Ocean water-leaving radiance spectra extending from the visible (blue) to the near-infrared (NIR) derived from Moderate Resolution Imaging Spectroradiometer (MODIS) measurements are reported here for the China east coastal region. The China east coastal region contains some of the most consistently highly turbid coastal waters to be found anywhere. For turbid waters, standard MODIS data processing often produces significant errors in the derived ocean color products due to significant ocean water-leaving radiance contributions at the two NIR bands. In this paper, we demonstrate an atmospheric correction approach using MODIS short wave infrared (SWIR) bands for deriving more accurate ocean color products in coastal regions. The ocean is generally still black at the SWIR bands even for very turbid waters due to much stronger water absorption at the SWIR wavelengths. The approach has been tested for and applied to various coastal regions (e.g., the U.S. and China east coastal regions) using MODIS data. In situ data collected along the China east coastal region have been used to validate performance of the algorithm. Using MODIS Aqua measurements from July of 2002 to December of 2005, seasonal variations in the visible and the NIR ocean water-leaving radiances for the Hangzhou Bay in the China east coastal region are derived and analyzed. Very significant ocean NIR contributions along various locations in the China east coastal region (e.g., Hangzhou Bay) are observed.
 For global open ocean waters, both the Sea-viewing Wide Field-of-view Sensor (SeaWiFS) and the Moderate Resolution Imaging Spectroradiometer (MODIS) have been producing high quality ocean color products [Bailey and Werdell, 2006; McClain et al., 2004]. These products have been used by scientists worldwide to study and understand global and regional ocean variability and changes, as well as their effects on climatic processes. However, there are still some outstanding issues related to the performance of the atmospheric correction in the coastal regions. In particular, the sensor-derived ocean water-leaving radiances at the blue wavelengths (e.g., 412 nm) are often biased low and sometimes even go negative in coastal regions. This problem often results from turbid waters, which have significant ocean radiance contributions at the near-infrared (NIR) bands [Ruddick et al., 2000; Siegel et al., 2000; Wang and Shi, 2005]. For the ocean-atmosphere system, the top-of-atmosphere (TOA) reflectance, ρt (λ), measured by the satellite sensor can be written as a linear sum from various contributions (ignore the whitecap and sun glint terms for brevity) [Gordon and Wang, 1994]:
where ρr(λ), ρA(λ), and ρw(λ) are the reflectance contributions from molecules (Rayleigh scattering), aerosols (including Rayleigh-aerosol interactions), and ocean waters, respectively. t(λ) is the diffuse transmittance of the atmosphere. The important part of the ocean color data processing is to derive the water-leaving reflectance ρw(λ) by removing atmosphere and ocean surface effects, ρr(λ) and ρA(λ) in equation (1) [Gordon, 1997; Gordon and Wang, 1994]. In the current MODIS ocean color data processing, the ocean is basically assumed to be black at the two NIR bands (748 and 869 nm) for the open ocean with modifications for productive ocean waters. However, the NIR black ocean assumption (and modification) is invalid for turbid waters, leading to significant errors in MODIS-derived ocean color products. Recently, Wang and Shi  demonstrated an approach to derive the NIR ocean contributions for turbid waters using the MODIS short wave infrared (SWIR) bands, and Wang  proposed the SWIR atmospheric correction for ocean color remote sensing in coastal regions. At the SWIR wavelengths, ocean water has much stronger absorption than that at the NIR bands [Hale and Querry, 1973], thus the black ocean assumption is generally valid at the SWIR bands even for very turbid ocean waters. The SWIR atmospheric correction approach has been developed and integrated into the MODIS ocean color data processing [Wang, 2007], and tested for the U.S. east and west coastal regions. In this paper, we report ocean color products derived from MODIS Aqua measurements in the China east coastal region using the SWIR atmospheric correction algorithm. We show that, in this region, very significant ocean NIR contributions are evident all year around. In fact, the China east coastal region contains some of the most consistently turbid coastal waters to be found anywhere. Some in situ data collected along the China east coastal region are also provided for the purpose of algorithm validation.
2. Ocean Color Products From the China East Coastal Region
2.1. MODIS Aqua Measurements
 As indicated above, along the China east coastal region, ocean waters are highly turbid nearly all the time. Very high concentrations of suspended sediment are consistently present along the various China east coastal regions, e.g., the Hangzhou Bay, the Yangtze River estuary, as well as their adjacent waters [Chen et al., 2003]. Figure 1a provides an example of a MODIS Aqua true color image along the China east coastal region on October 19, 2003. This image shows very turbid waters along the China east coastal region, demonstrating the impact of large concentrations of suspended sediment on the color of these waters (grayish-yellow ocean waters). Particularly, the China coastal regions along the Hangzhou Bay, the Yangtze River estuary, and the ocean to the north of the Yangtze River mouth show extremely high sediment loadings [Chen et al., 2003]. MODIS measurements in the region show reasonably consistent optical and biological characteristics on a year-around basis. In Figure 1a, the four marked locations indicate where in situ optical data were collected in the spring and fall of 2003 (see discussions in section 2.3).
2.2. MODIS Ocean Color Products
 MODIS Aqua data have been processed for the China east coastal region using the SWIR atmospheric correction algorithm. Specifically, using the same 12 aerosol models as for the current MODIS algorithm, the aerosol lookup tables were generated for the MODIS SWIR bands [Wang, 2007]. The algorithm is basically operated in the same way as that of Gordon and Wang , but replacing the two NIR bands (748 and 869 nm) with the two SWIR bands at 1240 and 2130 nm for atmospheric correction. For the data processing, the new SWIR cloud masking scheme has been used [Wang and Shi, 2006]. The MODIS-derived products include the normalized water-leaving radiance spectra nLw(λ) (or [Lw(λ)]N, or equivalently, the normalized water-leaving reflectance [ρw(λ)]N), chlorophyll-a concentration, and aerosol optical property data, e.g., aerosol optical thickness. Definitions of the normalized water-leaving radiance nLw(λ) and reflectance [ρw(λ)]N can be found in the work of Gordon .
Figures 1b–1f provide examples of the MODIS-derived ocean color and aerosol products along the China east coastal region, corresponding to MODIS data acquired on October 19, 2003. Figures 1b–1f are images of the MODIS-derived chlorophyll-a (Chl-a), normalized water-leaving radiance nLw(λ) at wavelengths 443, 531, and 869 nm, and aerosol optical thickness at 869 nm, τa(869), respectively. The Chl-a concentration (Figure 1b) is scaled logarithmically from 0.1 to 32 (mg/m3), while the values of nLw(λ) for bands 443 and 531 nm (Figures 1c and 1d) are scaled linearly from −1 to 6 (mW cm−2μm−1 sr−1) and from −0.5 to 2 (mW cm−2μm−1 sr−1) for the 869 nm band (Figure 1e). Value of τa(869) in Figure 1f is scaled from 0 to 0.3. It is noted that, along the China east coastal region, the MODIS nLw(λ) value increases with increase of the wavelength from the blue to the green band, and nLw(λ) values for some regions actually peak at the red band (see results in section 2.3). This is a typical characteristic of the sediment-dominated waters. The NIR ocean contributions are quite significant, e.g., at the Hangzhou Bay region the average nLw(λ) values can reach to ∼2.5 (mW cm−2μm−1 sr−1) at 859 nm in the boreal winter (see results in section 2.4). However, for the derived Chl-a values, there may have uncertainties relating to the model limitations for such complex waters. Results in Figure 1 show no obvious correlations between MODIS-derived nLw(λ) and τa(869) data. In addition, aerosol Ångström exponent data have been analyzed and shown no correlations with ocean color products.
2.3. MODIS Data Compared With In Situ Measurements
 During the spring and fall of 2003, there were extensive field campaigns along the East China Sea and the Yellow Sea regions for purposes of collecting various in situ physical, biological, and optical ocean data [Tang et al., 2004]. In particular, in situ ocean water-leaving reflectance spectrum data from 350–1050 nm were measured using an Analytical Spectral Devices, Inc. (ASD) FieldSpec Dual UV/VNIR spectrometer, for which the instrument spectral sampling interval is 1.4 nm for its entire spectral coverage (350–1050 nm). The in situ data collection and processing were carried out following the procedures outlined in the NASA ocean optics protocols [Mueller and Fargion, 2002]. Specifically, in the data processing, the radiance that is contributed by the ocean surface reflection from the sky radiance to the instrument detector has been calculated and removed. Here, we compare the MODIS-derived [ρw(λ)]N spectra with those from the in situ measurements. Figure 2 provides four examples of these validation comparisons. Figures 2a–2d are [ρw(λ)]N data that were collected at the locations indicated in Figure 1a (from the North to the South), i.e., the corresponding latitudes for data obtained in Figures 2a–2d are 36°N (March 22, 2003), 33°N (April 5, 2003), 31.5°N (September 25, 2003), and 30.5°N (September 23, 2003), respectively. In Figure 2, the MODIS [ρw(λ)]N data were obtained by averaging of 9 × 9 pixels surrounding the data location.
 The results in Figure 2 show that the SWIR atmospheric correction performed reasonably well for these turbid ocean waters. Indeed, both the in situ and MODIS data show significantly high NIR ocean water-leaving reflectance values, i.e., the NIR [ρw(λ)]N values range from ∼0.3%–1.0%. For these cases, the MODIS-derived [ρw(λ)]N at the visible bands compare quite well with the in situ measurements. Figure 3 shows comparisons between the MODIS and in situ measured [ρw(λ)]N values at the selected eight MODIS wavelengths. These comparisons include in situ data acquired in the China east coastal regions for the time periods of March 22–April 6 and September 19–23 in 2003 with a total of 21 measurements. Although there are some variations (noise), the results in Figure 3 show a reasonably good matchup between MODIS and in situ reflectance measurements. These are indicated by a linear fit for all data sets with a slope of 0.896, intercept (Int) of 0.005, and correlation coefficient (R) of 0.951. Results of linear fits for spectral bands 443, 488, 531, 645, and 859 nm are also shown in the plot, indicating a little more noise at the blue than green bands and excellent results at the NIR band.
2.4. Seasonal Variation of the MODIS nLw(λ) in the Hangzhou Bay Region
 Along the China east coastal region, MODIS Aqua data have been processed using the SWIR atmospheric correction algorithm for the time period from July of 2002 to December of 2005. These data may be used to study the seasonal variation of the ocean optical and biological properties in the region. Figure 4 shows the MODIS seasonal averages of nLw(λ) for the boreal winter (December, January, and February), spring (March, April, and May), summer (June, July, and August), and fall (September, October, and November) as a function of wavelength for a location inside of Hangzhou Bay (30.5°N and 121.8°E) indicated in Figure 1b. Note that nLw(λ) at bands 667, 748, and 869 nm are frequently subject to sensor saturation in the Hangzhou Bay, thus the results for these bands are not shown here. These results (Figure 4) show that in the Hangzhou Bay region the nLw(λ) values often peaked at the red spectral band. The highest nLw(λ) at the red and NIR bands are observed in the boreal winter, while the lowest nLw(λ) values at the red and NIR bands appeared in the boreal summer. These results are consistent with other studies in which significantly higher suspended sediment concentration is observed in the boreal winter than that in the boreal summer for the Hangzhou Bay region due to seasonal monsoon winds and the ocean surface layer mixing effects [Chen et al., 2003]. High loading of the suspended sediment concentration leads to a significantly large backscattering by suspended sediment particles, and thus the large water-leaving radiances. It is interesting to note that in the winter season the average NIR nLw(λ) value at 859 nm reaches ∼2.5 (mW cm−2μm−1 sr−1), significantly larger than the nLw(λ) values at the blue bands. Average aerosol optical thicknesses at 859 nm for four seasons are also presented in the plot. Obviously, for retrieval of the aerosol optical properties over oceans, the NIR ocean contributions need to be accurately accounted for, in particular, in coastal regions.
 We have derived the MODIS ocean water-leaving radiance spectra along the China east coastal region using the SWIR atmospheric correction algorithm and compared the MODIS products with those from in situ measurements. In the China east coastal region, the ocean waters are consistently highly turbid and have extremely large NIR water-leaving radiances. Therefore, the standard MODIS data processing often fails to produce valid ocean color products in these regions. The validation results from this effort show a reasonably good agreement between satellite-derived values and in situ data when using the SWIR atmospheric correction algorithm. Therefore, we have demonstrated that the SWIR atmospheric correction approach produces reasonably good quality radiance/reflectance estimates in optically complex turbid ocean waters. It is worth noting that, for the remote retrieval of the aerosol optical properties over oceans, the effects of the NIR ocean contributions need to be accurately accounted for.
 This research was supported by the NASA NPP and MODIS grants NNG05HL40I and NNG05HL35I (MW and WS), and the NOAA Internal Government Studies (IGS) C2NF1NP (MW). The in situ data collection in the China east coastal region was supported by the China HY-1 satellite program and the China National High Technology 863 Program (project 2001AA636010) (JT). The views, opinions, and findings contained in this paper are those of the authors (MW and WS) and should not be construed as an official NOAA or U.S. Government position, policy, or decision.