Airborne MSS data to estimate GLAI

Abstract

Airborne multispectral scanner ( MSS) data, collected in June 1984, were used to estimate the green leaf area index ( GLAI) for 60 km2 of grassland. The methodology involved (i) radiometric and atmospheric correction, (ii) production of a vegetation index image, (iii) calculation of a calibration relationship between a vegetation index and GLAI, (iv) production of an image of estimated GLAI by inversion of the calibration relationship in (iii), and (v) accuracy assessment. The initial accuracy of GLAI estimation was ± 0-75 GLAI for an area and 17-40 per cent at the 95 per cent confidence level, for a six-class classification. Refinements to the methodology were evaluated by their effect upon the accuracy of GLAI estimation. In order of increasing importance these refinements were: suppression of environmental effects on the remotely sensed data, processing on a per-field rather than a per-pixel basis, calculation of the calibration relationship between a vegetation index and GLAI, using ground-based radiometric data and a modified least-squares regression up to the asymptote of the vegetation index and allowance for error in the ground data. By utilizing all of these refinements the accuracy of GLAI estimation increased to + 009 GLAI for an area and 60-82 per cent at the 95 per cent confidence level, for a five-class classification. This methodology was then applied to MSS data recorded at the same place, but at a different time, with a different sensor look direction,and under different atmospheric conditions. Despite these differences, the accuracy of GLAI estimation was not significantly different, at the 95 per cent confidence level, from that obtained with the previous data set. The accuracy was + 0-10 GLAI for an area and 67-97 per cent at 95 per cent confidence level, for a five-class classification.