Satellite Ocean Color Data Research Articles

Sea-surface reflectance factor: replicability of computed values.

The sea-surface reflectance factor ρ required for the determination of the water- leaving radiance from above-water radiometric measurements is derived from radiative transfer simulations relying on models of the sky-radiance distribution and sea-surface statistics. This work primarily investigates the impact on ρ of various sky-radiance and sea-surface models. A specific replicability analysis, restricted to the 550 nm wavelength, has been performed with the Monte Carlo code for Ocean Color Simulations (so-called MOX) accounting for the measurement geometry recommended in protocols for the validation of satellite ocean color data and commonly applied for operational measurements. Results indicate that the variability of ρ increases with wind speed and reaches the largest values for sun elevations close to the zenith or approaching the horizon. In particular, a variability up to about 2% is observed for wind speeds below 4 ms-1 and sun zenith angles larger than 20°. Finally, the benchmark of the ρ values from this study with those formally determined with the Hydrolight code and widely utilized by the ocean color community, exhibits systematic differences. The source of these differences is discussed and the implications for field measurements are addressed.

A Deep Learning Model Using Satellite Ocean Color and Hydrodynamic Model to Estimate Chlorophyll-a Concentration

In this study, we used convolutional neural networks (CNNs)—which are well-known deep learning models suitable for image data processing—to estimate the temporal and spatial distribution of chlorophyll-a in a bay. The training data required the construction of a deep learning model acquired from the satellite ocean color and hydrodynamic model. Chlorophyll-a, total suspended sediment (TSS), visibility, and colored dissolved organic matter (CDOM) were extracted from the satellite ocean color data, and water level, currents, temperature, and salinity were generated from the hydrodynamic model. We developed CNN Model I—which estimates the concentration of chlorophyll-a using a 48 × 27 sized overall image—and CNN Model II—which uses a 7 × 7 segmented image. Because the CNN Model II conducts estimation using only data around the points of interest, the quantity of training data is more than 300 times larger than that of CNN Model I. Consequently, it was possible to extract and analyze the inherent patterns in the training data, improving the predictive ability of the deep learning model. The average root mean square error (RMSE), calculated by applying CNN Model II, was 0.191, and when the prediction was good, the coefficient of determination (R2) exceeded 0.91. Finally, we performed a sensitivity analysis, which revealed that CDOM is the most influential variable in estimating the spatiotemporal distribution of chlorophyll-a.

Using the Red Band Difference Algorithm to Detect and Monitor a Karenia spp. Bloom Off the South Coast of Ireland, June 2019

During the months of May, June, July and August 2019 the Red Band Difference algorithm was tested over Irish waters to assess its suitability for the Irish harmful algal bloom alert system. Over the 4 weeks of June an extensive localised surface phytoplankton bloom formed in the Celtic Sea, south of Ireland. Satellite imagery from the Sentinel-3a’s Ocean and Land Colour Instrument, processed using the Red Band Difference algorithm detected the bloom in surface shelf waters and helped monitor its movement. Daily satellite images indicated that the bloom appeared at the sea surface on the 2nd June 2019 and peaked in size and surface abundance in offshore shelf waters within 4 weeks, remnants remained at the surface into July. A particle tracking approach was used to replicate oceanic circulation patterns in the vicinity of the observed algal bloom and estimate its trajectory. The initial horizontal distribution of particles in the tracking model were based on a satellite imagery polygon of the bloom when it first appeared in surface waters. Good agreement was observed between satellite imagery of the bloom and the particle tracking model. In situ sampling efforts from a research cruise and the national inshore phytoplankton monitoring programme confirmed that Karenia mikimotoi was the causative organism of the bloom. This pilot study shows great potential to use the Red Band Difference algorithm in the existing Irish harmful algal bloom alert system. In addition, satellite ocean colour data combined with particle tracking model estimates can be a useful tool to monitor high biomass harmful algal bloom forming species, such as Karenia mikimotoi, in surface coastal waters around Ireland and elsewhere.

Cross-calibration of MODIS and VIIRS long near infrared bands for ocean color science and applications

Generation of consistent multi-sensor datasets is critical to the assessment of long-term global changes using satellite-borne instruments. Recent research suggests, however, that a fundamental assumption in satellite ocean color data processing concerning the calibration of the long near infrared band (i.e., 865 nm for MODIS) may introduce sensor-specific biases in space and/or time, which may also contribute to cross-sensor inconsistency in the derived reflectance data products. As such, it is necessary to assess the calibration of this band across sensors – performed here for MODIS/Aqua and VIIRS/SNPP using ‘simultaneous same view’ matchups (SSV; similar to simultaneous nadir overpass, but allowing for non-nadir measurements). Towards that end, we assess geometric, temporal, and spatial homogeneity metrics to identify SSVs, and develop a band-shifting approach applicable within standard satellite data processing routines to resolve expected spectral differences in the radiometry. We find top-of-atmosphere (TOA) radiance data from VIIRS/SNPP long near infrared band to be approximately 3% higher than the corresponding MODIS/A data. With the expectation that cross-calibrating the NIRL should improve cross-sensor continuity of downstream geophysical products (e.g., chlorophyll-a), we reprocessed VIIRS data using updated calibration coefficients. While we noticed many minor improvements in cross-sensor continuity in such data products, large-scale geographic and temporal biases between these two datasets still remain. These discontinuities may be the result of disparate errors in polarization correction or atmospheric correction, both of which are modulated by radiant path geometry.

Characterization and absolute calibration of an AERONET-OC radiometer.

The Ocean Color component of the global Aerosol Robotic Network (AERONET-OC) utilizes CE-318 sun photometers modified for above-water radiometry from fixed structures such as oil rigs, lighthouses, and service platforms. Primarily, AERONET-OC measurements allow determination of the water-leaving radiance required for the validation of ocean color satellite data products. One instrument from the AERONET-OC network, identified as AERONET #080, was studied in this work. A laser-illuminated integrating sphere of known radiance enabled determination of the linearity with flux and absolute radiance responsivity at multiple wavelengths within seven of the AERONET #080 filter bands. We compared the results to calibrations from the AERONET facility at the Goddard Space Flight Center of the National Aeronautics and Space Administration and from the Joint Research Centre of the European Commission. These results agree within the estimated mean comparison uncertainty of 1.88 % (k=2). We also assessed these results using calibrated lamp-illuminated integrating spheres and observed a spectral dependence to the comparison results that is unexplained.

Evaluation of regions suitable for vicarious calibration of ocean color satellite sensors in the South China Sea.

Accurate retrieval of biogeochemical components of the ocean at a global scale from space requires accurately calibrated top-of-atmosphere (TOA) radiance, which is usually achieved by deriving a vicarious gain coefficient (g-factor) through a process called system vicarious calibration (SVC). Currently, only two SVC sites, Marine Optical Buoy (MOBY) and BOUée pour l'acquiSition d'une Série Optique à Long termE (BOUSSOLE), are routinely operated to support the SVC process for all on-orbit ocean color satellite payloads. However, high-quality matchups between satellite observations and in situ measurements are rare because of the strict requirements of the SVC process. Meanwhile, a stable g-factor is usually computed by averaging sufficient gain measurements. Therefore, more SVC sites are required to derive a stable g-factor in a short duration, particularly for the initial calibration of newly launched satellite sensors. In this study, nearly twenty years of well-calibrated ocean color satellite data were used to calculate the mean and standard deviation of physical and optical properties of waters and the atmosphere in the South China Sea (SCS) to evaluate the feasibility of establishing a SVC site. A region was identified that meets all requirements that were used to evaluate the MOBY and BOUSSOLE sites. Two in situ measurements within this region were used to derive a g-factor for MODIS-Terra and MODIS-Aqua and were compared with the g-factor derived using MOBY data. The consistence of the two g-factors indicates that the identified region in the SCS could be a potential area for establishing a long-term moored SVC site.

Analysis of the Monthly and Spring-Neap Tidal Variability of Satellite Chlorophyll-a and Total Suspended Matter in a Turbid Coastal Ocean Using the DINEOF Method

Missing spatial data is one of the major concerns associated with the application of satellite data. The Data INterpolating Empirical Orthogonal Functions (DINEOF) method has been proven to be an effective tool for filling spatial gaps in various satellite data products. The Ariake Sea, which is a turbid coastal sea, shows the large spatial and temporal variability of chlorophyll-a (Chl-a) and total suspended matter (TSM). However, ocean color satellite data for this region usually have large gaps, which affects the accurate analysis of Chl-a and TSM variability. In this study, we applied the DINEOF method to fill the missing pixels from the regionally tuned Moderate Resolution Imaging Spectroradiometer (MODIS)-Aqua (hereafter, MODIS) Chl-a and MODIS-derived TSM datasets for the period 2002–2017. The validation results showed that the DINEOF reconstructed data were accurate and reliable. Furthermore, the Empirical Orthogonal Functions (EOF) analysis based on the reconstructed data was used to quantitatively analyze the spatial and temporal variability of Chl-a and TSM at both monthly and individual events of spring-neap tidal scales. The first three EOF modes of Chl-a showed seasonal variability mainly caused by precipitation, the sea surface temperature (SST), and river discharge for the first EOF mode and the sea level amplitude for the second. The first three EOF modes of TSM exhibited both seasonal and spring-neap tidal variability. The first and second EOF modes of TSM displayed spring-neap tidal variability caused by the sea level amplitude. The second EOF mode of TSM also showed seasonal variability caused by the sea level amplitude. In this study, we first applied the DINEOF method to reconstruct the satellite data and to capture the major spatial and temporal variability of Chl-a and TSM for the Ariake Sea. Our results demonstrate that the DINEOF method can reconstruct patchy oceanic color datasets and improve spatio-temporal variability analysis.

Derivation of Red Tide Index and Density Using Geostationary Ocean Color Imager (GOCI) Data

Red tide causes significant damage to marine resources such as aquaculture and fisheries in coastal regions. Such red tide events occur globally, across latitudes and ocean ecoregions. Satellite observations can be an effective tool for tracking and investigating red tides and have great potential for informing strategies to minimize their impacts on coastal fisheries. However, previous satellite-based red tide detection algorithms have been mostly conducted over short time scales and within relatively small areas, and have shown significant differences from actual field data, highlighting a need for new, more accurate algorithms to be developed. In this study, we present the newly developed normalized red tide index (NRTI). The NRTI uses Geostationary Ocean Color Imager (GOCI) data to detect red tides by observing in situ spectral characteristics of red tides and sea water using spectroradiometer in the coastal region of Korean Peninsula during severe red tide events. The bimodality of peaks in spectral reflectance with respect to wavelengths has become the basis for developing NRTI, by multiplying the heights of both spectral peaks. Based on the high correlation between the NRTI and the red tide density, we propose an estimation formulation to calculate the red tide density using GOCI data. The formulation and methodology of NRTI and density estimation in this study is anticipated to be applicable to other ocean color satellite data and other regions around the world, thereby increasing capacity to quantify and track red tides at large spatial scales and in real time.

Extending satellite ocean color remote sensing to the near-blue ultraviolet bands

Ultraviolet (UV) radiation has a profound impact on marine life, but historically and even currently, most ocean color satellites cannot provide radiance measurements in the UV, and thus UV penetration, in the global ocean. We develop a system (termed as UVISRdl) in this study, based on deep learning, to estimate remote sensing reflectance (Rrs) at 360, 380, and 400 nm (collectively termed as near-blue UV bands, nbUV) from Rrs in the visible bands that are obtained by ocean color satellites. This system is tested using both synthetic and field-measured data that cover a wide range and large number of values, with the resulted coefficient of determination close to 1.0 and bias close to 0 between UVISRdl estimated and known Rrs(nbUV). These results indicate excellent predictability of Rrs(nbUV) from Rrs(visible) via UVISRdl. The system was further applied to VIIRS (the Visible Infrared Imaging Radiometer Suite) data with the estimated Rrs(nbUV) evaluated using matchup field measurements, and obtained a mean absolute relative difference (MARD) at 360 nm of ~14% for oceanic waters and ~ 50% for coastal waters. These results are equivalent to those reported in the literature for satellite Rrs(visible) in oceanic and coastal waters. Examples of the global distribution of Rrs(nbUV), and subsequently the diffuse attenuation coefficient at the nbUV bands (Kd(nbUV)), are generated after applying UVISRdl to Rrs(visible) from the VIIRS data. The system lays the groundwork to generate decade-long Rrs(nbUV) and Kd(nbUV) from satellite ocean color data, which will be useful and important for both ocean color remote sensing and biogeochemical studies.

Application of Parallel Processing Techniques to Satellite Ocean Color Data Processing

Recent advances demand that remote-sensing satellites efficiently process massive amounts of ocean color data. This paper compares the open multi-processing (OpenMP), the open computing language (OpenCL), the Message Passing Interface (MPI), the hybrid MPI/OpenMP, and the hybrid MPI/OpenCL in the parallel implementation of ocean color processing algorithms using data from the Geostationary Ocean Color Imager (GOCI), which is the first ocean color remote sensor operated in geostationary orbit. Since 2010, GOCI has observed ocean color around the Korean Peninsula and has generated hundreds of terabytes of big data. When any of the data-processing algorithms are updated, all preexisting data is required to be reprocessed, which can take hundreds of days because GOCI data are currently processed sequentially. Therefore, we attempted to develop an efficient parallel processing methodology for GOCI data. We tested well-known GOCI data-processing algorithms, like the chlorophyll (CHL) and total suspended solid (TSS) concentration estimation algorithms, using a cluster system. This cluster uses the Red Hat Linux operating system with two Intel Xeon 8-core processors (CPU), an AMD Radeon HD 7970 (GPU), and InfiniBand 4x QDR (network). As a result of this study we were able to improve the GOCI ocean color algorithms

Toward Determining the Spatio-Temporal Variability of Upper-Ocean Ecosystem Stoichiometry From Satellite Remote Sensing

The elemental stoichiometry of particulate organic carbon (C), nitrogen (N), and phosphorus (P) connects the C fluxes of biological production to the availability of the limiting nutrients in the ocean. It also influences the marine food-web by modulating zooplankton’s feeding behavior and organic matter decomposition by bacteria. Despite its importance, there is a general paucity of information on how the global C:N:P ratio evolves seasonally and interannually, and large parts of the global ocean remain devoid of observational data. Here, a new method combining satellite ocean-color data with a cellular-trait-based model to characterize the spatio-temporal variability of the phytoplankton stoichiometry in the surface mixed layer of the ocean is presented. This new method is demonstrated specifically for the C:P ratio. The approach was applied to phytoplankton growth rates and chlorophyll-to-carbon ratios derived from MODIS-Aqua and maps of temperature-dependent nutrient limitation to generate global and seasonal maps of upper-ocean phytoplankton C:P. Taking it a step further, we determined the C:P of the bulk particulate organic matter, using MODIS-Aqua estimates of particulate organic carbon and phytoplankton biomass. Our results are within 95% confidence interval of available data for both horizontal distributions and time series, indicating our new method’s viability in accurately quantifying seasonally resolved global ocean bulk C:P. We anticipate the new hyperspectral capabilities of the NASA PACE (Plankton, Aerosol, Cloud, ocean Ecosystem) mission will facilitate the determination of phytoplankton stoichiometry for different size classes and further enhance the predictability of marine-ecosystem stoichiometry from space.

Estimating particulate organic carbon flux in a highly dynamic estuary using satellite data and numerical modeling

Quantitative estimates of the flux, transportation, and burial of riverine particulate organic carbon (POC) in estuaries and shelf areas are crucial for understanding the carbon cycle in marginal seas. The complexity of the hydrodynamic environment and estuarine processes in the Changjiang River estuary (CRE) results in a POC flux with high spatiotemporal variability. In this study, we proposed a new method for time series monitoring of riverine POC flux by combining satellite ocean color and numerical modeling data. We used 500-m resolution geostationary satellite data from the Geostationary Ocean Color Imager (GOCI) to retrieve surface POC concentrations in the CRE and used the numerical Finite Volume Community Ocean Model (FVCOM) to simulate three-dimensional (3D) current and sediment distribution. Diurnal, seasonal, and annual variations in POC flux through the lower reaches (Datong Hydrological Station), upper estuary (Xuliujing Hydrological Station), and outlets of the CRE, as well as transportation and deposition outside the CRE, were examined from July 2011 to June 2018. The annual mean POC fluxes through the Datong, Xuliujing, and outlet sections were 1.16 ± 0.16 Tg C/yr, 1.29 ± 0.11 Tg C/yr, and 1.17 ± 0.11 Tg C/yr, respectively. After passing through the estuary, 35.2% of POC was deposited in the sandbar outside the outlets, 52.9% was transported to the south, and a small amount entered the eastern shelf. Thus, these results indicated that POC flux through the river and estuary sections differed from the effective riverine POC flux into the sea. The proposed method combining high spatiotemporal-resolution satellite data and numerical modeling not only makes full use of their advantages, but also reduces the uncertainty of their individual estimates. This approach also supports long-term monitoring of riverine fluxes to the sea and helps clarify the effects of terrestrial inputs on marginal seas under multiple stresses from human activities and climate change.

Atmospheric correction of geostationary satellite ocean color data under high solar zenith angles in open oceans

With a revisit time of 1 h, spatial resolution of 500 m, and high radiometric sensitivity, the Geostationary Ocean Color Imager (GOCI) is widely used to monitor diurnal dynamics of oceanic phenomena. However, atmospheric correction (AC) of GOCI data with high solar zenith angle (>70°) is still a challenge for traditional algorithms. Here, we propose a novel neural network (NN) AC algorithm for GOCI data under high solar zenith angles. Unlike traditional NN AC algorithms trained by radiative transfer-simulated dataset, our new AC algorithm was trained by a large number of matchups between GOCI-observed Rayleigh-corrected radiance in the morning and evening and GOCI-retrieved high-quality noontime remote-sensing reflectance (Rrs). When validated using hourly GOCI data, the new NN AC algorithm yielded diurnally stable Rrs in open ocean waters from the morning to evening. Furthermore, when validated by in-situ data from three Aerosol Robotic Network-Ocean Color (AERONET-OC) stations (Socheongcho, Gageocho and Ieodo), the GOCI-retrieved Rrs at visible bands obtained using the new AC algorithm agreed well with the in-situ values, even under high solar zenith angles. Practical application of the new algorithm was further examined using diurnal GOCI observation data acquired in clear open ocean waters. Results showed that the new algorithm successfully retrieved Rrs for the morning and evening GOCI data. Moreover, the amount of Rrs data retrieved by the new algorithm was much higher than that retrieved by the standard AC algorithm in SeaDAS. Our proposed NN AC algorithm can not only be applied to process GOCI data acquired in the morning and evening, but also has the potential to be applied to process polar-orbiting satellite ocean color data at high-latitude ocean that also include satellite observation with high solar zenith angles.

ESTIMATION OF OCEANIC PARTICULATE ORGANIC CARBON WITH MACHINE LEARNING

Abstract. Understanding and quantifying ocean carbon sinks of the planet is of paramount relevance in the current scenario of global change. Particulate organic carbon (POC) is a key biogeochemical parameter that helps us characterize export processes of the ocean. Ocean color observations enable the estimation of bio-optical proxies of POC (i.e. particulate backscattering coefficient, bbp) in the surface layer of the ocean quasi-synoptically. In parallel, the Argo program distributes vertical profiles of the physical properties with a global coverage and a high spatio-temporal resolution. Merging satellite ocean color and Argo data using a neural networkbased method has already shown strong potential to infer the vertical distribution of bio-optical properties at global scale with high space-time resolution. This method is trained and validated using a database of concurrent vertical profiles of temperature, salinity, and bio-optical properties, i.e. bbp, collected by Biogeochemical-Argo (BGC-Argo) floats, matched up with satellite ocean color products. The present study aims at improving this method by 1) using a larger dataset from BGC-Argo network since 2016 for training, 2) using additional inputs such as altimetry data, which provide significant information on mesoscale processes impacting the vertical distribution of bbp, 3) improving the vertical resolution of estimation, and 4) examining the potential of alternative machine learning-based techniques. As a first attempt with the new data, we used some feature-specific preprocessing routines followed by a Multi-Output Random Forest algorithm on two regions with different ocean dynamics: North Atlantic and Subtropical Gyres. The statistics and the bbp profiles obtained from the validation floats show promising results and suggest this direction is worth investigating even further at global scale.

Sensitivity of Satellite Ocean Color Data to System Vicarious Calibration of the Long Near Infrared Band

Satellite ocean color missions require accurate system vicarious calibrations (SVC) to retrieve the relatively small remote-sensing reflectance ( $R_{\mathrm {rs}}$ , sr−1) from the at-sensor radiance. However, the current atmospheric correction and SVC procedures do not include calibration of the “long” near infrared band (NIRL—869 nm for MODIS), partially because earlier studies, based primarily on simulations, indicate that accuracy in the retrieved $R_{\mathrm {rs}}$ is insensitive to moderate changes in the NIRL vicarious gain ( $g$ ). However, the sensitivity of ocean color data products to $g$ (NIRL) has not been thoroughly examined. Here, we first derive 10 SVC “gain configurations” (vicarious gains for all visible and NIR bands) for MODIS/Aqua using current operational NASA protocols, each time assuming a different $g$ (869). From these, we derive a suite of ~1.4E6 unique gain configurations with $g$ (869) ranging from 0.85 to 1.2. All MODIS/A data for 25 locations within each of five ocean gyres were then processed using each of these gain configurations. Resultant time series show substantial variability in dominant $R_{\mathrm {rs}}$ (547) patterns in response to changes in $g$ (869) (and associated gain configurations). Overall, mean $R_{\mathrm {rs}}$ (547) values generally decrease with increasing $g$ (869), while the standard deviations around those means show gyre-specific minima for $0.97 (869) $g$ (869) using such time series, finding $g$ (869) = 1.025 most closely comports with expectations. This approach is broadly applicable to other ocean color sensors, and highlights the importance of rigorous cross-sensor calibration of the NIRL bands, with implications on consistency of merged-sensor data sets.

A Layer Removal Scheme for Atmospheric Correction of Satellite Ocean Color Data in Coastal Regions

The radiance received by satellite sensors viewing the ocean is a mixed signal of the atmosphere and ocean. Accurate decomposition of the radiance components is crucial because any inclusion of atmospheric signal in the water-leaving radiance leads to an incorrect estimation of the oceanic parameters. This is especially true over the turbid coastal waters, where the estimation of the radiance components is difficult. A layer removal scheme for atmospheric correction (LRSAC) has been developed to take the atmospheric and oceanic components as the layer structure according to the sunlight passing in the Sun-Earth-satellite system. Compared with the normal coupled atmospheric column, the uncertainty of the layer structure of Rayleigh and aerosols has a relatively small error with a mean relative error (MRE) of 0.063%. As the aerosol layer was put between Rayleigh and ocean, a new Rayleigh lookup table (LUT) was regenerated using 6SV (Second Simulation of a Satellite Signal in the Solar Spectrum, Vector version 3.2) based on the zero reflectance at the ground to produce the pure Rayleigh reflectance without the Rayleigh-ocean interaction. The accuracy of the LRSAC was validated by in situ water-leaving reflectance, obtaining an MRE of 6.3%, a root-mean-square error (RMSE) of 0.0028, and the mean correlation coefficient of 0.86 based on 430 matchup pairs over the East China Sea. Results show that the LRSAC can be used to decompose the reflectance at the top of each layer for the atmospheric correction over turbid coastal waters.

Modulation of chlorophyll concentration by downwelling Rossby waves during the winter monsoon in the southeastern Arabian Sea

Signatures of Rossby waves that are evident in satellite ocean colour data in the southeastern Arabian Sea (SEAS) are more prominent during the winter than the summer monsoon. During the winter, the sea level is high, and the conditions are less conducive for the injection of nutrients through the eddy-pumping mechanism. In this study, we use satellite observations and a physical-ecosystem model to examine the role of horizontal advection associated with Rossby waves in enhancing chlorophyll concentration in the SEAS. Zonal advection dominates over meridional advection, and its effect is more prominent in the south of Indian subcontinent and west of the Maldives. Except along the coast and around the islands, the vertical processes are weak and do not lead to a substantial increase in chlorophyll. Model simulations show that the currents associated with the Rossby wave also transport nutrients and other physicochemical properties of water along the path of the wave. The advected water decreases the concentration of Arabian Sea High Salinity water mass and replenishes the oxygen in the oxygen minimum zone in the SEAS.

Influence of the Suspended Particulate Matter on the Satellite Radiance in the Sunglint Observation Geometry in Coastal Waters

The analysis of satellite ocean color data that are acquired over coastal waters is highly relevant to gain understanding of the functioning of these complex ecosystems. In particular, the estimation of the suspended particulate matter (SPM) concentrations is of great interest for monitoring the coastal dynamics. However, a high number of pixels of satellite images could be affected by the surface-reflected solar radiation, so-called the sunglint. These pixels are either removed from the data processing, which results in a loss of information about the ocean optical properties, or they are subject to the application of glint correction techniques that may contribute to increase the uncertainties in the SPM retrieval. The objective of this study is to demonstrate the high potential of exploiting satellite observations acquired in the sunglint viewing geometry for determining the water leaving radiance for SPM dominated coastal waters. For that purpose, the contribution of the water leaving radiance Lw to the satellite signal LTOA is quantified for the sunglint observation geometry using forward radiative transfer modelling. Some input parameters of the model were defined using in-situ bio-optical measurements performed in various coastal waters to make the simulations consistent with real-world observations. The results showed that the sunglint radiance is not sufficiently strong to mask the influence of the oceanic radiance at the satellite level, which oceanic radiance remains significant (e.g., 40% at 560 nm for a SPM concentration value of 9 g m−3). The influence of the sunglint radiance is even weaker for highly turbid waters and/or for strong wind conditions. In addition, the maximum radiance simulated in the sunglint region for highly turbid waters remains lower than the saturation radiances specified for the current ocean color sensors. The retrieval of Lw and SPM should thus be feasible from radiances measured in the sunglint pattern by satellite sensors, thus increasing the number of exploitable pixels within a satellite image. The results obtained here could be used as a basis for the development of inverse ocean color algorithms that would interestingly use the radiance measured in sunglint observation geometry as it has been done for other topics than the field of ocean color research.

An Ocean Color Algorithm Based on Power Functions to Retrieve Inherent Optical Properties From Remotely Sensed Data Off the Baja California Peninsula, Mexico

Ocean color is determined by the different components of seawater. Satellite ocean color data are commonly used to infer the inherent optical properties (IOPs) of seawater and concentration of chlorophyll $a$ (Chl- $a$ ). Many of the algorithms that relate satellite data to the components of seawater use global data to determine their parameters. Regionalizing the algorithms is recommended to improve the retrieval of IOPs from satellite data. In this article, we use the in situ and satellite data from the southern part of the California Current System (CCS) off the Baja California Peninsula during 1999 to test the Garver and Siegel IOPs’ inversion model (GSM). The strong seasonal variability observed in the data was not reproduced by this algorithm. Thus, we propose a new regional algorithm whose order and coefficient depend on seasonal variability. This algorithm includes the linear form of the GSM model but allows changes in two parameters according to the season. These two parameters are the magnitude and spectral shape of the relationship of the remote sensing reflectance with the ratio of the backscattering coefficient and the sum of backscattering and absorption coefficients. In this area, both parameters increase gradually from January to October. The new algorithm improves the retrieval of the IOPs’ ratio from an ${r}^{2}$ below 0.5 with the GSM model to an ${r}^{2}$ that ranges from 0.87 in April and October to 0.93 and 0.94 in January and August, respectively. Furthermore, Chl- $a$ data were used as a proxy to evaluate the new algorithm. Absorption data retrieved from remotely sensed data for 1999, 2008, and 2011 relate well to the in situ Chl- $a$ data.

Modeling the Chlorophyll-a from Sea Surface Reflectance in West Africa by Deep Learning Methods: A Comparison of Multiple Algorithms

Deep learning provide successful applications in many fields. Recently, machines learning are involved for oceans remote sensing applications. In this study, we use and compare about eight (8) deep learning estimators for retrieval of a mainly pigment of phytoplankton. Depending on the water case and the multiple instruments simultaneouslyobserving the earth on a variety of platforms, several algorithm are used to estimate the chlolophyll-a from marine reflectance. By using a long-term multi-sensor time-series of satellite ocean-colour data, as MODIS, SeaWifs, VIIRS, MERIS, etc, we make a unique deep network model able to establish a relationship between sea surface reflectance and chlorophyll-a from any measurement satellite sensor over West Africa. These data fusion take into account the bias between case water and instruments. We construct several chlorophyll-a concentration prediction deep learning based models, compare them and therefore use the best for our study. Results obtained for accuracy training and test are quite good. The mean absolute error are very low and vary between 0,07 to 0,13 mg/m3.