Introduction

Forests are major biomes present throughout the world and are often located upstream of agricultural and urban areas (Sawano et al. 2005; Trisurat et al. 2010). Thus, the water cycle of forest is related to catchment runoff and, therefore, water resources (Macfarlane et al. 2010; Ford et al. 2011).

Aging, natural disturbances (e.g., insect infestation) and forestry practices (e.g., thinning) affect forest structure and possibly its water cycle (Komatsu et al. 2008; Licata et al. 2011). Transpiration by canopy trees (E) is a major component of a forest water cycle (Note that the symbols used in this paper are summarized in Table 1). E changes greatly with changes in forest structure, suggesting changes in catchment runoff (Alsheimer et al. 1998; Vertessy et al. 2001). For example, Vertessy et al. (2001) analyzed changes in E with age for mountain ash forests in Australia. E is the multiplicative product of the sapwood area at stand scale (A) and the mean sap flux per unit sapwood area (J) for the stand. A decreased with age as a result of a decrease in stand density (N), while J did not change systematically with age. E thus decreased with age, suggesting that runoff from catchments covered with mountain ash forests increases with forest age.

Table 1 Symbols used in this paper
Full size table

In Japan, there is a strong need to clarify the effect of thinning on components of the water cycle for coniferous plantations (Komatsu 2010; Onda et al. 2010). Coniferous plantations cover approximately 40 % of the forested area in Japan or 25 % of Japan’s land surface (Japan Forestry Agency 2013). These plantations need to be thinned twice or three times before they are harvested at an age of ~50 years. Plantations in Japan appear to have been generally thinned before 1980 because of relatively high prices of timber (Japan Forestry Agency 2013). However, such thinning has since been phased out mainly because of an increase in the importation of cheap timbers and woody products from other countries and an increase in employment costs (Fujimori 2000). Several leading researchers in Japan (Tsukamoto 1998; Kuraji 2003) pointed out that plantations that have not been thinned could consume more water by evapotranspiration from the dense canopies and, consequently, reduce catchment runoff and water resources. Primarily to secure water resources, many local governments in Japan impose taxes and distribute subsidies to forest owners to aid thinning of coniferous plantations (Imawaka and Sato 2008; Morotomi and Numao 2012).

Our knowledge concerning changes in the water cycle due to thinning of coniferous plantations in Japan is limited despite its necessity (Komatsu et al. 2010; Dung et al. 2011). This is especially true for changes in E. There have been two studies (Morikawa et al. 1986; Komatsu et al. 2013) that measured E for two successive growing seasons just before and after thinning on the basis of the sap-flux method. The change in E due to thinning relative to E before thinning was comparable to that in A for these studies, suggesting the importance in the change in A rather than that in J (Komatsu et al. 2013). Thinning of coniferous plantations in Japan could enhance radial stem growth (Fukada et al. 2009; Miyamoto et al. 2009), suggesting an enhanced increase in sapwood area at single-tree scale (a) due to the correlation between stem diameter at breast height (d) and a (Kumagai et al. 2005; Kume et al. 2010). An enhanced increase in a may enhance an increase in A. However, it is unclear if a thinned plantation retains a lower A as compared to a non-thinned plantation for several years after thinning.

To clarify this, this study examined changes in A for several years after thinning of Japanese cedar (Cryptomeria japonica) and cypress (Chamaecyparis obtusa) plantations. These species are very common and occupy ~70 % of the total plantation area in Japan (Japan Forestry Agency 2013). Previous studies (Takeuchi et al. 1975; Tamai et al. 1983; Kawahara et al. 1989) reported the mean d (\( \bar{d} \)) and N for a plot just before thinning and for four to six years after thinning of Japanese cedar and cypress plantations. We estimated changes in A during the experimental period on the basis of the data presented in these studies. To calculate A, we use empirical equations of the relationships between d and a for Japanese cedar and cypress trees developed by Tsuruta et al. (2011).

Materials and methods

Data used

We used data for \( \bar{d} \) and N for two Japanese cedar site (TCE and KCE) and two cypress sites (KCY and TCY). A general description of the sites is given in Table 2. Four or five plots were installed for each site. \( \bar{d} \) and N just before thinning did not differ greatly among the plots for each site. Thinning was performed for all plots except the control plot at each site. The ratio of N just after thinning for a thinned plot to that for its control plot (R N) differed among plots at a site, indicating differences in thinning intensity. Characteristics of the trees cut by thinning were different among plots. Suppressed trees were cut by thinning for Plots I–IV of TCE and Plots III of KCE, KCY and TCY, resulting in an increase in \( \bar{d} \) for these plots. Sub-dominant trees were cut for Plots II of KCE, KCY and TCY, resulting in a slight increase in \( \bar{d} \) for these plots. Dominant trees were cut for Plots I of KCE and KCY, resulting in a decrease in \( \bar{d} \). Line-thinning was performed for Plot I of TCY and only slightly affected \( \bar{d} \) for this plot. \( \bar{d} \) and N were measured for four to six years after thinning. Note that Tamai et al. (1983) actually reported data for 12 years after thinning. We used data for the first six years, because a second round of thinning was performed six years after the first round.

Table 2 Age, mean diameter at breast height (\( \bar{d} \)) and stem density (N) for plots of the four sites
Full size table

Change in \( \bar{d} \)

We first examined whether the change in \( \bar{d} \) for a thinned plot was greater than that for its control plot. For this, we calculated the difference (\( \Delta \bar{d} \)) between \( \bar{d} \) just after thinning and at the end of the experiment for each plot. We then calculated the index

$$ R_{{\Delta \bar{d}}} = \frac{{\Delta \bar{d}_{\text{t}} }}{{\Delta \bar{d}_{\text{c}} }} \times 100 , $$
(1)

where \( \Delta \bar{d}_{\text{t}} \) and \( \Delta \bar{d}_{\text{c}} \) are values of \( \Delta \bar{d} \) for the thinned and control plots, respectively (Fig. 1a). \( R_{{\Delta \bar{d}}} \) indicates the percentage of the \( \Delta \bar{d} \) increase for a thinned plot relative to that for its control plot. \( R_{{\Delta \bar{d}}} \) greater than 100 % indicates greater radial stem growth for the thinned plot than for its control plot.

Fig. 1
figure 1

Schematic drawings of \( \Delta \bar{d} \) for a thinned plot (\( \Delta \bar{d}_{\text{t}} \)) and its control plot (\( \Delta \bar{d}_{\text{c}} \)) (a) and δA just after thinning (δA s) and that at the end of the experiment (δA e) (b)

Full size image

Change in A

To examine the change in A, we calculated A for each plot from \( \bar{d} \) and N data. A can be expressed as

$$ A = \frac{{\sum\nolimits_{i = 1}^{NS} {a_{i} } }}{S} , $$
(2)

where S is the ground area of the plot. NS indicates the number of trees included in the plot. According to Tsuruta et al.’s (2011) empirical equations for the relationships between d (in cm) and a (in cm2) for Japanese cedar and cypress trees (detailed below), a is expressed as

$$ a = k_{1} d + k_{2} , $$
(3)

where k 1 and k 2 are empirical constants. k 1 and k 2 are 10.0 cm and −41.5 cm2 for cedar, respectively. k 1 and k 2 are 8.2 cm and −27.4 cm2 for cypress, respectively. Substituting Eq. (3) into Eq. (2), it follows that

$$ A = \frac{{\sum\nolimits_{i = 1}^{NS} {(k_{1} } d + k_{2} )}}{S} $$
(4)
$$ \quad= N(k_{1} \bar{d} + k_{2} ) $$
(5)
$$ \quad= N \times \bar{a} , $$
(6)

where \( \bar{a} \) is the value of a for a tree with d being equal to \( \bar{d} \). Note that \( \bar{a} \) is equal to the mean a for the trees in the plot (i.e., A/N) because of the linear relationship between d and a.

Tsuruta et al.’s (2011) equations were developed on the basis of a observed for 81 cedar trees in six plots and for 109 cypress trees in nine plots in Japan. Despite relatively wide ranges of tree age and N for the plots, they did not find any effects of these factors on the relationships between d and a. Thus, Tsuruta et al. (2011) simply regressed the relationships using linear equations and determined k 1 and k 2 for cedar and cypress separately. The relationships between d and a were actually non-linear for a broad range of d. However, the relationships could be approximated using linear equations when d was no more than ~30 cm (Tsuruta et al. 2011). We could not examine whether the data we used in this study satisfied these conditions, because frequency distributions of d for the plots during the experimental periods were not always available in Takeuchi et al. (1975), Tamai et al. (1983) and Kawahara et al. (1989). However, we presumed that the data would generally satisfy these conditions for the following reason. The coefficient of variation (CV) for d of the trees in a plot, calculated as the standard deviation (σ) divided by \( \bar{d} \), is generally less than 30 % for cedar plantations and 25 % for cypress plantations (Kunisaki 2001). The values of \( \bar{d} \) + 2σ for the plots we examined were less than 30 cm according to these CV values, even when using \( \bar{d} \) values recorded at the end of the experiments. Thus, d for most trees in the plots would be less than 30 cm, suggesting the validity of using Eq. (3) in this study. There are scatterings in the relationships between d and a (Fig. 1 of Tsuruta et al. 2011), suggesting uncertainty in k 1 and k 2 determined by Tsuruta et al. (2011). We confirmed that even when using k 1 and k 2 for the upper/lower envelopes of the relationships, our results did not change qualitatively.

On the basis of the A data calculated in the manner described above, we examined whether the difference in A between a thinned plot and its control plot (δA; i.e., A for the control plot minus that for the thinned plot) reduced to zero during the experimental period. To evaluate the rate of δA reduction, we then calculated the index R δA for each plot:

$$ R_{\delta A} = \frac{{ (\delta A_{\text{s}} - \delta A_{\text{e}} )/T_{\text{e}} }}{{\delta A_{\text{s}} }} \times 100 , $$
(7)

where δA s and δA e are respectively, δA just after thinning and at the end of the experiment, and T e is the period (in years) from the thinning to the end of the experiment (Fig. 1b). R δA indicates the percentage of the decrease in δA per year during the experiment relative to δA just after thinning. A positive R δA indicates a decrease in δA. To give an indication of the rate of a decrease in δA, we also calculated the time required for δA to fall to zero (T δA=0, in years) under the assumption of a constant rate of change in δA. T δA=0 is equal to 100 divided by R δA . T δA=0 was not calculated for plots with negative R δA (i.e., plots where δA increased).

Results

Change in \( \bar{d} \)

Figure 2a, b respectively, show changes in \( \bar{d} \) and N after thinning of the TCE site. Results for the other sites are shown in Fig. S1 of the supplementary material. \( \Delta \bar{d} \) values were different among plots; \( \bar{d} \) for Plot I increased from 12.1 to 17.5 cm during the experiment (Fig. 2a); i.e., \( \Delta \bar{d} \) was 5.4 cm. This \( \Delta \bar{d} \) value was greater than that for Plot C (2.6 cm). On the other hand, \( \bar{d} \) for Plot II increased from 10.2 to 12.9 cm; i.e., \( \Delta \bar{d} \) was 2.7 cm. This \( \Delta \bar{d} \) value was nearly the same as that for Plot C. N for some plots (i.e., Plots III and C) decreased slightly after thinning owing to natural mortality (Fig. 2b).

Fig. 2
figure 2

Changes in the mean diameter at breast height (\( \bar{d} \)) (a) and stem density (N) (b) for the TCE site

Full size image

\( \Delta \bar{d} \) values for many thinned plots (i.e., Plots I–IV of TCE, Plots I–III of KCE, and Plots I and II of TCY) were greater than those for their control plots (Fig. 3a). However, this was not always the case (i.e., Plots I–III of KCY and Plot III of TCY). \( R_{{\Delta \bar{d}}} \) varied among plots and the range was between 86 and 208 % (Fig. 3b). \( R_{{\Delta \bar{d}}} \) negatively correlated (P < 0.01, two-tailed Pearson’s correlation coefficient test) with R N (Fig. 4). Thus, greater \( \Delta \bar{d} \) values were observed only for plots with low R N due to intensive thinning. We did not observe clear differences in \( R_{{\Delta \bar{d}}} \) for a given R N according to the differences in the trees to be cut (data not shown), suggesting minor effects of the differences in the trees to be cut on determination of \( R_{{\Delta \bar{d}}} \).

Fig. 3
figure 3

Difference (\( \Delta \bar{d} \)) between the mean diameter at breast height just after thinning and that at the end of the experiment period for each plot (a); percentage of the \( \Delta \bar{d} \) increase for a thinned plot relative to that for its control plot (\( R_{{\Delta \bar{d}}} \)) (b)

Full size image
Fig. 4
figure 4

Relationship of the percentage of the \( \Delta \bar{d} \) increase for a thinned plot relative to that for its control plot (\( R_{{\Delta \bar{d}}} \)) with stem density just after thinning for a thinned plot relative to that for its control plot (R N). The regression line, determined employing the least-squares method, is y = −2.01x + 248

Full size image

Change in A

Figure 5a–c respectively, show changes in \( \bar{a} \), A, and δA after thinning of the TCE site. Results for the other sites are shown in Fig. S2 of the supplementary material. The change in \( \bar{a} \) for each plot (Fig. 5a) was qualitatively the same as that in \( \bar{d} \) (Fig. 2a), because \( \bar{a} \) was calculated from \( \bar{d} \) using Eq. (3). The increase in A after thinning did not differ greatly among plots, including the control plot (Fig. 5b), resulting in relatively small changes in δA after thinning (Fig. 5c). δA did not reduce to zero for the plots, which was also the case for all the thinned plots of the other sites (Figs. S2c, f and i of the supplementary material).

Fig. 5
figure 5

Changes in the sapwood area (\( \bar{a} \)) for a tree with d being equal to \( \bar{d} \) in the plot (a), the sapwood area at stand scale (A) (b), and the difference in A between a thinned plot and its control plot (δA) (c) for the TCE site. Note that \( \bar{a} \) is equal to the mean sapwood area for the trees in the plot

Full size image

R δA ranged between −2.9 and 7.2 % (Fig. 6a). T δA=0 was not available for 7 of the 13 plots because of negative R δA values (Fig. 6b). T δA=0 ranged between 13.9 and 119 years for the other plots. We did not observe a significant, positive correlation (P > 0.10, one-tailed Pearson’s correlation coefficient test) between \( R_{{\Delta \bar{d}}} \) and R δA (Fig. 7a). This suggests that despite increased radial stem growth for plots with low R N due to intensive thinning (Fig. 4), the effect of increased radial stem growth on the increase in A was not large enough to cancel the effect of small N due to intensive thinning of the plots. On the other hand, we observed a significant, positive correlation (P < 0.10, two-tailed Pearson’s correlation coefficient test) between R N and R δA (Fig. 7b), although the correlation coefficient was relatively low. This suggests that a decrease in δA was actually observed for plots with high R N, although the rate of radial stem growth was not high for these plots (Fig. 4). Relatively high R δA for a given R N was observed for plots satisfying either or both of two conditions: (1) \( R_{{\Delta \bar{d}}} \) for the plot was higher than that for other plots with similar R N (i.e., Plot II of the KCE site); and (2) an N decrease after thinning due to natural mortality was less considerable for the plot than for its control plot (i.e., Plots I–III of the TCY site).

Fig. 6
figure 6

Percentage of the decrease in δA per year during the experiment relative to δA just after thinning (R δA ) (a); time required for δA to reach zero (T δA=0) (b)

Full size image
Fig. 7
figure 7

Relationships of the decrease in δA per year during the experiment relative to δA just after thinning (R δA ) with the percentage of the \( \Delta \bar{d} \) increase for a thinned plot relative to that for its control plot (\( R_{{\Delta \bar{d}}} \)) (a); and with the ratio of stem density for a thinned plot to that for its control plot (R N) (b). The regression lines, determined via the least-squares method, are y = −0.0277x + 4.00 and y = 0.0877x−4.86 in Fig. 7a, b, respectively

Full size image

Discussion

Implications for the effectiveness of thinning in reducing transpiration

δA did not reduce to zero during the experiment for any plots in this study (Fig. 5c and Fig. S2 of the supplementary material). δA, rather, increased for more than half of the plots (Fig. 6a). δA decreased for the other plots, but the rate of decrease was relatively low (i.e., T δA=0 ranged between 13.9 and 119 years).

These results imply that a lower E for thinned plantations as compared to plantations not thinned remains for several years (more than 10 years according to the data for T δA=0) after thinning, given that the change in E after thinning was primarily caused by a change in A. This is especially true for plantations with intensive thinning. R δA was smaller for plots with lower R N (Fig. 7b), indicating a smaller decrease in δA for the plots. R δA was even negative for four of the five plots with R N < 60 %, indicating an increase in δA for these plots. Thus, our results imply that thinning (especially intensive thinning) is effective for reducing water consumption by canopy trees. Note that intensive thinning with R N ranging 40–60 % has been proposed for environmental control (e.g., securing water resources), because intensive thinning reduces stem density with lower economic costs than repetition of moderate thinning (Forestry and Forest Products Research Institute 2010).

Relationship between changes in the sapwood area and basal area

R δA was negative for most plots with R N less than 60 %; i.e., δA increased during the experiment. This result seems to contradict the results, derived by Miyamoto et al. (2009), of changes in the basal area at a stand scale (B A) after thinning. Miyamoto et al. (2009) examined changes in B A for five years after thinning of Japanese cypress plantations. The increase in B A during the five years for plots with R N less than 60 % tended to be greater than that for the control plot, indicating a decrease in the difference in B A between the thinned and control plots (δB A).

A and B A are different in that A is proportional to d (Eq. 3) but B A is proportional to d 2. A unit increase in d thus results in a greater increase in B A than in A. This difference between A and B A is the primary cause for the difference in the results between our study and Miyamoto et al. (2009). The following information also supports the above discussion. B A data are available for two of the five plots with R N less than 60 % (i.e., Plot III of KCE and Plot II of KCY). δB A for Plot III of KCE and Plot II of KCY indeed decreased during the experimental period (Kawahara et al. 1989), although δA increased during the same period (i.e., R δA was negative, Fig. 6a). These results are similar to those of Miyamoto et al. (2009). The difference between the increase in A and that in B A after thinning might suggest a difference in water use efficiency between thinned and control plots, although there have been very few studies examining this difference for Japanese cedar and cypress.

Applicability of our results to older stands

Our results (i.e., δA did not disappear during the experiment) were derived from data recorded at plantation sites having an age of 10–30 years. Thinning of coniferous plantations in Japan is also performed at an age of more than 30 years. This is especially true in recent years. Plantations in Japan were originally harvested at an age of ~50 years. These plantations have not been harvested at an age of ~50 years recently because of the stagnation of forestry. Consequently, thinning has also been performed for stands having an age of no less than 50 years for the purpose of environmental control with the aid by local governments (Japan Forestry Agency 2013).

It is, at this stage, unclear whether our results are applicable to stands having an age of more than 30 years, necessitating further experimental studies. However, the following discussion implies their applicability. Thinning alleviates light competition among trees in the stand and enhances radial stem growth as a result of a change in carbon allocation among the trees (Hayashi and Sakai 1972; Kawahara et al. 1989). Light competition would be less severe for stands having an age of more than 30 years than for stands having an age no more than 30 years because the leaf area indices for the former stands are generally lower than those for the latter stands (Kiyono 1988; Murakami et al. 2000). This implies that the increase in the rate of radial stem growth caused by thinning would not be considerable enough to cancel the effect of low N for stands having an age of more than 30 years, similar to our results obtained for stands having an age of 10–30 years. Additionally, Yashiro et al. (2010) reported that the net primary production for Japanese cedar stands reaches a maximum at 20–25 years of age and then declines with age. This implies that an increase in the rate of radial stem growth caused by thinning would not be considerable for stands older than 30 years, even when assuming no difference in light competition between stands older than 30 and those younger than 30.

Recommendations for further studies

No studies have examined changes in E employing the sap-flux method for no less than two years after thinning of Japanese cedar and cypress plantations (Morikawa et al. 1986; Komatsu et al. 2013). This is also the case for plantations comprising other tree species (Bréda et al. 1995; Lagergren et al. 2008). Measurements at a longer time scale are highly recommended for clarifying the changes in E after thinning cedar and cypress plantations, although there are technical problems applying the sap-flux method for measuring long-term changes in sap flux (e.g., reduction in the sap-flux caused by a sensor implantation; Moore et al. 2010). Our results for the change in A after thinning are useful for developing a hypothesis for changes in E for a longer time scale. Once data for changes in E after thinning of various plantations are accumulated, information on changes in A and other structural parameters (e.g., the leaf area index) after thinning will be useful in comparing and contrasting changes in E among different plantations.

In addition to examing E over a longer time frame, we recommend assessing changes in other components of evapotranspiration due to thinning of coniferous plantations in Japan (i.e., interception evaporation, understory transpiration and soil surface evaporation). These components are affected by thinning and therefore affect catchment runoff and water resources (Komatsu et al. 2009; Komatsu 2010).