Remote Sensing Statistical Inference for Colored Dissolved Organic Matter in Inland Water: Case Study in Qiandao Lake

Compared to traditional remote sensing classification and inversion techniques, remote sensing statistical inference is a novel method for rapidly estimating the statistical properties of ground objects. Despite some initial work, this method has not been thoroughly evaluated for water quality assessment. In this study, using field-measured data from Qiandao Lake, we tested over 240 000 inference models for determining the mean, median, standard deviation, minimum, and maximum of colored dissolved organic matter using a bootstrap approach and various combinations of bands, variables, and functions. The results indicated that all five statistical parameters could be inferred accurately with errors of less than 10%. The best models used two band ratios, three statistical variables, and polynomial functions. The study also demonstrated the importance of redistributing the raw field-measured data for improved performance, as models based on the redistributed data outperformed those based on the raw data.