Species distribution models are valuable tools for conservation management. However, there remain challenges in developing and interpreting these models in the marine environment, such as the nature of the species used for the modelling process. When working with mobile species in dynamic environments, lack of observation is usually interpreted as an observation of absence, which can result in the introduction of biases by methodological (false) absences. Here, we explore the role of absences when modelling marine megafauna distributions. To better understand how the use of absences (or equivalent) affects the niche modelling algorithms, we used a set of 20 virtual species with different relations to the habitat (generalist static, specialist static, generalist dynamic and specialist dynamic) with different encounter rates. We tested six different modelling techniques divided into three distinct groups: presence-only, presence-background and presence-absence. We compared the outputs of the models using traditional validation metrics and overlap metrics in the geographical and environmental spaces. Algorithms characterized the ecological niche for the simulated species differently. Approaches using background data generally outperformed the other methods, suggesting that the non-observation of a species in a given location and time should not be considered as an absence. A very intense (practically unrealistic) sampling schema would be required to obtain a genuine unbiased absence when working with these species and habitats. For highly mobile species, a precautionary approach would be to consider the non-observation of a species as part of the background (a sample of the conditions available in the study area) rather than an absence. A good starting point would be to use presence-background models, complemented with presence-absence and/or presence-only models, comparing outputs from the different algorithms tested in the geographic and environmental space. Improving model performance for highly mobile marine species should lead to better-informed decision making for conservation.

Best,

Marc Fernandez
Postdoctoral fellow
MARE – Marine and Environmental Sciences Centre
Agência Regional para o Desenvolvimento da Investigação Tecnologia e Inovação (ARDITI)
Madeira Island, Portugal