基于遥感融合数据的玉米叶片叶绿素含量反演方法  

Chlorophyll Content Inversion Method in Maize Leaves Based onRemote Sensing Fusion Data

作　　者：李宣 宋开山[2] 刘吉平[1] 朱冰雪 LI Xuan;SONG Kai-shan;LIU Ji-ping;ZHU Bing-xue(College of Geographic Science and Tourism,Jilin Normal University,Siping 136000,China;Northeast Institute of Geography and Agroecology,Chinese Academy of Sciences,Changchun 130102,China)

机构地区：[1]吉林师范大学地理科学与旅游学院,四平136000 [2]中国科学院东北地理与农业生态研究所,长春130102

出　　处：《科学技术与工程》2025年第11期4428-4437,共10页Science Technology and Engineering

基　　金：中国科学院战略重点研究项目(XDA28050400);吉林省科技发展计划重点研发项目(20230202040NC);国家民用空间基础设施陆地观测卫星共性应用支撑平台项目(2017-000052-73-01-001735)。

摘　　要：玉米是中国重要的粮食储备作物之一,其产量直接影响着国家的粮食安全。玉米的叶绿素含量与光合能力密切相关,且显著影响叶片的光合速率和植被生产力,是作物生长监测、病虫害监测和成熟度预测中重要的作物参数。实时、准确监测对于玉米参数和产量预测具有重要意义。以吉林省四平市梨树县典型黑土区为研究区,为解决Sentinel-2卫星重访周期间可能会出现的有效影像缺失问题,提出一种基于Sentinel-2与MODIS影像融合数据的玉米叶片叶绿素反演方法。基于融合影像提取叶绿素敏感特征波段,通过3种机器学习算法:随机森林(random forest, RF),梯度提升树(gradient boosting decision tree, GBDT)和极限梯度提升(extreme gradient boosting, XGBOOST)构建玉米叶片叶绿素含量估测模型,估测玉米叶片叶绿素含量并验证模型精度。得到结论如下:(1)使用ESTARFM数据融合算法模拟的数据与真实影像保持了高度的相关性;(2)影像缺失日期叶片叶绿素反演模型构建中,以融合影像波段反射率和植被指数为输入变量,XGBOOST模型拟合精度有较好的效果。研究结果表明,结合融合影像特征波段和机器学习算法可以实现在影像缺失日期叶片叶绿素含量精准估算,这有效提升了玉米叶绿素含量获取的时间精度,为影像缺失等情况的逐日或大范围叶片叶绿素含量反演研究提供了新的方法,同时为时间间隔更短、更多种类作物生理生化参数的精细化监测提供新思路。Corn is one of the important grain reserve crops in China,and its yield directly impacts national food security.The chlorophyll content of corn is closely related to its photosynthetic capacity and significantly affects the photosynthetic rate of the leaves and vegetation productivity.It is an important crop parameter for monitoring crop growth,pest and disease surveillance,and maturity prediction.Real-time and accurate monitoring is of great significance for corn parameters and yield prediction.This study was conducted in the typical black soil area of Lishu County,Siping City,Jilin Province.To solve the problem of missing effective images that may occur during the revisit period of Sentinel-2 satellites,a method for retrieving corn leaf chlorophyll based on the fusion data of Sentinel-2 and MODIS images was proposed.Using fused imagery,three machine learning algorithms were employed:random forest(RF),gradient boosting decision tree(GBDT),and extreme gradient boosting(XGBOOST)to construct a model for estimating corn leaf chlorophyll content,and the accuracy of the model was verified.The conclusions obtained were as follows.The data simulated using the ESTARFM data fusion algorithm maintained a high correlation with the real imagery.Among the leaf chlorophyll inversion models for missing image dates,where input variables included fused image band reflectance and vegetation index,the XGBOOST model showed good fitting accuracy The research demonstrates that accurate estimation of leaf chlorophyll content can be achieved even on days with missing imagery,when fusion image feature bands are integrated with machine learning algorithms.This notably improves the temporal precision of corn chlorophyll content measurement,presenting a novel method for daily or large-scale inversion studies of leaf chlorophyll content,particularly in scenarios involving image gaps.Furthermore,it illuminates the potential for refined monitoring of physiological and biochemical parameters across a wider range of crops,with shortened time int

关 键 词：玉米 叶绿素含量 遥感反演 机器学习 Sentinel2-MODIS融合 ESTARFM 

分 类 号：P237[天文地球—摄影测量与遥感]