Conceptual model of autumn phenological responses of temperate trees to early-season development and late-season temperature.

Autumn phenology, represented in this study by the timing of primary growth cessation (bud set), is influenced by two opposing factors: early-season development and late-season temperature. a) In our model, early-season development, which is driven by temperature, has an advancing effect on autumn phenology that lasts until shortly after the summer solstice (green curve). Higher temperatures cause trees to complete their annual life cycles faster, allowing them to set buds. After the summer solstice, as days shorten, trees become increasingly sensitive to cooling conditions, so late-season warming slows the progression of bud set and senescence, delaying autumn phenology (red curve). The compensatory point, where the advancing effect of early-season development is balanced by the delaying effect of late-season warming, is represented by the green circle. b) According to the model, the timing of this effect reversal is flexible and varies between years based on the speed of development. When development is slow or starts late (blue curve), the effect reversal occurs later than under fast or early development (green curve). Therefore, shortly after the solstice, the effect of temperature on autumn phenology differs between fast/early and slow/late developing individuals. For example, point 1 (green circle) shows no net temperature effect on phenology in fast/early trees. By contrast, point 2 (blue circle) shows that in slow/late developing trees, warmer temperatures shortly after the solstice (in July) still advance autumn phenology. However, as the growing season progresses and days shorten, trees become more responsive to cooling regardless of prior developmental speed. By August, both fast/early and slow/late trees should therefore exhibit similar phenological responses, with warming consistently delaying autumn phenology (points 3 and 4).

Effects of early-season development and late-season temperature on the timing of autumn bud set in Fagus sylvatica (experiment 1).

a) Bud set dates for early- (green) and late- (blue) leafing trees including all treatments. Late-leafing trees were cooled (2-7°C) in climate chambers from 4 April to 24 May to arrest their development and delay leaf-out. b) Effects of July (22 June to 23 July) and August (24 July to 25 August) moderate cooling (8-13°C) on bud set date for early- (green) and late- (blue) leafing trees (see Fig. S3 for extreme cooling effects). Analyses show effect size means ± 95% confidence intervals from linear mixed models, including treatment as fixed effect and bud-type (apical vs lateral) as random effect. Early-leafing effects are calculated against the early-leafing control and late-leafing effects are calculated against the late-leafing control. Number labels (1-4) above each point are shown to aid comparison between points 1-4 in the conceptual model (Fig. 1b) and the observed effects. Positive values indicate advances in bud set and negative values indicate delays.

The pre- and post-solstice effects of night, full-day and day cooling on the timing of autumn primary growth cessation in Fagus sylvatica (experiment 2).

Effects of pre-solstice (22 May to 21 June) and post-solstice (22 June to 21 July) cooling on bud set date. Full-day cooling trees were continuously cooled to 8°C, day cooling trees were cooled to 8°C in the day and kept at 20°C at night, night cooling trees were cooled to 8°C at night and kept at 20°C in the day. Analyses show effect size means ± 95% confidence intervals from linear mixed models, including treatment as fixed effect and bud-type (apical vs lateral) as random effect. Positive values indicate advances to bud set and negative values indicate delays.

Diel patterns of relative growth, photosynthesis and theoretical thresholds for cold-induced bud set in Fagus sylvatica.

Vertical red lines indicate the start of the day, and vertical blue lines indicate the start of the night, marking the boundaries between the 12-hour treatment windows used in experiment 2. a) The green curve shows relative carbon assimilation rate measured under controlled conditions (Urban et al., 2014), with values linearly interpolated between measurements. The black curve shows the relative probability for growth (Zweifel et al., 2021). Both curves were smoothed by taking the running mean of the target value, the previous value and the following value. At night, low temperatures slow the trees’ developmental processes, possibly leading to delayed primary growth cessation, while low temperatures during the day reduce photosynthetic activity. b) As days shorten after the solstice, autumn bud set becomes increasingly responsive to cooling. Temperatures below a certain threshold induce overwintering responses, advancing autumn phenology. Our results indicate that daytime cooling of 8°C is below this threshold. However, because daily temperatures reach their minimum during the night, trees’ induction thresholds should be lower at night than during the day. Post-solstice night-time cooling of 8°C may thus have delayed senescence by slowing development rather than inducing it.

Description of the temperature treatments applied in experiment 1.

July treatments were from 22 June to 23 July. August treatments were from 24 July to 25 August. Ambient means outside under natural conditions. Temperature ranges indicate the daily minimum and maximum temperatures experienced by trees inside climate chambers. The early-leafing and late-leafing trees were experimentally generated by placing potted trees in climate chambers from 4 April to 24 May and cooling them to 2°C at night and 7°C during the day to arrest spring development.

Description of the temperature treatments applied in experiment 2.

June treatments were from 22 May to 21 June. July treatments were from 22 June to 21 July. Ambient means outside under natural conditions. The remaining temperature regimes are in the format day/night and refer to the temperatures applied to trees in climate chambers.

Seasonal soil water content (%) for each treatment (experiment 1).

Five individuals per treatment were measured at each time step.

Effects of slow spring development on the timing of autumn bud set in Fagus sylvatica (experiment 1).

Spring development was slowed by cooling the late-leafing trees from 4 April to 24 May in climate chambers. The late-leafing trees experienced a simulated day-night cycle of temperature and light availability, with minimum (2°C) and maximum (7°C) temperatures reached during the night- and daytime, respectively and otherwise simulated ambient conditions. Analyses show effect size means ± 95% confidence intervals from linear mixed models, including treatment as fixed effect and bud-type (apical vs lateral) as random effect. For each late-leafing treatment, the effect is calculated against the equivalent early-leafing treatment. Positive values indicate advances to bud set and negative values indicate delays.

Effects of early-season development and extreme late-season cooling on the timing of autumn primary growth cessation in Fagus sylvatica (experiment 1).

Effects of July (22 June to 23 July) and August (24 July to 25 August) extreme cooling (2-7°C) on bud set date for early- (grey) and late- (blue) leafing trees. Analyses show effect size means ± 95% confidence intervals from linear mixed models, including treatment as fixed effect and bud-type (apical vs lateral) as random effect. Early-leafing effects are calculated against the early-leafing control and late-leafing effects are calculated against the late-leafing control. Positive values indicate advances to bud set and negative values indicate delays.

Effects of early-season development and late-season cooling on total bud growth in Fagus sylvatica (experiment 1).

Effects of July (22 June to 23 July) and August (24 July to 25 August) moderate and extreme cooling (8-13°C and 2-7°C, respectively) on total seasonal (4 July to bud set date) bud growth for early- (grey) and late- (blue) leafing trees. Analyses show effect size means ± 95% confidence intervals from linear mixed models, including treatment as fixed effect and bud-type (apical vs lateral) as random effect. Early-leafing effects are calculated against the early-leafing control and late-leafing effects are calculated against the late-leafing control.

Effects of early-season development and late-season cooling on relative bud growth in Fagus sylvatica (experiment 1).

Effects of July (22 June to 23 July) and August (24 July to 25 August) moderate and extreme cooling (8-13°C and 2-7°C, respectively) on relative seasonal (4 July to bud set date) bud growth for early- (grey) and late- (blue) leafing trees. Analyses show effect size means ± 95% confidence intervals from linear mixed models, including treatment as fixed effect and bud-type (apical vs lateral) as random effect. Early-leafing effects are calculated against the early-leafing control and late-leafing effects are calculated against the late-leafing control.

Relationship between end of season bud length and the timing of primary growth cessation in Fagus sylvatica (experiment 1).

Effect of bud set DOY (day-of-year) on the final bud length at the end of the growing season from a linear model with 95% confidence interval.

Bud set timing (day-of-year) for each control treatment in experiment 2.

One-way ANOVA revealed no significant differences in bud set timing between the three control groups (F2,136 = 0.346, p=0.708). Ambient control trees were placed outside throughout the experiment, Pre-Chamber control trees were in a climate chamber under simulated ambient conditions between 22 May and 21 June, Post-Chamber control trees were placed in a climate chamber under simulated ambient conditions between 22 June and 21 July.

Effects of pre- and post-solstice daytime, nighttime and full-day cooling on relative bud growth in Fagus sylvatica (experiment 2).

Pre-solstice (22 May to 21 June) and post-solstice (22 June to 21 July) cooling treatments were applied. Full-day cooling trees were continuously cooled to 8°C, day cooling trees were cooled to 8°C in the day and kept at 20°C at night, night cooling trees were cooled to 8°C at night and kept at 20°C in the day. Analyses show effect size means ± 95% confidence intervals from linear mixed models, including treatment as fixed effect and bud-type (apical vs lateral) as random effect.

Photosynthesis rates of Fagus sylvatica trees under pre-solstice ambient conditions, daytime, nighttime and full-day cooling in July (experiment 2).

Absolute leaf-level photosynthesis (net CO2 assimilation).

Conceptual model of pre- and post-solstice effects of day and nighttime temperatures on the autumn phenology of Fagus sylvatica.

The colours of the shaded areas represent the effect of cooling to different temperatures at specific times of the day. Above a certain threshold, plants function normally, and cooling has little effect on the timing of primary growth cessation (green; here depicted as 10°C, n.b. this is an intended simplification and does not necessarily accurately represent the reality). Light blue areas show combinations that lead to arrested development, which delays primary growth cessation. Pale green areas depict combinations that reduce photosynthesis and potentially growth but do not affect phenology. Dark purple areas depict combinations that lead to the induction of senescence and overwintering processes, therefore advancing primary growth cessation. a) Before the solstice, lengthening days give temperate trees a reliable cue of the favourable conditions to come, so there is no temperature that can induce senescence at this time of year (Kramer, 1936). Nighttime cooling (< ∼10°C) slows down beech’s developmental processes such as meristematic activity, tissue expansion and maturation, which leads to a growth and development debt (Zweifel et al., 2021). As the days are sufficiently warm, photosynthesis is unaffected. This means that the trees accumulate excess carbon that can only be used later in the growing season to make up for this developmental debt, thus delaying primary growth cessation. Daytime cooling (< ∼10°C) slows development to a lesser extent, however, it also reduces carbon gain by inhibiting photosynthesis but not respiration. Due to carbon limitation, affected trees may prioritise development over growth, so buds may be shorter but there is no resulting delay to primary growth cessation. b) After the solstice and due to shortening days, trees become increasingly sensitive to cooling, especially during the day. Therefore, daytime temperatures < ∼10°C are likely to induce overwintering responses and therefore advance autumn phenology. However, due to typical daily cycles of temperature always reaching their minimum during the night, trees are less susceptible to cooling in the night than in the day. At night, temperatures between ∼4°C and ∼10°C slow the trees’ developmental processes, leading to delayed primary growth cessation.

Mean and 95% confidence intervals of 50% leaf-out (day-of-year), bud set (day-of-year) and absolute bud growth (cm) for each treatment in experiment 1.

Mean and 95% confidence intervals of bud set (day-of-year) and absolute bud growth (cm) for each treatment in experiment 2.