基于Sentinel-2卫星数据的水稻叶片叶绿素含量反演研究

叶绿素含量是评价农作物健康状况、生产能力和环境胁迫的重要指标，实时、快速、准确获取农作物叶片叶绿素含量对监测农作物生长状况具有重要意义。遥感是获取区域和全球农作物叶片叶绿素含量的有效途径，但已有的作物叶片叶绿素含量遥感反演研究未充分考虑下垫面背景的干扰，影响了反演精度。为此，以Sentinel-2遥感卫星影像为数据源，结合典型水稻田的观测数据，使用PROSAIL辐射传输模型建立了水稻田叶片叶绿素含量反演查找表，评估了利用绿光波段和不同红边波段构建的叶绿素指数（CI）和两个不同红边波段构建的Zarco and Miller指数（ZM）反演叶片叶绿素含量的差异，引入G（Greenness index）指数减小背景干扰对叶片叶绿含量反演的影响。研究结果表明：（1）基于不同波段构建的光谱指数反演的叶片叶绿素含量精度存在差异，其中CI₇₄₀(R2=0.79, RMSE=9.02 μg·cm-2) 反演精度最高，其次为ZM(R2=0.71, RMSE=10.53 μg·cm-2)、CI₇₀₅(R2=0.69, RMSE=9.17 μg·cm-2) 和CI₇₈₃(R2=0.67, RMSE=10.84 μg·cm-2)；（2）水稻叶片叶绿素含量反演结果受背景影响明显，特别在水稻生长早期，由于背景干扰较大，反演结果明显偏低平均相对误差（MRE）为-18.87%~-31.94%]；（3）引入G指数构建的CI/G和ZM/G可以有效消除背景的影响，提高水稻叶片叶绿素含量反演精度（MRE为8.11%~18.11%）。结果对提高水稻不同叶面积指数水平下的叶片叶绿素含量遥感反演精度具有重要参考意义。

Inversion of Rice Leaf Chlorophyll Content Based on Sentinel-2 Satellite Data

Chlorophyll content is an important indicator of crop health, plant productivity, and environmental stress. Real-time, fast and accurate acquisition of leaf chlorophyll content of crops is of significant for monitoring crop growth. Remote sensing is an effective way to retrieve leaf chlorophyll content of crops at regional and global scales. However, previous studies retrieving leaf chlorophyll content of crops does not fully consider the impact of underlying surface background, limiting retrieval accuracy. To this end, this paper aims at the inversion of rice leaf chlorophyll content from Sentinel-2 remote sensing satellite data using a look-up table based approach. The look-up table was simulated using the PRAOSAIL radiation transfer model. The applicability of chlorophyll indices (CI) calculated from the reflectance of the green band and different red-edge bands and the spectral index (Zarco and Miller, ZM) constructed by two different red edge bands in inverting leaf chlorophyll content was evaluated using field measurements. The greenness index (G) was integrated with CI and ZM to constrain the impact of background on the inversion of leaf chlorophyll content. The main findings of this study are: (1) The accuracy of leaf element content inversion based on the spectral index constructed in different bands is different, and CI₇₄₀ performed the best (R2=0.79, RMSE=9.02 μg·cm-2), followed by ZM (R2=0.71, RMSE=10.53 μg·cm-2), CI₇₀₅(R2=0.69, RMSE=9.17 μg·cm-2), and CI₇₈₃(R2=0.67, RMSE=10.84 μg·cm-2); (2) The inverted leaf chlorophyll content is significantly affected by the background, especially at the early stage of rice growth. The inverted leaf chlorophyll content was systematically lower than observations (mean relative error (MRE) in the range from -18.87% to -31.94%) owing to strong background interference; (3) CI/G and ZM/G can effectively eliminate the influence of background and improve the accuracy of rice leaf chlorophyll inversion. At the early stage of rice growth, inversion based on CI/G and ZM/G significantly improves agreement between inverted and observed leaf chlorophyll content (MRE in the range from 8.11% to 18.11%). These findings are of great significance for improving the inversion of leaf chlorophyll content under different leaf area index levels of rice from remote sensing data.