节点文献

基于冠层时序植被指数的冬小麦单产预测

Using Canopy Time-Series Vegetation Index to Predict Yield of Winter Wheat

  • 推荐 CAJ下载
  • PDF下载
  • 不支持迅雷等下载工具,请取消加速工具后下载。

【作者】 项方林李鑫格马吉锋刘小军田永超朱艳曹卫星曹强

【Author】 XIANG FangLin;LI XinGe;MA JiFeng;LIU XiaoJun;TIAN YongChao;ZHU Yan;CAO WeiXing;CAO Qiang;Nanjing Agricultural University/National Engineering and Technology Center for Information Agriculture/Key Laboratory for Crop System Analysis and Decision Making, Ministry of Agriculture and Rural Affairs/Engineering and Research Center for Smart Agriculture, Ministry of Education/Jiangsu Key Laboratory for Information Agriculture;

【通讯作者】 曹强;

【机构】 南京农业大学/国家信息农业工程技术中心/农业农村部农作物系统分析与决策重点实验室/智慧农业教育部工程研究中心/江苏省信息农业重点实验室

【摘要】 【目的】研究冬小麦冠层时序植被指数的动态变化规律并基于其构建单产预测模型,为田间实时、准确获取作物单产信息提供有效的技术手段。【方法】本研究于2017—2019年在江苏省兴化市万亩粮食产业园开展不同品种及氮肥水平的田间小区试验,利用主动传感器RapidSCAN CS-45获取冠层归一化红边植被指数(normalized difference red edge,NDRE)和归一化植被指数(normalized difference vegetation index,NDVI),基于双Logistic函数拟合时序植被指数并提取曲线特征参数,进而分析各特征参数与单产的相关关系,并以独立试验数据对单产预测模型进行验证。【结果】NDRE在孕穗期和抽穗期与单产关系最好,R~2达到0.84以上;通过多元逐步线性回归法发现,利用2个或多个时期NDRE预测单产的效果较单生育时期有所提高,且第一和第二被选择的时期分别为拔节期和孕穗期。基于全生育时期相对NDRE(relative NDRE,RNDRE)和相对NDVI(relative NDVI,RNDVI)构建时序曲线,并利用曲线特征参数建立单产预测模型,其中RNDRE和RNDVI的最大值、累积值及增长速率与单产关系较好。利用独立试验数据对上述单产预测模型进行检验,结果表明基于RNDRE时序曲线最大值和累积值所构建的单产模型验证效果较好,R2~大于0.80,相对均方根误差和相对误差均小于10%,其验证效果优于单时期或多时期基于NDRE的预测模型,且优于基于NDVI构建的单产模型。【结论】基于冠层时序植被指数提取的特征参数RNDRE最大值和累积RNDRE具有良好估测单产的潜力,研究结果为田间进行实时、准确预测冬小麦单产提供了技术支持。

【Abstract】 【Objective】The research elaborated the dynamic change trend of canopy time series vegetation index in winter wheat. The yield prediction model was constructed based on time series vegetation index in wheat, which provided effective technical support for obtaining crop yield information timely and accurately. 【Method】From 2017 to 2019, the field experiments involving different nitrogen(N) rates and varieties were conducted in Ten Thousand Acres Grain Industrial Park located in Xinghua, Jiangsu province. The normalized difference red edge(NDRE) and normalized difference vegetation index(NDVI) were obtained from the active canopy sensor RapidSCAN CS-45. The curve of time-series vegetation index was fitted based on the double logistic function, and the characteristic parameters of curve were extracted. The correlation between each characteristic parameters and yield were analyzed. The yield estimation models were verified with independent test data. 【Result】The results of the study indicated that the relationship between NDRE and yield was performed well at booting and heading stage, and R~2 of them was 0.86 and 0.85, respectively. The results of multiple stepwise linear regression showed that the yield prediction models could be improved using NDRE of two or more growth stages, compared with using single growth stage information. The first and second selected periods were jointing and booting stage, respectively. Based on the relative NDRE(RNDRE) and relative NDVI(RNDVI) of the whole growth period, the time-series curve was constructed and the yield prediction models were developed using the characteristic parameters of the curve. The maximum value, accumulative value and growth rate of RNDRE and RNDVI time series curve had a good relationship with the yield. The yield prediction models based on the maximum and accumulative values of RNDRE performed satisfactorily with validation using independent data, the R~2 was greater than 0.80, and the relative root mean square error and relative error were less than 10%. The validation effect was better than NDRE-based prediction model with the single-period or multi-period, which was better than NDVI-based yield predicted model. 【Conclusion】The maximum and the accumulative RNDRE extracted from the canopy time series vegetation index had a good potential to estimate the yield, which provided technical support for real-time and accurate yield prediction in the field.

【基金】 国家重点研发计划(2016YFD0300608);国家自然科学基金青年科学基金(31601222);中央高校基本科研业务费专项资金项目(KJQN201725);江苏现代农业产业技术体系建设专项资金(JATS[2019]433,JATS[2019]141)
  • 【文献出处】 中国农业科学 ,Scientia Agricultura Sinica , 编辑部邮箱 ,2020年18期
  • 【分类号】S512.11
  • 【被引频次】7
  • 【下载频次】413
节点文献中: 

本文链接的文献网络图示:

本文的引文网络