Abstract

Remotely sensed radiance recorded in red and near, middle and thermal infrared wavebands is usually correlated with the green leaf area index (GLAI) of a vegetation canopy. Empirical models based upon airborne multispectral scanner data recorded in these wavebands were inverted and the GLAI of grassland was estimated, but to an accuracy of only 18-51% (95% confidence level) for a five-class classification and +- 0.37 +- 0.75 GLAI at a point. Refinement of the methodology by suppression of environmental influences on the remotely sensed data and by allowing for both the asymptote in the empirical model and error in the ground data increased significantly the accuracy of GLAI estimation to 60-88% (95% confidence level) for a five class-classification and +-0.06 +- 0.10 GLAI at a point. It was concluded that future studies of this kind should use red and near infrared radiance; the refined methodology and an improved sampling scheme for the collection of ground data.