the east African Dik-dik

GPS Uncertainty

 

The animal location data collected in this study did not exclude uncertainty in the GPS location. Research has demonstrated that although home-range estimates appear generally robust to standard levels of location imprecision and bias, habitat selection estimates are less accurate.  Moreover, the determination of whether an animal selects or avoids a location, and the magnitude of either option, are sensitive to the precision of GPS data.

 

As displayed in the results, the animal location error is visually evident. Dik-diks are not found in large bodies of water, or flowing rivers like the one in individual 8’s range. It should also be noted that the location error is approximately the same length as some of the habitat features (shrubs) which are assumed the dik-dik selects for. Previous studies have determined that telemetry error is influenced by raster patch properties. Moreover, that ignoring this uncertainty may still lead to inaccurate estimates of selection for small resource path size.

 

Classification Error

 

Classification error occurs when a map class is incorrectly assigned to a specific feature. During unsupervised classification, multi band spectral response patterns are grouped into clusters that are statistically distinguishable. However, given the processing limitations of ArcMap, different features which showed the same spectral response, were grouped together, and could not be separated. However, given that the classification in this study is only relevant to the dik-dik location, the error should not have much effect. The locations did not overlap with any specific land use areas, and for the purposes of this study trees and grass can both be considered vegetation. However, one should proceeded with caution if using the classified tiles in different locations then where the dik-diks were present.

 

Secondly, the initial Quickbird satellite image tiles all showed different spectral responses, some more drastic in their differentiation. Because of this, two tiles gave a maximum of 7 to 8 categories could be classified, and did not match well with those  derived from tiles with 12 categories. In this study however, the two tiles the dik-diks were present on, where similar enough to provide 12 categories that transitioned in the tile overlap.

 

Resource Selection

 

In this study, resources were assumed to be ‘in use’ or selected for when the location of a given individual yielded one of the classified habitat categories of the raster they were located on. These were considered observed selections For the purpose of comparison, expected selections were required, and acquired by the generation of random points over the same home range.

 

The approximation of home range is critical as it provides some understanding of what resources are available to the individual and the size and range of resources it willing to defend. The kernel density proved to be useful in this home range approximation as only the area where the points occurred most often , was considered. Moreover, as the majority of over 1500 location points occurred in a specific small area, the few points relatively far removed, with no points leading to it, can safely be assumed incorrect.

 

Previous studies regarding the behavior and life history of dik-diks indicate that their habitat plays an important role in their survival. Given their small size, they are eligible prey to a host of species and require adequate cover.  This cover is commonly in the form of long grass, shrubs and boulders. Not only do these features provide cover from predator, but also provide a shaded area to prevent overheating. Finally, the dik-dik is a herbivore, and is known to consume a variety of vegetation as part of their diet. Therefore, the selection of grass and shrubs is consistent with their known behavior. With respect to roads, and bare earth,  because the dik-dik spends the majority of it’s time feeding  in the early morning and later afternoon; that resource may offer a quick movement pathway between different food sources.

 

Not surprisingly, the diurnal resource selection was determined to be independent of the expected use, assuming random location, and no selection. A result of independence, from a Chi-Squared analysis, occurred for all but two results over 7 individual dik-diks. A larger sample size, and further statistical analysis is recommended to determine which resources are more heavily selected for, and the specific differences between night and day. Also, further analysis is required to determine how the strongly specific resources are selected for, given what is available in their home range.