Abstract

Satellite remote sensing of ocean color has the potential to map global surface phytoplankton concentrations at rates of up to once per day, providing wide-area data on a number of fundamental ocean processes relating to biological production, air–sea exchange of biogenic greenhouse gases, climate change, and living resources. There remain a number of problems in the technique, including the possible decline of satellite-sensor sensitivity over time and the difficulty of correcting for light detected by the satellite sensor that originated from the atmosphere. To overcome these problems, the new generation of ocean-color sensors must be coupled with an extensive calibration and validation program. In this study, the authors report on progress to develop a methodology to measure water-leaving radiance and incident irradiance from a moored buoy in order to effect vicarious calibration and validation of ocean-color satellite data at a rate of up to twice daily. The Plymouth Marine Bio-Optical Data Buoy, which became operational on 19 April 1997, is assessed against rigorous specifications for surface and in-water radiometers and is shown to be capable of making high-quality optical measurements for a fraction of the cost associated with other calibration–validation projects.

1. Introduction

The first space-based ocean-color sensor, the Coastal Zone Color Scanner (CZCS), was a proof-of-concept mission with a planned lifetime of 1 year. It was nonoperational (data were acquired irregularly) but continued with intermittent operation until 1986, 7.5 years after launch. The spatially extensive data on near-surface phytoplankton distributions led to remarkable advances in our understanding of a number of fundamental oceanic processes, and proved the potential of ocean-color remote sensing. Abbot and Chelton (1991) give a review of CZCS literature, and Aiken et al. (1992) summarize the relevance of ocean color to climate-change research. The process of satellite ocean-color remote sensing is as follows.

  1. Light at the top of the atmosphere is measured by optical sensors onboard a satellite.

  2. This signal includes light leaving the sea (water-leaving radiance, Lw) and light that has been scattered into the sensor by molecules and aerosols in the intervening atmosphere. The signal arising from scattering in the atmosphere must be removed to give Lw in a process called “atmospheric correction.” Water-leaving radiance is then converted to normalized water-leaving radiance, Lwn, a quantity that is largely independent of the time of day, date, latitude, and atmospheric transmittance when the measurement was made. This quantity was first defined by Gordon and Clark (1981), and an equivalent definition is given in Mueller and Austin (1995), as (1): 
    formula
    where Lw is the water-leaving radiance (upwelling nadir radiance just above the sea surface) in μW cm−2 sr−1 nm−1; Es is the downwelling irradiance (in μW cm−2 nm−1) just above the sea surface at the same instant of time as Lw was measured; and Fo is the mean extraterrestrial solar irradiance (Neckel and Labs 1984) converted to the same irradiance units. Some atmospheric correction procedures produce Lwn directly rather than estimating Lw and Es separately.
  3. Bio-optical algorithms are used to retrieve the most accurate values of bio-optical parameters (such as the surface concentration of phytoplankton) from Lwn measured at different wavelengths since Lwn depends significantly on the concentrations of material in the water.

Studies to develop algorithms for step 3, above, are well advanced and will not be considered here. As most bio-algorithms rely on the satellite accurately measuring Lwn, errors in steps (1) and (2) will lead to errors in surface phytoplankton concentrations. Errors in measuring Lwn arise from two main sources. Only when these two problems have been addressed can we expect to have measurements of water-leaving radiance from space to an accuracy of ±1% relative and ±5% absolute, and thence, estimates of phytoplankton concentration to within 35%, which is the aim of the Sea-viewing Wide Field-of-view Sensor (SeaWiFS), (Hooker et al. 1992). First, optical sensors in space tend to lose sensitivity with time: it was estimated that sensor degradation of between 10% and 40% had occurred by the end of the CZCS mission (Gordon et al. 1983). The optical sensors used in the new generation of ocean-color sensors may also degrade over time, and despite using onboard lamps and periodically viewing the sun (via plaques) and moon, calibration uncertainties could remain. Second, as up to 95% of the signal measured at the satellite originates from the atmosphere, imperfect atmospheric correction leads to significant error into the measurements of Lwn (Gordon 1993). The use of near-infrared bands on the new ocean-color satellites will allow atmospheric correction schemes to more accurately estimate scattering from atmospheric aerosols and multiple-scattering effects. These schemes must be validated in order to confirm that they are performing as expected and to derive improved methods.

a. Vicarious calibration

End-to-end validation of Lwn can be effected by measuring Lw(λ) and Es(λ) for a given pixel using in situ instrumentation, converting this to Lwn(λ) [(1)], and comparing this with Lwn(λ) for the same pixel, derived by atmospherically correcting the signal measured by the satellite sensor. If the calibrations of the sensors onboard the spacecraft are assumed to be good, this procedure will give a check on the atmospheric correction scheme. Further, if the absorption and scattering parameters of the atmosphere are measured simultaneously (for example, using a sun photometer), accurate atmospheric correction can be achieved using multiple-scattering modeling techniques. The calibration of the satellite sensors themselves can then be validated to within (but no better than) the accuracy of the in situ measurements.

2. Methods

a. The moored buoy approach

The rationale for using a moored buoy for making in situ measurements of Lwn for vicarious calibration–validation is as follows. Every time the satellite overflies the buoy site, if there are no clouds, a pair of observations will be generated: that is, a measurement of Lwn made at sea level by the buoy and an estimate of Lwn, at the same location and at the same time, produced from data measured by the satellite. Regular daily measurement of Lwn from a buoy allows calibration and validation of the satellite data at the maximum possible rate. For SeaWiFS, this is at a rate of up to two per day at U.K. latitudes. This rate is greater than can be achieved by aircraft or boat-based calibration experiments that are expensive and experience difficulties making concurrent measurements with the satellite overpass, a factor that is critical if the phytoplankton is patchy. Optical measurements from small buoys will also generally be subject to less self-shading error than measurements from a more robust offshore site, such as a tower. A moored optical data buoy will not allow validation of the satellite-measured Lwn over such a wide variety of optical provinces as a drifting optical buoy but has a number of advantages:

  1. A moored buoy can be expected to return a higher proportion of data to shore because high-capacity telemetry is simpler and because maintenance of the buoy is possible.

  2. The calibration of a moored buoy can be ongoing, with retrieval at certain intervals to check sensor degradation and biofouling, giving it a longer useful lifetime than a drifting buoy.

  3. A moored buoy can carry more sensors (e.g., optical sensors at multiple depths to facilitate the extrapolation of light to the surface) because a higher data transmission rate to shore is possible and because the short lifetime and eventual loss of a drifting buoy means expensive sensors may not give enough data to justify their cost.

  4. Simultaneous measurements of atmospheric parameters are possible from a nearby site.

A time series of measurements of Lwn made by a mooring can also be used to investigate if ocean-color observations from a satellite are likely to show a systematic bias. As satellite data are obtained only under one type of condition (cloud free) at one time of day (near local noon), if near-surface chlorophyll concentration varies with cloud cover or time of day, composite satellite data, produced by averaging satellite measurements over a period of time, will be biased with respect to the actual mean. Repeated measurements at a point under all weather conditions over a relatively long period of time (annual scales) can allow such a bias to be identified.

b. Comparability

For the vicarious calibration approach to be effective, we must be sure that the in situ measurements of Lw and Es (and hence Lwn) are truly comparable with those made by the satellite system; note the following requirements.

  1. The Lw and Es measured by the buoy are representative of the area close to buoy. The performance of the moored data buoy in measuring Lw and Es must be assessed against the rigorous SeaWiFS specifications for surface and in-water radiometers, associated biogeochemical measurements, modes of deployment, measurement protocols, and data analysis procedures (Mueller and Austin 1995). The performance of the buoy is assessed theoretically and checked experimentally using a profiling radiometer, as explained below.

  2. The Lw and Es measured in the immediate vicinity of the buoy are representative of the mean values for the pixel containing the buoy (i.e., an area of about 1 km × 1 km for SeaWiFS in Local Area Coverage mode). For Lw, this can be achieved in four ways. 1) Make frequent bio-optical surveys, by boat, around the site to build up an understanding of the scales of variability of the optical properties of the water: for example, using the Undulating Oceanographic Recorder (Aiken and Bellan 1990). 2) Select a site that is assumed to have low variation in optical properties: for example, open case 1 waters with low pigment concentrations. 3) Obtain a large number of corresponding points of in situ and satellite observation measurements and assume that any single measurement of Lwn in the area close to the buoy is different from the mean Lwn for the pixel by a random (i.e., unbiased) amount. 4) Note that sampling from a moored buoy is Eulerian and that tidal currents will translate the spatial variation of bio-optical properties into a temporal variation that can be measured by the buoy.

  3. An Lw(λ) value measured by buoy sensors that have bandwidths of ∼10 nm is comparable with a measurement made by the SeaWiFS sensor centered on λ, which has a bandwidth of ∼20 nm (λ < 700 nm). For calibration–validation purposes, Mueller and Austin (1995) recommend using optical bands that are narrower than SeaWiFS channels and entirely contained within them to reduce skewing errors arising if the absorption of light by material in the water changes rapidly with wavelength. They state that new methods must be developed to reconcile in-water measurements with wider-band SeaWiFS measurements. Such work is ongoing at the Plymouth Marine Laboratory (PML) as part of this research, but will not be discussed here.

c. The Plymouth Marine Bio-Optical Data Buoy (PlyMBODy)

PlyMBODy is a moored optical data buoy for vicarious calibration–validation of ocean-color data developed at PML. It is one of three calibration moorings that will be used for SeaWiFS calibration and validation (McClain et al. 1996), along with the Marine Optical Buoy (MOBY) systems of Clark [National Oceanic and Atmospheric Administration (NOAA) and Moderate Resolution Imaging Spectroradiometer (MODIS) Project] and Broenkow [Moss Landing Marine Laboratory, San Jose State University, San Jose, California). Details of the NOAA–MODIS MOBY can be found in McClain et al. (1992). The design of PlyMBODy is summarized in Fig. 1 and Table 1, and Fig. 2 shows its location.

Fig. 1.

The Plymouth Bio-Optical Data Buoy (PlyMBODy).

Fig. 1.

The Plymouth Bio-Optical Data Buoy (PlyMBODy).

Table 1.

Summary of PlyMBODy design.

Summary of PlyMBODy design.
Summary of PlyMBODy design.
Fig. 2.

The site of the PlyMBODy in the western English Channel.

Fig. 2.

The site of the PlyMBODy in the western English Channel.

The buoy carries a suite of multispectral radiometers and sensors for physical and biological parameters, producing data that are digitized, compressed, and transmitted back to the laboratory over a cellular telephone link. Communication is automatically established with the buoy approximately once every two hours, allowing the sampling strategy of the buoy to be changed throughout the day. The solar panels and reservoir battery system obviate the need to recover the buoy for battery replacement. The major methodological challenges include the following.

  1. Minimization of self-shading due to the buoy itself. The optical sensors and the buoy itself will perturb the underwater light field we are trying to measure despite the fact that the buoy is small and black and the sensors are held on arms more than 1 m away from the body to reduce this. It is necessary to show that measurements of Lw made by the buoy are not significantly affected by self-shading, and this can be achieved by showing that Lw measurements from the buoy agree closely (<5% difference) with Lw measurements made by an optical profiler following SeaWiFS protocols. The results of this test are given in section 3. The yaw of the buoy, measured by an onboard compass, is merged with the modeled sun position to give the position of the sensors relative to the sun. This allows for data to be excluded if it is found that the sensors suffer from excessive self-shading when the optical sensors are behind the buoy. Tests show that this is not necessary.

  2. Acceptable buoy tilt. An adjustable float on the top sensor arm balances the weight of the arms and makes the buoy sit level in the water in calm conditions. Tidal drag and steady winds cause the buoy to lean. An onboard dual-axis inclinometer allows measurements, taken when the sensors are angled off vertical by more than 20°, to be excluded.

  3. Long-term safety. The buoy has been awarded U.K. Department of Transport consent, and PML continues to liaise with local mariners.

  4. Mitigation of biofouling of the optical sensor windows. Throughout the 1997 deployment, the sensors were cleaned by divers once every 7 to 14 days. An automatic cleaning system has been developed for subsequent deployments in 1998 and 1999.

  5. Periodic recalibration of the sensors over time to manage sensitivity drift. The buoy design allows the sensors to be removed and replaced by divers, for periodic recalibration at PML.

d. Buoy optical sensors

The three seven-band optical sensors on the buoy are Satlantic OCI-200 (irradiance) or OCR-200 (radiance) units. The specifications of these sensors together with the data acquisition equipment are given in Tables 2a–d.

Table 2a. Optical sensor spectral specifications from Mueller and Austin (1995) and the PlyMBODy sensors. Nonconformance is shown in bold.

Table 2a. Optical sensor spectral specifications from Mueller and Austin (1995) and the PlyMBODy sensors. Nonconformance is shown in bold.
Table 2a. Optical sensor spectral specifications from Mueller and Austin (1995) and the PlyMBODy sensors. Nonconformance is shown in bold.

The optical sensor system of PlyMBODy conforms very closely to SeaWiFS specifications. The small mismatch in bands (Table 2a) was resolved for the 1998 deployment. The nonconformances to SeaWiFS specifications in Table 2b are not significant, as the buoy sensors have been optimized for use at 50°N, whereas the Mueller and Austin (1995) specifications cover radiometers to be used at all latitudes. The minimum underwater upwelling radiance specifications in Mueller and Austin (1995) relate to sensor profiles through the water and were fixed to ensure that upwelling radiance could be measured to an optical depth of at least 3. As the radiance sensors on the buoy will always be near surface, the minimum underwater upwelling radiance limits can be higher. The departure from true cosine response of the longer wavelength collectors (Table 2c) is not thought to be significant. Table 2d shows that calibration of the PlyMBODy sensors in the PML optical calibration facility conform fully to the SeaWiFS standards.

Table 2b. Optical sensor responsivity specifications from Mueller and Austin (1995) and the PlyMBODy sensors. Nonconformance is shown in bold.

Table 2b. Optical sensor responsivity specifications from Mueller and Austin (1995) and the PlyMBODy sensors. Nonconformance is shown in bold.
Table 2b. Optical sensor responsivity specifications from Mueller and Austin (1995) and the PlyMBODy sensors. Nonconformance is shown in bold.

e. Sun-tracking photometer

As an extension to the buoy work, from the spring of 1997, a sun-tracking photometer system was mounted at Rame Head, Plymouth, close to the optical data buoy. It measured atmospheric correction parameters (ozone and water vapor absorption and aerosol optical thickness) simultaneously with buoy measurements of Lwn to allow accurate atmospheric correction of the satellite data to be carried out. By minimizing errors in Lwn due to poor atmospheric correction, the calibration of the satellite sensors alone can be monitored over time.

3. PlyMBODy test results

a. Incident irradiance test

The buoy was deployed in Plymouth Sound on 11 December 1996, and over the course of 70 min two sets of data were recorded concurrently.

  1. Buoy incident irradiance, buoy tilt, and buoy orientation. A set of 16 readings of all sensor channels at 2 Hz was taken once every 4 min.

  2. Incident downwelling irradiance above the sea surface was measured using a seven-band Biospherical Profiling Reflectance Radiometer (PRR), which was located high on the bow of the research vessel Squilla (hereafter referred to as the “deck cell”). The irradiance sensor of this instrument has optical characteristics comparable to those of the buoy (Table 2) and was calibrated at PML in the same way as the buoy optical sensors.

b. Incident irradiance results

Figure 3 shows the variation in incident irradiance centered on 490 nm as measured by the deck cell and PlyMBODy, after screening both sets of data for excessive tilt angles (taken as greater than 20° for the purpose of this test, which excluded 4% of the data). Figure 4 shows the spectral comparison between the incident irradiance measured by the deck cell and PlyMBODy, and the range of measurements through the test. A summary of the difference in the measurements is given in Table 3.

Fig. 3.

Incident irradiance Es at 490 nm (in μW cm−2 nm−1) from the deck cell and from PlyMBODy.

Fig. 3.

Incident irradiance Es at 490 nm (in μW cm−2 nm−1) from the deck cell and from PlyMBODy.

Fig. 4.

Spectral incident irradiance Es(λ) (in μW cm−2 nm−1) from the deck cell and from PlyMBODy. The vertical bars show the range of the variation over the 70-min test.

Fig. 4.

Spectral incident irradiance Es(λ) (in μW cm−2 nm−1) from the deck cell and from PlyMBODy. The vertical bars show the range of the variation over the 70-min test.

Table 3.

Test of incident irradiance measured by PlyMBODy sensors relative to the PRR deck cell. A positive error implies that the PlyMBODy measurement is too high, and vice versa.

Test of incident irradiance measured by PlyMBODy sensors relative to the PRR deck cell. A positive error implies that the PlyMBODy measurement is too high, and vice versa.
Test of incident irradiance measured by PlyMBODy sensors relative to the PRR deck cell. A positive error implies that the PlyMBODy measurement is too high, and vice versa.

c. Underwater optics test

With the buoy deployed at 50°13.0′N, 04°05.1′W, two sets of data were recorded over 11 min on 17 June 1997.

  1. Buoy underwater optics (upwelling radiance at two depths), buoy tilt, and buoy orientation. Each channel was sampled every 4 s.

  2. Ten profiles of upwelling radiance were measured using the PML–Satlantic optical profiler as detailed in Pinkerton et al. (1999). The optical measurements were made according to SeaWiFS protocols (Mueller and Austin 1995), and these data may be used as a high-quality standard against which to test the buoy. All profiles were measured between 15 and 200 m from the buoy.

d. Processing of underwater data

Following the exclusion of all PlyMBODy upwelling radiance data recorded when the buoy was tilted by more than 20° (less than 1% of the data), the following model [(2)] was applied to optical data from both the buoy and the profiler:

 
Lu(z, λ) = Lu(0, λ)ekLu(λ)z,
(2)

where Lu(z, λ) is the upwelling radiance at nadir at depth z m, Lu(0, λ) is the upwelling radiance at nadir just below the surface (both in μM cm−2 sr−1 nm−1), and kLu(λ) is the diffuse attenuation coefficient of upwelling light (m−1) that is assumed to be constant over the surface 10 m. This model allows a kLu(λ) and Lu(0, λ) value to be calculated from each measurement made by the deep and shallow upwelling radiance sensors on the buoy. The data from the profiler was split into 10 separate profiles, and a kLu value was calculated for each of the bands. These attenuation coefficients were then used to project each measurement of Lu(z, λ) made by the profiler to the surface for comparison with the PlyMBODy measurements made at the same time. Water-leaving radiance Lw(λ) was then calculated for both buoy and profiler data as (3):

 
formula

where ρ is the Fresnel reflectance of the sea surface for normal incidence (usually taken as 2.0%) and n is the refractive index of water with respect to air (usually taken as 1.34). Both ρ and n are taken as constant for this test, though it is noted that ρ varies slightly with sea state [by less than 1% up to a wind speed of 10 m s−1 (Austin 1974)] and n varies by about 1% over the spectral range used. Figure 5 shows the results for Lw(490) over the duration of the test, and Fig. 6 shows the spectral results for Lw(λ).

Fig. 5.

Water-leaving radiance Lw at 490 nm (in μW cm−2 sr−1 nm−1) from the optical profiler and from PlyMBODy.

Fig. 5.

Water-leaving radiance Lw at 490 nm (in μW cm−2 sr−1 nm−1) from the optical profiler and from PlyMBODy.

Fig. 6.

Spectral water-leaving radiance Lw(λ) (in μW cm−2 sr−1 nm−1) from the optical profiler and from PlyMBODy. The vertical bars show the range of the variation over the 11-min test.

Fig. 6.

Spectral water-leaving radiance Lw(λ) (in μW cm−2 sr−1 nm−1) from the optical profiler and from PlyMBODy. The vertical bars show the range of the variation over the 11-min test.

4. Discussion and conclusions

a. Incident irradiance test

The test of the incident irradiance sensor of PlyMBODy was conducted in poor weather conditions (as shown by the low incident irradiance values), with an overcast sky, rain, and mist. The test was conducted in Plymouth Sound, which means that land appeared in the fields of view of both the PlyMBODy sensor and the deck cell, albeit near the horizon where the cosine response means that sensor sensitivity was low. At the time of the test, a float on the top sensor arm was not present, and the mean tilt of the buoy was of the order of 10°. This float is now present, and the mean tilt is generally reduced to less than 5°. There was also a mismatch between the longer-wavelength bands of the two instruments, and this would lead to some difference between the measurements. Despite these factors, the mean error between data from the deck cell mounted on the research vessel and data from the Es sensor of PlyMBODy is always less than 6%, which is extremely encouraging. Further tests are planned to confirm that the performance of the buoy Es has now been improved.

b. Underwater optics test

The agreement between Lw measured by the buoy and by the optical profiler (taken as a SeaWiFS standard) is excellent for 412–620-nm bands (error less than 5%) but poorer for the PlyMBODy 670-nm band (error of 11%). During this test, the sensor arms of the buoy were situated behind the buoy body relative to the sun, and the sky was clear. These figures, hence, represent the case of maximum self-shading by the buoy, and imply that it is not necessary to exclude optical data measured by the buoy under any conditions because of self-shading. Further work in a variety of lighting conditions is required to confirm this.

c. Overall assessment

Data from the two tests indicate that a small and relatively low-cost moored buoy, such as PlyMBODy, has the potential to provide water-leaving radiance data of a quality suitable for the vicarious calibration and validation of remotely sensed ocean-color data from sensors such as SeaWiFS. Such measurement has been thought of as the preserve of extremely high-cost projects, but the results of this research project indicate the feasibility of operating a number of small moorings in a variety of oceanographic regimes, a program that will significantly contribute to precision atmospheric correction and allow ocean-color missions to deliver truly accurate estimates of near-surface phytoplankton concentrations.

d. Future plans

SeaWiFS was slated to be launched in November 1993, but rocket system and spacecraft problems led to a number of delays, and SeaWiFS was launched successfully on 1 August 1997. The first data from SeaWiFS was received on 14 September 1997, and PlyMBODy has been measuring data during satellite overpasses since then. The buoy was recovered in December 1997 to avoid the worst of the winter storms in the western English Channel and was redeployed between April and December 1998. Future work will focus on the following:

  • Merging concurrent buoy data with SeaWiFS data to fulfill the calibration–validation promise.

  • Improving the conformance of the buoy optical sensor bands to the Mueller and Austin (1995) standards.

  • Reducing the frequency at which the buoy sensors need to be cleaned.

  • Carrying out safe, regular, high-quality data collection through 1999.

Table 2c. Optical sensor angular response specifications from Mueller and Austin (1995) and the PlyMBODy sensors. Nonconformance is shown in bold.

Table 2c. Optical sensor angular response specifications from Mueller and Austin (1995) and the PlyMBODy sensors. Nonconformance is shown in bold.
Table 2c. Optical sensor angular response specifications from Mueller and Austin (1995) and the PlyMBODy sensors. Nonconformance is shown in bold.

Table 2d. Optical sensor calibration specifications from Mueller and Austin (1995) and the PlyMBODy sensors.

Table 2d. Optical sensor calibration specifications from Mueller and Austin (1995) and the PlyMBODy sensors.
Table 2d. Optical sensor calibration specifications from Mueller and Austin (1995) and the PlyMBODy sensors.
Table 4.

Test of underwater upwelling radiance measured by PlyMBODy sensors relative to optical profiler. A positive error implies that the PlyMBODy estimate is too high, and vice versa.

Test of underwater upwelling radiance measured by PlyMBODy sensors relative to optical profiler. A positive error implies that the PlyMBODy estimate is too high, and vice versa.
Test of underwater upwelling radiance measured by PlyMBODy sensors relative to optical profiler. A positive error implies that the PlyMBODy estimate is too high, and vice versa.

Acknowledgments

This work forms part of the PML Strategic Research Project 1 and was supported by a Natural Environment Research Council SeaWiFS Exploitation Initiative Special Topic Grant GST/02/836 to M. H. Pinkerton, and by logistic support from the National Aeronautics and Space Administration. This work could not have been carried out without the professionalism and kind assistance of the PML research vessel crews and PML divers.

REFERENCES

REFERENCES
Abbott, M. R., and D. B. Chelton, 1991: Advances in passive remote sensing of the ocean. U.S. National Report to International Union of Geodesy and Geophysics 1897–1990, Contributions in Oceanography, Amer. Geophys. Union, 571–589.
Aiken, J., and I. Bellan, 1990: Optical oceanography: An assessment of towed measurement. Light and Life in the Sea, P. J. Herring, et al., Eds., Cambridge University Press, 39–57.
——, G. F. Moore, and P. M. Holligan, 1992: Remote sensing of ocean biology in relation to global climate change. J. Phycol., 28, 579–590.
Austin, R. W., 1974: The remote sensing of spectral radiance from below the ocean surface. Optical Aspects of Oceanography, N. G. Jerlov and E. S. Nielsen, Eds., Academic Press, 317–344.
Gordon, H. R., 1993: Radiative transfer in the atmosphere for correction of ocean colour remote sensors. Ocean Colour: Theory and Applications in a Decade of CZCS Experience, V. Barale and P. M. Schlittenhardt, Eds., ECSC, EEC, EAEC, 33–77.
——, and D. K. Clark, 1981: Clear water radiances for atmospheric correction of Coastal Zone Color Scanner imagery. Appl. Opt., 20, 4175–4180.
——, J. W. Brown, O. B. Brown, R. H. Evans, and D. K. Clark, 1983: Nimbus-7 CZCS: Reduction of its radiometric sensitivity with time. Appl. Opt., 22(24), 3929–3931.
Hooker, S. B., W. E. Esaias, G. C. Feldman, W. W. Gregg, and C. R. McClain, 1992: An overview of SeaWiFS and ocean colour. NASA Tech. Memo. 104566, Vol. 1, SeaWiFS Tech. Report Series, 24 pp.
McClain, C. R., W. E. Esaias, W. Barnes, B. Guenther, D. Endres, S. B. Hooker, B. G. Mitchell, and R. Barnes, 1992: SeaWiFS calibration and validation plan. NASA Tech. Memo. 104566, Vol. 3, SeaWiFS Tech. Report Series, 41 pp.
——, and Coauthors, 1996: SeaWiFS calibration and validation quality control procedures. NASA Tech. Memo. 104566, Vol. 38, SeaWiFS Tech. Report Series, 64 pp.
Mueller, J. L., and R. W. Austin, 1995: Ocean optics protocols for SeaWiFS validation, revision 1. NASA Tech. Memo. 104566, Vol. 25, SeaWiFS Tech. Report Series, 67 pp.
——, C. Johnson, L. Cromer, S. B. Hooker, J. T. McLean, and S. F. Biggar, 1996: The third SeaWiFS intercalibration round-robin experiment (SIRREX-3), 19–30 September 1994. NASA Tech. Memo. 104566, Vol. 34, SeaWiFS Tech. Report Series, 78 pp.
——, ——, ——, ——, ——, ——, 1999: The fifth SeaWiFS intercalibration round-robin experiment (SIRREX-5). NASA Tech. Memo. 104566, SeaWiFS Tech. Report Series, in press.
Neckel, H., and D. Labs, 1984: The solar radiation between 3,300 and 12,500 Å. Sol. Phys., 90, 205–258.
Pinkerton, M. H., C. C. Trees, J. Aiken, A. J. Bale, G. F. Moore, R. G. Barlow, and D. G. Cummings, 1999: Retrieval of near-surface bio-optical properties of the Arabian Sea from remotely-sensed ocean colour data. Deep-Sea Res., 46, 549–570.

Footnotes

Corresponding author address: Matt H. Pinkerton, Plymouth Marine Laboratory, Prospect Place, Plymouth, Devon PL1 3DH, United Kingdom.