Department of Software Technology
Vienna University of Technology
Using neural networks to predict individual tree mortality
Within forest growth modeling it is customary to employ LOGIT models
to predict individual tree mortality.
In this paper we use Learning Vector Quantization and the self-organizing map
as different formalisms to predict individual tree mortality.
The data set for this study came from permanent sample plots in
uneven-aged Norway spruce (Picea abies L. Karst) stands in Austria.
After parameterizing the LOGIT model and training the two different network types
we evaluate the differences in the resulting mortality predictions using an
independent test data set.
The results indicate that the LVQ performs slightly better than the conventional
LOGIT approach as well as the self organizing map.
Up
Comments: rauber@ifs.tuwien.ac.at