Lyu Lili, Song Kaishan, Wen Zhidan, Liu Ge, Fang Chong, Shang Yingxin, Li Sijia, Tao Hui, Wang Xiang, Li Yong, Wang Xiangyu
Northeast Institute of Geography and Agroecology, Chinese Academy of Sciences, Changchun 130102, China; Jilin Jianzhu University, Changchun, China.
Northeast Institute of Geography and Agroecology, Chinese Academy of Sciences, Changchun 130102, China; School of Environment and Planning, Liaocheng University, Liaocheng 252000, China.
Sci Total Environ. 2023 Nov 15;899:166363. doi: 10.1016/j.scitotenv.2023.166363. Epub 2023 Aug 19.
In recent years, under the dual pressure of climate change and human activities, the cyanobacteria blooms in inland waters have become a threat to global aquatic ecosystems and the environment. Phycocyanin (PC), a diagnostic pigment of cyanobacteria, plays an essential role in the detection and early warning of cyanobacterial blooms. In this context, accurate estimation of PC concentration in turbid waters by remote sensing is challenging due to optical complexity and weak optical signal. In this study, we collected a comprehensive dataset of 640 pairs of in situ measured pigment concentration and the Ocean and Land Color Instrument (OLCI) reflectance from 25 lakes and reservoirs in China during 2020-2022. We then developed a framework consisting of the water optical classification algorithm and three candidate algorithms: baseline height, band ratio, and three-band algorithm. The optical classification method used remote sensing reflectance (R) baseline height in three bands: R(560), R(647) and R(709) to classify the samples into five types, each with a specific spectral shape and water quality character. The improvement of PC estimation accuracy for optically classified waters was shown by comparison with unclassified waters with RMSE = 72.6 μg L, MAPE = 80.4 %, especially for the samples with low PC concentration. The results show that the band ratio algorithm has a strong universality, which is suitable for medium turbid and clean water. In addition, the three-band algorithm is only suitable for medium turbid water, and the line height algorithm is only suitable for high PC content water. Furthermore, the five distinguished types with significant differences in the value of the PC/Chla ratio well indicated the risk rank assessment of cyanobacteria. In conclusion, the proposed framework in this paper solved the problem of PC estimation accuracy problem in optically complex waters and provided a new strategy for water quality inversion in inland waters.
近年来，在气候变化和人类活动的双重压力下，内陆水体中的蓝藻水华已成为全球水生生态系统和环境的一大威胁。藻蓝蛋白（PC）作为蓝藻的一种诊断色素，在蓝藻水华的检测和预警中发挥着重要作用。在此背景下，由于光学复杂性和微弱的光学信号，通过遥感准确估计浑浊水体中的PC浓度具有挑战性。在本研究中，我们收集了2020年至2022年期间来自中国25个湖泊和水库的640对现场测量的色素浓度与海洋陆地颜色仪器（OLCI）反射率的综合数据集。然后，我们开发了一个由水光学分类算法和三种候选算法组成的框架：基线高度、波段比值和三波段算法。光学分类方法利用三个波段（R(560)、R(647)和R(709)）的遥感反射率（R）基线高度将样本分为五种类型，每种类型具有特定的光谱形状和水质特征。与未分类水体（RMSE = 72.6 μg/L，MAPE = 80.4%）相比，光学分类水体的PC估计精度有所提高，尤其是对于低PC浓度的样本。结果表明，波段比值算法具有很强的通用性，适用于中度浑浊和清洁水体。此外，三波段算法仅适用于中度浑浊水体，线高度算法仅适用于高PC含量水体。此外，PC/Chla比值差异显著的五种不同类型很好地表明了蓝藻的风险等级评估。总之，本文提出的框架解决了光学复杂水体中PC估计精度的问题，并为内陆水体水质反演提供了一种新策略。
