Spectral mixture analysis of potato crops under different irrigation regimes

In an agricultural remote sensing image, the digital reflectance value for each pixel is a result of the combined spectral contributions of the various scene components, namely the plant, soil and shadow. Traditional remote sensing image processing methods such as vegetation indices do not separate these components explicitly, yet it is only the plants for which information is sought. The technique of spectral mixture analysis (SMA) is designed to derive the fraction of each component that is contributing to a pixel´s reflectance. In this paper, SMA and vegetation indices are compared in a remote sensing experiment to monitor moisture stress in potatoes at a test site near Lethbridge, Alberta Canada. Differential irrigation treatments were implemented at the test site to induce various levels of moisture stress on the potato crop. In 1998, ground-based and airborne remote sensing data were collected in June, July and August. This paper addresses the ground-based August dataset using SMA to quantify the abundance of plant, soil, and shadow at sub-pixel scales towards improved extraction of plant biophysical and structural information. The impact of moisture stress on the crop in August was significant. The strength of the relationship to biophysical parameters was similar for both the SMA fractions and the vegetation indices and was somewhat lower than anticipated. A number of factors are discussed that may have affected the predictive capability of both remote sensing image processing methods.