Estimation of nitrogen and carbon content from soybean leaf reflectance spectra using wavelet analysis under shade stress

Changes in leaf carbon and nitrogen traits are effective parameters for evaluating whether plants are growing normally and for guiding field management measures. Therefore, a nondestructive estimation of soybean leaf carbon content (LCC) and leaf nitrogen content (LNC) is important for managing soybean production. This study aims to investigate the properties of soybean LCC, LNC and leaf carbon and nitrogen ratio (LCNR) under different light conditions and to construct the estimation models for soybean LCC, LNC and LCNR on the basis of hyperspectral reflectance. A 3-year field trial was conducted from 2013 to 2015 to investigate the effect of five shade treatments (from 57% to 18% of normal photosynthetic active radiation) by regulating the distance between maize and soybean rows in an intercropping system. Results showed that the LNC is increased, and the LCC is decreased with the increase in shading level. A typical leaf hyperspectral reflectance showed that the changes in near infrared and around the green peak are approximately proportional to the LCC and analogously inverse to the LNC. The corresponding first-order derivative presented the changes in the red-edge positions as a 'blue shift' with the increase in shading level. General vegetation indices (VIs), uninformative variable elimination-partial least squares (UVE-PLS) and continuous wavelet transform (CWT) were used to develop the models. The linear models constructed using the CWT, compared with the VIs and LTVE-PLS, had the highest determination coefficient and the lowest root mean square error (RMSE). The models were validated using independent data from 2015 with RMSE values of 1.9789, 0.6132 and 2.1587 for the LCC, LNC and LCNR, respectively. Moreover, these models obtained relative errors of 4.5%, 13.2% and 20% for the LCC, LNC and LCNR, respectively. Results indicated that soybean LNC and LCC can be accurately estimated using the CWT for assessing the physiological situation.