POLDER and Ocean Color
Anne Lifermann (Anne.Lifermann@cnes.fr)
CNES, POLDER Project Scientist
Pierre-Yves Deschamps (firstname.lastname@example.org), Jean-Marc Nicolas (Nicolas@loaser.univ-lille1.fr), CNRS, Laboratoire d’Optique Atmosphérique
François-Marie Bréon (email@example.com), Cyril Moulin (firstname.lastname@example.org), CEA, Laboratoire des Sciences du Climat et de l’Environnement
Annick Bricaud (email@example.com)
CNRS, Laboratoire de Physique et Chimie Marine
POLDER (POLarization and Directionality of the Earth’s Reflectances) is a multispectral imaging radiometer providing unique measurements of the anisotropy and polarization of the solar radiation reflected by the Earth-atmosphere system. POLDER was developed by the Centre National d’Etudes Spatiales (Toulouse, France) and was flown on board NASDA’s ADEOS-1 platform between August 1996 and June 1997. POLDER is also part of the ADEOS-2 payload, to be launched in 2000.
POLDER is a multidisciplinary programme addressing four aspects of earth observation: (i) aerosols, (ii) clouds and water vapor, (iii) ocean color, and (iv) land surfaces. With only three bands dedicated to ocean color (443, 490, and 565 nm, and 670, 765 and 865 nm for atmospheric correction) and a 6 x 7 km resolution, POLDER is often seen as an outsider for ocean color compared to other sensors such as OCTS (also on ADEOS) or SeaWiFS. Clearly, coastal and real-time applications are outside the scope of POLDER. Still, POLDER has specific advantages and is able to provide a state of the art global ocean-color data set for studying the dynamics of ocean biology and ocean primary production.
POLDER Measurement Concept
POLDER is a camera with a two dimensional CCD detector array, a wide field of view telecentric optics and a rotating wheel carrying spectral and polarized filters. A series of images (one per filter) is acquired every 20 seconds. The wide bi-dimensional field of view of the sensor (2400 x 1800 km) allows the observation of a single target for 14 successive images during the same orbit pass. The target can be viewed under different geometries from adjacent orbits of the 41 day ADEOS cycle, thus allowing for a unique comprehensive sampling of the target Bi-directional Reflectance Distribution Function.
The wide field of view provides a daily quasi-global coverage. Moreover, POLDER’s multiple viewing capability allows observation of the glitter pattern for calibration without loss of coverage for ocean color.
An Alternative Approach for Calibration
Calibration is recognized as a critical issue for analysis of long-term trends and is of paramount importance for ocean color. An accuracy of 2-3% is necessary for the absolute calibration parameters. This requirement could not be fulfilled with an on-board calibration device (up to now, no remote sensing instruments have flown with a completely reliable onboard calibration device). Also, the calibration of POLDER must account for the bi-dimensional wide field of view and the polarized light. For these reasons, the POLDER calibration team have developed an alternative approach, based on natural targets with known absolute, spectral or directional reflectance characteristics.
There are two components to the POLDER calibration procedure:
(i) a pre-flight component consisting of intensive measurement campaigns with ad hoc devices (integrating sphere, polarizing system and optical references generator) resulting in a comprehensive characterization of the sensor (both radiometric and geometric) before launch and,
(ii) an in-flight calibration procedure, which is a combination of vicarious methods using a variety of natural targets such as sunglint, molecular scattering, desert sites and clouds. Basically, for the absolute calibration, the shorter wavelength channels are calibrated with respect to molecular scattering; the glitter is then used to transfer the calibration to longer wavelengths. Other methods allow cross checking of the results.
An operational calibration processing unit has been developed for POLDER. Results to date show very good agreement between all methods for the spectral bands 565 to 910 nm, with an estimated accuracy of better than 3%. At 490 nm, the performance is believed to be better than 4%, and at 443 nm better than 5%. An inconsistency has been found in the 443 nm calibration, with a significant disagreement between results obtained from the molecular scattering method and that from clouds. To be consistent with the atmospheric correction algorithm, ocean color estimates are obtained using the values of the molecular scattering method. This point, which deserves more research, may stem from the present limitations in the understanding of radiative transfer.
The POLDER project has invested a significant amount of time and effort (more than 10 man-years) in developing new methods for operational, vicarious calibration of sensors. The skills developed will be tested on SeaWiFS and are ready for upcoming missions such as EOS, ENVISAT and obviously POLDER on ADEOS-2.
Atmospheric Correction and Chlorophyll Concentration Estimate
POLDER measurements are initially corrected for atmospheric absorption. The water vapor absorption correction makes use of the 910 nm channel which is centered on a water vapor absorption band, while the oxygen absorption correction makes use of the absorption band centered around 763 nm. The ozone absorption correction makes use of stratospheric ozone concentration estimate from TOMS, also onboard ADEOS.
Next, a chlorophyll concentration is initialized to obtain an estimate of the water reflectance at 670 nm (assumed to be zero at 765 nm and 865 nm). Aerosol reflectance is then derived from the longer wavelength measurements after correction for the molecular, water, glitter and foam contributions. This step makes use of pre-computed reflectance as a function of viewing geometry and surface wind speed. Finally, an aerosol optical thickness and its spectral variation is derived from the estimated aerosol contribution, making use of pre-computed radiances for a set of optical thicknesses and for 12 aerosol models. The "best" model is selected from the available directions (only those which are not affected by the direct glitter are used), by averaging the measured spectral signature weighted by the aerosol radiance.
Once a model and an optical thickness have been derived, the aerosol contribution is extrapolated to the lower wavelengths using pre-computed radiance and transmission tables. The reflectance measurements are corrected for molecular, foam, glitter, and aerosol contributions to derive an estimate of the water reflectance.
Radiative transfer simulations as well as observations have shown directional effects on the water column reflectance. The results of these simulations are used to correct POLDER estimates of the water reflectance for directional effects. After this correction, the various POLDER estimates (for a given pixel) are averaged (there is no attempt at this stage to derive information from the directional signature).
An estimate of the chlorophyll concentration can then be derived from the water reflectance at 443, 490 and 565 nm. The classical algorithm is based on either the 443/565 or 490/565 reflectance ratio, depending on the concentration. An alternative is to use a new index proposed by Robert Frouin (Scripps Institution for Oceanography), the Normalized Difference Pigment Index :
NDPI = (R443-R565)/R490
which seems to be less sensitive to residual noise on the reflectances after atmospheric corrections. Depending on the differences between the initial and the retrieved values of the optical thickness and the pigment concentration, the algorithm described above can be iterated with new values.
Ocean Color Validation
From the beginning, the POLDER project has made a clear distinction between calibration and validation activities. The ocean color validation itself is a stepwise approach including:
-validation of the calibration
-validation of marine reflectances
-validation of marine parameters (pigment concentration and water type)
The availability of coincident in-situ measurements is a major difficulty of the validation procedure. National opportunities as well as cooperation with the US have been used to compile two separate validation data sets: one of in-situ optical measurements and one of pigment concentrations.
A dedicated instrumentation, SIMBAD, has been developed by LOA for at sea en route measurements of aerosol optical thickness and marine reflectances. The expansion of these operations to a network of merchant ships had to be terminated at an early stage after the loss of ADEOS, but the approach has proved to be very useful for the provision of timely and abundant measurements.
The 532 optical measurements resulted in 38 match-up data sets out of which 9 ‘golden cases’ have been selected to reconstruct the top of the atmosphere (TOA) signal and to check the calibration. The other measurements have been used to validate the directional marine reflectances, yielding a relative accuracy of 38% at 443 nm and 31% at 565 nm. This large error in the retrieved marine reflectances requires further investigation to determine the relative contribution of instrument, algorithm and natural bio-optical variability.
The 1138 chlorophyll measurements resulted in 256 match-up data sets in the range 0.3 to 2 mg m-3 leading to a relative accuracy of 58 %, equivalent to a resolution of four classes per decade of pigment concentration (Figure 1).
In summary, a significant effort has been made by the POLDER project for a quantitative assessment of the ocean color products. Further improvements are expected in the context of preparation of refined algorithms for POLDER on ADEOS 2.
POLDER Ocean Color Products
The POLDER data set covers the period November 1, 1996 to June 29, 1997.
Two kinds of Level 2 products are generated per orbit:
- directional parameters with marine reflectances which are useful for validation of directional models or to characterize some species of phytoplankton,
- non-directional parameters with underwater diffuse marine reflectances, water type, and pigment concentration (Figure 2).
In addition, one monthly level 3 product provides temporal syntheses per ten-day period and per month (Figures 2 & 3).
POLDER Comprehensive Data Set
To complement marine pigment concentrations, POLDER also provides the first vegetation index corrected for directional effects, thus allowing for a monitoring of the global biosphere. Moreover, other innovating parameters such as aerosol type over the oceans, cloud properties including cloud phase and bi-directional functions (BRDF) of land and clouds are retrieved at the global scale from POLDER measurements. POLDER thus provides a comprehensive set of key geophysical parameters for climate research and environmental monitoring applications.
POLDER Data Plan and Distribution
POLDER daily measurements are processed operationally and delivered by the POLDER Processing Center (CPP) located in CNES, Toulouse. The systematic processing of the complete archive (over the ocean, land and cloud) will start in September 1998 and will be finished before the end of 1998. In the frame of the extended ocean color validation phase recommended by the Validation Review panel, the delivery of POLDER ocean color products is limited to Principal Investigators (PI’s) and other contributors. The call for contributors, already made during the validation phase, is renewed. Investigators willing to contribute to the validation work by supplying additional in situ measurements for comparison with satellite measurements, or by their regional expertise in ocean color are invited to make proposals to the project. POLDER data may be available for contributors with the same rules as for PI’s (no charge, PI license agreement stating that publications should be sent to CNES before submission, feedback on the product quality etc.).
POLDER products, as well as POLDER user services and other information are available by contacting the order desk at:
Hagolle, O., P. Goloub, P.Y. Deschamps, H. Cosnefroy, X. Briottet, T. Bailleul, J.M. Nicolas, F. Parol, B. Lafrance and M. Herman (1998). Results of POLDER in-flight calibration, IEEE Transactions on Geoscience and Remote Sensing, submitted.
Fougnie, B., P.Y. Deschamps, R.Frouin (1998). Vicarious calibration of the POLDER ocean color spectral bands using in-situ measurements, IEEE Transactions on Geoscience and Remote Sensing, submitted.
Fougnie, B., R. Frouin, P. Lecomte, and P.Y. Deschamps (1998). Reduction of skylight reflection effects in the above-water measurements of diffuse marine reflectance, JGR-Oceans, submitted.
Thanks are due to the POLDER Ocean Color Team at LOA, LSCE, LPCM for the development and validation of the algorithms, to R. Frouin for his support in collecting in-situ measurements and to all in-situ data providers. The results presented in this paper were obtained using data from CNES’s POLDER instrument onboard NASDA’s ADEOS satellite.
Figure 1 : Validation of POLDER estimates of pigment concentration against coincident in-situ measurements. The two red lines indicate a factor of two between the satellite estimate and the in-situ measurement.
Figure 2 : Illustration of POLDER Ocean Color products: from daily (Level 2) to 10-day and monthly syntheses (Level 3) over the Mediterranean basin. Image courtesy of LOA/ LPCM/ LSCE/ CNES/ NASDA.
Figure 3: POLDER global chlorophyll concentration (June 1997). Image courtesy of LOA/ LPCM/ LSCE/ CNES/ NASDA.