Height Growth of Korean Pine Seedlings Planted under Strip-Cut Larch Plantations in Northeast China

1The University of Tokyo Hokkaido Forest, Graduate School of Agricultural and Life Sciences, The University of Tokyo, 9-61 Yamabe-Higashimachi, Furano 079-1563, Japan 2Graduate School of Environment and Information Science, Yokohama National University, 79-7 Tokiwadai, Hodogaya-ku, Yokohama 240-8501, Japan 3Dabiangou Forest Farm of Qingyuan County, 303 Provincial Road, Fushun, Liaoning 113311, China 4College of Forestry, Shenyang Agricultural University, 120 Dongling Road, Shenyang, Liaoning 110161, China


Introduction
Larches (Larix spp.) are important fast-growing and highyield tree species in Northeast China, and government policy has keenly promoted larch plantations in the region [1].According to the National Forest Resource Survey of 1988, there were 1.2 million hectares of larch plantations in the region [2].However, short-rotation management of pure larch plantations may have some disadvantages, such as soil degradation, loss of biodiversity, high vulnerability to disease, and insect outbreaks [3,4].To prevent water shortages and other environmental threats, a large proportion of the planted larch forest has been designated "ecological public welfare forest" where clear-cutting is prohibited [5].
To enhance ecosystem functioning while maintaining the economic benefits of larch plantations, a pressing need exists to develop alternative forest management regimes in Northeast China [5,6].
Forest management whereby even-aged plantation monocultures are transformed to two-storied mixed-species stands has been practiced in many forests around the world [7][8][9].Underplanting is combined with the partial cutting of the existing stand to promote the development of complex stand structure with multiple canopy layers and various tree sizes [9].Two-storied forest management in which overstory larch trees are strip-cut and Korean pine (Pinus koraiensis Sieb.et Zucc.)seedlings are underplanted has been deemed a viable option in Northeast China [6,10].Korean pine is an ecologically key species in the region, where mixed coniferbroad-leaved forests dominated by Korean pine are naturally distributed [11,12].Korean pine has great economic value, because it produces high-quality timber and commercial seeds for eating and medical use [13,14].
Korean pine is shade-tolerant in the early stages of growth but becomes gradually shade-intolerant as it ages [11,15].In Northeast China, Korean pine has frequently been planted in broad-leaved secondary forests to restore a climax-like vegetative cover; a large area of mixed forest has been established using this approach [12,16].Previous studies have analyzed the growth characteristics of Korean pine planted under overstory canopies with results indicating that the light environment is an important factor influencing its growth [12,[16][17][18][19].Consequently, additional research has examined the photosynthetic traits of Korean pine seedlings and saplings [6,11,16,[20][21][22][23]. Limited knowledge exists on the growth patterns of Korean pine seedlings planted under the larch canopy openings created by strip-cutting operations.
The purpose of this study was to examine the height growth of Korean pine seedlings planted under the larch canopy openings created by different types of strip-cutting treatments in Northeast China.A case analysis was conducted at two-storied forest stands established in eastern Liaoning Province.We employed a statistical modeling technique to quantify the height growth of underplanted Korean pine seedlings.Overstory larch canopies and the width of strip-cut were used as surrogates of light conditions, and the possible effects on seedling size were analyzed based on the derived growth model.Our analysis aimed to find optimal levels of strip-width for securing adequate height growth.We concluded with a brief discussion of management implications for the practice of underplanting Korean pine in strip-cut larch plantations.

Study Site.
We conducted this study at the Dabiangou Forest Farm of Qingyuan County in Liaoning Province (42 ∘ 00-02  N, 125 ∘ 03-04  E, 480-520 m above sea level).The forest is located in the mid to lower slopes of the mountainous area of the Changbai Mountain Longgang branch.The region has a marked continental monsoon climate featuring cold, dry winters and hot, rainy summers [24].The mean annual temperature and mean annual precipitation are 5 ∘ C and 700-800 mm, respectively, with a maximum temperature of 36.5 ∘ C, a minimum temperature of −37.7 ∘ C, and a frost-free period of 120-125 days [6].The major soil type is brown forest soil, with a soil depth of 30-50 cm [10].The main plantation species in the region are conifers such as Larix kaempferi (Lamb.)Carr., Larix olgensis Henry, Pinus koraiensis, and Pinus tabulaeformis Carr.[25].
We conducted field surveys in five larch plantations that had been established previously by the Dabiangou Forest Farm as experimental stands [6,26].The stands were composed of either L. olgensis or L. kaempferi.Based on the stand ages and the types of strip-cutting, we hereafter refer to these five plots as A12-W4.5,A12-W6.0,A17-W4.5, A17-W6.0, and A37-W7.5 (Table 1).The larches had originally been planted at a spacing of 1.5 m × 1.5 m in each stand, although some trees had been thinned or were dead at the time of survey.
In 2008, Korean pine seedlings were underplanted in the openings created by the strip-cuts [26].The seedlings were planted by the Forest Farm according to the strip width: in a line with 2.0 m spacing at A12-W4.5, in a line with 1.2 m spacing at A17-W4.5, in two lines with 2.0 × 1.2 m spacing at A12-W6.0, and 1.2 × 1.2 m spacing at A17-W6.0.Despite the widest strip the seedlings were sparsely planted in a center line with 1.2 m spacing at A37-W7.5, because natural tree regeneration was also expected to occur.
We set up one rectangular plot (234-612 m 2 ) in each stand.The plot A12-W4.5 spanned two strip openings while the other plots had one strip, in which Korean pine seedlings had been planted.The retained rows of larches on both sides of the strip(s) were included in the plot area.In each plot, the diameter at breast height (DBH) and height for all larch trees had been measured by the Dabiangou Forest Farm every year.Table 1 summarizes the stand parameters of overstory larch trees within each plot (tree density, mean DBH, mean tree height, and stand basal area).
Table 1 also shows the mean canopy openness (%), which was measured in June 2012 by taking hemispherical photographs at four or five randomly selected locations above the underplanted Korean pine seedlings in each plot.The photographs were taken at breast height (130 cm) using a digital camera (GX200, Ricoh Company, Ltd., Tokyo, Japan) fitted with a fisheye lens (UWC-0195, FIT Corporation, Nagano, Japan).We used imaging software (Gap Light Analyzer ver.2.0) to estimate the canopy openness [27].Significant differences were found among the plots ( < 0.001; oneway ANOVA).Generally, canopy openness was greater in the plots having older stand ages and wider strips after 4 years of the strip-cutting treatments.
In addition to the five plots mentioned above, we set up one 320 m 2 control plot in a young plantation hereafter referred to as OPEN to examine seedling growth in open sites.The plot OPEN was located at a distance of <5 km from the other plots and at the mid part of an east-facing slope with a slope angle of 44% where Korean pine and Korean spruce (Picea koraiensis Nakai) had been planted in 2007 alternately at a spacing of 2.0 × 2.0 m.Shrubs and herbs covered most of the forest floor.International Journal of Forestry Research 2.2.Data Collection and Analysis.Field measurement was conducted in mid-June 2012; four growing seasons after the Korean pine seedlings were underplanted in the larch plantations and five growing seasons after the planting in OPEN [26].The annual height growth of Korean pine seedlings was estimated by measuring the distance between the successive terminal bud scars (internodes or branch whorls) along the main stems downward [28,29] to a precision of 0.1 cm.We assumed that the current-year shoot growth had completed by early June [15] and thus regarded the length of the new shoots as the height growth in 2012.The number of Korean pine seedlings in a given plot ranged from 9 to 25, with a total of 85.The number of seedlings varied among the plots due to differences in plot size, initial planting density, and mortality.The mean rate of seedling survival was 66% (85/128), with the range between 39 and 79%.The number of annual height growth measurements in a given plot ranged from 36 to 100, with a total of 339.We estimated the mean height of Korean pine seedlings for each plot and each growing season by subtracting the annual height growths backward from the current tree height.Because the homogeneity of variance assumption was questionable ( < 0.05; Bartlett's test), we employed a nonparametric Scheffe's test for multiple comparisons to analyze differences between the plots.Next, a generalized linear mixed model (GLMM) analysis [30] was conducted to examine the annual height growth of Korean pine seedlings under different canopy conditions.GLMM provides a flexible modeling framework that allows the handling of data with a nonnormal distribution and the incorporation of random effects [31].We assumed that the logarithm of the height growth followed a normal distribution, because growth is a nonnegative continuous variable [32]: where   is the observed annual height growth (cm year −1 ) of seedling  in plot  at the th growing season,   is the expected annual height growth (cm year −1 ), and  2 is the variance.
Using the logarithmic link function [31], we expressed annual seedling growth as a function of three fixed effects, seedling height at the beginning of the growing season, basal area of overstory larch trees, and strip-cut width.We assumed that the effect of strip-cut treatment on height growth would vary according to residual overstory basal area.We included random effects to take into account the pseudoreplication within individual seedlings and plots.The growing years were considered as a temporal random effect: where   is the initial height (cm) of seedling  in plot  at the beginning of the th growing season;   is the basal area (m 2 ha −1 ) of overstory larch trees in plot  at the th growing season;   is a categorical variable that represents the stripcut width in plot  (i.e., 4.5, 6.0, and 7.5 m;   = 0 at OPEN); tree  , plot  , and year  are the random-effect parameters for individual seedling , plot , and the th growing season; and  0 ,  1 ,  2 , and  3 are fixed-effect parameters.
Backward stepwise selection method based on Akaike's information criterion (AIC) [33] was used to select the best model [34].We evaluated the goodness of fit of the selected model by determining the coefficient of determination  2 for GLMMs [35].We further used a leave-one-out crossvalidation approach, because neither independent data nor enough data to split into two groups were available [36].At each iteration, one observation was excluded from the data set and the selected model was fitted to the remaining observations [37].The analyses were conducted using the "lme4" [38], "lmerTest" [39], and "MuMIn" [40] packages in the "" 3.0.3statistical software [41].
Using the selected model with the estimated parameters, the annual height growth of Korean pine seedlings was predicted.We further simulated changes in seedling height in the early stages of growth by accumulating the predicted annual growth for each year.In the simulation, the initial seedling height at the time of planting was set at 20.0 cm.Our computations were continued for 7 growing years after planting.
Significant positive correlations were found between the initial seedling height at the beginning of each growing season and the observed annual height growth at the growing season for each plot ( < 0.01; Figure 2), with Pearson correlation coefficients of 0.40-0.77.The mean annual height growth was the largest in OPEN (23.7 cm year −1 ), followed by A37-W7.5 (20.1 cm year −1 ).All other plots (i.e., other than A37-W7.5)had significantly lower height growth (12.1-17.3cm year −1 ) than OPEN ( < 0.01; Scheffe's test).

Model Selection, Parameter Estimation, and Goodness of
Fit.The GLMM inclusive of all three fixed effects (i.e., initial seedling height, stand basal area of overstory larch trees, and strip-cut width) was selected as the best model (Table 2), with the lowest AIC value of 167.4.The second smallest AIC value was 172.6, when basal area of overstory larch trees was excluded as a fixed effect.According to the parameter estimates, initial seedling height had a positive effect on the annual height growth.Additionally, the height growth of   underplanted seedlings in the narrowest strip-cut width (i.e., 4.5 m) was significantly smaller than that in OPEN.The selected model and its associated parameter estimates provided an acceptable fit to the data, with a marginal  2 GLMM() (i.e., variance explained by fixed factors; [35]) of 0.49 and a conditional  2 GLMM() (i.e., variance explained by fixed and random factors) of 0.55.The leave-one-out cross-validation predictions fitted adequately to the observed height growth of seedlings ( 2 = 0.51; Figure 3), having a root mean square error of 4.7 cm.

Predicted Height Growth of Korean Pine Seedlings.
Figure 4 shows the predicted annual height growth of Korean pine seedlings as a function of initial seedling height and Figure 4: Relationships between the initial height and annual height growth of Korean pine seedlings, predicted by the generalized linear mixed model.In each strip-cut width, the basal area of overstory larch trees (BA) within a range of the plot data (Table 1) was used for the prediction.
stand basal area of overstory larch trees in each strip-cut width using the selected model (random-effect parameters were not used for the prediction).Seedlings planted in OPEN showed the largest predicted growth.The predicted growth of 120-cm seedlings (42.4 cm year −1 ) was more than twice the predicted growth of 20-cm seedlings (17.4 cm year −1 ).Generally, underplanted seedlings in wider strip-cuts tended to have relatively large height growth than those in narrower strip-cuts.That is, seedlings in the widest strip-cut (i.e., 7.5 m) had the largest predicted growth, whereas height growth was strongly suppressed at the narrowest strip-cut (i.e., 4.5 m).
The annual height growths in the 4.5 m strip were 36−47% smaller than in OPEN.In the wider strip-cuts (6.0-7.5 m), 13-36% reductions in annual height growth were predicted versus OPEN.
Figure 5 shows the predicted change in seedling height over a period of 7 years after planting in each strip-cut width using the selected model.The seedlings in OPEN were predicted to reach 301.6 cm after 7 years.The predicted seedling heights in the underplanted sites were smaller than those in OPEN.Seedling heights after 7 years in the 4.5 m strip were 109.4-112.9cm, which corresponded to 36-37% of the predicted seedling height in OPEN.In the wider strip-cuts (6.0-7.5 m), seedlings had greater predicted heights (136.8-222.9cm) after 7 years compared to the narrowest stripcut.

Discussion
Our GLMM analysis successfully derived a statistical model that allows the assessment of annual height growth of Korean pine seedlings at the early stages of growing.The selected model and estimated parameters fit the observed height growth with an adequate predictive accuracy.The exponential growth pattern delineated in Figure 5 is consistent with our field observations (Figure 1), as well as a previous study by Liu and Wang [42], which found an exponential relationship between the height and age of Korean pine seedlings for 5 growing years after planting in the same region.
According to our results, the initial seedling height at the beginning of the growing year positively influenced the annual height growth of Korean pine seedlings (Table 2).Although dealing with different coniferous tree species, positive correlations have also been observed at plantations of Japanese sugi (Cryptomeria japonica D. Don) and hinoki (Chamaecyparis obtusa (Sieb.et Zucc.)Endl.) when they were small in tree size [43].Kiyono et al. [44] examined the early height growth of underplanted Japanese hinoki using a multiple regression analysis and achieved a better predictive accuracy of the model by including initial seedling height as an explanatory variable.In the United States, Moores et al. [45] found that initial seedling height had a significant effect on predicting the annual height growth of three shade-tolerant coniferous species, including balsam fir (Abies balsamea (L.) Mill), red spruce (Picea rubens Sarg.), and eastern hemlock (Tsuga canadensis (L.) Carr.).Similarly, in the case of Korean pine, the initial height of seedlings should be considered an important predictor for height growth modeling.
Overstory larch canopies significantly affected the seedling height growth of underplanted Korean pine International Journal of Forestry Research (Table 2).As presented in Figure 4, seedlings planted in the open site (OPEN) showed superior height growth versus the underplanted seedlings.A relatively low light regime created by partial canopy closure may inhibit the height growth of underplanted Korean pine seedlings [12,19].In eastern Liaoning, Liu and Wang [42] reported a seedling height of 82.2 cm at 4 years after planting in an open site, which is relatively small compared to our control site (114.2cm at OPEN).This is probably due to their investigations being conducted at a higher elevation site (684 m above sea level) where relatively small-sized seedlings were initially planted (13.3 cm on average).
Our results also indicated that the type of strip-cut has a significant effect on the height growth of underplanted Korean pine seedlings.We observed a general increasing trend of canopy openness as the strip-cut width got wider (Table 1).In the stand with the narrowest strip (i.e., 4.5 m), the annual height growth was predicted to be approximately onehalf and the seedling height after 7 growing years was predicted to be one-third of those in the open site (Figures 4 and  5).Such a cramped strip-cut could not greatly improve the understory light conditions [16].Although stand age affected canopy openness on equal strip-cut width (Table 1, probably through a change in stand density; [46]), its effect on seedling height growth was ambiguous.Despite its larger canopy openness, A17-W6.0 produced smaller height growth than A12-W6.0(Figure 1).Available light for tree regeneration establishment (natural or artificial) can be increased through strip-cutting, but not so much as to promote competition by understory shrubs or herbaceous plants [9].In addition to larch canopy conditions, the competitive effect of understory vegetation may be a factor for consideration in growth predictions.
We assume that the seedling height growth observed in the 4.5 m strip-cut is inadequate for forest management purposes.Because larger pine trees generally produce more cone crops [47], smaller height growth may cause economic loss by delaying cone production and the harvest of nuts.To secure preferred height growth, forest managers are recommended to consider wider strip-cuts for two-storied management of larch and Korean pine plantations.Although seedlings in the wider strip-cuts (6.0-7.5 m) had smaller predicted growths than in OPEN (Figure 4), the differences were not statistically significant (Table 2).By creating 6 m or wider strip-cuts, underplanted seedlings were predicted to reach breast height within 1-2 years of seedlings planted at open sites (Figure 5).Strip-cuts that are too wide, however, may negatively affect the growth and photosynthetic capacities of Korean pine as it increases in size [16].The planting of Korean pine seedlings in wide openings may entail risks including understory competition [9], physiological drought [48], and early stem divergence [17].Even though our model predicted greater height growth for Korean pine seedlings in wider strip-cuts, forestry practitioners should apply this research carefully.
This study was intended to quantitatively illustrate the height growth of underplanted Korean pine seedlings in strip-cut larch plantations in Northeast China.Our results, however, are limited to predicting height growth at the early growing stages; longitudinal studies within a site are needed for a more complete understanding of growth dynamics under partial larch canopies.Data of understory vegetation would be useful to examine the competitive effect on underplanted seedlings.Furthermore, site-specific conditions may have affected the results of this study.Further research should examine growth patterns under various other site conditions.Future studies should provide for a more fully developed understanding of Korean pine growth for the sustainable management of larch plantations in Northeast China.

Figure 1 : 1 )Figure 2 :
Figure 1: The observed mean height of Korean pine seedlings for 4 years after planting.

1 )Figure 3 :
Figure 3: Relationships between observed and predicted annual height growth of Korean pine seedlings ( = 339).Leave-oneout cross-validation was used for the prediction.The dotted line represents the point of identical observed/predicted height growth.

Figure 5 :
Figure5: Predicted height of Korean pine seedlings for 7 growing years after plantation.The initial seedling height when planted was set at 20 cm for all plots, and the area of overstory larch trees (BA) was assumed to be constant over those 7 years.

Table 1 :
Stand parameters and canopy openness for the plots with a larch overstory.

Table 2 :
Parameter estimates from the analysis of the annual height growth of Korean pine seedlings using a generalized linear mixed model.