Extraction of forest stand parameters from panchromatic and multispectral SPOT-1 data

Abstract

The main objective of this paper is to determine the extent to which forest structural variables can be estimated from SPOT-1 panchromatic and multispectral radiance. The test site is the state-owned Pynven forest, featuring a large amount of evenly-aged pine stands from which those with Corsican pine (Pinus nigra Am. var. Calabrica Loud.) are selected. Investigated image parameters include waveband, first-order statistics estimators and sample window size. The forest stand parameters are stand density, stand age, average tree diameter, stand basal area, average canopy height and stand volume. Empirical linear models are constructed based on Bartlett's three-group method, followed by inversion and validation using an independent test data set. The green and red SPOT-1 waveband are useless for forest stand parameter estimation, and the arithmetic mean of sample window radiance is the only usable estimator from the set of tested statistical measures. Stand density and average canopy height can satisfactorily be estimated from the SPOT-1 data, with the panchromatic band yielding consistently better results than the infrared waveband. Radiance sampling in larger windows is usually more likely to capture structural variation of the forest on stand basis and results in more accurate estimation of the ground data.