|
|
Landscape-level plans are required in today's forest
planning environment. Geographic information systems
linked with relational data bases that store inventory
information about landscape elements provide the data
tools needed to create the plan. Despite the availability
of sophisticated display and database tools, a comprehensive
landscape-wide plan is still difficult to create, because
the inventory is only partially complete. Most Similar
Neighbor inference (MSN) provides a method for objectively
linking actual ground inventory data to unsampled landscape
units. For a unit in a landscape that doesn't have a
detailed inventory available, MSN finds the "most
similar" sampled unit and creates a link to its inventory
attributes. In the seminar, I will describe the
statistical theory and discuss recent applications. In a
current study, MSN analysis is applied to tree-level
information collected on forest survey plots that are
spatially referenced to processed satellite imagery, to
estimate the distribution and relative importance of 5
major tree species over a 180 x 120 km area in the Finger
Lakes region of New York. The degree to which the
interpolation maintains both spatial and attribute
structure present in the sample data is discussed. The
ability to retain both types of structure should result
in more realistic interpolations, and lead to an improved
tool for resource managers concerned with interaction
among multiple ecosystem components across a landscape.
|
|