Atmospheric correction algorithm for hyperspectral remote sensing of ocean color from space

Existing atmospheric correction algorithms for multichannel remote sensing of ocean color from space were designed for retrieving water-leaving radiances in the visible over clear deep ocean areas and cannot easily be modified for retrievals over turbid coastal waters. We have developed an atmospheric correction algorithm for hyperspectral remote sensing of ocean color with the near-future Coastal Ocean Imaging Spectrometer. The algorithm uses lookup tables generated with a vector radiative transfer code. Aerosol parameters are determined by a spectrum-matching technique that uses channels located at wavelengths longer than 0.86 mum. The aerosol information is extracted back to the visible based on aerosol models during the retrieval of water-leaving radiances. Quite reasonable water-leaving radiances have been obtained when our algorithm was applied to process hyperspectral imaging data acquired with an airborne imaging spectrometer.     

Atmospheric correction of SeaWiFS imagery for turbid coastal and inland waters

The standard SeaWiFS atmospheric correction algorithm, designed for open ocean water, has been extended for use over turbid coastal and inland waters. Failure of the standard algorithm over turbid waters can be attributed to invalid assumptions of zero water-leaving radiance for the near-infrared bands at 765 and 865 nm. In the present study these assumptions are replaced by the assumptions of spatial homogeneity of the 765:865-nm ratios for aerosol reflectance and for water-leaving reflectance. These two ratios are imposed as calibration parameters after inspection of the Rayleigh-corrected reflectance scatterplot. The performance of the new algorithm is demonstrated for imagery of Belgian coastal waters and yields physically realistic water-leaving radiance spectra. A preliminary comparison with in situ radiance spectra for the Dutch Lake Markermeer shows significant improvement over the standard atmospheric correction algorithm. An analysis is made of the sensitivity of results to the choice of calibration parameters, and perspectives for application of the method to other sensors are briefly discussed.     

Atmospheric correction of satellite ocean color imagery: the black pixel assumption

The assumption that values of water-leaving radiance in the near-infrared (NIR) are negligible enable aerosol radiative properties to be easily determined in the correction of satellite ocean color imagery. This is referred to as the black pixel assumption. We examine the implications of the black pixel assumption using a simple bio-optical model for the NIR water-leaving reflectance [rho(w)(lambda(NIR))](N). In productive waters [chlorophyll (Chl) concentration >2 mg m(-3)], estimates of [rho(w)(lambda(NIR))](N) are several orders of magnitude larger than those expected for pure seawater. These large values of [rho(w)(lambda(NIR))](N) result in an overcorrection of atmospheric effects for retrievals of water-leaving reflectance that are most pronounced in the violet and blue spectral region. The overcorrection increases dramatically with Chl, reducing the true water-leaving radiance by roughly 75% when Chl is equal to 5 mg m(-3). Relaxing the black pixel assumption in the correction of Sea-Viewing Wide Field-of-View Sensor (SeaWiFS) satellite ocean color imagery provides significant improvements in Chl and water-leaving reflectance retrievals when Chl values are greater than 2 mg m(-3). Improvements in the present modeling of [rho(w)(lambda(NIR))](N) are considered, particularly for turbid coastal waters. However, this research shows that the effects of nonzero NIR reflectance must be included in the correction of satellite ocean color imagery.     

Evaluation of a reflectance model used in the SeaWiFS ocean color algorithm: implications for chlorophyll concentration retrievals

For the atmospheric correction of ocean-color imagery obtained over Case I waters with the Sea-Viewing Wide Field-of-View Sensor (SeaWiFS) instrument the method currently used to relax the black-pixel assumption in the near infrared (NIR) relies on (1) an approximate model for the nadir NIR remote-sensing reflectance and (2) an assumption that the water-leaving radiance is isotropic over the upward hemisphere. Radiance simulations based on a comprehensive radiative-transfer model for the coupled atmosphere-ocean system and measurements of the nadir remote-sensing reflectance at 670 nm compiled in the SeaWiFS Bio-optical Algorithm Mini-Workshop (SeaBAM) database are used to assess the validity of this method. The results show that (1) it is important to improve the flexibility of the reflectance model to provide more realistic predictions of the nadir NIR water-leaving reflectance for different ocean regions and (2) the isotropic assumption should be avoided in the retrieval of ocean color, if the chlorophyll concentration is larger than approximately 6, 10, and 40 mg m(-3) when the aerosol optical depth is approximately 0.05, 0.1, and 0.3, respectively. Finally, we extend our scope to Case II ocean waters to gain insight and enhance our understanding of the NIR aspects of ocean color. The results show that the isotropic assumption is invalid in a wider range than in Case I waters owing to the enhanced water-leaving reflectance resulting from oceanic sediments in the NIR wavelengths.     

Moderate-resolution imaging spectroradiometer ocean color polarization correction

The polarization correction for the Moderate-Resolution Imaging Spectroradiometer (MODIS) instruments on the Terra and Aqua satellites is described. The focus is on the prelaunch polarization characterization and on the derivation of polarization correction coefficients for the processing of ocean color data. The effect of the polarization correction is demonstrated. The radiances at the top of the atmosphere need to be corrected by as much as 3.2% in the 412 nm band. The effect on the water-leaving radiances can exceed 50%. The polarization correction produces good agreement of the MODIS Aqua water-leaving radiance time series with data from another, independent satellite-based ocean color sensor, the Sea-Viewing Wide Field-of-View Sensor (SeaWiFS).     

SeaWiFS on-orbit gain and detector calibrations: effect on ocean products

The NASA Ocean Biology Processing Group's Calibration and Validation Team has analyzed the mission-long Sea-Viewing Wide Field-of-View Sensor (SeaWiFS) on-orbit gain and detector calibration time series to verify that lunar calibrations, obtained at nonstandard gains and radiance ranges, are valid for Earth data collected at standard gains and typical ocean, cloud, and land radiances. For gain calibrations, a constant voltage injected into the postdetector electronics allows gain ratios to be computed for all four detectors in each band. The on-orbit lunar gain ratio time series show small drifts for the near infrared bands. These drifts are propagated into the ocean color data through the atmospheric correction parameter epsilon, which uses the 765/865 nm band ratio. An anomaly analysis of global mean normalized water-leaving radiances at 510 nm shows a small decrease over the mission, while an analysis of epsilon shows a corresponding increase. The drifts in the lunar time series for the 765 and 865 nm bands were corrected. An analysis of the revised water-leaving radiances at 510 nm shows the drift has been eliminated, while an analysis of epsilon shows a reduced drift. For detector calibrations, solar diffuser observations made by the individual detectors in each band allows the response of the detectors to be monitored separately. The mission-long time series of detector calibration data show that the variations in the response of the individual detectors are less than 0.5% over the mission for all bands except the 865 nm band, where the variations are less than 1%.     

Improving atmospheric correction for highly productive coastal waters using the short wave infrared retrieval algorithm with water-leaving reflectance constraints at 412 nm

The recently developed short wave infrared (SWIR) atmospheric correction algorithm for ocean color retrieval uses long wavelength channels to retrieve atmospheric parameters to avoid bright pixel contamination. However, this retrieval is highly sensitive to errors in the aerosol model, which is magnified by the higher variability of aerosols observed over urban coastal areas. While adding extra regional aerosol models into the retrieval lookup tables would tend to increase retrieval error since these models are hard to distinguish in the IR, we explore the possibility that for highly productive waters with high colored dissolved organic matter, an estimate of the 412 nm channel water-leaving reflectance can be used to constrain the aerosol model retrieval and improve the water-leaving reflectance retrieval. Simulations show that this constraint is particularly useful where aerosol diversity is significant. To assess this algorithm we compare our retrievals with the operational SeaWiFS Data Analysis System (SeaDAS) SWIR and near infrared retrievals using in situ validation data in the Chesapeake Bay and show that, especially for absorbing aerosols, significant improvement is obtained. Further insight is also obtained by the intercomparison of retrieved remote sensing reflectance images at 443 and 551 nm, which demonstrates the removal of anomalous artifacts in the operational SeaDAS retrieval.     

Role of oceanic air bubbles in atmospheric correction of ocean color imagery

Ocean color is the radiance that emanates from the ocean because of scattering by chlorophyll pigments and particles of organic and inorganic origin. Air bubbles in the ocean also scatter light and thus contribute to the water-leaving radiance. This additional water-leaving radiance that is due to oceanic air bubbles could violate the black pixel assumption at near-infrared wavelengths and be attributed to chlorophyll in the visible. Hence, the accuracy of the atmospheric correction required for the retrieval of ocean color from satellite measurements is impaired. A comprehensive radiative transfer code for the coupled atmosphere--ocean system is employed to assess the effect of oceanic air bubbles on atmospheric correction of ocean color imagery. This effect is found to depend on the wavelength-dependent optical properties of oceanic air bubbles as well as atmospheric aerosols.     

Effects of ocean surface reflectance variation with solar elevation on normalized water-leaving radiance

Effects of the ocean surface reflection for solar irradiance on the normalized water-leaving radiance in the visible wavelengths are evaluated and discussed for various conditions of the atmosphere, solar-zenith angles, and wind speeds. The surface reflection effects on water-leaving radiance are simply due to the fact that the radiance that is backscattered out of the water is directly proportional to the downward solar irradiance just beneath the ocean surface. The larger the solar-zenith angle, the less the downward solar irradiance just beneath the ocean surface (i.e., more photons are reflected by the ocean surface), leading to a reduced value of the radiance that is backscattered out of the ocean. For cases of large solar-zenith angles, the effects of surface irradiance reflection need to be accounted for in both the satellite-derived and in situ measured water-leaving radiances.     

Calibration of SeaWiFS. II. Vicarious techniques

We present an overview of the vicarious calibration of the Sea-Viewing Wide Field-of-view Sensor (SeaWiFS). This program has three components: the calibration of the near-infrared bands so that the atmospheric correction algorithm retrieves the optical properties of maritime aerosols in the open ocean; the calibration of the visible bands against in-water measurements from the Marine Optical Buoy (MOBY); and a calibration-verification program that uses comparisons between SeaWiFS retrievals and globally distributed in situ measurements of water-leaving radiances. This paper describes the procedures as implemented for the third reprocessing of the SeaWiFS global mission data set. The uncertainty in the near-infrared vicarious gain is 0.9%. The uncertainties in the visible-band vicarious gains are 0.3%, corresponding to uncertainties in the water-leaving radiances of approximately 3%. The means of the SeaWiFS/in situ matchup ratios for water-leaving radiances are typically within 5% of unity in Case 1 waters, while chlorophyll a ratios are within 1% of unity. SeaWiFS is the first ocean-color mission to use an extensive and ongoing prelaunch and postlaunch calibration program, and the matchup results demonstrate the benefits of a comprehensive approach.     

Atmospheric correction for NO2 absorption in retrieving water-leaving reflectances from the SeaWiFS and MODIS measurements

The absorption by atmospheric nitrogen dioxide (NO2) gas in the visible has been traditionally neglected in the retrieval of oceanic parameters from satellite measurements. Recent measurements of NO2 from spaceborne sensors show that over the Eastern United States the NO2 column amount often exceeds 1 Dobson Unit (approximately 2.69x10(16) molecules/cm2). Our radiative transfer sensitivity calculations show that under high NO2 conditions (approximately 1x10(16) molecules/cm2) the error in top-of-atmosphere (TOA) reflectance in the blue channels of the sea-viewing wide field-of-view sensor (SeaWiFS) and moderate-resolution imaging spectroradiometer (MODIS) sensors is approximately 1%. This translates into approximately 10% error in water-leaving radiance for clear waters and to higher values (>20%) in the coastal areas. We have developed an atmospheric-correction algorithm that allows an accurate retrieval of normalized water-leaving radiances (nLws) in the presence of NO2 in the atmosphere. The application of the algorithm to 52 MODIS scenes over the Chesapeake Bay area show a decrease in the frequency of negative nLw estimates in the 412 nm band and an increase in the value of nLws in the same band. For the particular scene reported in this paper, the mean value of nLws in the 412 nm band increased by 17%, which is significant, because for the MODIS sensor the error in nLws attributable to the digitization error in the observed TOA reflectance over case 2 waters is approximately 2.5%.     

Phytoplankton pigment concentrations in the Middle Atlantic Bight: comparison of ship determinations and CZCS estimates

The processing algorithms used for relating the apparent color of the ocean observed with the Coastal-Zone Color Scanner on Nimbus-7 to the concentration of phytoplankton pigments (principally the pigment responsible for photosynthesis, chlorophyll a) are developed and discussed in detail. These algorithms are applied to the shelf and slope waters of the Middle Atlantic Bight and also to Sargasso Sea waters. In all, four images are examined, and the resulting pigment concentrations are compared to continuous measurements made along ship tracks. The results suggest that over the 0.08-1.5-mg/m3 range the error in the retrieved pigment concentration is of the order of 30-40% for a variety of atmospheric turbidities. In three direct comparisons between ship-measured and satellite-retrieved values of the water-leaving radiance the atmospheric correction algorithm retrieved the water-leaving radiance with an average error of approximately 10%. This atmospheric correction algorithm does not require any surface measurements for its application.     

Oceanic radiance model development and validation: application of airborne active-passive ocean color spectral measurements

It is shown that airborne active-passive (laser-solar) ocean color data can be used to develop and validate oceanic radiance models. The two principal inputs to the oceanic radiance model, chlorophyll pigment and incident solar irradiance, are obtained from a nadir-viewing laser-induced fluorescence spectrometer and a zenith-viewing radiometer, respectively. The computed water-leaving radiances are validated by comparison with the calibrated output of a separate nadir-viewing radiometer subsystem. In the North Atlantic Ocean, the calculated and the observed airborne radiances are found to compare very favorably for the 443-, 520-, and 550-nm wavelengths over an ～ 170-km flight track east of St. John's, Newfoundland. The results further suggest that the semianalytical radiance model of ocean color, the airborne active (laser) fluorescence spectrometer, and the passive (solar) radiometric instrumentation are all remarkably precise.     

Contribution of Raman scattering to water-leaving radiance: a reexamination

We have reexamined the contribution of Raman scattering to the water-leaving radiance in case 1 waters by carrying out radiative transfer simulations that combine the latest reported measurements of the absorption coefficient of pure water with direct measurements of the spectral variation of the Raman-scattering coefficient. The resulting contribution of Raman scattering is then compared with experimental measurements of the water-leaving radiance, and the fractional contribution of radiance produced by Raman scattering to the total radiance measured at a given wavelength is determined. The results show that (1) the contribution of Raman scattering to the water-leaving radiance in an ocean of pure seawater is as much as 50-100% larger than earlier predictions, and (2) the Raman contribution does not decay as rapidly with increasing concentrations of chlorophyllouslike pigments C as predicted earlier. In fact, the Raman fraction for C 8% at wavelengths of interest in ocean color remote sensing and therefore cannot be ignored in ocean color modeling.     

Factor analysis of multispectral radiances over coastal and open ocean water based on radiative transfer calculations

Factor analysis is applied to multispectral (seventeen wavelengths) radiances simulated by a radiative transfer model (matrix-operator method) in and above coastal and open ocean waters. The calculated radiances were compared with measured radiances before applying factor analysis. They agree well for different sun elevations and even for turbid coastal waters. The factor analysis technique allows us to extract the characteristic signatures of phytoplankton, suspended matter, and yellow substance. The fluorescence of chlorophyll at lambda = 685 nm is found to be a clear signal for phytoplankton, also in the presence of other suspensions and yellow substance. A comparison of different algorithms for the extraction of the fluorescence peak favors the addition of chlorophyll absorption at lambda = 670 nm. The blue-green ratio is found to be useless for chlorophyll detection in coastal waters. Suspended matter and yellow substance can also clearly be seen in the factor loading for all multispectral radiances analyzed. However, suspended matter is reflected more strongly than yellow substance.     

Validation study of the SeaWiFS oxygen A-band absorption correction: comparing the retrieved cloud optical thicknesses from SeaWiFS measurements

Atmospheric correction in ocean-color remote sensing corrects more than 90% of signals in the visible contributed from the atmosphere measured at satellite altitude. The Sea-viewing Wide Field-of-view Sensor (SeaWiFS) atmospheric correction uses radiances measured at two near-infrared wavelengths centered at 765 and 865 nm to estimate the atmospheric contribution and extrapolate it into the visible range. However, the SeaWiFS 765-nm band, which covers 745-785 nm, completely encompasses the oxygen A-band absorption. The O(2) A-band absorption usually reduces more than 10-15% of the measured radiance at the SeaWiFS 765-nm band. Ding and Gordon [Appl. Opt. 34, 2068-2080 (1995)] proposed a numerical scheme to remove the O(2) A-band absorption effects from the atmospheric correction. This scheme has been implemented in the SeaWiFS ocean-color imagery data-processing system. I present results that demonstrate a method to validate the SeaWiFS 765-nm O(2) A-band absorption correction by analyzing the sensor-measured radiances at 765 and 865 nm taken looking at the clouds over the oceans. SeaWiFS is usually not saturated with cloudy scenes because of its bilinear gain design. Because the optical and radiative properties of water clouds are nearly independent of the wavelengths ranging from 400 to 865 nm, the sensor-measured radiances above the cloud at the two near-infrared wavelengths are comparable. The retrieved cloud optical thicknesses from the SeaWiFS band 7 measurements are compared for cases with and without the O(2) A-band absorption corrections and from the band 8 measurements. The results show that, for air-mass values of 2-5, the current SeaWiFS O(2) A-band absorption correction works reasonably well. The validation method is potentially applicable for in-orbit relative calibration for SeaWiFS and other satellite sensors.     

Remote sensing of ocean color: a methodology for dealing with broad spectral bands and significant out-of-band response

A methodology for delineating the influence of finite spectral bandwidths and significant out-of-band response of sensors for remote sensing of ocean color is developed and applied to the Sea-viewing Wide-Field-of-view Sensor (SeaWiFS). The basis of the method is the application of the sensor's spectral-response functions to the individual components of the top-of-the-atmosphere (TOA) radiance rather than the TOA radiance itself. For engineering purposes, this approach allows one to assess easily (and quantitatively) the potential of a particular sensor design for meeting the system-sensor plus algorithms-performance requirements. In the case of the SeaWiFS, two significant conclusions are reached. First, it is found that the out-of-band effects on the water-leaving radiance component of the TOA radiance are of the order of a few percent compared with a sensor with narrow spectral response. This implies that verification that the SeaWiFS system-sensor plus algorithms-meets the goal of providing the water-leaving radiance in the blue in clear ocean water to within 5% will require measurements of the water-leaving radiance over the entire visible spectrum as opposed to just narrow-band (10-20-nm) measurements in the blue. Second, it is found that the atmospheric correction of the SeaWiFS can be degraded by the influence of water-vapor absorption in the shoulders of the atmospheric-correction bands in the near infrared. This absorption causes an apparent spectral variation of the aerosol component between these two bands that will be uncharacteristic of the actual aerosol present, leading to an error in correction. This effect is dependent on the water-vapor content of the atmosphere. At typical water-vapor concentrations the error is larger for aerosols with a weak spectral variation in reflectance than for those that display a strong spectral variation. If the water-vapor content is known, a simple procedure is provided to remove the degradation of the atmospheric correction. Uncertainty in the water-vapor content will limit the accuracy of the SeaWiFS correction algorithm.     

Effects of Spectral Bandpass on SeaWiFS-Retrieved Near-Surface Optical Properties of the Ocean

A simple correction method to remove the spectral bandpass effects of the Sea-viewing Wide Field-of-view Sensor (SeaWiFS) on the derived normalized water-leaving radiances and ocean-near-surface chlorophyll concentration is developed and implemented in the SeaWiFS data-processing system. SeaWiFS has not only in-band response structures but also significant sensor out-of-band contributions. The effects of the SeaWiFS out-of-band contribution at the green bands is particularly significant for the derived normalized water-leaving radiances and therefore for the retrieved ocean-near-surface chlorophyll concentration. With the sensor spectral bandpass corrections, the low chlorophyll concentration is even lower in the clear ocean regions, whereas there are almost no changes for the oceans with a chlorophyll concentration of >0.2 mg/m(3).     

Assessment of a bidirectional reflectance distribution correction of above-water and satellite water-leaving radiance in coastal waters

Water-leaving radiances, retrieved from in situ or satellite measurements, need to be corrected for the bidirectional properties of the measured light in order to standardize the data and make them comparable with each other. The current operational algorithm for the correction of bidirectional effects from the satellite ocean color data is optimized for typical oceanic waters. However, versions of bidirectional reflectance correction algorithms specifically tuned for typical coastal waters and other case 2 conditions are particularly needed to improve the overall quality of those data. In order to analyze the bidirectional reflectance distribution function (BRDF) of case 2 waters, a dataset of typical remote sensing reflectances was generated through radiative transfer simulations for a large range of viewing and illumination geometries. Based on this simulated dataset, a case 2 water focused remote sensing reflectance model is proposed to correct above-water and satellite water-leaving radiance data for bidirectional effects. The proposed model is first validated with a one year time series of in situ above-water measurements acquired by collocated multispectral and hyperspectral radiometers, which have different viewing geometries installed at the Long Island Sound Coastal Observatory (LISCO). Match-ups and intercomparisons performed on these concurrent measurements show that the proposed algorithm outperforms the algorithm currently in use at all wavelengths, with average improvement of 2.4% over the spectral range. LISCO's time series data have also been used to evaluate improvements in match-up comparisons of Moderate Resolution Imaging Spectroradiometer satellite data when the proposed BRDF correction is used in lieu of the current algorithm. It is shown that the discrepancies between coincident in-situ sea-based and satellite data decreased by 3.15% with the use of the proposed algorithm. This confirms the advantages of the proposed model over the current one, demonstrating the need for a specific case 2 water BRDF correction algorithm as well as the feasibility of enhancing performance of current and future satellite ocean color remote sensing missions for monitoring of typical coastal waters.     

Analysis of the influence of O(2) A-band absorption on atmospheric correction of ocean-color imagery

Two satellite-borne ocean-color sensors scheduled for launch in the mid 1990's each have a spectral band (nominally 745-785 nm) that completely encompasses the O(2) A band at 762 nm. These spectral bands are to be used in atmospheric correction of the color imagery by assessment of the aerosol contribution to the total radiance at the sensor. The effect of the O(2) band on the radiance measured at the satellite is studied with a line-by-line backward Monte Carlo radiative transfer code. As expected, if the O(2) absorption is ignored, unacceptably large errors in the atmospheric correction result. The effects of the absorption depend on the vertical profile of the aerosol. By assuming an aerosol profile-the base profile-we show that it is possible to remove most of the O(2)-absorption effects from atmospheric correction in a simple manner. We also investigate the sensitivity of the results to the details of the assumed base profile and find that, with the exception of situations in which there are significant quantities of aerosol in the stratosphere, e.g., following volcanic eruptions or in the presence of thin cirrus clouds, the quality of the atmospheric correction depends only weakly on the base profile. Situations with high concentrations of stratospheric aerosol require additional information regarding vertical structure for this spectral band to be used in atmospheric correction; however, it should be possible to infer the presence of such aerosol by a failure of the atmospheric correction to produce acceptable water-leaving radiance in the red. An important feature of our method for removal of the O(2)-absorption effects is that it permits the use of lookup tables that can be prepared in the absence of O(2) absorption by the use of more efficient radiative transfer codes.