Multisite analyses of spectral-biophysical data for wheat

Reflectance factors and biophysical plant measurements for wheat (Triticum aestivum L.) experiments conducted at Lubbock, Texas, Manhattan, Kansas, Phoenix, Arizona, Sidney, Montana, and Weslaco, Texas were fit by various equation forms for six currently used vegetation indices (VI): n-space greenness (GVI), perpendicular (PVI), NIR/RED ratio (RVI), soil adjusted (SAV12), normalized difference (NDVI), and transformed soil adjusted (TSAVI). The objective was to produce relations from the data pooled across all locations that could be recommended for general use for wheat. Data were analyzed by premaximum leaf area (pre-Lmax), post-maximum leaf area (post-Lmax), and whole season portions of the growing seasons, Leaf area index (L) was best estimated from RVI and TSAVI by linear equations, from NDVI and TSAVI by exponential equations, and from the orthogonal indices GVI and PVI equally well by power and quadratic equation forms. These equation forms gave coefficients of determination (R2) that ranged from 0.72 to 0.86, and root mean square errors (RMSE) in estimating L that ranged from 0.63 to 0.90 across locations and measurements that differed in soils, sun angles, cultivars, agronomic treatments, and radiometers. The single best equation form for estimating L from VI across all VI was the power form. The orthogonal indices GVI and PVI were more responsive to canopy architecture than the ratio vegetation indices, and GVI ranked as the best single index. Equations for estimating fractional absorbed photosynthetically active radiation (FPAR) from VI given herein agree well with empirical and semitheoretical equations and their coefficients found in the literature. Our results demonstrate the robustness of vegetation indices across experiment variables and measurement conditions, and provide general functional relations that should be useful for wheat.