A simple semi-analytical model (SAB) was developed for computing a(560) and bb(550) from HJ-1A/CCD images. By comparison with field measurements, the SAB model produces 5.3-23.5% uncertainty for a(560) and bb(550) retrievals. The a(560) and bb(550) are also retrieved from satellite images. The match-up analysis results indicate that a(560) and bb(550) may be derived from the HJ-1A/CCD images with respective uncertainties of 29.84 and 21.35%. These findings imply that, provided that an atmospheric correction scheme for the green bands is available, the extensive database of HJ-1A/CCD imagery may be used for the quantitative monitoring of optical properties in coastal waters.
© 2013 OSA
The interface between terrestrial and ocean environments, coastal waters are among the most turbid and productive components of the world’s oceans . For example, coastal ecosystems have been estimated to account for 14-30% of total marine organic matter production, and recent studies have highlighted the large degree of uncertainty surrounding the role of coastal seas in carbon dioxide sequestration . In addition to their role in the global carbon cycle, coastal seas support a diverse and vital range of ecosystem services, including fisheries and industries such as recreation and tourism . However, coastal regions are dynamic environments where events and processes operate over short time and space scales, often causing conventional station samples to be insufficient for the purpose of mapping patterns . New observational methods with the capability to rapidly sample at high resolution are required. Over the past 30 years, ocean color data have primarily been used to provide information regarding marine primary productivity, coastal sedimentation and sediment dispersal, harmful algal blooms, organic matter content, raw sewage disposal, and pollution [5–8]. Recent advances in optical sensor technology have allowed scientists to utilize ocean color satellite images to synoptically investigate large-scale surface features in coastal regions .
Theoretical studies have shown that remote sensing reflectance may be related to the inherent optical properties (IOPs) of the water column, which in turn are composed of individual absorbers and scatters, including pure water, particulates, and dissolved materials ; based on the results of these studies, it may be seen that the total absorption spectrum is the basis for estimating the contribution of the different absorbers in the water . More generally, an accurate estimation of the scattering and attenuation characteristics of the upper layer ocean may provide insight into the nature of the particles in suspension . Total absorption and backscattering coefficients are also directly associated with a proper bio-optical treatment of the light regime of the ocean mixed layer and the related heat budget . Moreover, the total absorption and backscattering coefficients define the upwelling light field, and are the most logical targets for the retrieval of the water-leaving reflectance spectrum determined from satellite observations . In spite of these observations, optical data have been used for coastal studies and the optical classification of water bodies , thus IOPs, such as total absorption and backscattering coefficients at a specific wavelength, are essential for characterizing the marine optical environment and remote sensing applications.
Methods which may be used for the accurate remote sensing of total absorption and backscattering coefficients in coastal waters have been under investigation for several decades, and algorithms ranging from empirical to full-spectral optimization have been proposed. Empirical models apply simple or multiple regressions between the property of interest and the ratios of irradiance reflectance or remote sensing reflectance . Such approaches do not require a full understanding of the relationship between remote sensing reflectance and the absorption and backscattering coefficients . Due to the statistical nature of regression, however, the accuracy of these models decreases when the bio-optical conditions depart from these representative data sets used to empirically derive the covariance relationships . In order to improve the limitations and local specificity of the empirical model, in a recent study, Lee et al.  developed a multi-band quasi-analytical algorithm (QAA) for deriving the absorption and backscattering coefficients from remote sensing reflectance of open ocean and coastal areas, further analyzed the influential factors of the model, and suggested a method for the optimization of the reference positions. The model is composed of two empirical models, three semi-analytical models, and two analytical models, and has been well validated using in situ observations from a variety of oceanic and coastal environments with a wide range of water types, covering waters such as the Gulf of Mexico, Bering Sea, Arabian Sea, Sargasso Sea, and the area near Key West. Lee et al.  demonstrated that the QAA model may be used to assess absorption coefficients in various turbid waters without re-parameterization. Recently, Smyth et al.  presented a semi-analytical approach to the problem of determining inherent optical properties from satellite and in situ ocean color data. The model uses empirical derived spectral slopes between neighboring wavebands, in combination with the hydrological radiative transfer model, to determine the spectral absorption and backscattering. The model has been thoroughly tested using the NASA bio-optical marine algorithm data set, indicating that, compared with over 400 in situ data points, the model produces an accurate estimation of the total absorption and backscattering across the entire spectrum, with regression slopes close to unity, with little or no bias, a high percentage of variance explained, and a low level of estimation error (<20%).
Remote sensing techniques may be used to optimize our efforts to improve the information content of the ocean, while reducing costs. One issue is the necessity for a better spatial knowledge of environmental variables . However, most current efforts are attributed to developing models for quantifying the total absorption and backscattering coefficients from low spatial resolution remote sensing systems, such as the Moderate Resolution Imaging Spectroradiometer (MODIS), Sea-viewing Wide Field-of-view Sensor (SeaWiFS), and Medium Resolution Imaging Spectrometer (MERIS) [4, 15, 18], but at the same time these methods lack appropriate wavelengths or bandwidths for high spatial resolution remote sensing systems. Additionally, it is widely known that coastal areas are dynamic aquatic regions where events and processes operate over very high spatial scale and short temporal scales [8, 20]. Therefore, in the future broadband sensors will be of greater importance in many coastal applications, due to the limited spatial resolution of ocean satellite sensors such as MODIS.
Environment and disaster monitoring and forecasting with a small satellite constellation (HJ-1A) represents a new generation of small Chinese civilian Earth-observing optical remote sensing satellite. For example, due to high spatial resolution and short revisited time, it is very convenient to use HJ-1A data for water body mapping , water surface monitoring , and the effect of non-point source pollution on water quality forecasting . According to the Chinese government’s plans, the small optical satellite HJ-1A was successfully launched into an ascending orbit at 10:30 on September 6, 2008. The satellite provides remote sensing information in the visible and infrared spectral bands with mid-high spatial resolution, high time resolution, and wide bandwidth, which are widely used in environmental monitoring, protection and reduction. Although two Charge Coupled Device (CCD) cameras were attached to the HJ-1A satellite, only the CCD1 images scanned on 12 September, 2012 met the requirements of the study; the images of both CCD1 and CCD2 from other dates were not adequate in quality, due to heavy cloud cover or lack of satellite overpass. The CCD1 camera has three bands in the visible range at 485, 560, and 660 nm with bandwidths of 60 nm, and one band in the near-infrared range at 825 nm with a bandwidth of 150 nm. It was placed symmetrically to the sub-astral point, dividing the field of observation equally, and allowing a side-by-side observation and joint push broom imaging. The ground swath width of HJ-1A is 700 km. It has a 30 m ground pixel resolution, ± 30° side observation ability, and a calibration function . Although the HJ-1A/CCD sensors were primarily designed to study land targets, when used properly, they may also be applied in coastal aquatic environments to assess the concentration of waterborne optically active constituents (suspended sediment, chlorophyll-a particle, and colored dissolved organic matters), bathymetry bottom vegetation cover, and coral reef .
The objectives of this study are to construct and test a simple semi-analytical total absorption and backscattering coefficients retrieval model (SAB model). The specific goals are as follows: (1) To discuss the optical properties of coastal waters in the Oujiang River estuary; (2) To construct and calibrate the SAB model using the data set collected from the Oujiang River estuary; (3) To test and evaluate the accuracy of the SAB model in predicting total absorption and backscattering coefficients from the HJ-1A/CCD data of the Oujiang River estuary; and (4) To further evaluate the performance of the SAB model using the data sets collected from the West Florida Shelf, USA, and Bohai Sea, China.
2 Data, methods and techniques
2.1 Study area
The Oujiang River estuary (Fig. 1) is located between longitudes 120°48′E and 121°32′E, and latitudes 27°36′N and 28°12′N, near the center of Wenzhou City, Zhejiang Province, a city which is well known in China for its rapid economic development over the past two decades. The optical properties of coastal waters such as the Oujiang River estuary are vital to local human activities and needs, and play a critical role in the regional ecosystem, which may also impact climatic changes. Due to the rapid economic development and population growth of this area, enormous quantities of nutrients and other pollutants are transported from land to the Oujiang River estuary, which may be among the main causes of the increasing number and scale of harmful algal bloom events in the coastal waters there . Thus, there is an urgent need to effectively monitor and manage the aquatic environment in the Oujiang River estuary, and to better understand the optical, biological and ecological processes and phenomena which occur there . In order to obtain useful bio-optical information from the estuary, it is necessary to perform research on the respective developments of specific, regional, and high resolution remote sensing algorithms in this area, in order to effectively monitor and manage water color in coastal regions.
2.2 Accuracy assessment
In this study, the Root-Mean-Square (RMS) of the ratio of modeled-to-measured values will be used to assess the accuracy of atmospheric correction. This statistic will be referred to as RMS and is described by the following equation:
where, xmod,i is the modeled value of the ith element, xobs,i is the observed value of the ith element, and n is the number of elements.
2.3 Data used and in situ measurements
In order to evaluate the accuracy of the algorithms for predicting the total absorption and backscattering coefficients in the Oujiang River estuary, in this study bio-optical data sets (Fig. 2 and Table 1) consisting of simultaneous measurements of above-water remote sensing reflectance, total backscattering coefficient, and the total (minus pure water) absorption were measured. The bio-optical data sets were collected during eight independent cruises in the Oujiang River estuary, respectively on September 9, 10, 12, 16, 18, 19, 21 and 25, 2012. The bio-optical data sets were divided into two smaller groups randomly, a calibration data set and a validation data set. The calibration data set, which included 61 samples, was used to initialize the SAB models respectively collected on September 9, 10, 18, 19, 21 and 25, 2012. The validation data set, which included 21 samples, was used to estimate the stability and accuracy of the SAB models collected on September 12 and 17, 2012. Furthermore, to validate the applicability of the SAB models in other coastal waters, two independent data sets, including simultaneous measurements of above-water remote sensing reflectance and total absorption coefficients, were collected during the period of 1999-2003 in the West Florida Shelf, USA and in 2005 in the Bohai Sea, China. Each data set was divided into two smaller data sets, namely the calibration data set and validation data set. It is worth noting that the West Florida Shelf data set was provided by the SeaWiFS Bio-optical Archive and Storage System (SeaBASS), which was funded by NASA for data product validation, algorithm development, satellite data comparison and inter-calibration, and data merger studies and time series analyses.
Measurements made during the cruise included above-surface remote sensing reflectance, total absorption coefficients, and total backscattering coefficients. The total backscattering coefficient, bb(λ), was measured at 442, 488, 550, 676, 700 and 852 nm, using a Hydroscat-6 (Hydro-Optics, Biology, and Instrumentation Laboratories). The Hydroscat data was corrected for absorption and scattering in the sample volume using the “sigma correction method”, as recommended by the manufacturer, with coefficients calculated from coincident data recorded by a WETLab AC-9 during Hydroscat deployments. The vertical profiles of total (minus pure) absorption, at-w(λ), were measured using WETLabs AC-9. The WETLabs AC-9 was calibrated annually by the manufacturer. Clear water calibrations were carried out at each measurement, and air tracking was performed with every cleaning, before and after every cruise . Above-surface optical measurements were made by means of a field spectroradiometer (Spectral Devices, Boulder, CO, ASD) simultaneous with in-water optical measurements, and following the NASA ocean-optics protocols . The spectroradiometer had been calibrated annually by the manufacturer. Various measurements (three repeated measurements in a short time) were repeated at each station in order to estimate the uncertainty (RMS values of three repeated measurements) associated with each measurement, and the average measurements (mean values of the three repeated measurements) with <5% RMS at each station were selected for model calibration and validation.
2.4 SAB model
Due to the fact that satellites and many other sensors measure remote sensing reflectance from above the water surface, it is necessary to convert above-surface remote sensing reflectance spectral Rrs(λ) into below-surface spectral rrs(λ) [15, 29], as expressed in the following formula:30]. As Rrs(λ) is small for most oceanic and coastal waters, the variation of the Q value exerts only a small influence on the conversion between Rrs(λ) and rrs(λ) . From the calculated HYDROLIGHT (a radiative transfer numerical model that computes radiance distributions and derived quantities for natural water bodies ) Rrs(λ) and rrs(λ) values, it is found that T≈0.52 and κQ≈1.7 for optically deep waters and a nadir-viewing sensor .
The IOP inversion model assumes that the relationship between rrs(λ) and the ratio of backscattering to absorption is known and stable, for example :34] found that l0≈0.0949 and l1≈0.0794 for Case I waters, while Lee et al.  suggested that l0 of 0.084 and l1 of 0.17 work better for higher scattering coastal waters. Actually the values of l0 and l1 may vary with particle phase function , which is not known remotely. Without local bio-optical information or model to predetermine the values of l0 and l1 in an semi-analytical algorithm, the Gordon et al.  suggested values are used in Case I waters, while Lee et al.  determined values work in Case II waters. Therefore, knowing rrs(λ), Lee et al.  proposed that the ratio of bb(λ) to a(λ) + bb(λ) can be expressed using Eq. (4) that is derived from Eq. (3), by solving second order quadratic.
In order to solve Eq. (4), which has two unknowns, a(λ) and bb(λ), if l0 and l1 are known, then two equations are required. This is achieved by using the spectral slopes of both absorption and backscattering between two neighboring wavelengths, λi and λj (where λi>λj and assuming linearity) . The parameter εa, the spectral ratio for total absorption, and εb, the spectral ratio for total backscattering, are defined as follows:
In addition to backscattering by all particulate matter, u(λ) is also affected by absorption by tripton, CDOM, chlorophyll-a, and waters. If the absorbing varies between samples, u(λ) output would be different for the same bb(λ). To account for this, a second spectral band has been used. To isolate bb(λ) from u(λ), substituting Eq. (5) into the difference, u−1(λi)- εaεbu−1(λj), yields the following:
The parameter εa depends on the slope of absorption caused by phytoplankton, suspended particles, and colored dissolved organic matter (CDOM). Variations in chlorophyll-a, which are the main causes of variations in ocean color spectra in coastal regions, have the greatest effect between 400 and 560 nm, in terms of both change in shape and magnitude of the absorption spectrum. Fortunately, Smyth et al.  indicated that some small changes based on the ratio of photoprotectant to photosynthetic carotenoids may have some bearing in the 490-550 spectral range. However, there is no HJ-1A/CCD wavelength wholly belong to the 490-550 spectral range. Fortunately, there are two HJ-1A/CCD wavelengths that are quite closed to 490-550 spectral range, i.e., 485 and 560 nm that partially located in 490-550 spectral range. It may work well for assuming that the ratio of at-w(560) to at-w(485) is relatively unaffected by changes in chlorophyll-a and may be chosen as the pairing to calculate the εa. Hale and Querry  proposed that the water absorption coefficients for HJ-1A/CCD bands at 485 and 560 nm are approximately 0.0183 and 0.0904 m−1, respectively. Therefore, reforming Eq. (6), we may obtain the conceptual model for total backscattering coefficient deriving as follows:37, 38]. Provided that bb(λi) is known, the conceptual model for a(λi) retrievals can be expressed as following formula:
It is noteworthy that in this study the proportional symbol (as opposed to the equal symbol) is used to construct the SAB model. There are at least two reasons for which the proportional symbol is more suitable than the equal symbol. First, the parameters such as l0 and l1 are site-specific parameters . This is to say that these parameters are generally only appropriate to waters with optical characteristics similar to those used in the parameters determination. Secondly, the structure of Eq. (3) originates from Gordon , and revised by Lee et al. . Morel and Gentili  proposed that the possibility that the particle’s volume scattering function (VSF) may change is not accounted for the approach of Eq. (3). As a result, Eq. (3) may be violated while the VSF of optically active constituent or solar zenith angle is different.
3.1 Reflectance spectra
Remote-sensing reflectance data collected from the Oujiang River estuary and Bohai Sea exhibit a large variability over the visible and NIR spectral regions. The magnitude and shape of the reflectance curves in the two data sets (Fig. 2) all differ from each other, clearly indicating that they represent very different optical environments, ranging from low-turbid to high-turbid waters. In general, the remote sensing reflectance is highly variable over the visible and near-infrared spectral regions in turbid Case II waters. In the Oujiang River estuary and Bohai Sea, spectra are very low in the blue range (400-500 nm), which is lower than 0.025 sr−1. Reflectance in the green range (500-600 nm) is much higher than in the blue and red range, with the exception of samples with high suspended sediment concentration. In the red region (600-700 nm) reflectance showed several spectral features. A second minimum of approximately 674 nm corresponds to the red chlorophyll-a absorption maximum. However, this spectral trough is not very distinct when the chlorophyll-a concentration is low but the suspended sediment concentration is high. A distinct peak is located between 690 and 710 nm, which is the result of both high backscattering and a minimum in absorption by all optically active constituents including pure water . As a result, the reflectance values of waters with high suspended sediment concentration are larger than the reflectance values around 570 nm. The reflectance in the near-infrared range (700-800 nm) showed wide variation with low reflectance, the spectral curves being very flat at this range due to the low backscattering and absorption by optically active constituents, except for pure waters. Finally, a third distinct peak around 815 nm corresponds to the high backscattering caused by suspended sediment concentration. Due to the high absorption by pure waters, the reflectance values of this peak are usually lower than those at around 570 and 700 nm.
Remote-sensing reflectance in the West Florida Shelf exhibits characteristics which are greatly different from those of the Oujiang River estuary and Bohai Sea. The magnitude and shape of the reflectance curves in the West Florida Shelf all differ from each other, clearly indicating that they represent very different optical environments, ranging from low-cha, blue waters to productive coastal waters. On the one hand, the spectral curves with a negative slope at the blue regions are quite similar in magnitude and shape to typical reflectance in Case I waters , indicating that some West Florida Shelf waters are typical low-cha, blue waters. In these waters, the reflectance in the blue range was shown to be very low (<0.01 sr−1), was even lower at the green-red regions (<0.005 sr−1), and reached its lowest point at the NIR regions (<0.0002 sr−1). On the other hand, some spectral curves with positive slope at the blue regions are quite similar to typical reflectance in Case II waters, illuminating that some of the West Florida Shelf waters fall into the productive Case II water type . In these waters, while remote sensing reflectance in the range from 400 to ~480 nm remained below 0.01 sr−1, remote sensing reflectance in the green was much higher, reaching 0.026 sr−1, and the peak around 690 nm at many stations was quite lower than that at the green reflectance peak.
3.2 Relationship between IOPs and rrs(λ)
The structure of Eq. (3) originates from a study by Gordon et al. ; the authors developed a reflectance model, in which l0 and l1 were given the respective mean values of 0.0949 and 0.0794 for Case I waters. Recently, Lee et al.  suggested that an l0 of 0.084 and l1 of 0.17 are more suitable for higher scattering coastal waters. As shown in Morel and Gentilli , the coefficients l0 and l1 are not constant and vary in an orderly manner along with changes in the water optical properties and with the illumination conditions. This signifies that the parameters of l0 and l1 are local specific coefficients, which change with the bio-optical properties of the local waters. Thus, in order to accurately retrieve the total absorption and backscattering coefficients at specific wavelengths from remote sensing reflectance, local bio-optical information is required to improve the values of l0 and l1, while the bio-optical properties of given aquatic environments are different from those used to develop models by Gordon et al. , Morel and Gentilli , and Lee et al. .
Gordon et al.  indicated that the values of l0 and l1 do not change with wavelength, at least for solar zenith angles varying from 0 to 20°, except perhaps for the band centered at 667 nm, which is not typically used. Thus, it may be assumed that l0 and l1 are independent of wavelengths for all Hydroscat-6 wavebands of interest. This signifies that if the values of l0 and l1 are known at given wavelengths, the relationship between IOPs and rrs(λ) at other wavebands of interest may also be constructed. In this study, the reflectance around the spectral peak is used to estimate the values of l0 and l1, due to the fact that these wavebands are beneficial to minimizing the impacts of the measurements’ noise on the retrieval results. Although there are three candidate wavebands of 570, 700, and 815 nm available for computing the values of l0 and l1, Hydroscat-6 only provides the waveband at 700 nm for this purpose. Figure 3 indicates a prediction of l0 and l1 using field measurements at 700 nm, the correlation coefficient of which is 0.9281. It was found that l0 = 0.0364 and l1 = 0.3484 are more suitable for the coastal waters of the Oujiang River estuary, which is quite different from the study results shown by both Gordon et al.  and Lee et al. .
3.3 SAB Model calibration
For the HJ-1A/CCD sensor, εb cannot be accurately computed using the measurements provided by Hydroscat-6, due to the fact that the spectral parameters of Hydroscat-6, such as spectral resolution and wavelength, are not matched with the spectral characteristics of the HJ-1A/CCD sensor. Additionally, Lee et al.  indicated that the higher specific backscattering is achieved by smaller suspended particles, due to their larger cross-sectional area to volume ratio. This signifies that εb is a local specific parameter, which does vary with particle size. As a result, although there are various field and theoretical studies regarding spectral slope of backscattering coefficient [18, 43, 44], the values in these studies cannot be used to approximately calculate the default values of ε in the Oujiang River estuary. In this study, the ε is deemed as an “unknown” and may be computed from Eq. (7) using the non-linear least square method proposed by Chen and Quan .
By inputting the bio-optical data set containing u(λ), a(λ) and bb(λ), as shown in Fig. 2(a) and Table 1a, the optimal SAB model may be determined by using the non-linear iterative method proposed by Chen and Quan . Figure 4 shows the optimal SAB model regressed from the calibration data set collected on September 9, 10, 18, 19, 21 and 25, 2012 in the Oujiang River estuary. From the data, it is found that the SAB model initialized with the HJ-1A/CCD spectral response function is an effective predictor in deriving the total backscattering and absorption coefficient at specific wavelengths of interest from turbid coastal waters, the correlation coefficients of which are 0.9421 and 0.9265.
3.4 Evaluation of SAB model accuracy
In this section, the performance evaluation of the SAB model with the HJ-1A/CCD spectral bands is presented. The evaluation was based on comparison of the IOPs concentrations predicted by the SAB model, with IOPs measured analytically for the independent validation data set. A comparison of the measured and predicted estimates of inherent optical properties by the SAB model is presented in Fig. 5. From the results, it is found that for the range of a(560) from 0.16 to 2.89 m−1 the RMS of the a(560) prediction was 23.68%, and with the range of from 0.07 to 1.34 m−1 the RMS of the bb(550) prediction was 25.52%. This example indicates that, provided that an atmospheric correction scheme for the green bands is available, the extensive database of HJ-1A/CCD imagery may be used for the quantitative monitoring of absorption coefficients in coastal waters.
3.5 IOPs retrieval from HJ-1A/CCD imagery
3.5.1 Atmospheric correction
Due to heavy cloud cover or lack of satellite overpass, only one HJ-1A/CCD image taken from over the Oujiang River estuary was selected during the experiments. The complex path signal (with photons scattered due to air molecules, aerosols, and the air-sea interface) contributing approximately 80% to the total signal recorded at the top of the atmosphere, was effectively removed from the HJ-1A/CCD data using an improved “clear water” atmospheric correction model for broadband sensors (the ICAC model) proposed by Chen et al. . The ICAC model made the following assumptions : (1) Given the surface atmospheric pressure (to determine the value of τr(λ)) and the surface wind speed (to define the roughness of the water surface), the Rayleigh scattering contributions, even those accounting for polarization effects, may be removed using the lookup table method ; (2) The remote sensing reflectance associated with aerosol scattering at 825 nm was estimated using the “clear water” method after Rayleigh scattering correction; (3) The Angstrom exponential model was used to extrapolate the aerosol scattering contribution from 825 nm into 485 and 560 nm; and (4) The empirical spectral relationship between 485 and 560 nm and the aerosol scattering concentration retrieved from 825 nm were used to determine the exponential coefficient of the Angstrom exponential model . The satellite-derived Rrs(485) and Rrs(560) are shown in Fig. 6.
The accuracy of the atmospheric correction algorithms was evaluated through the comparison of the retrieved and observed water-leaving reflectance. The observation stations located within a ± 3 hour time window of satellite overpass and measurement were selected. The atmospheric conditions within this ± 3 hour period are reasonably stable . For the data match-up analysis, the procedure described by Bailey and Werdell  was used to produce the satellite data for comparison with in situ measurements. For a given satellite-derived remote sensing reflectance, pixels with a 3 × 3 box centered at the location of the in situ measurement were extracted, and the retrievals of the 3 × 3 pixels were averaged for the validation. In order to evaluate the accuracy of Rrs(485) and Rrs(560) retrievals, 12 field independent measurements were collected from the Oujiang River estuary within a ± 3 hour time window of satellite overpass on September 12, 2012.
Figure 7 provides comparisons of HJ-1A/CCD-derived remote sensing reflectance plotting against field measurements. It was found that the ICAC algorithm produces good performance in predicting remote sensing reflectance from HJ-1A/CCD images, the RMS values of which are <15%. The study results produced by Chen et al.  showed that the ICAB algorithm performs well in retrieving remote sensing reflectance from Landsat-5 images in Taihu Lake, the RMS values of which are <11%. By comparison, the performance of the ICAC model in Taihu Lake is slightly superior to that in the Oujiang River estuary. This is mainly because since the launching of HJ-1A satellite, the data from satellite sensor were more or less problematic because of degradation of the optics . The poor calibration would reduce the signal-to-noise ration and enhance the effect of inherent noise in the satellite-recorded signal.
In addition to these conclusions, in order to evaluate the applicability of the ICAC model in other coastal waters which have bio-optical characteristics differing from those in the Oujiang River estuary, the model was further calibrated using a data set taken in Jiaozhou Bay, on March 24, 2011. Figure 7 shows the HJ-1A/CCD-derived remote sensing reflectance plotted against the field measurements in Jiaozhou Bay. From the results it is found that when the ICAC model is used to derive remote sensing reflectance from the HJ-1A/CCD data it produces ~15% uncertainty in remote sensing reflectance retrieval. These findings indicate that the ICAC model is both stable and accurate enough for deriving remote sensing reflectance at the blue and green bands from broadband satellite data for the purpose of remote sensing application in coastal waters.
3.5.2 IOPs retrieval
Using the results from the atmospheric correction algorithm, the total absorption and backscattering coefficients were calculated using the SAB model (Fig. 8). The comparison between the satellite-derived and measured a(560) and bb(550) on September 12, 2012 in the Oujiang River estuary using the procedure proposed by Bailey and Werdell  is shown in Fig. 9. The results indicate that the respective RMS values of a(560) and bb(550) retrievals are 29.84 and 21.35%. These findings imply that, provided that an atmospheric correction scheme for the green bands is available, the extensive database of HJ-1A/CCD imagery may be used for the quantitative monitoring of specific optical properties in coastal waters.
Suspended sediment concentration plays an important role in the optical properties of estuarine environments. Sediment sources in the estuary vary widely, including imports from the ocean, retention of riverine material, and recycling of salt-marsh and tidal flat sediments by erosion . Each of these sources appears to be locally important and contributes to the bio-optical properties of the estuarine waters. Figure 8 reveals that a(560) and bb(550) in inshore areas and around islands are higher than those in other areas of the Oujiang River estuary, and the highest a(560) and bb(550) values are found in the estuarine turbidity maximum zone near the estuarine mouth. The values of a(560) and bb(550) decrease from the estuary toward all directions, with the sharpest decline to the east. These findings imply that the processes of the mass and energy exchange have a strong influence on the bio-optical properties of estuarine aquatic environments, and thus the bio-optical properties are very complex due to the complicated hydrodynamic conditions in these regions.
In order to evaluate the applicability of the SAB model in other coastal waters which possess bio-optical characteristics different from those in the Oujiang River estuary, the model was further calibrated using the data sets taken over the West Florida Shelf, USA, and the Bohai Sea, China. Due to the fact that the data sets of total backscattering coefficients for these two data sets are not available, only the specific form of the SAB model in predicting total absorption coefficient is given in Figs. 10(a)-10(b). It was found that the SAB model still provides a high accuracy in predicting the a(560) in the West Florida Shelf and Bohai Sea, despite the fact that the bio-optical properties of these two regions differ greatly from that used for model development (Table 1). However, the site-specific parameterization of ε requires further optimization, due to the different bio-optical characteristics among the Oujiang River estuary, West Florida Shelf, and Bohai Sea (Table 1 and Fig. 2). This signifies that local bio-optical information is required to improve the site-specific parameterizations of the retrieval model when the bio-optical properties are different from these used for model development. The study results indicate that the respective site-specific parameterizations of ε for the Oujiang River estuary, West Florida Shelf, and Bohai Sea are 0.228, 0.342, and 0.313. Moreover, the stability and accuracy of the SAB model in a(560) retrievals were evaluated using the validation data sets collected from the West Florida Shelf in 2003 and from the Bohai Sea in June, 2005. The results indicates (Fig. 10(c)) that the respective RMS values of the a(560) retrievals for the West Florida Shelf and Bohai Sea are 5.23 and 22.37%. Therefore, it is concluded that the SAB model may be used to derive the specific IOPs in coastal waters, despite the fact that some local bio-optical information is required to reinitialize the parameters of the SAB model.
The SAB model produces a superior performance in the a(560) retrievals in the West Florida Shelf in comparison to those in both the Oujiang River estuary and the Bohai Sea (Fig. 10(c)). On one hand, the a(485) in the West Florida Shelf (Table 1) ranges from 0.0301 to 0.3601 m−1, with the average being 0.087 m−1. In addition, the study results produced by Vargo et al.  indicate that the West Florida Shelf is a productive water. Therefore, the coastal waters in West Florida Shelf are blue coastal waters, and may be grouped into the category of productive blue coastal water. On the other hand, the a(485) in the Oujiang River estuary and Bohai Sea (Table 1) vary from 0.0359 to 12.54 m−1 and from 0.0938 to 0.7004m−1, respectively, and the respective average values are 1.7868 and 0.2983 m−1. Cui et al.  showed that due to the fact that larger rivers, including the Yellow River, pour into the Bohai Sea, carrying a large amount of inorganic and organic suspended matter, with an average annual run-off of up to 4.86 × 1010 m3 and an annual sediment transport of 1.59 × 109 tones, the optical properties of the Bohai Sea are complex, and the majority of the Bohai Sea consists of typical of Case II waters. Similar to the Bohai Sea, the bio-optical properties in the Oujiang River estuary are greatly impacted by large amounts of nutrients and other pollutants transported from land , which leads to the waters in the Oujiang River estuary being very turbid and classified as Case II waters . By comparison, the bio-optical properties of the Oujiang River estuary and Bohai Sea are more complicated than those of the West Florida Shelf. These findings imply that the SAB model may be used to derive specific optical properties from highly complicated or productive coastal waters, but the retrieval accuracy is greatly dependent on a complex degree of optical properties in aquatic environments.
Importantly, the wavelength considered in this paper is 560 nm which is only really useful for looking at turbidity and not productivity [37, 51]. Thus, it would be meaningful if the total absorption coefficient at 560 nm may be expanded to other HJ-1A/CCD visible wavelengths. Fortunately, the choice of the reference wavelength (550 nm) is arbitrary given the fact that bb(λ) at other wavelengths may be derived mathematically using power law exponent proposed by Zawada et al. . If bb(λ) spectrum is known, the a(λ) spectrum may then be easily calculated using Eq. (4). As a case study in the Oujiang River estuary, Fig. 11 shows the performance of a(485) and a(660) retrievals expanding from a(560) using the procedures proposed by Lee et al. . It is found that Lee’s procedures allow a(485) and a(660) to be predicted quite accurately (Fig. 11). They account for less than 36.62% of the RMS values in both a(485) and a(660), and thus may be used to accurately estimate a(485) and a(660) in the highly turbid waters of the Oujiang River estuary.
It is noteworthy that the ε is a site-specific parameter. It is well known that the site-specific parameter generally only appropriates to waters with optical characteristics similar to those used in the model development, thus its applicability may be quite limited, and significant errors may occur. In some complicated bio-optical waters which differ from the aquatic environment conditions for model developments used in this study, local information may still require reinitialization of the site-specific parameters of the SAB model. Another limitation of this study is that the calibration and validation data sets only contain a narrow range of optical properties of natural coastal waters, as these data sets were only taken from the Oujiang River estuary in 2012, the West Florida Shelf waters during the period of 1999-2003, and the Bohai Sea in 2005. It is insufficient to completely validate the accuracy of the model in other waters with different bio-optical properties, thus it is concluded that the SAB model should be used for specific bio-optical properties in coastal waters, although it may be essential to accordingly optimize the site-specific parameterization for the given aquatic bio-optical conditions. The researchers also suggest the calibration and validation of the algorithms based on more in situ measurements of waters with different optical properties.
Presented in this study is a total absorption and backscattering coefficients retrieval based on the observations made in September 2012 in the Oujiang River estuary, China. Then, the HJ-1A/CCD data was processed to derive the a(560) and bb(550). Finally, two independent bio-optical data sets were used to calibrate and evaluate the stability and accuracy of the SAB model in predicting the a(560) and bb(550) in coastal waters. These two bio-optical data sets were taken in the West Florida Shelf, USA, during the period of 1999-2003 and in the Bohai Sea, China, in 2005. The findings of the study are summarized as follows:
- (1) The rrs(λ) may be denoted as the function u(λ), but the parameters of that function are local specific coefficients, which vary with the bio-optical properties of the aquatic environment. The results of this study suggest that l0 = 0.0364 and l1 = 0.3484 are more suitable for the coastal waters of the Oujiang River estuary, China.
- (2) The applicability of the ICAC model with the HJ-1A/CCD band was validated in this study. Compared with the field measurements collected from the Oujiang River estuary on September 12, 2012, the remote sensing reflectance may be derived from the HJ-1A/CCD sensor within 15% RMS value.
- (3) In this study a semi-analytical model was proposed to retrieve the a(560) and bb(550) from coastal waters using HJ-1A/CCD images. The study results suggest that the SAB model may be used to retrieve the a(560) and bb(550) from coastal waters. This signifies that if the atmospheric correction scheme is available at the green bands, the HJ-1A/CCD imagery may be used for the quantitative monitoring of the a(560) and bb(550) in coastal waters (RMS<30%).
Finally, there are still some aspects of this study which require improvement in future research, e.g. although the SAB model may be used to derive total absorption and backscattering coefficients from highly turbid or productive waters, the retrieval accuracy is greatly dependent on a complex degree of optical properties of aquatic environments. The observations of this study suggest that local information or an optimized method may still require reinitialization of the site-specific parameters of the SAB model, due to the fact that the bio-optical conditions differ from those used for model development.
This study is supported by the Science Foundation for 100 Excellent Youth Geological Scholars of China Geological Survey, open fund of Key Laboratory of Marine Hydrocarbon Resources and Environmental Geology (MRE201109), High-tech Research and Development Program of China (No. 2007AA092102), and Dragon 3 Project (ID 10470). We would like to just express our gratitude to Prof. K. L. Carder (download from SeaBASS) for providing the bio-optical data set for this study. We would also like to express our gratitude to three anonymous reviewers for their useful comments and suggestions.
References and links
1. V. Vantrepotte, H. Loisel, D. Dessailly, and X. Mériaux, “Optical classification of contrasted coastal waters,” Remote Sens. Environ. 123, 306–323 (2012). [CrossRef]
2. E. H. Shadwick, H. Thomas, A. Comeau, S. E. Craig, C. W. Hunt, and J. E. Salisbury, “Air-sea CO2 fluxes on the scotian shelf: seasonal to multi-annual variability,” Biogeosciences 7(11), 3851–3867 (2010). [CrossRef]
3. S. E. Craig, C. T. Jones, W. K. W. Li, G. Lazin, E. Horne, C. Caverhill, and J. J. Cullen, “Deriving optical metrics of coastal phytoplankton biomass from ocean colour,” Remote Sens. Environ. 119, 72–83 (2012). [CrossRef]
4. R. W. Gould Jr and R. A. Arnone, “Remote sensing estimates of inherent optical properties in a coastal environment,” Remote Sens. Environ. 61(2), 290–301 (1997). [CrossRef]
5. S. Mishra and D. R. Mishra, “Normalized difference chlorophyll index: a novel model for remote estimation of chlorophyll-a concentration in turbid productive waters,” Remote Sens. Environ. 117, 394–406 (2012). [CrossRef]
6. E. Spyrakos, L. González Vilas, J. M. Torres Palenzuela, and E. D. Barton, “Remote sensing chlorophyll a of optically complex waters (rias Baixas, NW Spain): Application of a regionally specific chlorophyll a algorithm for MERIS full resolution data during an upwelling cycle,” Remote Sens. Environ. 115(10), 2471–2485 (2011). [CrossRef]
7. C. Hu, R. Luerssen, F. E. Muller-Karger, K. L. Carder, and C. A. Heil, “On the remote monitoring of Karenia brevis blooms of the west Florida shelf,” Cont. Shelf Res. 28(1), 159–176 (2008). [CrossRef]
8. M. W. Matthews, S. Bernard, and L. Robertson, “An algorithm for detecting trophic status (chlorophyll-a), cyanobacterial-dominance, surface scums and floating vegetation in inland and coastal waters,” Remote Sens. Environ. 124, 637–652 (2012). [CrossRef]
9. G. C. Chang and R. W. Gould, “Comparisons of optical properties of the coastal ocean derived from satellite ocean color and in situ measurements,” Opt. Express 14(22), 10149–10163 (2006). [CrossRef] [PubMed]
10. S. A. Garver and D. Siegel, “Inherent optical properties inversion of ocean color spectral and its biogeochemical interpretation. 1. Time series from the Sargasso Sea,” J. Geophys. Res. 102(C8), 18607–18625 (1997). [CrossRef]
11. F. Mélin, J.-F. Berthon, and G. Zibordi, “Assessment of apparent and inherent optical properties derived from SeaWiFS with field data,” Remote Sens. Environ. 97(4), 540–553 (2005). [CrossRef]
12. V. Volpe, S. Silvestri, and M. Marani, “Remote sensing retrieval suspended sediment concentration in shallow waters,” Remote Sens. Environ. 115(1), 44–54 (2011). [CrossRef]
13. J. C. Ohlmannn, D. A. Siegel, and C. D. Mobley, “Ocean radiant heating: Part I. Optical influences,” J. Phys. Oceanogr. 30(8), 1833–1848 (2000). [CrossRef]
14. T. S. Moore, J. W. Campbell, and M. D. Dowell, “A class-based approach to characterizing and mapping the uncertainty of the MODIS ocean chlorophyll product,” Remote Sens. Environ. 113(11), 2424–2430 (2009). [CrossRef]
15. Z. P. Lee, K. L. Carder, and R. A. Arnone, “Deriving inherent optical properties from water color: a multiband quasi-analytical algorithm for optically deep waters,” Appl. Opt. 41(27), 5755–5772 (2002). [CrossRef] [PubMed]
16. Z. P. Lee, J. Sandidge, and M. R. Zhang, “Ocean-color inversion: a combined approach by analytical solution and neural networks,” Presented at Ocean Remote Sensing and Imaging, SPIE, 3–8 August 2003, San Diego, CA, USA., 2003.
17. K. L. Carder, F. R. Chen, Z. P. Lee, S. K. Hawes, and J. P. Cannizzaro, MODIS Ocean Science Team Algorithm Theoretical Basis Document: Case 2 chlorophyll a, ATBD 19, Version 7., 2003.
18. T. J. Smyth, G. F. Moore, T. Hirata, and J. Aiken, “Semi-analytical model for the derivation of ocean color inherent optical properties: description, implementation, and performance assessment,” Appl. Opt. 45(31), 8116–8131 (2006). [CrossRef] [PubMed]
21. L. Gao, D. Fan, D. Li, and J. Cai, “Fluorescence characteristics of chromophoric dissolved organic matter in shallow water along the Zhejiang coasts, southeast China,” Mar. Environ. Res. 69(3), 187–197 (2010). [CrossRef] [PubMed]
22. J. Chen, W. Quan, Z. Wen, and T. Cui, “A simple “clear water” atmospheric correction algorithm for Landsat-5 sensors,” Int. J. Remote Sens. (2012), doi:. [CrossRef]
23. J. R. V. Zanefeld, J. C. Kitchen, A. Bricaud, and C. C. Moore, “Analysis of in-situ spectral absorption meter data,” Ocean Optics XI, Proceedings, SPIE, 1750 (1992).
24. J. L. Mueller and G. S. Fargion, “Ocean optics protocols for satellite ocean color sensor validation,” SeaWiFS Technical Report Series, Revision 3(Part II), 171–179 (2002).
25. Q. Wang, C. Q. Wu, Q. Li, and J. S. Li, “Chinese HJ-1A/B satellites and data characteristics,” Sci. China Earth Sci. 53(S1), 51–57 (2010). [CrossRef]
26. K. Hu, F. Chen, and S. Liang, “Application of HJ-1B Data in Monitoring Water Surface Temperature,” Proc. Environ. Sci. Part C, 2042–2049 (2011).
27. J. Li, H. Li, B. Shen, and Y. Li, “Effect of non-point source pollution on water quality of the Weihe River,” Int. J. Sediment Res. 26(1), 50–61 (2011). [CrossRef]
28. C. M. Hu, F. E. Muller-Karger, S. Andrefouet, and K. L. Carder, “Atmospheric correction and cross-calibration of Landsat-7/ETM+ imagery over aquatic environments: A multiplantform approach using SeaWiFS/MODIS,” Remote Sens. Environ. 78(1-2), 99–107 (2001). [CrossRef]
29. Z. P. Lee, K. L. Carder, R. G. Steward, T. G. Peacock, C. O. Davis, and J. S. Patch, “An empirical ocean color algorithm for light absorption coefficients of optically deep waters,” J. Geophys. Res. 103(27), 967–978 (1998).
31. C. D. Mobley, Hydrolight 6.0 User's Guide, Final Report, SRI International (Menlo Park, Calif), (2008).
32. Z. P. Lee, K. L. Carder, C. D. Mobley, R. G. Steward, and J. S. Patch, “Hyperspectral remote sensing for shallow waters. 2. deriving bottom depths and water properties by optimization,” Appl. Opt. 38(18), 3831–3843 (1999). [CrossRef] [PubMed]
33. H. R. Gordon, O. B. Brown, and M. M. Jacobs, “Computed relationships between the inherent and apparent optical properties of a Flat Homogeneous Ocean,” Appl. Opt. 14(2), 417–427 (1975). [CrossRef] [PubMed]
34. H. R. Gordon, O. B. Brown, R. H. Evans, J. W. Brown, R. C. Smith, K. S. Baker, and D. K. Clark, “A Semianalytic Radiance Model of Ocean Color,” J. Geophys. Res. 93(D9), 10909–10924 (1988). [CrossRef]
35. Z. P. Lee and K. L. Carder, “Absorption spectrum of phytoplankton pigments derived from hyperspectral remote-sensing reflectance,” Remote Sens. Environ. 89(3), 361–368 (2004). [CrossRef]
37. D. G. Bowers and C. E. Binding, “The optical properties of mineral suspended particles: A review and synthesis,” Estuar. Coast. Shelf 67(1-2), 219–230 (2006). [CrossRef]
38. A. Bricaud, A. Morel, and L. Prieur, “Absorption by dissolved organic matter of sea (yellow substance) in the UV and visible domains,” Limnol. Oceanogr. 40, 393–410 (1981). [CrossRef]
39. M. Ondrusek, E. Stengel, C. S. Kinkade, R. L. Vogel, P. Keegstra, C. Hunter, and C. Kim, “The development of a new optical total suspended matter algorithm for the Chesapeake Bay,” Remote Sens. Environ. 119, 243–254 (2012). [CrossRef]
40. A. A. Gitelson and M. N. Merzlyak, “Spectral Reflectance Changes Associated with Autumn Senescence of Aesculus Hippocastanum L. and Acer Platanoides L. Leaves. Spectral Features and Relation to Chlorophyll Estimation,” J. Plant Physiol. 143(3), 286–292 (1994). [CrossRef]
41. A. Morel and L. Prieur, “Analysis of variances in ocean color,” Limnol. Oceanogr. 22(4), 709–722 (1977). [CrossRef]
42. E. M. Lee, D. G. Bowers, and E. Kyte, “Remote sensing of temporal and spatial patterns of suspended particle size in the Irish Sea in relation to the Kolmogorov microscale,” Cont. Shelf Res. 29(9), 1213–1225 (2009). [CrossRef]
43. E. J. D'Sa, R. L. Miller, and B. A. McKee, “Suspended particulate matter dynamics in coastal waters from ocean color: Application to the northern Gulf of Mexico,” Geophys. Res. Lett. 34(23), L23611 (2007). [CrossRef]
44. M. W. Zhang, J. W. Tang, Q. Dong, Q. T. Song, and J. Ding, “Retrieval of total suspended matter concentration in the Yellow and East China Seas from MODIS imagery,” Remote Sens. Environ. 114(2), 392–403 (2010). [CrossRef]
45. H. R. Gordon and M. H. Wang, “Retrieval of water-leaving radiance and aerosol optical thickness over the oceans with SeaWiFS: a preliminary algorithm,” Appl. Opt. 33(3), 443–452 (1994). [CrossRef] [PubMed]
46. S. W. Bailey and P. J. Werdell, “A multi-sensor approach for the on-orbit validation of ocean color satellite data products,” Remote Sens. Environ. 102(1-2), 12–23 (2006). [CrossRef]
47. X. L. Yu and Z. C. Wu, “The comparison between hj satellite's ccd sensors field calibration and cross calibration,” Chin. J. Sen. Act. 24, 1435–1439 (2011).
48. J. O. Blanton, H. Seim, C. Alexander, J. Amft, and G. Kineke, “Transport of salt and suspended sediments in a curving channel of a coastal plain estuary: Satilla River, GA,” Estuar. Coast. Shelf Sci. 57(5-6), 993–1006 (2003). [CrossRef]
49. G. A. Vargo, C. A. Heil, K. A. Fanning, L. K. Dixon, M. B. Neely, K. Lester, D. Ault, S. Murasko, J. Havens, J. Walsh, and S. Bell, “Nutrient availability in support of Karenia brevis blooms on the central West Florida Shelf: What keeps Karenia blooming?” Cont. Shelf Res. 28(1), 73–98 (2008). [CrossRef]
50. T. Cui, J. Zhang, S. Groom, L. Sun, T. Smyth, and S. Sathyendranath, “Validation of MERIS ocean-color products in the Bohai Sea: A case study for turbid coastal waters,” Remote Sens. Environ. 114(10), 2326–2336 (2010). [CrossRef]
51. R. P. Bukata, J. H. Jerome, K. Y. Kondratyev, and D. V. Pozdnyakov, Optical properties and remote sensing of inland and coastal waters 1st ed. (New York, CRC Press, 1995).
52. D. G. Zawada, C. M. Hu, T. Y. Clayton, Z. Q. Chen, J. C. Brock, and F. E. Muller-Karger, “Remote sensing of particle backscattering in Chesapeake Bay: A 6-year SeaWiFS retrospective view,” Estuar. Coast. Shelf 73(3-4), 792–806 (2007). [CrossRef]