SPECTRAL MIXTURE MODELING TO ESTIMATE WOOD VOLUME IN NORTH OF SPAIN FROM OPTICAL SATELLITE DATA

Resumo

Information from satellite imagery is an important data source to Forest management. Remote sensing techniques provide information about volume, biomass and other biophysical parameters of forest stands. The main goal of this work is to map the conifer stand volume from Landsat Thematic Mapper (TM) data using Spectral Mixture Analysis (SMA). Fraction images from spectral unmixing show biophysics properties more easily than original bands because they represent physics aspects of ground covers. The work área was El Alto Valle del Ebro (Spain). Two mini-scenes, (50x50km) acquired on March 12th 1996 and July 13th 1996, were used. The applied methodology had three main steps: 1) unmixing the original imagery, the bands 3, 4, 5 and 7 (the most employed in vegetation studies) were used in SMA; 2) these fraction images were related with the over bark volume (OBV) variable (extracted from Second Spanish National Forest Inventory NFI2 ) by means of regression techniques and multivariate analysis; 3) a volume map was obtained using interpolation techniques from the obtained allometric model and the basal area (BA), of considered stand. The main conclusion of this work was the possibility to obtain a model (adjusted R²=0.75) which permitted to estimate the stand volume from medium spatial resolution satellite data. Moreover, the use of this model and with the support of digital ortophotographs allowed an estimated volume map to be obtained.