Estimation of Colored Dissolved Organic Matter From Landsat-8 Imagery for Complex Inland Water: Case Study of Lake Huron

Abstract

Spectral reflectance data, including irradiance reflectance (R <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">t</sub> ) and remote sensing reflectance (R <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">rs</sub> , sr <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">-1</sup> ), and colored dissolved organic matter (CDOM) absorption coefficients a <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">CDOM</sub> (440), were collected in the Saginaw River and Kawkawlin River plume regions of Lake Huron. We developed an empirical band ratio algorithm to derive a <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">CDOM</sub> (440) that could be directly applicable to Landsat-8 imagery. A model ranking method is used to determine the best band ratios as well as their empirical functions. One problem of previous CDOM estimations from Landsat imagery is that they usually use R <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">t</sub> or R <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">t</sub> /π rather than the real R <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">rs</sub> as the input data, but as a result of our study, algorithms derived using R <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">rs</sub> performed much better than using Rt. The green/red band ratio gave the best accuracies by fitting with power and exponential models (power model: R <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> = 0.819 and RMSE = 0.889 m <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">-1</sup> and exponential model: R <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> = 0.829 and RMSE = 0.863 m <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">-1</sup> ). The power and exponential models were further validated using an independent data group, achieving excellent results with the RMSE of 0.642 and 0.504 m <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">-1</sup> , respectively. The application of the exponential model a <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">CDOM</sub> (440) = 40.75e <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">-2.463x</sup> , x = R <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">rs</sub> (B3)/R <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">rs</sub> (B4) on two Landsat-8 images demonstrated that estimating CDOM from to Landsat-8 imagery has potential applications for monitoring water quality in long term and large scale.