Using co-clustering to analyze spatio-temporal patterns: a case study based on spring phenology

Clustering is a widespread method to explore patterns in large spatio-temporal datasets. Most clustering studies are, however, performed either from a spatial or from a temporal point of view. This is sub-optimal because patterns explored from a spatial perspective cannot describe the time-varying behavior present in the dataset and vice versa. Here we illustrate a co-clustering-based analysis that enables the simultaneous analysis of spatial and temporal patterns. In particular we show that by combining the Bregman block average co-clustering with I-divergence and the k-means algorithms we can extract the main patterns of leaf onset over Europe, northern Africa, Turkey and the Middle East. Our results indicate that four main spatial patterns exist in the period under study (1950 to 2011). These patterns were visualized using maps and a timeline was used to indicate the years of occurrence of these patterns.