An Inherent Optical Properties Data Processing System for Achieving Consistent Ocean Color Products From Different Ocean Color Satellites

Abstract

We used field measurements and multimission satellite data to evaluate how well an inherent optical properties (IOPs) data processing system performed at correcting the residual error of the atmospheric correction in satellite remote sensing reflectance (R rs ) and how well the system simultaneously minimized intermission biases between different remote sensing systems. We developed the IOPs data processing system as a semianalytical algorithm called IDAS. Our results show that IDAS generates accurate and consistent IOPs products from two ocean color missions: Sea-viewing Wide Field-of-View Sensor (SeaWiFS) and Moderate Resolution Imaging Spectroradiometer Aqua (MODISA). Specifically, with "high-quality" SeaWiFS and MODISA R rs data, IDAS provided temporally consistent IOPs products for the oligotrophic open ocean resulting in an annual mean intermission difference of less than 3%, which is significantly lower than what a quasi-analytical algorithm (QAA) provided. We used IDAS to generate a long time series of b b (555) from the Northwest Atlantic Subtropical Gyre using SeaWiFS (1998SeaWiFS ( to 2002 and MODISA (2003MODISA ( to 2017 images. Our results show that the IDAS-derived annual b b (555) decreased monotonically by 2.81% per decade from 1998 to 2017. Comparing the IDAS-generated annual trend for b b (555) to the same data processed with the QAA algorithm, we found that the QAA results differed because of impacts of the residual errors of the atmospheric correction and intermission biases. The differences in the annual trends existed despite the same temporal changing patterns of in situ particulate organic carbon existing in the Sargasso Sea and in the satellite chlorophyll-a concentration in the Northwest Atlantic Subtropical Gyre.

An Improved Inherent Optical Properties Data Processing System for Residual Error Correction in Turbid Natural Waters

“…Because the IDASv2 algorithm was designed to semianalytically derive IOPs from satellite Rrs, and, at the same time, to account for the residual errors in the satellite Rrs [20,21], the IDASv2 algorithm tolerates noise better than the QAA algorithm does. For example, many points in the scatterplots of the QAA results deviated from the 1:1 line, but the IDASv2 algorithm results tended to gather around the 1:1 line.…”

“…Due to imperfect data processing systems and sensor performance, satellite-observed Rrs always includes substantial residual errors and inter-mission biases [21], which can cause inconsistencies in the IOPs derived from different ocean color data. The IOPs derived from the data of one satellite-observed Rrs would be highly consistent with the IOPs derived from other satellite-observed Rrs data if the IDASv2 algorithm could absorb the residual errors and inter-mission bias in the data.…”

“…As a result, for practical applications we always have to adjust the retrieval procedures according to the satellites' spectral characteristics. In addition, it is well known that different algorithms always respond differently to the residual errors and inter-mission bias in satellite data [20], which can cause inconsistent outputs among the different algorithms even for the same water-leaving signals [21]. This is a disadvantage for generating smooth and long-term series of IOPs data on global or local scales.…”

“…Second-generation satellites have four "common bands" around 443, 490, 551, and 670 nm at visible wavelengths. By using the remote sensing reflectance (Rrs) in these four "common bands" as inputs, a quasi-analytical algorithm (QAA) [18] and an IOP data processing system (Version 1, abbreviated to IDASv1) [21] can estimate the IOPs from any visible band as long as the Rrs for this visible band is known. For most of the open ocean, because Rrs(551) and Rrs(670) are relatively stable, and Rrs(443) and Rrs(490) are sensitive to phytoplankton and covarying optically significant constituents [22], it is possible to effectively derive IOPs from Rrs at these four "common bands".…”

“…In the QAA algorithm, because the Rrs data contained substantial uncertainties, when a at reference wavelength (λ0) is "fixed" with high accuracy, the uncertainties in Rrs(λ0) propagates to bb(λ0) following the IOPs-Rrs relationship [27]. Thus, it is impossible to simultaneously obtain highly accurate absorption and backscattering coefficients from satellite Rrs that is contaminated with errors using the QAA or IDASv1 algorithms unless the uncertainties are removed by correcting for the residual error [21]. Furthermore, to accurately derive absorption and backscattering coefficients, it is necessary to reconstruct the method of the QAA or IDASv1 algorithm.…”