Recognizing sources of uncertainty in disease vector ecological niche models: An example with the tick Rhipicephalus sanguineus sensu lato

Abstract

Epidemiology is one of many fields that use ecological niche modeling to assess potential distributions or potential range expansions of species. When such models are transferred in space and time, it is important to understand sources and location of uncertainty in their predictions. Here, we used the tick species Rhipicephalus sanguineus sensu lato (distributed in different areas around the world) as an example; for the first time, we characterized its global geographic distribution using ecological niche modeling, and explore the uncertainty involved in transferring models in space and time. We assessed uncertainties based on risks of strict extrapolation and amounts and patterns of variation in our predictions. We integrated occurrence records and climate data to calibrate models for 5 world regions, and to project them to 11 general circulation models (GCMs) and two representative concentration pathway emissions scenarios (RCPs) for 2050. Models created in different calibration areas showed high agreement of suitable areas among model predictions from the eastern United States, southern Mexico, South America, Europe, North Africa, sub-Saharan countries, Asia, and Australia. The global potential distributions of R. sanguineus sensulato were very similar between the two RCPs, but GCMs, model replicates, and model parametrizations contributed importantly to the overall variation detected. Patterns of uncertainty (strict extrapolation areas and variation) in our model predictions depended on the calibration area, and underlined the important implications of not considering variability and extrapolation risk in interpretations of ecological niche model projections.

Keywords

• Model calibration
• Model extrapolation
• Model variation
• Hierarchical partitioning
• Ecological niche modeling