Changes in canopy transpiration due to thinning of a Cryptomeria japonica plantation

: There is a strong pressing to clarify the effect of thinning coniferous plantations on components of the forest water cycle in Japan. This study evaluates changes in canopy transpiration ( E ) due to thinning of a Cryptomeria japonica plantation, the most common type of plantation in Japan. Using E derived with the sap-flux method, we modeled canopy conductance ( G c ) for periods before thinning and after thinning with the input of the vapor pressure deficit separately. We hypothetically calculated E values using these G c models, respectively, under the same meteorological conditions. The ratio of E estimated using the G c model after thinning to that before thinning was 56%. This value was comparable to the ratio of the total sapwood area for the stand ( A s ) before thinning to that after thinning (63%). This suggests the possibility of predicting the relative change in E due to thinning using data for the mean diameter at breast height and stem density for the target stand before and after thinning, which are readily available for most japonica plantations in Japan and are used to predict A s .


INTRODUCTION
Forests are often situated upstream of urban and agricultural areas, and the water cycle in forested areas affects water resources downstream (Vertessy et al., 2001). The effect of forestry practices (e.g., clearcutting, afforestation, and thinning) on the water cycle of a forest is an important topic in forest hydrology. Numerous studies (e.g., Komatsu et al., 2008a) have examined changes in catchment runoff due to forestry practices. Besides these studies, recent studies (e.g., Forrester et al., 2012) have examined changes in various components of the forest water cycle (e.g., transpiration and interception evaporation) due to forestry practices. Such examinations enhance our understanding of processes underlying changes in catchment runoff.
In Japan, there is a strong need to clarify the effect of thinning on components of the water cycle for coniferous plantations (Komatsu et al., 2007). Coniferous plantations cover approximately 40% of the forested area in Japan. These plantations need to be thinned twice or three times before they are harvested at an age of ~50 years. Such thinning has been phased out since 1980 mainly because of an increase in the importation of cheap timbers and woody products from other countries and an increase in employment costs. 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 reduce catchment runoff and water resources.
Among components of evapotranspiration, several studies (Hattori and Chikaarashi, 1988;Murai and Kumagai, 1989) have examined changes in interception evaporation due to thinning of coniferous plantations in Japan. On the other hand, information on changes in canopy transpiration (E) induced by thinning of coniferous plantations in Japan is quite limited. There has been only one study (Morikawa et al., 1986) that examined changes in E due to thinning. Further, there have been no studies examining changes in E due to thinning of Cryptomeria japonica plantations, although C. japonica is the most dominant plantation species.
This study aimed to evaluate changes in E due to thinning of a C. japonica plantation. We used data for meteorological factors and E derived using the sap-flux method for two months in growing seasons before thinning and after thinning. As E is the multiplicative product of the sapwood area (A s ) and the mean sap flux per unit sapwood area (J s ) for the stand (Kumagai et al., 2007), the change in E relates to changes in A s and J s . Our hypothesis was that the change in E would be primarily due to the change in A s , and therefore, the relative change in E due to thinning would be comparable to that in A s . We developed this hypothesis on the basis of the results of Kumagai et al.'s (2007) study. Kumagai et al. (2007) examined differences in E between two C. japonica stands with different stand structure (i.e., stem density, the diameter at breast height (DBH), and tree height). They reported that the difference in E between stands was primarily due to the difference in A s . There are several studies (e.g., Lagergren et al., 2008) reporting relatively small differences in E between thinned and control plots (or before and after thinning) especially for drought periods, because the practice of thinning could relax soil-water competition among trees by reducing interception evaporation and canopy transpiration. This process was not considered in our hypothesis. Soil-water competition is not expected to be severe in Japan, because precipitation is generally higher than potential evaporation (Komatsu et al., 2008b).

Site description
We used data for meteorological factors and E obtained at the Yamanokami site. The site is situated 12 km east of Fukuoka city, Japan (33°38'N, 130°31'E, 100 m a.s.l.). The mean air temperature, recorded between 1995 and 2005 at a meteorological observatory situated 15 km from the site, was approximately 17.2°C. The mean annual precipitation was 1790 mm.
The site was approximately 0.12 hectare in area and located on an east-facing slope of a small hill. The site was covered with C. japonica plantations with an age of 39 years in 2010. Surrounding vegetation was broadleaf forest. Thinning was performed uniformly at the site in October 2010. Note that thinning of coniferous plantations in Japan is generally performed during fall-winter. Trees to be cut were selected randomly and pruning of the remaining trees was not performed.
Data on meteorological factors were recorded at a station in an open space situated adjacent to the site (Supplement Text S1). Data for E were recorded at a 10 m × 10 m plot located in the middle of the slope. Stand structure for the plot before and after thinning is summarized in Table I. The ratio of A s after thinning to that before thinning was 63%. (Measurements of the sapwood area are detailed later in the text.)

Data
Data for E were determined from sap flux measurements using the thermal dissipation method with Granier-type sensors (Granier, 1987, Supplement Text S2). Sap flux measurements were made for all trees in the plot during August 2010-September 2011. A sensor was inserted, before thinning, on the northern inside of each tree at a depth of 0-20 mm to measure sap flux density (F d ), i.e., radial and circumference variations in F d were not considered in this study. We aimed to assess changes in E due to thinning relative to E recorded before thinning. This assessment would be possible even when neglecting radial and circumference variations in F d because of the correlation between F d values for different directions/depths for C. japonica trees (Shinohara et al., 2013). Furthermore, the following information could legitimize our experimental design. Consideration of radial and circumference variations in F d is less important than that of tree-to-tree variations for estimating E of a C. japonica plantation (Shinohara et al., 2013). F d decreases with depth for C. japonica trees (Kumagai et al., 2007) and the sapwood area at a depth of 0-20 mm for the plot was approximately 80% of the total sapwood area for the plot.
Using F d data, we estimated E during August-September 2010 and August-September 2011 (i.e., the same months of the year before and after thinning) because we aimed to model canopy conductance (G c ) before and after thinning under similar meteorological conditions. When F d data for a tree were unavailable in a specific period, we filled the data gap using the relationship for F d between the tree and another tree. Using these data, E was estimated as (Kume et al., 2010) (1) where a s is the tree sapwood area, n is the number of the trees included in the plot, and A G is the ground area. a s was determined on the basis of measurements of sapwood thickness. Sapwood thickness was determined for all individuals in the plot using a ruler on a core extracted with a 5-mm increment borer at 1.3 m above the ground and assessed as the mean of two orthogonal measurements. Distinct color differences were used to identify the boundary between sapwood and heartwood. a s was obtained from the difference between the heartwood area and the stem crosssectional area beneath the bark, where we assumed that the stem cross-sections were circular. In sapwood of C. japonica, a "white zone", the water content of which is much lower than that of the heartwood, exists adjacent to the heartwood. There is no water movement in the white zone (Kumagai et al., 2005). The width of the white zone was assumed to be 10 mm, as in Kumagai et al. (2007). The area of the white zone was subtracted from the sapwood area to determine the sapwood area effective for water movement (i.e., a s )

Methods of analysis
We first calculated and modeled canopy conductance (G c ) with the input of meteorological factors separately for the periods before and after thinning. We then hypothetically calculated E using the G c models before and after thinning under the same meteorological conditions and assessed the change in E due to thinning.
We calculated G c using the simplified Penman-Monteith equation (McNaughton and Black, 1973): Table I. Stand structure of the plot for transpiration measurements * 1 We calculated the mean DBH, total basal area, mean tree height, and total sapwood area after thinning from measurements made on March 3, 2010; i.e., before thinning. We thus assumed that temporal changes in these parameters due to tree growth were negligible. * 2 Measured using the plant-canopy analyzer (LI-2000, Li-Cor Inc., Lincoln, Nebraska, USA) on September 28, 2010 and August 31, 2011 (i.e., before and after thinning, respectively). * 3 Diameter at breast height * 4 A numeral in the parentheses indicates the standard deviation.

Parameter
Before thinning After thinning* 1 The number of trees 11 6 Stem density (stems ha −1 ) 1100 600 Leaf area index* 2 (m 2 m −2 ) 4.7 2.3 Mean DBH* 3 (cm) 28.9 (6.1)* 4 31.6 (5.4) Total basal area (m 2 ha −1 ) 75. where γ is the psychrometric constant, λ is the latent heat of water vaporization, c p is the specific heat of air, and ρ is the air density. This equation is derived from the Penman-Monteith equation under the assumption of complete coupling between the canopy and atmosphere (see Supplement Text S3). G c was calculated as a daily average conductance using mean daytime T and D, and E summed over 24 hr but divided by daylight hours (Phillips and Oren, 1998). G c calculations were made only for days without rain, because F d data could be subject to noise on rainy days (Kumagai et al., 2008).
Using G c data, we developed separate G c models before and after thinning. We assumed that D and solar radiation (S) were the possible factors to be considered in our G c models. As D is generally the most important factor determining G c , we regressed the relationship between D and G c , on the basis of the least-squares method, to determine the function expressing the effect of D on G c (f(D)) (Oren et al., 1999): where G cref and s express the reference value for G c and the relative sensitivity of G c to D, respectively. We used G c data recorded when S was no less than 400 W m −2 for this regression, because G c was light-saturated under this condition for C. japonica stands examined by Kumagai et al. (2008). We then examined correlation between S and observed G c divided by f(D) to determine whether S needed to be considered. For the assessment of the change in E due to thinning, we calculated E using G c models separately determined using data before and after thinning. For these hypothetical calculations, we used meteorological data for August and September 2011 as input (Supplement Text S4). We confirmed that our results did not change qualitatively when using meteorological data obtained in August and September 2010. Figure 1 shows time series of meteorological factors and E in August and September 2010 and in August and September 2011. There was frequent precipitation in both years (Figure 1a and 1b). The total P values were 304.0 and 581.5 mm, respectively. S did not show clear temporal trends in either year (Figure 1c and 1d). Day-to-day variations in S generally corresponded to P. T tended to be higher in August than in September for both years (Figure 1e and 1f), although a reduction in T was observed in mid-August 2011. Day-to-day variations in D generally corresponded to those in T (Figure 1g and 1h). E was higher for 2010 than for 2011 (Figure 1i and 1j). Day-to-day variations in E generally corresponded to those in D in both years. The total E values were 71.0 and 37.3 mm, respectively.

RESULTS
We observed significant (p < 0.01) negative (R = −0.716 and R = −0.590, respectively) correlations between D and G c with S no less than 400 W m −2 for 2010 and 2011 according to a two-tailed Pearson's correlation coefficient test (Figure 2). Regressing the relationship between D and G c for 2010, G cref and s for 2010 were determined as 0.00341 m s −1 and 0.432 ln(kPa) −1 , respectively. Similarly, G cref and s for 2011 were determined as 0.00190 m s −1 and 0.430 ln(kPa) −1 , respectively. s was nearly the same for the two years, indicating that the difference in G c between 2010 and 2011 for a given meteorological condition was primarily due to the difference in G cref . The correlation between observed G c divided by f(D) and S was not significant (p >  (Figure 3), suggesting that taking S into account did not improve the predictability of G c models developed in this study. Additionally, correlation between observed G c divided by f(D) and the number of successive days with P less than 5 mm day −1 before the G c data recorded was not significantly negative (p > 0.10), suggesting unimportance of soil water content in determination of G c . Thus, G c (in m s −1 ) for 2010 and 2011 were respectively modeled as (see Supplement Text S5) and G c = 0.00190 (1.00 − 0.430 ln(D)).
(5) Figure 4 shows E values calculated using the G c models for 2010 and 2011, respectively, with input of the same meteorological data (i.e., those recorded in August and September 2011). E calculated using the G c model for 2010 was higher than that calculated using the G c model for 2011 throughout the period. The total E was 67.0 mm for the G c model for 2010 and 37.3 mm for 2011. The ratio of the latter relative to the former was 56%. This ratio was comparable to the ratio of A s before thinning to that after thinning (63%). E for each day determined using the G c model for 2011 strongly correlated with that determined using the G c model for 2010 (R > 0.99, p < 0.01), which agrees with the fact that s was nearly the same for 2010 and 2011.

DISCUSSION
The relative change in E due to thinning was comparable to that in A s for our case. This is expected on the basis of the assumption that P is generally higher than potential evaporation (E p ) in Japan. The data used in this study indeed satisfy this assumption. The total E p values for August and September 2010 and for August and September 2011 were estimated as 283 and 254 mm, respectively (Supplement Text S6), on the basis of observed meteorological data ( Figure 1). The total P values for these periods were 304.0 and 581.5 mm, respectively. Thus, P is higher than E p in both periods.
The assumption of higher P than E p is generally valid in an average year in Japan, implying that our results would be reflective of most cases in Japan. This is supported by the data presented by Morikawa et al. (1986), who carried out another study on changes in E due to thinning for a coniferous plantation in Japan. Morikawa et al. (1986) did not explicitly present A s data for their plot and therefore did not compare the relative change in E with that in A s . However, it is possible to estimate A s for their plot using data for the DBH and stem density (Table 1 of Morikawa et al., 1986) and the relationship between DBH and a s (Figure 1 of Morikawa et al., 1986). A s values before and after thinning are estimated as 23.3 and 17.4 m 2 ha −1 , respectively. Thus, the relative change in A s is 74%, which is comparable to that in E (79%). There are very few studies reporting both changes in E and A s due to thinning of forests not only in Japan but also in other countries. However, the results of Gebauer et al. (2011) examining transpiration at a single-tree scale seem to support our discussion (Supplement Text S7).
It is possible that the assumption of P being higher than E p is not satisfied for specific periods in a low-precipitation year in Japan. The relative change in E due to thinning might be less than that in A s in such cases. Thinning would relax soil-water competition among trees by reducing interception evaporation and canopy transpiration. This results in less significant reduction in transpiration during drought periods for a thinned plot than for a control plot. Several studies (Simonin et al., 2007;Lagergren et al., 2008) have observed such a phenomenon. Simonin et al. (2007) measured E for a pine forest. They reported considerable differences in E between thinned and control plots when soil water content was high. However, the difference was much less when the soil water content was low. Lagergren et al. (2008) conducted measurements in a pine-spruce forest and compared E between thinned and control plots. E for the thinned plot was generally lower than that for the control plot in the first year after thinning. However, E for the thinned plot was rather higher than that for the control plot in the drought period (July-September) of the year when soil water content was low owing to low precipitation. We thus recommend further studies examining E for coniferous plantations in Japan before and after thinning using E data recorded in low-precipitation years.
Thinning could improve a light environment in the stand and increase E relative to A s or tree-scale transpiration (Tang et al., 2003). In reality, we did not obtain any evidence of such an increase for our site. This result would not change even when the intensity and/or method of thinning were different from those for this study. We obtained no evidence of an effect of S on G c even for the case before thinning (Figure 3a). This suggests insignificant light competition among trees for our site, because the relationship between S and G c reflects the change in stomatal conductance with S, and therefore our results suggest light saturation of stomatal conductance even for the case before thinning. Thus, the improved light environment induced by thinning would not contribute to increasing stomatal conductance for our site. When we analyzed the data at hourly (or halfhourly) time resolution, we might observe evidence of an effect of S on G c (Komatsu et al., 2006). However, this effect is unimportant in determination of G c and E at daily or longer time scales (Figure 3).
Evidence of an effect of S on G c for forests has been reported in previous studies including that made by Kumagai et al. (2008) on C. japonica stands. Thinning might increase E relative to A s or tree-scale transpiration for such cases. However, we presume that this increase would not be considerable. Variations in the relationship between S and G c cause slight differences in E for forests at daily or longer time-scales (Komatsu, 2004). Indeed, most studies examining changes in E due to thinning at these scales did not observe evidence of an increase in E due to the improved light-environment by thinning (Whitehead and Kelliher, 1991;Lagergren et al., 2008).
This study did not attempt to examine gradual changes in E after thinning that might persist for several years. Previous studies for other species (Bréda et al., 1995;Lagergren et al., 2008) reported such gradual changes in E for several years after thinning. Measurement studies at a multi-year scale are required to examine whether such gradual changes in E could occur for C. japonica plantations in Japan. Our results have implications for such changes. Water and light competition just after thinning would be less significant than that for the succeeding years. However, the change in E was comparable to that in A s , i.e., J did not increase due to thinning, according to data recorded just after thinning. This suggests that J for the succeeding years would be nearly the same as that just after thinning and that the change in E for the years would be caused primarily by the change in A s , if meteorological factors for the years did not differ greatly from those for the experimental period of this study.
An increase in leaf area for the succeeding years could cause an increase in E. However, the increase in leaf area would accompany an increase in a s and therefore that in A s according to the pipe-model theory (Shinozaki et al., 1964a,b) and observation results for other species (Medhurst and Beadle, 2002). Therefore, an increase in E caused by the increase in leaf area would be regarded as corresponding to the increase in A s . Here, we assumed that the relationship between leaf area and a s for an unthinned tree does not change by thinning. Although this assumption is valid for several species (Medhurst and Beadle, 2002), it is unclear whether it is the case for C. japonica trees. Further studies examining this uncertainty are required.
It is, at this stage, unclear whether the relative change in E is generally comparable to that in A s for C. japonica plantations in Japan. If this is the case, it suggests the possibility of predicting the relative change in E due to thinning on the basis of data for stem density and the mean DBH for a stand before and after thinning. Tsuruta et al. (2011) found a linear relationship between the DBH and a s for C. japonica trees (and a similar relationship for C. obtusa trees). If we input data for the mean DBH and stem density for a stand before and after thinning, we can obtain data for A s before and after thinning of the stand. On the basis of these A s data and the assumption that the relative change in E is equal to that in A s , we can predict the relative change in E due to thinning. This method for predicting the relative change in E is practically useful. Data required for the prediction are the mean DBH and stem density. These data are generally available for most C. japonica (and also C. obtusa) plantations in Japan. Besides A s , the leaf area index (LAI), stem density, and basal area could be used to predict the relative change in E due to thinning ( Table I). The use of A s or LAI is more process-based than the use of the stem density or basal area, because A s and LAI directly relate to E. The use of A s is more practical than the use of LAI, because the error in A s estimates would be generally less than that in LAI estimates (Tsuruta et al., 2012).

ACKNOWLEDGEMENT
We acknowledge staff at Kasuya Research Forest, Kyushu University for their thinning practice. Thanks are also due to Dr. Shin'ichi Iida (Forestry and Forest Products Research Institute, Japan) and two anonymous reviewers for providing critical comments. This work is funded by a CREST project (Field and Modeling Studies on the Effect of Forest Devastation on Flooding and Environmental Issues).

SUPPLEMENTS
Text S1. Measurements of meteorological factors Text S2. Granier-type sensors Text S3. Use of the simplified Penman-Monteith equation Text S4. Interaction between transpiration and meteorological factors Text S5. Comparison of the model parameters determined Text S6. Potential evapotranspiration estimates Text S7. Change in tree-transpiration with thinning Figure S1. Relationships between the vapor pressure deficit (D) and canopy transpiration (E) Figure S2. Relationships between the vapor pressure deficit (D) and tree transpiration