Hyperspectral estimation of leaf chlorophyll content in mycorrhizal inoculated soybean under drought stress

Abstract: Researchers have done a lot of studies on spectral characteristics to estimate chlorophyll content of crops both in China and abroad by using hyperspectral technology. However, there are few studies on hyperspectral estimation of leaf chlorophyll content of arbuscular mycorrhizal inoculated soybean under drought stress. A series of soybean pot experiments were set up to determine the visible spectrum/ near infrared (VIS/NIR) reflectance spectral characteristics of arbuscular mycorrhizal inoculated and non-inoculated soybean under different drought stress. Three stress levels were set up, and each stress level included inoculated and non-inoculated treatments. Sixty six samples were collected after 30, 45, and 64 days of inoculation, the spectral reflectance were measured using ASD Fieldspec3 spectroradiometer in darkroom. Meanwhile, leaf chlorophyll content of each sample was measured using SPAD-502 device. Forty six random samples were selected to establish the model. The remaining samples were used to test model (i.e. comparing model-simulated results with measured results). To establish the model, correlation analysis was conducted between the first derivative values of original reflectance spectrum and leaf chlorophyll content of soybean. The bands with correlation coefficient of leaf chlorophyll content greater than 0.8 were selected as the sensitive bands of chlorophyll content inversion. After that, the differential area based on curved estimation method and linear combination of sensitive bands was calculated. The regression models with the independent variables of the differential area and the first derivatives as dependent variables were established using the stepwise regression approaches. The results showed that: 1) soybean shoot dry weight and leaf chlorophyll content were reduced with the strengthening of drought stress, indicating that drought stress inhibited soybean growth. But at each stress level, the inoculated soybean grew better and its leaf chlorophyll content were higher (P<0.05) than non-inoculated ones after 45 and 64 days inoculation, highlighting that inoculation could reduce the impact of drought on plant growth to some extent; 2) these differences were also shown on the leaf reflectance spectral characteristics of the inoculated and non-inoculated soybeans, especially on the visible waveband in the late two measuring dates at which the reflectance increased gradually with the increment of drought stress, and at the same drought stress level, the reflectance of inoculated soybean was lower (P<0.05) than non-inoculated; 3) the bands of 512-523, 589-621, 630-666, 679-697 and 748-761 nm were sensitive bands of estimating chlorophyll content of soybean. In addition, soybean leaf chlorophyll content was significantly correlated with first derivative value and differential area variables of these bands; and 4) the first derivative of reflectance model based on stepwise regression method was the optimal inversion model with the determination R2 and prediction R2 of 0.90 and 0.84, respectively, indicating possibility of estimating inoculated soybean leaf chlorophyll content. This study provided technical support for promoting mycorrhizal technology in arid and semi-arid agricultural regions, and for monitoring growth of arbuscular mycorrhizal inoculated soybean using remote sensing in combination with hyperspectral technology.