The influence of climate warming on flowering phenology in relation to historical annual and seasonal temperatures and plant functional traits

Global average temperature has risen by about 1.1 °C since the late 19th century, with sustained warming observed across the United States and specifically in eastern Pennsylvania. Such warming can alter phenology—the timing of biological events—leading to cascading ecological effects, including potential mismatches between flowering plants and their pollinators or between prey and predators. Experimental warming studies, elevational and latitudinal gradients, and urban heat island gradients have all shown temperature-linked phenological shifts. Herbarium collections provide valuable historical records to link flowering dates with concurrent temperature data across long time spans. However, herbarium-based studies can be affected by collection biases (location, time, collector preferences, duplicates) and inconsistent phenophase assessment, though large, curated datasets and consistent criteria mitigate these issues. Prior work shows many species flower earlier with warming (often 2–6 days earlier per 1 °C), but responses vary and some species flower later, possibly due to differential sensitivity to seasonal temperatures (winter vs spring) and species-specific cues such as chilling and vernalization. Functional traits may explain interspecific variation in responses, including native status, growth habit (woody vs herbaceous), fruit type (fleshy vs dry), and seasonality of blooming (spring vs summer). Objectives were to quantify long-term changes in flowering time relative to annual and seasonal temperatures in eastern Pennsylvania using herbarium specimens (1884–2015), to compare the roles of yearly, winter, and spring temperatures, and to test whether native status, growth form, fruit type, and blooming season modulate phenological responses. The authors predicted earlier flowering with warming, differential sensitivity to seasonal temperatures, and stronger responses in non-native, herbaceous, dry-fruited, and spring-blooming species.

The paper synthesizes evidence that plant phenology commonly advances with warming, based on experimental warming studies, elevational/latitudinal gradients, and analyses of herbarium records. Earlier studies reported advances of 2–6 days per 1 °C historically and even larger shifts in recent decades. Yet species show heterogeneous responses, with some delaying flowering. Potential mechanisms include opposing effects of warmer winters (reducing chilling or altering dormancy/vernalization) and warmer springs (advancing phenology). Herbarium studies can be biased by collection practices and phenophase assessment, but large datasets, duplicate removal, and standardized flowering criteria enhance reliability; collection dates have been validated against field observations for flowering. Functional traits may mediate responses: non-natives, especially invasives, may be more plastic; woody species may rely more on photoperiod than temperature; fruit type may constrain developmental schedules; and seasonality (spring vs summer flowering) may alter sensitivity to warming. These prior findings motivate testing trait-based differences and seasonal temperature effects.

Study region: Berks, Bucks, Lehigh, and Northampton counties (Lehigh and Delaware River watersheds) in eastern Pennsylvania, characterized by fragmented mature deciduous forest within agricultural and suburban landscapes. Growing season ~177 days; last frost early April, first frost mid-October. Climate data: Daily average temperatures (1884–2015) were obtained from NOAA (Climate Data Online; Global Summary of the Month CSV) for 11 stations in the study area with extensive coverage. Daily temperatures were averaged across stations to reduce station-specific gaps and variability, then aggregated to monthly averages (TAVG). From these, annual average temperatures (YearTAVG) were calculated. Seasonal metrics were defined as WinterTAVG (mean of Dec of prior year, Jan, Feb) and SpringTAVG (mean of Mar, Apr, May). Phenology data: Physical and digital herbarium specimens were sourced from Muhlenberg College, The Academy of Natural Sciences, and The Morris Arboretum (via Mid-Atlantic Megalopolis Project/Mid-Atlantic Herbaria Consortium). Inclusion criteria: insect-pollinated species with at least 30 specimens collected within the study area during peak flowering (≥75% flowers open), collection years 1884–2015, and removal of duplicate records with identical date and location. Specimens from large urban/industrial centers were excluded. This yielded 36 species from 28 families. Each specimen’s collection date was converted to day-of-year. Species were classified by native status (USDA NRCS), growth form (woody vs herbaceous), fruit type (dry vs fleshy), and blooming season (spring vs summer based on majority of blooming months). Records outside ±2 SD of species’ mean day-of-year were removed to retain the central 95% of observations. Analyses: Linear mixed-effects models (species as random effect) were fit with flowering day-of-year as the response and one of three temperature predictors: YearTAVG, WinterTAVG, or SpringTAVG. Slopes represent change in flowering date per 1 °C change. When residuals were non-normal or heteroscedastic, non-parametric bootstrap 95% confidence intervals were computed. Additional mixed-effects models tested categorical moderators (native status, growth type, fruit type, and seasonality) and their interaction with each temperature variable. Interaction term confidence intervals assessed whether temperature effects differed between categories. To assess sensitivity to long flowering durations, analyses were repeated excluding species with flowering periods >5 months. Analyses were conducted in R (RStudio) using packages: mosaic, readr, tidyr, dplyr, lme4, ggpubr, rstatix.

- Across all species, flowering advanced with higher annual and spring temperatures, but not winter temperatures: - Yearly average temperature (YearTAVG): −2.26 days per 1 °C (95% CI: −3.27 to −1.26), significant. - Winter average temperature (WinterTAVG): −0.16 days per 1 °C (95% CI: −0.59 to 0.38), not significant. - Spring onset temperature (SpringTAVG): −2.93 days per 1 °C (95% CI: −3.62 to −2.27), significant. - Excluding species with extended flowering periods (>5 months): YearTAVG −2.72 (95% CI: −3.96 to −1.74); SpringTAVG −3.16 (95% CI: −3.83 to −2.38); WinterTAVG 0.09 (95% CI: −0.39 to 0.55), reinforcing patterns. - Native vs non-native: No significant differences in temperature–phenology relationships for YearTAVG, WinterTAVG, or SpringTAVG (interaction CIs included 0). - Woody vs herbaceous: Significant difference only for YearTAVG; woody species advanced flowering more with increasing annual temperatures than herbaceous species (β_temp*woody = −2.852; 95% CI: −4.919 to −0.822). No significant differences for WinterTAVG or SpringTAVG. - Fruit type (fleshy vs dry): No significant differences for any temperature period (interaction CIs included 0). - Seasonality (spring- vs summer-blooming): Significant difference for YearTAVG only. Spring-blooming species advanced −3.284 days per 1 °C, summer-blooming −1.014 days per 1 °C; difference = 2.27 days earlier per 1 °C for spring vs summer (95% CI: 0.485 to 4.662). No significant differences for WinterTAVG or SpringTAVG. - Overall, 36 species (1884–2015, eastern Pennsylvania) show phenological advancement primarily driven by spring temperatures; winter temperatures did not significantly predict flowering dates.

Findings support the hypothesis that climate warming advances flowering phenology, with stronger effects linked to spring onset temperatures than annual means, indicating the primacy of spring warming in triggering dormancy termination, germination, and floral development. The lack of a significant winter temperature effect, with some indications of later flowering under warmer winters, suggests that reduced chilling or altered dormancy requirements can delay phenology for some species, potentially offsetting spring warming effects and explaining interspecific variability observed in the literature. Contrary to the prediction that non-natives would respond more strongly due to higher plasticity, no native vs non-native difference was detected, aligning with mixed results in previous studies. Possible reasons include underrepresentation of invasive species among non-natives in the dataset and the role of adaptive selection reducing differences attributed to plasticity. Unexpectedly, woody species exhibited a stronger advance than herbaceous species with higher annual temperatures, challenging assumptions about photoperiod dominance in woody phenology and suggesting complex interactions among cues that may vary by latitude, life history, and seasonality. Fruit type did not modulate responses, potentially due to limited representation of fleshy-fruited species or weak linkage between fruit type and flowering timing. Spring-blooming species advanced more than summer-blooming species with rising annual temperatures, consistent with earlier seasonal cues exerting stronger influence and with reduced risk of late winter damage under warming facilitating selection for earlier flowering. Potential urban heat island effects were minimized in this dataset (few urban specimens), which may partly explain effect sizes on the lower end of reported ranges. Declines in herbarium collection rates in recent decades may also understate recent phenological shifts. Overall, results highlight substantial interspecific variation and the importance of seasonal temperatures and functional traits in shaping phenological responses to climate change.

Climate change is advancing flowering time in eastern Pennsylvania, with mean advances of about 2.3 days per 1 °C for annual temperatures and about 2.9 days per 1 °C for spring temperatures; winter temperatures showed no significant effect. Responses vary among species and are influenced by functional traits, including stronger annual-temperature responses in woody versus herbaceous species and greater advances among spring- versus summer-blooming species, while native status and fruit type did not differentiate responses. Future research should integrate historical and experimental approaches to disentangle mechanisms, explicitly consider seasonal temperature effects, precipitation, snowmelt, extreme weather events, urban heat island influences, trait interactions, and phylogenetic signals, and improve representation of invasive species to better predict community and ecosystem consequences of phenological change.

- Herbarium-based data can be biased due to non-random collection (locations, times, collector preferences) and duplicate specimens; mitigated here by large sample size and duplicate removal but not eliminated. - Phenophase assessment variability and species with long flowering periods can obscure patterns; addressed by standardized peak flowering criteria and sensitivity analyses excluding species with >5-month flowering periods. - Declines in herbarium collection rates in recent decades may underestimate recent phenological shifts. - Climate data continuity issues across stations were mitigated by averaging across 11 stations, but station coverage varied over time. - Underrepresentation of invasive species among non-natives may mask differences in plasticity-related responses. - Limited number of fleshy-fruited species may reduce power to detect fruit-type effects. - Sampling error and limited sample sizes for some species may contribute to interspecific variability.