一种湿地河流叶绿素a遥感反演方法
刘卫华(1993— ),男,河南周口人,硕士生,主要从事水色和生态遥感研究。E-mail:liuwh@radi.ac.cn |
收稿日期: 2019-09-25
要求修回日期: 2019-12-11
网络出版日期: 2020-12-25
基金资助
中国科学院战略性先导科技专项(A类)地球大数据科学工程(XDA19030501)
第二次青藏高原综合科学考察专题(2019QZKK0204)
版权
A Remote Sensing Method for Retrieving Chlorophyll-a Concentration from River Water Body
Received date: 2019-09-25
Request revised date: 2019-12-11
Online published: 2020-12-25
Supported by
The Strategic Priority Research Program of the Chinese Academy of Sciences: CAS Earth Big Data Science Project(XDA19030501)
The Second Comprehensive Scientific Investigation of the Tibetan Plateau(2019QZKK0204)
Copyright
叶绿素a作为一项重要的水质安全评价指标,其浓度的准确监测对水产行业发展、水生态系统平衡和人类饮水安全等有着重要意义。随着对地观测卫星传感器空间和光谱分辨率的提高,遥感技术在河流水质时空变化监测中发挥着越来越重要的作用。本文以新疆巴音布鲁克湿地河流水体为研究对象,同步采集了水体反射光谱和水样,并在实验室对叶绿素a、浊度等水质参数进行测定。首先,基于光谱波段对叶绿素a浓度的敏感性分析,构建了多种光谱指数模型;然后,提出以4.50 mg/m3作为水体叶绿素a浓度分级阈值,利用三波段半分析模型因子D3B与叶绿素a的线性关系建立水体叶绿素a浓度分级标准,进而对比评估了11种经验、半分析模型分别在全部样本数据集和两级叶绿素a浓度数据集中的精度表现;其次,根据各模型精度结果选用三波段半分析模型D3B和蓝绿波段比模型OC2V4,组成叶绿素a分级反演算法OC2-D3B,其精度(R2=0.96,RMSE=0.32 mg/m3,MAE=0.24 mg/m3,MRE=5.71%)相比以上2种单一算法提高了50%以上;最后,本文利用Sentinel-2影像,对湿地河流水体叶绿素a浓度的空间分布特征和季节时序模式进行了分析,得到该水域夏季叶绿素a含量最高,春秋季次之,冬季最低的结论。此外,本研究还发现气温相比其他环境因子对水体Chl-a浓度的控制作用更加明显。
刘卫华 , 王思远 , 马元旭 , 申明 , 游永发 , 海凯 , 吴林霖 . 一种湿地河流叶绿素a遥感反演方法[J]. 地球信息科学学报, 2020 , 22(10) : 2062 -2077 . DOI: 10.12082/dqxxkx.2020.190547
Chlorophyll-a (Chl-a) is an important indicator to evaluate water quality security. The accurate estimation of its concentration is of major significance to aquaculture development, aquatic ecosystem sustainability, and human drinking water safety. With the enhancement of the spatial and spectral resolution of earth-observed satellite sensor, remote sensing technology is exerting a growing important effect on monitoring the spatiotemporal changes of water quality in rivers. In this study, we synchronously measured water spectrum and collected water samples along the upper and middle reaches of the Kaidu River and around some small lakes in the Bayanbulak Wetland. Chlorophyll-a concentration and turbidity were measured for each sample in the laboratory. Based on the water reflectance spectrum and measured chlorophyll-a, we initially performed the sensitivity analysis of spectrum band to the concentration of chlorophyll-a, and then established various spectral index models, including band differences, ratios, and difference-sum ratios. Then Chl-a=4.50 mg/m3 was proposed as a hierarchical threshold for dividing waters samples into two groups and 11 empirical and semi-analytical chlorophyll-a retrieval models after calibration were applied to all sample datasets and the two separate datasets with relatively high and low chlorophyll-a concentrations to evaluate their accuracy. The optimal linear relationship between the independent variable (D3B) of three-band semi-analytical model and chlorophyll-a determined that D3B=-0.051 could be regarded as an indicator to classify waters with different chlorophyll-a concentrations. According to the performance of all the models, we ultimately selected the D3B model for high chlorophyll-a concentration waters and the blue-green band ratio model for low chlorophyll-a concentration waters, resulting in the hierarchical retrieval algorithm OC2-D3B. Its accuracy (R2=0.96, RMSE=0.32 mg/m3, MAE=0.24 mg/m3, and MRE=5.71%) was greatly improved compared with other single algorithms. Finally, we analyzed the spatial distribution and seasonal pattern of chlorophyll-a concentration in Bayanbulak Wetland using Sentinel-2 images from 2016 to 2019. The results indicate that the chlorophyll-a concentration in lake was higher than that in river, and the highest chlorophyll-a concentration usually appeared in summer, followed by spring and autumn, while the lowest chlorophyll-a concentration occurred in winter. Based on observational data from the Bayanbulak meteorological station, we further analyzed the effects of three environmental factors of temperature, precipitation, and sunshine duration on the chlorophyll-a concentration in the wetland river. The results show that the correlation coefficient between temperature and chlorophyll-a concentration reached 0.88, which was much higher than the other two factors. Thus, it seems that temperature was the main factor affecting chlorophyll-a concentration to some extent. In addition, this study could provide technical support for water environmental protection and water resource regulation in Bayanbulak.
表1 2018年7月实测水样Chl-a浓度和浊度的统计分析Tab. 1 Descriptive statistics of the concentration of water constituents in July 2018 |
水质参数 | 最小值 | 最大值 | 平均值 | 标准差 | 变异系数 | 采样数 | |
---|---|---|---|---|---|---|---|
叶绿素a/(mg/m3) | 总体 | 2.53 | 8.72 | 4.29 | 1.65 | 0.39 | 38 |
干流 | 2.78 | 8.06 | 4.16 | 1.37 | 0.33 | 12 | |
湖泊 | 2.53 | 8.72 | 5.33 | 1.98 | 0.37 | 13 | |
支流 | 2.67 | 5.24 | 3.37 | 0.67 | 0.20 | 13 | |
浊度(NTU) | 总体 | 2.57 | 93.42 | 34.23 | 26.56 | 0.78 | 38 |
干流 | 4.33 | 93.42 | 49.54 | 29.81 | 0.60 | 12 | |
湖泊 | 2.57 | 44.71 | 25.77 | 13.92 | 0.54 | 13 | |
支流 | 2.89 | 75.38 | 28.56 | 16.36 | 0.57 | 13 |
图3 基于浓度分级的Chl-a反演流程Fig. 3 Framework for Chl-a estimation based on concentration classification |
表3 基于回归分析的水体反射率与Chl-a浓度的最佳拟合方程Tab. 3 The optimal fitting equation between reflectance and Chl-a concentration based on regression analysis |
模型简写 | 自变量 | 拟合方程 | R2 |
---|---|---|---|
X1 | 0.56 | ||
X2 | 0.48 | ||
X3 | 0.36 | ||
X4 | 0.34 | ||
X5 | 0.63 | ||
X6 | 0.82 | ||
X7 | 0.73 |
表4 以D3B=-0.051和Chl-a=4.50 mg/m3为水体分级阈值的水样统计Tab. 4 Descriptive statistics of the water samples with D3B=-0.051 and Chl-a=4.5 mg/m3as hierarchical threshold |
D3B分组 | Chl-a最小值 | Chl-a最大值 | Chl-a均值 | 标准差 | 变异系数 | 采样点数 |
---|---|---|---|---|---|---|
D3B≤-0.051 | 2.53 | 4.25 | 3.40 | 0.50 | 0.15 | 24 |
D3B>-0.051 | 2.67 | 8.72 | 6.05 | 1.73 | 0.28 | 12 |
Chl-a分组 | D3B最小值 | D3B最大值 | D3B均值 | 标准差 | 变异系数 | 采样点数 |
Chl-a ≤4.50 mg/m3 | -0.092 | -0.041 | -0.072 | 0.014 | -0.19 | 26 |
Chl-a>4.50 mg/m3 | -0.047 | 0.026 | -0.014 | 0.023 | -1.69 | 10 |
[1] |
|
[2] |
|
[3] |
|
[4] |
|
[5] |
|
[6] |
|
[7] |
|
[8] |
|
[9] |
|
[10] |
|
[11] |
|
[12] |
潘洋洋. SVM模型在叶绿素a非线性定量遥感反演中的应用研究[D]. 武汉:华中科技大学, 2017.
[
|
[13] |
石绥祥, 王蕾, 余璇, 等. 长短期记忆神经网络在叶绿素a浓度预测中的应用[J]. 海洋学报, 2020,42(2):134-142.
[
|
[14] |
|
[15] |
|
[16] |
郭宇龙, 李云梅, 李渊, 等. 一种基于GOCI数据的叶绿素a浓度三波段估算模型[J]. 环境科学, 2015,36(9):3175-3185.
[
|
[17] |
徐升, 顾长梅, 钱贞兵, 等. 基于四波段模型的巢湖水体藻蓝素浓度反演[J]. 绿色科技, 2016(16):18-22,25.
[
|
[18] |
|
[19] |
刘文雅, 邓孺孺, 梁业恒, 等. 基于辐射传输模型的巢湖叶绿素a浓度反演[J]. 国土资源遥感, 2019,31(2):102-110.
[
|
[20] |
李云梅, 黄家柱, 韦玉春, 等. 用分析模型方法反演水体叶绿素的浓度[J]. 遥感学报, 2006,10(2):27-33.
[
|
[21] |
|
[22] |
|
[23] |
|
[24] |
陈芸芝, 郑高强, 汪小钦, 等. 基于GSM01融合的多传感器数据叶绿素a浓度反演[J]. 地球信息科学学报, 2013,15(6):911-917.
[
|
[25] |
|
[26] |
张运林, 冯胜, 马荣华, 等. 太湖秋季光学活性物质空间分布及其遥感估算模型研究[J]. 武汉大学学报·信息科学版, 2008,33(9):967-972.
[
|
[27] |
|
[28] |
阎福礼, 刘韶菲, 王世新, 等. 太湖浮游藻类的后向散射分离及其叶绿素a浓度反演[J]. 地球信息科学学报, 2014,16(6):989-996.
[
|
[29] |
|
[30] |
|
[31] |
高庆, 艾里西尔·库尔班, 肖昊. 巴音布鲁克地区植物物候时空动态变化及其驱动分析[J]. 干旱区研究, 2018,35(6):1418-1426.
[
|
[32] |
陈文玲, 何雨. 草原生态旅游资源开发评价体系构建与应用研究——以新疆巴音布鲁克草原为例[J]. 资源开发与市场, 2016,32(11):1394-1397.
[
|
[33] |
徐晓龙, 王新军, 朱新萍, 等. 1996-2015年巴音布鲁克天鹅湖高寒湿地景观格局演变分析[J]. 自然资源学报, 2018,33(11):39-53.
[
|
[34] |
|
[35] |
唐军武, 田国良, 汪小勇, 等. 水体光谱测量与分析Ⅰ:水面以上测量法[J]. 遥感学报, 2004,8(1):37-44.
[
|
[36] |
|
[37] |
|
[38] |
|
[39] |
|
[40] |
|
[41] |
European Space Agency. ESA Copernicus Open Access Hub[DB/OL]. https://scihub.copernicus.eu/. 2019-08-25.
|
[42] |
苏伟, 张明政, 蒋坤萍, 等. Sentinel-2卫星影像的大气校正方法[J]. 光学学报, 2018,38(1):322-331.
[
|
[43] |
国家气象信息中心. 中国地面气候资料日值数据集(V3.0) [DB/OL]. http://data.cma.cn/data/cdcdetail/dataCode/SURF_CLI_CHN_MUL_DAY_V3.0.html. 2019-09-01.
[ National Meteorological Information Center. Dataset of Daily Climate Data from Chinese Surface Stations (V3.0)[DB/OL]. http://data.cma.cn/data/cdcdetail/dataCode/SURF_CLI_CHN_MUL_DAY_V3.0.html. 2019-09-01.]
|
[44] |
|
[45] |
|
[46] |
|
[47] |
|
[48] |
|
[49] |
|
[50] |
慈晖, 张强. 新疆NDVI时空特征及气候变化影响研究[J]. 地球信息科学学报, 2017,19(5):662-671.
[
|
[51] |
周梦甜, 李军, 朱康文. 近15a新疆不同类型植被NDVI时空动态变化及对气候变化的响应[J]. 干旱区地理, 2015,38(4):779-787.
[
|
/
〈 | 〉 |