RESEARCH LETTER Climatic Warming Increases Spatial Synchrony in Spring 10.1029/2018GL081370 Vegetation Phenology Across the Northern Hemisphere Key Points: • The spring phenology will become Qiang Liu1 , Shilong Piao1,2,3 , Yongshuo H. Fu1,4 , Mengdi Gao1 , Josep Peñuelas5,6 , more spatially synchronous across and Ivan A. Janssens4 the Northern Hemisphere 1 • Larger advances in spring phenology Sino‐French Institute for Earth System Science, College of Urban and Environmental Sciences, Peking University, at high latitudes than the low Beijing, China, 2Key Laboratory of Alpine Ecology and Biodiversity, Institute of Tibetan Plateau Research, Chinese latitudes are estimated under future Academy of Sciences, Beijing, China, 3Center for Excellence in Tibetan Earth Science, Chinese Academy of Sciences, climatic warming scenarios • The increased synchrony is due to Beijing, China, 4Department of Biology, University of Antwerp, Wilrijk, Belgium, 5CREAF, Cerdanyola del Valles, both fast spring warming and slow Barcelona, Spain, 6CSIC, Global Ecology Unit CREAF‐CSIC‐UAB, Barcelona, Spain decline in temperature sensitivity of spring phenology at high latitudes Abstract Climatic warming has advanced spring phenology across the Northern Hemisphere, but the Supporting Information: spatial variability in temperature sensitivity of spring phenology is substantial. Whether spring phenology • Supporting Information S1 will continue to advance uniformly at latitudes has not yet been investigated. We used Bayesian model • Table S1 averaging and four spring phenology models, and demonstrated that the start of vegetation growing season across the Northern Hemisphere will become substantially more synchronous (up to 11.3%) under future Correspondence to: climatic warming conditions. Larger start of growing season advances are expected at higher than lower S. Piao, latitudes (3.7–10.9 days earlier) due to both larger rate in spring warming at higher latitudes and larger slpiao@pku.edu.cn decreases in the temperature sensitivity of start of growing season at low latitudes. The consequent impacts on the northern ecosystems due to this increased synchrony may be considerable and thus Citation: worth investigating. Liu, Q., Piao, S., Fu, Y. H., Gao, M., Peñuelas, J., & Janssens, I. A. (2019). Plain Language Summary Recent climatic warming has not only triggered a notable advance of Climatic warming increases spatial spring phenology over the past decades but also changed the spatial pattern of its temperature sensitivity. synchrony in spring vegetation phenology across the Northern Whether the shifts in spring phenology would continue to follow the latitude gradient remains unclear. Hemisphere. Geophysical Research Based on simulations over the end of this century, we concluded that the advance of spring phenology at Letters, 46, 1641–1650. https://doi.org/ high latitudes is larger than the low latitudes, consequently resulting in more synchronous spring 10.1029/2018GL081370 phenology. Further analysis suggests that besides greater spring warming at high latitudes, the larger Received 23 NOV 2018 decrease in temperature sensitivity of spring phenology possibly due to shorter day length and chilling loss at Accepted 11 JAN 2019 low latitudes also contributes to this finding. Our study, therefore, reports the possibility of improving Accepted article online 18 JAN 2019 phenological modules in dynamic vegetation models and thus promoting our understanding of the response Published online 5 FEB 2019 of northern ecosystem to ongoing climate change. 1. Introduction The rising spring temperatures in recent decades have led to a notable advance of spring phenology over the Northern Hemisphere (NH; Fu et al., 2016; Menzel et al., 2006; Peñuelas & Filella, 2001; Piao et al., 2015; Schwartz et al., 2006), which has substantially affected terrestrial water, carbon, and nutrient balances (Keenan et al., 2014; Peñuelas & Filella, 2009; Piao et al., 2007, 2017; Richardson et al., 2010). A recent study reported a declining response of spring phenology to warming temperature, attributed to winter chilling loss (Fu et al., 2015) and questioning whether spring phenology across different regions would continue to advance as climate warms. Extrapolating current knowledge on the bioclimatic gradient in spring phenology (e.g., a progressive delay in spring phenology with increasing latitude, “Hopkin's bioclimatic law”; Hopkins, 1920, 1938) to the future is therefore challenging. Recent studies have reported an increase in synchrony of spring phenology (i.e., reduced differences between the start of growing season (SOS) among species) in recent decades from in situ observations at regional scales (C. Wang et al., 2016) and along altitudinal gra- dients (Vitasse et al., 2018), attributed to spatial differences in warming speed and phenological responses to rising temperature. Whether such increasing synchrony continues to occur across the entire NH under warmer scenarios, however, has not yet been investigated. We quantified the changes in the spatial variance ©2019. American Geophysical Union. of spring phenology across the NH under current and future conditions based on model simulations to All Rights Reserved. explore the changes in the synchrony of SOS. LIU ET AL. 1641 Geophysical Research Letters 10.1029/2018GL081370 Models of spring phenology have been widely used to reproduce SOS in previous studies (Jeong et al., 2013; Liu et al., 2018; Melaas et al., 2015), but considerable uncertainties nonetheless remain within and among models due to limited understanding of mechanisms underlying spring phenology (Chuine & Régnière, 2017; Richardson et al., 2012). Multiple model predictions, combining the strengths of individual models and reducing the computed uncertainties, are therefore required, although comprehensive understanding of spring phenology still need well‐designed experimental studies. Previous multimodel studies simply assigned equal weights to all models, regardless of their performance, and calculated their ensemble mean (i.e., simple model averaging (SMA)). We, however, applied Bayesian model averaging (BMA) to deal with individual model uncertainty by assigning higher weights to models with higher accuracies in the observa- tional period. Four models of spring phenology (see section 2), were applied to simulate SOS over the NH under current (1990–2009) and future (2080–2099; four Representative Concentration Pathways (RCPs)) conditions. 2. Materials and Methods 2.1. Climatic Data Sets Daily mean temperature (T) was obtained from the CRU‐NCEP (Climatic Research Unit (CRU) and National Centers for Environmental Protection (NCEP)) v5 data set at a spatial resolution of 0.5 × 0.5° for 1982–2012. This data set combines the CRU‐TS (time series) v3.1 0.5 × 0.5° monthly climatological data (1901–2012) and the NCEP reanalysis at 2.5 × 2.5° and a 6‐hr time interval since 1948 (Mitchell & Jones, 2005; New et al., 2000). For future projections, we acquired data for daily T from predictions by Coupled Model Intercomparison Project phase 5 models (Table S1). The Coupled Model Intercomparison Project phase 5 model predictions for the historical (1980–2005), RCP2.6, RCP4.5, RCP6.0, and RCP8.5 (2006– 2009 and 2070–2099) scenarios were processed and interpolated to a spatial resolution of 1 × 1° using Climate Data Operators (Schulzweida, 2013). 2.2. Satellite‐Derived SOS In this study, we extracted SOS from the latest Normalized Difference Vegetation Index (NDVI) data set by NASA's GIMMS group (NDVI3g.v1) at a spatial resolution of 1/12 × 1/12° spanning the period 1982–2012. Apart from the issues which have already been addressed in NDVI3g (e.g., update of satellite sensors, atmo- spheric interference, and nonvegetation dynamics; Pinzon & Tucker, 2014), artifacts due to snow coverage and changes in calibration were further processed in NDVI3g.v1. The SOS extraction methods include two main steps: (1) reconstructing NDVI time series to a daily basis using data filter function and (2) determining SOS from predefined thresholds or changing characteristics of the filtered NDVI curve. Following Liu, Fu, Zeng, et al. (2016) and Liu, Fu, Zhu, et al. (2016), we estimated the date of SOS using four commonly used methods (i.e., Hants‐Mr (Jakubauskas et al., 2001), Polyfit‐Mr (Piao et al., 2006), double logistic (Julien & Sobrino, 2009), and piecewise logistic (Zhang et al., 2003)). To reduce the uncertainty resulted from inter- method differences, we applied SOS averaged from four methods in our analysis. 2.3. Models of Spring Phenology We used four models of spring phenology involving both endodormancy and ecodormancy phases. These models assume that chilling and forcing (subsequent to chilling) are required to break endodormancy and ecodormancy, respectively, and define the day when the accumulation of forcing becomes critical as the date of leaf unfolding (Cannell & Smith, 1983; Murray et al., 1989). The four models include three temperature‐ driven models, that is, the Sequential Model (Hänninen, 1990; Kramer, 1994), Unified Model, and UniChill Model (I. Chuine, 2000). Besides, the DORMPHOT model additionally considered the effect of photoperiod and assumed that reduced temperature and photoperiod co‐triggered dormancy, and a long day length would promote the accumulation of forcing (Caffarra, Donnelly, & Chuine, 2011). These models were cali- brated using satellite‐derived SOS and daily temperature at plant function type (MacBean et al., 2015) scale for the period 1982–2012 using Bayesian optimization technique (Martinez‐Cantin, 2014). The detailed model implementation could be found in Table S2 and Text S1. 2.4. BMA Previous studies have mostly been based on results from either a single model of spring phenology or the average of multiple models. Simple model averaging (SMA) is superior to a single model but does not LIU ET AL. 1642 Geophysical Research Letters 10.1029/2018GL081370 identify the best model. BMA, however, assigns higher weights to models that have previously performed better and is therefore able to obtain the minimum root‐mean‐square error between the weighted model ensemble and observations (Raftery & Zheng, 2003). We applied BMA to determine the optimal weight for each model of spring phenology at a pixel level for 1982–2012. Supposing that SOS anomaly was forecasted by on the basis of training data SOST anomaly (randomly selected as 80% from the entire study period), using models of spring phenology (M1, … , Mk), then the forecast probability density function, p(SOS), is written as K  pðSOSÞ ¼ ∑ pðSOSjM k Þp M k SOST (1) k¼1 where p(SOS|Mk) is the normal conditional probability density function based on model Mk and p(Mk|SOST) is the posterior probability of model Mk, which is corrected by the training data and also indicates how well model Mk fits the training data (i.e., weight of each model). This term can be calculated as   p SOST jM k pðM k Þ p M k SOST ¼ k (2)  ∑ p SOST jM l pðM l Þ l¼1 where p(Mk) is the prior probability of model Mk. We defined the probability of the training data (e.g., SOST anomaly) as the likelihood function following Raftery et al. (2005), the optimal weights (e.g., p(Mk|SOST)) were determined when the likelihood function reached its maximum. More details about BMA are provided by Raftery et al. (2005) and Vrugt and Robinson (2007). 2.5. Sensitivity of Spring Phenology to Temperature We calculated the sensitivity (ST) of SOS to mean spring temperature (Tspr) as the ratio between changes in SOS (ΔSOS) and changes in Tspr (ΔTspr) during 2080–2099 and 1990–2009: ST ¼ ΔSOS=ΔT spr (3) We calculated BMA‐based SOS for each year over these two periods. Tspr for individual years was calculated as the average of daily mean temperature from 1 March to 31 May. We assessed the influence of future warming on SOS by calculating ST across the NH under four RCPs. We first calculated ΔSOS and ΔTspr for each model and then applied their ensemble mean to reduce the uncertainties within predictions by the Coupled Model Intercomparison Project phase 5 models (Table S1). 2.6. Analysis We ran the models of spring phenology under five independent scenarios to investigate whether the response of SOS to temperature might be constrained by shorter photoperiod as SOS continues to advance or chilling loss due to winter warming. We chose the DORMPHOT model to test the effect of photoperiod. We first ran this model with the actual photoperiod (S1), then increased the photoperiod during spring (March–May) by 1 and 3 hr, individually (S2). The remaining simulations were carried over all four models. We calculated winter temperature from 1 September of the previous year to 31 January of the current year, because the accumulation of chilling begins on 1 September of the previous year in these models. In the third scenario (S3), we randomly selected winter temperature from 1982 to 2009 and kept it constant during 2080– 2099. Similarly, we considered varying winter temperature but kept spring temperature (March–May) con- stant during 2080–2099 (S4). To reduce the sampling bias, we additionally ran S3 and S4 for 9 times and applied the average of 10 simulations in our analysis. In the fifth scenario (S5), we considered varying winter and spring temperature for 2080–2099. We thereafter calculated BMA‐based SOS projections for 2080–2099 and then the differences among these scenarios. We conducted this analysis based on temperature projec- tions from four RCPs to test the influence of longer photoperiod (S2‐S1), winter warming (S3‐S5), and spring warming (S4‐S5) on the date of SOS. LIU ET AL. 1643 Geophysical Research Letters 10.1029/2018GL081370 Figure 1. Bayesian model averaging (BMA) of simulated start of the growing season (SOS) for 1982–2012 and its comparison with simple model averaging (SMA) and four models. (a) The optimal selection among models based on weight inferred from BMA. (b and c) Root‐mean‐square error (RMSE) and (e and f) temporal correlations between simulated and satellite‐derived SOS. (d) The box plots of RMSEs for the individual models, SMA and BMA. (g) The probability distribution of correlation coefficients across the NH. Different letters in (d) indicate significant differences at p < 0.05 using nonparametric Kruskal‐Wallis tests followed by Bonferroni post hoc tests. 3. Results and Discussion 3.1. BMA The weight assigned by BMA served as a good criterion for model evaluation for 1982–2012. The Unified Model performed best in ~30% of the NH, that is, northeastern Siberia and western North America. The UniChill and Sequential Models produced the best agreement with satellite‐derived SOS in ~25 and 24% of the NH, respectively. The UniChill Model outperformed the other models in central Eurasia, and the Sequential Model performed the best in northeastern Europe and northern Siberia. The DORMPHOT model agreed best in ~21% of the NH, albeit fragmentally distributed (Figure 1a). Plants differ across the regions, and species differ in their response to chilling, photoperiod, and spring warming, so different models are likely to be better for different regions. Benefiting from the strength of individual models, the BMA‐based SOS simulations were in better agreement with satellite observations across the NH. As expected, root‐ mean‐square errors were consistently lower compared to simulations based on either SMA or the single models, particularly over southwestern North America and central Eurasia (Figures 1b–1d and S1). The median root‐mean‐square error decreased substantially from 9.3–10.2 days for the individual models to 8.8 days for SMA and 5.8 days for BMA (Figure 1d). These differences were less apparent in the temporal correlations between the simulated and satellite‐derived SOSs (Figures 1e–1g and S2). 3.2. Projected Changes in Spring Phenology and Its Synchrony SOS across the NH was obtained for 2080–2099 by extrapolating the same weights of the phenological models to future simulations. Regionally different responses to the projected climate change were obtained (Figures 2 and S4; based on ensemble mean of model projections under each RCP, because the intermodel differences showed similar magnitude) (Figure S3). SOS was generally significantly earlier in most of the NH LIU ET AL. 1644 Geophysical Research Letters 10.1029/2018GL081370 Figure 2. Changes of SOS and spring temperature between 1980–2009 and 2080–2099. The top panels show the changes of BMA‐based SOS predictions under two representative concentration pathways (RCPs), (a) RCP2.6 and (b) RCP8.5. The middle panels show changes in mean spring temperature (Tspr) under (c) RCP2.6 and (d) RCP8.5. Regions marked with black dots indicate statistically significant (p < 0.05, t test) changes in SOS (a and b) and spring temperature (c and d). Figures (e) and (f) indicate the temperature sensitivity (ST) of SOS to spring warming under two RCPs. (ranging from 75% in RCP2.6 to 91% in RCP8.5, p < 0.05) between 2080–2099 and 1990–2009. SOS, however, was sometimes also delayed, especially under warmer scenarios in the low latitudes (e.g., 30–50°N; Figures 2a, 2b, 3a, and 3b). SOS across the NH was 7.9 ± 6.8 days (RCP2.6) to 19.1 ± 12.3 days (RCP8.5) earlier (average ± SD) during 2080–2099 than the current period (1990–2009). SOS under RCP2.6 scenario advanced more (e.g., >12 days) at high latitudes (e.g., northwestern North America, northern Europe, north- ern Siberia) and the Tibetan Plateau, followed by low latitudes (e.g., southeastern Siberia and central Eurasia; Figure 2a). This spatial pattern of faster SOS advances at higher than lower latitudes was more apparent under warmer RCPs. For example, the difference in the average SOS advancement between higher and low latitudes increased from 4 days under RCP2.6 to 11 days under RCP8.5 (Figures 2b, S4a, S4b, 3a, 3b, S5a, and S5b). LIU ET AL. 1645 Geophysical Research Letters 10.1029/2018GL081370 Figure 3. Changes in BMA‐based SOS and spring warming for 2080–2099 compared to 1980–2009 for various latitudinal bands. Box plots indicate the distribution of changes in SOS, changes in Tspr, and temperature sensitivity (ST) at low (30–50°N), medium (50–60°N), and high latitudes (>60°N) under RCP2.6 and RCP8.5, with black circles denoting their averages. Different letters indicate significant differences at p < 0.05 using nonparametric Kruskal‐Wallis tests followed by Bonferroni post hoc tests. The weakening latitudinal gradient of SOS under future climatic warming indicated the increasing syn- chrony of spring phenology across the NH. The spatial standard deviation of SOS decreased by 4.0% from 32.7 days (~99% of 33.0 days based on satellite observations) for 1990–2009 to 31.4 days for 2090–2099 (RCP2.6) and continued to decrease under warmer RCPs, that is, 30.8 days (5.8%, RCP4.5), 30.4 days (7.0%, RCP6.0), and 29.0 days (11.3%, RCP8.5; Figure 4c). We found that the decrease in the spatial variation of SOS was related to the regional differences in the response of vegetation to the future climatic warming. Average SOS for 1990–2009 was significantly negatively correlated (p < 0.05) with its projected future change (Figures 4a, 4b, S6a, and S6b), implying that regions with later SOSs at current condition (usually at high latitudes) would likely have larger SOS advances under future climatic warming. These findings were also confirmed by each of the four spring phenology models (Figure S7) and thus indicate that the vegetation across the NH will initiate spring growth more synchronously in the future, especially between the low and high latitudes. 3.3. Determinations of the Increasing Synchrony of Spring Phenology The increasing synchrony of spring phenology between 2080–2099 and 1990–2009 is co‐determined by spring temperature (Tspr) and the sensitivity of SOS to spring warming (ST), which both differ between lower and higher latitudes. ST was higher at low latitudes (e.g., 30–50°N) than at high latitudes (e.g., >60°N, p < 0.05 under most RCPs, Kruskal‐Wallis tests followed by Bonferroni post hoc tests; Figures 3e, 3f, S5e, and S5f), in line with results from previous observational studies (Menzel et al., 2006; Wolkovich et al., LIU ET AL. 1646 Geophysical Research Letters 10.1029/2018GL081370 Figure 4. Scatterplots showing the relationship between the average SOS for 1990–2009 and its future change for 2080–2099. (a) Scatterplot based on SOS predictions under RCP2.6. The black line and the two dashed lines indicate the linear regression and its 95% confidence intervals. The red line shows the average changes in SOS within a one‐week moving window. The upper and right insets are the frequency distribution of SOS for 1990–2009 and its change from 1990–2009 to 2080–2099. (b) The similar result based on RCP8.5. (c) The standard deviation of SOS across the Northern Hemisphere under current and future conditions. 2012), implying that vegetation at lower latitudes (or with earlier SOS) is more sensitive to changes in Tspr. The changes in SOS between these periods nonetheless did not always match the spatial pattern of ST, because of the much larger increase in Tspr at higher latitudes (e.g., 0.2–0.7 °C higher than at low latitudes; Figures 3c, 3d, S5c, and S5d). For example, ST was lower for northern Siberia than the Tibetan Plateau, but the advance in SOS was similar between these regions due to the faster increase in Tspr across northern Siberia (e.g., RCP2.6; Figures 2a and 2e). This finding was consistently confirmed under the remaining RCPs (Figures 2 and S4). Although the ST was lower at higher latitudes, the faster increasing in Tspr could also lead to larger advancement in SOS, thereby reducing the spatial differences in predicted SOS across the NH (e.g., RCP2.6; Figure 2). Moreover, ST decreased more at lower than higher latitudes across different RCPs. The decrease in ST between the results under RCP2.6 and RCP8.5 was largest at lower latitudes (30–50°N), averaging 4.1 days/°C (Figures 3e and 3f). The average decrease in ST at higher latitudes (>60°N), however, was only 2.7 days/°C (Figures 3e and 3f). The comparison between ST estimated from RCP2.6 and the two intermediate RCPs confirmed this finding (Figures 3e, S5e, and S5f). Greater synchrony in spring phenology is thus projected across the NH (Figure 4c), accompanied with the faster spring warming at the high latitudes (Figures 3c, 3d, S5c, and S5d). Several hypotheses have been proposed to explain the faster decrease in ST at low latitudes. A first hypothesis correlates the reduced ST with the variance of local spring temperature (Tspr). A recent ground‐based study found that plants at locations with larger Tspr variances were less sensitive to warming (T. Wang et al., 2014), LIU ET AL. 1647 Geophysical Research Letters 10.1029/2018GL081370 because plants under such conditions tend to depend on more stable environmental cues, such as photoper- iod (Körner & Basler, 2010) and/or higher chilling requirement (Schwartz & Hanes, 2010), to avoid prema- ture exposure to damaging spring frost. Such reduction in ST due to higher Tspr variances, however, was not directly implemented in the temperature‐based models of phenology and could not be tested. The hypothesized role for a negative impact of a shorter photoperiod on ST, however, could be tested across the NH using the DORMPHOT model (Table S2), which involves the effect of photoperiod (Caffarra, Donnelly, & Chuine, 2011). We compared SOS projections over 2080–2099 with actual (S1) and increased (S2) photoperiod (see section 2). The difference between S2 and S1 indicated that increasing photoperiod alone would exacerbate the advance of SOS across most of the NH (>98% (Figures S8a–S8d) and >99% (Figures S8e–S8h), across all RCPs), especially at low latitudes (e.g., southern North America, the Tibetan Plateau) and in northern Europe. The effects of increased photoperiod on advancing SOS decreased signifi- cantly toward higher latitudes (Figure S8). This finding not only confirmed the photoperiodic control of SOS (Caffarra, Donnelly, Chuine, & Jones, 2011; Körner & Basler, 2010; Zohner et al., 2016) but also suggested that the shorter days due to earlier SOS might restrict the advance of SOS in the future, particularly at lower latitudes, thus shaping the spatial pattern of a faster decrease in ST at low latitudes. Photoperiod does not exclusively account for our finding; the other models considering chilling and forcing effects also produced a similar spatial pattern of faster decrease in ST at lower latitudes (Figure S9). The control of SOS by cold temperatures during dormancy is a third hypothesis that could account for the regional difference in the future decrease in ST. We tested this hypothesis by applying four chilling‐based models for 2080–2099 but kept the winter temperature (S3) or the spring temperature (S4) constant and then assessing their difference to simulations based on varying winter and spring temperatures (S5; Figures S10a–S10h; see section 2). The warming winter led to a delayed SOS, albeit primarily at lower latitudes (except for the Tibetan Plateau; Figures S10a–S10d). The predicted SOSs across most of the NH (mainly at higher latitudes) did not differ significantly between these two simulations (p > 0.05). These findings imply that increasing winter temperatures may indeed lead to insufficient chilling at lower latitudes (except for the Tibetan Plateau), which would subsequently minimize the response of SOS to warming temperatures because of the increasing requirement for forcing temperatures with decreased chilling (Fu et al., 2015; Laube et al., 2014). In contrast, vegetation at high latitudes would still experience sufficient chilling due to the colder and longer winter, and the increasing temperature might even increase its exposure to chilling in regions where temperature was typically below the base temperature for chilling accumulation (Vandvik et al., 2018). Experimental studies comparing phenology in regions at lower and higher elevations have reported similar findings (Gusewell et al., 2017; Vitasse et al., 2018). Large‐scale manipulative field or virtual numerical experiments are still required to further test these hypotheses, because the scenarios designed to explain the faster decrease in ST were less realistic. It should also be noted that our analysis assumes that model bias in the historical period is representative of model performance in the future, and model weights determined under current conditions are preserved for the future. This assumption needs to be tested a wide range of plant species and climatic conditions, because the underlying mechanisms for spring phenology are still not entirely elucidated. For example, photoperiod might influence the spring phe- nology of some species more strongly under a warmer climate (Körner & Basler, 2010; Way & Montgomery, 2015). In this study, we applied four empirical models, involving chilling, forcing, and photoperiod effects, to account for changes in SOS and observed an increased synchrony in spring phenology. Nonetheless, more efforts should also be placed on experimental studies focused on revealing mechanisms behind the pheno- logical processes (e.g., temperature, photoperiod, and precipitation controls) and improving the predictabil- ity of spring phenology models (e.g., Figures 1e and 1f). 4. Conclusions We report overall advancing, but heterogeneously distributed, trends in spring phenology across the NH between 1990–2009 and 2080–2099 (under four climatic scenarios). The synchrony of spring phenology is expected to increase (i.e., a decrease in spatial variation) across the NH in the future, because projected advances in SOS for 2080–2099 are larger for regions with later SOSs during 1990–2009 (e.g., higher lati- tudes). This phenomenon is due mainly to the larger increase in spring temperature at high latitudes, which even compensates for the influence of the generally lower ST at high versus low latitudes. Besides, ST under LIU ET AL. 1648 Geophysical Research Letters 10.1029/2018GL081370 the warmer climatic scenarios decreased more at low latitudes than at high latitudes, probably due to the shorter day length and/or loss of winter chilling, thus amplifying the phenological synchrony as climatic warming proceeds. Changes in the synchrony of spring phenology may result in synchronous change of land‐surface physical conditions (e.g., canopy conductance, albedo, and surface aerodynamic roughness) over large areas (Peñuelas & Filella, 2009; Richardson et al., 2013), and thereby surface water fluxes, energy balance, and atmospheric circulation during spring (Peñuelas & Filella, 2009; Stéfanon et al., 2012). To date, most phenological modules in dynamic global vegetation models relied on growing degree days (Sitch et al., 2003; Thornton et al., 2002) or carbon allocation (Ostle et al., 2009) and yielded poor predictions in spring phenology (Richardson et al., 2012). Therefore, updating the bioclimatic gradient of spring phenology (Hopkins, 1920, 1938) across the latitude due to climate change and improving the representation of pheno- logical synchrony in dynamic global vegetation models could enhance our understanding and predictive capacity of how northern ecosystems will respond to a future warmer climate. Data Availability All data sets and materials used in this study are freely available online, as described in Table S3. Acknowledgments References This study was supported by the National Natural Science Foundation of Caffarra, A., Donnelly, A., & Chuine, I. (2011). Modelling the timing of Betula pubescens budburst. II. Integrating complex effects of China (41530528), National Youth Top‐ photoperiod into process‐based models. Climate Research, 46(2), 159–170. https://doi.org/10.3354/cr00983 notch Talent Support Program in Caffarra, A., Donnelly, A., Chuine, I., & Jones, M. B. (2011). Modelling the timing of Betula pubescens budburst. I. Temperature and China, and the 111 Project (B14001). photoperiod: A conceptual model. Climate Research, 46(2), 147–157. https://doi.org/10.3354/cr00980 Cannell, M., & Smith, R. (1983). Thermal time, chill days and prediction of budburst in Picea sitchensis. Journal of Applied Ecology, 20(3), 951–963. https://doi.org/10.2307/2403139 Chuine, I. (2000). A unified model for budburst of trees. Journal of Theoretical Biology, 207(3), 337–347. https://doi.org/10.1006/ jtbi.2000.2178 Chuine, I., & Régnière, J. (2017). Process‐based models of phenology for plants and animals. Annual Review of Ecology, Evolution, and Systematics, 48(1), 159–182. https://doi.org/10.1146/annurev‐ecolsys‐110316‐022706 Fu, Y. H., Liu, Y., De Boeck, H. J., Menzel, A., Nijs, I., Peaucelle, M., et al. (2016). Three times greater weight of daytime than of night‐time temperature on leaf unfolding phenology in temperate trees. New Phytologist, 212(3), 590–597. https://doi.org/10.1111/nph.14073 Fu, Y. H., Zhao, H., Piao, S., Peaucelle, M., Peng, S., Zhou, G., et al. (2015). Declining global warming effects on the phenology of spring leaf unfolding. Nature, 526(7571), 104–107. https://doi.org/10.1038/nature15402 Gusewell, S., Furrer, R., Gehrig, R., & Pietragalla, B. (2017). Changes in temperature sensitivity of spring phenology with recent climate warming in Switzerland are related to shifts of the preseason. Global Change Biology, 23(12), 5189–5202. https://doi.org/10.1111/ gcb.13781 Hänninen, H. (1990). Modelling bud dormancy release in trees from cool and temperate regions. Acta Forestalia Fennica, (213), 1–47. Hopkins, A. D. (1920). The bioclimatic law. Journal of the Washington Academy of Sciences, 10(2), 34–40. Hopkins, A. D. (1938). Bioclimatics: A science of life and climate relations. US Department of Agriculture. Jakubauskas, M. E., Legates, D. R., & Kastens, J. H. (2001). Harmonic analysis of time‐series AVHRR NDVI data. Photogrammetric Engineering and Remote Sensing, 67(4), 461–470. Jeong, S. J., Medvigy, D., Shevliakova, E., & Malyshev, S. (2013). Predicting changes in temperate forest budburst using continental‐scale observations and models. Geophysical Research Letters, 40, 359–364. https://doi.org/10.1029/2012Gl054431 Julien, Y., & Sobrino, J. (2009). Global land surface phenology trends from GIMMS database. International Journal of Remote Sensing, 30(13), 3495–3513. https://doi.org/10.1080/01431160802562255 Keenan, T. F., Gray, J., Friedl, M. A., Toomey, M., Bohrer, G., Hollinger, D. Y., et al. (2014). Net carbon uptake has increased through warming‐induced changes in temperate forest phenology. Nature Climate Change, 4(7), 598–604. https://doi.org/10.1038/ nclimate2253 Körner, C., & Basler, D. (2010). Phenology under global warming. Science, 327(5972), 1461–1462. https://doi.org/10.1126/science.1186473 Kramer, K. (1994). Selecting a model to predict the onset of growth of Fagus sylvatica. Journal of Applied Ecology, 31(1), 172–181. https:// doi.org/10.2307/2404609 Laube, J., Sparks, T. H., Estrella, N., Hofler, J., Ankerst, D. P., & Menzel, A. (2014). Chilling outweighs photoperiod in preventing preco- cious spring development. Global Change Biology, 20(1), 170–182. https://doi.org/10.1111/gcb.12360 Liu, Q., Fu, Y. H., Liu, Y., Janssens, I. A., & Piao, S. (2018). Simulating the onset of spring vegetation growth across the northern hemi- sphere. Global Change Biology, 24(3), 1342–1356. https://doi.org/10.1111/gcb.13954 Liu, Q., Fu, Y. H., Zeng, Z., Huang, M., Li, X., & Piao, S. (2016). Temperature, precipitation, and insolation effects on autumn vegetation phenology in temperate China. Global Change Biology, 22(2), 644–655. https://doi.org/10.1111/gcb.13081 Liu, Q., Fu, Y. H., Zhu, Z., Liu, Y., Liu, Z., Huang, M., Janssens, I. A., et al. (2016). Delayed autumn phenology in the northern hemisphere is related to change in both climate and spring phenology. Global Change Biology, 22(11), 3702–3711. https://doi.org/10.1111/gcb.13311 MacBean, N., Maignan, F., Peylin, P., Bacour, C., Bréon, F. M., & Ciais, P. (2015). Using satellite data to improve the leaf phenology of a global terrestrial biosphere model. Biogeosciences, 12(23), 7185–7208. https://doi.org/10.5194/bg‐12‐7185‐2015 Martinez‐Cantin, R. (2014). BayesOpt: A Bayesian optimization library for nonlinear optimization, experimental design and bandits. Journal of Machine Learning Research, 15(1), 3735–3739. Melaas, E. K., Friedl, M. A., & Richardson, A. D. (2015). Multi‐scale modeling of spring phenology across deciduous forests in the eastern United States. Global Change Biology, 22(2), 792–805. Menzel, A., Sparks, T. H., Estrella, N., Koch, E., Aasa, A., Ahas, R., et al. (2006). European phenological response to climate change matches the warming pattern. Global Change Biology, 12(10), 1969–1976. https://doi.org/10.1111/j.1365‐2486.2006.01193.x LIU ET AL. 1649 Geophysical Research Letters 10.1029/2018GL081370 Mitchell, T. D., & Jones, P. D. (2005). An improved method of constructing a database of monthly climate observations and associated high‐ resolution grids. International Journal of Climatology, 25(6), 693–712. https://doi.org/10.1002/joc.1181 Murray, M., Cannell, M., & Smith, R. (1989). Date of budburst of fifteen tree species in Britain following climatic warming. Journal of Applied Ecology, 26(2), 693–700. https://doi.org/10.2307/2404093 New, M., Hulme, M., & Jones, P. (2000). Representing twentieth‐century space‐time climate variability. Part II: Development of 1901–96 monthly grids of terrestrial surface climate. Journal of Climate, 13(13), 2217–2238. https://doi.org/10.1175/1520‐0442(2000)013<2217: RTCSTC>2.0.CO;2 Ostle, N. J., Smith, P., Fisher, R., Woodward, F. I., Fisher, J. B., Smith, J. U., et al. (2009). Integrating plant–soil interactions into global carbon cycle models. Journal of Ecology, 97(5), 851–863. https://doi.org/10.1111/j.1365‐2745.2009.01547.x Peñuelas, J., & Filella, I. (2001). Responses to a warming world. Science, 294(5543), 793–795. https://doi.org/10.1126/science.1066860 Peñuelas, J., & Filella, I. (2009). Phenology feedbacks on climate change. Science, 324(5929), 887–888. https://doi.org/10.1126/ science.1173004 Piao, S., Fang, J., Zhou, L., Ciais, P., & Zhu, B. (2006). Variations in satellite‐derived phenology in China's temperate vegetation. Global Change Biology, 12(4), 672–685. https://doi.org/10.1111/j.1365‐2486.2006.01123.x Piao, S., Friedlingstein, P., Ciais, P., Viovy, N., & Demarty, J. (2007). Growing season extension and its impact on terrestrial carbon cycle in the northern hemisphere over the past 2 decades. Global Biogeochemical Cycles, 21, GB3018. https://doi.org/10.1029/2006GB002888 Piao, S., Liu, Z., Wang, T., Peng, S., Ciais, P., Huang, M., et al. (2017). Weakening temperature control on the interannual variations of spring carbon uptake across northern lands. Nature Climate Change, 7(5), 359–363. https://doi.org/10.1038/nclimate3277 Piao, S., Tan, J., Chen, A., Fu, Y. H., Ciais, P., Liu, Q., et al. (2015). Leaf onset in the northern hemisphere triggered by daytime temperature. Nature Communications, 6(1), 6911. https://doi.org/10.1038/ncomms7911 Pinzon, J. E., & Tucker, C. J. (2014). A non‐stationary 1981–2012 AVHRR NDVI3g time series. Remote Sensing, 6(8), 6929–6960. https://doi. org/10.3390/rs6086929 Raftery, A. E., Gneiting, T., Balabdaoui, F., & Polakowski, M. (2005). Using Bayesian model averaging to calibrate forecast ensembles. Monthly Weather Review, 133(5), 1155–1174. https://doi.org/10.1175/MWR2906.1 Raftery, A. E., & Zheng, Y. (2003). Discussion: Performance of Bayesian model averaging. Journal of the American Statistical Association, 98(464), 931–938. https://doi.org/10.1198/016214503000000891 Richardson, A. D., Anderson, R. S., Arain, M. A., Barr, A. G., Bohrer, G., Chen, G., et al. (2012). Terrestrial biosphere models need better representation of vegetation phenology: Results from the North American carbon program site synthesis. Global Change Biology, 18(2), 566–584. https://doi.org/10.1111/j.1365‐2486.2011.02562.x Richardson, A. D., Black, T. A., Ciais, P., Delbart, N., Friedl, M. A., Gobron, N., et al. (2010). Influence of spring and autumn phenological transitions on forest ecosystem productivity. Philosophical Transactions of the Royal Society, B: Biological Sciences, 365(1555), 3227–3246. https://doi.org/10.1098/rstb.2010.0102 Richardson, A. D., Keenan, T. F., Migliavacca, M., Ryu, Y., Sonnentag, O., & Toomey, M. (2013). Climate change, phenology, and phe- nological control of vegetation feedbacks to the climate system. Agricultural and Forest Meteorology, 169(2013), 156–173. https://doi.org/ 10.1016/j.agrformet.2012.09.012 Schulzweida, U. (2013). Climate data operators: A large tool set for working on climate and NWP model data. Retrieved from http://code. zmaw.de/projects/cdo Schwartz, M. D., Ahas, R., & Aasa, A. (2006). Onset of spring starting earlier across the northern hemisphere. Global Change Biology, 12(2), 343–351. https://doi.org/10.1111/j.1365‐2486.2005.01097.x Schwartz, M. D., & Hanes, J. M. (2010). Continental‐scale phenology: Warming and chilling. International Journal of Climatology, 30(11), 1595–1598. https://doi.org/10.1002/joc.2014 Sitch, S., Smith, B., Prentice, I. C., Arneth, A., Bondeau, A., Cramer, W., et al. (2003). Evaluation of ecosystem dynamics, plant geography and terrestrial carbon cycling in the LPJ dynamic global vegetation model. Global Change Biology, 9(2), 161–185. https://doi.org/ 10.1046/j.1365‐2486.2003.00569.x Stéfanon, M., Drobinski, P., D'Andrea, F., & Noblet‐Ducoudré, N. (2012). Effects of interactive vegetation phenology on the 2003 summer heat waves. Journal of Geophysical Research, 117, D24103. https://doi.org/10.1029/2012JD018187 Thornton, P. E., Law, B. E., Gholz, H. L., Clark, K. L., Falge, E., Ellsworth, D. S., et al. (2002). Modeling and measuring the effects of dis- turbance history and climate on carbon and water budgets in evergreen needleleaf forests. Agricultural and Forest Meteorology, 113(1‐4), 185–222. https://doi.org/10.1016/S0168‐1923(02)00108‐9 Vandvik, V., Halbritter, A. H., & Telford, R. J. (2018). Greening up the mountain. Proceedings of the National Academy of Sciences of the United States of America, 115(5), 833–835. https://doi.org/10.1073/pnas.1721285115 Vitasse, Y., Signarbieux, C., & Fu, Y. H. (2018). Global warming leads to more uniform spring phenology across elevations. Proceedings of the National Academy of Sciences of the United States of America, 115(5), 1004–1008. https://doi.org/10.1073/pnas.1717342115 Vrugt, J. A., & Robinson, B. A. (2007). Treatment of uncertainty using ensemble methods: Comparison of sequential data assimilation and Bayesian model averaging. Water Resources Research, 43, W01411. https://doi.org/10.1029/2005WR004838 Wang, C., Tang, Y., & Chen, J. (2016). Plant phenological synchrony increases under rapid within‐spring warming. Scientific Reports, 6(1), 25460. https://doi.org/10.1038/srep25460 Wang, T., Ottlé, C., Peng, S., Janssens, I. A., Lin, X., Poulter, B., et al. (2014). The influence of local spring temperature variance on tem- perature sensitivity of spring phenology. Global Change Biology, 20(5), 1473–1480. https://doi.org/10.1111/gcb.12509 Way, D. A., & Montgomery, R. A. (2015). Photoperiod constraints on tree phenology, performance and migration in a warming world. Plant, Cell & Environment, 38(9), 1725–1736. https://doi.org/10.1111/pce.12431 Wolkovich, E. M., Cook, B., Allen, J., Crimmins, T., Betancourt, J., Travers, S., et al. (2012). Warming experiments underpredict plant phenological responses to climate change. Nature, 485(7399), 494–497. https://doi.org/10.1038/nature11014 Zhang, X., Friedl, M. A., Schaaf, C. B., Strahler, A. H., Hodges, J. C., Gao, F., et al. (2003). Monitoring vegetation phenology using MODIS. Remote Sensing of Environment, 84(3), 471–475. https://doi.org/10.1016/S0034‐4257(02)00135‐9 Zohner, C. M., Benito, B. M., Svenning, J.‐C., & Renner, S. S. (2016). Day length unlikely to constrain climate‐driven shifts in leaf‐out times of northern woody plants. Nature Climate Change, 6(12), 1120–1123. https://doi.org/10.1038/nclimate3138 LIU ET AL. 1650