Exploring the environmental drivers of vegetation seasonality changes in the northern extratropical latitudes: a quantitative analysis

Vegetation seasonality in the northern extratropical latitudes (NEL) has changed dramatically, but our understanding of how it responds to climate change (e.g. temperature, soil moisture, shortwave radiation) and human activities (e.g. elevated CO2 concentration) remains insufficient. In this study, we used two remote-sensing-based leaf area index and factorial simulations from the TRENDY models to attribute the changes in the integrated vegetation seasonality index (S), which captures both the concentration and magnitude of vegetation growth throughout the year, to climate, CO2, and land use and land cover change (LULCC). We found that from 2003 to 2020, the enhanced average S in the NEL (MODIS: 0.0022 yr−1, p < 0.05; GLOBMAP: 0.0018 yr−1, p < 0.05; TRENDY S3 [i.e. the scenario considering both time-varying climate, CO2, and LULCC]: 0.0011 ± 7.5174 × 10−4 yr−1, p < 0.05) was primarily determined by the elevated CO2 concentration (5.3 × 10−4 ± 3.8 × 10−4 yr−1, p < 0.05) and secondly controlled by the combined climate change (4.6 × 10−4 ± 6.6 × 10−4 yr−1, p > 0.1). Geographically, negative trends in the vegetation growth concentration were dominated by climate change (31.4%), while both climate change (47.9%) and CO2 (31.9%) contributed to the enhanced magnitude of vegetation growth. Furthermore, around 60% of the study areas showed that simulated major climatic drivers of S variability exhibited the same dominant factor as observed in either the MODIS or GLOBMAP data. Our research emphasizes the crucial connection between environmental factors and vegetation seasonality, providing valuable insights for policymakers and land managers in developing sustainable ecosystem management strategies amidst a changing climate.


Introduction
Vegetation plays a pivotal role in mitigating anthropogenic climate change through biophysical and biogeochemical feedback.The former (e.g.changes in evapotranspiration) may slow down the warming rate (Jackson et al 2008, Zeng et al 2017), reduce the temperature range (He et al 2022, Xu et al 2022), and boost the water availability across water-limited and high-elevation regions (Yu et al 2017, Cui et al 2022).The latter (e.g.changes in carbon uptake) may enhance seasonal CO 2 exchange (Forkel et al 2016, Yuan et al 2018) and increase the amplitude of CO 2 (Barichivich et al 2013, Zeng et al 2014).The enhancing effect of vegetation on climate extremes is also non-negligible.For instance, reduced vegetation cover can directly intensify droughts and floods in Brazil (Chagas et al 2022), and grasslands in Europe can also intensify heatwaves by accelerating soil moisture depletion (Teuling et al 2010).Such biophysical and biogeochemical impacts of vegetation on climate are also seasonally dependent.In boreal summer, the biophysical impacts of northern vegetation tend to dampen warming, whereas, in spring, it can amplify warming (Lian et al 2022).The spring extensional growing season enhances CO 2 uptake, while the autumn extension intensifies CO 2 release (Barichivich et al 2013).
Many studies have focused on the drivers of partial changes in vegetation growth throughout the year (e.g.spring, summer, autumn, and the growing season), especially in northern high latitudes.This is due to the intrinsic characteristics of vegetation, which undergoes seasonal changes.In spring, warming typically advances the onset date of vegetation growth in the northern extratropical latitudes (NEL) (Xia et al 2015, Park et al 2016, Wang et al 2022), although the temperature sensitivity of spring vegetation phenology has been declining (Fu et al 2015, Meng et al 2020).During the summer season, droughts usually enhance gross primary production (GPP) in Southwestern China (Song et al 2019b), reduce GPP in non-forest sites in North America (Wu and Chen 2013), and cause asymmetric ecosystem carbon fluxes in Europe (Bastos et al 2020).The delayed ending date in autumn has also been widely revealed (Jeong et al 2011, Zhao et al 2015, Ma et al 2022), likely associated with climate change, spring phenology, and elevated CO 2 concentration (Liu et al 2016).
Various metrics have been developed to measure the seasonal growth of vegetation, such as the length of the growing season (Cui andShi 2021, Jiang et al 2022) and the difference between the maximum and minimum vegetation index during the growing season (Chen et al 2021).These metrics aim to capture the specific condition of vegetation growth throughout the year, rather than focusing on overall changes; however, it is important to note that vegetation growth is a continuous cycle process.Advanced vegetation onset date may lead to declined summer productivity (Buermann et al 2013) and the influence of vegetation carryover effect (i.e. the impact of current vegetation states on subsequent growth) is even stronger than climate on peak-to-late vegetation growth (Lian et al 2021).To better characterize vegetation cycling/seasonality throughout a year, previous studies also used the Markham index (i.e. the index that quantifies the degree of dissimilarity between two sets of data) to represent the seasonality of fruiting (Zimmerman et al 2007) or litterfall (Chave et al 2010).Although the Markham index describes the distribution of vegetation growth over 12 months, it mainly captures the concentration (i.e. the disparities between the distribution of actual vegetation growth and a uniform distribution), instead of the magnitude, of vegetation growth.To comprehensively investigate vegetation activity, including both the concentration and magnitude, another study examined the changes of 21 vegetation metrics, such as leaf-on date and leaf-off date, and revealed the role that each metric played in vegetation activity (Buitenwerf et al 2015).However, the natural and anthropogenic driving mechanisms underlying the changes of individual seasonal metrics remain unexplored.
In NEL, temperature is a crucial factor influencing vegetation seasonality, as supported by satellite observations (Piao et al 2013, Chen et al 2018) and manipulative experiments (Richardson et al 2018, Song et al 2019a).It is worth noting that the impact of warming on spring phenology is often associated with the rate of warming within the spring season (Wang et al 2015, Meng et al 2020).On the other hand, winter warming in Europe has been linked to an increased thermal requirement for leaf unfolding (Zhang et al 2022).The complexity of vegetation response to temperature may be regulated by changes in the optimum temperature for vegetation growth (Huang et al 2019), which is further modulated by the availability of water or light.For example, droughts may cause a declined influence of temperature on vegetation growth (Piao et al 2014) and seasonal precipitation patterns largely control the lagged effect of spring warming on vegetation growth (Buermann et al 2018).Moreover, seasonal flowering and fruiting are limited by seasonal irradiance in wet regions of neotropical forests (Zimmerman et al 2007).Therefore, we hypothesize that the concentration of vegetation growth, which represents the distribution of vegetation growth, is mainly determined by climate change.
In terms of the magnitude of vegetation growth, which refers its overall quantity, we hypothesize that various factors may matter.Climate change has a profound impact on the trend and variability of vegetation growth (Seddon et al 2016).Temperature controls more than 40% of NEL vegetation growth (Chen et al 2018), while water availability is increasingly becoming a limiting factor for vegetation growth in NEL (Jiao et al 2021).Additionally, the importance of solar radiation on vegetation growth is intensifying in NEL's moist regions (Yuan et al 2022).Anthropogenic effects (e.g.CO 2 fertilization, land use and land cover change [LULCC]) on the magnitude of vegetation growth are also remarkable.CO 2 physiological effects enhance GPP by expediting the process of carboxylation and improving water use efficiency but may decrease GPP through vegetation-climate feedback by reducing precipitation (Andrews et  To advance the understanding of NEL (>30 • N) vegetation growth activities, we applied an integrated seasonality index (i.e.equation ( 1)) (Feng et al 2013) in this study and examined the spatiotemporal changes in vegetation seasonality from 2003 to 2020.We further disentangled the individual effects of climate change and human activities on trends and variabilities of vegetation seasonality using semi-factor simulations from the TRENDYv10 project (https://blogs.exeter.ac.uk/trendy/).We also examined changes in relative entropy (concentration of vegetation growth) and normalized annual leaf area index (LAI) (magnitude of vegetation growth), respectively, and verified the above-mentioned process-based hypotheses.

Datasets
LAI, a proxy of vegetation growth, is widely used because of its clear physical interpretation and availability for both observations and models (Piao et al 2020).In this study, we used two remote-sensing LAI for the period 2003-2020: the gap-filled Moderate Resolution Imaging Spectroradiometer LAI (MODIS C6) (http://globalchange.bnu.edu.cn/research/laiv6) and GLOBMAP LAI (version3) (Liu et al 2012).
Monthly LAI, temperature, radiation and soil moisture data (table S1) were also derived from the process-based ecosystem models compiled by the TRENDYv10 project.The physical interpretation of the underlying mechanism in these models is based on the fundamental processes of photosynthesis, respiration, and carbon allocation in plants, as well as soil biogeochemistry (https://blogs.exeter.ac.uk/ trendy/).The TRENDY models simulate how these processes are affected by environmental factors such as temperature, soil moisture, LULCC, and atmospheric CO 2 concentrations.The climate and CO 2 fields in TRENDY models were generated from observations (Friedlingstein et al 2022, Lin et al 2023), which were used to drive and evaluate the models.Each model has four scenarios: S0 was driven by steady-state environmental forcings; S1 was forced by time-varying CO 2 with other factors kept constant; S2 was driven by time-varying CO 2 and climate and steady-state LULCC; and S3 was forced by time-varying CO 2 , climate, and LULCC.Therefore, the effects of CO 2 , climate and LULCC on the vegetation seasonality can be estimated from S1-S0, S2-S1, and S3-S2, respectively.Moreover, the combined effects of CO 2 and climate can be estimated from S2-S0, and the combined effects of CO 2 , climate and LULCC can be estimated from S3-S0.
All observational and model datasets were remapped into a spatial resolution of 0.5 • × 0.5 • , and the study area in NEL (>30 • N) was defined by NDVI > 0.125 and by the fraction of irrigation coverage ⩽ 10% (Data were collected around 2005; www.fao.org/aquastat/en/geospatial-information/globalmaps-irrigated-areas/latest-version/;Li et al 2022).

Analysis
The vegetation seasonality index (S) was calculated by combining multiplicatively relative entropy (D) and normalized annual LAI ( L Lmax ), and similar seasonality methodology was used in the analysis of rainfall seasonality in a previous study (Feng et al 2013): where l t is monthly LAI, p t is monthly probability distribution (p t = lt L ), q is the uniform distribution (q = 1 12 ), L max is the maximum accumulated annual LAI from 2003 to 2020.
D represents the concentration of vegetation growth within a calendar year.A higher D value indicates a greater difference between the real distribution of LAI within 1 year and the uniform distribution, suggesting a greater variability or change in LAI.When the actual probability distribution is uniform, D reaches a minimum value of 0, while it reaches a maximum value of log 2 12 when L is concentrated in a single month.L Lmax represents the magnitude of vegetation growth throughout a year (figure S1).Overall, D captures the seasonal variability in vegetation growth, and trends in L Lmax capture the patterns of greening or browning in annual cumulative LAI.
All the trends were obtained using linear regression coefficients.The responses of vegetation seasonality variabilities to climate were examined by partial correlation coefficients.Specifically, we calculated the seasonality indices of temperature (S TMP ), soil moisture (S SM ), and radiation (S RAD ) using the same method as in calculating S.Then, we removed the long-term trends in the anomalies of these seasonality indices.Finally, we used partial correlation coefficients to examine the sensitivities of S to the climate seasonality indices.

Trends in NEL averaged vegetation seasonality and their drivers
Averaged time series of S anomalies from satellite observations and the multi-model ensemble mean (MMEM) consistently revealed significant positive trends in NEL during the period 2003-2020 (figures 1(a) and (b)).For MODIS, the trend was 0.0022 yr −1 (p = 0.000 03); for GLOBMAP, it was 0.0018 yr −1 (p = 0.0002); and for TRENDY S3, it was 0.0011 ± 7.5174 × 10 −4 yr −1 (p = 0.000 04).Both remote-sensing S anomalies were mostly within one standard deviation of the MMEM under the S3 scenario (figure 1(a)).Further consistencies were found when comparing the observed trends of S anomalies with those simulated by individual TRENDY models (figure S2(a)).
All remote-sensing and S3-forced D anomalies showed decreasing trends, indicating that the NEL LAI probability distribution became closer to uniform (figures 1(c) and (d)).Climate change induced large but insignificant D reduction, whereas LULCC and CO 2 showed small, statistically insignificant contributions to the D evolution (figure 1(d)).In contrast to D changes, the average NEL L Lmax for individual products demonstrated evident increasing trends, with both CO 2 and climate change significantly and almost equally driving the tends (figures 1(e) and (f); figures S2(b)-(f)).

Spatial trend patterns of vegetation seasonality and their drivers
The spatial trend distributions of S anomalies, derived from the two satellite LAI (figures 2(a) and (b)), showed positive values over a large portion of the NEL vegetated area (figure 2(d)).Both remote-sensing products demonstrated significantly positive trends of S largely in central and northern North America, northeastern Russia, and North China.However, they exhibited a different sign of local changes mainly in central Europe and central-eastern Asia.
The S trends of MMEM with all the drivers considered (TRENDY S3) largely agreed with the satellite estimations, in terms of the magnitude and changing sign (figure 2(c)).The MMEM S changes in central Europe were closer to those estimated by MODIS, while the simulated S trends in central-eastern Asia resembled the GLOBMAP results.Moreover, all three products showed that the cover fraction with statistically significant positive trends (p < 0.1) (MODIS: 78.2%; GLOBMAP: 73.8%; TRENDY S3: 81.6%) was higher than that with statistically significant negative trends (p < 0.1) (MODIS: 21.8%; GLOBMAP: 26.2%; TRENDY S3: 18.4%) (figure 2(d)).
Across the NEL vegetated area, D anomalies showed significant negative trends (p < 0.1) (MODIS:

Sensitivity of vegetation seasonality to climate change
For both MODIS and TRENDY estimations, the coverage area dominated by S TMP in the variation of S constituted approximately 40% of the study area (figures 4(a) and (c)).However, the GLOBMAP estimations slightly underestimated the role of S TMP , wherein S TMP was the dominant factor in about 30% of the study area (figure 4(b)).Additionally, GLOBMAP results exaggerated the spatial heterogeneity in the dominant climate drivers compared to the MODIS and TRENDY estimations (figure 4).Furthermore, while all products exhibited a significant positive influence of S SM on S across southwestern North America and a significant negative influence of S RAD on S across central North America, they varied in the influencing factors observed in different regions.Diverse affecting factors mainly existed across the Arctic region (>53 • N), western Europe, eastern North America, and northeast China.
In the Arctic, S variation produced by MODIS was mainly positively significantly associated with S RAD (40.8%) (figure 4(a)), showing a similar pattern to the S RAD response of S in GLOBMAP (29.1%) (figure 4(b)).However, the results from TRENDY overestimated the positive contribution of S SM (28.4%) (figure 4(c)).The main locations with significant positive impacts of S TMP on TRENDY S were western Europe (80.2%) and eastern North America (90.0%) (i.e.western Europe: 45 4(c)), similar to the estimates from MODIS (western Europe: 84.4%; eastern North America: 72.8%) (figure 4(a)).In the northeast China region (45 • -53 • N, 118 • -130 • E), S variation simulated by TRENDY was significantly negatively associated with S RAD (74.4%) (figure 4(c)), resembling the S RAD response of S in GLOBMAP (47.3%) (figure 4(b)).Overall, despite the differences observed among the various datasets, some degree of comparability still existed between them (figure 4(d)).It is crucial to emphasize that approximately 60% of the study area displayed the same dominant factor as observed in either the MODIS or GLOBMAP data when the contribution sign was disregarded.These areas were scattered across the entire study area.

Discussion
Vegetation seasonality plays a critical role in assessing vegetation growth and in understanding the impacts of climate change and human activities on the health This study investigated compound changes in the concentration and magnitude of vegetation seasonality (figures 2, S3 and S4), and explored possible natural and anthropogenic drivers behind such annual evolutions (e.g.enhanced S, reduced D, enhanced L Lmax ) (figures 1, 3 and S5-S7).From 2003 to 2020, both CO 2 and climate change were essential to increased NEL averaged S or L Lmax anomalies, whereas decreased NEL averaged D anomalies were primarily caused by climate (figure 1).We also examined the If the dominant driver (did not consider the sign of contribution) estimated from TRENDY S3 was consistent with that from MODIS or GLOBMAP, the corresponding pixel was filled in green; otherwise, it was filled in gray.The spatial pattern was smoothed using a nine-point smoothing method.spatial patterns of major climatic drivers of S variabilities at an annual scale.We found that all products exhibited a significant positive influence of S SM on S across southwestern North America and a significant negative influence of S RAD on S across central North America.Additionally, sensitivity differences of S to climatic variables were observed in the Arctic, western Europe, eastern North America, and northeast China existing in diverse LAI products (figure 4).
Spatially, in most NEL vegetated lands, elevated CO 2 concentration induced positive trends in S, D, and L Lmax (figures S5(a)-S7(a)), confirming the enhancement of NEL vegetation growth due to higher CO 2 (Zhu et al 2016, Huang et al 2018, O'Sullivan et al 2022).Nevertheless, different from previous studies suggesting the dominant influence of CO 2 on greening (Piao et al 2006, Los 2013, Haverd et al 2020), slightly stronger effects of climate than CO 2 on the trends of L Lmax (the greening) were noticed in this study (figures S7(a) and (b)).This result was reasonable given multi-line evidence showed declining CO 2 fertilization effects on vegetation growth during a similar study period.For example, weakening CO 2 effects on photosynthesis were observed by (Wang et al 2020(Wang et al , 2021)), and declining CO 2 fertilization effects were revealed when investigating ecosystem carbon sequestration (Norby et al 2010, Penuelas et al 2017).Notably, climate primarily induced negative trends in L Lmax anomalies (the browning) in Europe's mid-latitudes region (figure 3(c)), potentially caused by combined warming and decreased soil moisture (Kong et al 2017, Pan et al 2018).LULCC demonstrated much weaker influences than CO 2 or climate on the S or L Lmax trends (figures S5(c) and S7(c)).However, LULCC showed significant local impacts, especially across the latitude band between 30 • -50 • N (figure 3), likely attributed to the regional intensification and expansion of agriculture activities (Bastos et al 2019).
The regions with significant changes in S trends were mainly located in central North America, North China, northeastern Russia, and western Europe (figure 2).According to remote-sensing products, both D and L Lmax were strengthened in the former three regions, whereas TRENDY D was weakened in northeastern Russia (figures S3 and S4).In central North America and North China, positive trends of vegetation growth in summer were caused by CO 2 fertilization, while decreased vegetation activities in spring and winter were regulated by climate change (figures S10 and S12).In northeastern Russia, CO 2 fertilization also induced enhanced vegetation growth in summer, whereas climate change caused the greening during green-up and browning during senescence (figure S11), and thus the effect of climate may be responsible for the TRENDY D trend deviation.Vegetation growth in western Europe increased in spring but decreased in summer (figure S13), leading to reduced D but enhanced L Lmax (figures S3 and S4); climate change also determined these seasonal changes in vegetation growth (figure S13 The S variabilities were determined by climate variabilities (figures 1(a) and S9), in agreement with previous research that showed climate-induced alteration of both the concentration and magnitude of vegetation growth (Zhu et al 2016, Leblans et al 2017).Regionally, S variations in southwestern North America, where grasslands dominated (figure S8), were identified to be driven by S SM (figure 4); this agrees with studies that exhibited strong sensitivity of grassland growth to soil moisture (Flanagan andJohnson 2005, Fang et al 2018).In the Arctic, both MODIS LAI and GLOBMAP LAI demonstrated a notably positive impact of S TMP on S variabilities, while TRENDY S3 LAI showed weak temperature responses of S variabilities (figure 4).This implies that TRENDY S3 LAI may be biased in its temperature sensitivity in the Arctic, given previous studies that suggested remarkable warming influences on Arctic vegetation growth (Elmendorf et al 2012, Bjorkman et al 2020).Strong relationships between S variabilities and temperature seasonality variabilities were observed in western Europe and eastern North America for both the TRENDY and MODIS results, while GLOBMAP showed diverse spatial patterns of driving factors (figure 4).Inconsistences in these temperature impacts are probably due to the differences between LAI definitions, retrieval algorithms, and input data (Fang et al 2013); changes in NOAA satellite orbit and MODIS sensor degradation may also be contributing factors (Jiang et al 2017).Furthermore, the sparse temporal frequency of satellite data may introduce further uncertainty, making seasonality analysis susceptible to the influence of extreme values (Wilby et al 2023).
The TRENDY models generally performed well in simulating multi-year trends of S (figure 2).However, when separating S into D and L Lmax , the magnitudes of simulated D with negative trends were generally smaller than those found in the remote-sensing products (figure S3).This observation suggested that from 2003 to 2020, although both TRENDY and remote-sensing LAI distributed more and more uniformly throughout the year in regions with negative D trends, the TRENDY LAI appeared to underestimate the tendency towards uniformity compared to the remote-sensing LAI.The deviations of modeled interannual S variabilities were also mainly contributed by biases in D, and the TRENDY models showed much lower D variabilities than the remote-sensing based estimations (figures 1(a), (c), (e)).These model biases in the concentration of vegetation growth are likely due to the inability of TRENDY models to capture the phenological processes and their environmental responses (Richardson et al 2012, Murray-Tortarolo et al 2013, Forkel et al 2014, Meng et al 2021, Peano et al 2021).For example, the models typically overestimated the growing season length (Murray-Tortarolo et al 2013) but underestimated maximum LAI during the year (Richardson et al 2012).In terms of phenological sensitivities, models tended to underestimate the response of vegetation growth to water availability (Humphrey et al 2018, Forzieri et al 2020), and simulate weaker sensitivity of spring onset to warming (Chen et al 2016, Meng et al 2021).Future research is thus needed to reduce the uncertainties in the modeled phenology and its connections with various environmental cues.
In this study, we primarily investigated the potential impacts of contemporaneous environmental variables on S trends and variabilities.However, it is important to consider the lag and accumulation effects of climate change on vegetation growth, which could improve the climatic explanation of vegetation changes by more than 10% (Wu et al 2015, Ding et al 2020).Moreover, a recent study suggests that the previous vegetation state could have a stronger influence on the present vegetation state than previous climate conditions (Lian et al 2021).Other factors that were not examined in this study may also play a role in the changes of vegetation seasonality, such as the tradeoff process between CO 2 fertilization and water saving (Terrer et al 2016, Jiang et al 2021), and the balance between temperature and CO 2 on photosynthesis (Ainsworth andRogers 2007, Mathur et al 2014).Therefore, future studies are needed to comprehensively assess the variations and drivers of S and its components, and to improve the model representations of vegetation seasonality and their environmental feedbacks.

Conclusions
This study investigated the influence of various environmental factors on the trends and variabilities of vegetation seasonality in NEL.Our findings reveal that climate change and elevated CO 2 concentration jointly controlled the increasing trend in NELaveraged S, whereas climate change mainly regulated the S variabilities.Spatially, NEL enhanced S was primarily affected by climate (32.9%), followed by CO 2 (29.3%) and LULCC (7.6%).Our inter-product comparisons also raised concerns about uncertainties in TRENDY simulated trends and environmental sensitivity of D, and in the varying S sensitivity to climatic variables among diverse remote-sensing and model products.Given continued warming and increasing human activities, changes in NEL vegetation seasonality are expected to be amplified in the future.Overall, this study advanced our understanding of NEL vegetation activities throughout the year and the ways in which they respond to multi-factor environmental changes.

Figure 1 .
Figure 1.Trend and variation in vegetation seasonality.(a)/(c)/(e), Time series of seasonality index anomaly/relative entropy anomaly/normalized annual LAI anomaly (mean ± one standard deviation) estimated by two satellite LAI data sets and MMEM LAI over the period 2003-2020.(b)/(d)/(f), Trend in seasonality index anomaly/relative entropy anomaly/normalized annual LAI anomaly derived by satellite observation (OBS) and modeled trends driven by CO2 fertilization (CO2), climate change (CLI), and land use and land cover change (LULCC).Error bars show the standard deviation of trends derived from 14 model simulations.The asterisk indicates that the MODIS, GLOBMAP, or MMEM trend is statistically significant (p < 0.05, two asterisks; p < 0.1, one asterisk).

Figure 3 .
Figure 3. Spatial pattern of dominant drivers of the trend in vegetation seasonality anomalies.(a)-(c), Dominant driving factors of the trend in vegetation seasonality index anomalies (a), relative entropy anomalies (b), or normalized annual LAI anomalies (c), defined as the driver that induced the highest absolute trend in each grid cell.Regions labeled by black dots indicate trends induced by the dominant driving factor that are statistically significant (p < 0.1).Trend in seasonality index anomalies/relative entropy anomalies/normalized annual LAI anomalies derived by CO2 fertilization (CO2), climate change (CLI), and land use and land cover change (LULCC).

Figure 4 .
Figure 4. Spatial pattern of dominant drivers of the sensitivity between climate and vegetation seasonality.(a)-(c), Dominant driving factors of the sensitivity between climate variables and vegetation seasonality, defined as the driver that has the highest absolute partial correlation coefficient in each grid cell.A, calculated by MODIS LAI.B, calculated by GLOBMAP LAI.C, calculated by TRENDY S3 data.Regions labeled by black dots indicate sensitivities between dominant climate variables and vegetation seasonality that are statistically significant (p < 0.1).The climate variables included temperature seasonality (S TMP ), soil moisture seasonality (SSM), and radiation seasonality (SRAD).(d), Agreement in the dominant drivers of sensitivity.If the dominant driver (did not consider the sign of contribution) estimated from TRENDY S3 was consistent with that from MODIS or GLOBMAP, the corresponding pixel was filled in green; otherwise, it was filled in gray.The spatial pattern was smoothed using a nine-point smoothing method.
(e)).Such climate-modulated vegetation variations in western Europe confirm the findings of temperature-induced earlier start of the growing season (Hamunyela et al 2013, Jin et al 2019, Gao et al 2020) and weakened vegetation activities by intensified droughts in summer (Ciais et al 2005, Peters et al 2018, Lin et al 2020).
'Sullivan M et al 2022 Process-oriented analysis of dominant sources of uncertainty in the land carbon sink Nat.Commun.13 4781 Pan N, Feng X, Fu B, Wang S, Ji F and Pan S 2018 Increasing global vegetation browning hidden in overall vegetation greening: insights from time-varying trends Remote Sens. Environ.214 59-72 Pan Y et al 2011 A large and persistent carbon sink in the World's forests Science 333 988-93 Park T, Ganguly S, Tømmervik H, Euskirchen E S, Høgda K-A, Karlsen S R, Brovkin V, Nemani R R and Myneni R B 2016 Changes in growing season duration and productivity of northern vegetation inferred from long-term remote sensing data Environ.Res.Lett.11 084001 Pazzagli P T, Weiner J and Liu F 2016 Effects of CO2 elevation and irrigation regimes on leaf gas exchange, plant water relations, and water use efficiency of two tomato cultivars Agric.Water Manage.169 26-33 Peano D et al 2021 Plant phenology evaluation of CRESCENDO land surface models -Part 1: Start and end of the growing season Biogeosciences 18 2405-28 Penuelas J, Ciais P, Canadell J G, Janssens I A, Fernandez-Martinez M, Carnicer J, Obersteiner M, Piao S, Vautard R and Sardans J 2017 Shifting from a fertilization-dominated to a warming-dominated period Nat.Ecol.Evol. 1 1438-45 Peters W et al 2018 Increased water-use efficiency and reduced CO (2) uptake by plants during droughts at a continental-scale Nat.Geosci.11 744-8 Piao S et al 2013 Evaluation of terrestrial carbon cycle models for their response to climate variability and to CO2 trends Glob.Change Biol.19 2117-32 Piao S et al 2014 Evidence for a weakening relationship between interannual temperature variability and northern vegetation activity Nat.Commun.5 5018 Piao S et al 2020 Characteristics, drivers and feedbacks of global greening Nat.Rev. Earth Environ. 1 14-27 Piao S, Friedlingstein P, Ciais P, Zhou L and Chen A 2006 Effect of climate and CO2 changes on the greening of the Northern Hemisphere over the past two decades Geophys.Res.Lett.33 L23402 Richardson A D et al 2012 Terrestrial biosphere models need better representation of vegetation phenology: results from the North American carbon program site synthesis Glob.Change Biol.18 566-84 Richardson A D et al 2018 Ecosystem warming extends vegetation activity but heightens vulnerability to cold temperatures Nature 560 368-71 Seddon A W, Macias-Fauria M, Long P R, Benz D and Willis K J 2016 Sensitivity of global terrestrial ecosystems to climate variability Nature 531 229-32 Sha Z, Bai Y, Li R, Lan H, Zhang X, Li J, Liu X, Chang S and Xie Y 2022 The global carbon sink potential of terrestrial vegetation can be increased substantially by optimal land management Commun.Earth Environ. 3 8 Song J et al 2019a A meta-analysis of 1,119 manipulative experiments on terrestrial carbon-cycling responses to global change Nat.Ecol.Evol. 3 1309-20 Song L, Li Y, Ren Y, Wu X, Guo B, Tang X, Shi W, Ma M, Han X and Zhao L 2019b Divergent vegetation responses to extreme spring and summer droughts in Southwestern China Agric.For.Meteorol.279 107703 Terrer C, Vicca S, Hungate B A, Phillips R P and Prentice I C 2016 Mycorrhizal association as a primary control of the CO2 fertilization effect Science 353 72-74 Teuling A J et al 2010 Contrasting response of European forest and grassland energy exchange to heatwaves Nat.Geosci.3 722-7 O