The following methods were all
carried out in ArcMap in
ArcGIS Desktop 10.
Satellite
Classification
The satellite bands were first
rearranged as: red
–band 4, green- band 3, blue – band 2
A preliminary unsupervised
classification, done with the
‘ISO unsupervised classification’ tool, was ran with 40 classes, to
represent
the maximum number of possible classes given that there are 4 input
raster
bands (#bands x 10).Two of the eight tiles did not yield the assumed
max number
of classes, and only gave 7 and 8. From there, the 40 classes were
manually
selected and filled to determine if the relevant habitat features were
represented.
Following this, as second
unsupervised classification was carried
out with the number of known relevant classes. There were 10, and
included:
shrubs, water, herbaceous growth, trees, grass, earth, roads, other
land use
areas (such as bomas), the map edge, and shadow. For tiles that
contained visible
cloud cover, a second two classes were added to account for the cloud
and its
shadow.
The remaining 6 tiles (as two only
yielded 7and 8 classes
initially) were then classified with 10 or 12 categories (depending on
presence
of clouds). The result was examined for errors, and it was found that
certain
categories; such as trees, and grass were grouped together as one
class.
However, when the 40 class result was examined, the two categories
still were
not separate. Therefore, it was concluded given the processing
capability of
ArcMap, they would be left combined.
The final result was a classified
raster, with the following
categories of interest; shrubs, herbaceous growth, trees/grass and
earth and
roads. These were filled with different colors, and the remaining
categories
were combined via reclassification, and left hollow.
Animal
Location
The location of 7 dik-diks (labeled 1
through 8, as one died
before the study was complete) were originally given by latitude and
longitude
co-ordinates. The location for each animal was recorded approximately
2000
times over 5 days. These data were added as x-y coordinates, separately
for
each individual. The coordinates were then converted to a point feature
layer,
and were finally spatially referenced by their UTM zone, based on the
satellite
projection. The final result was a shape file for each individual, with
a
projected coordinate system of: WGS 84 UTM zone 37 N.
Defining a
home range
A ‘Kernel Density’ tool was used to
approximate the home
range of each individual. Each yielded 9 animal density ranges. For
each
individual, the lowest density range was omitted, and assumed to
reflect error
in the telemetry data; as it incorporated the outlying few points not
in
proximity to the majority of points.
As a final step the Kernel Density
estimations were converted
to a polygon. This was done by first using the ‘Extract by Attributes”
tool to
isolate the majority density areas, and was then reclassified. Lastly,
the
‘Raster to Polygon’ tool was used.
Resources
used by
each animal
The resource use was examined by
taking into account various
attributes. First, the data was corrected for poor GPS precision. This
is referred
to as the dilution of precision, DOP. This was done by eliminating all
values
of horizontal accuracy over 5.
Next, both the original data, and a
set corrected for high
values of DOP were separated by the diurnal value (day and night) for
each
individual. An
‘Extract by Attributes’
tool was then used to determine the amount of each habitat category,
for each
diurnal set, for each individual. That is, for every location (animal
point)
the raster cell value (classified habitat) was determined and summed.
In order to compare the results of
this analysis, a set of
random points, produced through the ‘Create Random Points’ tool was
done for
each data set. The amount of each habitat category was also determined
by the
‘Extract by Attributes’ tool.
The home range for two individuals
spanned over two
satellite tiles and therefore the process mentioned above were carried
out
twice for these animals, each time with the second classified raster.
The
extract tool gave a null raster value for the points which were not
included in
the specified raster, so the valid values were separated from each
input raster
and then combined.
The result was a resource use data
set for each individual,
for both day and night, and for both precise DOP values, and the
original data;
with a random set of points for each. Also, the total amount of each
habitat
category was calculated in each home range polygon through the ‘Extract
by
Polygon’ tool. This yielded the amount of each raster category over the
given
home range.
