Impact of projected climate changes on future gorilla distribution

Thorne et al. (2013) used correlative species distribution models (SDM) to predict shifts in climate suitability for the mountain gorillas throughout their range. They estimated the

54

current climate suitable areas for the mountain gorillas by associating geo-referenced occurrence data for the entire range, which consists of Bwindi (Uganda) and the three contiguous parks in the Virunga Volcanoes area: Mgahinga Gorilla National Park (Uganda), Parc National des Volcans (Rwanda), and the Mikeno Sector of Parc National des Virunga (DR Congo; Figure 2.3) with sets of current climatic variables [monthly temperature minimum, maximum and precipitation] from WorldClim (Hijmans, et al., 2005). The derived model was then used to predict geographic areas that will have the same climate conditions as observed mountain gorilla sites for different time periods with future climate change for three SRES (IPCC 2000) greenhouse gas emission scenarios:

A2, A1B and B1 for two time periods: 2041-60 (‘2050´), and 2081-2100 (‘2090´); using five global climate models (GCMs). The SDMs used represent four sets of biological assumptions: 1) the “standard” approach, which used mountain gorilla occurrence sites and climate data with two SDM algorithms (Bioclim and Maxent) and two sets of climatic predictor variables (either 18 or three); 2) the “niche conservatism” models, which are the same as the standard models except that they use the entire eastern gorilla species as unit of analysis, i.e., combining the presence of mountain gorilla with that of the eastern lowland gorilla, G. b. graueri; 3) the “behavioral” model, a climate driven gorilla time-management model; and 4) the “limiting-factor” model, which predicts that mountain gorilla can survive in any environment where plant productivity is at least as high as the areas where they currently occur, i.e., at least as warm and humid as these areas.

The ‘standard’ models showed that the mountain gorilla parks, and by extension the whole of the Albertine Rift ecoregion, will not be climatically suitable for the

55

subspecies by 2050. But Thorne et al. (2013) were quick to add that this result could be because these models assume that a species can only occur in areas with a climate that is similar to that of the sites where it is currently observed. This is unlikely to be the case with mountain gorillas as we know that their range is presently restricted by non-climatic factors, particularly expanding human populations and other human-related dispersal barriers. So the standard SDMs may not reflect the entire ecological niche of a species.

Because the mountain gorilla has been effectively isolated in its mountain refuges by human encroachment and by the geographic barrier of the Rift Valley, it is possible that the species currently occurs at the cold end of what it can tolerate. If that is so, then the mountain gorilla might not be affected by some warming as long as rainfall remains high and well distributed, such that net primary productivity is maintained throughout the year.

The Niche conservatism models and the limiting factor modes predicted that suitability of the parks would not be much affected in 2050 or 2090. The behavior model showed intermediate predictions, and had the largest decrease of suitable area between 2050 and 2090. The differences among the model predictions for areas immediately outside (less than 5km) were comparable to those for the parks. The limiting factor and behavior models predicted much more area to be suitable in the future in other areas of the Albertine Rift ecoregion.

Rather than relying on a single “best” model, some authors (e.g., Thuiller, 2003) have argued for using many models and applying some sort of model averaging. When Thorne et al. (2013) applied the model averaging, a number of areas that might be climatically suitable in 2090 were revealed. These include most of the parks and areas

56

outside the parks, especially areas to the south and west. Additional areas appear to be suitable now and in the future. These include areas in the Ruwenzori Mountains in Uganda and the DR Congo, to the west of the Mikeno Sector of the Virunga National Park (Mt Nyamulagira area in Virunga National Park, DR Congo), Nyungwe National Park in Rwanda and Kibira National Park in Burundi. Lands suitable for park expansion were found in most directions from existing protected areas. Areas downslope of the current parks in the Virungas appear to currently be, and will remain, suitable,

particularly regions in Rwanda. For Bwindi, the area most suitable for expansion appears to be the west of the southern lobe of that park. In addition, Gishwati forest in Rwanda, a relic ancient rainforest, directly south of the Virunga Massif, was deemed suitable in about 60% of the future model runs.

Differences in model predictions point to gaps in our knowledge that could be used to guide more research. For example, areas identified as unsuitable in one model but suitable in another, like the northern part of Bwindi which is at the lower and warmer areas of the park and is currently available but unoccupied by gorillas, could become a focus of further study (Thorne et al. 2013).

2.10. Synthesis

There is much variation in the physical environment of Bwindi, therefore gorillas utilize a range of habitats in the park. There are also different types of human disturbances in the park that occur singly in particular locations or in combination in other locations. Human disturbance occurs in the whole park but is unevenly distributed and different

disturbances are concentrated in different areas. Historical disturbances like logging and

57

mining still have a legacy in terms of wildlife distribution. It is therefore difficult to point out a single environmental factor responsible for gorilla distribution. Given that wild gorillas rapidly flee from people and may increase the distance they travel per day (Cipolletta 2003), the gorillas might respond by shifting their home ranges away from areas with high levels of human presence and direct disturbance. I therefore would postulate that levels of human activity in and around Bwindi are responsible for the present distribution of wild gorillas in the park rather than ecological factors.