Combining two water type classification schemes for semi-analytical estimation of suspended particulate matter concentrations in various water bodies

Abstract

Retrieval of suspended particulate matter concentration (SPM) from remote-sensing reflectance (Rrs) is useful for frequent and widespread monitoring of water quality. However, Rrs values vary not only with SPM but also with particle composition (organic-dominated or mineral-dominated) and colored dissolved organic matter (CDOM), making it difficult to accurately estimate SPM in diverse aquatic environments using a single algorithm. In this study, two water type classification schemes: optical water type classification scheme and particle composition classification scheme, were integrated into a semi-analytical method to improve the accuracy of SPM estimation in various water bodies. By combining these two classification schemes, we classified water bodies around the world into 12 water types and developed an SPM estimation algorithm for each water type. Using 4,513 in situ Rrs-SPM measurements, the performance of the new SPM estimation algorithm was compared to that of 11 existing SPM estimation algorithms, and the results show that the median absolute percentage error (MdAPE) was reduced from 51.3 to 58.9% to 43.2%. The performance of the proposed method was also evaluated using 226 satellite matchups, with an MdAPE of 43.4%. Further comparative analysis and showcases based on several satellite images demonstrate that the two water type classification schemes play different roles that can effectively enhance the accuracy of SPM estimation.