Fernandez, M., Yesson, C., Gannier, A., Miller, P.I. & Azevedo, J.M.N. (2017) The importance of temporal resolution for niche modelling in dynamic marine environments.
Journal of Biogeography, 44(12), 2816-2827. DOI:10.1111/jbi.13080 (IF2017 4,154; Q1 Ecology)Highly dynamic ocean environments can experience dramatic changes over relatively short timeframes, affecting the spatial distribution of resources and therefore the presence or absence of highly mobile species. We use simulation studies to investigate how different temporal resolutions might affect the results of species distribution models for highly mobile species (e.g. cetaceans) in marine environments.
Azores archipelago, Portugal.
We developed three virtual species with different habitat preferences influenced by (1) only static (topographic), (2) only dynamic (oceanographic), and (3) both dynamic and static variables. Assuming that species would reposition themselves daily according to these preferences (as has been observed for large marine foragers such as cetaceans), we used two different approaches (generalized linear model and generalized boosted model) to test the effect of using daily, weekly and monthly environmental datasets to model distributions.
The results showed that the selection of different temporal scales has a very important effect on model predictions. When dynamic variables are important components of habitat preference, models based on daily or weekly timeframes performed best at reconstructing the known niche.
It is important that we consider temporal resolution when applying species distribution models. Several factors (e.g. species ecology and oceanographic characteristics of the ecosystem) should be taken into consideration when selecting an adequate temporal scale for niche modelling. For fine scale applications (e.g. dynamic ocean management), highly dynamic ecosystems, and highly mobile species, our results suggest exploring temporal resolution of 7–8 days rather than coarser temporal scales. For some applications annual, seasonal or even monthly averages may produce inferior or inaccurate models.
