A hybrid framework for regional climate seasonality study and trend analysis

Abstract

One of the profound effects produced by climate change is shifting the seasons in terms of both duration and start/end dates. It is important for sustainable management to detect and predict any such seasonal changes as they may trigger earlier-than-usual timing of plant phenology, animal migration, and other ecological, environmental, economic, and social implications. In this study, we are using meteorological data recorded in four cities across Southern Ontario, Canada over the past 70 years (1953–2022) to explore regional relationship between climate variables and seasonal shifts. Applying a combination of statistical and machine learning (ML) algorithms, a novel hybrid framework is suggested for detecting, quantifying, and visualizing seasonal clusters and trends. A comparative analysis of different ML clustering algorithms to identify variations in seasonality timing and to establish phenological seasons is conducted. The resultant seasonal clusters are then used to detect shifts in seasonality dynamics and trends in climate parameters.