Red Tide Detection Based on High Spatial Resolution Broad Band Optical Satellite Data
Rongjie Liu,Yanfang Xiao,Yi Ma,Tingwei Cui,Jubai An
DOI: https://doi.org/10.1016/j.isprsjprs.2021.12.009
IF: 12.7
2022-01-01
ISPRS Journal of Photogrammetry and Remote Sensing
Abstract:Red tide, one of the major ecological disasters in the world, exerts great impact on the marine environment. The ocean color satellite data with low spatial resolution and high spectral resolution is often used for red tide detection, but is insufficient for the fine-scale red tide detection due to its coarse spatial resolution. To address this problem, a new red tide detection method based on pseudo hue angle (PHA-RI) for high spatial resolution broad band satellite data is proposed in this paper. Different from the standard International Commission on Illumination (CIE) system, the false-color bands of near infrared (NIR), red and green are used to calculate the CIE tristimulus X, Y, Z, instead of the true-color bands of red, green and blue. With this method, the red tide can be easily differentiated from non-red tide water, and the distinction between red tide and non-red tide water is doubled compared to that of true-color bands composite. Experiment results show that the PHA-RI method can effectively detect red tide, with an averaged overall accuracy, F1-score and precision of 92%, 0.92 and 0.92 respectively. Compared with the traditional machine learning method support vector machine (SVM) and spectral index algorithm Gaofen-1 (GF-1) red tide detection index (GF-RI), the proposed algorithm has obvious advantages in red tide detection, especially for the strip distributed red tide. Besides, the proposed algorithm has good applicability, and it is proved to be suitable for the detection of red tides with different dominant species including Noctiluca scintillans, Skeletonema costatum and Heterosigma akashiide. Also, it is applicable to different broad band sensors, such as Chinese satellites GF-1 Wide Field of View (WFV), Haiyang 1D (HY-1D) Coastal Zone Imager (CZI), Sentinel-2 MultiSpectral Instrument (MSI) and Landsat 8 Operational Land Imager (OLI). The application experiment indicates that the proposed algorithm is effective in red tide detection, and it can successfully detect fine-scale red tide events, which have not been detected by Moderate Resolution Imaging Spectroradiometer (MODIS) Aqua due to its coarse spatial resolution. The results also shows the advantages of high spatial resolution satellites in fine-scale red tide detection. This work provides a method for red tide detection using high spatial resolution broad band satellite data.