CASE STUDY 6 : Modelling movements and habitat preference of Grevy's Zebra


Tracking data of Grevy’s zebra was analyzed in combination with NDVI data and a habitat classification to model movements and habitat preference of Grevy’s Zebra has been made. This research was conducted by Eline Hostens (UGent student) for her master’s thesis. The thesis can be downloaded using following link:

The tracking of Grevy’s zebras is part of the overall GSM GPS tracking project supported by Safaricom and executed by Save the Elephants. The tracking of Grevy’s Zebras is jointly executed by Save the Elephants and Lewa Wildlife Conservancy with further collaboration with the Northern Rangelands Trust (NRT) and other stakeholders.

As the Grevy’s zebra is a threatened species, it is important to know as much as possible about their habitat use and migration pattern. This thesis had two main objectives: the creation of a habitat classification and the analysis of the Grevy’s zebras migration. The habitat classification was based on Landsat and MODIS images. Both Maximum Likelihood and Neural Networks were used to conduct the classification. To analyse the migration, data obtained from the GPS-tracking of sixteen Grevy’s zebras was used. Several factors with a possible influence on the migration were examined: distribution of biomass, water, livestock and towns. The final step was to make an integration of all these factors to predict the areas within the study area that are most suitable for Grevy’s zebras.

The first objective of this thesis was to make a habitat classification of the study area. The use of Landsat satellite images was abandoned as no good result was obtained using these images. Instead time series of MODIS images were used which enhanced the distinction between different classes providing information on the plant phenology. The Maximum Likelihood classification method only made a good separation of the forest class from the other habitat classes. Using the Neural Networks classification technique, a better distinction between the different savanna sub-classes was obtained.

The best classification result was obtained with NN using all MODIS spectral images, all NDVI images and all EVI images as input.


Land cover classification

The second objective was to model the migration of the Grevy’s zebras. The most important factor influencing the migration of the Grevy’s zebras was the available biomass as food source. NDVI was used as a proxy for available biomass. The Grevy’s zebras almost always used areas with significantly higher NDVI values than in the surroundings. Only during the first rainy season they preferred areas with significantly lower NDVI values and in the second rainy season there was no significant difference between the NDVI values in pixels where zebras were absent or present. The fact that in the first rainy season areas with lower NDVI values were chosen can be explained by the very wet rainy season.


Comparison of average NDVI in areas where Grevy’s zebra were observed and areas where Grevy’s zebras were not observed

The other factors influencing Grevy’s zebra migration are proximity to water and livestock density. The zebras mostly prefer areas between 0–15km of water. They are most present within the range of 2.5–4.5km from the nearest water point. Areas very close to water are less preferred as there is more competition in these areas with other wildlife and livestock. In this study, all zebras were always in relatively close proximity to water, as they can go without water for 2–5 days and can travel between 10 and 15km per day.


Relationship between number of zebra observations and distance to the nearest water

When comparing the tracking data and livestock density it was found that Grevy’s zebras avoid areas with high livestock density. This can be explained by the direct competition between zebras and livestock for water and food. The relationship of the Grevy’s zebras and the distance to the nearest town resembles the relationship between the zebras and the distance to the nearest water point. Their migration and occurrence is probably not very affected by the towns in the study area.


Relationship between livestock density and number of zebra observations

Based on the MODIS and Africover classification, a habitat preference ranking for the Grevy’s zebras was performed. First it was tested whether there was a random use of habitat or not. In the case of a random use, the zebras use the available habitat in proportion to the area of each habitat type. In case of a non-random use of habitats, a ranking was made per zebra of which habitat they preferred. Finally, the result of all sixteen zebras was integrated to obtain an overall habitat preference ranking for all Grevy’s zebras tracked in the study area. From the preference ranking based on the MODIS classification, it could only be concluded that Grevy’s zebras avoid forest habitat. Between the other habitat types no significant distinction in preference could be made. A possible explanation is that the classification does not correspond with reality very well.

From the preference ranking based on the Africover classification could be concluded that in the first level comparison, between the composition of the study area and that of the home ranges of each animal, there is a significant preference of the habitats settlements and shrubs & herbaceous, followed by a preference for open-sparse shrubs. All other habitat types could not be ranked in a significant order. For the second level comparison, this is between the home range compositions and the GPS data, there is a significant preference of the habitat types herbaceous & shrubs, and herbaceous. The next habitat types in the preference ranking are open woody and open-sparse shrubs. The other habitat types could be left out as most of the MCPs were composed of these four habitat types.

Finally an integration of all the factors influencing the migration was made based on the obtained results. The areas not suitable for Grevy’s zebras were determined. For the other areas the influence of the distance to the nearest water point and of the NDVI was taken into account to divide these areas into different preference classes. The result showed a 2.4 times more usage of the most suitable areas by the Grevy’s zebras than would be expected from the area of this class. However, there are a lot more factors influencing the occurrence and migration of the Grevy’s zebras. For instance, there is an influence of predators, other ungulates and reproductive state of the Grevy’s zebra females. Data about all these influences and maybe even more should be collected and taken into account to get a better idea of the areas preferred and used by Grevy’s zebras.


Modeled habitat preference of Grevy’s Zebra within the study area (0=avoided,5=most preferred

 <Back to top> <Back to case studies>

Nov/Dec 2010
Final workshop in Nairobi
October 2010
Download Newsletter 6
May 2010
DEMO Endeleo in Nairobi
April 2010
Smoothed graphs available
March 2010
° Download Newsletter 5
° Case studies elaborated by users
February 2010
NEW in image viewer: toggle on/off borders
January 2010
Help desk available

Related Links
In order to further improve this website, we would like to know your opinion. Please fill in the feedback form.