Latitudinal variation in radial growth phenology of Cryptomeria japonica D. Don trees in Japan

Latitudinal variation in radial growth phenology of Cryptomeria japonica D. Don trees in Japan Abstract We investigated the latitudinal variation in radial growth phenology (onset, cessation, and duration of radial growth) of Cryptomeria japonica D. Don trees in Japan. The radial growth was observed using dendrometers in 16 stands at multiple latitudes throughout Japan. The onset of radial growth showed a clear latitudinal gradient: at high-latitude sites, C. japonica growth started later than at low-latitude sites. However, cessation of radial growth was independent of latitude. Consequently, the duration of radial growth (defined based on the onset and cessation dates) showed a weak but significant latitudinal gradient: at high-latitude sites, C. japonica trees grew for a shorter period than trees at low-latitude sites. The onset and duration of growth were respectively strongly and weakly affected by temperature, although growth cessation was not explained by temperature or precipitation. The latitudinal gradient for duration of radial growth weakly but significantly supports the hypothesis that regional variation in growth phenology is a key factor responsible for the regional variation in long-term growth. Furthermore, we found that growth duration was more variable at low latitudes than at high latitudes. Introduction Long-term forecasting of timber yield, based on understanding the pattern of age-related change in growth of forest stands, is basic knowledge required to manage forests sustainably. Many studies have been conducted to understand the growth trends in trees and forests and to develop yield prediction tools. In Japan, Cryptomeria japonica D. Don is a major forestry species (Fukuda et al., 2003). Many C. japonica trees have been planted, and C. japonica forests are widely distributed throughout Japan. Therefore, many studies have examined the growth and yield of this species in Japan. Recently, the authors (Nishizono et al., 2013, 2015) reported that the long-term pattern of height and diameter growth of C. japonica at a regional scale could be classified into two regional groups, and that these groups were distributed geographically in a uniform manner. Thus, in high-latitude regions with cold temperatures, C. japonica showed a pattern of late maturity, with slow initial growth but a large maximum size. In low-latitude regions, with warmer temperatures, the species showed a pattern of early maturity, with fast initial growth but a smaller maximum size. In other words, we found a clear latitudinal gradient in long-term growth for C. japonica in Japan. However, it was not clear how the regional difference in long-term growth was generated. Long-term tree growth, spanning periods of several decades to centuries, is a consequence of the accumulation of short-term annual tree growth over long periods. Furthermore, annual tree growth represents the accumulation of short-term daily growth. In temperate and boreal forests, the daily growth shows cyclical seasonal fluctuation within a year (Delpierre et al., 2016b); tree growth typically starts in the spring and ceases in late summer or autumn. (Hereafter, we refer to this seasonal fluctuation as ‘growth phenology’.) By considering the typical climatic variation along a latitude gradient in the northern hemisphere and hypothesizing that growth phenology will be controlled by temperature, we can predict a clear latitudinal gradient for growth phenology; that is, at low-latitude sites with warmer temperatures, growth typically starts earlier and stops later than at high-latitude sites with cooler temperature, resulting in longer growth duration. In fact, studies of xylogenesis (secondary xylem growth) for several conifers, not including C. japonica, showed both such latitudinal gradients (e.g. Jyske et al., 2014) and temperature-dependency of growth phenology (e.g. Rossi et al., 2016). Moreover, recent papers (Delpierre et al., 2016a, b; Guillemot et al., 2017) have indicated that growth phenology is one of the key processes in forest dynamics and should be included as a component in forest ecosystem models to permit accurate and reliable prediction of forest dynamics. On this basis, we hypothesized that the regional variation in growth phenology is an important component of the processes that lead to regional variation in long-term growth. Based on this hypothesis, we predicted that the regional variation in long-term growth and in growth phenology would show similar patterns; specifically, we hypothesized that a latitudinal gradient would exist for growth phenology. The study described in this paper tested this hypothesis for radial growth phenology. The radial growth phenology for C. japonica trees has been described previously (Yamato et al., 1989; Imagawa et al., 1990; Noda et al., 1991; Komiyama et al., 1992; Yamashita et al., 2006; Nanami et al., 2010; Saito et al., 2016). These studies examined the effects of rainfall or temperature on the phenology of a species or on interspecific differences in the phenology based on measurements of the radial growth phenology in a forest stand or several stands within a region. However, no study quantitatively analysed the regional variation in growth phenology on a large spatial scale across multiple regions. The previous studies for high-latitude and low-latitude regions found that radial growth at high-latitude sites (~40°N) started in late April (Imagawa et al., 1990), whereas the radial growth at low-latitude sites (~32°N) started in early March (Yamato et al., 1989; Noda et al., 1991). This implies the existence of a latitudinal gradient in growth phenology of C. japonica trees. However, these previous studies did not adopt a uniform quantitative definition of the components of growth phenology (onset, cessation and duration of growth), making it difficult to quantitatively compare their results. Therefore, experimental data and scientific knowledge of the growth phenology of C. japonica trees are insufficient to let us detect a latitudinal gradient in tree growth phenology on a large spatial scale. For species other than C. japonica, many papers (e.g. Tardif et al., 2001; Rossi et al., 2006, 2008, 2016; Deslauriers et al., 2007, 2008; Mäkinen et al., 2008; Jyske et al., 2014; Delpierre et al., 2016a, b; Guillemot et al., 2017) described radial growth phenology, especially in the northern hemisphere, and examined the effect of several factors (measurement method, temperature, moisture, photoperiod and others) on the phenology. However, only a few studies (e.g. Delpierre et al., 2016a, b; Guillemot et al., 2017) examined the link between regional variations in growth phenology and forest growth dynamics. In addition, because the latitudinal gradients (e.g. Jyske et al., 2014) and temperature-dependency (e.g. Rossi et al., 2016) of growth phenology have been mainly examined in high-latitude areas with cool temperatures, we have little information on the gradients in low-latitude areas with warm temperatures, as is the case for much of Japan. In the present study, our goal was to provide some of the missing information. We measured and quantitatively analysed the radial growth phenology of C. japonica trees at 16 sites ranging from southern to northern Japan. Based on the measurement data, we examined the following questions: (1) Is there a latitudinal gradient, analogous to the gradient found for long-term growth, in the radial growth phenology of C. japonica trees in Japan? (2) If such a pattern exists, are temperature and precipitation responsible for the observed variation? The answers to these questions will improve our understanding of the link between growth phenology and long-term growth patterns, as well as the likely effects of climate on tree growth, and will provide the basic information required for long-term forecasting of timber yield. Methods Growth data Study sites and measurement methods We collected radial growth phenology data from 16 C. japonica stands for 1–4 years (Figure 1). Stands were chosen to provide coverage of most of Japan. Table 1 summarizes the characteristics of the stands and measurements. This sample included 15 single-storey stands and 1 two-storey stand (stand 4, Iwate 3). For our measurements, we used both automatic dendrometers (for 3 stands) and manual dendrometers (for 13 stands). We installed the dendrometers on 5–10 trees for each stand, at a height of ~1.2 m. Using the automatic dendrometer (Type DC2, Ecomatik, Munich, Germany; thermal elongation coefficient 1.4 × 10−6 K−1), we measured and recorded the stem circumference at an interval of 30 min. Using the manual dendrometer, we measured the stem circumference or diameter at breast height (DBH) by visually reading the scale, mainly at intervals of 7 days (growing season) or 15 days (winter). Of the 13 stands measured with the manual dendrometer, we used a commercially available band dendrometer made from Astralon plastic (Type D1, UMS GmbH, Munich, Germany; thermal elongation coefficient 75 × 10−6 K−1) to measure DBH in one stand (7, Toyama). In the other 12 stands, we used home-made stainless-steel dendrometers (thermal elongation coefficient 11.5 × 10−6 K−1) to measure the stem circumference. For some of the automatic dendrometer measurements (one for the Hokkaido site in 2016, two and one for the Kochi site in 2015 and 2016, and one for the Miyazaki site in 2016), the measurement failed for various reasons, including damage to the sensor caused by a falling branch, tree growth reaching the maximum measurable size of the dendrometer, or unknown reasons. Therefore, we excluded the data obtained from these dendrometers from our subsequent analyses. Consequently, the number of sample trees used for the analyses were 5 and 4 trees for the Hokkaido site in 2015 and 2016, 3 and 4 trees for the Kochi site in 2015 and 2016 (respectively), and 4 trees for the Miyazaki site in 2016. Table 1 Summary of the studied forests and the measured trees Site no.  Site name  Site property  Stand attributes3  Sample tree3  Measurement data  Latitude (°N)  Longitude (°E)  Elevation (m a.s.l.)  Annual mean temperature1 (°C)  Annual precipitation1 (mm)  Mean maximum snow depth1 (cm)  Age (years)  Stem density (trees/ha)  Sample size  Mean DBH (cm)  Mean height (m)  First date (year/month/day)  Last date (year/month/day)  Interval between measurements  Dendrometer material and data reading method  1  Hokkaido  41.846  140.733  65  8.4  1170.6  43  44  1285  5  29.2  22.0  2015/3/13  2016/11/1  30 min  Steel, Automatic  2  Iwate 1  39.770  141.140  207  9.3  1261.1  41  38  2026  10  24.8  20.2  2009/4/6  2009/12/24  7 days (15 days for winter season)  Steel, Manual  3  Iwate 2  39.790  141.153  269  9.1  1223  41  109  648  10  45.9  29.7  2009/4/9  2009/12/25  7 days (15 days for winter season)  Steel, Manual  4  Iwate 32  39.792  141.156  216  9.1  1227.8  39  109  157  10  48.4  28.3  2009/4/9  2009/12/25  7 days (15 days for winter season)  Steel, Manual  18  666  10  9.9  9.0  5  Yamagata 1  38.937  140.264  168  9.6  2249.3  148  40  2046  10  27.9  23.4  2009/4/16  2009/12/24  15 days  Steel, Manual  6  Yamagata 2  38.939  140.265  168  9.6  2249.3  148  96  769  10  37.5  27.5  2009/4/16  2009/12/24  15 days  Steel, Manual  7  Toyama  36.305  137.329  227  10.4  1895.3  86  26  885  10  25.1  16.4  2013/1/11  2013/11/14  15 days  Plastic, Manual  8  Ibaraki 1  36.184  140.217  35  13.5  1184.2  6  19  2141  8  14.1  10.8  2014/1/17  2015/9/15  7 days for growing season (more than 15 days for winter season)  Steel, Manual  9  Ibaraki 2  36.008  140.133  21  13.6  1223.1  5  36  1373  10  30.3  19.9  2013/2/28  2016/12/28  7 days  Steel, Manual  10  Chiba 1  35.192  140.144  222  13.7  1897.9  3  39  2306  10  24.4  18.3  2015/1/30  2016/12/26  15 days  Steel, Manual  11  Chiba 2  35.192  140.144  227  13.7  1897.9  3  40  3695  10  20.7  18.0  2015/2/6  2016/12/26  15 days  Steel, Manual  12  Chiba 3  35.171  140.165  299  13.2  1902.9  3  24  1927  10  22.1  16.7  2016/1/25  2016/12/26  15 days  Steel, Manual  13  Chiba 4  35.167  140.161  311  13.3  1896.4  3  71  1276  10  42.0  26.4  2016/2/10  2016/12/26  15 days  Steel, Manual  14  Chiba 5  35.161  140.146  293  13.3  1920.2  3  110  1259  10  45.8  23.9  2016/2/10  2016/12/26  15 days  Steel, Manual  15  Kochi  33.541  133.478  47  16.1  2550  1  46  1140  5  35.4  24.7  2015/1/1  2016/11/16  30 min  Steel, Automatic  16  Miyazaki  31.867  131.302  193  15.8  2563.7  1  46  936  5  31.1  21.6  2016/1/1  2016/12/6  30 min  Steel, Automatic  Site no.  Site name  Site property  Stand attributes3  Sample tree3  Measurement data  Latitude (°N)  Longitude (°E)  Elevation (m a.s.l.)  Annual mean temperature1 (°C)  Annual precipitation1 (mm)  Mean maximum snow depth1 (cm)  Age (years)  Stem density (trees/ha)  Sample size  Mean DBH (cm)  Mean height (m)  First date (year/month/day)  Last date (year/month/day)  Interval between measurements  Dendrometer material and data reading method  1  Hokkaido  41.846  140.733  65  8.4  1170.6  43  44  1285  5  29.2  22.0  2015/3/13  2016/11/1  30 min  Steel, Automatic  2  Iwate 1  39.770  141.140  207  9.3  1261.1  41  38  2026  10  24.8  20.2  2009/4/6  2009/12/24  7 days (15 days for winter season)  Steel, Manual  3  Iwate 2  39.790  141.153  269  9.1  1223  41  109  648  10  45.9  29.7  2009/4/9  2009/12/25  7 days (15 days for winter season)  Steel, Manual  4  Iwate 32  39.792  141.156  216  9.1  1227.8  39  109  157  10  48.4  28.3  2009/4/9  2009/12/25  7 days (15 days for winter season)  Steel, Manual  18  666  10  9.9  9.0  5  Yamagata 1  38.937  140.264  168  9.6  2249.3  148  40  2046  10  27.9  23.4  2009/4/16  2009/12/24  15 days  Steel, Manual  6  Yamagata 2  38.939  140.265  168  9.6  2249.3  148  96  769  10  37.5  27.5  2009/4/16  2009/12/24  15 days  Steel, Manual  7  Toyama  36.305  137.329  227  10.4  1895.3  86  26  885  10  25.1  16.4  2013/1/11  2013/11/14  15 days  Plastic, Manual  8  Ibaraki 1  36.184  140.217  35  13.5  1184.2  6  19  2141  8  14.1  10.8  2014/1/17  2015/9/15  7 days for growing season (more than 15 days for winter season)  Steel, Manual  9  Ibaraki 2  36.008  140.133  21  13.6  1223.1  5  36  1373  10  30.3  19.9  2013/2/28  2016/12/28  7 days  Steel, Manual  10  Chiba 1  35.192  140.144  222  13.7  1897.9  3  39  2306  10  24.4  18.3  2015/1/30  2016/12/26  15 days  Steel, Manual  11  Chiba 2  35.192  140.144  227  13.7  1897.9  3  40  3695  10  20.7  18.0  2015/2/6  2016/12/26  15 days  Steel, Manual  12  Chiba 3  35.171  140.165  299  13.2  1902.9  3  24  1927  10  22.1  16.7  2016/1/25  2016/12/26  15 days  Steel, Manual  13  Chiba 4  35.167  140.161  311  13.3  1896.4  3  71  1276  10  42.0  26.4  2016/2/10  2016/12/26  15 days  Steel, Manual  14  Chiba 5  35.161  140.146  293  13.3  1920.2  3  110  1259  10  45.8  23.9  2016/2/10  2016/12/26  15 days  Steel, Manual  15  Kochi  33.541  133.478  47  16.1  2550  1  46  1140  5  35.4  24.7  2015/1/1  2016/11/16  30 min  Steel, Automatic  16  Miyazaki  31.867  131.302  193  15.8  2563.7  1  46  936  5  31.1  21.6  2016/1/1  2016/12/6  30 min  Steel, Automatic  1The values were obtained from the mesh climate data provided by the Japan Meteorological Agency (2002), which provides monthly means from 1971 to 2000 throughout Japan at a spatial resolution of ca. 1 km. The ‘mean maximum snow depth’ indicates the average value of the maximum snow depth in each year from 1971 to 2000. 2The Iwate 3 plot was in a two-storey forest stand. The upper and lower rows of data show the attributes of trees in the upper and lower stories of the stand, respectively. 3The values were obtained at the beginning of the measurements. Figure 1 View largeDownload slide Location of the 16 Cryptomeria japonica study sites. Numbers correspond to the site numbers in Table 1. Figure 1 View largeDownload slide Location of the 16 Cryptomeria japonica study sites. Numbers correspond to the site numbers in Table 1. Data processing First, we converted the stem circumference into DBH based on the assumption that the stem cross-section was circular. Second, we applied a temperature correction to all DBH data using the thermal elongation coefficient of the dendrometer band and mean daily or hourly temperatures obtained from the Japan Meteorological Agency (see the next section for details). Third, for the measurements obtained using the automatic dendrometer, we calculated the daily mean DBH using only DBH values measured between 8:00 a.m. and 4:00 p.m., which is the range of times when the measurements using the manual dendrometer were usually conducted. Excluding the night data should reduce the bias compared with the manual dendrometers due to differences in the time of day used for the measurement. Finally, we converted the measurement date into day of the year (DOY) values for the data from each year. We used the relationship between DBH and DOY for each sample tree in our analyses of the radial growth phenology. Meteorological data We obtained daily or hourly meteorological data (temperature and precipitation) during the measurement period for each test site from the nearest meteorological station. Meteorological data for each station were obtained from the Japan Meteorological Agency (http://www.data.jma.go.jp/gmd/risk/obsdl/index.php). The temperature was corrected for elevation differences using a standard temperature lapse rate of −0.60°C per 100 m in elevation. Analysis method Quantifying growth phenology We used three indices to quantify the growth phenology: the onset time (To), cessation time (Tc), and duration (Pg) of the radial growth. We calculated these indices for each plot and year using a growth curve at a stand level fitted to the DBH–DOY relationships for the trees in each plot. In many previous studies of growth phenology, the researchers used a sigmoidal growth function (e.g. the Gompertz function; Jyske et al., 2014) for this fitting. However, in the present study we used a more flexible spline curve for the fitting because the sigmoidal function was not applicable to the bimodal growth pattern, with growth peaks in the spring and autumn, exhibited by trees growing in a Mediterranean climate (e.g. Vieira et al., 2014) or in south-western Japan (e.g. Noda et al., 1991; Kawasaki and Takeuchi, 1993). To fit the spline curve, we applied a smoothing-splines mixed-effects model, with DOY as the fixed factor and individual trees as the random factor, to define the DBH–DOY relationships for each year and plot. We conducted the fitting using the sme function of the sme package (Berk, 2013) for the R software (http://www.R-project.org/). For the smoothing parameters, we obtained the optimal values based on the value of Akaike’s information criterion (AIC) using a Nelder-Mead search (Berk, 2013). To obtain a suitable fitting, we used trial and error to transform the variables and calibrate the tolerance. Because our data were obtained by repeated measurements from the same individuals, the data had non-independent errors. This violates the important requirement of independence of error for standard statistical analysis, indicating that standard methods of analysis could not be applied to our data (Crawley, 2007). However, a mixed-effects model can account for such temporal pseudoreplication (Crawley, 2007), and allowed us to model the average pattern of radial growth phenology at a stand level. To obtain the three phenology indices from the fitted curve, we first calculated the annual growth of the average tree in the stand as the maximum DBH minus the minimum DBH from the fitted curve at a stand level for a given year. Next, we calculated To and Tc as the DOY when the tree attained 10 and 90 per cent, respectively, of the total annual growth, as defined by Nanami et al. (2010). Finally, we calculated Pg as Tc – To (i.e. the difference between the onset and cessation times). Based on our preliminary assessment of the fitting curves (Figure 2), DBH in several stands (e.g. stands 3, 5 and 6; Figure 2B) decreased during the period between dendrometer installation and the onset of radial growth. This decrease may have resulted from incorrect measurements due to slack after installation of the dendrometer band (Drew and Downes, 2009). Therefore, the observed onset date for these stands may be later than the actual value. To confirm the reliability of the observed onset, we also measured the DBH change for these stands during the spring of the next year. Comparison of the resulting graphs indicated that the onset of radial growth in the next growing season appeared similar to the observed onset for the first growing season in these stands (results not shown). Therefore, we conclude that the onset dates were reliable for these stands. In the Hokkaido stand (no. 1), stem shrinkage was probably induced by low temperatures (Cocozza et al., 2009), which occurred during the cold winter season (DOY < 70, with daily mean temperatures less than ~0°C) in 2016 (Figure 2B). This shrinkage resulted in an unreliable estimate of the phenology indices. We therefore excluded the part of the fitted curve for this period (DOY < 70) from the analysis in our calculation of the indices. Figure 2 View largeDownload slide Fitted spline curves for the DBH changes at the 16 study sites. DOY represents the day of the year (from 1 January). The DBH change was calculated as D(t) – Dmin, where D(t) is the DBH (cm) on DOY t (days), and Dmin is the minimum DBH in a stand in each year. Text in the legend represents the site number (1–16 in Table 1) followed by the measurement year; for stand 4, U represents the upper layer of trees in the two-storey stand and L represents the lower layer. Figure 2 View largeDownload slide Fitted spline curves for the DBH changes at the 16 study sites. DOY represents the day of the year (from 1 January). The DBH change was calculated as D(t) – Dmin, where D(t) is the DBH (cm) on DOY t (days), and Dmin is the minimum DBH in a stand in each year. Text in the legend represents the site number (1–16 in Table 1) followed by the measurement year; for stand 4, U represents the upper layer of trees in the two-storey stand and L represents the lower layer. Effects of climate, latitude, tree size and stand attributes on the growth phenology To determine which factors affected the radial growth phenology, we conducted single regression analysis between the growth phenology indices and individual variables using a linear mixed-effects model (LMM). In our dataset, the locations of the stands were somewhat clustered (Figure 1). To account for the geographical clustering and the resulting spatial pseudoreplication (Crawley, 2007), the LMM had the following form:   yijk=β0+βxijk+αi+αij+εijkαi≅N(0,σα12)αij≅N(0,σα22)εijk≅N(0,σεΛij) (1)where yijk is the phenology index in plot j (nested within location i) and year k; β0 is the intercept; β is the slope coefficient for the fixed effect; xijk is the independent climatic, geographical, or other variable (defined later in this section) in plot j (nested within location i) and year k; εijk is an error term for plot j (nested within location i) and year k; and αi and αij are random effects terms for location i and plot j (nested within the location). We assumed normality and homogeneity of variance (σα12 and σα22) for αi and αij. We used the eight prefectures (described in ‘Site name’ of Table 1: Hokkaido, Iwate, Yamagata, Toyama, Ibaraki, Chiba, Kochi and Miyazaki) as the location term. Based on examination of the standardized residuals, we defined the variance structure (σεΛij) for To, Tc and Pg. Examination of the standardized residuals revealed a homoscedastic error structure (σε2) for To and Tc, and the variance structure could be described using an exponential function of mean daily precipitation to correct for heteroscedasticity for Pg. We tested the significance of the fixed-effects term (β) using a conditional t-test (Pinheiro and Bates, 2000), and defined significance as P < 0.05. We also calculated the marginal R2GLMM (Nakagawa and Schielzeth, 2013) to examine the explanatory power of the independent variable. The marginal R2GLMM represents the proportion of variance in the response variable that is explained entirely by the fixed effect. It ranges from 0 to 1, which represent no fit and the best possible fit, respectively. We calculated the annual and quarterly (January to March, April to June, July to September, and October to December) means of the mean daily temperature and daily precipitation during the measurement year for each stand. We used these annual and quarterly means as explanatory variables in the regression analysis. In addition, we used the mean initial sizes of the measured trees for each stand (mean DBH and mean height), initial stand age, initial stem density and the latitude of each stand as explanatory variables in the regression analysis. We excluded data from the lower storey in the Iwate 3 stand from our analyses of stem density vs. phenology indices because the trees in the lower storey must be affected by trees in the upper storey; it was therefore difficult to quantify the stand density for the lower storey on the same basis as stem density in the single-storey stands. The fitting was performed using the lme function of the NLME package (Pinheiro and Bates, 2000) for the R software. The marginal R2GLMM was calculated by using the r.squaredGLMM function of the MuMIn package (Barton, 2016) for the R software. Results Radial growth phenology Overall, the DBH change (the difference between DBH on a given DOY and the minimum DBH during the year) increased with increasing DOY, and then stabilized (Figure 2). The DBH growth rate (cm/day) increased with increasing DOY after onset of growth in most plots (Figure 2B–D), and then decreased to almost zero by the summer, although some plots showed fluctuations of DBH change (without a significant increase) after the summer. However, these fluctuations may have been an artefact of the use of splines for the models, and may not have biological significance. However, the DBH change showed a different pattern in several plots (Figure 2A). The DBH growth rate in these plots first increased, then decreased to near zero and remained at that level until mid-summer, but from late summer until autumn, increased again with increasing DOY. We estimated the phenology indices for 25 curves (16 stands in at least 1 year, more than 1 year for some stands, and the understorey and overstorey curves in stand 4). The estimated To values ranged from DOY 81 to 139, and averaged DOY 107.8. The estimated Tc values ranged from DOY 155 to 233, and averaged DOY 195.7. The estimated Pg values ranged from 51 to 144 days, and averaged 88.0 days. Latitudinal variation and gradient in growth phenology Higher-latitude sites had a significantly larger To (Figure 3, Table 2). Thus, radial growth started later at higher latitudes than at lower latitudes. R2GLMM was 0.719, indicating a model with strong explanatory power for the effect of latitude on To. Latitude did not significantly affect Tc (Figure 3, Table 2). Thus, the relationship between Tc and latitude is unclear. Higher latitude sites had a significantly smaller Pg (Figure 3, Table 2). Thus, the duration of radial growth decreased with increasing latitude. R2GLMM was 0.338, indicating a model with significant but weak explanatory power for the effect of latitude on Pg. Table 2 Parameter estimates and statistics for the linear mixed-effects model using equation [1]1,2 Explanatory variables  Timing of onset (DOY)  Timing of cessation (DOY)  Growth duration (days)  β0  β  P  R2GLMM  β0  β  P  R2GLMM  β0  β  P  R2GLMM  Latitude (°N)  −109.3910  5.9810  <0.001  0.719  175.150  0.6457  0.796  0.006  249.972  −4.4786  0.031  0.338  Mean daily mean temperature (°C)   Whole year  192.8064  −6.2552  <0.001  0.789  224.138  −1.9913  0.354  0.058  53.737  2.3516  0.201  0.112   Jan.–Mar.  129.8004  −5.4690  <0.001  0.843  205.234  −2.6285  0.095  0.146  79.259  1.3655  0.300  0.057   Apr.–June  221.4001  −6.8680  <0.001  0.742  219.390  −1.2874  0.597  0.018  29.510  3.4108  0.092  0.167   July–Sep.  279.6027  −7.3514  <0.001  0.723  191.785  0.3235  0.885  0.001  1.476  3.6562  0.041  0.175   Oct.–Dec.  157.5543  −4.7811  <0.001  0.704  211.177  −1.2889  0.396  0.034  97.892  −1.6176  0.027  0.058  Mean daily precipitation (mm)   Whole year  107.2507  0.5650  0.838  0.003  176.976  3.7260  0.148  0.095  73.932  1.8848  0.412  0.041   Jan.–Mar.  97.6598  2.7955  0.294  0.084  188.665  2.2634  0.341  0.045  83.759  −0.1125  0.948  0.000   Apr.–June  121.1767  −1.6701  0.220  0.060  196.878  0.3814  0.710  0.002  78.738  0.9411  0.352  0.018   July–Sep.  103.6441  0.8449  0.301  0.013  192.321  0.8361  0.204  0.013  82.434  0.1267  0.818  0.000   Oct.–Dec.  120.9711  −1.9121  0.282  0.029  203.960  −0.9060  0.502  0.006  82.698  0.1406  0.880  0.000  Mean size of sample trees   DBH (cm)  109.3901  0.0460  0.761  0.000  188.685  0.3594  0.372  0.023  79.255  0.1332  0.773  0.005   Tree height (m)  108.5475  0.1054  0.741  0.001  186.831  0.6007  0.469  0.019  83.565  −0.0122  0.989  0.000  Stand attributes   Age (years)  109.4301  0.0277  0.630  0.001  197.073  0.0416  0.784  0.003  86.146  −0.0514  0.756  0.006   Stem density (trees/ha)  110.9242  0.0001  0.958  0.000  215.097  −0.0124  0.085  0.226  85.234  −0.0010  0.882  0.002  Explanatory variables  Timing of onset (DOY)  Timing of cessation (DOY)  Growth duration (days)  β0  β  P  R2GLMM  β0  β  P  R2GLMM  β0  β  P  R2GLMM  Latitude (°N)  −109.3910  5.9810  <0.001  0.719  175.150  0.6457  0.796  0.006  249.972  −4.4786  0.031  0.338  Mean daily mean temperature (°C)   Whole year  192.8064  −6.2552  <0.001  0.789  224.138  −1.9913  0.354  0.058  53.737  2.3516  0.201  0.112   Jan.–Mar.  129.8004  −5.4690  <0.001  0.843  205.234  −2.6285  0.095  0.146  79.259  1.3655  0.300  0.057   Apr.–June  221.4001  −6.8680  <0.001  0.742  219.390  −1.2874  0.597  0.018  29.510  3.4108  0.092  0.167   July–Sep.  279.6027  −7.3514  <0.001  0.723  191.785  0.3235  0.885  0.001  1.476  3.6562  0.041  0.175   Oct.–Dec.  157.5543  −4.7811  <0.001  0.704  211.177  −1.2889  0.396  0.034  97.892  −1.6176  0.027  0.058  Mean daily precipitation (mm)   Whole year  107.2507  0.5650  0.838  0.003  176.976  3.7260  0.148  0.095  73.932  1.8848  0.412  0.041   Jan.–Mar.  97.6598  2.7955  0.294  0.084  188.665  2.2634  0.341  0.045  83.759  −0.1125  0.948  0.000   Apr.–June  121.1767  −1.6701  0.220  0.060  196.878  0.3814  0.710  0.002  78.738  0.9411  0.352  0.018   July–Sep.  103.6441  0.8449  0.301  0.013  192.321  0.8361  0.204  0.013  82.434  0.1267  0.818  0.000   Oct.–Dec.  120.9711  −1.9121  0.282  0.029  203.960  −0.9060  0.502  0.006  82.698  0.1406  0.880  0.000  Mean size of sample trees   DBH (cm)  109.3901  0.0460  0.761  0.000  188.685  0.3594  0.372  0.023  79.255  0.1332  0.773  0.005   Tree height (m)  108.5475  0.1054  0.741  0.001  186.831  0.6007  0.469  0.019  83.565  −0.0122  0.989  0.000  Stand attributes   Age (years)  109.4301  0.0277  0.630  0.001  197.073  0.0416  0.784  0.003  86.146  −0.0514  0.756  0.006   Stem density (trees/ha)  110.9242  0.0001  0.958  0.000  215.097  −0.0124  0.085  0.226  85.234  −0.0010  0.882  0.002  DOY, day of year. 1Grey areas indicate statistically significant relationships. 2As a goodness-of-fit statistic for the mixed-effects model, we used the marginal R2GLMM statistic (Nakagawa and Schielzeth, 2013), which represents the proportion of the variance explained exclusively by the fixed effect. Figure 3 View largeDownload slide Relationships between latitude and the onset, cessation and duration of radial growth (DOY, day of year). Text in the legend represents the site number (1–16 in Table 1); for stand 4, U represents the upper layer of trees in the two-storey stand and L represents the lower layer. Figure 3 View largeDownload slide Relationships between latitude and the onset, cessation and duration of radial growth (DOY, day of year). Text in the legend represents the site number (1–16 in Table 1); for stand 4, U represents the upper layer of trees in the two-storey stand and L represents the lower layer. Effects of climatic conditions, tree size and stand attributes on growth phenology Higher temperatures, both for the year as a whole and for the four periods, were significantly negatively associated with To (Figure 4, Table 2); that is, warmer mean temperatures resulted in earlier onset of growth. R2GLMM ranged from 0.704 to 0.843, indicating a model with strong explanatory power for the effect of temperature on To. Precipitation did not significantly affect To (Table 2). These results indicated that temperature was the climatic variable with the strongest effect on To and could by itself explain most of the variation in To. Figure 4 View largeDownload slide Relationships between the mean daily mean temperature for the whole year and the onset, cessation, and duration of radial growth. Text in the legend represents the site number (1–16 in Table 1); for stand 4, U represents the upper layer of trees in the two-storey stand and L represents the lower layer. Figure 4 View largeDownload slide Relationships between the mean daily mean temperature for the whole year and the onset, cessation, and duration of radial growth. Text in the legend represents the site number (1–16 in Table 1); for stand 4, U represents the upper layer of trees in the two-storey stand and L represents the lower layer. Neither temperature nor precipitation significantly affected Tc (Figure 4, Table 2). Higher mean temperatures from July to September and from October to December were significantly positively and negatively, respectively, associated with Pg; that is, warmer temperatures in these months increased and decreased, respectively, the growth duration (Table 2). For mean temperature from July to September, the R2GLMM values for these means were 0.175, indicating a model with significant but weak explanatory power for the effect of temperature on Pg. However, for mean temperature from October to December, the R2GLMM values for these means were 0.058, indicating a negligible effect. Precipitation was not significantly associated with Pg (Table 2). Thus, at warm sites, the growth duration is longer than at cold sites. The size parameters (DBH and height) for the sample trees, stand age, and stem density were not significantly associated with To, Tc or Pg (Table 2). These results indicate that tree size, stand age and stand density did not affect the observed variation of growth phenology. Discussion Latitudinal gradient in growth phenology We found a clear latitudinal gradient for the onset of radial growth of C. japonica trees (Table 2, Figure 3). At high-latitude sites, radial growth started later than at low-latitude sites. This association was strong and could explain most (71.9 per cent) of the variation in onset time. However, we found no significant latitudinal gradient for the cessation of radial growth. Thus, latitude could not explain the observed variation in cessation time. Furthermore, we found a latitudinal gradient for the duration of radial growth: the duration of radial growth was shorter at high-latitude sites than at low-latitude sites. The latitude was weakly but significantly associated with the growth duration, as it explained only 33.8 per cent of the variation in the growth duration. This study provides the first quantitative evidence of latitudinal gradients in the onset and the duration of growth for C. japonica, although such latitudinal gradients have been found for other conifers (e.g. Jyske et al., 2014). The C. japonica trees in stands at higher latitudes had a late onset time, leading to a short growth duration (Figures 2 and 4). For the C. japonica trees in stands at lower latitudes, the onset times were early for all stands, but the growth duration varied widely within the stands. The growth duration was longest for the C. japonica trees in some stands at lower latitudes, which continued to grow until autumn. On the other hand, the growth duration was shortest for the trees of some stands at lower latitudes, which stopped growing by mid-May. These results show that C. japonica trees at low latitudes have variable growth duration, whereas those at high latitudes have a relatively uniform growth duration. Effects of temperature on growth phenology Temperature clearly affected the radial growth phenology (Table 2, Figure 4), whereas precipitation had no impact. Furthermore, age, tree size and stem density had no impact. These results show that the effects of environment (temperature) were much stronger than the effects of ontogeny and forest management. Therefore, in the following discussion, we will focus mainly on the effects of temperature. Previous studies (Oribe and Kubo, 1997; Begum et al., 2010) showed that heating of the stem breaks the winter cambial dormancy of C. japonica, and results in cambial reactivation. Thus, rising temperatures trigger the onset of radial growth. In addition, Rossi et al. (2016) showed that the onset of growth became earlier with increasing mean annual temperature for coniferous forest stands in the northern hemisphere. In the present study, the onset of radial growth was strongly associated with the mean temperature at the site (Table 2, Figure 4). Thus, at sites with higher mean temperatures, radial growth started earlier than at sites with lower mean temperatures. These results agreed with the previous findings. Furthermore, because our study included data from stands at lower latitudes and lower elevations than the data analysed by Rossi et al. (2016), the maximum temperature in the range that we analysed was higher than that analysed by Rossi et al. (2016). However, Rossi et al. (2016) included sites with colder temperatures than our coldest sites. Therefore, our study extended the previous finding (the association between higher temperature and earlier onset) to regions with higher temperatures. On the other hand, the factor(s) that determined the date of cessation of radial growth is unclear (Delpierre et al., 2016b). Some studies (Rossi et al., 2007, 2016) reported that climate factors affected the cessation of growth. For example, Rossi et al. (2007, 2016) reported that temperature triggers the cessation of growth. Additionally, some researchers noted that drought (e.g. Gruber et al., 2010; Eilmann et al., 2011) or photoperiod (Ford et al., 2017) were linked with the timing of cessation. However, other studies reported that the cessation of growth is more strongly affected by the total number of xylem cells produced in a season than by climatic factors (Lupi et al., 2010), and is only weakly explained by climatic factors (Duchesne et al., 2012). In the present study (Table 2, Figure 4), the cessation of radial growth was not significantly associated with the temperature and precipitation variables. This was because the range of cessation dates mostly overlapped for the stands in high- and low-temperature areas (Figure 4). These results indicated that the growth cessation was not triggered solely by these two climatic factors. Rossi et al. (2016), who had a large sample size that included trees from sites with a wide range of temperature, showed that spring (onset-related) and autumn (cessation-related) phenological events became earlier and later, respectively, with increasing mean temperature. Simultaneously, they found that mean temperature explained a smaller part of the overall variation in cessation date than of the variation in onset date. These previous findings suggest that temperature has a weak effect on the cessation of growth, and that such an effect would be difficult to detect with a small sample size or a narrow temperature gradient; thus, other factors are likely to affect the cessation of growth by acting together with temperature (Delpierre et al., 2016b). For example, Ford et al. (2017) reported that high temperature under a long photoperiod in the early summer was linked with early growth cessation. Based on these previous and present results, we hypothesize that the cessation of radial growth is determined by multiple factors acting together in a complicated mechanism, and that this complexity explains the unclear relationships we observed between the cessation of radial growth and the climatic factors and latitude in the present study. The mean temperatures from July to September had a weak but significant positive effect on the duration of growth, and explained a small but significant portion of the variation in the duration (Table 2). These results show that the duration of growth is longer at warm sites than at cold sites. The present study provides the first quantitative evidence of temperature-dependency in the duration of growth for C. japonica trees, although such temperature-dependency has been found for other conifers (e.g. Rossi et al., 2016). Furthermore, our study extended the previous finding (the association between higher temperature and longer duration) to regions with higher temperatures. Cryptomeria japonica trees in high-temperature areas had highly variable growth duration, whereas those in low-temperature areas had a more uniform growth duration (Figure 4). This low degree of variation of growth phenology in low-temperature (high-latitude) areas may be attributed to the severe climatic conditions in winter. Woody plants undergo a cold-acclimation process to increase their freezing tolerance in winter (Sakai and Larcher, 1987). The first step in this process is growth cessation (Horiuchi and Sakai, 1973; Sakai and Larcher, 1987). Therefore, to survive the winter, C. japonica trees at high-latitude sites with cold conditions may need to cease growth earlier to avoid freezing damage. This adaptation to cold conditions would lead to an earlier upper limit on the cessation of growth at high latitudes than at low latitudes. In addition, actively growing tissues are more vulnerable to cold temperatures (Sakai and Larcher, 1987), which occur frequently during the spring. Avoidance of these temperatures may explain the delayed onset of radial growth at high latitudes (Figure 4). Therefore, the onset and cessation dates in high-latitude areas necessarily fall in the narrow range between the later lower limit and earlier upper limit. This low variation would increase the number of trees that survive the severe winters in high-latitude areas. In contrast, the high variation of growth phenology in high-temperature areas can be attributed to the mild climatic conditions. The warmer temperatures potentially permit earlier onset and later cessation of growth, thereby leading to a longer growth duration. However, some site-specific factors other than temperature, such as soil conditions (e.g. moisture, nutrients), may limit growth, resulting in an earlier cessation and a shorter duration than the potential values based only on temperature. In addition, a shorter growth duration will not necessarily be linked to tree mortality if the trees can ensure continuous growth, even if only a small amount of growth. Thus, the mild climate permits greater variation in the growth phenology of C. japonica trees in low-latitude areas. Based on this discussion, latitudinal variation in radial growth phenology appears to originate from the temperature-dependence of growth, particularly for onset and somewhat for duration. However, temperatures will exhibit potentially high inter-annual variability. In addition, factors other than temperature (e.g. drought) also have high inter-annual variability. These variations may therefore affect the duration of growth (e.g. Gruber et al., 2010; Eilmann et al., 2011). Therefore, researchers should note that the year selected for their measurements may affect the relationship of latitude to radial growth phenology. In addition, It should be noted that several papers (e.g. Yamashita et al., 2006; Deslauriers et al., 2007; Mäkinen et al., 2008) have noted a discrepancy between the growth phenology inferred from dendrometer measurements and microscopic observations of wood formation. For C. japonica, Yamashita et al. (2006) reported that the onset of growth was similar, but the cessation of growth was different, between these two methods. This uncertainty of determining the cessation of growth using different methods may have affected the relationships we observed between the cessation of radial growth and the climatic factors and latitude in the present study, because we adopted dendrometer measurements rather than more accurate microscopic methods. Therefore, in the future, we should re-assess the phenology data by means of microscopic examination using micro-core sampling, as this approach probably provides the most reliable estimate of growth phenology, although the method has not (to our knowledge) been applied to C. japonica trees. Relationship between growth phenology and long-term growth patterns In this study, we hypothesized that the latitudinal gradient that exists for long-term growth might be caused by a parallel gradient in growth phenology. We tested this hypothesis and found significant latitudinal gradients for the onset and duration of radial growth (Table 2, Figure 3). In our previous research on long-term growth (Nishizono et al., 2013, 2015), we found a pattern of late maturity combined with slow initial growth and a large maximum size in high-latitude areas, and a pattern of early maturity combined with fast initial growth and a small maximum size in low-latitude areas. In the present study, we found a pattern of short growth duration combined with late onset in high-latitude areas and a pattern of long duration combined with early onset in low-latitude areas. However, we found no significant effects of age (ontogeny) on the latitudinal gradient (Table 2). These results seem compatible. A pattern of fast initial growth and long duration was found in low-latitude areas, and a pattern of slow initial growth and short growth duration was found in high-latitude areas. This combination of patterns is plausible, because faster initial growth at a young age can be explained by a longer duration of growth. Thus, in low-latitude regions, the longer duration of growth permits large annual growth and results in fast initial growth at a young age. Furthermore, a pattern of early maturity combined with small maximum size and long growth duration was found in low-latitude areas, and a pattern of late maturity combined with large maximum size and short growth duration was found in high-latitude areas. Trees with a large annual growth rate can reach the maximum size for their species earlier than trees with a small annual growth rate, indicating that both large annual growth and small maximum size may be linked with a pattern of early maturity. Thus, a long growth duration may lead to large annual growth, resulting in an early-maturity pattern in low-latitude regions. However, we do not propose that the variation of growth phenology explains the variation of the maximum tree size, because we are unaware of a mechanism that could link growth phenology with maximum tree size. The maximum size appears to be determined by factors other than the duration of growth, indicating that the maturity pattern is also strongly affected by factors other than duration. Thus, variations of the growth duration do not explain the variations of the maximum tree size, but may partly explain the variation in maturity patterns. Taken together, based on the results we have discussed in this section, we hypothesize that the geographical variation in growth phenology generates, at least in part, comparable geographical variation in long-term tree growth. This hypothesis is consistent with the idea that growth phenology is one of the key processes in forest dynamics (Delpierre et al., 2016a, b). Because long-term tree growth is one of the most important factors for determining the optimal rotation length (Nishizono, 2010; Nishizono et al., 2013), regional variation of tree phenology will play an important role for planning the forest management for C. japonica if our hypothesis is true. The effects of growth phenology would be prominent during the younger developmental stages. However, during older developmental stages, the effect of maximum tree size will be more prominent. Additional research will be required to identify the factors and mechanisms responsible for determining the maximum tree size and the growth pattern during older developmental stages. Conclusions In this study, we analysed the radial growth phenology of C. japonica trees at sites throughout Japan. We found that the trees in high-latitude regions begin their growth later than those in low-latitude regions and grow for a shorter duration. Most of the regional variation in the onset of growth and some of the regional variation in the duration of growth could be explained by temperature. Precipitation, tree size, tree age and stem density had no significant effect. The temperature results generally corresponded to our expectations and confirm that variations of growth phenology are key factors responsible for generating a latitudinal gradient in long-term growth. In contrast, the regional variation of growth cessation was not significantly associated with latitude, tree size, tree age, stem density, or the other climatic factors we investigated, which contradicts our expectations. Further research will be required to identify the factors that most strongly affect the cessation of growth. We plan to develop a simple growth model by explicitly incorporating the growth duration into various forms of the classical growth curve (e.g. Gompertz equation, Mitscherlich equation). It will also be necessary to examine a wider range of factors (e.g. photoperiod, soil moisture and nutrient contents) and use techniques such as multiple regression to reveal any interactions among the factors. Such studies will improve our understanding of the link between growth phenology and long-term growth trends. Conflict of interest statement The authors declare no conflicts of interest. Funding This work was supported by the Japan Society for the Promotion of Science KAKENHI programme [grant number 90353797]. Acknowledgements We thank Dr Yoshiyuki Inagaki (Shikoku Research Center, Forestry and Forest Products Research Institute) and Dr Yuichiro Oribe (Tohoku Regional Breeding Office, Forestry and Forest Products Research Institute) for providing information about their studies on growth phenology and xylogenesis. We also thank Dr Hisashi Sugita and Dr Shoji Noguchi (Forestry and Forest Products Research Institute) for providing forest census data for several test plots. We also thank Dr Nicolas Delpierre and two anonymous reviewers for their careful reading of an earlier version of this article and their constructive comments. We also thank Yasuyuki Tsurumi (University of Tokyo Forests) for his support during the fieldwork at the Chiba sites. References Barton, K. 2016 MuMIn: Multi-model inference. R Packag. version 1.10.0. Available at: https://cran.r-project.org/web/packages/MuMIn/index.html (accessed on April 22, 2016). Begum, S., Nakaba, S., Oribe, Y., Kubo, T. and Funada, R. 2010 Cambial sensitivity to rising temperatures by natural condition and artificial heating from late winter to early spring in the evergreen conifer Cryptomeria japonica. Trees - Struct. Funct.  24, 43– 52. Google Scholar CrossRef Search ADS   Berk, M. 2013 Smoothing-splines Mixed-effects Models [R package sme version 0.8]. Available at: https://cran.r-project.org/web/packages/sme/index.html (accessed on 16 June, 2017]. Cocozza, C., Lasserre, B., Giovannelli, A., Castro, G., Fragnelli, G. and Tognetti, R. 2009 Low temperature induces different cold sensitivity in two poplar clones (Populus×canadensis Mönch ‘I-214’ and P. deltoides Marsh. ‘dvina’). J. Exp. Bot.  60, 3655– 3664. Google Scholar CrossRef Search ADS PubMed  Crawley, M.J. 2007 The R Book . Wiley, p. 942. Google Scholar CrossRef Search ADS   Delpierre, N., Berveiller, D., Granda, E. and Dufrene, E. 2016a Wood phenology, not carbon input, controls the interannual variability of wood growth in a temperate oak forest. New. Phytol.  210, 459– 470. Google Scholar CrossRef Search ADS PubMed  Delpierre, N., Vitasse, Y., Chuine, I., Guillemot, J., Bazot, S., Rutishauser, T., et al.   2016b Temperate and boreal forest tree phenology: from organ-scale processes to terrestrial ecosystem models. Ann. For. Sci.  73, 5– 25. Google Scholar CrossRef Search ADS   Deslauriers, A., Rossi, S. and Anfodillo, T. 2007 Dendrometer and intra-annual tree growth: What kind of information can be inferred? Dendrochronologia  25, 113– 124. Google Scholar CrossRef Search ADS   Deslauriers, A., Rossi, S., Anfodillo, T. and Saracino, A. 2008 Cambial phenology, wood formation and temperature thresholds in two contrasting years at high altitude in southern Italy. Tree. Physiol.  28, 863– 871. Google Scholar CrossRef Search ADS PubMed  Drew, D.M. and Downes, G.M. 2009 The use of precision dendrometers in research on daily stem size and wood property variation: a review. Dendrochronologia  27, 159– 172. Google Scholar CrossRef Search ADS   Duchesne, L., Houle, D. and D’Orangeville, L. 2012 Influence of climate on seasonal patterns of stem increment of balsam fir in a boreal forest of Québec, Canada. Agric. For. Meteorol.  162–163, 108– 114. Google Scholar CrossRef Search ADS   Eilmann, B., Zweifel, R., Buchmann, N., Graf Pannatier, E. and Rigling, A. 2011 Drought alters timing, quantity, and quality of wood formation in Scots pine. J. Exp. Bot.  62, 2763– 2771. Google Scholar CrossRef Search ADS PubMed  Ford, K.R., Harrington, C.A. and St. Clair, J.B. 2017 Photoperiod cues and patterns of genetic variation limit phenological responses to climate change in warm parts of species’ range: modeling diameter-growth cessation in coast Douglas-fir. Glob. Chang. Biol.  23, 3348– 3362. Google Scholar CrossRef Search ADS PubMed  Fukuda, M., Iehara, T. and Matsumoto, M. 2003 Carbon stock estimates for sugi and hinoki forests in Japan. For. Ecol. Manage.  184, 1– 16. Google Scholar CrossRef Search ADS   Gruber, A., Strobl, S., Veit, B. and Oberhuber, W. 2010 Impact of drought on the temporal dynamics of wood formation in Pinus sylvestris. Tree. Physiol.  30, 490– 501. Google Scholar CrossRef Search ADS PubMed  Guillemot, J., Francois, C., Hmimina, G., Dufrêne, E., Martin-StPaul, N.K., Soudani, K., et al.   2017 Environmental control of carbon allocation matters for modelling forest growth. New. Phytol.  214, 180– 193. Google Scholar CrossRef Search ADS PubMed  Horiuchi, T. and Sakai, A. 1973 Effect of temperature on the frost hardiness of Cryptomeria japonica (in Japanese with English summary). J. Japanese For. Soc.  55, 46– 51. Imagawa, K., Oikawa, N. and Kitada, K. 1990 Simple method of measurement of change in stem diameter of Cryptomeria japonica trees (in Japanese). Trans. 42th Annu. Meet. Tohoku Branch Japanese For. Soc., 130–131. Japan Meteorological Agency 2002 Mesh Climate Data 2000 (CD-ROM) . Japan Meteorological Agency. Jyske, T., Mäkinen, H., Kalliokoski, T. and Nöjd, P. 2014 Intra-annual tracheid production of Norway spruce and Scots pine across a latitudinal gradient in Finland. Agric. For. Meteorol.  194, 241– 254. Google Scholar CrossRef Search ADS   Kawasaki, T. and Takeuchi, I. 1993 Diameter growth of trees with various sized in a Chamaecyparis Obtusa Plantation (in Japanese). Trans. 2 Annu. Meet. Kansai Branch Japanese For. Soc., 157–158. Komiyama, A., Higo, M., Kurumado, K. and Matsuhashi, T. 1992 Seasonality in radial growth of several trees (in Japanese). Trans. 40th Annu. Meet. Chubu Branch Japanese For. Soc., 73–74. Lupi, C., Morin, H., Deslauriers, A. and Rossi, S. 2010 Xylem phenology and wood production: resolving the chicken-or-egg dilemma. Plant. Cell. Environ.  33, 1721– 1730. Google Scholar CrossRef Search ADS PubMed  Mäkinen, H., Seo, J.W., Nöjd, P., Schmitt, U. and Jalkanen, R. 2008 Seasonal dynamics of wood formation: a comparison between pinning, microcoring and dendrometer measurements. Eur. J. For. Res.  127, 235– 245. Google Scholar CrossRef Search ADS   Nakagawa, S. and Schielzeth, H. 2013 A general and simple method for obtaining R2 from generalized linear mixed-effects models. Methods Ecol. Evol.  4, 133– 142. Google Scholar CrossRef Search ADS   Nanami, S., Kawaguchi, H. and Yamakura, T. 2010 Seasonal patterns of stem diameter growth in woody species with different life forms and geographical distributions at Mt. Mikasa, Nara, Japan. Appl. For. Sci  19, 27– 34. Nishizono, T. 2010 Effects of thinning level and site productivity on age-related changes in stand volume growth can be explained by a single rescaled growth curve. For. Ecol. Manage.  259, 2276– 2291. Google Scholar CrossRef Search ADS   Nishizono, T., Kitahara, F., Iehara, T. and Mitsuda, Y. 2013 Geographical variation in age–height relationships for dominant trees in Japanese cedar (Cryptomeria japonica D. Don) forests in Japan. J. For. Res  19, 305– 316. Google Scholar CrossRef Search ADS   Nishizono, T., Kitahara, F., Iehara, T. and Mitsuda, Y. 2015 Effect of climatic condition on regional differences in age-diameter relationships for dominant trees in Japanese cedar (Cryptomeria japonica D. Don) forests in Japan. (in Japanese with English summary). Kanto J. For. Res  66, 5– 8. Noda, T., Nakao, T. and Kuroki, Y. 1991 Analysis of diameter growth curve of trees in south Kyusyu (in Japanese). Trans. 44th Annu. Meet. Kyushu Branch Jpn. For. Soc., 61–62. Oribe, Y. and Kubo, T. 1997 Effect of heat on cambial reactivation during winter dormancy in evergreen and deciduous conifers. Tree. Physiol.  17, 81– 87. Google Scholar CrossRef Search ADS PubMed  Pinheiro, J.C. and Bates, D.M. 2000 Mixed-effects Models in S and S-PLUS . Springer, p. 528. Google Scholar CrossRef Search ADS   Rossi, S., Anfodillo, T., Čufar, K., Cuny, H.E., Deslauriers, A., Fonti, P., et al.   2016 Pattern of xylem phenology in conifers of cold ecosystems at the northern hemisphere. Glob. Chang. Biol.  22, 3804– 3813. Google Scholar CrossRef Search ADS PubMed  Rossi, S., Deslauriers, A., Anfodillo, T. and Carraro, V. 2007 Evidence of threshold temperatures for xylogenesis in conifers at high altitudes. Oecologia  152, 1– 12. Google Scholar CrossRef Search ADS PubMed  Rossi, S., Deslauriers, A., Anfodillo, T., Morin, H., Saracino, A., Motta, R., et al.   2006 Conifers in cold environments synchronize maximum growth rate of tree-ring formation with day length. New. Phytol.  170, 301– 310. Google Scholar CrossRef Search ADS PubMed  Rossi, S., Deslauriers, A., Griçar, J., Seo, J.W., Rathgeber, C.B.K., Anfodillo, T., et al.   2008 Critical temperatures for xylogenesis in conifers of cold climates. Glob. Ecol. Biogeogr.  17, 696– 707. Google Scholar CrossRef Search ADS   Saito, S., Kabeya, D., Araki, M., Tanaka, K., Migita, C. and Kawasaki, T. 2016 Phenology of diameter and height growth for Japanese Cedar (in Japanese with English summary). Kanto J. For. Res  67, 203– 206. Sakai, A. and Larcher, W. 1987 Frost Survival of Plants: Responses and Adaptation to Freezing Stress . Springer Berlin Heidelberg, p. 321. Google Scholar CrossRef Search ADS   Tardif, J., Flannigan, M. and Bergeron, Y. 2001 An analysis of the daily radial activity of 7 boreal tree species, northwestern Quebec. Environ. Monit. Assess.  67, 141– 160. Google Scholar CrossRef Search ADS PubMed  Vieira, J., Rossi, S., Campelo, F. and Nabais, C. 2014 Are neighboring trees in tune? Wood formation in Pinus pinaster. Eur. J. For. Res.  133, 41– 50. Google Scholar CrossRef Search ADS   Yamashita, K., Okada, N. and Kamo, K. 2006 Application of the wire dendrometer for monitoring the radial growth of trees. Mokuzai Gakkaishi  52, 8– 18. Google Scholar CrossRef Search ADS   Yamato, K., Nakao, T. and Kuroki, Y. 1989 Seasonal diameter growth of trees in South Kyusyu (in Japanese). Trans. 42th Annu. Meet. Kyushu Branch Japanese For. Soc., 145–146. © Institute of Chartered Foresters, 2018. All rights reserved. For Permissions, please e-mail: journals.permissions@oup.com. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Forestry: An International Journal Of Forest Research Oxford University Press

Latitudinal variation in radial growth phenology of Cryptomeria japonica D. Don trees in Japan

Loading next page...
 
/lp/ou_press/latitudinal-variation-in-radial-growth-phenology-of-cryptomeria-b5XOJ0w1zZ
Publisher
Institute of Chartered Foresters
Copyright
© Institute of Chartered Foresters, 2018. All rights reserved. For Permissions, please e-mail: journals.permissions@oup.com.
ISSN
0015-752X
eISSN
1464-3626
D.O.I.
10.1093/forestry/cpx055
Publisher site
See Article on Publisher Site

Abstract

Abstract We investigated the latitudinal variation in radial growth phenology (onset, cessation, and duration of radial growth) of Cryptomeria japonica D. Don trees in Japan. The radial growth was observed using dendrometers in 16 stands at multiple latitudes throughout Japan. The onset of radial growth showed a clear latitudinal gradient: at high-latitude sites, C. japonica growth started later than at low-latitude sites. However, cessation of radial growth was independent of latitude. Consequently, the duration of radial growth (defined based on the onset and cessation dates) showed a weak but significant latitudinal gradient: at high-latitude sites, C. japonica trees grew for a shorter period than trees at low-latitude sites. The onset and duration of growth were respectively strongly and weakly affected by temperature, although growth cessation was not explained by temperature or precipitation. The latitudinal gradient for duration of radial growth weakly but significantly supports the hypothesis that regional variation in growth phenology is a key factor responsible for the regional variation in long-term growth. Furthermore, we found that growth duration was more variable at low latitudes than at high latitudes. Introduction Long-term forecasting of timber yield, based on understanding the pattern of age-related change in growth of forest stands, is basic knowledge required to manage forests sustainably. Many studies have been conducted to understand the growth trends in trees and forests and to develop yield prediction tools. In Japan, Cryptomeria japonica D. Don is a major forestry species (Fukuda et al., 2003). Many C. japonica trees have been planted, and C. japonica forests are widely distributed throughout Japan. Therefore, many studies have examined the growth and yield of this species in Japan. Recently, the authors (Nishizono et al., 2013, 2015) reported that the long-term pattern of height and diameter growth of C. japonica at a regional scale could be classified into two regional groups, and that these groups were distributed geographically in a uniform manner. Thus, in high-latitude regions with cold temperatures, C. japonica showed a pattern of late maturity, with slow initial growth but a large maximum size. In low-latitude regions, with warmer temperatures, the species showed a pattern of early maturity, with fast initial growth but a smaller maximum size. In other words, we found a clear latitudinal gradient in long-term growth for C. japonica in Japan. However, it was not clear how the regional difference in long-term growth was generated. Long-term tree growth, spanning periods of several decades to centuries, is a consequence of the accumulation of short-term annual tree growth over long periods. Furthermore, annual tree growth represents the accumulation of short-term daily growth. In temperate and boreal forests, the daily growth shows cyclical seasonal fluctuation within a year (Delpierre et al., 2016b); tree growth typically starts in the spring and ceases in late summer or autumn. (Hereafter, we refer to this seasonal fluctuation as ‘growth phenology’.) By considering the typical climatic variation along a latitude gradient in the northern hemisphere and hypothesizing that growth phenology will be controlled by temperature, we can predict a clear latitudinal gradient for growth phenology; that is, at low-latitude sites with warmer temperatures, growth typically starts earlier and stops later than at high-latitude sites with cooler temperature, resulting in longer growth duration. In fact, studies of xylogenesis (secondary xylem growth) for several conifers, not including C. japonica, showed both such latitudinal gradients (e.g. Jyske et al., 2014) and temperature-dependency of growth phenology (e.g. Rossi et al., 2016). Moreover, recent papers (Delpierre et al., 2016a, b; Guillemot et al., 2017) have indicated that growth phenology is one of the key processes in forest dynamics and should be included as a component in forest ecosystem models to permit accurate and reliable prediction of forest dynamics. On this basis, we hypothesized that the regional variation in growth phenology is an important component of the processes that lead to regional variation in long-term growth. Based on this hypothesis, we predicted that the regional variation in long-term growth and in growth phenology would show similar patterns; specifically, we hypothesized that a latitudinal gradient would exist for growth phenology. The study described in this paper tested this hypothesis for radial growth phenology. The radial growth phenology for C. japonica trees has been described previously (Yamato et al., 1989; Imagawa et al., 1990; Noda et al., 1991; Komiyama et al., 1992; Yamashita et al., 2006; Nanami et al., 2010; Saito et al., 2016). These studies examined the effects of rainfall or temperature on the phenology of a species or on interspecific differences in the phenology based on measurements of the radial growth phenology in a forest stand or several stands within a region. However, no study quantitatively analysed the regional variation in growth phenology on a large spatial scale across multiple regions. The previous studies for high-latitude and low-latitude regions found that radial growth at high-latitude sites (~40°N) started in late April (Imagawa et al., 1990), whereas the radial growth at low-latitude sites (~32°N) started in early March (Yamato et al., 1989; Noda et al., 1991). This implies the existence of a latitudinal gradient in growth phenology of C. japonica trees. However, these previous studies did not adopt a uniform quantitative definition of the components of growth phenology (onset, cessation and duration of growth), making it difficult to quantitatively compare their results. Therefore, experimental data and scientific knowledge of the growth phenology of C. japonica trees are insufficient to let us detect a latitudinal gradient in tree growth phenology on a large spatial scale. For species other than C. japonica, many papers (e.g. Tardif et al., 2001; Rossi et al., 2006, 2008, 2016; Deslauriers et al., 2007, 2008; Mäkinen et al., 2008; Jyske et al., 2014; Delpierre et al., 2016a, b; Guillemot et al., 2017) described radial growth phenology, especially in the northern hemisphere, and examined the effect of several factors (measurement method, temperature, moisture, photoperiod and others) on the phenology. However, only a few studies (e.g. Delpierre et al., 2016a, b; Guillemot et al., 2017) examined the link between regional variations in growth phenology and forest growth dynamics. In addition, because the latitudinal gradients (e.g. Jyske et al., 2014) and temperature-dependency (e.g. Rossi et al., 2016) of growth phenology have been mainly examined in high-latitude areas with cool temperatures, we have little information on the gradients in low-latitude areas with warm temperatures, as is the case for much of Japan. In the present study, our goal was to provide some of the missing information. We measured and quantitatively analysed the radial growth phenology of C. japonica trees at 16 sites ranging from southern to northern Japan. Based on the measurement data, we examined the following questions: (1) Is there a latitudinal gradient, analogous to the gradient found for long-term growth, in the radial growth phenology of C. japonica trees in Japan? (2) If such a pattern exists, are temperature and precipitation responsible for the observed variation? The answers to these questions will improve our understanding of the link between growth phenology and long-term growth patterns, as well as the likely effects of climate on tree growth, and will provide the basic information required for long-term forecasting of timber yield. Methods Growth data Study sites and measurement methods We collected radial growth phenology data from 16 C. japonica stands for 1–4 years (Figure 1). Stands were chosen to provide coverage of most of Japan. Table 1 summarizes the characteristics of the stands and measurements. This sample included 15 single-storey stands and 1 two-storey stand (stand 4, Iwate 3). For our measurements, we used both automatic dendrometers (for 3 stands) and manual dendrometers (for 13 stands). We installed the dendrometers on 5–10 trees for each stand, at a height of ~1.2 m. Using the automatic dendrometer (Type DC2, Ecomatik, Munich, Germany; thermal elongation coefficient 1.4 × 10−6 K−1), we measured and recorded the stem circumference at an interval of 30 min. Using the manual dendrometer, we measured the stem circumference or diameter at breast height (DBH) by visually reading the scale, mainly at intervals of 7 days (growing season) or 15 days (winter). Of the 13 stands measured with the manual dendrometer, we used a commercially available band dendrometer made from Astralon plastic (Type D1, UMS GmbH, Munich, Germany; thermal elongation coefficient 75 × 10−6 K−1) to measure DBH in one stand (7, Toyama). In the other 12 stands, we used home-made stainless-steel dendrometers (thermal elongation coefficient 11.5 × 10−6 K−1) to measure the stem circumference. For some of the automatic dendrometer measurements (one for the Hokkaido site in 2016, two and one for the Kochi site in 2015 and 2016, and one for the Miyazaki site in 2016), the measurement failed for various reasons, including damage to the sensor caused by a falling branch, tree growth reaching the maximum measurable size of the dendrometer, or unknown reasons. Therefore, we excluded the data obtained from these dendrometers from our subsequent analyses. Consequently, the number of sample trees used for the analyses were 5 and 4 trees for the Hokkaido site in 2015 and 2016, 3 and 4 trees for the Kochi site in 2015 and 2016 (respectively), and 4 trees for the Miyazaki site in 2016. Table 1 Summary of the studied forests and the measured trees Site no.  Site name  Site property  Stand attributes3  Sample tree3  Measurement data  Latitude (°N)  Longitude (°E)  Elevation (m a.s.l.)  Annual mean temperature1 (°C)  Annual precipitation1 (mm)  Mean maximum snow depth1 (cm)  Age (years)  Stem density (trees/ha)  Sample size  Mean DBH (cm)  Mean height (m)  First date (year/month/day)  Last date (year/month/day)  Interval between measurements  Dendrometer material and data reading method  1  Hokkaido  41.846  140.733  65  8.4  1170.6  43  44  1285  5  29.2  22.0  2015/3/13  2016/11/1  30 min  Steel, Automatic  2  Iwate 1  39.770  141.140  207  9.3  1261.1  41  38  2026  10  24.8  20.2  2009/4/6  2009/12/24  7 days (15 days for winter season)  Steel, Manual  3  Iwate 2  39.790  141.153  269  9.1  1223  41  109  648  10  45.9  29.7  2009/4/9  2009/12/25  7 days (15 days for winter season)  Steel, Manual  4  Iwate 32  39.792  141.156  216  9.1  1227.8  39  109  157  10  48.4  28.3  2009/4/9  2009/12/25  7 days (15 days for winter season)  Steel, Manual  18  666  10  9.9  9.0  5  Yamagata 1  38.937  140.264  168  9.6  2249.3  148  40  2046  10  27.9  23.4  2009/4/16  2009/12/24  15 days  Steel, Manual  6  Yamagata 2  38.939  140.265  168  9.6  2249.3  148  96  769  10  37.5  27.5  2009/4/16  2009/12/24  15 days  Steel, Manual  7  Toyama  36.305  137.329  227  10.4  1895.3  86  26  885  10  25.1  16.4  2013/1/11  2013/11/14  15 days  Plastic, Manual  8  Ibaraki 1  36.184  140.217  35  13.5  1184.2  6  19  2141  8  14.1  10.8  2014/1/17  2015/9/15  7 days for growing season (more than 15 days for winter season)  Steel, Manual  9  Ibaraki 2  36.008  140.133  21  13.6  1223.1  5  36  1373  10  30.3  19.9  2013/2/28  2016/12/28  7 days  Steel, Manual  10  Chiba 1  35.192  140.144  222  13.7  1897.9  3  39  2306  10  24.4  18.3  2015/1/30  2016/12/26  15 days  Steel, Manual  11  Chiba 2  35.192  140.144  227  13.7  1897.9  3  40  3695  10  20.7  18.0  2015/2/6  2016/12/26  15 days  Steel, Manual  12  Chiba 3  35.171  140.165  299  13.2  1902.9  3  24  1927  10  22.1  16.7  2016/1/25  2016/12/26  15 days  Steel, Manual  13  Chiba 4  35.167  140.161  311  13.3  1896.4  3  71  1276  10  42.0  26.4  2016/2/10  2016/12/26  15 days  Steel, Manual  14  Chiba 5  35.161  140.146  293  13.3  1920.2  3  110  1259  10  45.8  23.9  2016/2/10  2016/12/26  15 days  Steel, Manual  15  Kochi  33.541  133.478  47  16.1  2550  1  46  1140  5  35.4  24.7  2015/1/1  2016/11/16  30 min  Steel, Automatic  16  Miyazaki  31.867  131.302  193  15.8  2563.7  1  46  936  5  31.1  21.6  2016/1/1  2016/12/6  30 min  Steel, Automatic  Site no.  Site name  Site property  Stand attributes3  Sample tree3  Measurement data  Latitude (°N)  Longitude (°E)  Elevation (m a.s.l.)  Annual mean temperature1 (°C)  Annual precipitation1 (mm)  Mean maximum snow depth1 (cm)  Age (years)  Stem density (trees/ha)  Sample size  Mean DBH (cm)  Mean height (m)  First date (year/month/day)  Last date (year/month/day)  Interval between measurements  Dendrometer material and data reading method  1  Hokkaido  41.846  140.733  65  8.4  1170.6  43  44  1285  5  29.2  22.0  2015/3/13  2016/11/1  30 min  Steel, Automatic  2  Iwate 1  39.770  141.140  207  9.3  1261.1  41  38  2026  10  24.8  20.2  2009/4/6  2009/12/24  7 days (15 days for winter season)  Steel, Manual  3  Iwate 2  39.790  141.153  269  9.1  1223  41  109  648  10  45.9  29.7  2009/4/9  2009/12/25  7 days (15 days for winter season)  Steel, Manual  4  Iwate 32  39.792  141.156  216  9.1  1227.8  39  109  157  10  48.4  28.3  2009/4/9  2009/12/25  7 days (15 days for winter season)  Steel, Manual  18  666  10  9.9  9.0  5  Yamagata 1  38.937  140.264  168  9.6  2249.3  148  40  2046  10  27.9  23.4  2009/4/16  2009/12/24  15 days  Steel, Manual  6  Yamagata 2  38.939  140.265  168  9.6  2249.3  148  96  769  10  37.5  27.5  2009/4/16  2009/12/24  15 days  Steel, Manual  7  Toyama  36.305  137.329  227  10.4  1895.3  86  26  885  10  25.1  16.4  2013/1/11  2013/11/14  15 days  Plastic, Manual  8  Ibaraki 1  36.184  140.217  35  13.5  1184.2  6  19  2141  8  14.1  10.8  2014/1/17  2015/9/15  7 days for growing season (more than 15 days for winter season)  Steel, Manual  9  Ibaraki 2  36.008  140.133  21  13.6  1223.1  5  36  1373  10  30.3  19.9  2013/2/28  2016/12/28  7 days  Steel, Manual  10  Chiba 1  35.192  140.144  222  13.7  1897.9  3  39  2306  10  24.4  18.3  2015/1/30  2016/12/26  15 days  Steel, Manual  11  Chiba 2  35.192  140.144  227  13.7  1897.9  3  40  3695  10  20.7  18.0  2015/2/6  2016/12/26  15 days  Steel, Manual  12  Chiba 3  35.171  140.165  299  13.2  1902.9  3  24  1927  10  22.1  16.7  2016/1/25  2016/12/26  15 days  Steel, Manual  13  Chiba 4  35.167  140.161  311  13.3  1896.4  3  71  1276  10  42.0  26.4  2016/2/10  2016/12/26  15 days  Steel, Manual  14  Chiba 5  35.161  140.146  293  13.3  1920.2  3  110  1259  10  45.8  23.9  2016/2/10  2016/12/26  15 days  Steel, Manual  15  Kochi  33.541  133.478  47  16.1  2550  1  46  1140  5  35.4  24.7  2015/1/1  2016/11/16  30 min  Steel, Automatic  16  Miyazaki  31.867  131.302  193  15.8  2563.7  1  46  936  5  31.1  21.6  2016/1/1  2016/12/6  30 min  Steel, Automatic  1The values were obtained from the mesh climate data provided by the Japan Meteorological Agency (2002), which provides monthly means from 1971 to 2000 throughout Japan at a spatial resolution of ca. 1 km. The ‘mean maximum snow depth’ indicates the average value of the maximum snow depth in each year from 1971 to 2000. 2The Iwate 3 plot was in a two-storey forest stand. The upper and lower rows of data show the attributes of trees in the upper and lower stories of the stand, respectively. 3The values were obtained at the beginning of the measurements. Figure 1 View largeDownload slide Location of the 16 Cryptomeria japonica study sites. Numbers correspond to the site numbers in Table 1. Figure 1 View largeDownload slide Location of the 16 Cryptomeria japonica study sites. Numbers correspond to the site numbers in Table 1. Data processing First, we converted the stem circumference into DBH based on the assumption that the stem cross-section was circular. Second, we applied a temperature correction to all DBH data using the thermal elongation coefficient of the dendrometer band and mean daily or hourly temperatures obtained from the Japan Meteorological Agency (see the next section for details). Third, for the measurements obtained using the automatic dendrometer, we calculated the daily mean DBH using only DBH values measured between 8:00 a.m. and 4:00 p.m., which is the range of times when the measurements using the manual dendrometer were usually conducted. Excluding the night data should reduce the bias compared with the manual dendrometers due to differences in the time of day used for the measurement. Finally, we converted the measurement date into day of the year (DOY) values for the data from each year. We used the relationship between DBH and DOY for each sample tree in our analyses of the radial growth phenology. Meteorological data We obtained daily or hourly meteorological data (temperature and precipitation) during the measurement period for each test site from the nearest meteorological station. Meteorological data for each station were obtained from the Japan Meteorological Agency (http://www.data.jma.go.jp/gmd/risk/obsdl/index.php). The temperature was corrected for elevation differences using a standard temperature lapse rate of −0.60°C per 100 m in elevation. Analysis method Quantifying growth phenology We used three indices to quantify the growth phenology: the onset time (To), cessation time (Tc), and duration (Pg) of the radial growth. We calculated these indices for each plot and year using a growth curve at a stand level fitted to the DBH–DOY relationships for the trees in each plot. In many previous studies of growth phenology, the researchers used a sigmoidal growth function (e.g. the Gompertz function; Jyske et al., 2014) for this fitting. However, in the present study we used a more flexible spline curve for the fitting because the sigmoidal function was not applicable to the bimodal growth pattern, with growth peaks in the spring and autumn, exhibited by trees growing in a Mediterranean climate (e.g. Vieira et al., 2014) or in south-western Japan (e.g. Noda et al., 1991; Kawasaki and Takeuchi, 1993). To fit the spline curve, we applied a smoothing-splines mixed-effects model, with DOY as the fixed factor and individual trees as the random factor, to define the DBH–DOY relationships for each year and plot. We conducted the fitting using the sme function of the sme package (Berk, 2013) for the R software (http://www.R-project.org/). For the smoothing parameters, we obtained the optimal values based on the value of Akaike’s information criterion (AIC) using a Nelder-Mead search (Berk, 2013). To obtain a suitable fitting, we used trial and error to transform the variables and calibrate the tolerance. Because our data were obtained by repeated measurements from the same individuals, the data had non-independent errors. This violates the important requirement of independence of error for standard statistical analysis, indicating that standard methods of analysis could not be applied to our data (Crawley, 2007). However, a mixed-effects model can account for such temporal pseudoreplication (Crawley, 2007), and allowed us to model the average pattern of radial growth phenology at a stand level. To obtain the three phenology indices from the fitted curve, we first calculated the annual growth of the average tree in the stand as the maximum DBH minus the minimum DBH from the fitted curve at a stand level for a given year. Next, we calculated To and Tc as the DOY when the tree attained 10 and 90 per cent, respectively, of the total annual growth, as defined by Nanami et al. (2010). Finally, we calculated Pg as Tc – To (i.e. the difference between the onset and cessation times). Based on our preliminary assessment of the fitting curves (Figure 2), DBH in several stands (e.g. stands 3, 5 and 6; Figure 2B) decreased during the period between dendrometer installation and the onset of radial growth. This decrease may have resulted from incorrect measurements due to slack after installation of the dendrometer band (Drew and Downes, 2009). Therefore, the observed onset date for these stands may be later than the actual value. To confirm the reliability of the observed onset, we also measured the DBH change for these stands during the spring of the next year. Comparison of the resulting graphs indicated that the onset of radial growth in the next growing season appeared similar to the observed onset for the first growing season in these stands (results not shown). Therefore, we conclude that the onset dates were reliable for these stands. In the Hokkaido stand (no. 1), stem shrinkage was probably induced by low temperatures (Cocozza et al., 2009), which occurred during the cold winter season (DOY < 70, with daily mean temperatures less than ~0°C) in 2016 (Figure 2B). This shrinkage resulted in an unreliable estimate of the phenology indices. We therefore excluded the part of the fitted curve for this period (DOY < 70) from the analysis in our calculation of the indices. Figure 2 View largeDownload slide Fitted spline curves for the DBH changes at the 16 study sites. DOY represents the day of the year (from 1 January). The DBH change was calculated as D(t) – Dmin, where D(t) is the DBH (cm) on DOY t (days), and Dmin is the minimum DBH in a stand in each year. Text in the legend represents the site number (1–16 in Table 1) followed by the measurement year; for stand 4, U represents the upper layer of trees in the two-storey stand and L represents the lower layer. Figure 2 View largeDownload slide Fitted spline curves for the DBH changes at the 16 study sites. DOY represents the day of the year (from 1 January). The DBH change was calculated as D(t) – Dmin, where D(t) is the DBH (cm) on DOY t (days), and Dmin is the minimum DBH in a stand in each year. Text in the legend represents the site number (1–16 in Table 1) followed by the measurement year; for stand 4, U represents the upper layer of trees in the two-storey stand and L represents the lower layer. Effects of climate, latitude, tree size and stand attributes on the growth phenology To determine which factors affected the radial growth phenology, we conducted single regression analysis between the growth phenology indices and individual variables using a linear mixed-effects model (LMM). In our dataset, the locations of the stands were somewhat clustered (Figure 1). To account for the geographical clustering and the resulting spatial pseudoreplication (Crawley, 2007), the LMM had the following form:   yijk=β0+βxijk+αi+αij+εijkαi≅N(0,σα12)αij≅N(0,σα22)εijk≅N(0,σεΛij) (1)where yijk is the phenology index in plot j (nested within location i) and year k; β0 is the intercept; β is the slope coefficient for the fixed effect; xijk is the independent climatic, geographical, or other variable (defined later in this section) in plot j (nested within location i) and year k; εijk is an error term for plot j (nested within location i) and year k; and αi and αij are random effects terms for location i and plot j (nested within the location). We assumed normality and homogeneity of variance (σα12 and σα22) for αi and αij. We used the eight prefectures (described in ‘Site name’ of Table 1: Hokkaido, Iwate, Yamagata, Toyama, Ibaraki, Chiba, Kochi and Miyazaki) as the location term. Based on examination of the standardized residuals, we defined the variance structure (σεΛij) for To, Tc and Pg. Examination of the standardized residuals revealed a homoscedastic error structure (σε2) for To and Tc, and the variance structure could be described using an exponential function of mean daily precipitation to correct for heteroscedasticity for Pg. We tested the significance of the fixed-effects term (β) using a conditional t-test (Pinheiro and Bates, 2000), and defined significance as P < 0.05. We also calculated the marginal R2GLMM (Nakagawa and Schielzeth, 2013) to examine the explanatory power of the independent variable. The marginal R2GLMM represents the proportion of variance in the response variable that is explained entirely by the fixed effect. It ranges from 0 to 1, which represent no fit and the best possible fit, respectively. We calculated the annual and quarterly (January to March, April to June, July to September, and October to December) means of the mean daily temperature and daily precipitation during the measurement year for each stand. We used these annual and quarterly means as explanatory variables in the regression analysis. In addition, we used the mean initial sizes of the measured trees for each stand (mean DBH and mean height), initial stand age, initial stem density and the latitude of each stand as explanatory variables in the regression analysis. We excluded data from the lower storey in the Iwate 3 stand from our analyses of stem density vs. phenology indices because the trees in the lower storey must be affected by trees in the upper storey; it was therefore difficult to quantify the stand density for the lower storey on the same basis as stem density in the single-storey stands. The fitting was performed using the lme function of the NLME package (Pinheiro and Bates, 2000) for the R software. The marginal R2GLMM was calculated by using the r.squaredGLMM function of the MuMIn package (Barton, 2016) for the R software. Results Radial growth phenology Overall, the DBH change (the difference between DBH on a given DOY and the minimum DBH during the year) increased with increasing DOY, and then stabilized (Figure 2). The DBH growth rate (cm/day) increased with increasing DOY after onset of growth in most plots (Figure 2B–D), and then decreased to almost zero by the summer, although some plots showed fluctuations of DBH change (without a significant increase) after the summer. However, these fluctuations may have been an artefact of the use of splines for the models, and may not have biological significance. However, the DBH change showed a different pattern in several plots (Figure 2A). The DBH growth rate in these plots first increased, then decreased to near zero and remained at that level until mid-summer, but from late summer until autumn, increased again with increasing DOY. We estimated the phenology indices for 25 curves (16 stands in at least 1 year, more than 1 year for some stands, and the understorey and overstorey curves in stand 4). The estimated To values ranged from DOY 81 to 139, and averaged DOY 107.8. The estimated Tc values ranged from DOY 155 to 233, and averaged DOY 195.7. The estimated Pg values ranged from 51 to 144 days, and averaged 88.0 days. Latitudinal variation and gradient in growth phenology Higher-latitude sites had a significantly larger To (Figure 3, Table 2). Thus, radial growth started later at higher latitudes than at lower latitudes. R2GLMM was 0.719, indicating a model with strong explanatory power for the effect of latitude on To. Latitude did not significantly affect Tc (Figure 3, Table 2). Thus, the relationship between Tc and latitude is unclear. Higher latitude sites had a significantly smaller Pg (Figure 3, Table 2). Thus, the duration of radial growth decreased with increasing latitude. R2GLMM was 0.338, indicating a model with significant but weak explanatory power for the effect of latitude on Pg. Table 2 Parameter estimates and statistics for the linear mixed-effects model using equation [1]1,2 Explanatory variables  Timing of onset (DOY)  Timing of cessation (DOY)  Growth duration (days)  β0  β  P  R2GLMM  β0  β  P  R2GLMM  β0  β  P  R2GLMM  Latitude (°N)  −109.3910  5.9810  <0.001  0.719  175.150  0.6457  0.796  0.006  249.972  −4.4786  0.031  0.338  Mean daily mean temperature (°C)   Whole year  192.8064  −6.2552  <0.001  0.789  224.138  −1.9913  0.354  0.058  53.737  2.3516  0.201  0.112   Jan.–Mar.  129.8004  −5.4690  <0.001  0.843  205.234  −2.6285  0.095  0.146  79.259  1.3655  0.300  0.057   Apr.–June  221.4001  −6.8680  <0.001  0.742  219.390  −1.2874  0.597  0.018  29.510  3.4108  0.092  0.167   July–Sep.  279.6027  −7.3514  <0.001  0.723  191.785  0.3235  0.885  0.001  1.476  3.6562  0.041  0.175   Oct.–Dec.  157.5543  −4.7811  <0.001  0.704  211.177  −1.2889  0.396  0.034  97.892  −1.6176  0.027  0.058  Mean daily precipitation (mm)   Whole year  107.2507  0.5650  0.838  0.003  176.976  3.7260  0.148  0.095  73.932  1.8848  0.412  0.041   Jan.–Mar.  97.6598  2.7955  0.294  0.084  188.665  2.2634  0.341  0.045  83.759  −0.1125  0.948  0.000   Apr.–June  121.1767  −1.6701  0.220  0.060  196.878  0.3814  0.710  0.002  78.738  0.9411  0.352  0.018   July–Sep.  103.6441  0.8449  0.301  0.013  192.321  0.8361  0.204  0.013  82.434  0.1267  0.818  0.000   Oct.–Dec.  120.9711  −1.9121  0.282  0.029  203.960  −0.9060  0.502  0.006  82.698  0.1406  0.880  0.000  Mean size of sample trees   DBH (cm)  109.3901  0.0460  0.761  0.000  188.685  0.3594  0.372  0.023  79.255  0.1332  0.773  0.005   Tree height (m)  108.5475  0.1054  0.741  0.001  186.831  0.6007  0.469  0.019  83.565  −0.0122  0.989  0.000  Stand attributes   Age (years)  109.4301  0.0277  0.630  0.001  197.073  0.0416  0.784  0.003  86.146  −0.0514  0.756  0.006   Stem density (trees/ha)  110.9242  0.0001  0.958  0.000  215.097  −0.0124  0.085  0.226  85.234  −0.0010  0.882  0.002  Explanatory variables  Timing of onset (DOY)  Timing of cessation (DOY)  Growth duration (days)  β0  β  P  R2GLMM  β0  β  P  R2GLMM  β0  β  P  R2GLMM  Latitude (°N)  −109.3910  5.9810  <0.001  0.719  175.150  0.6457  0.796  0.006  249.972  −4.4786  0.031  0.338  Mean daily mean temperature (°C)   Whole year  192.8064  −6.2552  <0.001  0.789  224.138  −1.9913  0.354  0.058  53.737  2.3516  0.201  0.112   Jan.–Mar.  129.8004  −5.4690  <0.001  0.843  205.234  −2.6285  0.095  0.146  79.259  1.3655  0.300  0.057   Apr.–June  221.4001  −6.8680  <0.001  0.742  219.390  −1.2874  0.597  0.018  29.510  3.4108  0.092  0.167   July–Sep.  279.6027  −7.3514  <0.001  0.723  191.785  0.3235  0.885  0.001  1.476  3.6562  0.041  0.175   Oct.–Dec.  157.5543  −4.7811  <0.001  0.704  211.177  −1.2889  0.396  0.034  97.892  −1.6176  0.027  0.058  Mean daily precipitation (mm)   Whole year  107.2507  0.5650  0.838  0.003  176.976  3.7260  0.148  0.095  73.932  1.8848  0.412  0.041   Jan.–Mar.  97.6598  2.7955  0.294  0.084  188.665  2.2634  0.341  0.045  83.759  −0.1125  0.948  0.000   Apr.–June  121.1767  −1.6701  0.220  0.060  196.878  0.3814  0.710  0.002  78.738  0.9411  0.352  0.018   July–Sep.  103.6441  0.8449  0.301  0.013  192.321  0.8361  0.204  0.013  82.434  0.1267  0.818  0.000   Oct.–Dec.  120.9711  −1.9121  0.282  0.029  203.960  −0.9060  0.502  0.006  82.698  0.1406  0.880  0.000  Mean size of sample trees   DBH (cm)  109.3901  0.0460  0.761  0.000  188.685  0.3594  0.372  0.023  79.255  0.1332  0.773  0.005   Tree height (m)  108.5475  0.1054  0.741  0.001  186.831  0.6007  0.469  0.019  83.565  −0.0122  0.989  0.000  Stand attributes   Age (years)  109.4301  0.0277  0.630  0.001  197.073  0.0416  0.784  0.003  86.146  −0.0514  0.756  0.006   Stem density (trees/ha)  110.9242  0.0001  0.958  0.000  215.097  −0.0124  0.085  0.226  85.234  −0.0010  0.882  0.002  DOY, day of year. 1Grey areas indicate statistically significant relationships. 2As a goodness-of-fit statistic for the mixed-effects model, we used the marginal R2GLMM statistic (Nakagawa and Schielzeth, 2013), which represents the proportion of the variance explained exclusively by the fixed effect. Figure 3 View largeDownload slide Relationships between latitude and the onset, cessation and duration of radial growth (DOY, day of year). Text in the legend represents the site number (1–16 in Table 1); for stand 4, U represents the upper layer of trees in the two-storey stand and L represents the lower layer. Figure 3 View largeDownload slide Relationships between latitude and the onset, cessation and duration of radial growth (DOY, day of year). Text in the legend represents the site number (1–16 in Table 1); for stand 4, U represents the upper layer of trees in the two-storey stand and L represents the lower layer. Effects of climatic conditions, tree size and stand attributes on growth phenology Higher temperatures, both for the year as a whole and for the four periods, were significantly negatively associated with To (Figure 4, Table 2); that is, warmer mean temperatures resulted in earlier onset of growth. R2GLMM ranged from 0.704 to 0.843, indicating a model with strong explanatory power for the effect of temperature on To. Precipitation did not significantly affect To (Table 2). These results indicated that temperature was the climatic variable with the strongest effect on To and could by itself explain most of the variation in To. Figure 4 View largeDownload slide Relationships between the mean daily mean temperature for the whole year and the onset, cessation, and duration of radial growth. Text in the legend represents the site number (1–16 in Table 1); for stand 4, U represents the upper layer of trees in the two-storey stand and L represents the lower layer. Figure 4 View largeDownload slide Relationships between the mean daily mean temperature for the whole year and the onset, cessation, and duration of radial growth. Text in the legend represents the site number (1–16 in Table 1); for stand 4, U represents the upper layer of trees in the two-storey stand and L represents the lower layer. Neither temperature nor precipitation significantly affected Tc (Figure 4, Table 2). Higher mean temperatures from July to September and from October to December were significantly positively and negatively, respectively, associated with Pg; that is, warmer temperatures in these months increased and decreased, respectively, the growth duration (Table 2). For mean temperature from July to September, the R2GLMM values for these means were 0.175, indicating a model with significant but weak explanatory power for the effect of temperature on Pg. However, for mean temperature from October to December, the R2GLMM values for these means were 0.058, indicating a negligible effect. Precipitation was not significantly associated with Pg (Table 2). Thus, at warm sites, the growth duration is longer than at cold sites. The size parameters (DBH and height) for the sample trees, stand age, and stem density were not significantly associated with To, Tc or Pg (Table 2). These results indicate that tree size, stand age and stand density did not affect the observed variation of growth phenology. Discussion Latitudinal gradient in growth phenology We found a clear latitudinal gradient for the onset of radial growth of C. japonica trees (Table 2, Figure 3). At high-latitude sites, radial growth started later than at low-latitude sites. This association was strong and could explain most (71.9 per cent) of the variation in onset time. However, we found no significant latitudinal gradient for the cessation of radial growth. Thus, latitude could not explain the observed variation in cessation time. Furthermore, we found a latitudinal gradient for the duration of radial growth: the duration of radial growth was shorter at high-latitude sites than at low-latitude sites. The latitude was weakly but significantly associated with the growth duration, as it explained only 33.8 per cent of the variation in the growth duration. This study provides the first quantitative evidence of latitudinal gradients in the onset and the duration of growth for C. japonica, although such latitudinal gradients have been found for other conifers (e.g. Jyske et al., 2014). The C. japonica trees in stands at higher latitudes had a late onset time, leading to a short growth duration (Figures 2 and 4). For the C. japonica trees in stands at lower latitudes, the onset times were early for all stands, but the growth duration varied widely within the stands. The growth duration was longest for the C. japonica trees in some stands at lower latitudes, which continued to grow until autumn. On the other hand, the growth duration was shortest for the trees of some stands at lower latitudes, which stopped growing by mid-May. These results show that C. japonica trees at low latitudes have variable growth duration, whereas those at high latitudes have a relatively uniform growth duration. Effects of temperature on growth phenology Temperature clearly affected the radial growth phenology (Table 2, Figure 4), whereas precipitation had no impact. Furthermore, age, tree size and stem density had no impact. These results show that the effects of environment (temperature) were much stronger than the effects of ontogeny and forest management. Therefore, in the following discussion, we will focus mainly on the effects of temperature. Previous studies (Oribe and Kubo, 1997; Begum et al., 2010) showed that heating of the stem breaks the winter cambial dormancy of C. japonica, and results in cambial reactivation. Thus, rising temperatures trigger the onset of radial growth. In addition, Rossi et al. (2016) showed that the onset of growth became earlier with increasing mean annual temperature for coniferous forest stands in the northern hemisphere. In the present study, the onset of radial growth was strongly associated with the mean temperature at the site (Table 2, Figure 4). Thus, at sites with higher mean temperatures, radial growth started earlier than at sites with lower mean temperatures. These results agreed with the previous findings. Furthermore, because our study included data from stands at lower latitudes and lower elevations than the data analysed by Rossi et al. (2016), the maximum temperature in the range that we analysed was higher than that analysed by Rossi et al. (2016). However, Rossi et al. (2016) included sites with colder temperatures than our coldest sites. Therefore, our study extended the previous finding (the association between higher temperature and earlier onset) to regions with higher temperatures. On the other hand, the factor(s) that determined the date of cessation of radial growth is unclear (Delpierre et al., 2016b). Some studies (Rossi et al., 2007, 2016) reported that climate factors affected the cessation of growth. For example, Rossi et al. (2007, 2016) reported that temperature triggers the cessation of growth. Additionally, some researchers noted that drought (e.g. Gruber et al., 2010; Eilmann et al., 2011) or photoperiod (Ford et al., 2017) were linked with the timing of cessation. However, other studies reported that the cessation of growth is more strongly affected by the total number of xylem cells produced in a season than by climatic factors (Lupi et al., 2010), and is only weakly explained by climatic factors (Duchesne et al., 2012). In the present study (Table 2, Figure 4), the cessation of radial growth was not significantly associated with the temperature and precipitation variables. This was because the range of cessation dates mostly overlapped for the stands in high- and low-temperature areas (Figure 4). These results indicated that the growth cessation was not triggered solely by these two climatic factors. Rossi et al. (2016), who had a large sample size that included trees from sites with a wide range of temperature, showed that spring (onset-related) and autumn (cessation-related) phenological events became earlier and later, respectively, with increasing mean temperature. Simultaneously, they found that mean temperature explained a smaller part of the overall variation in cessation date than of the variation in onset date. These previous findings suggest that temperature has a weak effect on the cessation of growth, and that such an effect would be difficult to detect with a small sample size or a narrow temperature gradient; thus, other factors are likely to affect the cessation of growth by acting together with temperature (Delpierre et al., 2016b). For example, Ford et al. (2017) reported that high temperature under a long photoperiod in the early summer was linked with early growth cessation. Based on these previous and present results, we hypothesize that the cessation of radial growth is determined by multiple factors acting together in a complicated mechanism, and that this complexity explains the unclear relationships we observed between the cessation of radial growth and the climatic factors and latitude in the present study. The mean temperatures from July to September had a weak but significant positive effect on the duration of growth, and explained a small but significant portion of the variation in the duration (Table 2). These results show that the duration of growth is longer at warm sites than at cold sites. The present study provides the first quantitative evidence of temperature-dependency in the duration of growth for C. japonica trees, although such temperature-dependency has been found for other conifers (e.g. Rossi et al., 2016). Furthermore, our study extended the previous finding (the association between higher temperature and longer duration) to regions with higher temperatures. Cryptomeria japonica trees in high-temperature areas had highly variable growth duration, whereas those in low-temperature areas had a more uniform growth duration (Figure 4). This low degree of variation of growth phenology in low-temperature (high-latitude) areas may be attributed to the severe climatic conditions in winter. Woody plants undergo a cold-acclimation process to increase their freezing tolerance in winter (Sakai and Larcher, 1987). The first step in this process is growth cessation (Horiuchi and Sakai, 1973; Sakai and Larcher, 1987). Therefore, to survive the winter, C. japonica trees at high-latitude sites with cold conditions may need to cease growth earlier to avoid freezing damage. This adaptation to cold conditions would lead to an earlier upper limit on the cessation of growth at high latitudes than at low latitudes. In addition, actively growing tissues are more vulnerable to cold temperatures (Sakai and Larcher, 1987), which occur frequently during the spring. Avoidance of these temperatures may explain the delayed onset of radial growth at high latitudes (Figure 4). Therefore, the onset and cessation dates in high-latitude areas necessarily fall in the narrow range between the later lower limit and earlier upper limit. This low variation would increase the number of trees that survive the severe winters in high-latitude areas. In contrast, the high variation of growth phenology in high-temperature areas can be attributed to the mild climatic conditions. The warmer temperatures potentially permit earlier onset and later cessation of growth, thereby leading to a longer growth duration. However, some site-specific factors other than temperature, such as soil conditions (e.g. moisture, nutrients), may limit growth, resulting in an earlier cessation and a shorter duration than the potential values based only on temperature. In addition, a shorter growth duration will not necessarily be linked to tree mortality if the trees can ensure continuous growth, even if only a small amount of growth. Thus, the mild climate permits greater variation in the growth phenology of C. japonica trees in low-latitude areas. Based on this discussion, latitudinal variation in radial growth phenology appears to originate from the temperature-dependence of growth, particularly for onset and somewhat for duration. However, temperatures will exhibit potentially high inter-annual variability. In addition, factors other than temperature (e.g. drought) also have high inter-annual variability. These variations may therefore affect the duration of growth (e.g. Gruber et al., 2010; Eilmann et al., 2011). Therefore, researchers should note that the year selected for their measurements may affect the relationship of latitude to radial growth phenology. In addition, It should be noted that several papers (e.g. Yamashita et al., 2006; Deslauriers et al., 2007; Mäkinen et al., 2008) have noted a discrepancy between the growth phenology inferred from dendrometer measurements and microscopic observations of wood formation. For C. japonica, Yamashita et al. (2006) reported that the onset of growth was similar, but the cessation of growth was different, between these two methods. This uncertainty of determining the cessation of growth using different methods may have affected the relationships we observed between the cessation of radial growth and the climatic factors and latitude in the present study, because we adopted dendrometer measurements rather than more accurate microscopic methods. Therefore, in the future, we should re-assess the phenology data by means of microscopic examination using micro-core sampling, as this approach probably provides the most reliable estimate of growth phenology, although the method has not (to our knowledge) been applied to C. japonica trees. Relationship between growth phenology and long-term growth patterns In this study, we hypothesized that the latitudinal gradient that exists for long-term growth might be caused by a parallel gradient in growth phenology. We tested this hypothesis and found significant latitudinal gradients for the onset and duration of radial growth (Table 2, Figure 3). In our previous research on long-term growth (Nishizono et al., 2013, 2015), we found a pattern of late maturity combined with slow initial growth and a large maximum size in high-latitude areas, and a pattern of early maturity combined with fast initial growth and a small maximum size in low-latitude areas. In the present study, we found a pattern of short growth duration combined with late onset in high-latitude areas and a pattern of long duration combined with early onset in low-latitude areas. However, we found no significant effects of age (ontogeny) on the latitudinal gradient (Table 2). These results seem compatible. A pattern of fast initial growth and long duration was found in low-latitude areas, and a pattern of slow initial growth and short growth duration was found in high-latitude areas. This combination of patterns is plausible, because faster initial growth at a young age can be explained by a longer duration of growth. Thus, in low-latitude regions, the longer duration of growth permits large annual growth and results in fast initial growth at a young age. Furthermore, a pattern of early maturity combined with small maximum size and long growth duration was found in low-latitude areas, and a pattern of late maturity combined with large maximum size and short growth duration was found in high-latitude areas. Trees with a large annual growth rate can reach the maximum size for their species earlier than trees with a small annual growth rate, indicating that both large annual growth and small maximum size may be linked with a pattern of early maturity. Thus, a long growth duration may lead to large annual growth, resulting in an early-maturity pattern in low-latitude regions. However, we do not propose that the variation of growth phenology explains the variation of the maximum tree size, because we are unaware of a mechanism that could link growth phenology with maximum tree size. The maximum size appears to be determined by factors other than the duration of growth, indicating that the maturity pattern is also strongly affected by factors other than duration. Thus, variations of the growth duration do not explain the variations of the maximum tree size, but may partly explain the variation in maturity patterns. Taken together, based on the results we have discussed in this section, we hypothesize that the geographical variation in growth phenology generates, at least in part, comparable geographical variation in long-term tree growth. This hypothesis is consistent with the idea that growth phenology is one of the key processes in forest dynamics (Delpierre et al., 2016a, b). Because long-term tree growth is one of the most important factors for determining the optimal rotation length (Nishizono, 2010; Nishizono et al., 2013), regional variation of tree phenology will play an important role for planning the forest management for C. japonica if our hypothesis is true. The effects of growth phenology would be prominent during the younger developmental stages. However, during older developmental stages, the effect of maximum tree size will be more prominent. Additional research will be required to identify the factors and mechanisms responsible for determining the maximum tree size and the growth pattern during older developmental stages. Conclusions In this study, we analysed the radial growth phenology of C. japonica trees at sites throughout Japan. We found that the trees in high-latitude regions begin their growth later than those in low-latitude regions and grow for a shorter duration. Most of the regional variation in the onset of growth and some of the regional variation in the duration of growth could be explained by temperature. Precipitation, tree size, tree age and stem density had no significant effect. The temperature results generally corresponded to our expectations and confirm that variations of growth phenology are key factors responsible for generating a latitudinal gradient in long-term growth. In contrast, the regional variation of growth cessation was not significantly associated with latitude, tree size, tree age, stem density, or the other climatic factors we investigated, which contradicts our expectations. Further research will be required to identify the factors that most strongly affect the cessation of growth. We plan to develop a simple growth model by explicitly incorporating the growth duration into various forms of the classical growth curve (e.g. Gompertz equation, Mitscherlich equation). It will also be necessary to examine a wider range of factors (e.g. photoperiod, soil moisture and nutrient contents) and use techniques such as multiple regression to reveal any interactions among the factors. Such studies will improve our understanding of the link between growth phenology and long-term growth trends. Conflict of interest statement The authors declare no conflicts of interest. Funding This work was supported by the Japan Society for the Promotion of Science KAKENHI programme [grant number 90353797]. Acknowledgements We thank Dr Yoshiyuki Inagaki (Shikoku Research Center, Forestry and Forest Products Research Institute) and Dr Yuichiro Oribe (Tohoku Regional Breeding Office, Forestry and Forest Products Research Institute) for providing information about their studies on growth phenology and xylogenesis. We also thank Dr Hisashi Sugita and Dr Shoji Noguchi (Forestry and Forest Products Research Institute) for providing forest census data for several test plots. We also thank Dr Nicolas Delpierre and two anonymous reviewers for their careful reading of an earlier version of this article and their constructive comments. We also thank Yasuyuki Tsurumi (University of Tokyo Forests) for his support during the fieldwork at the Chiba sites. References Barton, K. 2016 MuMIn: Multi-model inference. R Packag. version 1.10.0. Available at: https://cran.r-project.org/web/packages/MuMIn/index.html (accessed on April 22, 2016). Begum, S., Nakaba, S., Oribe, Y., Kubo, T. and Funada, R. 2010 Cambial sensitivity to rising temperatures by natural condition and artificial heating from late winter to early spring in the evergreen conifer Cryptomeria japonica. Trees - Struct. Funct.  24, 43– 52. Google Scholar CrossRef Search ADS   Berk, M. 2013 Smoothing-splines Mixed-effects Models [R package sme version 0.8]. Available at: https://cran.r-project.org/web/packages/sme/index.html (accessed on 16 June, 2017]. Cocozza, C., Lasserre, B., Giovannelli, A., Castro, G., Fragnelli, G. and Tognetti, R. 2009 Low temperature induces different cold sensitivity in two poplar clones (Populus×canadensis Mönch ‘I-214’ and P. deltoides Marsh. ‘dvina’). J. Exp. Bot.  60, 3655– 3664. Google Scholar CrossRef Search ADS PubMed  Crawley, M.J. 2007 The R Book . Wiley, p. 942. Google Scholar CrossRef Search ADS   Delpierre, N., Berveiller, D., Granda, E. and Dufrene, E. 2016a Wood phenology, not carbon input, controls the interannual variability of wood growth in a temperate oak forest. New. Phytol.  210, 459– 470. Google Scholar CrossRef Search ADS PubMed  Delpierre, N., Vitasse, Y., Chuine, I., Guillemot, J., Bazot, S., Rutishauser, T., et al.   2016b Temperate and boreal forest tree phenology: from organ-scale processes to terrestrial ecosystem models. Ann. For. Sci.  73, 5– 25. Google Scholar CrossRef Search ADS   Deslauriers, A., Rossi, S. and Anfodillo, T. 2007 Dendrometer and intra-annual tree growth: What kind of information can be inferred? Dendrochronologia  25, 113– 124. Google Scholar CrossRef Search ADS   Deslauriers, A., Rossi, S., Anfodillo, T. and Saracino, A. 2008 Cambial phenology, wood formation and temperature thresholds in two contrasting years at high altitude in southern Italy. Tree. Physiol.  28, 863– 871. Google Scholar CrossRef Search ADS PubMed  Drew, D.M. and Downes, G.M. 2009 The use of precision dendrometers in research on daily stem size and wood property variation: a review. Dendrochronologia  27, 159– 172. Google Scholar CrossRef Search ADS   Duchesne, L., Houle, D. and D’Orangeville, L. 2012 Influence of climate on seasonal patterns of stem increment of balsam fir in a boreal forest of Québec, Canada. Agric. For. Meteorol.  162–163, 108– 114. Google Scholar CrossRef Search ADS   Eilmann, B., Zweifel, R., Buchmann, N., Graf Pannatier, E. and Rigling, A. 2011 Drought alters timing, quantity, and quality of wood formation in Scots pine. J. Exp. Bot.  62, 2763– 2771. Google Scholar CrossRef Search ADS PubMed  Ford, K.R., Harrington, C.A. and St. Clair, J.B. 2017 Photoperiod cues and patterns of genetic variation limit phenological responses to climate change in warm parts of species’ range: modeling diameter-growth cessation in coast Douglas-fir. Glob. Chang. Biol.  23, 3348– 3362. Google Scholar CrossRef Search ADS PubMed  Fukuda, M., Iehara, T. and Matsumoto, M. 2003 Carbon stock estimates for sugi and hinoki forests in Japan. For. Ecol. Manage.  184, 1– 16. Google Scholar CrossRef Search ADS   Gruber, A., Strobl, S., Veit, B. and Oberhuber, W. 2010 Impact of drought on the temporal dynamics of wood formation in Pinus sylvestris. Tree. Physiol.  30, 490– 501. Google Scholar CrossRef Search ADS PubMed  Guillemot, J., Francois, C., Hmimina, G., Dufrêne, E., Martin-StPaul, N.K., Soudani, K., et al.   2017 Environmental control of carbon allocation matters for modelling forest growth. New. Phytol.  214, 180– 193. Google Scholar CrossRef Search ADS PubMed  Horiuchi, T. and Sakai, A. 1973 Effect of temperature on the frost hardiness of Cryptomeria japonica (in Japanese with English summary). J. Japanese For. Soc.  55, 46– 51. Imagawa, K., Oikawa, N. and Kitada, K. 1990 Simple method of measurement of change in stem diameter of Cryptomeria japonica trees (in Japanese). Trans. 42th Annu. Meet. Tohoku Branch Japanese For. Soc., 130–131. Japan Meteorological Agency 2002 Mesh Climate Data 2000 (CD-ROM) . Japan Meteorological Agency. Jyske, T., Mäkinen, H., Kalliokoski, T. and Nöjd, P. 2014 Intra-annual tracheid production of Norway spruce and Scots pine across a latitudinal gradient in Finland. Agric. For. Meteorol.  194, 241– 254. Google Scholar CrossRef Search ADS   Kawasaki, T. and Takeuchi, I. 1993 Diameter growth of trees with various sized in a Chamaecyparis Obtusa Plantation (in Japanese). Trans. 2 Annu. Meet. Kansai Branch Japanese For. Soc., 157–158. Komiyama, A., Higo, M., Kurumado, K. and Matsuhashi, T. 1992 Seasonality in radial growth of several trees (in Japanese). Trans. 40th Annu. Meet. Chubu Branch Japanese For. Soc., 73–74. Lupi, C., Morin, H., Deslauriers, A. and Rossi, S. 2010 Xylem phenology and wood production: resolving the chicken-or-egg dilemma. Plant. Cell. Environ.  33, 1721– 1730. Google Scholar CrossRef Search ADS PubMed  Mäkinen, H., Seo, J.W., Nöjd, P., Schmitt, U. and Jalkanen, R. 2008 Seasonal dynamics of wood formation: a comparison between pinning, microcoring and dendrometer measurements. Eur. J. For. Res.  127, 235– 245. Google Scholar CrossRef Search ADS   Nakagawa, S. and Schielzeth, H. 2013 A general and simple method for obtaining R2 from generalized linear mixed-effects models. Methods Ecol. Evol.  4, 133– 142. Google Scholar CrossRef Search ADS   Nanami, S., Kawaguchi, H. and Yamakura, T. 2010 Seasonal patterns of stem diameter growth in woody species with different life forms and geographical distributions at Mt. Mikasa, Nara, Japan. Appl. For. Sci  19, 27– 34. Nishizono, T. 2010 Effects of thinning level and site productivity on age-related changes in stand volume growth can be explained by a single rescaled growth curve. For. Ecol. Manage.  259, 2276– 2291. Google Scholar CrossRef Search ADS   Nishizono, T., Kitahara, F., Iehara, T. and Mitsuda, Y. 2013 Geographical variation in age–height relationships for dominant trees in Japanese cedar (Cryptomeria japonica D. Don) forests in Japan. J. For. Res  19, 305– 316. Google Scholar CrossRef Search ADS   Nishizono, T., Kitahara, F., Iehara, T. and Mitsuda, Y. 2015 Effect of climatic condition on regional differences in age-diameter relationships for dominant trees in Japanese cedar (Cryptomeria japonica D. Don) forests in Japan. (in Japanese with English summary). Kanto J. For. Res  66, 5– 8. Noda, T., Nakao, T. and Kuroki, Y. 1991 Analysis of diameter growth curve of trees in south Kyusyu (in Japanese). Trans. 44th Annu. Meet. Kyushu Branch Jpn. For. Soc., 61–62. Oribe, Y. and Kubo, T. 1997 Effect of heat on cambial reactivation during winter dormancy in evergreen and deciduous conifers. Tree. Physiol.  17, 81– 87. Google Scholar CrossRef Search ADS PubMed  Pinheiro, J.C. and Bates, D.M. 2000 Mixed-effects Models in S and S-PLUS . Springer, p. 528. Google Scholar CrossRef Search ADS   Rossi, S., Anfodillo, T., Čufar, K., Cuny, H.E., Deslauriers, A., Fonti, P., et al.   2016 Pattern of xylem phenology in conifers of cold ecosystems at the northern hemisphere. Glob. Chang. Biol.  22, 3804– 3813. Google Scholar CrossRef Search ADS PubMed  Rossi, S., Deslauriers, A., Anfodillo, T. and Carraro, V. 2007 Evidence of threshold temperatures for xylogenesis in conifers at high altitudes. Oecologia  152, 1– 12. Google Scholar CrossRef Search ADS PubMed  Rossi, S., Deslauriers, A., Anfodillo, T., Morin, H., Saracino, A., Motta, R., et al.   2006 Conifers in cold environments synchronize maximum growth rate of tree-ring formation with day length. New. Phytol.  170, 301– 310. Google Scholar CrossRef Search ADS PubMed  Rossi, S., Deslauriers, A., Griçar, J., Seo, J.W., Rathgeber, C.B.K., Anfodillo, T., et al.   2008 Critical temperatures for xylogenesis in conifers of cold climates. Glob. Ecol. Biogeogr.  17, 696– 707. Google Scholar CrossRef Search ADS   Saito, S., Kabeya, D., Araki, M., Tanaka, K., Migita, C. and Kawasaki, T. 2016 Phenology of diameter and height growth for Japanese Cedar (in Japanese with English summary). Kanto J. For. Res  67, 203– 206. Sakai, A. and Larcher, W. 1987 Frost Survival of Plants: Responses and Adaptation to Freezing Stress . Springer Berlin Heidelberg, p. 321. Google Scholar CrossRef Search ADS   Tardif, J., Flannigan, M. and Bergeron, Y. 2001 An analysis of the daily radial activity of 7 boreal tree species, northwestern Quebec. Environ. Monit. Assess.  67, 141– 160. Google Scholar CrossRef Search ADS PubMed  Vieira, J., Rossi, S., Campelo, F. and Nabais, C. 2014 Are neighboring trees in tune? Wood formation in Pinus pinaster. Eur. J. For. Res.  133, 41– 50. Google Scholar CrossRef Search ADS   Yamashita, K., Okada, N. and Kamo, K. 2006 Application of the wire dendrometer for monitoring the radial growth of trees. Mokuzai Gakkaishi  52, 8– 18. Google Scholar CrossRef Search ADS   Yamato, K., Nakao, T. and Kuroki, Y. 1989 Seasonal diameter growth of trees in South Kyusyu (in Japanese). Trans. 42th Annu. Meet. Kyushu Branch Japanese For. Soc., 145–146. © Institute of Chartered Foresters, 2018. All rights reserved. For Permissions, please e-mail: journals.permissions@oup.com.

Journal

Forestry: An International Journal Of Forest ResearchOxford University Press

Published: Apr 1, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 12 million articles from more than
10,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Unlimited reading

Read as many articles as you need. Full articles with original layout, charts and figures. Read online, from anywhere.

Stay up to date

Keep up with your field with Personalized Recommendations and Follow Journals to get automatic updates.

Organize your research

It’s easy to organize your research with our built-in tools.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

Monthly Plan

  • Read unlimited articles
  • Personalized recommendations
  • No expiration
  • Print 20 pages per month
  • 20% off on PDF purchases
  • Organize your research
  • Get updates on your journals and topic searches

$49/month

Start Free Trial

14-day Free Trial

Best Deal — 39% off

Annual Plan

  • All the features of the Professional Plan, but for 39% off!
  • Billed annually
  • No expiration
  • For the normal price of 10 articles elsewhere, you get one full year of unlimited access to articles.

$588

$360/year

billed annually
Start Free Trial

14-day Free Trial