Measuring crop status using multivariate analysis of hyperspectral field reflectance with application to disease severity and plant density

Abstract

Using spectral reflectance to estimate crop status is a method suitable for developing sensors for site-specific agricultural applications. When developing spectral analysis methods, it is important to know the influence of different crop parameters on the spectral reflectance profile. The objective of this report was to present and evaluate a multivariate method for objective hyperspectral analysis in the examination of how different parts of the reflectance spectrum are affected by disease severity and above ground plant density. Data from two field experiments were used; fungal disease severity assessments in wheat 1998 and above ground plant density measurements 2003. The analysis method consisted of two steps: a pre-processing step where the data was normalized and a classification step for estimating the crop variable. Using only 12% of the data as training data, the method resulted in coefficients of determination (R ²) of 94.3% for the disease severity data and 96.9% for the plant density data. The hyperspectral analysis method presented could also be used to extract spectral signatures of disease severity and plant density using the experimental data. In general, two types of spectral signatures for both data sets, with respect to increasing disease severity and decreasing plant density, were observed (1) a flattening of the green reflectance peak together with a general decrease in reflectance in the near infrared region and, (2) a decrease of the shoulder of the near infrared reflectance plateau together with a general increase in the visible region between 550 and 750 nm.