2.3 Forest evaluation and land cover classification
For the forest evaluation, we used a simple random sampling design using
the same four 1 ha random plots per condition as sampling units,
measuring every tree above 1.3 m height and breast height diameter (BHD)
in the plot (DBH >1cm). Additionally, we used partial data
from a paired A.araucana regeneration study project with the
Chilean national forest corporation described in (Montecino et al., In
Prep), which consisted of four more 1 ha random plots, which allowed us
to reduce error and to represent forest stand conditions better
Forest structure was evaluated by analyzing the observed stand diameter
distribution and fitting the most common probability density functions
such as lognormal, negative exponential, Weibull, and gamma functions
(Newton, 2007). Following goodness-of-fit indices such as the Akaike
Information Criterion (AIC) and Kolmogorov- Smirnov (K-S) test to
determine the best-fit diameter distribution models.
A high-resolution multispectral satellite image was used to generate an
unsupervised land cover classification of the area. The image was
obtained from the World View II satellite with a 0.46m resolution from
February 2015. These images supply great detail and geospatial accuracy,
allowing to conduct an unsupervised classification in ENVI 4.7 software,
using the ISODATA clustering algorithm.