Correcting Aquaculture Facility-Induced Spectral Distortions for Improved Satellite Water Quality Retrieval in Marine Ranching Areas

Marine ranching plays a vital role in sustainable fishery resource utilization and marine ecosystem protection. The accurate water quality monitoring through remote sensing is essential in these areas; however, aquaculture facilities such as floating buoys introduce additional reflected signals that can distort remote sensing reflectance [ $R_{\mathrm {rs}}(\lambda)$ ], leading to significant errors in water quality parameter retrievals. Despite this challenge, previous studies have neither systematically quantified this interference nor developed effective correction methods. This study proposes a novel correction method to mitigate aquaculture-induced distortions in $R_{\mathrm {rs}}(\lambda)$ , enhancing the accuracy of remote sensing-based water quality assessments. Using the mussel aquaculture ranching area off Gouqi Island, China, as a case study, we systematically analyze the spectral influence of aquaculture facilities on $R_{\mathrm {rs}}(\lambda)$ derived from Landsat observations. Comparative spectral analysis between affected and unaffected areas reveals that the high-reflectance characteristics of aquaculture facilities cause abnormally elevated $R_{\mathrm {rs}}(\lambda)$ values, particularly in the shortwave infrared (SWIR) bands. To address this issue, we introduce an aquaculture facility influence factor and develop a pixel-based dynamic correction approach that adjusts for varying degrees of aquaculture-induced distortions across different pixels. Validations using field-measured $R_{\mathrm {rs}}(\lambda)$ demonstrates that the mean absolute percentage errors (MAPEs) at wavelengths of 443, 483, 561, 655, 865, and 1609 nm decreased significantly from 17.6%, 19.8%, 14.8%, 26.8%, 50.9%, and 180.6% before correction to 9.5%, 11.0%, 9.2%, 10.4%, 7.7%, and 20.4%, respectively. The effectiveness of the correction method is further supported by improvements in the retrieval of water transparency ( $Z_{\mathrm {sd}}$ ). The sensitivity analysis further reveals that uncorrected $Z_{\mathrm {sd}}$ retrieval errors increase exponentially with aquaculture facility coverage, exceeding 60% when coverage reaches 10%, underscoring the necessity of correcting $R_{\mathrm {rs}}(\lambda)$ distortions caused by aquaculture facilities. Overall, the proposed correction method provides a robust and adaptable framework for improving satellite-based water quality monitoring in complex aquaculture regions, with potential applicability across diverse marine ranching environments and integration with various satellite sensors.