Imaging spectroscopy for early detection of nitrogen deficiency in grass swards

Abstract

The potential of an experimental imaging spectroscopy system with high spatial (0.16–0.28 mm²) and spectral resolution (5–13 nm) was explored for early detection of nitrogen (N) stress. From June through October 2000, a greenhouse experiment was conducted with 15 Lolium perenne L. mini-swards and 5 N treatments. Images were recorded twice a week. With the experimental system, spectra of grass leaves in the canopy can be obtained. Treatment effects on ground cover (GC) and changes in leaf spectral characteristics were studied separately. Leaf pixels with similar reflection intensity were grouped in intensity classes (IC). An index of reflection intensity (IRI) indicates the percentages of strongly reflecting grass pixels. Blue edge, green edge and red edge positions were calculated for each IC. Both GC and IRI increased until harvest, with largest increases for liberal N treatments. The width of the chlorophyll-dominated absorption band around 680 nm (CAW) increased up to a maximum of 133 nm for both liberal and limited N in the first two weeks after harvesting. CAW decreased for limited N in the second half of the growth period in contrast to liberal N. At harvest CAW explained 95% of the variation in relative dry matter (DM) yield between treatments. Principal component analyses showed an intertwined response of the principal components to both DM yield and N content. Edge positions changed strongly with IC. Possible effects of sensor characteristics, canopy geometry, leaf angle and changes in leaf characteristics with canopy position on the observed relation between IC and edge position are discussed.

Additional keywords:

• imaging spectrometry
• hyperspectral
• stress
• grassland
• leaf reflectance