A regional remote sensing algorithm for total suspended matter in the East China Sea

Abstract

The East China Sea (ECS) is well known for its high concentration of total suspended matter (TSM). Some regions of the ECS have concentrations higher than 5000gm−3, exceeding the valid ranges of many TSM remote sensing algorithms. To overcome the limitation of the existing algorithms, a new TSM model, the “complex proxy TSM model” (CPTSM), is developed in this study. The model is established on the basis of a complex proxy of remote sensing reflectance. The proxy is designed to convert the non-linear relationship between reflectance and TSM to a quasi-linear function over the entire range of TSM concentrations. This proxy is deduced from four indices defined by combinations of the reflectance at different bands. The four indices take advantage of the different relationships between the band combinations of the reflectance and total TSM concentrations. The band selections and model parameters are based on correlation coefficients and regression analysis between the indices and TSM. The results show that the correlation coefficient of 0.912 between the proxy and TSM is higher than that between any individual index and TSM. To validate the CPTSM model, TSM, turbidity, and reflectance data were collected in the ECS during 4 cruises in 2006 and 2007. The actual TSM concentration was measured by weighing the samples collected on filter papers. Turbidity was measured by a Seapoint Turbidity Meter. The turbidity data with values higher than 750FTU were re-calibrated using an empirical equation. All turbidity values were converted to TSM concentrations using a linear equation. The in situ reflectance was measured using the above-water method at 459 stations and the in-water method at 146 stations. A total of 87 pairs of reflectance measured by both methods were used for inter-comparison with a relative difference of 4.5%. The reflectance values were used to retrieve TSM concentrations using the CPTSM model. A comparison with in situ measurements gave a mean relative error of 23%. Applying the CPTSM model to the Sea-viewing Wide Field-of-view Sensor (SeaWiFS) data and analyzing the errors from a match-up dataset of SeaWiFS and in situ data, we found that the average relative error was 24.5%. We propose to use the CPTSM model to map TSM concentrations from satellite data in the ECS.