Case-based Reasoning to Predict Thinnings in Central European Forests
The motivation of a thinning is to influence the future development
and state of a forest. A thinning is basically the selective removal
of trees in a stand in order to achieve a certain future development
or state of a forest satisfying a number of economical and ecological
aims. Thinnings are regarded as the most important influences on the
development of forests in Central Europe. Therefore, foresters as well
as ecologists are interested in detailed information about the effects
of thinnings. Even though the information these two groups are
interested in is quite different, at present both have to deal with
simple models to predict the actual result of thinnings.
The domain covered in this project is complex and simple at the same
time. It is simple with respect to the limited stand and tree
information actually used to carry out a thinning, while its
complexity arises from the infinite number of possible combinations of
those pieces of information in actual forests.
The main problem in developing a KBS system to predict the outcome of
thinnings is the fact that thinnings are usually specified in terms of
abstract descriptions of goals to be achieved, which are not
operational guidelines, but rather a frame in which forest experts set
up their own individual thinning concepts. However, the forest expert,
usually working alone, is never required to express his individual
thinning heuristics explicitly. Those individual thinning concepts
are, however, worth collecting as they suggest
alternative ways of managing a forest. Our system is able to capture
different thinning concepts - providing
management and planning support based on individual thinning heuristics.
The system maintains libraries of
thinning situations with a known outcome. Each of these libraries
represents a particular thinning strategy. Faced with a new situation the
system is able to find a similar example in a library and use it to adapt
generic thinning rules to the new situation.
