Xiao Feng

Abstract
Aim: Ecological niche model (ENM) or species distribution model is a modeling technique broadly used in ecology and biogeography and is increasingly used in decision-making regarding land use and biodiversity conservation. The methodology behind ENM applications is critical for model accuracy. One critical question that every ENM study faces is how to define a model training area. Theories have suggested designing a training domain based on species’ dispersal ability for improved model performance (species’ mobility hypothesis). While this idea has been well perceived, there still lacks direct quantitative evidence that whether this approach leads to optimal model performance. Here I conducted a modeling experiment to investigate species’ mobility hypothesis.
Location: North and South America
Time period: 1950-present
Major taxa studied: hummingbird
Methods: The modeling experiment was based on a group of hummingbird species. A series of spatial buffers (from 5 to 5000km) were created around occurrences, where background data were sampled and used as input for model calibration. The models calibrated with spatial buffers were compared with models calibrated with training domains that considered species’ dispersal abilities (bioM).
Results: The experiment showed that model performance increased when the size of the training domain was larger, though the model performance reached saturation when size of the training domain passed a certain threshold. The model performance based on bioM was comparable to the saturation performance of models when spatial buffers were used.
Main conclusions: This study provided positive evidence that supports the species’ mobility hypothesis that designing a training domain based on species’ dispersal ability could lead to optimal or near-optimal model performance. When no information of dispersal is available, modelers may use a tuning strategy to identify the size of the training domain for optimized model performance.