18The phenology of spring leaf unfolding influences regional and hemispheric-scale carbon 19 balances 2 , the long-term distribution of tree species 9 , and plant-animal interactions 10 . Changes in 20 the phenology of spring leaf unfolding can also exert biophysical feedbacks on climate by 21 modifying the surface albedo and energy budget 11,12 . Recent studies have reported significant 22 advances in spring phenology as a result of warming in most northern hemisphere regions 1,3,4 . 23Climate warming is projected to further increase 13 , but the future evolution of the phenology of 24 spring leaf unfolding remains uncertain -in view of the imperfect understanding of how the 25 underlying mechanisms respond to environmental stimuli 12,14 . In addition, the relative 26 contributions of each environmental stimulus, which together define the apparent temperature 27 sensitivity of the phenology of spring leaf unfolding (advances in days per degree Celsius 28 warming, S T ), may also change over time 6,8,15 . An improved characterization of the variation in 29 3 phenological responses to spring temperature is thus valuable, provided that it addresses temporal 1 and spatial scales relevant for regional projections. 2Numerous studies have reported advanced spring leaf unfolding which matches warming trends 3 over recent decades 1,3,4 . However, there is still debate regarding the linearity of leaf unfolding 4 response to the climate warming 6,7 . Recent experimental studies of warming using saplings have 5 shown that S T weakens as warming increases 7 . Experimental manipulation of temperature for 6 saplings or twigs, however, might elicit phenological responses that do not accurately reflect the 7 response of mature trees 16,17 . We therefore investigated the temporal changes in S T in adult trees 8 monitored in situ and exposed to real-world changes in temperature and other climate variables. 9These long-term data series were obtained across Central Europe from the Pan European regression for the entire period and for two 15-year periods, namely 1980-1994 and 1999-2013, 25 that had slight difference in preseason lengths (Extended Data Fig. 3a). The leaf unfolding dates (Fig. 1a). But the surprising result is that S T 3 significantly decreased by 40.0% from 4.0 ± 1.8 days °C -1 during 1980-1994 to 2.3 ± 1.6 days °C -4 1 during 1999-2013 (t=-37.3, df=5473, P<0.001) (Fig. 1b). All species show similar significant 5 decreases in S T (Fig. 1a), although the extent of reduction was species-specific. For example, 6Aesculus hippocastanum (see caption to Fig. 1 for English common names) had the largest 7 decrease in S T (-2.0 days °C -1 ), while S T decreased only slightly (but still significantly) in Fagus 8 sylvatica (-0.9 days °C -1 ) (Fig. 1a). Similar results were also obtained using a fixed preseason 9 length determined either in the time period 1980-1994 or in 1999-2013 10 and 3c). The declining S T could, however, also have been an artifact resulting from the 11 ‗encroachment' of leaf unfolding dates...

Aim Although numerous studies have reported advanced Northern Hemisphere spring phenology since the 1980s, recent studies based on remote sensing have reported a reversal or deceleration of this trend. This study aimed (1) to fully understand recent spring phenology shifts using both in situ observations and satellite‐based normalized difference vegetation index (NDVI) datasets, and (2) to test whether the NDVI methods capture the trends observed in situ. Location Western Central Europe. Methods Temporal spring phenology trends (leaf unfolding dates) were examined using 1,001,678 in situ observations of 31 plant species at 3984 stations, as well as NDVI‐based start‐of‐season dates, obtained using five different methods, across the pixels that included the phenology stations. Results In situ and NDVI observations both indicated that spring phenology significantly advanced during the period 1982–2011 at an average rate of −0.45 days yr−1. This trend was not uniform across the period and significantly weakened over the period 2000–2011. Furthermore, opposite trends were found between in situ and NDVI observations over the period 2000–2011. Averaged over all species, the in situ observations indicated a slower but still advancing trend for leaf unfolding, whereas the NDVI series showed a delayed spring phenology. Main conclusions The recent trend reversal in the advancement of spring phenology in western Central Europe is likely to be related to the response of early spring species to the cooling trend in late winter. In contrast, late spring species continued to exhibit advanced leaf unfolding, which is consistent with the warming trend during spring months. Because remote sensing does not distinguish between species, the signal of growing‐season onset may only reflect the phenological dynamics of these earliest species in the pixel, even though most species still exhibit advancing trends.

Recent studies have revealed large unexplained variation in heat requirement-based phenology models, resulting in large uncertainty when predicting ecosystem carbon and water balance responses to climate variability. Improving our understanding of the heat requirement for spring phenology is thus urgently needed. In this study, we estimated the species-specific heat requirement for leaf flushing of 13 temperate woody species using long-term phenological observations from Europe and North America. The species were defined as early and late flushing species according to the mean date of leaf flushing across all sites. Partial correlation analyses were applied to determine the temporal correlations between heat requirement and chilling accumulation, precipitation and insolation sum during dormancy. We found that the heat requirement for leaf flushing increased by almost 50% over the study period 1980-2012, with an average of 30 heat units per decade. This temporal increase in heat requirement was observed in all species, but was much larger for late than for early flushing species. Consistent with previous studies, we found that the heat requirement negatively correlates with chilling accumulation. Interestingly, after removing the variation induced by chilling accumulation, a predominantly positive partial correlation exists between heat requirement and precipitation sum, and a predominantly negative correlation between heat requirement and insolation sum. This suggests that besides the well-known effect of chilling, the heat requirement for leaf flushing is also influenced by precipitation and insolation sum during dormancy. However, we hypothesize that the observed precipitation and insolation effects might be artefacts attributable to the inappropriate use of air temperature in the heat requirement quantification. Rather than air temperature, meristem temperature is probably the prominent driver of the leaf flushing process, but these data are not available. Further experimental research is thus needed to verify whether insolation and precipitation sums directly affect the heat requirement for leaf flushing.

Heat requirement, expressed in growing degree days (GDD), is a widely used method to assess and predict the effect of temperature on plant development. Until recently, the analysis of spatial patterns of GDD requirement for spring vegetation green-up onset was limited to local and regional scales, mainly because of the sparse and aggregated spatial availability of ground phenology data. Taking advantage of the large temporal and spatial scales of remote sensing-based green-up onset data, we studied the spatial patterns of GDD requirement for vegetation green-up at northern middle and high latitudes. We further explored the correlations between GDD requirement for vegetation green-up and previous winter season chilling temperatures and precipitation, using spatial partial correlations. We showed that GDD requirement for vegetation green-up onset declines towards the north at a mean rate of 18.8 °C-days per degree latitude between 35°N and 70°N, and vary significantly among different vegetation types. Our results confirmed that the GDD requirement for vegetation green-up is negatively correlated with previous winter chilling, which was defined as the number of chilling days from the day when the land surface froze in the previous autumn to the day of green-up onset. This negative correlation is a well-known phenomenon from local studies. Interestingly, irrespective of the vegetation type, we also found a positive correlation between the GDD requirement and previous winter season precipitation, which was defined as the sum of the precipitation of the month when green-up onset occur and the precipitation that occurred during the previous 2 months. Our study suggests that GDD requirement, chilling and precipitation may have complex interactions in their effects on spring vegetation green-up phenology. These findings have important implications for improving phenology models and could therefore advance our understanding of the interplay between spring phenology and carbon fluxes.

We evaluated the seasonality of CO 2 fluxes simulated by nine terrestrial ecosystem models of the TRENDY project against (1) the seasonal cycle of gross primary production (GPP) and net ecosystem exchange (NEE) measured at flux tower sites over different biomes, (2) gridded monthly Model Tree Ensembles-estimated GPP (MTE-GPP) and MTE-NEE obtained by interpolating many flux tower measurements with a machine-learning algorithm, (3) atmospheric CO 2 mole fraction measurements at surface sites, and (4) CO 2 total columns (X CO2 ) measurements from the Total Carbon Column Observing Network (TCCON). For comparison with atmospheric CO 2 measurements, the LMDZ4 transport model was run with time-varying CO 2 fluxes of each model as surface boundary conditions. Seven out of the nine models overestimate the seasonal amplitude of GPP and produce a too early start in spring at most flux sites. Despite their positive bias for GPP, the nine models underestimate NEE at most flux sites and in the Northern Hemisphere compared with MTE-NEE. Comparison with surface atmospheric CO 2 measurements confirms that most models underestimate the seasonal amplitude of NEE in the Northern Hemisphere (except CLM4C and SDGVM). Comparison with TCCON data also shows that the seasonal amplitude of X CO2 is underestimated by more than 10% for seven out of the nine models (except for CLM4C and SDGVM) and that the MTE-NEE product is closer to the TCCON data using LMDZ4. From CO 2 columns measured routinely at 10 TCCON sites, the constrained amplitude of NEE over the Northern Hemisphere is of 1.6 ± 0.4 gC m À2 d À1 , which translates into a net CO 2 uptake during the carbon uptake period in the Northern Hemisphere of 7.9 ± 2.0 PgC yr À1 .