Structure and Composition of Natural Gmelin Larch (Larix gmelinii var. gmelinii) Forests in Response to Spatial Climatic Changes

Background Many theoretical researches predicted that the larch species would decrease drastically in China under future climatic changes. However, responses of the structural and compositional changes of Gmelin larch (Larix gmelinii var. gmelinii) forests to climatic changes have rarely been reported. Methodology/Principal Findings Field survey was conducted to examine the structures and compositions of natural Gmelin larch forests along a climatic gradient. Stepwise linear regression analyses incorporating linear and quadratic components of climatic and non-climatic factors were performed on the structural and compositional attributes of those natural Gmelin larch forests. Isothermality, Max Temperature of Warmest Month (TempWarmestMonth), Precipitation of Wettest Month (PrecipWettestMonth), Precipitation Seasonality (PrecipSeasonality) and Precipitation of Driest Quarter (PrecipDriestQuarter) were observed to be effective climatic factors in controlling structure and composition of Gmelin larch forests. Isothermality significantly affected total basal area of larch, while TempWarmestMonth, PrecipWettestMonth and PrecipSeasonality significantly affected total basal area of Mongolian pine, and PrecipDriestQuarter significantly affected mean DBH of larch, stand density of larch and total basal area of spruce and fir. Conclusions/Significance The summer and winter temperatures and precipitations are all predicted to increase in future in Northeast China. Our results showed the increase of total basal area of spruce and fir, the suppression of regeneration and the decrease of stand density of larch under increased winter precipitation, and the decrease of total basal area of larch under increased summer temperature in the region of current Gmelin larch forest. Therefore, we suggest that larch would decrease and spruce and fir would increase in the region of future Gmelin larch forest.


Introduction
Gmelin larch (Larix gmelinii var. gmelinii, synonym: Larix dahurica) is a tree species native of Russia, Mongolia and part of Northeast China [1]. In the last decade, many theoretical researches predicted that this larch species would decrease drastically in China under future climatic changes, such as the rise in mean annual temperature [2], rises in monthly temperatures [3,4,5] and increases in annual precipitations [2,3]. However, responses of the structural and compositional changes of Gmelin larch forest to climatic change have rarely been reported [6,7,8,9].
In order to detect and depict plant responses to large-scale climatic change, experimental studies can use spatial climatic variations instead of temporal trends, and then build linear or quadratic response functions to quantify the plants responses to climatic change. Such studies have been conducted on Lodgepole pine, grasses, tropical trees and bryophytes [10,11,12,13,14,15,16]. Additionally, it also is necessary to recognize the effects of some non-climatic factors, such as soil nitrogen nutrition and natural succession. Different soil nitrogen levels [17,18] and the natural succession of community with age [19,20,21] can cause structural and compositional changes of communities. Usually, the stand age of natural larch forest can be represented by the maximum diameter of larch (Max.DBH Larch ) in the stand. First, stand age is the age of the oldest tree in the stand [22,23]. Second, the positive relationships between tree age and tree diameter are confirmed in larches [24,25]. Although the rapidly growing tree and slowly growing tree differ in age-DBH regression slopes within the same stand, that do not make much inter-stand difference [25].
In this paper, we surveyed the structure and composition of natural Gmelin larch forests along a climatic gradient in Heilongjiang Province, Northeast China. We used multiple regression analyses to detect the effects of climatic and nonclimatic factors on structure and composition of natural Gmelin larch forests. Our primary objective is to discover the climatic and non-climatic factors which control the structure and composition of Gmelin larch forests, and the second objective is to depict the responses of Gmelin larch forests to changes of those climatic and non-climatic factors.

Ethics Statement
All necessary permits were obtained for the described field studies. This study was approved by State Key Laboratory of Vegetation and Environmental Change, Institute of Botany, the Chinese Academy of Sciences; Forestry Administration of Mohe County (Mohe County Forestry Bureau); Huzhong National Nature Reserve; Shengshan National Nature Reserve and Liangshui National Nature Reserve.

Study area
This study was conducted in July and August of 2011 along a climatic gradient in Heilongjiang Province of Northeast China, including four sites: Mohe, Huzhong, Shengshan and Liangshui ( Fig. 1). Each site consists of three plots ( Table 1). The forests in the plots are natural Gmelin larch forests without any evidence of recent disturbance. The region has monsoon climate: July is the warmest month and January is the coldest. Summer is the wettest quarter and winter is the driest [26,27]. The bioclimate data of each plot downloaded from WorldClim [27] are at 1-km resolution ( Table 2). Isothermality is derived as MeanDiurnal-Range/TempAnnualRange, and TempAnnualRange is derived as TempWarmestMonth -TempColdestMonth. The TempWar-mestQuarter, TempColdestQuarter, PrecipWarmestQuarter and PrecipColdestQuarter are exactly the same as TempWettestQuarter, TempDriestQuarter, PrecipWettestQuarter and PrecipDriest-Quarter in the sites ( Table 2, Table 3).

Field survey, soil sample collection and nitrogen measurement
In each plot, every living tree higher than 1.4 m was identified to species and measured for circumference at breast height (1.3 m).
Soil samples were collected to a depth of 20 cm using a metal cylinder (8 cm diameter and 10 cm length). In each plot, soil cores were excavated from five random points and then mixed into one sample. Fresh soil samples were passed through a 2-mm sieve to be removed of visible plant tissues, then air-dried and kept in ziplock bags.
The air-dried soil samples were taken to the State Key Laboratory of Vegetation and Environmental Change for nitrogen analyses. The 0 -20 cm soil nitrogen contents were measured by the Semimicro-Kjeldahl method [28].

Data analyses
The circumference of each tree was converted into tree diameter at breast height (DBH) and into tree basal area using the circle formulas. Seven attributes of Gmelin larch forest (Max.DBH Larch , mean DBH of larch, stand density of larch, total basal area of larch, total basal area of Mongolian pine, total basal area of spruce and fir and total basal area of birch) were extracted from each plot.
One-way ANOVA and regression and correlation analyses were performed using SPSS 13.0. Significant differences (p,0.05) between sites were detected using One-way ANOVA with post-Duncan's test. Standard errors within sites were detected by oneway ANOVA with descriptive options.
Stepwise linear regression analyses incorporating linear and quadratic components of seventeen independent factors (fifteen climatic and two non-climatic) and stepwise linear regression analyses incorporating only linear components of them were performed on six dependent variables. The fifteen climatic factors  Table 2). The six dependent variables were mean DBH of larch (MeanDBH Larch ), stand density of larch (StandDensity Larch ), total basal area of larch (BasalArea Larch ), total basal area of spruce and fir (BasalArea Spruce-Fir ), total basal area of Mongolian pine (BasalArea MPine ) and total basal area of birch (BasalArea Birch ) ( Table 2). Regression models with non-significant (p.0.05) term(s) would be rejected. Among candidate models with all terms significant (p,0.05), the model with the lowest Small Sample Unbiased Akaike Information Criterion (AICc) would be selected [29].
Two-tailed partial correlation coefficients between every included dependent variable and the independent variables excluded from its regression model were tested, controlling for the independent variable(s) included in its regression model. Pearson's two-tailed correlation coefficients among all variables were also tested.

Gmelin larch forest structure and composition
There were significant differences in Max.DBH Larch , Mean-DBH Larch , StandDensity Larch , BasalArea Larch , BasalArea Spruce-Fir and BasalArea MPine among the four sites (P,0.05). The Mean-DBH Larch roughly increased while the StandDensity Larch roughly decreased from the northernmost site Mohe to the southernmost site Liangshui. BasalArea Larch decreased from the second north-

Responses of structure and composition in Gmelin larch forests to climatic and non-climatic factors
Although the excluded independent variables showed no significant partial correlation with the dependent variable when controlling for the included independent variable(s) (

Discussion
In Daxing'anling Mountains of China, the importance value of Mongolia pine decreases while the importance value of larch increases from the 19th year to the 100th year after fire disturbance in the succession layer of Gmelin larch forest [30]. In our result, the Max.DBH Larch was observed to have significant  negative relationship with total basal area of Mongolian pine (given that the Max.DBH Larch in Mohe didn't exceed 63.25 cm), but no significant relationship with total basal area of larch ( Table  6, Table 7). As the Max.DBH Larch represents the stand age, we suggest the natural decline of Mongolian pine in Gmelin larch forest with increasing stand age. Spruce (Picea spp.) and fir (Abies spp.) are more shade-tolerant than larch [31,32]. In the absence of spruce and fir in northeastern China, larch can maintain its canopy dominance via gap regeneration [33]. But when larch coexists with those more shade-tolerant spruce and fir, larch will not maintain itself under the canopy, and will be ultimately replaced by those more tolerant and self-maintaining evergreens (spruce and fir) [34,35]. Exposure to direct sunlight can cause winter injuries to conifers such as spruce [36]. The exposed conifer seedlings show more serious winter injuries and increase mortality compared with snowcovered seedlings [37,38]. However, the beneficial effects of snow cover (for example on dry matter productions) are more pronounced on spruce than on larch [39], while the larch has a higher resistance to winter injuries than fir has [40]. In our results, PrecipDriestQuarter was observed to have positive effect on total basal area of spruce and fir, negative effect on stand density of larch and positive effect on mean DBH of larch. As Precip-DriestQuarter is winter precipitation [26] (Table 3), a higher PrecipDriestQuarter can form a thicker snow-cover, protecting the evergreen conifer seedlings (spruce and fir) from winter injury and mortality, resulting in more evergreen conifers (spruce and fir) in Gmelin larch forest. The negative relationship between Precip-DriestQuarter and stand density of larch and the positive relationship between PrecipDriestQuarter and mean DBH of larch indicated that larch regeneration is suppressed in presence of those evergreen conifers (spruce and fir), resulting in a lower density of larch and a lack of younger (smaller) larch stems.
The positive effects of summer temperatures on radial growths of larch were observed in polar and alpine treelines (ecotones between forest and tundra) [6,7], but the negative effects of July temperatures on radial growths of larch were observed in China [41,42] and Central Yakutia [9]. In our results, the positive effect of Isothermality on total basal area of larch was observed. Thus, a higher TempWarmestMonth (hotter July) can cause a lower Isothermality, and result in a decrease in total basal area of larch. Although a higher TempColdestMonth can cause a higher Isothermality, the correlation analyses suggested the negative effect of TempColdestMonth on total basal area of larch instead of positive (Table 6).
Some studies found the negative effects of summer temperatures and positive effects of summer precipitations on radial growths of Mongolian pine (Pinus sylvestris var. mongolica) in China [43,44,45]. However, a hotter and dryer summer do more harm to larch than to pine (Pinus sylvestris), causing a superiority of pine in competition with larch [46]. In our study, the positive effect of TempWar-mestMonth and negative effect of PrecipWettestMonth on total basal area of Mongolian pine were observed, suggesting that a hot and dry summer may suppress larch and favor the regeneration of Mongolian pine in Gmelin larch forest.
Most of the excluded independent variables showed some significant correlations with the dependent variables, and every excluded independent variable showed some significant correlation(s) with the included independent variable(s) ( Table 6). The significant correlations indicated the redundancy of independent variables. Further research will be needed to tell whether the excluded independent variables are also effective in controlling structure and composition of Gmelin larch forests. Table 5. Matrix of two-tailed partial correlation coefficients between every included dependent variable and the independent variables excluded from its model, controlling for the independent variable(s) included in its model.  The summer and winter temperatures and precipitations in Northeast China will all increase in the 21st century under the SRES A1B, A2, B1 and B2 scenarios, as predicted by the CCCMA-GCM 3.1, CNRM-CM3, MPI-CHAM 5, UKMO-HadCM 3 and PRECIS models [47,48]. Based on the climatic predictions, our results suggested the decrease of total basal area and stand density of larch and the increase of total basal area of spruce and fir in the Gmelin larch forest in future. Many studies also predicted the decrease of Gmelin larch [2,3] or decrease of larch [4,5] in northeastern China under increased temperatures and precipitations. However, some studies predicted the larch would be replaced by pine and broad-leaved trees [4,5], while our study predicted the larch would be replaced by spruce and fir. Further research will be needed to tell which tree species will replace Gmelin larch and whether the total basal area of Mongolia pine will increase or decrease in future.

Conclusions
The summer and winter temperatures and precipitations are all predicted to increase in future in Northeast China [47,48]. Our results showed the increase of total basal area of spruce and fir, the suppression of regeneration and the decrease of stand density of larch under increased winter precipitation, and the decrease of total basal area of larch under increased summer temperature in the region of current Gmelin larch forest. Therefore, we suggest that larch would decrease and spruce and fir would increase in the region of future Gmelin larch forest.