Estimating leaf chlorophyll of barley at different growth stages using spectral indices to reduce soil background and canopy structure effects

摘要

Monitoring in situ chlorophyll (Chl) content in agricultural crop leaves is of great importance for stress detection, nutritional state diagnosis, yield prediction and studying the mechanisms of plant and environment interaction. Numerous spectral indices have been developed for chlorophyll estimation from leaf- and canopy-level reflectance. However, in most cases, these indices are negatively affected by variations in canopy structure and soil background. The objective of this study was to develop spectral indices that can reduce the effects of varied canopy structure and growth stages for the estimation of leaf Chl. Hyperspectral reflectance data was obtained through simulation by a radiative transfer model, PROSAIL, and measurements from canopies of barley comprising different cultivars across growth stages using spectroradiometers. We applied a comprehensive band-optimiration algorithm to explore five types of spectral indices: reflectance difference (RD), reflectance ratio (RR), normalized reflectance difference (NRD), difference of reflectance ratio (DRR) and ratio of reflectance difference (RRD). Indirectly using the multiple scatter correction (MSC) theory, we hypothesized that RRD can eliminate adverse effects of soil background, canopy structure and multiple scattering. Published indices and multivariate models such as optimum multiple band regression (OMBR), partial least squares regression (PLSR) and support vector machines for regression (SVR) were also employed. Results showed that the ratio of reflectance difference index (RRDI) optimized for simulated data significantly improved the correlation with Chl (R~2 = 0.98, p ＜ 0.0001) and was insensitive to LAI variations (1-8), compared to widely used indices such as MCARI/OSAVI (R~2 = 0.64, p ＜ 0.0001) and TCARI/OSAVI (R~2 = 0.74, p ＜ 0.0001). The RRDI optimized for barley explained 76% of the variation in Chl and outperformed multivariate models. However, the accuracy decreased when employing the indices for individual growth stages (R~2 ＜ 0.59). Accordingly, RRDIs optimized for open and closed canopies improved the estimations of Chl for individual stages before and after canopy closure, respectively, with R~2 of 0.65 (p ＜ 0.0001) and 0.78 (p ＜ 0.0001). This study shows that RRDI can efficiently eliminate the effects of structural properties on canopy reflectance response to canopy biochemistry. The results yet are limited to the datasets used in this study; therefore, transfer-ability of the methods to large scales or other datasets should be further evaluated.