Introduction

Wood anatomical features (e.g., conduit size and density) are regarded as a useful proxy that records environmental signals (Fonti et al. 2010). For conifers, tracheid lumen size and cell wall thickness are the main potential proxies for recording environmental signals (Eilmann et al. 2006; Vaganov et al. 2006). Intra-annual density fluctuations (IADFs) are identified as the key parameter for understanding tree responses to water stress (Battipaglia et al. 2014). For angiosperms, the vessel area and the vessel density are the main focus of attention in ring-porous species (Fonti and García-González 2004) and diffuse-porous species (Sass-Klaassen and Eckstein 1995). Notably, in ring-porous trees, the sizes (diameter/area) of the first row of earlywood vessels appear to be particularly sensitive to the local environment (García-González and Fonti 2006). At the intra-specific level, narrower earlywood vessels (EV) were reported to reduce the risk of cavitation under both dry and wet conditions (Corcuera et al. 2005; Fonti et al. 2009). Nevertheless, recent work has challenged this paradigm (Eilmann et al. 2011; Fernández-de-Uña et al. 2017), as similar EV diameters in Quercus robur were found under two different flood regimes (Gričar et al. 2013). The environment influences the EV diameter by inducing changes to the process of cambial cell division and earlywood differentiation, both in the growth rate and duration (Hacke et al. 2017). Therefore, it seems that not only the size of EV but also its formation process (rate and duration) may provide a better understanding of the tree’s performance under drought conditions.

Analysis of the intra-annual dynamics of wood formation can provide related information on xylem anatomy. The dynamics of wood formation describe the process of cambial phenology and cell production throughout the growing season, with specific growth timing, rate and duration (Cuny et al. 2012; Lupi et al. 2010), which reflect the effects of different climatic factors on trees with a high time resolution (Rossi et al. 2014).

The key role of temperature in cambial reactivation and cell production has recently been demonstrated (Deslauriers et al. 2008). The predicted increase in temperatures is expected to enhance xylem formation (King et al. 2013) by starting earlier and prolonging the growing season. However, the influence of temperature may weaken during the course of the growing season since wood formation is not completed at same time each year and the duration of formation is variable (Rossi et al. 2008). Several recent studies have proven that precipitation in late spring can also play a critical role in the onset of xylogenesis under cold drought-prone regimes (Ren et al. 2015; Zhang et al. 2018). Thus, recovering the appropriate cell turgor by water uptake is necessary to resume tree growth (Kozlowski and Pallardy 2002). Furthermore, long and severe drought in summer may slow down or even stop wood formation (Eilmann et al. 2011) since xylem cell differentiation (division and expansion) is also sensitive to water potential (Stojanović et al. 2017). Therefore, a deeper understanding of the environmental factors controlling the intra-annual progress of wood formation and wood anatomy would be useful to evaluate the impact of the predicted climate change on future tree growth.

Chinese cork oak (Quercus variabilis Blume) plays a key role in the development of the regional economy and ecosystem, which produces wood, cork, charcoal, and edible fungi. This ring-porous species is a common and widespread deciduous broad-leaved tree in eastern Asia, which grows in temperate–subtropical climate conditions (19°–42°N and 97°–140°E) (Ma et al. 2013). Due to its wide distribution, Q. variabilis is a good candidate for studying the influences of environmental factors on wood formation. Based on a previous dendrochronology analysis, the radial growth response of the oak can record climatic variations and is especially sensitive to precipitation in the early spring/summer (Ferrio et al. 2003; Lyu et al. 2017).

In this study, the wood formation of Q. variabilis was monitored at weekly/biweekly intervals, and the wood anatomy features were measured. Our study was conducted in two consecutive growing seasons (2015 and 2016), with different levels of monthly and seasonal dryness, in which 2016 was a warmer year with a drier spring. Therefore, the aims of this study were (1) to monitor the timing of intra-annual cambial phenology and the xylem differentiation dynamics of Q. variabilis; (2) to analyze the environmental influence on wood formation, especially the formation of the first row of earlywood vessels exposed to spring drought; and (3) to verify whether the first row of earlywood vessels could be a stable climate proxy in this monsoon-dominated region since most previous studies were conducted in Mediterranean climates with dry growing seasons. We hypothesized that (1) drought would induce a shorter growing season and that (2) the first row of earlywood vessels would be a sensitive indicator of the climate conditions, especially in the spring.

Materials and methods

Study site

The study area was located at Xiong’er Mountain (Central China, 33° 54′N–34° 31′N, 111° 10′E–112° 09′E, Fig. 1), in the Western Henan Mountain region. This mountain region is an extension of Qinling Mountain in Henan Province and transitions from a warm temperate zone to a subtropical zone, which is one of the most suitable habitats for Q. variabilis (Zhang et al. 2010). The mean annual temperature is 12.6 °C. The mean annual precipitation is 636 mm (1953–2014), and over 72% of this precipitation falls between May and September. These data were obtained from the nearest State standard meteorological station in Lushi (34° 03′N, 111° 02′E, 568 m a.s.l., linear distance ca. 30 km from the study site).

Fig. 1
figure 1

Geographical location of the study area in Central China

Full size image

The research was conducted on an uneven-aged Chinese cork oak secondary forest (Cao et al. 2003). Q. variabilis is the dominant woody species at low and middle elevations (< 1600 m a.s.l.), with an average height of 15 m and occupying more than 80% of the canopy coverage. The main soil type is sandy brown soil with an average soil depth of ca. 50 cm. The site is on a south-facing slope (20–30%) at an elevation of 970 m a.s.l.

Tree selection and section preparation

Intra-annual radial growth was monitored from late February until early December 2015 and 2016. Each year, we selected six healthy Q. variabilis sample trees without any visible injuries on the stems, branches or roots. To avoid wound effects from previous samplings, sampling positions on each tree were kept at least 5 cm apart in both horizontal and vertical direction. Thus, sampling could not be continued on the same trees in the second year, due to the high frequency of sampling dates and the need for multiple samples on some dates. The latter owed much to the high level of hardness in oak Chinese cork (level IV–V), which made it difficult to obtain intact samples. Trees sampled in 2015 and 2016 were distinct but with no significant differences (by T value test) in age, height and diameter at breast height (DBH), which were 31 ± 6 and 25 ± 4 years (p = 0.304), 16 ± 0.7 and 17 ± 2.3 m (p = 0.188), 20 ± 1.2 and 17 ± 1.5 cm (p = 0.465), respectively.

Micro-cores were initially collected at weekly intervals from late February to early June. Due to the slower growth rate of trees in summer, samples were processed at biweekly intervals from June to early December (González-González et al. 2013). Overall, ca. 1000 samples were collected using a Trephor tool (Rossi et al. 2006). Samples were stored in a solution (30% water and 70% equal parts ethanol and glycerol) at 4 °C to avoid tissue deterioration. After 1 month, the samples were dehydrated with increasing concentrations of ethanol and xylene and embedded in paraffin blocks (Paraplast Plus, ROTH, Karlsruhe, Germany). Transverse sections were cut to a thickness of 10–15 µm with a Leica RM 2235 rotary microtome (Leica Microsystems, Wetzlar, Germany). All sections were stained with 1% safranin (Solarbio, Beijing, China) (in water) and 0.5% fast green (Sigma-Aldrich, Steinheim, Germany) (in 95% ethanol). They were observed using a Nikon Ni-U light microscope (Tokyo, Japan) under a bright field and polarized light.

Cambial phenology and xylem growth

Stained sections for each tree were examined to distinguish the phenological phases of cambial activity and xylem cell differentiation as follows: (1) the onset of cambial cell production; (2) the onset of the enlargement of the first row of earlywood vessels (FEV); (3) the completion of FEV; (4) the transition from earlywood to latewood; (5) the end of cambial cell production; and (6) the end of xylem formation. Phases were assessed for each tree, which were calculated and expressed as the day of the year (DOY).

The observed phases were defined as follows (Plomion et al. 2001): (1) the onset of cambial cell production was defined as an increased number of thin-walled cambial cells compared to the dormant cambium (Fig. 2c) (Prislan et al. 2011); (2) the onset of FEV enlargement was characterized by large vessel cells and thin, non-lignified cell walls, which were generally located near the previous year’s growth ring boundary (González-González et al. 2014); (3) the completion of FEV was identified as the complete lignification of the FEV, which was indicated by completely red cell walls stained by the safranin–fast green procedure (Fig. 2d, e) (Gričar et al. 2017); (4) the transition from earlywood to latewood was defined as when small latewood vessels were no longer present in the rings (Gričar 2010); (5) the end of cambial cell production was determined as the time at which no new thin-walled cambial cells were observed in the cambium, and the number of cambial cells was comparable to the number before its reactivation in spring (Gričar et al. 2017); and (6) the end of xylem formation was identified by the complete lignification of the last formed latewood cells, which was indicated by completely red-stained cell walls (Fig. 2f). The duration of xylem formation was calculated as the number of days between the onset of cambial cell production (phase 1) and the end of xylem formation (phase 6) since cells in the enlargement phase were difficult to distinguish for Q. variabilis (Michelot et al. 2012).

Fig. 2
figure 2

Phases of cambial phenology in Quercus variabilis in 2016. a The structure of xylem rings in Q. variabilis. b The cambial zone was dormant on February 25; no new xylem cell production was observed. c Active cambium (black arrow) and newly formed xylem cells (enlargement, red arrow) on March 10. d, e Full completion of the first row of earlywood vessels on April 20; observed under a bright field (d) and polarized (e) light, respectively. f Newly formed xylem ring on September 5, 2016. a, b and c–f magnification of 50X, 200X and 100X, respectively. RW radial width, L latewood, E earlywood, CZ cambial zone, P phloem, X xylem, FEV the first row of earlywood vessels, LV latewood vessel, EV earlywood vessel

Full size image

Intra-annual xylem growth monitoring and fitting

The number of cambium cells was counted and the widths of xylem (µm) at each cell development phase were monitored and measured along three radial rows. This was accomplished with Nikon NIS-Elements D software (Tokyo, Japan).

The dynamics of the intra-annual cumulative xylem width per tree were fitted by the Gompertz function, which has proven its effectiveness in describing cellular-level growth (Lupi et al. 2014). Based on the estimated parameters from the Gompertz function, the date of the maximum rate of the new xylem width (tp, DOY) and the average new xylem width growth rate (rm, µm·day− 1) were computed. The growth-fitting procedures are given in detail in the Supplement.

Wood anatomy

The first row of earlywood vessels (FEV) has been found to contain the strongest record of environmental signals (García-González and Fonti 2006). The FEVs were considered to be (1) all vessels that were close to the ring boundary, and (2) those vessels whose front edges (i.e., the edges toward the pith) were radially positioned inward of the center of the vessels close to the ring boundary (González-González et al. 2014). The mean diameter of the first row of earlywood vessels (DFEV) and ring width (RW) on an inter-annual scale (2009–2016) were measured based on the micro-cores taken from every monitored tree in November 2016 (Table 1). Considering the substantially higher contribution of large conduits to hydraulic conductivity, only earlywood vessels with a lumen area > 7500 µm2 were considered to be FEVs (González-González et al. 2014). Anatomical sections were photographed with a Nikon Ni-U light microscope (Tokyo, Japan), and measurements were carried out using NIS-Elements D software.

Table 1 Several anatomical variables in the tree rings of Quercus variabilis
Full size table

Environmental factors

To assess the cumulative heat exposure before xylem formation, the cumulative growing degree days (GDD) from DateTa>5 (see below) to the date of cambial cell production onset and FEV completion were calculated. GDD was computed by a modified sine wave model based on daily air temperature (maximum and minimum) and threshold temperature (Seo et al. 2008). Here, 5 °C was chosen as the threshold temperature according to previous results performed on temperate species (Rossi et al. 2008). The date of DateTa>5 is defined as the first date on which a daily mean air temperature was warmer than 5 °C for five consecutive days.

To quantify environmental drought, we used the monthly Standardized Precipitation Evapotranspiration Index (SPEI). SPEI was calculated as the monthly difference between precipitation and potential evapotranspiration for the 1953–2016 period (Vicente-Serrano et al. 2010) using the SPEI package in R 3.4.2 (R Core Team 2017). Positive and negative SPEI values represent periods of wetness or dryness, respectively.

To produce corroborative environmental data from our study site, we measured the soil temperature (Ts, °C) and soil volumetric water content (SWC, m3 m− 3) at a soil depth of 10 cm using a HOBO data logger (Onset Computer Corp, Bourne, MA, USA) from 2015 to 2016. The recording frequency was every 2 h.

Data analysis

The differences in the timing and duration of cambial phenological phases between years (2015 and 2016) were determined using the Mann–Whitney U test (SPSS Inc., Chicago, IL, USA). Because the assumption of normality was not met by all parameters of critical dates and durations, analyses between years were followed by non-parametric Mann–Whitney U tests.

To investigate the effects of year and cambial age on DFEV and RW, we used linear mixed-effect models with the “lme4 package” in R 3.4.2 (Bates et al. 2012). The proposed models included the tree as a random effect and the year and cambial age as fixed effects. If year/cambial age had a significant influence on DFEV and RW (p < 0.05), we tested the differences using Tukey’s test. Additionally, we used Pearson correlations to assess the relationship between DFEV and RW.

The environment–xylem growth relationships were analyzed using Pearson correlation coefficients (Ren et al. 2015). Five environmental factors were included, i.e. daily mean air and soil temperatures, precipitation, relative humidity and soil volumetric water content. All factors were calculated as averaged (except for precipitation) or summed (precipitation) values between successive sampling dates (7–10 days). Xylem growth was measured by xylem width increments between successive samples, in which the xylem widths were predicted by the Gompertz model. Time-lag effects were also considered (Deslauriers and Morin 2005); however, the correlation coefficients between the environment and xylem growth gradually weakened when the lag time shifted from 0 to 7 days before each sampling date.

Results

Cambial phenology and xylem growth

The onset of cambial cell production occurred during the third week of March in 2015 (March 19, DOY 78), while it was ca. 1 week earlier in 2016 (March 10, DOY 70, Fig. 2c), with significant differences between the years (Table 2, p < 0.01). 7–10 days after the onset of cambial activity, the FEV became enlarged. Notably, the complete formation of FEV was nearly concurrent in 2015 (April 19, DOY 109) and 2016 (April 20, DOY 111, Fig. 2d), with no significant differences between the years (p = 0.734). However, the end of cambial cell production in 2015 was almost 3 weeks later than in 2016 (DOY 218), which was found as late as late-August (DOY 240). The end of xylem formation was detected on September 29 and September 5 (DOY 272 and 249, Fig. 2f) in 2015 and 2016, respectively.

Table 2 Phases of cambial phenology and xylem growth of Quercus variabilis in 2015 and 2016
Full size table

Thus, the duration of FEV formation was significantly different between the years (p < 0.05), which lasted an average of 31 days and 41 days in 2015 and 2016, respectively. Similarly, the duration of cambial cell production in 2015 lasted 163 days, compared with the shorter period of 148 days in 2016. As a result, the duration of xylem formation in 2015 was greater than 6 months (184 days), which was longer than the 172 days in 2016.

Dynamics of intra-annual xylem growth

Four to six cambium cells were present during the dormancy period, and after activity initiated, the number increased to a peak of 9 to 11 cells in May. The Gompertz function adequately fitted the intra-annual growth of Chinese cork oak (R2 = 0.88 ± 0.07, N = 12, p < 0.001; Fig. 3). During the two study years, > 60% of xylem growth was produced from May to July. The average width of xylem was 4311 ± 445 µm (min = 3896 µm; max = 5076 µm) and 3752 ± 891 µm (min = 2686 µm; max = 4952 µm) in 2015 and 2016, respectively (Fig. 3a, c). The date of maximum xylem growth rate (tp) occurred on June 1 in 2015 (DOY 152) and June 10 in 2016 (DOY 162). However, the average xylem width growth rate (rm) was 20 ± 2.9 µm day− 1 in 2015, which was less than 28 ± 8.4 µm day− 1 in 2016. Additionally, greater differences among the six trees were observed in 2016 than in 2015, which showed in both the date and the value of the maximum growth rate.

Fig. 3
figure 3

Xylem growth modeled by the Gompertz function for Quercus variabilis in 2015 and 2016. a, c Cumulative xylem growth (solid line), and b, d daily xylem width growth rate (dashed line). Dots represent micro-core data from six monitored trees in the 2015 and 2016 growing seasons

Full size image

Wood anatomy

No differences were found between the years (p = 0.235) or cambial ages (p = 0.159) in DFEV (Table 3), which were 390 ± 68.8 µm and 375 ± 82.3 µm in 2015 and 2016, respectively. In contrast, significant differences were found between years in RW (p < 0.01) but not for cambial ages (p = 0.114). RW in 2015 and 2016 was 2597 ± 1034.7 µm and 2583 ± 1207.0 µm, respectively. Moreover, the correlation coefficient between DFEV and RW was significant and positive (r = 0.326, N = 74, p < 0.01).

Table 3 Contribution of the year, cambial age, and their interaction on the mean diameter of the first row of earlywood vessels (DFEV) and ring width (RW) for Quercus variabilis (2009–2016)
Full size table

Environmental conditions in 2015 and 2016

The environment in 2015 and 2016 showed distinct differences at the start of the growing season in spring (February–April), and 2016 had a warmer but drier spring than 2015, including 0.4 and 2.2 °C higher monthly mean daily air temperatures in March and April, respectively, compared to 2015 (Fig. 4). Meanwhile, an extended drought period from December 2015 to early April 2016 (Fig. 5) caused an SPEI with continuously negative values that reached a minimum value of − 1.9 in March, when precipitation was only 0.5 mm. In contrast, 2015 had a much moister spring, with near-positive SPEI and a higher SWC value (ca. 20%) in March and April.

Fig. 4
figure 4

Standardized precipitation evapotranspiration index (SPEI), soil water content (SWC), precipitation and air temperature on a monthly scale in 2015 and 2016

Full size image
Fig. 5
figure 5

Daily environmental parameters at the study site in 2015 and 2016, including mean air and soil temperature (Ta, Ts), the sum of precipitation (P), mean relative humidity (RH), and mean soil water content (SWC). Data in a, c and d were obtained from the nearest State standard meteorological station in Lushi. Data in b and e were recorded by HOBO in the study site at Xiong’er Mountain. Vertical lines in (a) represent the DateTa>5 (the first date on which the daily mean air temperature was warmer than 5 °C for five consecutive days) in 2015 (black line) and 2016 (red line)

Full size image

In the main growing season (May–July), precipitation was slightly higher in 2016 (375.5 mm) than in 2015 (283.5 mm). However, its distribution was more uneven in 2016, in which 27% of the annual precipitation fell in June 2016 (14% in June 2015). High evaporation and low precipitation in August and September 2016 caused a negative SPEI value and lower SWC (Fig. 4).

DateTa>5 in 2015 and 2016, which represent the thermal condition, appeared on March 11 and March 2, respectively (Fig. 5a). Thermal requirements (GDD) for the onset of cambial cell production were similar in 2015 and 2016 (GDD2015 = 30.8 ± 0; GDD2016 = 37.0 ± 2.5; p = 0.050). However, different GDD requirements for FEV completion were observed between the 2 years (GDD2015 = 186.9 ± 34.0; GDD2016 = 266.0 ± 50.3; p < 0.05).

Environment–xylem growth relationship

Xylem growth showed positive and significant responses to air temperature, soil temperature and precipitation in 2016 (Table 4). However, xylem growth responded differently to precipitation between the years. In 2015, no significant correlation was observed with precipitation (r = 0.335, N = 23, p > 0.05), whereas positive and significant correlations were found only with air and soil temperature.

Table 4 Pearson correlation coefficients between environmental factors and xylem formation of Quercus variabilis in 2015 and 2016
Full size table

Discussion

Impact of thermal conditions on the onset of xylogenesis

Chinese cork oak exhibited similar thermal requirements for the onset of cambial cell production between the 2 years (GDD2015 = 30.8 ± 0; GDD2016 = 37.0 ± 2.5; p = 0.050), in spite of the warmer air temperature in 2016. DateTa>5 in 2015 and 2016 occurred at different times, March 2 (DOY 70) and March 11 (DOY 62), respectively (Fig. 5b). Interestingly, 8 days after DateTa>5, trees began synchronous cambial cell division in both years (Table 2).

This is consistent with some previous works that found thermal conditioning drives xylem growth resumption for similar oak species (Pérez-de-Lis et al. 2016). Temperature has been identified as the dominant environmental factor in triggering tree phenology at high altitudes/latitudes (Jiang et al. 2015; Körner 2006). However, some studies challenged this general belief that temperature is the only driver of growth reactivation. Various studies have demonstrated that thermal conditioning (Seo et al. 2008), precipitation (Ren et al. 2015) and/or photoperiod (Maurya and Bhalerao 2017) could determine the onset of cambial activity. On the other hand, an interaction between these environmental factors for the onset of xylogenesis was further reported, such as thermal status and precipitation in cold and drought-prone regions (Ren et al. 2018), and photoperiod and temperature in humid extra-tropical climates (Basler and Körner 2014). However, counterbalancing of thermal and precipitation factors during the onset was not confirmed in our study, as cambial cell division started under different precipitation conditions in the 2 years, probably suggesting weak precipitation influence on the onset of xylogenesis.

Linkage between spring precipitation and FEV formation

The earlier onset but delayed completion of FEV in 2016 led to the synchronous end of formation in 2015 and 2016 (Z = − 0.339, p = 0.734, Table 2). This probably suggests a weak thermal influence on the completion of FEV compared to the onset of cambial cell divisions because different GDD requirements for FEV completion were observed between the 2 years (GDD2015 = 186.9 ± 34.0; GDD2016 = 266.0 ± 50.3; p < 0.05).

In the drought year, Q. variabilis finished FEV formation in mid-April, when precipitation increased and SPEI changed from negative to positive (SPEIMarch 2016 = − 1.9, SPEIApril 2016 = 0.1, Fig. 4). Our results provide support for the hypothesis that the FEV, especially its duration, is sensitive to spring climate conditions since the secondary wall formation of FEV in 2016 completed 1–2 weeks after two moderate rain events (11.6 mm on April 5 and 18.8 mm April 14). As previously reported, to some extent, trees may slow down the growth rate of EV formation according to the moisture condition (Lavrič et al. 2017). Once EV formed, most of them are available to contribute to the water transport (Takahashi et al. 2013). Thus, under warm but severe water-deficit conditions, the earlier formation of EV would add to the risk of drought-induced embolisms in trees.

Besides the duration of FEV formation, the relationship between the diameter of FEV and precipitation has been widely studied. In drought-prone areas, narrow vessels are formed to reduce the risk of cavitation (Borghetti et al. 2017). However, recent work has challenged this paradigm by pointing out that similar or even larger size of EV was found under drought conditions (Gričar et al. 2013; Pérez-de-Lis et al. 2018). The size of vessels under drought stress may be a consequence of the duration and rate in the period of cell enlargement, rather than the period of secondary wall deposition (Cuny et al. 2014). This may explain why the mean diameter of FEV was similar in our study under different spring precipitation conditions. In the warmer spring of 2016, the larger the vessels due to the earlier reactivation, the more effective they are at water transport, but drought stress would impede growth rates, resulting in narrower vessels that are safer in hydraulic conductivity (Pérez-de-Lis et al. 2016). Therefore, vessel size is the trade-off between hydraulic efficiency and hydraulic safety (Zanne et al. 2006), which highlights the potential of vessel features (width/structure) time series for the study of intra-annual variation in climatic conditions (Verheyden et al. 2005).

Impact of drought on the end of xylem formation

As hypothesized, drought conditions in late summer/early autumn caused early completion of cambial cell production, which led to a shorter growing season in 2016. Significantly earlier cessation of cambial cell production and xylem formation in 2016 compared to 2015 (p < 0.05, Table 2) can be related to lower precipitation amounts occurring during late summer 2016. In the environment–xylem growth relationship analysis, precipitation was the main environmental factor driving xylem growth in the drought year (Table 4). Meanwhile, in the moist year, no significant correlation could be detected between precipitation and xylem growth, which may imply that precipitation in 2015 was appropriate for tree growth, without apparent limitations. On the other hand, June was the main growing period, during which 33.5% of xylem growth occurred in rainfall-abundant conditions in 2016 (180.5 mm), in contrast with the 21.6% growth observed in 2015 (95.5 mm).

Many studies have observed that water deficits during summer cause early cessation of cambial activity in conifers and broadleaves trees (Oberhuber et al. 2014; Scharnweber et al. 2011). The environmental factors controlling xylem formation at the beginning and the end of tree growth were different (Thibeault-Martel et al. 2008). The temperature at cessation is obviously still advantageous to growth, which indicates that different physiological mechanisms are involved in the onset and cessation periods. Several studies have revealed that tree growth cessation is mainly controlled by the photoperiod since simultaneous cessations are found along different altitudes (Moser et al. 2010) or in different years (Gruber et al. 2009). However, our results conflict with this synchronous cessation pattern, which likely means that the growth cessation of Q. variabilis is seldom determined by the photoperiod.

With respect to oak phenology, Hill et al. (1995) pointed out that oak trees rely on the stored carbohydrate resources (starch) from the previous year to initiate the radial growth before any leaves are present to produce new photosynthate for growth. Therefore, when under severe summer droughts, oak trees may trigger an earlier cessation of growth in response to decreased carbon gain (Bréda et al. 2006), and to adjust carbon reserves for the following year (Barbaroux and Bréda 2002). The impact of drought on the amount of carbohydrate reserve available for the growth of Chinese cork oak is, however, largely unknown but may also exist since earlier end of xylem growth under summer drought was observed.

High variability of growth among trees in drought year

Higher growth variability among six trees was observed in 2016 (Fig. 3), especially for the growth rate, which may be an adaptation of trees to different stress factors. According to other studies, these between-tree growth differences can be attributable to both abiotic and biotic factors (Lupi et al. 2014). Under drought conditions, water availability strongly affected tree growth and led to considerable variability between the trees (Jochner et al. 2017). We observed a significant correlation between precipitation and xylem growth in the 2016 drought year but not in the 2015 moist year. A similar result was observed by Latreille et al. (2017), who found no significant variability among individuals except for the response to summer precipitation in the current year.

On the other hand, tree genetics, cambial age, social status or neighborhood interaction, are also associated with the among-tree growth variability. As reported, species interaction was considered as a major factor in controlling the treeline dynamics of the Tibetan Plateau (Liang et al. 2016). Compared to intermediate and suppressed trees, dominant trees have a faster growth rate and produce more xylem cells (Rathgeber et al. 2011). In this study, these trees were selected from an even-aged group and were dominant or co-dominant; thus, the effects of cambial age or social status were expected to be very limited. However, the difference in selected trees between the 2 years may partly explain the differences in growth. Younger cambial-age trees were selected in 2016, the year when greater growth variations were shown.

Conclusions

To our knowledge, this is the first study to address the effect of spring precipitation on the formation of the first row of earlywood vessels (FEV) in ring-porous species under a monsoon climate. The results mainly confirmed our hypothesis that the growth rate of FEV in Q. variabilis is probably an adaptive response to water-stress conditions since there was an early onset of FEV formation, but completion was delayed until the spring drought was alleviated. However, the FEV diameter did not significantly differ under different spring precipitation conditions, suggesting that vessel size may be a trade-off between hydraulic efficiency and hydraulic safety. Earlier reactivation under warm temperature is expected to produce wider vessels, which are more efficient, whereas drought stress would impede growth rates, resulting in narrower vessels, which have greater hydraulic safety. Our results suggest that Q. variabilis has the ability to shift the process of earlywood formation under drought. However, to better understand how ring-porous species may respond to climatic instability, longer temporal and spatial monitoring is still needed.

Author contribution statement

JX designed study and started the microcore sampling. YZ did laboratory work. All authors contributed to data analyses, interpretation and writing.