Publications

Mitchell, Catherine; Cunningham, Alex; McKee, David (2014). Remote sensing of shelf sea optical properties: Evaluation of a quasi-analytical approach for the Irish Sea. REMOTE SENSING OF ENVIRONMENT, 143, 142-153.

Abstract
A quasi-analytical approach to deriving the coefficients of absorption, a(lambda), backscattering, b(b)(lambda), and the attenuation of planar irradiance, K-d(lambda), from sub-surface remote sensing reflectance, r(rs) (lambda), was investigated for the optically complex (Case 2) waters of the Irish Sea. The algorithms of Lee et al. (2005b, Journal of Geophysical Research Oceans, 110, C02017) were tuned for this region using radiative transfer calculations which incorporated locally determined specific inherent optical properties and constituent concentrations. The optimised algorithms were then applied to r(rs) spectra derived from profiling radiometry for 145 stations in the Irish Sea and adjacent waters, and their outputs compared with inherent optical property measurements for these stations. For the modelled data, recoveries of, a(lambda), b(b)(lambda) and K-d(lambda) in six wavebands between 412 nm and 667 nm had mean percentage errors below 3% and standard deviations of around 6%. For in situ data, the algorithms tended to over-estimate a(lambda) and under estimate b(b)(lambda) relative to measured values, but a contributing factor could be the difficulty of measuring optical properties accurately in turbid waters. Retrieved values of K-d(lambda) were highly correlated with those measured in situ, with regression slopes in the best-performing bands of 1.05 at 442 nm (r(2) = 0.91) and 0.92 at 488 nm (r(2) = 0.96). Three previously published relationships between satellite-derived optical properties at 490 nm and euphotic depth, z(1%PAR), produced roughly equivalent results with root mean square percentage errors, relative to in situ measurements, of around 20%. On the logarithmic scales usually employed in remote sensing, values of a(488), b(b)(488),K-d(488) and z(1%PAR) recovered using the optimised algorithms bore strong linear relationships to in situ measurements, with least square regression slopes in the range 0.89 to 1.03 and r(2) values from 0.87 to 0.94. Application of these algorithms to a MODIS image of the Irish Sea in early summer (25th May 2013) revealed marked spatial variations in optical properties which were well correlated with known patterns of tidal stirring in the region. (C) 2014 Elsevier Inc. All rights reserved.

DOI:
10.1016/j.rse.2013.12.011

ISSN:
0034-4257; 1879-0704