Imaging spectroscopy of foliar biochemistry in forestry environments
AbstractRemote sensing estimates of leaf biochemicals provide valuable information on ecosystem functioning, vitality and state at local to global spatial scales. This paper aims to give an overview of the state of the art of foliar biochemistry assessment in general and, where possible, attention is given to: (1) Eucalyptus forest environments, (2) use of hyperspectral remote sensing or imaging spectroscopy, and (3) the challenges towards operational application of such assessments. Estimation of foliar biochemicals has improved significantly from early broad-band sensor attempts, given the advent of hand-held, airborne and space-borne spectrometers. These instruments provide sensing in contiguous, narrow spectral bands in the visible to shortwave infrared, as compared to the small number of broad spectral bands provided by multispectral sensors. Chlorophyll, nitrogen, cellulose and lignin represent a sample of biochemicals that have been assessed successfully, particularly at leaf level and with varying success at the canopy scale. A major challenge is scaling of predictions of biochemicals from ground to airborne and ultimately space-borne levels. This entails development of algorithms that minimise the contributions of canopy structure, atmospheric conditions, sensor/illumination geometry and leaf water content variations. Some advances have been made in this direction including the derivation of new vegetation indices and the use of spectral transformations such as derivative analysis and continuum removal. Other studies have focused on developing physically based models, e.g. radiative transfer models (RTMs), which appear to be more robust when compared to statistical models. However, the application of RTMs needs to progress beyond the estimation of only chlorophyll and biochemicals in monoculture environments to other nutrients and adapted for more complex canopies. Furthermore, inversion techniques of these models need to be improved.
Keywords: artificial neural networks; foliar chemistry; hyperspectral; partial least squares; remote sensing; stepwise multiple linear regression
Southern Forests 2008, 70(3): 275–285