Jp. Gastelluetchegorry et al., AN ASSESSMENT OF CANOPY CHEMISTRY WITH AVIRIS - A CASE-STUDY IN THE LANDES FOREST, SOUTH-WEST FRANCE, International journal of remote sensing, 16(3), 1995, pp. 487-501
The capability of airborne (AVIRIS) and laboratory spectrometry was in
vestigated for assessing the chemical composition of foliar elements o
f a pine forest (The Landes, SW France). Simultaneously with AVIRIS ac
quisition, an atmospheric profile was carried out, and the forest vege
tation was sampled for chemical analyses and laboratory spectral measu
rements. Predictive relations between concentrations of nitrogen (r =
97 per cent), lignin (r = 89 per cent) and cellulose (r = 83 per cent)
and reflectances of pre-treated pine needles were determined through
stepwise regression analyses. A methodology was designed to assess the
ir extrapolation to remotely acquired spectrometric data: (1) geometri
c and atmospheric corrections, (2) registration within a biophysical d
ata base (e.g. LAI, biomass), and (3) comparative statistical analysis
of laboratory and airborne spectrometric information. The application
of laboratory derived relationships led to relatively large correlati
ons for nitrogen (74 per cent) and cellulose (79 per cent); poorer res
ults were obtained for lignin (55 per cent). The use of atmosphericall
y corrected reflectances led to slightly worse correlations: nitrogen
(73 per cent), cellulose (78 per cent) and lignin (44 per cent). It wa
s attempted to improve these results while taking into account the inf
luence of the canopy structure and total quantity of chemical compound
s. (1) Slightly poorer results were obtained when chemical concentrati
ons were weighted with local biomass and LAI values. (2) Predictive eq
uations based on laboratory measurements were applied to reflectances
of pine needles that were computed through the inversion of two reflec
tance models. This last approach improved correlations for lignin (74
per cent). No improvement was observed for nitrogen (70 per cent) and
cellulose (69 per cent). Finally, in order to provide suitable informa
tion to GIS based ecosystem models chemical concentrations were tentat
ively mapped.