Clustering analysis of EOF to identify seasons and essential habitats
To identify distinct essential habitats and to relate these with ecological season, a clustering analysis was performed on the loading factors and the EOF maps.
Through EOF or PCA, individuals (here time steps) and variables (here locations) are projected into two distinct spaces: the space of the individuals (time steps) and the space of the variables (locations). Often in standard PCA, clustering is realized in the space of the individuals only, but the same can be done in the space of the variables. While the first clustering allows to regroup individuals that have the same variable values, the second clustering allows to differentiate individuals.
This way, the clustering will regroup locations that have similar temporal patterns and time steps that have similar spatial patterns. Clusters of locations will be interpreted as distinct essential habitats and clusters of time steps will be interpreted as ecological seasons.
We performed clustering based on a Hierarchical Clustering on Principal Components (HCPC) through the package FactoMineR (LĂȘ et al., 2008).