BIO-OPTICS AND BIOGEOCHEMISTRY II

FAST AND ACCURATE RETRIEVAL OF AEROSOL AND MARINE PARAMETERS FOR COASTAL WATERS

Hamre, Børge¹; Stamnes, Jakob J.¹; Li, Wei²; Stamnes, Knut²; Spurr, Rob<sup>3

1</sup>University of Bergen Allegaten 55, Bergen, --, 5007 Bergen, Norway; ²Stevens Institute of Technology, Hoboken, NJ, 07030, United States; ³RT Solutions Inc, Cambridge, MA, 02138, United States

Recently the retrieval of two aerosol parameters and three marine parameters from eight SEAWifs channels was demonstrated [W. Li, K. Stamnes, R. Spurr, and J.J. Stamnes, Simultaneous retrieval of aerosol prperties and ocean properties by optimal estimation: SeaWiFS case studies for the Santa Barbara Channel, Int. J. Remote. Sens., in press (2008)]. The two aerosol parameters are the optical depth at 865 nm and the bimodal fraction of large vs small particles, and the three marine parameters are the chlorophyll content, the CDOM absorption, and the backscattering coefficient. The retrieval was based on forward simulations using a linearized discrete ordinate method for the coupled atmosphere-ocean system and optimal estimation theory to solve the inverse or retrieval problem. Since the forward simulations are time consuming, we demonstrate in this paper that the retrieval speed can be increased by a factor of about one thousand using interpolation by means of a radial basis function neural net. As a result, we obtain accurate retrievals of aerosol and marine parameters for coastal waters at an operational speed.

LONG-TERM GLOBAL OBSERVATIONS OF THE PARTICULATE BACKSCATTERING COEFFICIENT AND PARTICULATE ORGANIC CARBON IN THE FRAMEWORK OF THE GLOBPHY PROJECT

Duforet-Gaurier, Lucile¹; Loisel, Hubert¹; Dessailly, David¹; Nordkvist, Karin¹; Antoine, David<sup>2

1</sup>CNRS UMR 8187 LOG/ULCO 32 avenue Foch, Wimereux, --, 62930, France; ²LOV, Quai de la Darse, BP8 , Villefranche sur Mer, 06238, France

This study is proposed in the framework of the GLOBPHY project dedicated to the analysis of the decadal changes in the global ocean phytoplankton. A consistent time series of chlorophyll concentration, at global scale, was produced from the CZCS era (1979-1983) to the SeaWiFS era (1998-2002) by Antoine et al, 2005. The final data set showed an overall increase of the world ocean average phytoplankton chlorophyll (Chl) by as much as 22%. The general objective of the GLOBPHY project is to confirm and interpret these changes, by looking at the coherence between the phytoplankton changes and those of other physical and biogeochemical parameters. Therefore, we propose to derive, from the marine reflectance, other indicators than the chlorophyll concentration Chl. The geographic distribution and seasonal evolution of the particulate backscattering coefficient bbp is described over the whole ocean for the SeaWiFS and CZCS era. The monitoring of Chl and bbp is analyzed and compared to provide complementary information about the biogeo-chemical status of surface waters. Indeed, while Chl is a good index of the phytoplankton biomass, the particle backscattering coefficient bbp may be used to describe the spatio-temporal variability of the particulate organic carbon (POC) (Loisel et al, 2002). Until now, POC content obtained from space-borne observations are restricted to the surface. In this study, a relationship between the surface POC and the column-integrated POC is established from a statistical analysis of 563 in situ profiles collected in various oceanic areas. This empirical relationship is tested successfully against an independent in situ dataset composed of 197 vertical POC profiles.

ATMOSPHERIC CORRECTION OF MERIS DATA OVER CASE 2 WATERS

Brajard, Julien¹; Santer, Richard²; Thiria, Sylvie<sup>3

1</sup>ULCO LOCEAN, BC100, T45-55, pl. Jussieu, Paris, --, 75 005, France; ²ULCO, MREN av. Foch, Wimereux, --, 62930, France; ³LOCEAN, BC100, T45-55, pl. Jussieu, Paris, --, 75005, France

Radiometers on board satellite measure the solar radiation reflected by both ocean and atmosphere. One difficulty is that the signal is strongly polluted by the contribution of the atmosphere. Standard algorithms make the hypothesis that the contribution of the ocean is null in the near infrared bands. But, for case 2 waters, the contribution of the sediments is no more neglictible, and standards algorithms failed.

We propose here a new methodology, called NeuroVaria, based on a spectral optimization in the near infrared to estimate simultaneously the particles and the atmospheric contribution on the “top of atmosphere” signal. The aerosols are supposed to be non absorbing and to follow a Junge size distribution law. The water contribution of the water in the near infrared is determined by the spectral signature of the particles that is characterized by intensity and a shape parameter. This methodology combines a neural network for modelling the radiative transfer and a variational algorithm for the inversion of the spectra.

NeuroVaria was applied to MERIS data of the Adriatic Sea off the Venice bay from August 2003 to September 2005 where in-situ measurements of the water-leaving reflectance and the aerosol optical thickness are available. We present the comparison between NeuroVaria, the standard MERIS algorithm and in-situ measurements. We show that NeuroVaria performed a better estimation of the inherent optical water properties and improved the atmospheric correction of case 2 waters.

SECOND ORDER VARIABILITY IN SATELLITE OCEAN COLOR: THE ROLE OF NON-ALGAL MATTER AND BACKSCATTERING

Brown, Catherine A.¹; Huot, Yannick¹; Werdell, P. Jeremy²; Gentili, Bernard¹; Claustre, Hervé<sup>1

1</sup>Laboratoire d'Océanographie de Villefranche Quai de la darse, Villefranche sur Mer, --, 06230, France; ²NASA Goddard Space Flight Center, Greenbelt, Maryland, 20771, United States

Empirical algorithms that describe (global) first order relationships between spectral water-leaving radiances and near-surface chlorophyll concentrations, [Chl], are widely applied to satellite ocean color imagery, but do not explain dispersion or anomalies around their mean statistical trends. We develop an empirical approach that removes the first order effect of [Chl] from satellite ocean color imagery, allowing us to observe the influence of other optically significant water-column constituents (those other than [Chl]) and their distributions. We present statistical and modeling analyses to interpret the observed anomalies in terms of their optical sources (i.e. absorption and backscattering). With this information, we examine the global distributions of anomalies in colored detrital and dissolved organic matter (CDM) and backscattering by particles, both of which display significant regional and seasonal trends. We then use the anomalies to develop alternative remote sensing algorithms to estimate the backscattering coefficient of particles and, more tentatively, the absorption of CDM, and compare our results with other published studies. We further show how these anomalies explain dispersion in standard empirical chlorophyll algorithms and how they could be used to correct such approaches.

IMPROVING OCEAN COLOR DATA PRODUCTS USING A PURELY EMPIRICAL APPROACH: REDUCING THE REQUIREMENT FOR RADIOMETRIC CALIBRATION ACCURACY

gregg, watson<sup>1

1</sup>NASA/Goddard Space Flight Center Code 610.1, Greenbelt, MD, 20771, United States

Radiometric calibration is the foundation upon which ocean color remote sensing is built. Quality derived geophysical products, such as chlorophyll, are assumed to be critically dependent upon the quality of the radiometric calibration. Unfortunately, the goals of radiometric calibration are not typically met in global and large-scale regional analyses, and are especially deficient in coastal regions. The consequences of the uncertainty in calibration are very large in terms of global and regional ocean chlorophyll estimates. In fact, stability in global chlorophyll requires calibration uncertainty much greater than the goals, and outside of modern capabilities. Using a purely empirical approach, we show that stable and consistent global chlorophyll values can be achieved over very wide ranges of uncertainty. Furthermore, the approach yields statistically improved comparisons with in situ data, suggesting improved quality. The results suggest that accuracy requirements for radiometric calibration cab be reduced if alternative empirical approaches are used.