Characterization of Venezuelan Equine Encephalitis Emergence Sites Using GIS:
First Year Report
Introduction |
UTMB Main Page |
CHAART Projects |
CHAART Home
Use of Geographic Information Systems to Locate Enzootic Foci of Venezuelan
Equine Encephalitis Virus
Virus studies conducted between 1973-1981 by Walder and co-workers
demonstrated that the ID strain VEE virus persisted naturally in enzootic
foci located in western Venezuela. Because recent genetic studies have
shown the ID enzootic strain to be a likely progenitor of the epizootic
strains, a better understanding of the ecology of VEE virus is needed to
identify environmental factors responsible for virus maintenance in specific
foci. Our present effort concentrated on sites along the Catatumbo River
where virus was isolated in the earlier studies. We conducted a
retrospective analysis of the landscape using geographic information system
and remote sensing techniques. To define the region of interest, virus
isolation results from five sites within 12 km of the Catatumbo River were
used to develop a spatial model to predict the geographic space most likely
to have been occupied by the virus, as well as other components of the
maintenance cycle. In addition, the frequency of virus isolations was used
to create a spatial trend model to predict where virus activity was likely
to have been greatest in the general area of the study sites. Spatial
filtering methods were applied to create a polygon object to represent the
landscape that contained the primary focus of virus activity. The predicted
polygon object was used to define the
region of interest
in a Landsat Thematic Mapper (TM) scene
of the area acquired 31 December 1986. After extraction of the study area,
exploratory data analysis techniques were used to characterize the
reflectance values of habitat features in TM bands 1-5, and 7. In
addition, leaf area indices, vegetation indices, and tasseled cap
transformation procedures were used to interpret the data in preparation for
modeling by image classification methods. Several unsupervised
classification algorithms were compared with respect to their suitability
in characterizing the study area. Clustering the reflectance values around
15 class centers seemed to provide the best interpretation of the area.
Also, Adaptive Resonance, Fuzzy c Means, and ISODATA classifiers were best
in suggesting classes of what appear to be wetland features, possibly
associated with one or more mosquito habitats. Plans are now being made
to identify the landscape features associated with each reflectance class
and to validate our model.
Back to
VEE Project Introduction
Last updated: Mar 2000