Beyond seasonal means: ETCCDI thermal extreme indices and Robinia pseudoacacia flowering phenology in Piedmont (NW Italy)

Numerous studies have documented a significant relationship between seasonal average temperatures and flowering phenology across a wide range of species and regions However, seasonal averages flatten the temporal structure of thermal forcing, potentially obscuring the thermal patterns that drives biological processes. Building on evidence that spring mean Tmax is a primary predictor of Robinia pseudoacacia L. (black locust) flowering phenology, this study presents a novel framework to assess the independent contribution of ETCCDI-based thermal extreme indices, using Piedmont (NW Italy) as a case study.

Three indices were computed for all phenological observation cells across Piedmont using AgERA5 daily temperature data and ETCCDI-compliant 90th percentile thresholds (5-day centered window, 1991–2020 baseline): TX90p_spring (count of spring days with Tmax exceeding the 90th percentile, March 1 – April 15); Warm Spell Duration Index (WSDI, count of spring days belonging to runs of at least 6 consecutive days above the threshold); TX90p_winter (count of winter days with Tmax above the 90th percentile, December–February). Each index was then correlated with the corresponding seasonal mean Tmax and with observed first flowering Day of Year (DOY) to evaluate its predictive power. A residual analysis was subsequently conducted to isolate phenological variability not accounted for by seasonal means alone.

Results show that TX90p_spring retains a significant relationship with flowering DOY even after removing the effect of spring mean Tmax, confirming that the frequency of anomalously warm spring days carries phenological information independent of the seasonal mean. WSDI and TX90p_winter both showed significant direct associations with flowering timing, warm spell years and anomalously warm winter days were each associated with earlier flowering, but neither retained significance after controlling for the respective seasonal mean Tmax, suggesting their phenological signal is not independent of the seasonal mean.

These findings support the hypothesis that the temporal structure of spring warming, specifically the frequency of anomalously warm days, modulates flowering timing beyond what seasonal means alone can predict. This distinction carries direct implications for projecting phenological change under climate scenarios in which means and extremes may shift non-proportionally. Notably, while ETCCDI indices are typically interpreted as indicators of thermal stress, TX90p_spring behaves here as a biological forcing agent accelerating phenological development. This advance, however, in a season characterized by inherently unstable weather conditions, may increase exposure to late frost or precipitation events during a phenological sensitive phase, with potential consequences for nectar production and apicultural yields.