Chinese Phenological Observation Network (CPON)¹ was developed in 1963, which was widely used in phenological studies (Ge et al., 2015). In the present study, we selected sites where ginkgo has been observed for more than 10 years. In total, 10 sites in east China were selected and the spring budburst dates were used in our analyzation (Figure 1; Supplementary Table S1). The climate data were derived from China Meteorological Forcing Dataset², which was developed by Data Assimilation and Modeling Center for Tibetan Multi spheres, Institute of Tibetan Plateau Research, Chinese Academy of Sciences (He et al., 2020). In the current study, we defined the preseason as 2 months prior to the mean date of budburst for each site following previous study (Fu et al., 2016) and further calculated the mean temperature during the preseason.

To investigate spatial variation in the effect of temperature and daylength on gingko budburst, we conducted twig-cutting experiments in climate chambers. We collected twigs of ginkgo from three sites: a high-latitude site at Jiufeng Mountain (JF), Beijing in North China (366 m; 116° 28′ E, 39° 54′ N), and two low-latitude sites differing in elevation at Tianmu Mountain (TM), Zhejiang Province in East China (119° 26′ E, 30°19′ N, high-elevation site, TM_high = 1,105 m; low-elevation site, TM_low = 347 m). TM is characterized by subtropical evergreen and deciduous broad-leaved mixed forest, while JF is characterized by temperate broad-leaved deciduous forest.

Twigs were collected from adult ginkgo trees on January 17–22, 2021 at the above three sites. In total, 29 individual trees were selected, of which 11 and 12 individuals came from TM_low and TM_high, respectively, and 6 individuals from JF (this was the maximum possible number due to the management policy at the site; see details in Table 1). Six twigs, approximately 40 cm in length, were cut from each individual. The twigs were cleaned and disinfected with commercial hypochlorite solution following previous studies (Du et al., 2019; Wu et al., 2022b) and then cultivated in 395 ml plastic bottles filled with tap water. Every 2 weeks from the start of the experiment, the tap water was changed in bottles, the twigs were washed to remove mold grown, and their basal parts were trimmed by about 2 cm to avoid vessel occlusion.

Three climate chambers were used to manipulate air temperature, and daylength treatments were set up in each temperature treatment by covering half of the twigs with shade black cloth. Following a 2 × 3 full-factorial design, two daylength treatments (8-h [P8] and 16-h daylength [P16]) were combined with three temperature treatments (10°C [T10], 15°C [T15], and 20°C [T20]). To avoid plant variations within populations, the twigs of each specific individual were separately put into six environmental treatments (3 temperature × 2 daylength). In total, 174 twigs were used in this experiment, of which, per treatment, 6 twigs (replicates) came from JF, 12 twigs from TM_low, and 11 twigs from TM_high.

Temperature sensors (HOBO M2202) were installed within each environmental treatment. No significant difference in temperature was found between the two daylength treatments by using the ANCOVA analysis (Supplementary Figure S1). Following Vitasse (2013), the budburst date was defined as the date when buds start to open and leaves become partially visible. The twigs were put into the chambers on January 23, 2021. We monitored buds of each twig every 3 days during the treatment period. The days to achieve budburst (BBD) were defined as the number of days from treatment start (January 23, 2021) until budburst.

For the in situ phenological observations, changes in budburst date and the mean preseason temperature along the latitudinal gradient were estimated by using linear regression analysis. In addition, we conducted a correlation analysis to investigate the relationship between budburst date and preseason temperature at each site, and further explored the latitudinal variation of the correlation coefficients between budburst date and preseason temperature. As a stronger daylength effects also associated with a lower variation in budburst dates, we thus estimated the daylength limitation effect by using the standard deviation of budburst dates (Std) as a surrogate measure (Zohner et al., 2016; Geng et al., 2022). For the twig-cutting experiment, the heat requirement for budburst was calculated as the cumulative growing degree days (GDD) from January 23, when the twigs were put into the chambers, until budburst date:

where T_day is the mean daily temperature and T_base is the base temperature. Following previous studies, 0°C was used as T_base (Sarvas, 1972; Fu et al., 2016).

Differences of BBD and GDD between the temperature and daylength treatments were tested using independent samples t-tests. We also calculated chilling days (CHD) when the daily temperature was between −10 and 7°C from 1 September 2020 to the starting date of the experiment (Weinberger, 1950; Wang et al., 2020). The temperature sensitivity of budburst (S_T) was defined as the days advance of budburst date per degree warming (days °C⁻¹), which was calculated using linear regression analysis (Fu et al., 2015). Differences in temperature sensitivity between the daylength treatments were tested using ANCOVA (Fu et al., 2019a). All statistical analyses were conducted using R version 3.5.2.

Based on the in situ observations, we found that the spatial distribution of budburst dates followed a strong latitudinal pattern (Figure 1). Budburst dates (day of year, DOY) varied from DOY 89 at the southernmost Nanchang to DOY 122 at the northernmost Shenyang. For every 1° increase in latitude, DOY delayed by 1.98 days (Figure 1).

In our experiment, 83.3% (145) of the 174 twigs achieved budburst, and these were used for subsequent analysis. Under the same temperature and daylength regime, twigs collected from the high-latitude site (JF) were the first ones to flush, while twigs from the low-latitude/low-elevation site (TM_low) were the last ones to flush, with low-latitude/high-elevation twigs intermediate (TM_high, Figure 2A). In detail, the twigs collected at JF and TM_high showed, on average, 12.6 and 4.9 days earlier budburst, respectively, than those from TM_low.

Budburst became significantly earlier with the increase in temperature across treatments (Figure 2B). Compared with the T10 temperature treatment, budburst occurred 20.6 and 25.9 days (p < 0.05) earlier in the T15 and T20 treatments, respectively. The largest effect of temperature treatment was found for TM_low twigs (32.1 days difference between T10 and T20), followed by TM_high (23.5 days) and JF twigs (22.1 days), respectively.

Long daylength significantly advanced budburst in twigs collected at the two low-latitude sites, i.e., relative to 8-h short-day conditions, budburst occurred 9.6 and 6.7 days earlier under 16-h long days in TM_low and TM_high twigs, respectively (p < 0.05; Figure 3A). However, daylength did not affect budburst dates in twigs from the high-latitude site (JF), i.e., BBD = 23.6 and 25.1 days for the 16-h and 8-h treatments (p = 0.66). Under both 8-h and 16-h daylength, budburst occurred earliest in JF twigs, latest in TM_low twigs, with TM_high twigs intermediate (Figure 3B). Long-day conditions largely reduced the phenological differences among twigs from the three sites relative to 8-h short-day conditions (Figure 3B).

In agreement with the above results, increased temperature consistently advanced budburst, while longer daylength only advanced budburst in twigs from the two low-latitude sites for all three temperature treatments (Figures 4A–C). Interestingly, the daylength effect was greater under warmer conditions with a higher significance level in the T20 treatment (p < 0.001), compared to the T10 treatment (p < 0.05) at both TM_high and TM_low (Figures 4A,B). We further found that temperature sensitivity of budburst significantly decreased with latitude and elevation, ranging from 3.98 days °C⁻¹ (TM_low) to 2.53 days °C⁻¹ (TM_high) and 2.38 days °C⁻¹ (JF). However, no significant difference in the temperature sensitivity of budburst was found when compared between the two daylength treatments (Figure 4D).

As shown in Figure 5; Supplementary Figure S2, the chilling days increased from low elevation/latitude to high elevation/latitude (JF > TM_high > TM_low). Chilling significantly reduced the growing degree days (GDD) required for budburst under both daylength treatments (Figure 5). Among the three sites, twigs from JF had the lowest GDD requirement (336°C) to release budburst; the highest GDD requirement was found for the low-latitude/low-elevation site (TM_low: 500°C), with the low-latitude/high-elevation site being intermediate (TM_high: 389°C, Figure 5). Daylength significantly reduced the forcing requirement at the two low-latitude sites, yet no significant daylength effect was found at JF (Figure 5; Supplementary Figure S2). Long daylength reduced the GDD requirement by 24.0 and 21.6% for TM_low and TM_high, respectively (Figure 5).

Consistent with the above results, the in situ phenological observations showed that spring budburst and temperature are negatively correlated across all sites. The negative relationship (correlation coefficient) between spring budburst and temperature strengthens along latitudes at a rate of −0.02 °N⁻¹ (Figure 6A), suggesting a high temperature controls on ginkgo budburst at high latitudes.

In accordance with Hopkins’ bioclimatic law, the budburst timing of ginkgo was delayed with increasing latitude (Hopkins, 1920; Wang et al., 2015; Meng et al., 2021), driven by decreases in preseason temperature (Supplementary Figure S3). However, while Hopkins proposed that spring leaf unfolding delays by ~4 days for every degree increase in northern latitude (Hopkins, 1920), we observed a less pronounced response of only 1.98 days °N⁻¹. Cheng et al. (2021) and Liu et al. (2019) reported decreases in the latitudinal variation of spring phenology over time, which was mainly attributed to spatial differences in the temperature sensitivity of spring phenology, with only slight advances under warming at low latitudes and strong advances at high latitudes. Consequently, these asymmetric changes in spring phenology across latitudes lower the spatial variation in budburst dates, which was also supported by our experimental results. We found a greater daylength limitation on budburst at low latitudes and daylength insensitivity at high latitudes, which would lead us to predict a larger advance in budburst dates at high latitudes relative to low latitudes under climate warming. Climate warming is thus likely to reduce spatial differences in budburst dates both across latitudinal and elevational gradients.

Consistent with previous studies, we observed significant advances in budburst with warming (Fu et al., 2013; Hänninen, 2016; Wu et al., 2022b). In addition to temperature, daylength has been suggested as another dominant factor regulating budburst dates (Basler and Körner, 2012; Way and Montgomery, 2015; Fu et al., 2019a). Since frosts may unexpectedly occur until late in spring, reliance on daylength might help plants to prevent precocious leaf-out and frost damage to young leaves (Zohner et al., 2020a). The forcing requirements of daylength sensitive species decrease with increasing daylength, thus delaying budburst under early warm spells (Fu et al., 2019a). On the other hand, warmer winters might also increase plants’ forcing requirements as a result of reduced chilling accumulation. Long daylength can compensate for insufficient chilling and promote budburst, allowing plants to use favorable spring conditions for photosynthesis (Basler and Körner, 2012; Vitasse, 2013; Way and Montgomery, 2015). Daylength can thus have a dual role, both delaying and advancing budburst under certain conditions, reducing the overall variation in budburst dates over time.

Recent study demonstrated the decreased daylength sensitivity of spring phenology along elevation (Wu et al., 2022b), which was consistent with our observations. Interestingly, we found a high daylength sensitivity at low latitudes and daylength-independent at high latitudes. This might partly be explained by the high chilling accumulation at high latitudes (Zhang et al., 2007; Wenden et al., 2020), whereby chilling compensates for daylength, leaving spring temperature (GDD) as the dominant factor regulating budburst dates. On the contrary, chilling accumulation was substantially reduced at low latitudes, causing daylength to play a more important role by interacting with temperature. The chilling threshold to break dormancy might vary across space as well, and trees from high latitudes might exhibit higher chilling requirements (Sawamura et al., 2017). Since warming rates increase with latitude (IPCC, 2021), chilling accumulation might also become insufficient at high latitudes in the future, and thus, daylength may affect future budburst dates at both high- and low-latitude regions. In line with experimental results, daylength effect was mainly found at low latitudes using in situ observations. In details, we found that variation in budburst dates that could be an indirect index of daylength effect on budburst as suggested by Zohner et al. (2016) was significantly reduced toward low latitudes that suggests a larger daylength limitation on budburst at low latitudes (Figure 6).

In our experiment, we found that twigs collected at high latitude and elevation sites showed earlier budburst than the low-latitude/low-elevation twigs when kept under the same temperature and daylength conditions, which was in agreement with Zohner et al. (2020b). As shown in Figure 5; Supplementary Figure S2, this might be explained by the longer chilling period that twigs from high latitude or elevation sites experienced before the collection, which might have led to a reduction in the heat requirement for budburst and a shorter time to budburst (Laube et al., 2013; Du et al., 2019). In addition, environment-induced adaptive plasticity might lead to lower heat requirements to achieve budburst in colder environments (Vitasse et al., 2010; Firmat et al., 2017).

Interestingly, although the timing to achieve budburst was shortened for the twigs collected in high latitude and elevation, the temperature sensitivity of budburst was significantly lower in twigs from high latitude and elevations. Similar results were reported using a remote sensing-based dataset (Gao et al., 2020). Plants growing in high latitudinal and altitudinal regions with higher temperature variance may have adapted to unstable temperature conditions by developing a growth strategy with a lower temperature sensitivity of budburst (Lechowicz, 1984; Wang et al., 2014). Another possible mechanism is the temperature threshold differences in phenological responses. In detail, plants from high latitudes and elevations may have adapted to lower temperatures and might be more responsive to low temperatures (here ~10°C) than plants from low latitudes and elevations. Thus, while plants from low latitudes and elevations need a lot of time to leaf-out at 10°C, plants from high latitudes and elevations might quickly respond to 10°C.