Monitoring Large Conservation Areas with Imaging Spectroscopy

Monitoring of large conservation areas has to be accomplished to
fulfil the reporting commitment of the European FFH Directive. Aim
of this project was to develop a new monitoring approach for
area-wide mapping on a stand level. This approach was based on the
combination of numerical methods in vegetation ecology with imaging
spectroscopy. The study took place in the FFH conservation area
Murnauer Moos, Upper Bavaria. The imagery had been gathered using
the imaging spectrometer HyMap. In order to develop maps that
include spatial information on vegetation types as well as on
transitions, crisp field and image classifications were combined
with fuzzy methods in field and image data analysis. With
Non-metric Multidimensional Scaling (NMS) ordination technique for
the pre-processing of vegetation data and Partial Least Squares
(PLS) regression for extrapolation, we took account of occurring
mixed stands and gradual vegetation transitions. In contrast, crisp
supervised image classifications are suited to assign clear
categories, which are also needed in management practice. Certain
emphasis was given to the different possibilities of ground data
classification and endmember selection. Different applications of
endmember determination to Spectral Angle Mapper (SAM)
classification and Multiple Endmember Spectral Mixture Analysis
(MESMA) were compared. Synthesis maps for monitoring were produced
that deliver two-fold information on pixel basis: vegetation type
membership on the one side, stand position in the context of the
continuous field of the vegetation on the other. Hence, ecotones
can be monitored within habitats. This study shows that with the
use of high spatial and spectral resolution of the imagery, this
information is given in the same spatial detail for a large area,
and the quality of the given details is measurable.