Spatiotemporal Cluster Transitions in High-Resolution European Phenological Spring Indices

In recent decades, increasing attention has been devoted to studying the ecological and social impacts of our changing climate. In this study, we analyze temporal cluster changes in European phenology as captured by the so-called Extended spring indices. These indices consistently translate temperature records into a suite of biologically meaningful climate change indicators. More precisely, we use the European database of high spatial resolution Extended Spring Indices that we published in the 4TU.ResearchData repository (Izquierdo-Verdiguier et al. 2024) to analyze cluster transition. This database has a spatial resolution of 1 km² and covers the period from 1950 to 2020 at an annual temporal frequency. Four phenological indicators, namely the First Bloom, First Leaf, Last Freeze, and Damage Index were used in this study. The First Bloom, First Leaf, and Last Freeze indices are expressed as the day of year (DOY) on which the respective event occurred, while the Damage Index is computed as the difference between the dates of First Leaf and Last Freeze. To investigate long-term changes in these indices, two 30-year median composites (1960–1989 and 1990–2019) were generated, reflecting the commonly applied climatological assumption of climate variability occurring in 30-year cycles. The MiniBatch K-means clustering algorithm was subsequently applied to both composites, as it has proven effective for clustering large-scale datasets. After deriving cluster centroids for each period, the Hungarian algorithm was employed to align clusters between consecutive periods by matching centroids, ensuring consistent cluster labeling across time. Based on the aligned clusters, transition matrices were computed to quantify pixel-level transitions and changes in cluster characteristics (Atif et al. 2022). The resulting transition maps and cluster statistics enable a detailed visualization of spatial shifts between phenological regimes as well as changes in the internal properties of individual clusters. Our results indicate pronounced phenological changes across Europe between the two 30-years periods, most notably characterized by an earlier onset of First Leaf and First Bloom. By quantifying temporal transitions among phenoregions, this study provides a comprehensive and high-resolution baseline for understanding the rapid reorganization currently reshaping European ecosystems.

 

 

Izquierdo-Verdiguier, E., & Zurita-Milla, R. (2024). A multi-decadal 1 km gridded database of continental-scale spring onset products. Scientific Data, 11(1), 905.

Atif, M. & Leisch F. (2022). clusTransition: An R package for monitoring transition in cluster solutions of temporal datasets. PLOS one, 12 (17).