Off-campus UMass Amherst users: To download campus access dissertations, please use the following link to log into our proxy server with your UMass Amherst user name and password.

Non-UMass Amherst users: Please talk to your librarian about requesting this dissertation through interlibrary loan.

Dissertations that have an embargo placed on them will not be available to anyone until the embargo expires.

Document Type

Open Access Dissertation

Degree Name

Doctor of Philosophy (PhD)

Degree Program

Environmental Conservation

Year Degree Awarded

Summer 2014

First Advisor

Charles M Schweik

Second Advisor

Todd K Fuller

Third Advisor

Laurie R Godfrey

Subject Categories

Life Sciences


Context: The impacts of ecological, anthropogenic and future climate change on the distribution of wild mountain gorillas (Gorilla beringei beringei) are of ongoing concern. Knowing the factors that determine gorilla habitat suitability now and in future is essential for conservation planning. The mountain gorilla is recognized by IUCN Red Data Book as critically endangered and a great tourist attraction. However, the factors that impact on their spatial use of Bwindi are poorly understood.

Aims: I aimed at determining the major factors that determine gorilla distribution, predict the wild gorilla habitat suitability and establish the vulnerability index of the gorillas to future (2050) climate change throughout Bwindi using different tools.

Methods: I used seven independent environmental variables that are thought to affect gorilla distribution in Bwindi. I made a vegetation map from plant inventory data collected from stratified random transects, high resolution aerial photos and a 30m Aster Global Digital Elevation Model (GDEM). Slope steepness and surface curvature were derived the Aster GDEM. Variable ‘distance from roads within the park’, and ‘distance from park boundary’ were generated from the Bwindi GIS database, and levels from human activity were from the gorilla census data of 1997. Wild gorilla groups presence data points were compiled from Ranger Based Monitoring data of Uganda Wildlife Authority (1999 to 2011) and five-year interval gorilla census data of 1997, 2002, 2006 and 2011. Background points to describe a set of conditions available to the wild gorillas in the whole park was generated randomly in R program. The intention of providing background sample is not to pretend that the species is absent at the selected sites, but to provide a sample of conditions available to gorillas in the whole of Bwindi. Then the environments where the gorillas are known to occur were related to the environments across the rest of Bwindi (the ‘background’). Both wild gorilla presence and background data were randomly divided into training and testing data sets. Four algorithms – logistic regression, maximum entropy, random forest and boosted regression trees were used to fit the gorilla presence and background data, produce maps predicting wild gorilla habitat suitability and evaluate the accuracy of the prediction. I used the Nature Serve Climate Change Vulnerability tool (CCVI) to integrate information on gorillas to 18 natural and distribution factors that are associated with sensitivity to climate change and projections of climate changes for Bwindi area based on published literature to determine the vulnerability of gorillas to climate change.

Key results: All the four algorithms showed that vegetation and some form of human activity (roads, park edge, and level of human activity within the park) were the most important environmental factors in determining wild gorilla habitat suitability. All models performed better than random in the accuracy of their predictions (average area under the ROC curve, AUC = 0.7). The difference in their AUC scores was very small (≤0.02) meaning that all the algorithms had more or less the same predictive ability. However, model predictions of gorilla habitat suitability among the four algorithms differed substantially. Logistic regression model predicted that nearly the whole park was suitable for gorillas except part of the northeast. The other three models, however, predicted that most of Bwindi was unsuitable gorilla habitat with the northern sector and the edges of the southern sector being wholly unsuitable. The center of the south sector of the park was predicted to be the core habitat but with varying levels of suitability for each model. Logistic regression model predicted that much of the south sector was highly suitable wild gorilla habitat, while Maxent model showed that only the interior of the south sector was highly suitable. Random forest and boosted regression tree models gave the area in the interior of the south sector low suitability, with a few, small, scattered areas being highly suitable gorilla habitat. Using the climate projections for the A2 emission scenario and average ensemble of 16 global circulation models (GCMs), combined with sensitivity factor inputs, the CCVI tool ranked the gorillas “Not vulnerable/Presumed Stable (PS)” to climate change in Bwindi. This means that the available evidence did not suggest that the abundance and/or range extent within Bwindi assessment area will change (increase/decrease) substantially by year 2050. But the actual boundaries may change. This would make the gorillas adjust to climate-mediated changes in their habitat. This suggests that assisted migration of the gorillas may not be required. Factors that were identified as contributing to vulnerability included physiological thermal niche, physiological hydro niche and disturbance regime while those decreasing vulnerability were dispersal or movement, physical habitat restrictions and genetic variation. Seven factors were either unknown or irrelevant.

Conclusions: Gorillas may have occupied the south-east and south-west parts of Bwindi but were probably exterminated by hunters. They could be prevented or deterred from re-colonizing these areas because of feeding traditions or habitat preferences since the vegetation of Bwindi is spatially structured. The spread of gorillas in Bwindi is relatively recent and from south of the park. It is probable that they have not had enough time to occupy all the forested areas available to them. That could be one reason why there are no reports of gorillas in the north of the park. The northern part of the park also has the highest level of human disturbance and a road that separates it from the southern sector. These could have contributed to gorillas not occupying the northern parts of Bwindi.

Implications: Human disturbance seems to be a major important factor driving wild gorilla distribution in Bwindi and also likely to contribute significantly to the vulnerability of the gorillas to future climate change. Park management needs to increase law enforcement patrols, reexamine the multiple-use program and relocate the road outside the park. This could improve the prospects of long-term survival of the wild gorillas in an island habitat.

Included in

Life Sciences Commons