the east African Dik-dik

The initial satellite image was composed of 8 tiles. Of those, six yielded an unsupervised classification with 10 or 12 classes. The main classes (or habitat features) of interest were:  shrubs, trees, grass, herbaceous growth, bare earth and roads. These features are resources which dik-diks are assumed to select for. Although not each resource class was continuous across every tile, the dik-diks in this study were only present over two tiles.  These two tiles displayed similar spectral reflectance, which resulted in continuous resource categories. There were a total of 7 dik-dik individuals, whose locations were analyzed given the overlap with the classified satellite image to determine their resource selection.

See Satellite Imagery

A kernel density estimation was used to determine each individuals’ home range.  The smallest density range (1 of 9) was not included, as it most likely reflected error in the GPS location. A home range for 6 of the 7 dik-diks was calculated. One dik-dik was located beyond the satellite image, and was not included.

See Animal Location & Home Range

During the classification of the dik-dik habitat, and when the animal locations were added; uncertainty in the data was visible. The classification process was able to distinguish the main habitat features, though it was erroneous in that some features were grouped together which were not intended to be. Two main habitat features of interest, grass and trees, were grouped together as one category. This was due to their similar spectral reflectance, and could not be fixed in ArcMap. A second error was in relation to certain habitat features that were not of interest in this study.  Features such as boulders, boams, rivers and other land use areas sometimes was also grouped together with features such as herbaceous growth and shrubs.

See Data Uncertainty

Effect of Dilution of Precision (DOP)

The GSP co-ordinate of each location, for every individual was associated with a horizontal accuracy value, DOP. These ranged from 0 to 91. To determine if the location precision had an effect on the resource selection, every location with a value over 5, for individual 1, was excluded.  The resource use, when corrected for DOP did not differ to the resource use of the entire data.

See Resource Selection based on DOP

Diurnal Activity

The night and day resource selection of each dik-dik was compared to the ‘expected’ resource selection. These expected values were assumed to be equivalent to a set of random points. The numbers of points compared were the same for each night and day values.

A chi-squared test was conducted to determine if the observed points were independent from the expected (random) points. For each individual, the resultant P value was low, and for all except two comparisons they were under 0.05. For these comparisons, it was concluded that there is a selection process in the resource use, as there is an independence from the expected values.

P Values:

Although it is hard to determine which resources are selected for most heavily, it appears that areas of minimal grass, and grass and trees were the most used. Bare earth and shrubs were selected for next. However, the resource selection differs substantially between each individual.

See Diurnal Selection

Resource Selection and Resource Availability

For each individual, the proportion each resource was used, was compared to the availability of the resource, as determined by the proportion each resource existed in it’s home range. The results show some differences in use versus availability.

See Resource Availability 

 Therefore each individual was examined separately. Under closer examination, it appears that the dik-diks tend to re-select for some resources within their home range. There was not however, one resource that was selected for more often across all individuals.

See Individual Resource Selection