A Regionalization Approach Based on the Comparison of Different Clustering Techniques

For biodiversity conservation and the development of protected areas, it is essential to create strategic plans that ensure the preservation and sustainable use of natural resources. Biogeography plays a crucial role in supporting these efforts by identifying and categorizing geographic areas (regionalization) that represent different biotas, as well as recognizing patterns in biodiversity distribution. Another application of regionalization is in planning species sampling and inventories. Developing a species list is vital for monitoring and understanding diversity patterns. This study focuses on the Palearctic region, specifically the areas between Morocco, the Iberian Peninsula, and France. Its aim is to compare different clustering algorithms—such as K-means++, DBSCAN, PD-clustering, Infomap, and federated heuristic optimization based on fuzzy clustering—with a reference regionalization, using environmental and soil data. Various spatial contiguity approaches were applied, including the third-degree polynomial model and principal coordinates. The results demonstrated that the hybrid approach offers a robust solution in the construction of the regions and that K-means++ and PDC produced regions with strong spatial similarity to the reference regionalization, closely aligning with the expected number of regions, especially at the biome level. Our study shows that a purely statistical regionalization can approximate a global reference regionalization, making it reproducible.