Feng Chi, Jin Qi, Wang Yan-nan, Zhao Li-na, Lu Heng, Li Yun-mei
Huan Jing Ke Xue. 2015 May;36(5):1557-64.
Chlorophyll-a as one of the important water quality parameters is often used as a measure of the level of water eutrophication. The 326 measured data collected from Lake Taihu and Lake Dongting were classified based on their measured values of remote sensing reflectance spectra using an automatic clustering algorithm-two-step method, and three water types were finally classified. According to the location and width of GOCI satellite bands, the specific algorithm to estimate chlorophyll-a concentration for different water body types was developed. The bands at 490 nm and 555 nm were used for water body type I , while bands at 660 nm and 443 nm were selected for water body type II and bands at 745 nm and 680 nm were applied for water body type III. The accuracy assessment showed that the mean relative error decreased from 49. 78% to 38. 91% , 24. 19% and 22. 90% for water body type I , II and III, respectively, while the root mean square error decreased from 14.10 µg · L(-1) to 4.87 µg · L(-1), 8.13 µg · L(-1) and 11.66 µg · L(-1) for water body type I, II and III, respectively. The overall mean relative error decreased from 49. 78% to 29. 59% after classification, while the overall root mean square error was reduced from 14.10 µg · L(-1) to 9.29 µg · L(-1) after classification. The retrieval accuracy was significantly improved after classification. The chlorophyll-a concentration in Lake Taihu was retrieved using the GOCI image on May 13, 2013. The results showed that there was a significantly diurnal variation in the concentration of chllorophyll-a on May 13, 2013, and the regions with higher chlorophyll-a concentration were mainly distributed in the Zhushan Bay, Meiliang Bay and Gonghu Bay, while the regions with lower values were mainly located in the centre of the lake and the southern region. The chlorophyll-a concentration reduced significantly after 10:00 in the southwestern region of Lake Taihu. This method of retrieving, after classification played an important role in improving the model retrieval accuracy of case 2 water.
叶绿素a作为重要的水质参数之一，常被用作衡量水体富营养化程度的指标。利用自动聚类算法——两步法，对从太湖和洞庭湖采集的326个实测数据，根据其遥感反射光谱的测量值进行分类，最终划分出三种水体类型。根据GOCI卫星波段的位置和宽度，开发了针对不同水体类型估算叶绿素a浓度的具体算法。对于水体I型，使用490nm和555nm波段；对于水体II型，选择660nm和443nm波段；对于水体III型，应用745nm和680nm波段。精度评估表明，水体I型、II型和III型的平均相对误差分别从49.78%降至38.91%、24.19%和22.90%，而均方根误差分别从14.10μg·L⁻¹降至4.87μg·L⁻¹、8.13μg·L⁻¹和11.66μg·L⁻¹。分类后总体平均相对误差从49.78%降至29.59%，总体均方根误差从14.10μg·L⁻¹降至9.29μg·L⁻¹。分类后反演精度显著提高。利用2013年5月13日的GOCI影像反演了太湖的叶绿素a浓度。结果表明，2013年5月13日叶绿素a浓度存在明显的日变化，叶绿素a浓度较高的区域主要分布在竺山湾、梅梁湾和贡湖湾，而值较低的区域主要位于湖心和南部地区。太湖西南部地区10:00后叶绿素a浓度显著降低。这种分类后反演的方法对提高二类水体模型反演精度具有重要作用。
