Hyperspectral estimation model for chlorophyll concentrations in top leaves of rice
张玉森,姚霞,田永超,朱艳,杨杰,曹卫星
DOI: https://doi.org/10.3321/j.issn:1000-0933.2009.12.031
2009-01-01
Abstract:Leaf chlorophyll (Chl) status is a key index for evaluating crop photosynthetic efficiency and nutritional stress. Non-destructive and quick assessment of leaf Chl status is needed for growth diagnosis and nitrogen management in crop production. The objective of this study was to determine the relationship between leaf Chl concentration and spectral reflectance characteristics, and to develop useful hyperspectral indices for nondestructive and quick estimation of Chl in rice (Oryza sativa). Four field experiments with different nitrogen application rates and rice cultivars were conducted at different eco-sites over three years. Time course measurements were taken on hyperspectral reflectance of 350~2500 nm and Chl concentration in four top leaves, and the simple ratio spectral index (SR[λ1,λ2]) and normalized difference spectral index (ND[λ1,λ2]) were calculated with all combinations of two wavelengths (λ1 and λ2 nm) and other Chl sensitive parameters. Analysis showed that the best indicators for estimating Chl concentration in rice leaves were narrow-band hyperspectral indices calculated in the red edge region, and that the optimum wavelengths were different for specific Chl components. The best SR indices for estimation of chlorophyll a (Chla), total chlorophyll (Chla+b), and chlorophyll b (Chlb) were SR (724,709), SR (728,709) and SR (749,745), with coefficient of determination (R~2) values of 0.944, 0.946 and 0.905. The best ND indices were ND (780,709), ND (780,712) and ND (749,745), with R~2 values of 0.944, 0.946 and 0.905. Modifying the above spectral indices with the reflectance at 445 nm reduced the effect of differences in leaf surface reflectance, and increased the extrapolation potential for the model. Tests with another independent datasets showed that the modified SR indices: mSR (724,709), mSR (728,709) and mSR (749,745), and the modified ND indices: mND (780,709), mND (780,712) and mND (749,745) were better predictors of Chla、Chla+b and Chlb concentrations with RMSE values of 0.168, 0.190, 0.052, 0.159, 0.176 and 0.052, respectively, and RE values of 8.18%, 7.74%, 13.01%, 8.26%, 7.59%, 12.96%. These values indicated that the above spectral parameters can be used to estimate leaf Chl concentrations with good precision and accuracy.