南漪湖叶绿素a浓度荧光反演算法研究

doi:10.3964/j.issn.1000-0593(2022)12-3941-07

摘要
参考文献
相关文章 (15)

全文: PDF (4178 KB)
输出: BibTeX | EndNote (RIS)

摘要：叶绿素a（Chlorophyll-a, Chl-a）浓度是监测浮游植物和水质状况的代表性指标，对湖泊富营养化评价具有重要意义。为了探求南漪湖多时相的Chl-a浓度高光谱特征及反演方法，选取2020年—2021年间南漪湖8次走航式水体实验同步采集的98组高光谱数据和Chl-a浓度数据，分析不同Chl-a浓度条件下南漪湖实测光谱的变化特征，同时考虑水质组分变化和采样时间变化对光谱的影响，提取能反映Chl-a浓度信息的特征波段。然后引入峰谷距离法、荧光基线高度法、峰面积法和基于峰面积法改进的谷上峰面积法共同反演南漪湖Chl-a浓度，并利用5折交叉验证法比较不同反演算法的优劣。研究结果表明：（1）随着Chl-a浓度的增大，荧光峰位置向红外方向移动，Chl-a吸收谷和荧光峰分别有加深和升高的趋势，峰谷差异更加明显，荧光峰附近谱段能够有效反映Chl-a浓度变化；（2）利用5折交叉验证法将样本分为5组，依次作为验证集进行建模，对于不同组别的验证集各方法的RMSE和 MAPE极差平均值分别为0.437 5 μg·L^-1和28.27%，可见样本建模集与验证集的选取会显著影响精度评价结果，5折交叉验证的方法可以减小上述误差，在样本范围内最大程度地比较出各方法的优劣；（3）结合Chl-a浓度吸收谷极小值处水平切线提出的谷上峰面积法取得了最优的反演结果，其验证精度分别为R²=0.756 7，RMSE=1.653 1 μg·L^-1，MAPE=40.77%，相较于峰谷距离法、荧光基线高度法和峰面积法精度均有提升，为叶绿素a浓度荧光反演提供了新的思路。

关键词：南漪湖；叶绿素a；高光谱特征分析；荧光峰；k折交叉检验

Abstract：Serving as a representative indicator for phytoplankton and water quality monitoring, Chlorophyll a (Chl-a) is of great significance to evaluating lake eutrophication level. In order to explore the hyperspectral characteristics of multi-temporal Chl-a concentration and to select the best inversion methods of Nanyi Lake, 98 sets of hyperspectral data and Chl-a concentration data were collected simultaneously from 8 navigational water experiments in Nanyi Lake from 2020 to 2021 were selected. To extract the characteristic bands most sensitive to Chl-a concentration, measured spectrum data of Nanyi Lake under different Chl-a concentration levels were analyzed, considering the influence of changes in water quality at different timeson the spectrum. Then, the peak and valley distance method, the fluorescence line height method, the Normalized peak area method and the peak area above valley method was introduced to jointly invert the concentration of Chl-a in Nanyi Lake, followed by inter-comparing the results of the abovementioned algorithms based on the 5-fold cross-validation. The results areas follows: (1) As the concentration of Chl-a increases, the absorption valley and fluorescence peak of Chl-a tend to deepen and increase, respectively. At the same time, the position of the fluorescence peak moves towards the infrared part with increasing Chl-a concentration. The obvious difference between peak and valley under different Chl-a concentration levels indicates spectrum before and after fluorescence peak is highly sensitive to the change of Chl-a concentration. (2) Validation results using a 5-fold cross-validation method show that the mean values of RMSE and MAPE extreme differences for each method for different groups of validation sets were 0.437 5 μg·L^-1 and 28.27%. It can be seen that the sampling method of the modeling set and verification set will introduce evaluation error, which can effectively be reduced by the 5-fold cross-validation method, obtaining the pros and cons of each method to the greatest extent based on samples. (3) Best inversion results have been achieved by the peak area above valley method, which was proposed in combination with the horizontal tangent line at the minimum value of the absorption valley of Chl-a concentration, with R²=0.756 7, RMSE=1.653 1 μg·L^-1, and MAPE=40.77%. Compared with the peak and valley distance method, the fluorescence line height method and the Normalized peak area method witnessed significant improvement in the inversion accuracy and provided a new idea for the inversion of chlorophyll concentration based on fluorescence.

Key words：Nanyi Lake; Chlorophyll-a; Hyperspectral feature analysis; Fluorescence peak; K-fold cross validation

收稿日期: 2021-11-26 修订日期: 2022-03-07

中图分类号:

X87

基金资助: 国家自然科学基金项目（42176176）, 高分辨率对地观测系统重大专项(21-Y20B01-9001-19/22）, 国家民用空间基础设施项目资助

通讯作者: 谢勇 E-mail: xieyong@nuist.edu.cn

作者简介: 代前程，1997年生，南京信息工程大学地理科学学院硕士研究生 e-mail: qianchengdai@foxmail.com

引用本文:

代前程，谢勇，陶醉，邵雯，彭飞宇，苏逸，杨邦会. 南漪湖叶绿素a浓度荧光反演算法研究[J]. 光谱学与光谱分析, 2022, 42(12): 3941-3947.
DAI Qian-cheng, XIE Yong, TAO Zui, SHAO Wen, PENG Fei-yu, SU Yi, YANG Bang-hui. Research on Fluorescence Retrieval Algorithm of Chlorophyll a Concentration in Nanyi Lake. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2022, 42(12): 3941-3947.

链接本文:

https://www.gpxygpfx.com/CN/10.3964/j.issn.1000-0593(2022)12-3941-07 或 https://www.gpxygpfx.com/CN/Y2022/V42/I12/3941

[1] Wang Junlei, Zhang Yongjie, Yang Fei, et al. Environmental Earth Sciences, 2015, 73(8): 4063.
[2] ZHU Guang-wei, XU Hai, ZHU Meng-yuan, et al(朱广伟, 许海, 朱梦圆, 等). Journal of Lake Sciences(湖泊科学), 2019, 31(6): 1510.
[3] LUO Jie-chun-yi, QIN Long-jun, MAO Peng, et al(罗婕纯一, 秦龙君, 毛鹏, 等). Remote Sensing Technology and Application(遥感技术与应用), 2021, 36(3): 473.
[4] SHEN Wei, JI Qian, QIU Yao-wei, et al(沈蔚, 纪茜, 邱耀炜, 等). Journal of Hydroecology(水生态学杂志), 2021, 42(3): 1.
[5] WANG Lin, YANG Jian-hong, ZHAO Dong-zhi(王林, 杨建洪, 赵冬至). Journal of Applied Oceanography(应用海洋学学报), 2014, 33(1): 111.
[6] Tenjo C, Ruiz-Verdú A, Wittenberghe S V, et al. Remote Sensing, 2021, 13(2): 329.
[7] WANG Jin-liang, QIN Qi-ming, LI Jun, et al(王金梁, 秦其明, 李军, 等). Transactions of the Chinese Society of Agricultural Engineering(农业工程学报), 2014, 30(3): 128.
[8] Song Kaishan, Li Lin, Wang Zongming, et al. Environmental Monitoring & Assessment, 2012, 184(3): 1449.
[9] WEN Yan-sha, DENG Jian-ming, MAO Zhi-hua, et al(温颜沙, 邓建明, 毛志华, 等). Journal of Remote Sensing(遥感学报), 2018, 22(3): 424.
[10] GAO Chen, XU Jian, GAO Dan, et al(高晨, 徐健, 高丹, 等). Remote Sensing for Land and Resources(国土资源遥感), 2019, 31(1): 101.
[11] PENG Ling, MEI Jun-jun, WANG Na, et al(彭令, 梅军军, 王娜, 等). Spectroscopy and Spectral Analysis(光谱学与光谱分析), 2019, 39(9): 2922.
[12] Kim G, Baek I, Stocker M D, et al. Remote Sensing, 2020, 12(13): 2070.
[13] SONG Ting, ZHOU Wen-lin, LIU Jun-zhi, et al(宋挺, 周文鳞, 刘军志, 等). Acta Scientiae Circumstantiae(环境科学学报), 2017, 37(3): 888.
[14] HUANG Qi-hui, HE Zhong-hua, LIANG Hong, et, al(黄启会, 贺中华, 梁虹, 等). Environmental Science & Technology(环境科学与技术), 2019, 42(1): 134.
[15] MA Wan-dong, WANG Qiao, WU Chuan-qing, et al(马万栋, 王桥, 吴传庆, 等). Journal of Geo-information Science(地球信息科学学报), 2014, 16(6): 965.