Assessment of water constituents in highly turbid productive water by optimization bio-optical retrieval model after optical classification

• Published in: Journal of hydrology (Amsterdam) 2014-11, Vol.519, p.1572-1583
• Main Authors: Huang, Changchun, Li, Yunmei, Yang, Hao, Li, Junsheng, Chen, Xia, Sun, Deyong, Le, Chengfeng, Zou, Jun, Xu, Liangjiang
• Format: Article
• Language: English
• Subjects: Accuracy; Algorithms; Applied geophysics; Chlorophyll-a concentration; Classification; Computer simulation; Earth sciences; Earth, ocean, space; Exact sciences and technology; Genetic algorithm; Hydrology; Hydrology. Hydrogeology; Internal geophysics; Lakes; Non-linear quadratic programming; Optimization; Retrieval; Satellites; Taihu Lake; Total suspended matter
• ISSN: 0022-1694; 1879-2707
• DOI: 10.1016/j.jhydrol.2014.09.007

•We developed a new bio-optical optimizing retrieval model with two optimization steps.•Classification improved the retrieval accuracy of this new bio-optical model.•A large body of in situ data validated this algorithm.•This algorithm shows wide application to types of water and satellite sensors. Based on the bio-optical properties of highly turbid productive inland waterbody, Taihu Lake, a novel bio-optical optimization algorithm was developed to estimate chlorophyll-a concentration (Cchla) and total suspended matter concentration (CTSM) after optical classification. Between 2006 and 2013, 1080 in situ samples collected from four inland lakes in China were utilized to test this optimization algorithm. All data were classified into four classes based on a new bio-optical classification method. The retrieval results of CTSM and Cchla exhibit a good consistency with in situ measured CTSM and Cchla. CTSM retrieval accuracies (evaluated by the root mean square of percentage errors: RMSPs) of classes 1, 2, 3, and 4 were 35.77%, 16.09%, 28.42%, and 26.86% for data1 and were 32.15%, 33.14%, 47.71%, and 34.89% for data3, respectively. Cchla retrieval accuracies (RMSPs) of classes 1, 2, 3, and 4 were 32.49%, 20.05%, 42.01%, and 34.85% for data1, were 44.71%, 32.59%, 47.92%, and 38.11% for data2, and were 33.12%, 25.65%, 70.88%, and 23.57% for data3, respectively. The optimization algorithm was also tested by the simulated data of Medium Resolution Imaging Spectrometer, Moderate Resolution Imaging Spectroradiometer, and Sea-viewing Wide Field-of-view Sensor satellite sensors’ center wavelengths. The validation shows a good correlation with the measured CTSM and Cchla. All these examinations demonstrate that the bio-optical optimization algorithm and classification are valid and robust for both the in situ data and the simulated satellite data. The optical relationships of aph(440) to Cchla and bbp(440) to CTSM are reasonable and effective. In summary, the results present that the bio-optical optimization algorithm proposed in this study shows high potential application to various water types and satellite sensors. The retrieval accuracy of CTSM and Cchla derived by bio-optical optimization algorithm was significantly improved after classification.