| Resumen |
Species distribution models can be approached as a supervised learning problem, where the model’s independent or predictor variables correspond to known environmental variables that influence the species’ behavior, and the dependent variable is the presence of the species in a specific area. The quality of these models depends on various factors, with the quantity and quality of the input variables being crucial. Particularly for seabirds, obtaining in situ data for predictor variables is challenging due to the vast areas these species cover. Hence, satellite data presents itself as a viable and cost-effective option, providing continuous environmental variables over extensive areas and at different temporal resolutions. Since these data come from various sources, they are commonly found in different formats and with varied spatial and temporal resolutions. To standardize them as model input, it is essential to homogenize them. This work presents the methodology used for data acquisition and pre-processing before integration into the model. The process includes, first, bulk downloading of data through a semi-automated approach using command-line tools provided by data servers. Second, the homogenization of data from various sources with different characteristics is performed. This methodology is applied to the case study of predictor variables for modeling the distribution of seabird species. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024. |
