detector is maximized by using the fastest (lowest f/#) lens available. In practice, the system throughput Tλ is calculated, the diffuser transmission ρ, the spectral bandwidth Δλ, detector size d, and f -number f/# are known, and optical power Φ is measured.Given prior knowledge of the underwater light field distribution, i.e., magnitude of available spectral irradiance Eλ in the water column, the expected optical power Φ incident on the detector can be estimated. The desired measurement depths, along with the power estimates, directly define the required dynamic range of the detector. Specifically, the maximum measurement depth will be limited by the amount of available optical power relative to the noise power. The analysis shows that a larger detector size and smaller f/#, resulting in larger FOV, increases incident optical power. However, detector dark noise rate also increases with surface area [22]. The primary trade is then between system f/# and detector size d. When the system sensitivity is detector limited, as in the present case, detector dark noise rate must also be considered.
A small number of existing instruments have been developed specifically for measuring the deep-sea light field. The most well-known of these systems is the low-light auto-calibrating radiometer (LoLAR) [18]. At the time the LoLAR was developed, commonly available optical diffusers provided a transmission of less than 1%. The LoLAR system thus utilized a custom-designed r-θ lens, which maps incident ray angles to concentric rings, to provide high efficiency throughput with a cosine-corrected response. LoLAR was a logical progression from earlier systems that required even more extensive analog circuit design [12, 13]. All of these systems contained a photomultiplier tube (PMT) as the detector due to the relatively high sensitivity available. However, PMTs require special considerations, since they require high voltage power supplies and can be easily damaged by exposure to high light levels.
By taking advantage of recent advances in solid state detector technology, the low-light radiometers presented here greatly reduce the complexity of previous systems, resulting in lower cost and minimization of development time. Instead of a PMT, a silicon photomultiplier (SPM) is used as the detector. The SPM is in many ways the solid state equivalent of the PMT, but with certain advantages, including low input voltage requirements and insensitivity to magnetic fields. In addition, using an engineered diffuser yields a cosine response with high optical transmission. The combination of a high transmission diffuser and SPM detector allows for the use of relatively narrowband spectral filters, while still allowing for low-light measurement capability.
2.2. Hardware design
A key design choice was the selection of the detector as a thermoelectrically-cooled multi-pixel photon counter (MPPC) module from Hamamatsu Photonics (Hamamatsu, SZK, Japan). The MPPC is essentially an SPM in that it consists of an array of avalanche photodiodes (APDs) operating in Geiger mode, resulting in high gain [23]. Although the MPPC is a two-dimensional array, the signals from individual APDs are summed to provide a single value corresponding to the photon flux incident on the detector active area. To maximize light input to the system, a high-efficiency Lambertian (HiLAM) diffuser from RPC Photonics (Rochester, NY) was selected.
Two low-light radiometers were developed, each with a different spectral filter. Optical and detector system components are shown in Fig. 2. The basic measurement process is as follows. External light incident on the optical port P is transmitted and illuminates the diffuser D. Light is then transmitted by the diffuser D and emitted with a Lambertian profile. The first lens L1 is located at a distance near its focal length from the diffuser D. Light transmitted by the diffuser D is then collimated by lens L1. The collimated light is reduced and transmitted by the filter S, and arrives at the second lens L2. The remaining light is then focused by lens L2 onto the active area of detector module M. The signal received by detector M is recorded to disk by the control computer. The tilt sensor/digital compass simultaneously records instrument position and orientation.
Fig. 2 Schematic showing system components, including acrylic optical port (window) P, engineered diffuser D, lenses L1 and L2, spectral filter S, and detector module M.
Each radiometer system is enclosed in a cylindrical aluminum housing, pressure tested to 1000 m depth, with length 22.5” (57.15 cm) and diameter 4.5” (11.43 cm). The instrument housings were custom designed and machined from 6061-T6 aluminum alloy with Type III hard anodizing. External interfaces to the instruments were provided by wet-pluggable bulkhead connectors (SubConn, Pembroke, MA). Power is provided to each unit by a custom battery housing containing rechargeable lithium-ion battery packs. The instruments are controlled autonomously by small computers contained in the housings, which acquire and store data from the attached devices.
The following is a list of components with associated specifications used in the deep-sea low-light radiometer systems. Externally attachable neutral density (ND) filters were fabricated by waterjet machining of .125” (3.175 mm) thick Acrylite FF dark gray 2074 acrylic (Evonik CYRO, Parsippany, NJ) to 4.5” (11.43 cm) diameter. The optical port P was fabricated by waterjet machining of 1.5” (38.1 mm) thick Polycast Super Abrasion Resistant (SAR) acrylic (Spartech, Samford, CT) sheet to 3” (76.2 mm) diameter. The engineered diffuser D is a 1” (25.4 mm) diameter RPC-HiLAM diffuser consisting of two polymer-on-glass substrate components separated by a central air gap, with a transmission efficiency of ∼75% over a wide spectral range of 350–2000 nm. Cosine response data (not shown) provided by the manufacturer indicate <5% error from 0–65° and ∼15% error from 65–85°. The lens L1 is a ThorLabs LA1608 25 mm f/.33 plano-convex lens. The lens L2 is an Edmund Optics NT67-252 25 mm f/.6 aspheric lens. The spectral filters S are 25 mm diameter with optical density 6 (OD6) blocking, Edmund Optics NT67-033 472 nm with 30 nm bandpass and NT67-027 562 nm with 40 nm bandpass. The detectors M are 1 mm × 1 mm area Hamamatsu MPPC modules C11208-03 with 100 μm pixels and C11208-150 with 50 μm pixels. Note that two different detectors were used due to product availability limitations. The MPPC modules include internal temperature sensors. The computers are LP-170-G Pico-ITX with Intel Atom CPU (Global American, Hudson, NH). The solid state disks are Intel 530 series 120GB. The tilt sensor is an OS5000-US solid state tilt compensated 3-axis digital compass (OceanServer Technology, Fall River, MA).
2.3. Software design
The operating system for the internal instrument computers was Microsoft Windows 7. Custom C# software was developed to acquire data from the MPPC modules and the tilt sensor. Computer control and data recovery is performed by remote desktop via direct RJ-45 ethernet connection. The overall design strategy for the acquisition programs was to enable simple and reliable operation of connected devices, particularly while in the field.
2.4. System calibrations
It is necessary to relate the optical power received by the detector to the irradiance incident on the diffuser. That is, measured photon counts per second (cps) need to be converted to spectral irradiance (W/m2/nm) or spectral quantum irradiance (counts/s/m2/nm). There are light losses in the system due to optical throughput reduction corresponding to the total system spectral response. With the assumption that any losses are linear, which is reasonable following correction for detector nonlinear response, the system calibrations yield a multiplicative scale factor. The power incident on the detector can then be converted as desired.
Spectral transmission data for the optical port, spectral filters, along with detector photon detection efficiency (PDE) are shown in Fig. 3. Spectral transmission was measured for the acrylic ND filters (not shown) and optical port using a dual-beam spectrophotometer (PerkinElmer, Waltham, MA) with a 15 cm integrating sphere (Labsphere, North Sutton, NH). Spectral filter and detector response curves were generated using manufacturer supplied values. The total system spectral throughput Tλ was estimated by multiplying the transmission curves for the ND filters (if used), optical port, and spectral filters, by the PDE curve of the detector.
Fig. 3 Spectral response data for both radiometer systems. Response curves are included for the acrylic port (red line), blue and green filters (dark blue and green dashed lines), and MPPC detectors (light blue and green solid lines). Gray shaded regions represent total spectral response for each system (dark blue and green solid lines).
Each low-light radiometer would ideally have a linear response to incident optical power, with an offset caused by dark counts. That is, double the incident light yields double the photon count recorded by the detector. In general, the response is nonlinear and must be determined by laboratory calibration. The calibration is performed using a white light-emitting diode (LED) with a known spectrum. The low-light radiometers are mounted such that the diffusers are in a plane parallel to a known calibrated power sensor (ThorLabs, Newton, NJ). The magnitude of irradiance is reduced in steps, by placing calibrated neutral density filters in between the LED, and the radiometers and power sensor. Power supply current is adjusted to access the full range of available LED intensity, while simultaneously acquiring data with the low-light radiometers and power sensor. Response curves are constructed by plotting radiometer photon count against measured source power. Normalized response curves for both radiometer systems are shown in Fig. 4. Using the estimated system response, nonlinear effects can be removed from the acquired data using a lookup table. Note that information past the saturation point can be recovered, if it is known that the detector is operating in this regime, e.g., in the presence of slowly-varying light fields.
Fig. 4 Measured nonlinear response for each radiometer system. Input power is normalized to saturation level of the blue-filtered radiometer. Nonlinear curves are included for the blue-filtered radiometer (blue dash-cross line) with saturation level (blue solid line) and green-filtered radiometer (green dash-cross line) with saturation level (green solid line).
Following the previous formulation, spectral irradiance can be calculated as
where Φ is measured and Tλ, ρ, Δλ, d, and f/# are known quantities. The conversion between optical power Φ incident on the detector and spectral irradiance Eλ is then obtained using a scaling factor χλ that depends on the optical system. Estimated spectral irradiance Êλ is obtained by multiplying measured optical power Φ by the appropriate scaling factor, i.e., Êλ = χλ Φ where the units of spectral irradiance are cps/m2/nm, or W/m2/nm if the values have been converted from photon counts. Spectral throughput Tλ was taken as the average value over the filter bandwidth. Radiometer design specifications are shown in Table 1.
Table 1. Radiometer optical and detector design specifications.
An in situ cross-calibration of the low-light radiometers with a calibrated radiometer was performed to ensure an accurate measurement of absolute irradiance. The low-light radiometers with ND filters attached were deployed alongside a HyperPro spectroradiometer package from Satlantic (Halifax, NS, Canada) in an outdoor deep water tank at 3 m depth, and downward irradiance was measured. Simultaneously, the solar irradiance at the surface was monitored. The spectroradiometer was sensitive from 350 nm to 800 nm with 10 nm resolution. For proper comparison, spectroradiometer values were interpolated and reduced according to the spectral filter bandwidths of the low-light radiometers. Absolute scale factors for each low-light radiometer were obtained by minimization of the mean-square error between the scaled (by χλ) low-light radiometer and reduced (by Δλ) spectroradiometer data. Measured calibration and absolute scaled data from the low-light radiometers and spectroradiometer are shown in Fig. 5.
Fig. 5 Cross-calibration results from comparing downward irradiance measurements from Satlantic radiometer package (light blue and green solid lines) with low-light radiometer values (dark blue and green dashed lines). Low-light radiometer values are shown after adjustment by the estimated absolute scale factor.
A complementary method of nonlinear response calibration was developed to improve low-light-level nonlinearity estimation. The basic assumption is that, over some range of water depths, exponential light decay is observed. If a smooth irradiance depth profile is acquired, i.e., under constant sky conditions with no apparent scattering layers in the water column, a straight line can be fit to the log irradiance, as a function of depth, within the linear detector region. Using the presumed prior dependence, irradiance profile deviations from this line can then be used to determine the nonlinear system response to light input, as was done in the laboratory calibration shown in Fig. 4. Prior to nonlinear response estimation, smoothing of the profile is required to eliminate small irradiance fluctuations possibly resulting from surface waves or small vertical changes in the inherent optical properties, which is made possible by the high time resolution of the detectors. The method was performed with a downward irradiance depth profile acquired in the field. The model was validated by comparing the root-mean-square error (RMSE) of the cross-calibration minimization procedure, using both nonlinear estimation methods applied to the same raw data. The resultant RMSE was two to three times lower for the field method, yielding R2 values of 0.98 and 0.96 for blue and green detectors respectively. The field method was therefore used to process acquired irradiance data and greatly improved correction at low light levels.
2.5. Acquisition procedure
For acquisition of field data, the two low-light radiometers are deployed alongside other instruments, all contained in a mesopelagic light meter (MLM) optical package. As shown in Fig. 6, the complete MLM field system consists of:
- Low-light radiometer with 472±15 nm spectral filter
- Low-light radiometer with 562±20 nm spectral filter
- Absorption/attenuation meter with pressure (depth) sensor
- Data handler for absorption/attenuation meter
- Custom rechargeable battery pack
The absorption/attenuation meter is an ac-9 spectrophotometer with DH4 data handler (WET Labs, Philomath, OR). The ac-9 is depth-capable to 500 m and is used to measure absorption (a) and attenuation (c) coefficients at 9 discrete wavelengths. Note that the scattering coefficient (b) can then be estimated, since a + b = c [21]. Vertical changes in the inherent optical properties of seawater can be tracked using these measurements, which improves conclusions related to the distribution of irradiance in the water column. In addition, a Satlantic surface radiometer is operated simultaneously in order to obtain solar irradiance input.
Fig. 6 Mesopelagic light meter (MLM) field deployment package, including two low-light radiometers, absorption/attenuation meter with data handler, and rechargeable battery pack. (a) Actual photograph. (b) Component diagram.
The MLM package is connected to 1/4” non-conducting hydrowire using a swivel bolt and shackle, such that the wire spins, rather than the system. The system acquires data autonomously after being powered on at the surface. The package is deployed down to the target depth (with a maximum of 500 m due to pressure rating limitations of ac-9) at a winch rate of 10–15 m/min. Data is downloaded after the system is retrieved following the deployment.
3. Results
3.1. Field studies
A field study using the MLM system was performed during a research cruise on the R/V New Horizon in the eastern Pacific, near the coast of San Diego, California on April 1 and 2, 2014. The overall goal was to improve understanding of the low-light radiometer capabilities, specifically as related to measurements of downward irradiance and bioluminescent flash kinetics. Two deep stations were selected to eliminate bottom reflection effects. The first station was located in the San Diego Trough (1200 m bottom depth) and the second in the San Clemente Basin (2000 m bottom depth). Over the two field days, a variety of sea and sky conditions were experienced. Multiple depth profiles were acquired, including both day (irradiance) and night (bioluminescence) measurements.
A second field study was performed during a research cruise on the R/V Atlantic Explorer in the western Atlantic, with a transect from Barbados to Bermuda from May 31 to June 8, 2014. Four deep stations were selected for deployment of the MLM system. The irradiance profiles presented here were collected at a station east of Guadeloupe (5000 m bottom depth) under clear sky conditions. Note that the accessible range of measurement depths is dependent on water clarity, sky conditions, and sun position. The Atlantic stations were characterized by optically clearer waters than the Pacific stations.
The low-light radiometers were designed for deployment on a variety of instrument platforms, including free-falling profilers. In this study the primary objective was to measure downward irradiance at depths below 150 m, where ship shadowing is negligible. However, for shallow depths, e.g., from the surface to 50 m depth or less, depending on water clarity, the deployment system should be designed to drift away from the ship prior to profiling.
3.2. Downward irradiance
The MLM system was first deployed in the San Diego Trough down to 500 m depth under overcast sky conditions and moderate sea state with wave heights between 1 to 2 m. Based on radiative transfer simulation results [9], it was determined that two acrylic ND filters should be attached to the blue-filtered radiometer, such that the signal measured in each spectral band would be comparable. The detector gate times were set to 2 ms to enable high temporal (depth) resolution in the measurements. Ambient and internal instrument temperatures were expected to decrease with depth, but the detector temperature remained constant due to active cooling. Measurement-derived radiometer specifications of interest, i.e., saturation and noise levels, are shown in Table 2.
Table 2. Radiometer measurement-derived specifications.
Following acquisition, the data were processed to remove nonlinear effects and obtain absolute irradiance levels. Depth values were estimated using data from the pressure sensor connected to the ac-9 instrument. Example down-cast Ed (z, λ) profiles for both wavelengths over a depth range from 165 m to 235 m, which encompasses a transition from the epipelagic to mesopelagic zone, are shown in Fig. 7(a). Close inspection of the Ed (z, λ) profiles reveals apparent periodic motions of the instrument. Additional analysis (not shown) of the oscillations indicates that the amplitude decreases with depth, and the period fluctuates between 5 s to 10 s.
Fig. 7 Irradiance depth profiles for mesopelagic light meter system deployed in the San Diego Trough. (a) Log irradiance. (b) Diffuse attenuation functions. (c) Log ratio of blue and green irradiance profiles. (d) Optical signal-to-noise ratio calculated over 10 cm depth intervals.
The diffuse attenuation coefficient for spectral downward irradiance [21] is defined as
To obtain profiles of Kd (z, λ), it was necessary to calculate the derivative from the measured irradiance and depth data. The Ed (z, λ) profiles were first averaged over 1 m depth intervals, with the new values specified at the interval midpoints. Numerical differentiation was then performed using the central finite difference method. The final step was to smooth the differentiation result using a 10 m moving average filter. The estimated Kd (z, λ) profiles for both wavelengths, along with vertical lines indicating pure water absorption coefficients aw (λ) [24], are shown in Fig. 7(b). Whereas the Kd coefficients for both wavelengths are fairly constant within the investigated depth range, the magnitude of Kd (z, 472) is significantly higher than aw (472) and Kd (z, 562) is comparable to aw (562). This suggests considerable contributions of inelastic radiative processes, in particular Raman scattering by water molecules, to the green wavelength light field. The significance of these observations will be further elaborated upon in the discussion. The ratio of blue to green irradiance with depth is shown in Fig. 7(c). The ratio slope indicates that the quantity of blue, relative to green, irradiance increased by ∼0.5 orders of magnitude over a 70 m depth range.Optical signal-to-noise ratio (OSNR), defined as the ratio of signal mean to signal standard deviation for a Poisson process, was calculated from raw data (photon count rate) to avoid influence of the nonlinearity correction process on the statistics. Following subtraction of estimates of the dark noise rate, fine depth intervals of 10 cm were used to calculate mean, standard deviation, variance, and OSNR. Profiles of linear OSNR with depth are shown in Fig. 7(d). Higher variance observed in the OSNR estimation at shallower depths is related to the previously described fluctuations present in the Ed (z, λ) profiles.
For comparison, profiles from the Atlantic station east of Guadeloupe are shown for the depth range 220 m to 360 m in Fig. 8. In this case, the maximum available measurement depth was greater than at the Pacific station, as expected for the comparatively clear waters of this Atlantic region. Where reasonable, the axes limits were held the same as in Fig. 7. In this case, the Kd (z, λ) profiles indicate a gradual decrease in green attenuation over the initial 60 m depth, until an approximately constant value is reached. Importantly, Kd (z, 562) values are significantly smaller than aw (562) throughout the investigated range of mesopelagic depths. In contrast, this is not the case in the blue. These results again indicate a strong contribution of inelastic radiative processes to the mesopelagic light in the green spectral region. The irradiance ratio was initially similar to the previous measurement, but quickly changed to a rate of increase less than ∼0.3 orders of magnitude over a 140 m depth range. The effect of the observed transition is apparent in both the irradiance ratio and Kd (z, λ) profiles.
Fig. 8 Irradiance depth profiles for mesopelagic light meter system deployed in the western Atlantic, east of Guadeloupe. (a) Log irradiance. (b) Diffuse attenuation functions. (c) Log ratio of blue and green irradiance profiles. (d) Optical signal-to-noise ratio calculated over 10 cm depth intervals.
3.3. Bioluminescent flash kinetics
The MLM system was deployed in the San Clemente Basin down to 450 m depth under clear moonlit sky and calm sea state. The two acrylic ND filters remained attached to the blue-filtered radiometer. The detector gate times were set to 1 ms to increase temporal resolution in the measurements, such that fine detail in the bioluminescent flashes might be elucidated. Previous field studies in this particular oceanic region [17, 25] indicate that the bioluminescent copepod Gaussia princeps may be present at the specified measurement depths. However, no biological samples were collected here. It is also important to note that a complete study of in situ bioluminescence should examine various instrument configurations, such as horizontal or downward viewing sensors, to reduce possible surface effects.
The acquired light data was processed as previously described for downward irradiance measurements. To enable straightforward comparison with results from previous studies of bioluminescence, measured power was not converted to Watts from photon counts, such that estimated irradiance has units of photons/s/cm2 (cps/cm2). The time series profile was low-pass filtered and resampled prior to extraction of individual flashes. Example segments of the bioluminescence time series are shown in Fig. 9. Maximum flash response values range from 6.5 × 109 to 3.4 × 1010 cps/cm2. Measured flash durations range from 100 to 700 ms.
Fig. 9 Example segments from irradiance time series of in situ bioluminescence observed at depth under a moonlit sky. Subfigures are sorted in order of descending peak magnitude. Noise level represents limitations of the system as configured.
4. Discussion
A unique set of instruments has been developed for measuring light in the deep ocean. The most notable components of the mesopelagic light meter (MLM) system are the two custom low-light radiometers. The radiometers were designed to improve upon previous low-light instrumentation [12, 13, 18] by selection of solid-state silicon photomultiplier (SPM) devices as detectors. The use of thermoelectrically-cooled SPMs yields high light sensitivity (low dark count rate) with measurement gate times as low as 100 μs. In addition, engineered polymer diffusers were used to allow for maximum throughput while maintaining a cosine-corrected (Lambertian) response for transmitted light. The radiometers are autonomously controlled via internal computers and powered by a custom rechargeable battery pack. All components are contained in compact aluminum housings, capable of deployment to at least 1000 m depth. These qualities are ideal for the study of mesopelagic optical properties, e.g., measurement of downward irradiance and bioluminescent flashes.
A selected gate time of 2 ms allowed for excellent depth resolution of the available Ed (z, λ) profiles, as shown in Figs. 7(a) and 8(a). Periodic fluctuations about the mean irradiance were found to decrease in magnitude with depth. The cause of these oscillations has not yet been determined. However, the estimated periods of 5 s to 10 s are within the range of those for wind waves. It therefore seems reasonable to consider that the oscillations in measured irradiance may be correlated to variation of the sea state, i.e., surface waves. Another possible explanation for the observed fluctuations is that ship movement, due to surface waves, was coupled to the MLM system via the winch wire.
Examination of Ed (z, λ) profiles from Figs. 7(a) and 8(a) indicates a significant amount of green light present at mesopelagic depths. Specifically, downward irradiance at the green wavelength was observed at a magnitude on the order of only 103–104 less than at the blue wavelength. More importantly, the observed decrease in Kd (z, 562) with depth, shown in Fig. 8(b) for the Atlantic station, is consistent with Raman scattering effects at mesopelagic depths [9]. Such a transition is not observed in the data from the Pacific station, shown in Fig. 7(b), but may have occurred at shallower unmeasured depths. It is particularly important to note that the observed nearly-asymptotic value for Kd (z, 562) at the Atlantic station is less than the absorption coefficient for pure water near this wavelength, i.e., aw (562.5) = 0.0640 m−1 [24]. This result indicates that the rate of signal loss with depth is much lower than would be expected even for pure water. No known wavelength-dependent mechanism can explain this observation, other than the energy transfer described by Raman scattering, which acts to increase the amount of green light present in the water column. In contrast, Kd (z, 472), over all measured depths, is greater than the absorption coefficient for pure water near this wavelength, i.e., aw (472.5) = 0.0109 m−1 [24], since inelastic effects are insignificant in this blue spectral region.
Prior to OSNR calculations, the irradiance profiles were averaged over 10 cm depth intervals, which corresponds to a time average of 0.4 s. Mean and standard deviation values were used to calculate the OSNR as a function of depth. Examination of the OSNR profiles from Figs. 7(d) and 8(d) indicates a square-root decrease in OSNR with depth over ranges where the measurement was limited by signal shot noise. As depths were reached where the available light levels neared the detector dark count rate, the measurements became limited by dark noise, and OSNR approached unity. The radiometer noise characteristics are therefore consistent with Poisson statistics.
Given the specified noise irradiance levels of the instrument, as shown in Table 2, along with knowledge of the underwater light field [9], the maximum measurement depth can be estimated. For very clear surface waters with 0.02 mg Chl/m3 and clear sky with sun at 30° zenith, the expected irradiance just below the surface is on the order of 102 μW/cm2/nm for both (blue and green) system wavelengths. Note that the available irradiance would be slightly higher at exact noon. At 400 m depth, irradiance decreases to 10−2 μW/cm2/nm (blue) and 10−5 μW/cm2/nm (green). At 800 m depth, irradiance is reduced by another 5 orders of magnitude for both wavelengths. Under optimal conditions, it is therefore reasonable to expect maximum measurement depths of nearly 800 m (blue) and 600 m (green). At night under a full moon, the available irradiance levels are reduced by approximately 6 orders of magnitude [26]. In this case, the maximum measurement depths would be decreased to 300 m (blue) and 150 m (green). It is important to consider that reduced clarity of surface waters will greatly increase light attenuation, therefore decreasing the maximum measurement depths.
Results from the night profile exhibit several bioluminescent flashes, as shown in Fig. 9. Measured flash durations are an order of magnitude shorter than those observed in laboratory measurements of electrically-stimulated bioluminescence [16, 17]. Ideally, statistical calculations of maximum response, total quantum emission, rise time, and duration would be performed for the full set of flashes. However, there was not a large enough number of observed flashes to enable meaningful statistics. It is reasonable to assume that, in some of the segments, the complete flash is not being captured and is below the noise level of the instrument. Therefore, radiometer sensitivity should be increased by removing at least one of the acrylic ND filters, for the blue channel, prior to performing a night profile. Although it is difficult to make strong scientific conclusions with the present data set, the results clearly show the ability of the detectors to measure flashes of in situ bioluminescence. Acquisition of bioluminescence with high time resolution is important for many applications, including the taxonomic classification of zooplankton collected during the Arctic polar night [27], as well as exploration of the various ecological functions of bioluminescence, including defense, counterillumination, prey attraction, intraspecific communication, and so on [28].
Some improvements could be made to the current system. The most apparent difficulty, experienced during calibrations and field studies, was the inability to dynamically change the amount of light incident on the detector. Although the acrylic ND filters were designed to be easily attachable in the field, the number of filters used was driven by assumptions about the ambient light field [9], which are not accurate under all environmental conditions. The SPM detectors are not susceptible to being damaged by high irradiance at the surface, another advantage over PMTs, but saturation prevented near surface measurements. The addition of an electrically-addressable aperture in the optical path would maximize system dynamic range and allow for acquisition of irradiance throughout the water column, from surface to noise-equivalent depth. The use of sapphire, instead of acrylic, would reduce the optical port thickness, therefore maximizing the angle of incident light available to the diffuser.
A slightly larger housing for the radiometers would allow for inclusion of a filter wheel, or perhaps even a diffraction grating, which would increase the number of spectral bands measured. Using more narrowband filters would reduce system throughput. However, SPMs with larger active area are available, which would allow for the collection of more light (increased FOV). Unfortunately, dark count rate also increases with detector area. Ultimately, to improve maximum measurement depth, the ideal solution would be to use SPM arrays, which consist of multiple SPMs. Such devices are becoming available and have been recently implemented in a laser range-finding system [29]. Using an array of SPMs in the present application would increase the effective detection area and significantly improve the ability to measure lower light levels, yielding deeper measurement depths and higher spectral resolution. For radiance measurements, where acquisition of light incident from discrete solid angles is required, a single photon imaging detector [30] may be suitable, although it is particularly important to consider the trade between active pixel area and light available to the sensor in this case.
A primary goal of this study was to demonstrate the capability of SPM technology in deep-sea light applications. This includes exploring the ability to measure radiometric quantities, such as downward irradiance, as well as the peak intensity and pulse widths of bioluminescent flashes. Each of these areas has particular requirements that would drive the overall design of an instrument constructed specifically to examine a particular problem. At present, it is clear that spectrally-limited, high temporal resolution measurements of deep-sea light can be reliably acquired using systems with SPM detectors. It is expected that future development of radiometric and other non-imaging instrumentation for underwater applications will rapidly transition to using this new technology.
Acknowledgments
The authors would like to thank Dr. R. Reynolds (SIO/UCSD), Prof. S. Johnsen (Duke U.), and Prof. J. H. Cohen (U. Delaware) for helpful discussions, M. J. Bianco (SIO/UCSD) and C. Briseño-Avena (SIO/UCSD) for assistance with data collection, and the captain and crew of the R/V New Horizon (SIO), and the captain and crew of the R/V Atlantic Explorer (BIOS). This research was supported by ONR MURI grant number N00014-09-1-1053.
References and links
1. F. E. Nicodemus, J. C. Richmond, J. J. Hsia, I. W. Ginsberg, and T. Limperis, Geometrical Considerations and Nomenclature for Reflectance (National Bureau of Standards (U.S.), 1977), Monograph 160.
2. E. J. Denton, “On the organization of reflecting surfaces in some marine animals,” Philos. T. R. Soc. B 258, 285–313 (1970). [CrossRef]
3. E. J. Denton and M. F. Land, “Mechanism of reflexion in silvery layers of fish and cephalopods,” P. R. Soc. B 178, 43–61 (1971). [CrossRef]
4. S. Johnsen, “Cryptic and conspicuous coloration in the pelagic environment,” P. R. Soc. B 269, 243–256 (2002). [CrossRef]
5. S. Johnsen, “Lifting the cloak of invisibility: The effects of changing optical conditions on pelagic crypsis,” Integr. Comp. Biol. 43, 580–590 (2003). [CrossRef] [PubMed]
6. S. Johnsen and H. M. Sosik, “Cryptic coloration and mirrored sides as camouflage strategies in near-surface pelagic habitats: Implications for foraging and predator avoidance,” Limnol. Oceanogr. 48, 1277–1288 (2003). [CrossRef]
7. A. L. Holt, A. M. Sweeney, S. Johnsen, and D. E. Morse, “A highly distributed Bragg stack with unique geometry provides effective camouflage for Loliginid squid eyes,” J. R. Soc. Interface 8, 1386–1399 (2011). [CrossRef] [PubMed]
8. J. M. Haag, J. S. Jaffe, and A. M. Sweeney, “Measurement system for marine animal reflectance functions,” Opt. Express 21, 3603–3616 (2013). [CrossRef] [PubMed]
9. L. Li, D. Stramski, and R. A. Reynolds, “Characterization of the solar light field within the ocean mesopelagic zone based on radiative transfer simulations,” Deep Sea Res., Part I 87, 53–69 (2014). [CrossRef]
10. E. J. Warrant and N. A. Locket, “Vision in the deep sea,” Biol. Rev. 79, 671–712 (2004). [CrossRef] [PubMed]
11. G. L. Clarke and G. K. Wertheim, “Measurements of illumination at great depths and at night in the Atlantic Ocean by means of a new bathyphotometer,” Deep Sea Res. 3, 189–205 (1956). [CrossRef]
12. B. P. Boden, E. M. Kampa, and J. M. Snodgrass, “Underwater daylight measurements in the Bay of Biscay,” J. Mar. Biol. Assoc. U. K. 39, 227–238 (1960). [CrossRef]
13. E. M. Kampa, “Underwater daylight and moonlight measurements in the eastern North Atlantic,” J. Mar. Biol. Assoc. U. K. 50, 397–420 (1970). [CrossRef]
14. S. Johnsen, “The red and the black: bioluminescence and the color of animals in the deep sea,” Integr. Comp. Biol. 45, 234–246 (2005). [CrossRef] [PubMed]
15. D. E. Nilsson, E J. Warrant, S. Johnsen, R. Hanlon, and N. Shashar, “A unique advantage for giant eyes in giant squid,” Curr. Biol. 22, 683–688 (2012). [CrossRef] [PubMed]
16. M. I. Latz, T. M. Frank, M. R. Bowlby, E. A. Widder, and J. F. Case, “Variability in flash characteristics of a bioluminescent copepod,” Biol. Bull. 173, 489–503 (1987). [CrossRef]
17. M. R. Bowlby and J. F. Case, “Flash kinetics and spatial patterns of bioluminescence in the copepod Gaussia princeps,” Mar. Biol. 110, 329–336 (1991). [CrossRef]
18. E. A. Widder, F. M. Caimi, L. D. Taylor, and R. Tusting, “Design and development of an autocalibrating radiometer for deep sea biooptical studies,” in Proceedings of OCEANS ‘92, Mastering the Oceans Through Technology. (IEEE, 1992), pp. 525–530. [CrossRef]
19. E. A. Widder and S. Johnsen, “3D spatial point patterns of bioluminescent plankton: A map of the ‘minefield’,” ,J. Plankton Res. 22, 409–420 (2000). [CrossRef]
20. C. M. Herren, S. H. D. Haddock, C. Johnson, C. M. Orrico, M. A. Moline, and J. F. Case, “A multi-platform bathyphotometer for fine-scale, coastal bioluminescence research,” Limnol. Oceanogr.: Methods 3, 247–262 (2005). [CrossRef]
21. C. D. Mobley, Light and Water: Radiative Transfer in Natural Waters (Academic, 1994).
22. K. J. Kasunic, Optical Systems Engineering (McGraw-Hill, 2011).
23. P. Buzhan, B. Dolgoshein, L. Filatov, A. Ilyin, V. Kantzerov, V. Kaplin, A. Karakash, F. Kayumov, S. Klemin, E. Popova, and S. Smirnov, “Silicon photomultiplier and its possible applications,” Nucl. Instrum. Methods Phys. Res., Sect. A 504, 48–52 (2003). [CrossRef]
24. R. M. Pope and E. S. Fry, “Absorption spectrum (380–700 nm) of pure water. II. Integrating cavity measurements,” Appl. Opt. 36, 8710–8723 (1997). [CrossRef]
25. J. J. Childress, “Effects of pressure, temperature and oxygen on the oxygen-consumption rate of the midwater copepod Gaussia princeps,” Mar. Biol. 39, 19–24 (1977). [CrossRef]
26. E. Warrant, “Vision in the dimmest habitats on Earth,” J. Comp. Physiol. A 190, 765–789 (2004). [CrossRef]
27. G. Johnsen, M. Candeloro, J. Berge, and M. Moline, “Glowing in the dark: Discriminating patterns of bioluminescence from different taxa during the Arctic polar night,” Polar Biol. 37, 707–713 (2014). [CrossRef]
28. S. H. D. Haddock, M. A. Moline, and J. F. Case, “Bioluminescence in the sea,” Annu. Rev. Mar. Sci. 2, 443–493 (2010). [CrossRef]
29. K. T. Son and C. C. Lee, “Multiple-target laser rangefinding receiver using a silicon photomultiplier array,” IEEE Trans. Instrum. Meas. 59, 3005–3011 (2010). [CrossRef]
30. F. Guerrieri, S. Tisa, A. Tosi, and F. Zappa, “Two-dimensional SPAD imaging camera for photon counting,” IEEE Photon. J. 2, 759–774 (2010). [CrossRef]
