Spectral Mixture Analysis (SMA) of reflectance curves from Legume-Grass-Plots for the analysis of species composition in forage production.

This paper focuses on the detection of individual fractions of legume and grass species based on Spectral Mixture Analysis of remotely sensed. non-imaging data. In 005 one glasshouse-experiment and one field-experiment were built up and continuous measurements and samples were taken for subsequent comparison. Marked crop-plots were spectrally recorded with the fieldspectrometer ASD Fieldspec (c) Pro JR and subsequently harvested for laboratory analysis. In the first step the reflection signatures were smoothed by the adaptive Savitzky-Golay-Filter and evaluated for inaccurate data by means of descriptive statistics. After reduction to relevant wavelength-ranges the spectra were unmixed to their pure components (endmembers). One of the basic steps here is the selection of accurate endmembers. The endmembers have to represent one fraction in an appropriate manner. The results show the possibilities and limits of the method with fieldspectrometrical data with focus on their usage for site specific forage production. Good relations could be found for the data of the glasshouse-experiment, whereas the data from the field-experiment show deficiencies, mainly due to high weed rates and a more complex stand structure.