Long Island Sound Coastal Observatory: assessment of above-water radiometric measurement uncertainties using collocated multi and hyper-spectral systems: reply to comment

Abstract

Uncertainties associated with the derivation of the exact normalized water-leaving radiance (L WN ) from an above-water radiometric system were analyzed in Harmel et al. [Appl. Opt. 50, 5842 (2011)] based on collocated hyperspectral (HyperSAS) and multispectral (SeaPRISM) systems installed on the Long Island Sound Coastal Observational (LISCO) platform. Based on a 1.5 year time series of LISCO data, uncertainty contributors in the derivation of L WN were quantified in units of unbiased relative percentage differences (URPD) by applying the different steps of the respective data processing incrementally. Results showed that discrepancy between L WN data of two systems is significantly reduced when the average total sea radiance data of SeaPRISM is used in lieu of the standard one, which utilizes only the lowest total sea radiance measurements to remove the sky glint perturbations. The Zibordi comment [Appl. Opt. 51, 3888 (2012)] rejects the conclusion that attributes the sky glint removal step as the major uncertainty contributor in the SeaPRISM processing. It then states that the observed discrepancy might be due to an increased probability of sun-glint contamination in HyperSAS measurements because of its wider field of view and longer integration time. It was also underlined that observed dispersion between the atmospheric transmittance data derived from HyperSAS and SeaPRISM measurements can be attributed to probable contamination by stray light perturbation or issues with the noncosine response of the HyperSAS irradiance sensor. Finally, it was suggested to thoroughly investigate those instrumental perturbations. In this reply, impacts of nonperfect cosine response of the irradiance sensor are shown to be relatively low (<2% on average) and therefore can only partially explain the bias in atmospheric transmittance. Additional discrepancies between the HyperSAS and SeaPRISM downwelling irradiance derivation are attributed to the presence of absorbing aerosols. Intercomparisons of the total sea radiance and nonnormalized water-leaving radiance, complementary to those discussed in the LISCO paper, are analyzed, and this analysis shows that discrepancies in normalized water-leaving radiance retrievals arise from data processing and not from instrumental uncertainty. In addition, limitations in the standard data processing to meaningfully derive normalized water-leaving radiance for various appropriate viewing configurations are discussed. It is finally advocated that the issue of sky glint perturbation correction requires further analysis based on radiative transfer computations, including refined modeling of wave slope distributions.

Hyperspectral and multispectral above-water radiometric measurements to monitor satellite data quality over coastal area

“…As it has been shown in Harmel et al 10 , the impact of the successive processing steps on the retrieval uncertainties can be summarized as follows: (i) the sun glint removal step generates unbiased uncertainties of about 2.5%, with a slight positive bias in HyperSAS data, (ii) the sky glint removal step generates unbiased uncertainties of about 6%, (iii) the viewing angle dependence correction improves the data intercomparison by reducing the unbiased uncertainties by more than 1.5%, (iv) the atmospheric effect normalization generates approximately 5% of unbiased uncertainties, and induces a non-negligible bias, especially at the shorter wavelength, most likely due to an insufficiently accurate atmospheric transmittance derivation in the SeaPRISM processing. Ultimately, the exact normalized water-leaving radiances were then retrieved with an overall uncertainty of 19.5% and a positive bias of about 0.09 mW cm -2 sr -1 nm -1 in HyperSAS data ( Figure 5).…”

“…In conclusion, the significant HyperSAS data accuracy has been shown on the basis of statistics of daily measurements gathered over more than one year (i.e. October 2009 until January 2011, see Harmel et al 10 for a discussion on the seasonal variations of this accuracy) exhibiting uncertainties below 5% within consistent spectral and time ranges which are suitable for ocean color radiometry satellite validation activities. …”

“…Details on the multispectral data processing can be found in Zibordi et al 6 ; in addition the specificities of the hyperspectral data processing are detailed in Harmel et al 10 .…”

“…These fluctuations introduce geophysical noise that needs to be removed from SeaPRISM and HyperSAS data, respectively. Based on the retrieval scheme developed for SeaPRISM 4,9 , a hyperspectral-based procedure has been implemented to derive the normalized water-leaving radiance L WN from HyperSAS measurements 10 . Thanks to the ability of LISCO SeaPRISM and HyperSAS to provide two collocated and coincident datasets, the consistency and the efficacy of the above-water data processing were assessed over a period of more than one year encompassing the full natural annual variability of atmospheric and water conditions.…”

Impact of Aerosol Model Selection on Water-Leaving Radiance Retrievals from Satellite Ocean Color Imagery

“…Zibordi et al concluded that the overall n L w uncertainty budget, which is computed as the quadrature sum of the various individual, independent sources of uncertainty, indicates values typically below 5% in the 412-551 nm spectral range and approximately 8% at 667 nm, mostly because of environmental (sea surface) perturbations [30]. Other research has been conducted assessing and improving sky and sun glint methodologies, but the overall approximation of spectral AERONET-OC n L w uncertainties remains around 5% [31][32][33]. Also, AERONET-OC data are produced at wavelengths which are slightly different from site to site, as well as slightly different from the wavelengths used by MODIS and SeaWiFS.…”