基于Sentinel-2A影像的玉米冠层叶绿素含量估算

doi:10.3964/j.issn.1000-0593(2019)05-1535-08

摘要
参考文献
相关文章 (15)

全文: PDF (2127 KB)
输出: BibTeX | EndNote (RIS)

摘要：农作物叶片中的叶绿素通过吸收光能参与光合作用产生化学能，及时、准确地估算叶绿素含量对于农作物长势、养分含量监测、品质评价和产量估算具有重要意义。Sentinel-2卫星的重访周期为5 d，空间分辨率为10 m，具有13个光谱波段，其中包括三个波宽仅为15 nm对叶绿素含量变化敏感的红边波段，是叶绿素含量估算的理想数据源。植被指数是基于农作物在不同波段的反射特性，通过不同波段组合方式刻画长势和叶绿素含量的差异，可用于大区域范围内的玉米冠层叶绿素含量快速、精确估算。以Sentinel-2A影像为数据源，开展基于多种植被指数的玉米冠层叶绿素含量估算方法研究。课题组于2016年8月6—11日在河北省保定市（115°29′—116°14′E，39°5′—39°35′N）进行玉米冠层叶绿素含量的实地测量，并在每个采样位置上采用中绘i80 智能RTK(real-time kinematic)测量系统进行定位。Sentinel-2A影像预处理工作包括几何校正、辐射定标和大气校正，其中大气校正使用Sen2Cor模型和SNAP模型。首先，基于预处理后的Sentinel-2A遥感影像，分别计算CI_green(green chlorophyll index), CI_red-edge(red-edge chlorophyll index), DVI(difference vegetation index), LCI(leaf chlorophyll index), MTCI(MERIS terrestrial chlorophyll index), NAVI(normalized area vegetation index), NDRE(normalized difference red-edge), NDVI(normalized difference vegetation index), RVI(ratio vegetation index), SIPI(structure insensitive pigment index)植被指数。然后，建立样方位置上实测叶绿素含量与各植被指数的统计关系，从而构建玉米冠层叶绿素含量估算模型，并以野外实测玉米冠层叶绿素含量为依据，对基于各植被指数的估算结果进行精度评价。最后，利用筛选出的最优叶绿素含量估算模型，估算研究区内的玉米冠层叶绿素含量。研究的目标为：（1）通过比较分析，构建合适的玉米冠层叶绿素含量估算模型，估算精度以决定系数R²、均方根误差RMSE以及相对误差RE作为评价指标；（2）确定最优波段组合方案：在红边波段中选择与可见光、近红外波段组合效果更优的波段组合方案；（3）确定参与植被指数计算的红边波段的最优数量。精度评价结果表明：（1）选用的植被指数与玉米冠层叶绿素含量呈多项式拟合关系，其中使用红边波段计算的植被指数的估算结果明显优于未使用红边波段的估算结果；红边波段引入后明显提高了可见光、近红外波段对叶绿素含量的拟合的精度，CI_{green(560, 705)}指数比CI_{green(560, 842)}的回归模型R²提高0.516，红边波段参与计算的DVI相对于RVI来说，估算结果更稳定。（2）对于不同的植被指数，参与运算的Sentinel-2A影像的两个红边波段，估算精度的提高程度不同。对于可见光波段参与计算的植被指数来说，在红边波段1（中心波长为705 nm）的估算精度较高，如LCI，CI_green，DVI和RVI等；对于近红外波段参与计算的植被指数来说，在红边波段2（中心波长为740 nm）的估算精度较高，如CI_red-edge，NDRE和NAVI等。（3）对于Sentinel-2A影像来说，两个红边波段共同参与叶绿素含量估算时能取得最高的的估算精度。选用的植被指数中，MTCI_{(665, 705, 740)}指数与玉米冠层叶绿素含量估算精度最高，回归模型拟合精度R²为0.803，模型验证R²为0.665，RMSE为3.185，相对误差RE为4.819%。MTCI_{(665, 705, 740)}指数计算中使用了两个红边波段，突出红边波段反射率差值变化，与玉米冠层叶绿素含量表现出很好的相关性。最后，利用优选出的基于MTCI指数的叶绿素含量估算模型，对研究区范围内的叶绿素含量进行估算并完成空间制图。

关键词：Sentinel-2A；玉米；冠层叶绿素含量；红边波段；植被指数

Abstract：The chlorophyll within crop leaves and crop canopy produces energy and participates in photosynthesis process by absorbing sunlight. Therefore, it is important to estimate the crop canopy chlorophyll content timely and accurately for crop growth monitoring, nutrient content monitoring and crop quality evaluation. Sentinel-2 has a wide-swath sensor with 5-days revisit period, so the Sentinel-2 image is produced with high spatial resolution (10 m) and 13 spectral bands. Specially, there are three red edge bands in Sentinel-2 image, which are sensitive to crop canopy chlorophyll content and its change. So the Sentinel-2 image is an ideal remote sensing data source for chlorophyll content estimation. Vegetation indexes depict the difference for the crop between different growth conditions and different chlorophyll contents, through the band combinations based on the reflection characteristics of crops at different spectral bands. So the vegetation indexes from Sentinel-2 image can be used to estimate the corn canopy chlorophyll content timely and accurately in a regional area. Therefore, this study is focusing on estimating the corn canopy chlorophyll content using 10 kinds of vegetation indexes computing from Sentinel-2A remote sensing images. And the study area is located in three counties of Baoding City, Hebei Province, ranging from 115°29′E to 116°14′E, 39°5′N to 39°35′N. We measured the corn plant chlorophyll content in 24 sampling areas distributed randomly in the whole study area from 6 August to 11 August, 2016. And each sampling area was located using Huace i80 real-time kinematic (RTK) GPS receiver (Huace Ltd., Shanghai, China). The Sentinel-2A image was preprocessed including geometric correction, radiometric calibration and atmospheric correction, and Sen2Cor model and SNAP were used to do atmospheric correction. 10 vegetation indexes were computed including CI_green(Green Chlorophyll Index), CI_red-edge(Red-edge Chlorophyll Index), DVI(Difference Vegetation Index), LCI(Leaf Chlorophyll Index), MTCI(MERIS Terrestrial Chlorophyll Index), NAVI(Normalized Area Vegetation Index), NDRE(Normalized Difference Red-Edge), NDVI(Normalized Difference Vegetation Index), RVI(Ratio Vegetation Index), SIPI(Structure Insensitive Pigment Index). Secondly, the statistical correlativity was analyzed between these 10 vegetation indexes and measured chlorophyll content value for every sampling area. So the corn canopy chlorophyll content estimating was developed using this correlation analysis results. Lastly, the optimal chlorophyll content estimation model was selected to estimate the chlorophyll content in the whole study area. This study was focusing on (1) developing the estimation model for corn canopy chlorophyll content in the study area, and the accuracy was assessed using R², RMSE and RE; (2) deciding the optimal band combination; (3)deciding the optimal amount of red edge band participating in vegetation indexes calculation. The accuracy assessment results indicated that (1) there was polynomial correlation between measured chlorophyll content and the selected 10 vegetation indexes in this study, and the accuracy of estimated chlorophyll content using the vegetation indexes considering the red edge bands is better than the ones without red edge bands. The CI_{green(560, 705)}and DVI which were all considering red edge bands improved the chlorophyll content estimation accuracy, and the R² improved 0.516 for CI_{green(560, 705)}. The statistical relationship between the measured chlorophyll content and the vegetation index in the field work was established, and the relationship was extended to the whole study area. This study was about the estimation of corn canopy LAI and chlorophyll content using these ten vegetation indexes, which was focusing on the following four parts. Firstly, we compared if the vegetation with or without red-edge band could get accurate LAI and chlorophyll content estimated result. Secondly, we added two red-edge bands to the vegetation indexes without red-edge band originally. Thirdly, we added two red-edge bands to the vegetation indexes with one red-edge band originally only. Fourthly, we set up the vegetation index with two red-edge bands. The results showed that there are polynomial regression between the selection of multi-VI and the field survey of canopy chlorophyll content. Because the introductions of the red edge band, the fitting accuracy improved more than 0.3 between the vegetation index and corn canopy chlorophyll content, and the CI_{green (560, 705)} (Green Chlorophyll Index) improved 0.516 that is the highest. The index calculating between the visible light band and the first red edge band (705 nm), the near infrared band with the second red edge band (740 nm), both of which established the regression model with the field survey of corn canopy chlorophyll content, and promoted the best fitting precision. The MTCI (MERIS Terrestrial Chlorophyll Index) has the highest fitting precision in which the R² is 0.803, RMSE is 3.185, RE is 4.819%. It is shown that adding the red edge band will improve the fitting precision and it is suitable for crop growth monitoring.

Key words：Sentinel-2A; Corn; Canopy chlorophyll content; Red-edge band; Vegetation index

收稿日期: 2018-03-28 修订日期: 2018-08-05

中图分类号:

TP79

基金资助: 国家自然科学基金项目(41671433，41371327)，国家重点研发计划项目(2017YFB0503900-4-2)资助

通讯作者: 孙中平 E-mail: sunnybnu114@163.com

作者简介: 苏伟，1979年生，中国农业大学土地科学与技术学院副教授 e-mail: suwei@cau.edu.cn

引用本文:

苏伟，赵晓凤，孙中平，张明政，邹再超，王伟，史园莉. 基于Sentinel-2A影像的玉米冠层叶绿素含量估算[J]. 光谱学与光谱分析, 2019, 39(05): 1535-1542.
SU Wei, ZHAO Xiao-feng, SUN Zhong-ping, ZHANG Ming-zheng, ZOU Zai-chao, WANG Wei, SHI Yuan-li. Estimating the Corn Canopy Chlorophyll Content Using the Sentinel-2A Image. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2019, 39(05): 1535-1542.

链接本文:

https://www.gpxygpfx.com/CN/10.3964/j.issn.1000-0593(2019)05-1535-08 或 https://www.gpxygpfx.com/CN/Y2019/V39/I05/1535

[1] Peng Y, Nguy-Robertson A, Arkebauer T, et al. Remote Sensing, 2017, 9(3): 226.
[2] Chemura A, Mutanga O, Odindi J. IEEE Journal of Selected Topics in Applied Earth Observations & Remote Sensing, 2017, (99): 1.
[3] JIANG Hai-ling, YANG Hang, CHEN Xiao-ping, et al（姜海玲, 杨杭, 陈小平）. Spectroscopy and Spectral Analysis（光谱学与光谱分析）, 2015, 35(4): 975.
[4] DONG Heng, MENG Qing-ye, WANG Jin-liang, et al（董恒, 孟庆野, 王金梁, 等）. Journal of Infrared and Millimeter Waves（红外与毫米波学报）, 2012, 31(4): 336.
[5] Jesús D, Jochem V, Luis A, et al. Sensors, 2011, 11(7): 7063.
[6] Schlemmer M, Gitelson A, Schepers J, et al. International Journal of Applied Earth Observation &Geoinformation, 2013, 25(1): 47.
[7] ZHENG Yang, WU Bing-fang, ZHANG Miao（郑阳, 吴炳方, 张淼）. Journal of Remote Sensing（遥感学报）, 2017, 21(2): 318.
[8] WANG Xiao-xing, CHANG Qing-rui, LIU Meng-yun, et al(王晓星, 常庆瑞, 刘梦云, 等). Journal of Northwest A & F University·Nat. Sci. Ed(西北农林科技大学学报·自然科学版), 2016, 44(2): 48.
[9] DING Yong-jun, ZHANG Jing-jing, LI Xiu-hua, et al（丁永军, 张晶晶, 李修华，等）. Transactions of the Chinese Society for Agricultural Machinery（农业机械学报）, 2016, 47(3): 292.
[10] Gitelson A A, Keydan G P, Merzlyak M N. Geophysical Research Letters, 2006, 33(11): 431.
[11] Dash J, Curran P J. Evaluation of the MERIS Terrestrial Chlorophyll Index (MTCT), Geoscience and Remote Sensing Symposium, IGARSS’04, 2004.
[12] Dash J, Curran P J. The Meris Terrestrial Chlorophyll Index, International Journal of Remote Sensing, 2004, 25(23): 5403.
[13] Carmona F, Rivas R, Fonnegra D C. European Journal of Remote Sensing, 2015, 48: 319.
[14] Barnes E M, Clarke T R, Richards S E, et al. Coincident Detection of Crop Water Stress, Nitrogen Status and Canopy Density Using Ground-based Multispectral Data. International Conference on Precision Agriculture and Other Resource Management，2015.
[15] Penuelas J, Baret F, Filella I. Photosynthetica, 1995, 31(2): 221.