摘要
高光谱遥感能够快速无损地估测作物生长性状及产量,这为作物规模化育种的田间评价与选择提供了高效手段。选用生育时期相似、生长性状有差异的52份大豆品种(系)进行2年田间试验,在盛花期(R2)、盛荚期(R4)及鼓粒初期(R5)测定大豆冠层反射光谱,同步测定大豆叶面积指数(LAI)和地上部生物量(ABM),收获后测定产量。针对不同生育时期冠层光谱与生长性状及产量进行偏最小二乘回归(PLSR)分析。结果表明:不同生育时期LAI的PLSR模型可以解释LAI总变异的54.4%~61.0%;不同生育时期ABM的PLSR模型可以解释ABM总变异的65.5%~67.0%;R5期是利用冠层光谱估测产量的最佳生育时期,其PLSR模型可以解释产量总变异的66.1%。本研究结果可望为大豆规模化育种中大量试验材料的田间长势监测和产量估测提供快速无损预测的技术支持。
Hyperspectral remote sensing technique as a fast and non-destructive method can estimate growth traits and yield in crop,which provides an effective tool for field evaluation and selection in large-scale breeding programs. In the present study,a field experiment comparing 52 soybean varieties with similar flowering and maturity dates were tested a randomized blocks design with three replications in two years. The measurement of leaf area index( LAI) and aboveground biomass( ABM) was synchronized with the information collection of the canopy hyperspectral reflectance at R2,R4,and R5 growth stages. The seed yield was acquired after harvest. The partial least squares regression( PLSR) between canopy spectral reflectance at different growth stages and growth traits and seed yield showed that the PLSR models of ABM and LAI at different growth stages could explain65. 5% ~ 67. 0% and 54. 4% ~ 61. 0% of the total variance of ABM and LAI,respectively,and R5 stage performed as the best of the three growth stages for predicting yield using canopy spectral reflectance with an explanation up to 66. 1% of the total seed yield variance. The results can serve a quick and non-destructive technique for monitoring field growing status and predicting yield in large-scale soybean breeding programs.
出处
《大豆科学》
CAS
CSCD
北大核心
2015年第3期414-419,426,共7页
Soybean Science
基金
国家重点基础研究发展计划"973计划"(2011CB1093)
国家高技术研究发展计划"863计划"(2011AA10A105)
国家公益性行业(农业)专项经费项目(201203026-4)
教育部111项目(B08025)
教育部创新团队项目(PCSRT13073)
中央高校基本科研业务费项目(KYZ201202-8)
江苏省优势学科建设工程专项
江苏省JCIC-MCP项目资助
关键词
大豆
高光谱
叶面积指数
地上部生物量
产量
偏最小二乘回归
Soybean
Hyperspectral reflectance
Leaf area index(LAI)
Aboveground biomass(ABM)
Yield
Partial least squares regression(PLSR)