Assessment of remote sensing empirical algorithms to retrieve colored dissolved organic matter in the Lower Amazon River

There are many empirical algorithms described in the literature to estimate colored dissolved organic matter (CDOM) by remote sensing. When the measurements are acquired in turbid water, new algorithms might be required to better characterize regional optical characteristics. The aim of this study is the assessment of current empirical algorithms to retrieve CDOM absorption coefficient (acdom) by remote sensing in the Lower Amazon. This is a region where very turbid waters mix with clear water from tributary rivers. Surface water samples were collected concurrent with radiometric measurements in different stations along the study area, throughout the annual hydrological cycle: Rising, High, Falling and Low water. Remote sensing reflectance (R_rs) was estimated from in situ radiometric data presenting a low coefficient of variation in the range of 400-840 nm. Six empirical algorithms were used to estimate acdom (G11, K05, B05, M14, T11 and K15) from R_rs showing high values even in the green spectrum. The K15 (R2=0.57; RMSE = 0.09; p-value<0.05) showed the best performance when compared to all in situ acdom measurements. However, when the dataset is partitioned into clear water rivers and Amazon River the results are different. For clear water the algorithms presented a better performance (e.g. K15 R2=0.68; RMSE = 0.14, p-value<0.05), and no significant correlation was obtained for the Amazon River. The results suggest that other optically active constituents are interfering in the retrieve of CDOM, and it is likely to be suspended sediments. Therefore, different methodological approaches should be explored to correct for this interference.