Investigating climate-phenology relationships among the most common Italian forest species using Sentinel-2-derived vegetation phenology and productivity products
Climate change is profoundly altering forest phenology and productivity across Europe, with particularly strong impacts in Mediterranean regions characterized by high climatic heterogeneity. Understanding how climatic and site-specific drivers regulate the start, end, and length of the growing season, and how these phenological shifts translate into productivity responses, remains a key challenge for predicting forest carbon dynamics. In this study, we investigate phenological timing and total seasonal productivity across multiple Italian forest species spanning Mediterranean, temperate, and mountain environments, leveraging the new High-Resolution Vegetation Productivity and Phenology product from the Copernicus Land Monitoring Service, machine learning (random forests) modeling, and explainable artificial intelligence analysis (SHAP). Our results confirm a general lengthening of the growing season driven mainly by chilling accumulation and spring temperatures. Warmer conditions advance the start of the season by 1-10 days across species, while the combined effects of temperature, radiation, and moisture can extend the growing season by up to 20-30 days. End-of-season dynamics and season length are more strongly controlled by light and water availability than by temperature alone. In several Mediterranean species, the end of the season can advance by up to 40 days due to summer drought, high vapor pressure deficit, and site exposure. Mediterranean species often show compensatory shifts between season onset and senescence, maintaining a relatively stable length of the season, whereas mountain species exhibit a tighter coupling between delayed onset and shortened season length. Phenological shifts are frequently decoupled from productivity, which is mainly regulated by energy and water availability, highlighting species- and site-specific responses to climate change. The findings of this study highlight the substantial advantage of remote sensing data, coupled with machine learning approaches, for advancing the understanding of forest phenology and productivity across broad spatial and climatic gradients.