摘要
为建立不同生育时期冬小麦叶面积指数(LAI)的高光谱遥感预测模型,2017年在荥阳和鹤壁大田区域进行野外试验,利用便携式光谱仪ASD FieldSpec Handheld测量不同生育时期冬小麦冠层高光谱数据,使用LAI2200冠层分析系统采集冬小麦冠层LAI。通过对高光谱数据进行不同形式的变换以及高光谱特征变量的计算,并与叶面积指数进行相关分析。结果表明,在拔节-抽穗期,LAI与Dr(红边幅值)、SDr(红边面积)、VI_3(红边面积SDr与蓝边面积SDb的比值指数)、VI_5(红边面积SDr与蓝边面积SDb的归一化指数)、VI_6(红边面积SDr与黄边面积SDy的归一化指数)的相关性较大,相关系数均大于0.85;在开花-乳熟期,LAI与Rr(红谷反射率)、VI_1(绿峰反射率Rg与红谷反射率Rr的比值指数)、VI_2(绿峰反射率Rg与红谷反射率Rr的归一化指数)、VI_3、VI_5的相关性较大,相关系数均大于0.7,且均通过0.01水平显著性检验。因此,拔节-抽穗期选择变量Dr、SDr、VI_3、VI_5、VI_6作为估算模型的自变量;开花-乳熟期选择变量Rr、VI_1、VI_2、VI_3、VI_5作为估算模型的自变量。拔节-孕穗期叶面积指数单变量估算模型中大部分变量的二次模型决定系数较大,其中VI_3、VI_5、lg(1/ρ676)、dρ750/dλ750的二次模型决定系数超过0.6,拟合程度较高,同时dρ750/dλ750的RMSE值最小,因此认为以dρ750/dλ750为自变量的二次模型最优。开花-乳熟期单光谱变量建立的叶面积指数估算各类模型中大部分参数的指数模型决定系数较大,其中Rr、VI_3、VI_5的指数模型决定系数超过0.7,拟合程度最高,同时VI_5的RMSE值最小,因此认为以VI_5为自变量的指数模型最优。
To explore the relationship between hyperspectral reflectance and leaf area index(LAI),the experiment was conducted in Xingyang and Hebi in 2017.Winter wheat canopy hyperspectral data was measured at different growth stages by using the ASD FieldSpec HandHeld,and winter wheat LAI was collected at the same time by using LAI 2200 canopy analysis system.By correlation analysis between the countdown logarithms,first derivative,second derivative transformations of hyperspectral and LAI,the sensitivity spectral band and the index were obtained to estimate winter wheat LAI.The results showed that the correlation coefficients between LAI and Dr,SDr,VI3,VI5 and VI6 were relatively higher from jointing to heading stage,which were more than 0.85.The correlation coefficients between LAI and Rr,VI1,VI2,VI3 and VI5 were relatively higher from flowering to milking stage,which were more than 0.7,at 0.01 significance level.Therefore,from jointing to heading stage,vari-ables such as Dr,SDr,VI3,VI5 and VI6 were used as independent variables to estimate the model,and from flowering to milking stage,variables such as Rr,VI1,VI2,VI3 and VI5 were used as independent variables to estimate the model.From jointing to booting stage,the model of the most parameters in single spectral variables estimated LAI model had larger r2(more than 0.6),including VI3,VI5,lg(1/ρ676)and dρ750/dλ750,with an maximum fitting degree,while the RMSE of dρ750/dλ750 is minimum,which is the optimal variable model.From flowering to milking stage,the exponential model of the most parameters in single spectral variables estimated LAI model had larger r2(more than0.7),including Rr,VI3,VI5 models with maximum fitting degree,while the RMSE of VI5 is minimum,which is the optimal variable exponential model.
作者
李军玲
彭记永
LI Junling;PENG Jiyong(Henan Key Laboratory of Agrometeorological Ensuring and Applied Technique,CMA/Henan Institute of Meteorological Sciences,Zhengzhou,Henan 450003,China)
出处
《麦类作物学报》
CAS
CSCD
北大核心
2018年第8期979-987,共9页
Journal of Triticeae Crops
基金
国家自然科学基金联合基金项目(U1204406)
中国气象局农业气象保障与应用技术重点开放实验室开放研究基金项目(AMF201609)
河南省气象局科学技术研究项目(KM201814)
关键词
叶面积指数
高光谱数据
拔节-孕穗期
开花-乳熟期
模型精度
Leaf area index
Hyperspectral data
Jointing to heading stage
Flowering to milkingstage
Model accuracy