Subjective uncertainties in habitat suitability maps

Habitat suitability maps are very useful for planning management and conservation priorities. Epistemic uncertainty usually is an intrinsic part of the mapping process and yet, subjective uncertainties have rarely been addressed. Subjective uncertainty arises from subjective judgments, generally by experts, on how to interpret data and translate them into a model. This study examines how to encapsulate subjective uncertainty in a pair of 'bounding'habitat suitability maps. These maps represent extremes of the range of plausible maps obtained from alternative ways to model habitat suitability. Sambar deer (Cervus unicolor) was used to illustrate the process within the spatial context of the alpine areas of Victoria in southern Australia. Four habitat variables (presence of forest, distance to forest edge, presence of gullies, and amount of solar radiation) and three models were retained. A total of 22 equally plausible habitat suitability maps were compared with three different statistics (sum of suitability index, connectance and least-cost distance) to obtain bounding maps, each of them linked to a potential application of habitat suitability maps for this species. The three statistics gave different results, emphasizing the importance of considering how habitat suitability maps are used in subsequent analyses. Very large bounds were found when estimating Sambar population size through the sum of suitability values. With the connectance index, different pairs of bounding maps were found when the Alpine area or the State as a whole was considered. Using least-cost distances, we found that the uncertainty surrounding the choice of the suitability model was much more important than the uncertainty surrounding the ways of computing the habitat variables. Subjective uncertainty can be effectively propagated into subsequent analyses through the use of bounding habitat suitability maps, if the potential use of the maps in subsequent analysis is known. The importance of subjective uncertainties could therefore be assessed and decision-making in conservation ecology would be better informed.