A Semi-analytical Algorithm for Remotely Estimating Suspended Particulate Matter of Inland Waters

Abstract

Monitoring total suspended particulate matter (TSM) by remoting sensing is particularly challenging for inland waters. Although several models have been developed for TSM inversion, their accuracy varies evidently due to variability in inherent optical properties (IOPs) of the inland waters. To address this issue, we proposed a semi-analytical algorithm, which is based on Quasi-Analytical Algorithm (QAA), for remotely estimating TSM concentration in Maozhou River, located in Shenzhen, China. An innovative loss function was defined to represent homogeneity and rationality of IOPs within a small geographic region in this paper. Adaptive Moment Estimation (Adam) was employed to modify the QAA model by minimizing the value of the loss function. The proposed approach requires no in-situ IOPs data, which makes it practical. Furthermore, a TSM inversion model based on Normalized Difference TSM Index (NDTI) was developed and validated. The results of higher relevance between TSM concentrations and IOPs retrieved by the Modified QAA model showed that the approach we proposed was helpful for TSM inversion. The values of MRE, MAE and RMSE showed that the accuracy of the model is acceptable for inland waters.