Canonical correlation relationships among spectral and phytometric variables for twenty winter wheat fields

Abstract

To assess the adequacy of remote and conventional agricultural surveys, aircraft spectrometry of 20 commercial winter wheat fields was obtained in southern Moldova, CIS, and related to plant measurements using canonical correlation. Measurements were made over the 420–1200 nm range at intervals of 10 nm. A total of 510–575 reflectance spectra were obtained for each of the five phenophases studied and the extrema of the averaged curves of bidirectional reflectance factor ( b λ ) were determined. The blue and red minima, well as the relative maximum in the visible region, were positionally highly stable during the season (about 480 nm, 550 nm, and 670 nm, respectively). In the NIR region the b λ curve rose smoothly betwee 750 nm and 1070 nm, and its steepness was probably determined by soil background reflectance. The values of b 480, b 550, b 670 and b 860 were the s spectral variables chosen, and they were related, through canonical correlation analysis, to phytometric variables (dry aboveground phytomass, plant height, plant density, and percent plant cover) obtained from the corresponding ground-based agronomic measurements. There was a close relation between spectral and phytometric variables when the canopy was green (the first canonical correlation coefficient was 0.974 in the booting stage), but the closeness decreased with crop maturation. Canonical variables for spectral reflectance explained, at most, about 75% of the variance in phytometric variables while the first canonical variable for phytometric features accounted for 90% of the spectral variance. The NIR and red reflectances contributed most to canonical variables,and had the highest correlations with the plant measurements. Among the phytometric variables, percent plant cover and dry above-ground phytomass related more closely to spectral reflectance than did plant height and plant density. On the whole, the results obtained confirm the high information potential of aircraft spectrometry for estimating winter-wheat crop conditions.