Multivariate analysis of low cost airborne CIR imagery for the determination of forest canopy structure

摘要

Multivariate statistical analysis of airborne CIR imagery with a ground pixel size of 25 cm were used to develop a series of empirical models to estimate forest canopy structure in deciduous and mixed-wood stands. Using the linear combination of image spectral, textural and structural information, models were developed for predicting effective LAI, LAI, basal area, crown closure, and within-crown structural damage. Results from the best fit models suggest image spectral information was most important in modeling structure in the deciduous stand, while image structural and textural information were important for modeling the mixed-wood stand and within-crown structural damage in the deciduous stand. It was determined that the sensitivity of (1) image structural and textural measures to macroscale and microscale geometry of individual trees and (2) image spectral information to the amount of vegetation cover resulted in the ability to predict biophysical variables. Finally, this study demonstrated the effective use of low cost airborne CIR imagery for the retrieval of forest canopy structure.