Improving phenological event identification in trees using manually measured dendrometer data: conventional approaches vs. the novel two-stage threshold approach

Accurate detection of phenological events, such as growth onset, cessation, and seasonal transitions, is essential for understanding tree growth dynamics, particularly in Mediterranean forests where bimodal growth patterns are common. While microcore analysis remains the most precise method, its labor-intensive nature has led researchers to rely on dendrometer-based approaches. However, traditional methods using dendrometer data— such as parametric growth curve fitting with phenological events detected as relative thresholds—are often biased by hydration-related fluctuations and may fail to accurately capture complex growth patterns. In this study, we compared commonly used detection methods, including the Gompertz and Richards parametric functions, a generalized additive model (GAM), and the zero-growth (ZG) concept, with a novel approach: the two-stage threshold approach (2STA). Our results showed that the 2STA consistently outperformed existing methods in identifying spring onset, summer cessation, and autumn onset, exhibiting deviations within the expected measurement uncertainty (±7 days). The method’s reliance on direct growth rate transitions, rather than smoothed curves or arbitrary percentiles, reduced errors associated with hydration-related stem fluctuations. However, year cessation was best estimated using a 95% relative growth threshold derived from the Richards function. These findings highlight the potential of the 2STA as a robust and objective method for phenological event detection using manually collected band dendrometer data, particularly for species with complex intra-annual growth patterns. Future research should focus on refining climate-based thresholds for cambial activity and validating the method across a wider range of species and environmental conditions.