Cross-dating with master tree-ring oxygen isotope chronology
Cellulose δ18O variations of Ashiu cedar is shown in Fig. 2a, and the annual values vary from 25.8‰ to 28.8‰. To date tree-ring formation exactly, we performed cross dating by comparing between the measured cellulose δ18O with the master chronology of central Japan (Nakatsuka et al., 2020). The correlation coefficients were calculated with sliding year by year, and thus the highest value of the correlation coefficient was obtained when the formation year of the outermost ring was A.D. 2017 (r = 0.76), which consistent with the year of tree cutting. We were able to confirm that the outermost ring was formed in A.D. 2017. An age model of measured tree rings was constructed by counting the annual rings with A.D. 2017 as the year of formation of the outermost ring.
Reproducibility of intra-annual isotopic analysis
After dividing annual rings into six segments using an ophthalmic scalpel under a stereomicroscope, we analyzed oxygen isotopic ratios using a mass spectrometer. To evaluate the reproducibility of the procedure, we conducted dupricate analyses. As shown in Fig. 3, the repeated isotopic data agree within error (i.e., 0.20‰ or less), confirming that the partitioning and analysis are reproducible.
Intra-annual isotopic pattern and correlation among six segments of divided annual rings
The variations of cellulose δ18O are presented in Fig. 2a for annual and intra-annual scales. Intra-annual cellulose δ18O vary in the range of 24.1–32.8‰, which is a larger range than the annual variation.
Figure 2b presents a sequentially arranged time series by segment in the annual ring. Intra-annual isotopic pattern have high values early in the growing season; then decline to a seasonal minimum (Fig. 2b). In other Asian sites, cellulose δ18O values are known to be higher near the ring boundary and lower near the center of the annual rings (e.g., Managave et al. (2011) for India; Poussart et al. (2004) for Thailand; Zhu et al. (2012) for Cambodia; Xu et al. (2016) for southeastern China). Our intra-annual pattern documents an isotopic minimum at the final formed part of latewood, implying that the growing period of Japanese cedar at this site is more limited than those of earlier studies.
Table 1 presents the correlation coefficients of cellulose δ18O among six segments of divided annual rings. Significant positive correlations has been identified between adjacent samples in the six divisions of the annual rings. Additionally, for the outermost part of the latewood (i.e., sixth segment), the correlation is weak with those of proceeding period, suggesting that the growth periods are significantly different.
Table 1. Correlations of oxygen isotopic ratios among six segments divided annual rings
Correlation between tree-ring isotopic ratios and meteorological data
We examined the relationship between annual/intra-annual cellulose δ18O and meteorological data during the growing season (i.e., March–August). As shown in Fig. 4, the correlation coefficient is calculated every 30-days for meteorological data, and F, M, L of the horizontal axis respectively denote the first ten days of a month, the middle of a month, and the last ten days of a month. The colors of Fig. 4 depict the results of the respective segments as follows; red for first segment, orange for second segment, yellow for third segment, green for fourth segment, light blue for fifth segment, blue for sixth segment, and violet for the annual ring. The dashed line presents the p-value of 0.01.
From the inner (earlywood) to the outer (latewood) side of six segments, we found a slight shift from May to August in the time of highest inverse correlation with precipitation (Fig. 4a). A similar seasonal progression of correlation peaks was recognized by Xu et al. (2016), and our study clearly confirms this correlation. The isotopic values in the first, second, and third segments show the highest correlation with precipitation from late May to mid-June (Fig. 4a). Those of fourth and fifth segments are shifted gradually by about 10 days (Fig. 4a). The sixth segment is shifted by about 40 days from the fifth, during late July to mid-August (Fig. 4a). Overall the growth period of cedar trees is from April to August (Nanami et al., 2010; Nishizono et al., 2018). Significant correlation between intra-annual cellulose δ18O and precipitation is also observed during the period. It is noteworthy that the peak delay of the sixth segment (the outermost latewood) is consistent with growing later than the other segments.
The annual cellulose δ18O are found to have significant inverse correlations with precipitation during mid-May to early July (purple line of Fig. 4a). Similar correlations with summer precipitation were described in reports of previous studies of eastern and southeastern Asia (e.g., Xu et al., 2011; Sano et al., 2012; Xu et al., 2013; Li et al., 2015; Sakashita et al., 2016; Nakatsuka et al., 2020; Pumijumnong et al., 2020). The results of this study are consistent with those earlier studies. Moreover, compared to the conventional annual isotopic data, intra-annual isotopic data yield significant correlations over a wider period of time (Fig. 4a).
Correlations between cellulose δ18O and relative humidity are summarized in Fig. 5. Because the data of relative humidity are only available since A.D. 1961, correlation coefficients with relative humidity were calculated for 57 years from A.D. 1961 to A.D. 2017 (Fig. 5a). The annual oxygen isotopic ratios showed significant inverse correlations with relative humidity during mid-May to early July (purple of Fig. 5a). On intra-annual data, a tendency can be identified with the highest peaks of the correlations appear in segment order during mid-May to mid-August (Fig. 5a). Accordingly, intra-annual isotopic data obtain significant correlations with relative humidity over a wider period than conventional annual data, as similar to the results of precipitation.
Multiple regression analysis using cedar intra-annual oxygen isotopic ratios
A multiple regression analysis was performed using meteorological data (i.e., precipitation, relative humidity, or air temperature) as objective variables and cedar intra-annual cellulose δ18O as explanatory variables. In multiple regression analysis, multicollinearity is known to occur when correlation among explanatory variables is significant, which can engender inaccurate regression. To avoid multicollinearity, a principal component analysis was performed first for intra-annual cellulose δ18O, and thus six new variables were synthesized as principal components. Multiple regression analysis was then applied with the first, second, and third principal components as explanatory variables, and with precipitation/relative humidity as the objective variable. The sign-reversed multiple correlation coefficients are shown as black lines in Figs. 4a and 5a.
For most of the period, the multiple regression analysis is more accurate than a single regression analysis using the annual cellulose δ18O data (Figs. 4a and 5a). The difference is particularly pronounced for the period after July, with no significant correlation between annual cellulose δ18O and precipitation/relative humidity (purple of Figs. 4a and 5a), whereas the multiple regression yield significant correlations of less than 1% (black of Figs. 4a and 5a). The results therefore suggest that more detailed information related to past precipitation can be obtained by dividing annual rings into six segments.
Temporal change of radial growth phenology
In order to investigate the temporal change on the correlation between tree-ring intra-annual isotopic ratio and meteorological data, we attempted to divide the time series data into several intervals and perform a correlation analysis. The correlation coefficients of two periods (i.e., A.D. 1961–1989 and A.D. 1990–2017) are shown in Figs. 4bc and 5bc.
On the first half of analyzed duration, all segments within the annual ring have highest correlations with precipitation/relative humidity from late May to early July (Figs. 4b and 5b; except sixth segment), suggesting that most segments are prominently influenced by photosynthetic products formed in June.
On the other hand, for second half of analyzed duration, there is a clear difference among the seasons with highest correlation between intra-annual isotopic ratios and meteorological data (Figs. 4c and 5c). This suggests that the photosynthetic products used for each segment are not blended enough and it increases the influence of photosynthetic products formed in the preceding season before June. According to previous studies, heating of cedar stems promotes end of winter dormancy and cambial reactivation (Oribe & Kubo, 1997; Begum et al., 2010), and thus the temperature rise accelerate the initiation of radial growth. Additionally, according to Nishizono et al. (2018), the onset of radial growth of Japanese cedar is strongly influenced by mean annual temperature, with higher mean annual temperatures leading to earlier growth initiation. This study agrees with the results of previous studies as our isotopic records indicate that radial growth starts earlier than June after 1990s with higher annual mean temperatures (Figs. 4c and 5c). In is worth noting that the present study is novel in that it demonstrates the temporal change in radial growth phenology response to ongoing warming based on intra-annual isotopic data.
Cellulose δ18O of the third to fifth intra-annual divisions are characterized by higher correlations with March precipitation (Fig. 4c). According to previous study of ecosystem modeling by Hirano et al. (2021), temperature from winter to spring preceding radial growth affects photosynthesis of Japanese cedar and then the amounts of stored photosynthates, thereby leading to the variation of earlywood width. The results of this study are further consistent with the results of Hirano et al. (2021). It also might be a distinctive feature after 1990s, because there is no significant correlation between isotopic data and March precipitation prior to 1990s (Fig. 4b).
Based on the correlation analysis between intra-annual cellulose δ18O and meteorological data, we discussed the temporal change of radial growth phenology. Temperature rise around 1990 may have change the growth phenology of Japanese cedar in this site. The tree-ring phenology response to ongoing global warming has attracted much attention and investigated by many studies (e.g., Moser et al., 2010; Wang et al., 2015; Rossi et al., 2016; Nishizono et al., 2018), but little is known about temporal variation in the past several decade because of the difficulty to acquire reliable data. Our results points towards the potential that intra-annual isotopic records could be one of useful approach to assess tree-ring phenology response on global warming, because it is widely available across tree species and regions.