The Study On Prediction Chlorophyll Concentration Based On Leaf Hyperspectral Parameters Of Rice

Two experiments were conducted to investigate the changes of chlorophyll concentration and transmittance/absorptance/reflectance spectra of rice leaf at different fertilizer levels and growth stages. This study showed that leaf chlorophyll concentration and spectra characteristic parameters were sensitive to increased fertilizer levels.The trend of chlorophyll concentration especially chlorophyll a and total chlorophyll concentration can show the nitrogen nutration status of rice. As a result, it was more important to predict chlorophyll a and total chlorophyll concentration than to predict chlorophyll b concentration using hyperspectrum for the purpose of prediction nitrogen nutration status. The leaf hyperspectral features of different nitrogen fetilizer levels and growth stages could be used to explain chlorophyll concentration status. The correlations between chlorophyll concentration and spectrum of visible and NIR wavebands were analyzed to select chlorophyll concentration sensitive bands. Four absorption peak of chlorophyll (440.439nm/480.188nm/630.610nm/680.450nm) and the characteristic wavebands selected from the peak, dip and zero positions of the first derivative curve and spectral curve in the three edge(blue edge, yellow edge and red edge)regions, and models prdicting chlorophyll concentration were developed by these characteristic wavebands or their variables.The results indicated relationships between predicted and measured chlorophyll concentration were all significantly correlated at P<0.01 level. Accuracy was better predicting chlorophyll a and total chlorophyll concentration than chlorophyll b concentration. The algorithms of BTI/YTI/RTI (blue/yellow/red edge transmittance spectra chlorophyll index), BAI/YAI/RAI(blue/yellow/red edge absorptance spectra chlorophyll index), and BRI/YRI/RRI were proved to be better than spectra position variables, area variables and those reported vegetation indices. The accuracy of prediction chlorophyll a concetration using those models with YTI610.510.YTI570.169, BTI. RTI were 71.1%,73.1%,70.1% and 71.5% respectively, and the accuracy of prediction total chlorophyll concentration were 70.4%,75.1%,70.2% and 73.7% respectively. The investigation drew the preliminarily conclusion that it is feasible to predict chlorophyll concentration by these vegetation indices BTI/YTI/RTI, BAI/YAI/RAI, and BRI/YRI/RRI.