This is the accepted version of the journal article: Gao, Shan. «An earlier start of the thermal growing season enhances tree growth in cold humid areas but not in dry areas». Nature Ecology & Evolution, Published online February 2022. DOI 10.1038/s41559-022-01668-4 This version is available at https://ddd.uab.cat/record/257156 under the terms of the license 1 An earlier start of the thermal growing season enhances tree growth in 2 cold humid areas but not in dry areas 3 Shan Gao1, Eryuan Liang1*, Ruishun Liu1, Flurin Babst2,3, J. Julio Camarero4, Yongshuo H. Fu5, Shilong 4 Piao1,6, Sergio Rossi7,8, Miaogen Shen9, Tao Wang1, Josep Peñuelas 10,11 5 Affiliations: 1 6 State Key Laboratory of Tibetan Plateau Earth System, Resources and Environment (TPESRE), Institute 7 of Tibetan Plateau Research, Chinese Academy of Sciences, Beijing 100101, China 2 8 School of Natural Resources and the Environment, University of Arizona, Tucson, USA 3 9 Laboratory of Tree-Ring Research, University of Arizona, 1215 E. Lowell St., Tucson, AZ 85721, USA 4 10 Instituto Pirenaico de Ecología (IPE-CSIC), 50059 Zaragoza, Spain 5 11 College of Water Sciences, Beijing Normal University, Beijing 100875, China 6 12 Sino-French Institute for Earth System Science, College of Urban and Environmental Sciences, Peking 13 University, Beijing 100871, China 7 14 Département des Sciences Fondamentales, Université du Québec à Chicoutimi, Chicoutimi, Quebec, 15 Canada 8 16 Key Laboratory of Vegetation Restoration and Management of Degraded Ecosystems, Guangdong 17 Provincial Key Laboratory of Applied Botany, South China Botanical Garden, Chinese Academy of 18 Sciences, Guangzhou 510650, China 9 19 State Key Laboratory of Earth Surface Processes and Resource Ecology, Faculty of Geographical Science, 20 Beijing Normal University, Beijing 100875, China 10 21 CREAF, Cerdanyola del Valles, Barcelona 08193, Catalonia, Spain 11 22 CSIC, Global Ecology Unit CREAF-CSIC-UAB, Barcelona 08193, Catalonia, Spain 23 * Correspondence: Eryuan Liang. Email: liangey@itpcas.ac.cn 24 1 25 Abstract 26 Climatic warming alters the onset, duration and cessation of the vegetative season. While prior studies have 27 shown a tight link between thermal conditions and leaf phenology, less is known about the impacts of 28 phenological changes on tree growth. Here, we assessed the relationships between the start of the thermal 29 growing season (TSOS) and tree growth across the extratropical Northern Hemisphere using 3451 tree-ring 30 chronologies and daily climatic data for 1948-2014. An earlier TSOS promoted growth in regions with high 31 ratios of precipitation to temperature but limited growth in cold dry regions. Path analyses indicated that an 32 earlier TSOS enhanced growth primarily by alleviating thermal limitations on wood formation in boreal 33 forests and by lengthening the period of growth in temperate and Mediterranean forests. Semi-arid and dry 34 subalpine forests, however, did not benefit from an earlier onset of growth and a longer growing season, 35 presumably due to associated water loss and/or more frequent early spring frosts. These emergent patterns 36 of how climatic impacts on wood phenology affect tree growth at regional to hemispheric scales hint at 37 how future phenological changes may affect the carbon sequestration capacity of extratropical forest 38 ecosystems. 39 2 40 Main text 41 Introduction 42 An unprecedented increase in temperature has been recorded in recent decades, with higher rates of 43 warming outside than during the main growing season1. Such warming causes large changes in the timing, 44 duration and thermal conditions of the vegetative season in extratropical terrestrial biomes 2-5. The start of 45 the thermal growing season (TSOS) directly influences vegetation phenology and its advance closely 46 matches the interannual variability of spring green-up6-12. These phenological shifts influence the capacity 47 of the biosphere to take up carbon13-15 and affect the exchange of energy between the atmosphere and the 48 biosphere13, 16. It remains to be answered, whether the shifts in plant phenology would result in a negative 49 feedback to warming and an increase carbon uptake or alternatively exhibit additional ecological stress13. 50 Solving this issue may help reduce uncertainties associated with the forecasting and modeling of forest 51 productivity and global carbon cycling. 52 Satellite observations of forested areas provide evidence that recent climate change has shifted foliar 53 phenology and photosynthetic seasonality17. Ninety-five percent of the global land surface underwent 54 substantial changes in foliar phenology between 1980 and 2012, including changes in the timing of 55 phenological cycles and the vigor of vegetative activity7. In addition to the direct response of an advanced 56 foliar flush to an earlier TSOS18, peak photosynthesis occurs earlier and culminates higher in forests of the 57 extratropical Northern Hemisphere8, 19, 20. These phenological shifts may be strongly correlated with the 58 thermal conditions in spring, because satellite data indicate that the rate of phenological change slowed 59 under the warming hiatus of 1998-201221. 60 Changes in the timing and vigor of vegetation activity further affect when and how carbon is 61 assimilated by terrestrial ecosystems. The spring shifts of vegetation activity may increase ecosystem 62 productivity due to an earlier start of carbon uptake 19 and longer vegetative seasons with more vigorous 63 photosynthetic activity15, 22. Widespread and contrasting responses of productivity to shifts in foliar 64 phenology, however, have been detected across northern terrestrial ecosystems. The beneficial effects of 65 spring warmth on growing-season productivity can be offset by water stress due to higher 66 evapotranspiration in the summer23-25 and by increasing carbon losses due to higher respiration in the 3 67 autumn26. A long-term study of biomass also found that alpine plants grew earlier and faster, but the 68 increase in spring productivity was offset by a reduction in autumnal productivity due to increased water 69 stress27. Thus, any attempt to explain climatic influences on terrestrial carbon uptake solely based on 70 studies of shifting foliar phenology and photosynthetic seasonality remains challenging. 71 The carbon residence time in tree stems is much longer than in foliage, making the former a major 72 contributor to the long-term carbon sink in forests28. Tree radial growth represents the annual accumulation 73 and fixation of carbohydrates in the stem. Importantly, wood phenology, mainly in cold areas, is closely 74 related to temperature29. Wood formation in conifers begins when specific critical temperatures and 75 photoperiods are reached30-32. In addition to temperature, the length of the growing season determines the 76 available period for developing functional xylem through cell maturation and lignification, especially in 77 cold areas31, 33. In drier ecosystems, water availability for roots, rather than rainfall per se, is another 78 important driver of cambial reactivation34. Temperature and the availability and demand of water also co- 79 determine the rate of growth 35-37. In addition to climate, the phenology of wood formation is also 80 associated with physiological trade-offs with bud and foliar phenology, because phytohormones produced 81 in developing buds and foliage regulate the rate of cambial division29, 38 and can lead to changes in 82 priorities for allocating carbon within a tree. 83 Fundamental but still unresolved questions are thus whether and how the advance of the thermal 84 growing season in spring influences annual tree growth (and biomass accumulation) across environmental 85 gradients. We addressed these questions by investigating the influence of TSOS on tree radial growth 86 (represented by a ring-width index, RWI) in the extratropical Northern Hemisphere and by identifying the 87 dominant mechanisms controlling the relationship between TSOS and growth for several regions with 88 contrasting climates (northern Asia, northern Europe, Central Europe, the Mediterranean region, the 89 western and eastern coast of the US, and the Colorado and Tibetan Plateaus) with different forest biomes 90 (boreal, Mediterranean, temperate, semi-arid and dry subalpine forests). We tested the hypothesis that the 91 shift of TSOS influences tree growth by changing its timing, duration, and rate according to the influence 92 of climate on the processes of xylem formation35, 39, 40. We assumed that a shift of TSOS would lengthen 93 the growing season by modifying growing degree days and the availability of soil moisture and that such 4 94 phenological changes could affect growth through various ecophysiological mechanisms depending on the 95 ambient climatic conditions. 96 97 Results 98 Response of tree growth to TSOS changes 99 Most areas in the extratropical Northern Hemisphere had trends toward an earlier TSOS between 1948 and 100 2016. Correlation results show that 36.5% of these areas exhibited significant (p < 0.05 in a two-tailed 101 Student t-test) and 49.2% at least marginally significant (p < 0.1) advancing trends (Extended Data Fig. 1). 102 11.4% of the RWI chronologies had significant (p < 0.05, t-test) and 18.0% at least marginally significant 103 (p < 0.1) simple correlations with TSOS, and 7.7% had significant and 13.6% at least marginally significant 104 partial correlations (Extended Data Fig. 1). The correlations revealed distinct spatial patterns after gridding 105 onto a 2°×2° raster (Fig. 1). The area with negative TSOS-RWI correlations was generally larger than the 106 area with positive correlations (56% vs 33% in the simple correlation analyses and 46% vs 36% in the 107 partial correlation analyses; see histograms in Fig. 1C, D). Negative correlations dominated at high 108 latitudes (>60°N), central Europe, eastern and western coastal North America, indicating that the advancing 109 TSOS could benefit tree growth in these regions. Correlations were mainly positive for the Colorado and 110 Tibetan Plateaus, indicating that an advance in TSOS could reduce growth in these regions. Similar patterns 111 were found for both the simple and partial correlations. 112 We calculated the 30-year (1969-1998) mean growing degree days (GST) and the 30-year mean 113 growing-season precipitation (GSP) to compare the ambient climatic characteristics of the RWI sites with 114 contrasting responses to changes in TSOS. GST for the RWI chronologies with significant negative TSOS 115 correlations was distinctly lower than for RWI chronologies with positive correlations (Fig. 1E, F). Linear 116 regression analyses of GST and GSP further indicated a higher regression coefficient for RWI chronologies 117 with significant negative correlations than for RWI chronologies with positive correlations. These results 118 suggest that the advance in TSOS would likely benefit tree growth in cold areas with a lower number of 119 GST and/or a higher GSP:GST ratio. 120 Relationships between TSOS and RWI in climatically distinct regions 5 121 We conducted path analyses to decompose the effect of TSOS on RWI, so that the magnitude of the 122 underlying processes responsible for the emergent correlations could be compared. TSOS-RWI 123 relationships of each tree-ring chronology involved in path analyses for the eight selected regions (northern 124 Asia, northern Europe, Central Europe. the Mediterranean region, the western and eastern coast of the US, 125 the Colorado Plateau and the Tibetan Plateau) are shown in Extended Data Fig. 2. The path diagram was 126 reliable in all regions; the specific model fits for each region are presented in Supplementary Table 1. The 127 path effects showed distinct responses between regions (Fig. 2). TSOS had a negative total effect on RWI 128 (i.e., higher RWI under advanced TSOS) for boreal forests in northern Asia and Europe, temperate forests 129 in central Europe and the east coast of the US, and for forests in the Mediterranean region and along the 130 west coast of the US. In boreal forests, the most pronounced pathway affecting RWI is the “growing degree 131 days (GDD) effect” (i.e., the path effect through the “TSOS—GDD—RWI” path), suggesting that an 132 advance in TSOS increases tree growth mainly through the increase in GDD. TSOS also had a negative 133 total effect on RWI in temperate and seasonally dry Mediterranean forests, but the path effect was stronger 134 through the length of the thermal growing season (GSL) than GDD, suggesting that the beneficial effect of 135 an advanced TSOS on growth was due to the extension of the thermal growing season, without a clear 136 effect of drought due to reduced SM. In contrast, TSOS for semi-arid forests on the Colorado Plateau had a 137 strong positive total effect on RWI (i.e., lower RWI under advanced TSOS). The positive effect through 138 GSL combined with the effect through GDD and SM, suggests that an advance in TSOS could reduce tree 139 growth due to the longer growing season, the increase in GDD, and the decrease in SM (soil drought) 140 caused by the increased GDD. TSOS for dry subalpine forests on the Tibetan Plateau also had a positive 141 effect on RWI, with the main path through changes to GSL, suggesting that the unfavorable situation of an 142 advanced TSOS for growth was mainly caused by the lengthening of the growing season. 143 144 Discussion 145 Our study has demonstrated that spatiotemporal shifts in TSOS can significantly and variably affect 146 tree growth in the extratropical Northern Hemisphere. This conclusion is supported by our current 147 understanding of the physiological mechanisms that underlie wood formation. As shown by xylogenetic 6 148 studies, wood formation involves sequential processes of cambial cell division, cell enlargement and cell- 149 wall thickening41. The onset of wood formation is the main factor that directly or indirectly triggers all 150 subsequent phases of xylem maturation39. Small changes in the period of cell division can lead to 151 substantial increases in xylem cell production and growth40. The rate of increase in xylem size peaks when 152 the cambium is dividing vigorously and most cells are undergoing the enlarging phase. These physiological 153 processes culminate at the end of spring and slow down in late summer and autumn when the tree ring is 154 almost fully formed38, 41, 42. Therefore, tree growth would be enhanced by an earlier onset and also by 155 higher growth rates during the peak growing season in cold climates. Recent xylogenetic studies have also 156 demonstrated that a longer growing season induced by its earlier start will not benefit xylem formation in 157 trees located in drought-prone environments. Instead, warming induced drought could limit carbon 158 sequestration by reducing the rate of cell production35, 37. Based on these physiological mechanisms, we 159 assumed that growth changes caused by shifts in TSOS can be inferred from tree-ring data. 160 Our results revealed a clear spatial pattern in the response of tree growth (RWI) to TSOS (Fig. 1). 161 Areas with beneficial effects of TSOS on RWI (i.e., negative correlation) are generally located in high- 162 latitude (above 60°N), Europe, as well as in eastern and western coastal North America. These cold and 163 humid regions have no or minimal water limitation during the growing season. This spatial distribution 164 generally agrees with the distribution of areas that exhibit a clear advance in the timing of foliar onset and 165 peak photosynthetic activity11, 19, 43. This importantly suggests that enhanced carbon uptake induced by the 166 advance of TSOS promoted the production and accumulation of photosynthates and thus increased the 167 availability of resources for tree growth. Although a warmer autumn may offset the increased productivity 168 during spring due to a disproportionally larger increase in respiration compared with photosynthesis23, 24, 26 169 and can additionally cause earlier foliar senescence44, this is likely to affect carbon stored in pools with a 170 faster turnover rate such as shoots and leaves. However, the effect of autumnal warming was marginal for 171 “slow carbon”, i.e., that sequestered in the wood, compared to this canopy activity. The regions with 172 negative effects of TSOS on growth (i.e., positive correlation) were mainly located on the Colorado Plateau 173 and the Tibetan Plateau, corresponding to cold-dry conditions where forests are typically co-limited by the 174 availability of soil water and nutrients. Radial growth is more sensitive to low temperatures or drought than 7 175 photosynthesis45 and may cease long before carbon uptake in response to water shortage 46. Warming during 176 the growing season in these regions may intensify drought, inhibit woody tissue formation37, 45, and reverse 177 the positive effects of temperature on growth even in cold areas47. An extended growing season may also 178 increase the risk of tree expose to low temperature events such as spring frosts48. These effects are possible 179 causes of reduced tree growth and constrain carbon accumulation in the wood. 180 The shift in the timing of TSOS may have affected GSL, GDD and SM. The change of GSL would 181 extend the time when cambial activity and wood formation are possible. In contrast, the change of GDD 182 and SM would affect growth rates35, 49. All these factors can interact to modulate tree growth and the 183 resulting sequestration of carbon. Decomposing the effect of TSOS on radial growth in different forest 184 biomes – as we have done in this study – can therefore help advance our understanding of the effects of 185 TSOS on carbon sequestration and wood formation, and pave the way for improved forecasting of forest 186 carbon cycling. 187 The advance of TSOS benefited tree growth in the boreal forests of northern Asia and Europe, and the 188 path analyses indicated that the “GDD effect” was the primary responsible pathway (Fig. 2A). Our results 189 are consistent with previous studies of canopy processes reporting that an increase in vegetation greenness 190 was more pronounced across boreal ecosystems than in other regions50, which was mainly due to the 191 alleviation of the limitation of cold temperatures on vegetation growth under climatic warming 51, 52. The 192 advance of TSOS also benefited tree growth in temperate forests of central Europe and the east coast of the 193 US, and Mediterranean forests of the Mediterranean region and the west coast of the US. The “GSL effect” 194 was the primary path effect in those areas (Fig. 2B). In central Europe and along the east coast of the US, 195 precipitation is adequate to abundant, and the summers are generally warm and humid. A lengthened GSL 196 extends the growth duration and favors tree growth there. The Mediterranean climate is characterized by 197 dry and hot summers, with optimal conditions for vegetation growth occurring during the cool and rainy 198 springs and autumns, often leading to a bimodal pattern of growth with a temporary cessation of growth in 199 summer53. Photoperiods are longer in spring than in autumn, and an earlier reactivation of the cambium 200 after winter dormancy can harness this period for increasing production. A lengthening of the growing 201 season through the advance of TSOS may therefore benefit tree growth if spring droughts are not persistent 8 202 or severe. SM at the beginning of the growing season is also a major factor affecting tree radial growth54, 55, 203 but the advanced TSOS in our study may have had a limited effect on RWI via the “SM effects” (i.e. the 204 path effect through the “TSOS—GSL—GDD—SM—RWI” and “TSOS—SM—RWI” paths) in these 205 regions. The thermal conditions at the beginning of the growing season were mild and may not have caused 206 a severe loss of soil water through evaporation, but an advanced TSOS may accelerate snow melt and 207 increase the availability of soil water56. These remaining uncertainties need to be comprehensively 208 addressed in future studies. The path effects of northern and central Europe are small compared with other 209 regions, perhaps due to the difference in the distance from the ocean and the complexity in topography and 210 species composition, which also need to be studied with more detail in future work. 211 The advance of TSOS negatively affected growth in semi-arid forests on the Colorado Plateau and dry 212 subalpine forests on the Tibetan Plateau. Path analysis further indicated that growth reductions under 213 advanced TSOS were primarily caused by the “GSL effect” (Fig. 2). This result was not consistent with our 214 original hypothesis in the path diagram that an extended GSL would enhance tree radial growth 215 (Supplementary Fig. 1) and may involve more complex mechanisms. Extended GSL in these regions, 216 combined with higher heat accumulation (“GDD effect”) and/or evapotranspiration of soil water (i.e., the 217 path effect through the “TSOS—GSL—GDD—SM—RWI”), may induce both atmospheric and soil 218 droughts. Droughts will trigger stomatal closure, increase water tension in the xylem, and deplete the 219 contents of nonstructural carbohydrates in trees57-59, thus reducing the rate of wood production35. Forests in 220 these regions also suffered more from frost days than those in high latitude regions (see Supplementary Fig. 221 2). Earlier TSOS may increase tree exposure to spring frost and thereby reduce tree growth48, 60. The 222 specific mechanisms underlying these processes need to be addressed in further experimental studies. 223 Uncertainties in our analyses were mainly introduced by three sources: the spatial representativeness 224 of the tree-ring series, the determination of the TSOS thresholds and the establishment of the path diagram. 225 The ITRDB data set contains a large imbalance in the spatial distribution of sites and in its species 226 composition61, 62. Further, the local microenvironment, stand structure, or biotic and abiotic disturbances are 227 often unknown but can also impact tree radial growth and phenological responses10, 63, 64. To mitigate 228 potential biases associated with these caveats, we first gridded the correlation coefficients and then 9 229 displayed the percentage of the direction instead of the magnitude of the correlation coefficients. We were 230 thereby able to extract the dominant spatial patterns of the response of tree growth to shifts in the timing of 231 the thermal growing season. 232 The thermal threshold for growth of 5 °C is widely accepted and used 6, but debatable because the 233 choice of threshold may lead to different conclusions4. A more vegetative based threshold (for instance the 234 threshold from vegetation greenness) is, however, difficult to achieve due to the inconsistency in temporal 235 availability of tree-ring data and satellite-retrieved observations62. Biological evidence suggests that the 236 daily mean temperature threshold for the onset of xylem growth in conifers at high altitudes and in cold 237 climates is 5.6 to 8.0 °C32, 65. The critical threshold of mean air temperature at alpine treelines is about 3.9 238 °C31. With these premises, we assumed the threshold for TSOS and GDD range between 4 to 6 °C while 239 exploring the response of tree growth to TSOS at large-spatial scales and chose to present the results for the 240 5 °C cutoff. Reassuringly, the results of analyses with other cutoffs showed similar patterns, confirming the 241 robustness of the results. 242 The establishment of our path diagram was based on experimental studies; advanced TSOS would 243 extend growth duration (indicated by GSL) and affect growth rates (controlled by GDD and SM), thus 244 influencing annual tree growth. Path analysis is an extension of multiple linear regression. We therefore 245 assumed that the relationships among the variables were mainly linear, which is not always consistent with 246 our current understanding of the complex responses of tree growth to climate 66-68. Encouragingly, the 247 relationships between TSOS and RWI in the eight regions were mostly linear (Extended Data Fig. 2). We 248 therefore considered our use of path analyses to be appropriate. 249 We found that the impact of shifts in the timing and duration of the thermal growing season could be 250 detected in tree rings at regional to hemispheric scales. Our study thus allows for the further exploration of 251 the impact of climatic trends and variability on tree growth. Such information is essential for integrating 252 information regarding the responses of foliage and stems to climate change, and for predicting future 253 vegetation performance. Explaining the influence of plant phenology on carbon sequestration solely based 254 on the perspective of foliar phenology and photosynthesis seasonality (which drive carbon uptake) is 255 insufficient. Low temperatures and drought constrain growth more than photosynthesis45. A carbon sink 10 256 (i.e., wood) oriented view on phenological impacts is therefore essential for predicting carbon sequestration 257 capacity, because wood is the primary long-term carbon storage pool in forests. Wood formation, however, 258 is notoriously difficult to quantify using satellite observations or techniques of eddy covariance41. Our 259 study implies that the analysis of tree rings at regional to global scales could provide new solutions to 260 differentiate between shifts in the turnover of “slow” and “fast” carbon pools under a rapidly changing 261 climate69. 262 In summary, our study provides strong evidence that shifts in TSOS influence tree radial growth in the 263 extratropical Northern Hemisphere. The advance of TSOS is more likely to enhance tree growth in cold 264 humid areas with a higher water:heat ratio, whereas growth in cold dry areas may be reduced. Our results 265 also indicated that the primary path effect of TSOS on growth differed among forest biomes. The beneficial 266 effects in the boreal forests of northern Asia and Europe were mainly due to the alleviation of thermal 267 limitation on wood formation, so that higher growth rates were possible, but the primary beneficial effect in 268 the temperate forests of central Europe and the east coast of the US and Mediterranean forests involved a 269 lengthening of the growing season. The negative effects for semi-arid and dry alpine forests on the cold dry 270 Colorado Plateau and the Tibetan Plateau were primarily due to a longer period of growth, presumably due 271 to associated droughts driven by heat, as well as by an increased likelihood of spring frosts. This study 272 reveals how climate affects tree growth through wood phenology and contributes to improving our ability 273 to predict trends in the capacity of forests to sequester carbon at regional to global scales. 274 275 Methods 276 Experimental design 277 We raised fundamental but still unresolved questions of whether and how the advance of the thermal 278 growing season in spring influences tree growth across environmental gradients. We addressed these 279 questions by investigating the relationships between TSOS and tree radial growth across the extratropical 280 Northern Hemisphere with correlation analyses and by identifying the dominant mechanisms controlling 281 the relationships in path analyses for several regions with contrasting climates. A total of 3451 tree-ring 282 width chronologies and daily climatic data for 1948-2014 were used to conduct these analyses. 11 283 Data 284 Tree-ring width chronologies 285 Raw tree-ring width chronologies from 4219 sites across the extratropical Northern Hemisphere (20-75°N) 286 were selected from the reformatted data set of the International Tree-Ring Data Bank (ITRDB)61, as well as 287 83 sites on the Tibetan Plateau (Supplementary Table 2) from the tree-ring group of the Institute of Tibetan 288 Plateau Research Chinese Academy of Sciences (ITPCAS) (https://doi.org/10.11888/Terre.tpdc.271925). 289 We excluded chronologies shorter than 30 years after 1948 and those where TSOS varied little (i.e. no 290 change of TSOS for >20 years), for a total of 3451 sites retained for further analyses. Of these 291 chronologies, 73.6% (2540) were from evergreen conifers, 9.1% (314) from deciduous conifers (mainly 292 larch), 16.5% (569) from broadleaf species, and 0.2% (7) from shrubs at the boreal treeline. Twenty-one 293 chronologies lacked information about tree species. To transform the tree-ring width data into a ring-width 294 index (RWI) that accentuates the variability of annual to decadal growth, we removed long-term trends 295 caused by aging and increasing trunk diameter by fitting either a negative exponential curve or a cubic 296 smoothing spline (removing 50% of the variance for a period of 67% of series length) to the raw ring-width 297 series using the dplR package (version 1.7.1)70 in R71. Mean site chronologies of RWI after 1948 were 298 calculated using bi-weight robust means. 299 300 Climatic and soil-moisture data 301 Daily grids of mean air temperature and total precipitation for 1948 to 2016 were obtained from the Global 302 Meteorological Forcing Dataset of the Terrestrial Hydrology Research Group at Princeton University 303 (http://hydrology.princeton.edu/data.pgf.php) at a spatial resolution of 0.25°72. Daily soil-moisture content 304 (SM) in the root zone (0-100 cm) was obtained from the NASA Global Land Data Assimilation System 305 Version 2 (GLDAS-2) 306 (https://disc.gsfc.nasa.gov/datasets/GLDAS_CLSM025_D_2.0/summary?keywords=GLDAS2.0) at a 307 resolution of 0.25°. GLDAS-2 is forced entirely with the Princeton meteorological forcing input data and 308 provides a temporally consistent series from 1948 to 2014. 12 309 We extracted the timing and length of the thermal growing season for each year based on daily mean 310 air temperature. TSOS was defined as the first six uninterrupted days with daily mean temperatures >5 °C 311 at mid and high latitudes73. The end of the thermal growing season (TEOS) was defined as the first six 312 uninterrupted days after 1 July with daily mean temperatures <5 °C. GSL was calculated as the time 313 between TSOS and TEOS. 314 Growing degree days (GDD), which represent the effective accumulation of heat for vegetation 315 growth during the growing season, were calculated as the sum of daily mean temperatures >5 °C54: 316 𝐺𝐷𝐷 = ∑𝑇𝐸𝑂𝑆 𝑇𝑆𝑂𝑆 (𝑇𝑖 − 5) 𝑖𝑓 𝑇𝑖 > 5 (1) 317 where 𝑇𝑖 is the mean temperature on day i. 318 Growing-season precipitation was calculated as the sum of daily precipitation during the thermal growing 319 season. Mean SM during the growing season was the average of the daily content in the root zone during 320 the thermal growing season. The 30-year mean GDD (GST) and the 30-year mean growing-season 321 precipitation (GSP) were calculated for 1969-1998. When choosing an aridity metric for our study, we 322 decided to use a simple index that relies only on the most widely measured variables: temperature and 323 precipitation. We favored the GSP:GST ratio (similar to the Selyaninov hydrothermic coefficient 74) over 324 more complex indices because the latter often require input variables that are best measured locally. These 325 include atmospheric or even soil moisture content, which are not ubiquitously available in remote areas to 326 feed the data pipelines that produce global gridded climate products. We thus deemed the GSP:GST ratio to 327 be a robust, reliable and well-established aridity metric for our study. It was used to compare aridity 328 condition during the growing season among tree-ring sites. 329 330 Analyses 331 Correlations 332 We calculated both simple and partial Pearson correlations to explore the effects of TSOS on tree growth 333 for each site of tree-ring chronology. We eliminated the effects of GSL, GDD and SM when calculating 334 partial correlation coefficients between TSOS and RWI. Tree-ring width chronologies are likely co-driven 335 by local site factors such as microclimatic and soil conditions, forest composition and competition in 13 336 closed-canopy stands and a possible mismatch between the site location and the gridded climatic data (e.g. 337 elevation). To reduce the impact of these site-specific factors and identify general spatial pattern in the 338 correlation coefficients, we gridded the correlation coefficients by 2°×2° and displayed the percentage of 339 tree-ring series with positive coefficients within each grid. 340 341 Path analysis 342 Path analysis is an extension of multiple regression analyses used to evaluate causal models by examining 343 the linear relationships between independent and dependent variables75. Path analysis decomposes bivariate 344 correlation coefficients into path coefficients, which represent the relative importance of prespecified 345 hypotheses within the same path diagram. We used the existing information of how TSOS affects RWI (see 346 the Introduction section) to test a path diagram containing four hypothetical associations (Supplementary 347 Fig. 1). First, the advance in TSOS would extend GSL and thereby enhance tree radial growth (represented 348 by RWI). Second, the advance in TSOS would extend GSL and increase GDD, causing a positive change in 349 RWI. Third, the advance in TSOS would extend GSL, increase GDD and lead to a shortage of soil 350 moisture, with negative effects on RWI. Fourth, the change in TSOS could affect SM by accelerating snow 351 melt, by increasing the thawing of permafrost or by changing the proportion of precipitation during the 352 growing season56, thereby promoting tree growth and increasing RWI. 353 We used the “sem” package (version 3.1.9)76 in R to calculate the standardized path coefficients of the 354 preset path diagram. Path effects were then calculated as the product of the standardized path coefficients 355 along each pathway. We compared the bivariate correlation coefficients (i.e. TSOS and RWI) and the total 356 path effects (i.e. the sum of the four path effects) of all 3451 RWI series to determine the fit of the preset 357 path diagram to our data. The relationships between the bivariate correlation coefficients and the total path 358 effects were consistent (Supplementary Fig. 3). 359 We selected eight regions based on the spatial patterns identified by the correlation analyses and 360 climatological consistency to examine the general characteristics of the path effects. The definition of 361 northern Asia and Europe, central Europe, the Mediterranean region and the Tibetan Plateau referred to 362 IPCC climate reference regions77, the west and east coast of the US and the Colorado Plateau referred the 14 363 hydrologic and geographic unit. For the eight regions, general climatic conditions were presented in 364 Supplementary Fig. 4 and forest conditions were described in the supplementary text. Because we aimed to 365 decompose correlations into different processes for the interpretation of underlying mechanisms, only RWI 366 chronologies with at least marginally significant correlations (p < 0.1) were included in the regional path 367 analyses56. Anomalies of climatic variables (i.e., TSOS, GSL, GDD and SM) were calculated for each RWI 368 chronology in reference to its 30-year (1969-1998) mean climate condition. Then we used RWIs and their 369 corresponding climatic anomalies within the same region to conduct the path analysis. All variables were 370 standardized prior to path analyses. Many fitting measures can appraise a path diagram. We measured the 371 adequacy of the fitness of the path diagram in each region using the following criteria: goodness-of-fit 372 index (GFI) ≥ 0.95, comparative fit index (CFI) ≥ 0.90, root mean square error of approximation (RMSEA) 373 ≤ 0.10, nonnormed fit index (NNFI) ≥ 0.92 and standardized root mean square residual (SRMR) ≤ 0.08. 374 The path diagram was considered reliable when three of these five criteria were met78. 375 376 Results validation 377 In order to confirm the robustness of our results, we tested different thresholds of TSOS, as well as of GDD 378 at 4, 4.5, 5.5 and 6 °C, and conducted the full analysis for each of them. The results showed similar pattern 379 and are presented in the Supplementary Table 3. 380 381 Data availability 382 The reformatted data set of the International Tree-Ring Data Bank were obtained from 383 https://doi.org/10.5061/dryad.kh0qh06. Tree-ring width data from the ITPCAS tree-ring group are available 384 from https://doi.org/10.11888/Terre.tpdc.271925. The Global Meteorological Forcing Dataset of the 385 Terrestrial Hydrology Research Group at Princeton University were obtained from 386 http://hydrology.princeton.edu/data.pgf.php. The NASA Global Land Data Assimilation System Version 2 387 were obtained from 388 https://disc.gsfc.nasa.gov/datasets/GLDAS_CLSM025_D_2.0/summary?keywords=GLDAS2.0. 389 15 390 391 Code availability 392 Statistical analysis in this study were performed with publicly available packages in R (version 3.6.2, dplR 393 and sem packages) and Python (version 3.8, scipy package), and the figures were produced using Python 394 (matplotlib, cartopy and seaborn packages). The custom code for the analysis of the data are available from 395 https://doi.org/10.11888/Terre.tpdc.271925. 16 396 397 Acknowledgments 398 We acknowledge all contributors to the International Tree-Ring Databank for providing tree-ring data. This 399 study was supported by the Second Tibetan Plateau Scientific Expedition and Research Program (STEP) 400 (2019QZKK0301), the National Natural Science Foundation of China (41907387, 42030508, 41988101) 401 and the China Postdoctoral Science Foundation (2019M660813). JP was funded by Spanish Government 402 projects PID2019-110521GB-I00, Fundación Ramón Areces project ELEMENTAL-CLIMATE, and 403 Catalan government project SGR2017-1005. 404 405 Author Contributions Statement 406 S.G. and E.L. designed the research, S.G. and R.L. performed the analysis and S.G. drafted the manuscript. 407 E.L., F.B., J.J.C., Y.H.F., S.P., S.R., M.S., T.W. and J.P. contributed ideas, interpreted the results and were 408 involved in the editing and writing of the manuscript. 409 410 Competing Interest Statement 411 The authors declare no conflicts of interest. 412 413 414 17 415 416 Fig. 1 | Responses of tree growth to changes in the onset of the thermal growing season (TSOS) across 417 the extratropical Northern Hemisphere. Spatial patterns of the percentage of tree-ring series (represented 418 by RWI) with a positive simple correlation coefficient (A), partial correlation coefficient (B), significant (p 419 < 0.1) simple correlation coefficient (C) and significant partial correlation coefficient (D) between RWI 420 and TSOS within 2°×2° grids. The number of tree-ring width chronologies considered in each grid are 421 presented in Supplementary Fig. 5. The histograms in panels (A) to (D) present the frequency distributions 18 422 of the percentages. Climatic characteristics of tree-ring sites with a significant simple correlation 423 coefficient (E) and partial correlation coefficient (F) in the space of GST and GSP. The histograms located 424 at the top and right of panels (E) and (F) present the distributions of the tree-ring sites along the GST and 425 GSP gradients. The blue and red kernel density plots and histograms represent tree-ring chronologies with 426 negative and positive correlation coefficients, respectively. The lines in panels (E) and (F) are derived from 427 linear regression, the shown regression equations are all significant (p < 0.001) estimated using the F-test. 428 19 429 430 Fig. 2 | Path diagrams and path effects for northern Asia, northern and central Europe, the 431 Mediterranean region, the west and east coasts of the US, the Colorado Plateau, and the Tibetan 432 Plateau. In the geographic map, dots represent the location of tree-ring chronologies with significant (p < 433 0.1) positive (red dots) and negative (blue dots) simple correlation with TSOS; boxes delineate the eight 434 regions. The numbers in the path diagrams represent the mean and standard error of standardized path 435 coefficients in the regions, asterisks indicate the significance of the path coefficients (p < 0.05) and the 436 colors (negative and positive effects are presented as blue and red arrows, respectively) and widths of the 20 437 arrows represent the signs and magnitudes of the path coefficients, respectively. A, B, C and D in the 438 panels on the right represent the effect of four major paths, TE represents the total effect. The number of 439 tree-ring width chronologies for each region is presented in Supplementary Fig. 6. 21 440 441 Extended Data Fig. 1|Responses of tree growth to changes in TSOS in the extratropical Northern 442 Hemisphere. Spatial distributions of simple correlation coefficients (A), partial correlation coefficients 443 (B), significant (p < 0.1) simple correlation coefficients (C) and significant (p < 0.1) partial correlation 444 coefficients (D) of TSOS and RWI. (E) Areas with significant (p < 0.1, dark blue) and nonsignificant (light 445 blue) trends toward earlier TSOS between 1948 and 2016 in the extratropical Northern Hemisphere. (F) 446 Areas with significant (p < 0.05, blue shaded area) trends toward earlier TSOS overlapping tree-ring 447 chronologies with significant (p < 0.1) simple correlation coefficients of TSOS and RWI. The significance 448 of the correlation analyses is estimated by two-tailed Student’s t-test. This figure was generated using the 449 matplotlib and cartopy package in Python. 450 22 451 23 452 Extended Data Fig. 2 | Scatter plots of TSOS-RWI relationships in different regions. TSOS-RWI 453 relationships of tree-ring chronologies with significant (p < 0.1) simple correlations for northern Asia (A), 454 northern Europe (B), central Europe (C), the Mediterranean region (D), the west coast of the US (E), the 455 east coast of the US (F), the Colorado Plateau (G) and the Tibetan Plateau (H). The predicted mean (solid 456 lines) is bounded by the 95% confidence intervals (shaded areas). This figure was generated using the 457 seaborn package, “lmplot” function in Python. 458 24 459 460 References 461 1. K. E. Trenberth, P. D. Jones, Observations: Surface and atmospheric climate change, in Climate 462 change 2007: the physical science basis: contribution of working group I to the fourth assessment 463 report of the intergovernmental panel on climate change, (Cambridge Univ. Press, 2007), pp. 235–335. 464 2. H. W. Linderholm, Growing season changes in the last century. Agr. Forest Meteorol. 137, 1–14 465 (2006). 466 3. B. Yang, M. He, V. Shishov, I. Tychkov, E. Vaganov, S. Rossi, F.C. Ljungqvist, A. Bräuning, J. 467 Grießinger, New perspective on spring vegetation phenology and global climate change based on 468 Tibetan Plateau tree-ring data. Proc. Natl. Acad. Sci. USA. 114, 6966–6971 (2017). 469 4. M. Shen, Y. Tang, J. Chen, W. Yang, Specification of thermal growing season in temperate China 470 from 1960 to 2009. Clim. Change. 114, 783–798 (2012). 471 5. B. Zhou, P. Zhai, Y. Chen, R. Yu, Projected changes of thermal growing season over Northern Eurasia 472 in a 1.5°C and 2°C warming world. Environ. Res. Lett. 13, 35004 (2018). 473 6. J. Barichivich, K. R. Briffa, T. J. Osborn, T. M. Melvin, J. Caesar, Thermal growing season and 474 timing of biospheric carbon uptake across the Northern Hemisphere. Global Biogeochem. Cycles. 26, 475 B4015 (2012). 476 7. R. Buitenwerf, L. Rose, S. I. Higgins, Three decades of multi-dimensional change in global leaf 477 phenology. Nat. Clim. Change. 5, 364–368 (2015). 478 8. A. Gonsamo, J. M. Chen, Y. W. Ooi, Peak season plant activity shift towards spring is reflected by 479 increasing carbon uptake by extratropical ecosystems. Global Change Biol. 24, 2117-2128 (2018). 480 9. A. Menzel, T. H. Sparks, N. Estrella, E. Koch, A. Aasa, R. Ahas, K. Alm-Kübler, P. Bissolli, O. 481 Braslavská, A. Briede, F. M. Chmielewski, Z. Crepinsek, Y. Curnel, A. Dahl, C. Defila, A. Donnelly, 482 Y. Filella, K. Jatczak, F. Mage, A. Mestre, Ø. Nordli, J. Peñuelas, P. Pirinen, V. Remišová, H. 483 Scheifinger, M. Striz, A. Susnik, A.J.H. Van Vliet, F. Wielgolaski, S. Zach, A. Zust, European 484 phenological response to climate change matches the warming pattern. Global Change Biol. 12, 1969– 485 1976 (2006). 25 486 10. R. A. Montgomery, K. E. Rice, A. Stefanski, R. L. Rich, P. B. Reich, Phenological responses of 487 temperate and boreal trees to warming depend on ambient spring temperatures, leaf habit, and 488 geographic range. Proc. Natl. Acad. Sci. USA. 117, 10397–10405 (2020). 489 11. S. Piao, J. Tan, A. Chen, Y. H. Fu, P. Ciais, Q. Liu, I.A. Janssens, S. Vicca, Z. Zeng, S. Jeong, Y. Li, 490 R.B. Myneni, S. Peng, M. Shen, J. Peñuelas, Leaf onset in the northern hemisphere triggered by 491 daytime temperature. Nat. Commun. 6, 6911 (2015). 492 12. J. Barichivich, K. R. Briffa, R. B. Myneni, T. J. Osborn, T. M. Melvin, P. Ciais, S. Piao, C. Tucker, 493 Large-scale variations in the vegetation growing season and annual cycle of atmospheric CO2 at high 494 northern latitudes from 1950 to 2011. Global Change Biol. 19, 3167–3183 (2013). 495 13. J. Peñuelas, T. Rutishauser, I. Filella, Phenology Feedbacks on Climate Change. Science. 324, 887– 496 888 (2009). 497 14. S. Piao, Z. Liu, T. Wang, S. Peng, P. Ciais, M. Huang, A. Ahlstrom, J.F. Burkhart, F. Chevallier, I. A. 498 Janssens, S. Jeong, X. Lin, J. Mao, J. Miller, A. Mohammat, R. B. Myneni, J. Peñuelas, X. Shi, A. 499 Stohl, Y. Yao, Z. Zhu, P. P. Tans, Weakening temperature control on the interannual variations of 500 spring carbon uptake across northern lands. Nat. Clim. Change. 7, 359–363 (2017). 501 15. A. D. Richardson, T. A. Black, P. Ciais, N. Delbart, M. A. Friedl, N. Gobron, D. Y. Hollinger, W. L. 502 Kutsch, B. Longdoz, S. Luyssaert, M. Migliavacca, L. Montagnani, J. W. Munger, E. Moors, S. Piao, 503 C. Rebmann, M. Reichstein, N. Saigusa, E. Tomelleri, R. Vargas, A. Varlagin, Influence of spring and 504 autumn phenological transitions on forest ecosystem productivity. Philos. Trans. R. Soc., B. 365, 505 3227–3246 (2010). 506 16. G. B. Bonan, Forests and Climate Change: Forcings, Feedbacks, and the Climate Benefits of Forests. 507 Science. 320, 1444–1449 (2008). 508 17. S. Piao, Q. Liu, A. Chen, I. A. Janssens, Y. Fu, J. Dai, L. Liu, X. Lian, M. Shen, X. Zhu, Plant 509 phenology and global climate change: Current progresses and challenges. Global Change Biol. 25, 510 1922–1940 (2019). 26 511 18. Y. H. Fu, H. Zhao, S. Piao, M. Peaucelle, S. Peng, G. Zhou, P. Ciais, M. Huang, A. Menzel, J. 512 Peñuelas, Y. Song, Y. Vitasse, Z. Zeng, I.A. Janssens, Declining global warming effects on the 513 phenology of spring leaf unfolding. Nature. 526, 104–107 (2015). 514 19. T. Park, C. Chen, M. M. Fauria, H. Tømmervik, S. Choi, A. Winkler, U. S. Bhatt, D. A. Walker, S. 515 Piao, V. Brovkin, R. R. Nemani, R. B. Myneni, Changes in timing of seasonal peak photosynthetic 516 activity in northern ecosystems. Global Change Biol. 25, 2382–2395 (2019). 517 20. C. Xu, H. Liu, A. P. Williams, Y. Yin, X. Wu, Trends toward an earlier peak of the growing season in 518 Northern Hemisphere mid-latitudes. Global Change Biol. 22, 2852–2860 (2016). 519 21. X. Wang, J. Xiao, X. Li, G. Cheng, M. Ma, G. Zhu, M. A. Arain, T. A. Black, R.S. Jassal, No trends 520 in spring and autumn phenology during the global warming hiatus. Nat. Commun. 10, 2389 (2019). 521 22. S. Piao, P. Friedlingstein, P. Ciais, N. Viovy, J. Demarty, Growing season extension and its impact on 522 terrestrial carbon cycle in the Northern Hemisphere over the past 2 decades. Global Biogeochem. 523 Cycles. 21, B3018 (2007). 524 23. W. Buermann, P. R. Bikash, M. Jung, D. H. Burn, M. Reichstein, Earlier springs decrease peak 525 summer productivity in North American boreal forests. Environ. Res. Lett. 8, 24027 (2013). 526 24. W. Buermann, M. Forkel, M. O Sullivan, S. Sitch, P. Friedlingstein, V. Haverd, A. K. Jain, E. Kato, 527 M. Kautz, S. Lienert, D. Lombardozzi, J. E. M. S. Nabel, H. Tian, A. J. Wiltshire, D. Zhu, W. K. 528 Smith, A. D. Richardson, Widespread seasonal compensation effects of spring warming on northern 529 plant productivity. Nature. 562, 110–114 (2018). 530 25. X. Lian, S. Piao, L. Z. X. Li, Y. Li, C. Huntingford, P. Ciais, A. Cescatti, I. A. Janssens, J. Peñuelas, 531 W. Buermann, A. Chen, X. Li, R. B. Myneni, X. Wang, Y. Wang, Y. Yang, Z. Zeng, Y. Zhang, T. R. 532 McVicar, Summer soil drying exacerbated by earlier spring greening of northern vegetation. Sci. Adv. 533 6, eaax0255 (2020). 534 26. S. Piao, P. Ciais, P. Friedlingstein, P. Peylin, M. Reichstein, S. Luyssaert, H. Margolis, J. Fang, A. 535 Barr, A. Chen, A. Grelle, D. Y. Hollinger, T. Laurila, A. Lindroth, A. D. Richardson, T. Vesala, Net 536 carbon dioxide losses of northern ecosystems in response to autumn warming. Nature. 451, 49–52 537 (2008). 27 538 27. H. Wang, H. Liu, G. Cao, Z. Ma, Y. Li, F. Zhang, X. Zhao, X. Zhao, L. Jiang, N. J. Sanders, A. T. 539 Classen, J. S. He, Alpine grassland plants grow earlier and faster but biomass remains unchanged over 540 35 years of climate change. Ecol. Lett. 23, 701–710 (2020). 541 28. Y. Pan, R. A. Birdsey, J. Fang, R. Houghton, P. E. Kauppi, W. A. Kurz, O. L. Phillips, A. Shvidenko, 542 S. L. Lewis, J. G. Canadell, P. Ciais, R. B. Jackson, S. W. Pacala, A. D. McGuire, S. Piao, A. 543 Rautiainen, S. Sitch, D. Hayes, A Large and Persistent Carbon Sink in the World's Forests. Science. 544 333, 988–993 (2011). 545 29. N. Delpierre, Y. Vitasse, I. Chuine, J. Guillemot, S. Bazot, T. Rutishauser, C. B. K. Rathgeber, 546 Temperate and boreal forest tree phenology: from organ-scale processes to terrestrial ecosystem 547 models. Ann. Forest Sci. 73, 5–25 (2016). 548 30. J. Huang, Q. Ma, S. Rossi, F. Biondi, A. Deslauriers, P. Fonti, E. Liang, H. Makinen, W. Oberhuber, 549 C. Rathgeber, R. Tognetti, V. Treml, B. Yang, J. L. Zhang, S. Antonucci, Y. Bergeron, J. J. Camarero, 550 F. Campelo, K. Cufar, H. E. Cuny, M. De Luis, A. Giovannelli, J. Gricar, A. Gruber, V. Gryc, A. 551 Guney, X. Guo, W. Huang, T. Jyske, J. Kaspar, G. King, C. Krause, A. Lemay, F. Liu, F. Lombardi, 552 D. C. E. Martinez, H. Morin, C. Nabais, P. Nojd, R. L. Peters, P. Prislan, A. Saracino, I. Swidrak, H. 553 Vavrcik, J. Vieira, B. Yu, S. Zhang, Q. Zeng, Y. Zhang, E. Ziaco, Photoperiod and temperature as 554 dominant environmental drivers triggering secondary growth resumption in Northern Hemisphere 555 conifers. Proc. Natl. Acad. Sci. USA. 117, 20645–20652 (2020). 556 31. X. Li, E. Liang, J. Gričar, S. Rossi, K. Čufar, A. M. Ellison, Critical minimum temperature limits 557 xylogenesis and maintains treelines on the southeastern Tibetan Plateau. Sci. Bull. 62, 804–812 558 (2017). 559 32. S. Rossi, A. Deslauriers, J. Griçar, J. Seo, C. B. Rathgeber, T. Anfodillo, H. Morin, T. Levanic, P. 560 Oven, R. Jalkanen, Critical temperatures for xylogenesis in conifers of cold climates. Global Ecol. 561 Biogeogr. 17, 696–707 (2008). 562 33. A. Lenz, Y. Vitasse, G. Hoch, C Körner. Growth and carbon relations of temperate deciduous tree 563 species at their upper elevation range limit. J. Ecol. 102, 1537–1548 (2014). 28 564 34. Q. Zeng, S. Rossi, B. Yang, C. Qin, G. Li, Environmental drivers for cambial reactivation of Qilian 565 junipers (Juniperus przewalskii) in a semi-arid region of northwestern China. Atmosphere 11, 232 566 (2020). 567 35. P. Ren, E. Ziaco, S. Rossi, F. Biondi, P. Prislan, E. Liang, Growth rate rather than growing season 568 length determines wood biomass in dry environments. Agr. Forest Meteorol. 271, 46–53 (2019). 569 36. P. Sanginés De Cárcer, Y. Vitasse, J. Peñuelas, V. E. J. Jassey, A. Buttler, C. Signarbieux, Vapor– 570 pressure deficit and extreme climatic variables limit tree growth. Global Change Biol. 24, 1108–1122 571 (2017). 572 37. J. Zhang, X. Gou, M. R. Alexander, J. Xia, F. Wang, F. Zhang, Z. Man, N. Pederson. Drought limits 573 wood production of Juniperus przewalskii even as growing seasons lengthens in a cold and arid 574 environment. Catena 196, 104936 (2021). 575 38. J. Huang, A. Deslauriers, S. Rossi, Xylem formation can be modeled statistically as a function of 576 primary growth and cambium activity. New Phytol. 203, 831–841 (2014). 577 39. S. Rossi, H. Morin, A. Deslauriers. Causes and correlations in cambium phenology: towards an 578 integrated framework of xylogenesis. J. Exp. Bot. 63, 2117–2126 (2012). 579 40. S. Rossi, M. J. Girard, H. Morin. Lengthening of the duration of xylogenesis engenders 580 disproportionate increases in xylem production. Global Change Biol. 20, 2261–2271 (2014). 581 41. H. E. Cuny, C. B. K. Rathgeber, D. Frank, P. Fonti, H. Mäkinen, P. Prislan, S. Rossi, E. M. Del 582 Castillo, F. Campelo, H. Vavrčík, J. J. Camarero, M. V. Bryukhanova, T. Jyske, J. Gričar, V. Gryc, M. 583 De Luis, J. Vieira, K. Čufar, A. V. Kirdyanov, W. Oberhuber, V. Treml, J. Huang, X. Li, I. Swidrak, 584 A. Deslauriers, E. Liang, P. Nöjd, A. Gruber, C. Nabais, H. Morin, C. Krause, G. King, M. Fournier, 585 Woody biomass production lags stem-girth increase by over one month in coniferous forests. Nat. 586 Plants. 1, 15160 (2015). 587 42. E. Pasho, J. J. Camarero, S. M. Vicente-Serrano, Climatic impacts and drought control of radial 588 growth and seasonal wood formation in Pinus halepensis. Trees. 26, 1875–1886 (2012). 29 589 43. T. F. Keenan, J. Gray, M. A. Friedl, M. Toomey, G. Bohrer, D. Y. Hollinger, J. W. Munger, J. O 590 Keefe, H. P. Schmid, I. S. Wing, B. Yang, A. D. Richardson, Net carbon uptake has increased through 591 warming-induced changes in temperate forest phenology. Nat. Clim. Change. 4, 598–604 (2014). 592 44. L. Chen, H. Hänninen, S. Rossi, N. G. Smith, S. Pau, Z. Liu, G. Feng, J. Gao, J. Liu, Leaf senescence 593 exhibits stronger climatic responses during warm than during cold autumns. Nat. Clim. Change. 10, 594 777–780 (2020). 595 45. C. Körner. Paradigm shift in plant growth control. Curr Opin Plant Biol. 25, 107–114 (2015). 596 46. B. Muller, F. Pantin, M. Genard, O. Turc, S. Freixes, M. Piques, Y. Gibon. Water deficits uncouple 597 growth from photosynthesis, increase C content, and modify the relationships between C and growth 598 in sink organs. J. Exp. Bot. 62, 1715–1729 (2011). 599 47. N. D. Charney, F. Babst, B. Poulter, S. Record, V. M. Trouet, D. Frank, B. J. Enquist, M. E. K. Evans, 600 Observed forest sensitivity to climate implies large changes in 21st century North American forest 601 growth. Ecol. Lett. 19, 1119–1128 (2016). 602 48. Q. Liu, S. Piao, I. A. Janssens, Y. Fu, S. Peng, X. Lian, P. Ciais, R. B. Myneni, J. Peñuelas, T. Wang. 603 Extension of the growing season increases vegetation exposure to frost. Nat. Commun. 9, 426 (2018). 604 49. A. Deslauriers, H. Morin, Intra-annual tracheid production in balsam fir stems and the effect of 605 meteorological variables. Trees. 19, 402–408 (2005). 606 50. S. Piao, X. Wang, T. Park, C. Chen, X. Lian, Y. He, J. W. Bjerke, A. Chen, P. Ciais, H. Tømmervik, 607 R. R. Nemani, R. B. Myneni, Characteristics, drivers and feedbacks of global greening. Nat. Rev. 608 Earth Environ. 1, 14–27 (2020). 609 51. M. Huang, S. Piao, P. Ciais, J. Peñuelas, X. Wang, T. F. Keenan, S. Peng, J. A. Berry, K. Wang, J. 610 Mao, R. Alkama, A. Cescatti, M. Cuntz, H. De Deurwaerder, M. Gao, Y. He, Y. Liu, Y. Luo, R. B. 611 Myneni, S. Niu, X. Shi, W. Yuan, H. Verbeeck, T. Wang, J. Wu, I. A. Janssens, Air temperature 612 optima of vegetation productivity across global biomes. Nat. Ecol. Evol. 3, 772–779 (2019). 613 52. T. F. Keenan, W. J. Riley, Greening of the land surface in the world's cold regions consistent with 614 recent warming. Nat. Clim. Change. 8, 825–828 (2018). 30 615 53. J. J. Camarero, J. M. Olano, A. Parras, Plastic bimodal xylogenesis in conifers from continental 616 Mediterranean climates. New Phytol. 185, 471–480 (2010). 617 54. Y. H. Fu, S. Piao, H. Zhao, S. Jeong, X. Wang, Y. Vitasse, P. Ciais, I. A. Janssens, Unexpected role of 618 winter precipitation in determining heat requirement for spring vegetation green-up at northern middle 619 and high latitudes. Global Change Biol. 20, 3743–3755 (2014). 620 55. X. Wu, X. Li, H. Liu, P. Ciais, Y. Li, C. Xu, F. Babst, W. Guo, B. Hao, P. Wang, Y. Huang, S. Liu, Y. 621 Tian, B. He, C. Zhang, Uneven winter snow influence on tree growth across temperate China. Global 622 Change Biol. 25, 144–154 (2018). 623 56. X. Wang, T. Wang, H. Guo, D. Liu, Y. Zhao, T. Zhang, Q. Liu, S. Piao, Disentangling the 624 mechanisms behind winter snow impact on vegetation activity in northern ecosystems. Global Change 625 Biol. 24, 1651–1662 (2018). 626 57. H. D. Adams, A. D. Collins, S. P. Briggs, M. Vennetier, L. T. Dickman, S. A. Sevanto, N. Garcia- 627 Forner, H. H. Powers, N. G. McDowell, Experimental drought and heat can delay phenological 628 development and reduce foliar and shoot growth in semiarid trees. Global Change Biol. 21, 4210– 629 4220 (2015). 630 58. W. He, H. Liu, Y. Qi, F. Liu, X. Zhu, Patterns in nonstructural carbohydrate contents at the tree organ 631 level in response to drought duration. Global Change Biol. 26, 3627–3638 (2020). 632 59. A. P. Williams, C. D. Allen, A. K. Macalady, D. Griffin, C. A. Woodhouse, D. M. Meko, T. W. 633 Swetnam, S. A. Rauscher, R. Seager, H. D. Grissino-Mayer, J. S. Dean, E. R. Cook, C. 634 Gangodagamage, M. Cai, N. G. McDowell, Temperature as a potent driver of regional forest drought 635 stress and tree mortality. Nat. Clim. Change. 3, 292–297 (2012). 636 60. Y. Vitasse, A. Bottero, M. Cailleret, C. Bigler, P. Fonti, A. Gessler, M. Lévesque, B. Rohner, P. 637 Weber, A. Rigling, T. Wohlgemuth. Contrasting resistance and resilience to extreme drought and late 638 spring frost in five major European tree species. Global Change Biol 25, 3781–3792 (2019). 639 61. S. Zhao, N. Pederson, L. D'Orangeville, J. HilleRisLambers, E. Boose, C. Penone, B. Bauer, Y. Jiang, 640 R. D. Manzanedo, The International Tree-Ring Data Bank (ITRDB) revisited: Data availability and 641 global ecological representativity. J. Biogeogr. 46, 355–368 (2019). 31 642 62. F. Babst, B. Poulter, P. Bodesheim, M. D. Mahecha, D. C. Frank, Improved tree-ring archives will 643 support earth-system science. Nat. Ecol. Evol. 1, 8 (2017). 644 63. A. J. Elmore, S. M. Guinn, B. J. Minsley, A. D. Richardson, Landscape controls on the timing of 645 spring, autumn, and growing season length in mid-Atlantic forests. Global Change Biol. 18, 656–674 646 (2012). 647 64. S. A. Kannenberg, J. T. Maxwell, N. Pederson, L. D'Orangeville, D. L. Ficklin, R. P. Phillips Drought 648 legacies are dependent on water table depth, wood anatomy and drought timing across the eastern US. 649 Ecol. Lett. 22, 119–127 (2018). 650 65. S. Rossi, A. Deslauriers, T. Anfodillo, V. Carraro. Evidence of threshold temperatures for xylogenesis 651 in conifers at high altitudes. Oecologia 152, 1-12 (2007). 652 66. S. Gao, R. Liu, T. Zhou, W. Fang, C. Yi, R. Lu, X. Zhao, and H. Luo, Dynamic responses of tree-ring 653 growth to multiple dimensions of drought. Global Change Biol. 24, 5380–5390 (2018). 654 67. D. M. P. Peltier, K. Ogle, Tree growth sensitivity to climate is temporally variable. Ecol. Lett. 23, 655 1561–1572 (2020). 656 68. M. Wilmking, M. Maaten Theunissen, E. Maaten, T. Scharnweber, A. Buras, C. Biermann, M. 657 Gurskaya, M. Hallinger, J. Lange, R. Shetti, M. Smiljanic, M. Trouillier, Global assessment of 658 relationships between climate and tree growth. Global Change Biol. 26, 3212–3220 (2020). 659 69. K. Seftigen, D. C. Frank, J. Björklund, F. Babst, B. Poulter, The climatic drivers of normalized 660 difference vegetation index and tree-ring-based estimates of forest productivity are spatially coherent 661 but temporally decoupled in Northern Hemispheric forests. Global Ecol. Biogeogr. 27, 1352–1365 662 (2018). 663 70. A. G. Bunn, A dendrochronology program library in R (dplR). Dendrochronologia. 26, 115–124 664 (2008). 665 71. R Development Core Team. R: A language and environment for statistical computing. R Foundation 666 for statistical computing. (Vienna, Austria, 2019). https://www.R-project.org/ 667 72. J. Sheffield, G. Goteti, E. F. Wood, Development of a 50-Year High-Resolution Global Dataset of 668 Meteorological Forcings for Land Surface Modeling. J. Clim. 19, 3088-3111 (2006). 32 669 73. P. L. Frich, P. Alexander, B. Della-Marta, M. Gleason, A. Haylock, K. Tank, T. Peterson, Observed 670 coherent changes in climatic extremes during the second half of the twentieth century, Clim. Res., 19, 671 193–212 (2002). 672 74. G. T. Selyaninov, About climate agricultural estimation (in Russian). Proc. Agric. Meteorol., 20, 165– 673 177 (1928). 674 75. D. L. Streiner, Finding Our Way: An Introduction to Path Analysis. Can. J. Psychiatry. 50, 115–122 675 (2005). 676 76. J. Fox, Z. Nie, J. Byrnes, sem: Structural Equation Models. R package version 3.1-9. 2017 Available 677 from: https://CRAN.R-project.org/package=sem 678 77. M. Iturbide, J. M. Gutiérrez, L. M. Alves, J. Bedia, R. Cerezo-Mota, E. Cimadevilla, A. S. Cofifiño, 679 A. D. Luca, S. H. Faria, I. V. Gorodetskaya, M. Hauser, S. Herrera, K. Hennessy, H. T. Hewitt, R. G. 680 Jones, S. Krakovska, R. Manzanas, D. Martínez-Castro, G. T. Narisma, I. S. Nurhati, I. Pinto, S. I. 681 Seneviratne, B. van den Hurk, C. S. Vera. An update of IPCC climate reference regions for 682 subcontinental analysis of climate model data: definition and aggregated datasets. Earth Syst. Sci. 683 Data 12, 2959–2970 (2020). 684 78. R. P. Bagozzi, Y. Yi, Specification, evaluation, and interpretation of structural equation models. J. 685 Acad. Market. Sci. 40, 8–34 (2012). 33