Advances in global sensitivity analyses of demographic-based species distribution models to address uncertainties in dynamic landscapes

PeerJ

article

Ilona Naujokaitis-Lewis¹  Janelle M.R. Curtis² 

[1] National Wildlife Research Centre, Carleton University;Centre for Applied Conservation Research, University of British Columbia;Conservation Biology Section, Pacific Biological Station

关键词: Uncertainty;    Global sensitivity analysis;    Picea species;    Population viability analysis;    Disease;    Population dynamics;    Species distribution model;    Landscape dynamics;   

DOI  :  10.7717/peerj.2204

学科分类：社会科学、人文和艺术（综合）

来源: Inra

【 摘 要 】

Developing a rigorous understanding of multiple global threats to species persistence requires the use of integrated modeling methods that capture processes which influence species distributions. Species distribution models (SDMs) coupled with population dynamics models can incorporate relationships between changing environments and demographics and are increasingly used to quantify relative extinction risks associated with climate and land-use changes. Despite their appeal, uncertainties associated with complex models can undermine their usefulness for advancing predictive ecology and informing conservation management decisions. We developed a computationally-efficient and freely available tool (GRIP 2.0) that implements and automates a global sensitivity analysis of coupled SDM-population dynamics models for comparing the relative influence of demographic parameters and habitat attributes on predicted extinction risk. Advances over previous global sensitivity analyses include the ability to vary habitat suitability across gradients, as well as habitat amount and configuration of spatially-explicit suitability maps of real and simulated landscapes. Using GRIP 2.0, we carried out a multi-model global sensitivity analysis of a coupled SDM-population dynamics model of whitebark pine (Pinus albicaulis) in Mount Rainier National Park as a case study and quantified the relative influence of input parameters and their interactions on model predictions. Our results differed from the one-at-time analyses used in the original study, and we found that the most influential parameters included the total amount of suitable habitat within the landscape, survival rates, and effects of a prevalent disease, white pine blister rust. Strong interactions between habitat amount and survival rates of older trees suggests the importance of habitat in mediating the negative influences of white pine blister rust. Our results underscore the importance of considering habitat attributes along with demographic parameters in sensitivity routines. GRIP 2.0 is an important decision-support tool that can be used to prioritize research, identify habitat-based thresholds and management intervention points to improve probability of species persistence, and evaluate trade-offs of alternative management options.

【 授权许可】

CC BY