Hyperspectral characteristics and inversion model estimation of winter wheat under different elevated CO2 concentrations

Abstract

ABSTRACT As carbon dioxide (CO2) is required for plants photosynthesis, elevated CO2 (eCO2) concentrations have potential impacts on plant growth and development. The leaf area index (LAI) and soil and plant analysis development (SPAD), which are often used for characterize the chlorophyll content of plants, are important parameters for characterizing plant growth. The purpose of this study was to investigate the effects of different eCO2 concentrations on winter wheat growth and select sensitive spectral parameters to establish LAI and SPAD estimation models. A field experiment in which winter wheat was exposed to different eCO2 concentrations was performed using open-top chambers (OTCs) during the winter wheat growing season from 2017 to 2018. The experimental treatments consisted of exposure to the ambient CO2 concentration (CK), 80 μmol mol–1 CO2 above CK (T 1), and 200 μmol mol–1 CO2 above CK (T 2). The canopy spectral reflectance, LAI, and SPAD were measured during the main growth stages of the winter wheat. The results showed that no significant differences in LAI and SPAD were found under different treatments. Different eCO2 concentrations did not change the reflectance curves, solely affecting the reflectance. Elevated CO2 conditions induced the red edge parameters red shift first and then blue shift. The first derivative reflectance at 755 nm (R’ 755), red edge position (λ Red), ratio of the red edge area to the blue edge area (SDRed/SDBlue), and normalized value of the red edge area and blue edge area ((SDRed – SDBlue)/(SDRed + SDBlue)) were highly correlated with the LAI. The first derivative reflectance at 764 nm (R’ 764), SDRed/SDBlue, ratio of the red edge area to the yellow edge area (SDRed/SDYellow), (SDRed – SDBlue)/(SDRed + SDBlue), and normalized value of the red edge area and yellow edge area ((SDRed – SDYellow)/(SDRed + SDYellow)) were significantly correlated with the SPAD. The optimal regression model for the LAI was y = 0.49x 0.74, simulated by SDRed/SDBlue; that for the SPAD was y = – 0.035x 2–1.96x + 25.83, simulated by SDRed/SDYellow. The coefficient of determination (R 2) values were 0.49 and 0.61, respectively, and the root mean square error (RMSE) values were 0.38 and 1.51, respectively. Overall, our results indicate that inversion models based on SDRed/SDBlue and SDRed/SDYellow can be used to estimate the LAI and SPAD values under eCO2 concentration conditions during the winter wheat growth period.