Estimating Spatial Patterns of Biomass and Nitrogen Status in Grasslands through Imaging Spectrometry

摘要

In this paper, the potential of imaging spectrometry in mapping spatial patterns of canopy biomass and nitrogen status within grasslands is studied. A homogeneously managed grassland field was used as a test site. 20 plots were harvested and fresh and dry biomass as well as leaf nitrogen concentration was determined. Spectral measurements were performed on these plots using an ASD FieldSpec Pro spectroradiometer, covering the wavelength range from 350 - 2500 nm. A partial least squares analysis was applied to obtain insight in the prediction potential of spectrometry for biophysical variables using all wavelengths. In addition, we applied and compared several feature reduction techniques to reduce the total number of wavelengths used in the model. Methods included stepwise forward selection, SVM band-shaving and the use of vegetation indices. Results showed that a few bands already have a considerable predictive power. Results were also applied to whole fields using UltraCam images.