Experimental warming differentially affects vegetative and reproductive phenology of tundra plants

High-latitude and high-elevation regions are warming faster than the global average, which is altering tundra plant phenology (the timing and duration of vegetative and reproductive stages). Although spring advancement in green-up and flowering is well documented, it is unclear whether all phenophases shift synchronously or respond differently, particularly late-season phases like senescence. Understanding these responses is vital for predicting impacts on plant–pollinator interactions, herbivory, productivity, and carbon and energy balances in the tundra, a biome with relatively sparse phenology research. The study synthesizes experimental warming data from the International Tundra Experiment (ITEX) to test how warming affects multiple phenophases across the growing season. The authors pose five questions addressing magnitude and direction of phenological shifts, differences between reproductive and vegetative phenology, changes in phenoperiod duration, variation across spatial/temporal/climatic gradients, and temporal consistency of responses. They hypothesize: (1) differential responses by tissue type (reproductive vs vegetative) and by timing (early vs late season), (2) shifts in both timing and duration of growth, flowering, and fruiting periods, (3) stronger effects in colder climates (higher latitudes, colder years), (4) stronger effects in dry sites and with year-round chamber deployment, and (5) enhanced responses over time due to initial lags.

The paper situates its work within prior findings that spring phenology generally advances with warming while autumn responses (particularly senescence) are less resolved due to conflicting evidence and fewer studies. Differences between reproductive and vegetative phenology may stem from tissue-specific frost protection mechanisms and evolutionary drivers (pollinator interactions vs herbivory). Observational and remote-sensing studies suggest early-season advancement but mixed autumn trends; experimental warming studies (e.g., OTCs) help isolate temperature effects from confounded drivers like snowmelt and photoperiod. Prior work indicates potential stronger temperature sensitivity in colder climates and interactions with soil moisture and snow dynamics. The authors note gaps: limited species coverage, spatial extent, time periods, and inclusion of multiple phenophases in experimental studies, especially in tundra.

Design and sites: Synthesized phenology observations from 18 ITEX sites (46 subsites) across Arctic, sub-Arctic, and alpine ecosystems (1992–2019). Passive open-top chambers (OTCs) provided experimental warming of approximately 0.5–2.3 °C relative to controls. Subsites were categorized by soil moisture (dry <20% GWC; moist 20–60%; wet >60%). At year-round OTC sites, snowmelt dates were recorded when possible. Phenophases: Six standardized phenophases were analyzed: green up, start of flowering, end of flowering, fruiting, seed dispersal, and leaf senescence. Sites followed the ITEX manual; site-specific definitions were harmonized into these categories. Evergreen and deciduous leaf phenology were not separated, and male/female flowering times were not distinguished. Data handling: To address variable census intervals, each recorded day-of-year (DOY) event was paired with a prior-visit date to define an interval within which the event occurred. For cases lacking a prior visit, a conservative prior date was imputed (species-specific minimum across years minus 3 weeks; minimum DOY 100 for green up and 120 for flowering/end of flowering). Species×subsite×year combinations for green up and senescence with >20% missing observations were discarded. Species names were standardized (The Plant List via Taxonstand). Climate data: Daily mean air temperatures were compiled from local stations and infilled (ERA5-based methods) with strict limits (≤5 missing days per climate window). For each species and site, climate windows were defined as the 30 days prior to the average DOY of each phenophase. Two climate metrics were computed: site mean temperature over measurement years (within the phenophase window) and site-year temperature anomaly. Statistical analysis: A two-step approach was used.

1. Interval-censored regression (survreg) estimated replicate-level mean DOY and standard error for each treatment (OTC/control)×species×subsite×year replicate. Outliers where OTC-control differences exceeded 4 SD were removed; data were standardized by midpoint between DOY and prior-visit.
2. Bayesian hierarchical models (brms) estimated treatment effects with replicate-level DOY (and SE) as response. The core model included fixed effect of treatment (OTC vs control) and random slopes/intercepts for species, site, year-within-site, and subsite-within-site. Non-informative priors were used; variance components had half-Student-t priors; correlations used LKJ priors. Convergence was assessed via trace plots and Gelman-Rubin (<1.1). Effects were considered present when 90% and 95% Bayesian credible intervals did not cross zero. Interactions: Additional models tested treatment interactions with six spatiotemporal predictors: years of warming, latitude, soil moisture (dry/moist/wet), OTC deployment period (year-round vs summer-only), site mean temperature (spatial), and site-year temperature anomaly (temporal), using group-mean centering for climate predictors. Phenoperiod duration: Differences between posterior distributions for paired phenophases estimated changes in duration of growing (green up to senescence), flowering (start to end), and fruiting (fruiting to seed dispersal) periods.

• Experimental warming significantly shifted five of six phenophases (90% CIs not overlapping zero):
  • Green up: advanced by 0.7 ± 0.5 days (estimate -0.731 d; 90% CI -1.558 to -0.033).
  • Flowering: advanced by 2.4 ± 0.6 days (estimate -2.437 d; 90% CI -3.477 to -1.503; 95% CI -3.724 to -1.330).
  • End of flowering: advanced by 1.9 ± 0.6 days (estimate -1.877 d; 90% CI -2.788 to -0.992; 95% CI -3.031 to -0.760).
  • Fruiting: no consistent overall response (estimate -2.581 d; 90% CI -5.600 to 0.100; CI overlaps zero).
  • Seed dispersal: advanced by 2.9 ± 1.4 days (estimate -2.902 d; 90% CI -5.187 to -0.702; 95% CI -5.737 to -0.179).
  • Leaf senescence: delayed by 0.8 ± 0.4 days (estimate +0.766 d; 90% CI 0.174 to 1.340; 95% CI 0.032 to 1.452).
• Pattern by tissue and season: Reproductive phenophases advanced more than vegetative phenophases (tissue-type response). Early (green up) advanced while late (senescence) was delayed (early–late response).
• Growing season length: Net increase of 1.5 ± 0.6 days in species’ growing season (green up to senescence), equating to ~2.5–3.75% of the average 40–60 day growing season (~3% of 50 days). No significant change in flowering or fruiting period duration.
• Spatiotemporal interactions (3/42 significant):
  • Stronger flowering advancement in dry vs moist sites: OTC effect 1.31 ± 0.66 days earlier in dry sites (relative to moist).
  • Stronger flowering advancement with year-round vs summer-only OTC deployment: 2.19 ± 1.08 days earlier.
  • Seed dispersal advanced more at sites with warmer ambient temperatures during species’ dispersal windows: 0.88 ± 0.53 days earlier per °C of site mean temperature in the dispersal window.
• Consistency: Responses were relatively consistent across species, sites, years, and latitudes, with modest site-level variation and little evidence of change in response magnitude over time.
• Magnitude context: Observed shifts correspond to an average OTC warming of ~1.4 °C (range 0.5–2.3 °C), likely conservative relative to projected Arctic warming (3–13 °C by end of century).

The results show that experimental warming does not shift all tundra phenophases uniformly. Reproductive phases (flowering, end of flowering, seed dispersal) are more responsive than vegetative phases, supporting a tissue-type response, while opposing shifts of green up (earlier) and senescence (later) support an early–late response. Together these effects lengthen the vegetative growing season and shorten the interval between green up and flowering. Ecologically, even modest day-scale shifts may influence plant–pollinator synchrony, herbivore foraging, reproductive effort, and multi-trophic interactions in systems with brief growing seasons. A 1.5-day longer growing season could increase GPP in tundra, although net carbon storage outcomes may be moderated by increased autumn/winter respiration and belowground allocation dynamics. Warming effects were generally robust across spatial and temporal gradients, with minor modifications by soil moisture (stronger in dry sites) and OTC deployment period (year-round chambers producing greater advancement, likely via earlier snowmelt and greater thermal sums). The enhanced advancement of seed dispersal in warmer ambient dispersal windows suggests early vs late season phases may interact differently with ambient and experimental warming. Phenology responses remained stable over experimental durations, indicating sustained, not transient, effects under continued warming. Overall, the findings demonstrate that considering tissue type and seasonal timing is essential for predicting tundra phenological change and its consequences for trophic interactions and ecosystem function.

This synthesis of 18 ITEX sites, >100 species, and six phenophases provides robust experimental evidence that warming advances most tundra phenophases while delaying leaf senescence, lengthening species’ growing seasons by ~1.5 days (~3%) for ~1.4 °C of warming. Reproductive phases shift more than vegetative phases, and early vs late season phases can respond in opposite directions. Responses were broadly consistent across space, species, and time, with modest modulation by soil moisture, OTC deployment period, and ambient temperatures during dispersal. Future research directions proposed by the authors include:

• Direct physiological tests separating cues for vegetative vs reproductive phenology (e.g., warming leaves vs flowers).
• Concurrent monitoring of plant phenology, pollinators, and reproductive fitness (seed production/viability).
• Community-level phenology monitoring including all species.
• Experiments spanning a gradient of warming magnitudes to evaluate nonlinearity and thresholds.
• Linking phenological changes to ecosystem carbon fluxes via plot-level measurements and eddy covariance to refine process-based models and carbon budget predictions.

• OTC warming levels (0.5–2.3 °C) are modest relative to projected Arctic warming, likely yielding conservative effect sizes.
• Heterogeneous phenophase definitions across sites required harmonization; dioecious differences and evergreen/deciduous distinctions were not modeled separately (though preliminarily assessed).
• Variable census intervals necessitated interval-censoring and imputation of prior-visit dates, introducing uncertainty.
• Not all sites measured all phenophases; fruiting and seed dispersal had fewer observations, increasing uncertainty.
• Limited snowmelt data (subset of sites) constrained assessment of warming–snowmelt interactions.
• Species sampled represent a subset of community members at each site; species-level results may differ from community- or ecosystem-level responses.
• Inability to standardize temperature increase across all sites/years prevented estimation of per-degree phenological sensitivity from this dataset.