Site index for Chinese f ir plantations varies with climatic and soil factors in southern China

Xiaoyan Li 1 · Aiguo Duan 1,2 · Jianguo Zhang 1,2

Abstract Chinese f ir [ Cunninghamia lanceolata (Lamb.)Hook.] has a large native distribution range in southern China. Here, we tested diff erences in productivity of Chinese f ir plantations in diff erent climatic regions and screened the main environmental factors aff ecting site productivity in each region. Relationships of a Chinese f ir site index with climatic factors and the soil physiochemical properties of f ive soil layers were examined in a long-term positioning observation trial comprising a total of 45 permanent plots in Fujian (eastern region in the middle subtropics), Guangxi(south subtropics) and Sichuan (central region in the middle subtropics) in southern China. Linear mixed eff ects models were developed to predict the site index for Chinese f ir, which was found to vary signif icantly among different climatic regions. Available P, total N, bulk density and total K were dominant predictors of site index in three climatic regions. The regional linear mixed models built using these predictors in the three climatic regions f it well( R 2 = 0.86-0.97). For the whole study area, the available P in the 0-20-cm soil layer and total N in the 80-100-cm soil layer were the most indicative soil factors. MAP was the most important climatic variable inf luencing the site index.The model evaluation results showed that the f itting performance and prediction accuracy of the global site index model using the climatic region as the dummy variable and random parameters and the most important soil factors of the three climatic regions as predictors was higher than that of global site index model using the climatic variable and the most indicative soil variables of the whole study area.Our results will help with further evaluation of site quality of Chinese f ir plantations and the selection of its appropriate sites in southern China as the climatic changes.

Keywords Site productivity · Site index · Climate · Soil ·Chinese f ir

Introduction

Potential forest productivity in diff erent site conditions is one of the most important criteria for decision-making in forest management (Kayahara et al. 1998). Setting forest productivity potential as a reference criterion creates an opportunity for forest managers to select the most suitable tree species and allows foresters to accurately predict stand yield (Bergès et al. 2005). Forest site productivity may be def ined as the potential of a site to produce timber or forest biomass (Sharma et al. 2012), and site index, def ined as the average height of dominant trees at a given age, is widely used as a measure of forest site productivity (Seynave et al.2005; Shen et al. 2018).

Stand growth is dependent on numerous environmental factors such as climate, topography, soil and vegetation (Grant et al. 2010; Beaulieu et al. 2011). As the global climate changes, however, the relationship between forest growth and environmental variables becomes even more complex (Bravo-Oviedo et al. 2010). Because an accurate estimate of the relationship between forest growth or productivity and these environmental factors is essential for evaluating forest site quality as the climate changes, the site index has been indirectly predicted using ecological environmental factors and applied to evaluate site quality (Curt et al. 2001; ?zel et al. 2020). The most commonly used method for evaluating site productivity, based on various environmental factors, is to predict site index as a function of climatic, topography and soil factors (Wang et al. 2004;Gülsoy and ?inar 2019). Numerous studies have explored the relationship between site index and various site quality variables to select the best site factors for explaining the variation in site index, but have had varying degrees of success (Chen et al. 1998; Mitsuda et al. 2001; Fontes et al.2003; Wang 2011). Based on the site-growth relationship,many studies have focused on predicting the site index using climatic and soil variables. The site index has been shown to be related to soil type, lithology, depth, texture and pH and level of soil nutrients (Grant et al. 2010; Farrelly et al.2011a, b; Paulo et al. 2015; Subedi and Fox 2016). Therefore, soil variables have been considered as an important predictor of stand growth and productivity. Climatic factors are also important site factors inf luencing the site index(Albert and Schmidt 2010; Menéndez-Miguélez et al. 2015).Generally, hydrothermal conditions are the main climatic factors aff ecting forest growth (Kishchenko 2004; Tyukavina et al. 2019). At the same time, changes in temperature and precipitation patterns and increases in atmospheric CO2concentration could lead to major changes in forest structure and productivity (Kirilenko and Sedjo 2007; Wamelink et al.2009). Several studies predicted the site index as a function of various climatic factors (Monserud et al. 2006; Albert and Schmidt 2010; Menéndez-Miguélez et al. 2015). Because of diff erences in regions and tree species investigated in diff erent studies, the determined variables with important predictive ability for explaining the site index varied among these studies (McKenney and Pedlar 2003; Sabatia and Burkhart 2014). Because forest growth is aff ected by various site factors, it is absolutely critical to select suitable site factors for describing site characteristics in diff erent regions and revealing the relationship between site and forest growth.

Chinese f ir [Cunninghamia lanceolata(Lamb.) Hook.] is native to southern China and the most important timber tree species in this region, where it is widely distributed in areas with a subtropical climate. Because of its vast production area, its growth can vary greatly among the southern, northern and central production areas within the subtropical belt of southern China (Tong and Liu 2019). Given the increasing changes in climatic patterns worldwide, it becomes more urgent and critical to fully understand the relationships among forest growth, yield prediction and environmental variables. Developing a site index model for Chinese f ir based on various site factors is also critical and will help to quantitatively evaluate site quality of non-forest land. In previous studies, site quality for Chinese f ir was evaluated by classifying site types and compiling a site index table(Huang et al. 1989; Li 2017). However, the essential relationship between site productivity and site quality variables of Chinese f ir in diff erent climatic regions is not well understood, and the inf luence of climatic factors and soil physiochemical properties in diff erent soil layers, in particular, on site productivity has not been fully investigated. Therefore,in this study, we explored the importance of various climatic and soil factors in predicting the growth of Chinese f ir in southern China.

The main objectives of this study were to (1) explore the relationship of site index with climatic factors and the physical and chemical properties of the top f ive soil layers,(2) analyze the variation in site index in diff erent climatic regions, (3) select the dominant site factors aff ecting site productivity in each climatic region and examine the ability of soil factors in diff erent soil layers to predict the Chinese f ir site index in each region, and (4) develop linear mixed eff ects models to predict site index using these dominant site factors as predictors. The results of this study will help us better understand the response of Chinese f ir site index to various site factors, which will provide guidance or a reference for evaluating site quality for Chinese f ir plantations in southern China.

Materials and methods

Study area and site index data

Study sites in three provinces in southern China (Fig. 1)included Fujian (eastern region in mid-subtropics), Sichuan(central region in mid-subtropics) and Guangxi (southern subtropical climatic zone). The soil type is mainly red soil developed on granite in Fujian and Guangxi and red soil developed on shale in Sichuan.

In each of these three regions, long-term positioning observation test plantations of Chinese f ir established in 1982, with even-aged trees, were selected for this study. In each study area, 15 plots were installed in a random block arrangement with f ive initial planting densities: A (1667 trees ha -1 , 2 m × 3 m), B (3333 trees ha -1 , 2 m × 1.5 m),C (5000 trees ha -1 , 2 m × 1 m), D (6667 trees ha-1,1 m × 1.5 m), E (10,000 trees ha -1 , 1 m × 1 m). Each treatment level was replicated three times for a total of 15 plots.

Fig. 1 Map of distribution of Chinese f ir plantations and location of study sites in China

Each plot was 20 m × 30 m and surrounded by a 2-row buff er zone, comprising similarly treated trees. Three experiments were established using bare-root seedlings, and all trees were tagged. After aff orestation, each plot was surveyed in the winter at 1-3-year intervals. Height data for 18-20 years were obtained from each plot. The dominant height was computed as the average height of six tallest trees in each plot. The site index was calculated for each plot at the reference age of 20 years using the dominant height data (Fujian and Guangxi) or using the Richards model with three parameters (Sichuan).

Climatic data

ClimateAP is an online platform to generate annual, seasonal and monthly climatic data for historical and future periods in the Asia Pacif ic region ( http:// clima teap. net/; Wang et al.2012, 2017a, b). To explore the eff ects of climate factors on the site index of Chinese f ir plantations, climatic data of each site was obtained using ClimateAP through spatially interpolated estimations based on site longitude, latitude, and elevation. In this study, eight climatic variables including mean annual temperature (MAT, °C), mean annual precipitation (MAP; mm), degree-days above 5 °C (DD5), degreedays below 0 °C (DD_0), July maximum mean temperature (Tmax07; °C), summer mean maximum temperature(Tmax_JJA; °C), spring precipitation (PPT_MAM; mm),annual heat-moisture index (AHM) were chosen as candidate variables for model f itting of the site index. What’s more, AHM integrates MAT and MAP data into a single parameter, as shown below:

whereAHMis the annual heat-moisture index.MATis the mean annual temperature.MAPis the mean annual precipitation. LowerAHMvalues indicate relatively wetter conditions. The climatic data for each study site are shown in Table 1.

Soil data

Table 1 Summary of elevation, latitude, longitude and climatic variables at the three study areas

Soil samples were collected by digging a soil prof ile in mature Chinese fir plantations in Fujian, Guangxi and Sichuan. Three soil prof iles were selected and diagonally distributed in each plot. A total of 135 soil prof iles were manually dug at the three study sites. Each soil prof ile was 1 m deep and divided into f ive soil depths: 0-20, ＞ 20-40, ＞40 - 60, ＞ 60-80 and ＞ 80-100 cm. Bulk density was determined by inserting three cutting rings (5 cm height, 5 cm inner diameter and known volume) at each depth of the soil prof ile. At the same time, soil water content was determined from soil samples collected in three aluminum boxes placed at each depth. Approximately 1 kg of soil was sampled at each depth, stored in bags and transported to the laboratory.The soil samples were then air-dried, ground, sieved, then analyzed for pH, organic matter content (g kg -1 ), total N(g kg -1 ), alkali-hydrolyzable N (mg kg -1 ), total P (g kg -1 ),available P (mg kg -1 ), total K (g kg -1 ), available K (mg kg -1 ), bulk density (g cm -3 ), water content (%), C/N ratio,C/P ratio and N/P ratio. The soil pH was determined using the potentiometer method using a suspension of 1-part soil to 2.5 parts 1 M KCl. Soil organic matter content was measured using the K2Cr2O7-H2SO4oxidation method. Total N was determined using the Kjeldahl method and alkalihydrolyzable N using alkaline hydrolysis method. Total P was measured using the NaOH alkali solution-molybdenum antimony colorimetric method and available P using the NaHCO3alkali solution-molybdenum antimony colorimetric method. Total K and available K were determined using f lame photometry (Bao 2000; Venanzi et al. 2016). The soil data for the three provinces are summarized in Table 2 .

Data analyses

Analysis of variance (ANOVA) and multiple comparisons test were performed to compare the diff erences in site index among diff erent regions. The relationship between site index and soil factors at the diff erent soil depths in the three provinces and that between site index and climatic variables were examined by Pearson correlation analysis. To reveal the importance of site factors across regions with diff erent climates, while taking into account the interdependency among independent variables, stepwise regression analysis was used to limit the number of explanatory variables for the three climatic regions, and the most important site quality variables aff ecting the site index of Chinese f ir in each climatic region were determined.

Linear mixed eff ects model (LMM)

The linear mixed eff ects model was used to predict the site index as a function of the most important soil factors andclimatic factors. Three regional models were established based on the determined soil factors for each climatic region,a climatic model also developed using the most dominant climatic variable. The following basic model was used for modeling site index related to regional models and climatic model:

Table 2 Soil characteristics for the three study areas in Chinese f ir plantations in Fujian, Guangxi and Sichuan in China

whereSIis the site index,α0is the intercept,αis a vector of coeffi cients, X is a vector of independent variables, including various soil and climatic variables, andεis the error term,ε～N(0,σ2 ).

Considering the diff erences among diff erent climatic regions, we treated climatic region as a dummy variable. A global site index model for the whole study area was established, which used the dummy variable and the most important soil factors of the three climatic regions as predictors.The basic site index model could be given by:

whereSIis the site index,β0is the intercept,β1,β2,β3are vectors of coeffi cients for Fujian, Guangxi and Sichuan,respectively;X1,X2,X3are vectors of independent variables for Fujian, Guangxi and Sichuan, respectively;S1,S2,S3represent dummy variables for Fujian, Guangxi and Sichuan,respectively;S1 = 1 indicates the Fujian and 0 indicates the other climatic regions;S2 = 1 indicates the Guangxi region and 0 indicates the other climatic regions;S3= 1 indicates the Sichuan and 0 indicates the other climatic regions.

In addition, based on the important soil factors selected in the three climatic regions, the relationship between these soil factors and site index was explored in the whole study area.A polynomial relation between the soil factors and site index was obtained, from which the soil factors most related to site index in the whole study area were determined. Therefore,another global site index model was established by using the most important soil and climatic variables in the whole study area.We selected 30 plots from the 45 plots as modeling data,and the remaining 15 plots were used as vertif ication data.Since data collected from three regions and f ive planting densities in each region, the random eff ects of region and planting density were added to the intercept of the site index model. In addition, we used the independent equal variance structure for describing the variance-covariance structure of random eff ects. Parameters in the LMM models were estimated through restricted maximum likelihood approach implemented in ForStat2.2 software (Tang et al. 2009).

Model evaluation

The model evaluation and testing were based on the mean absolute error (MAE), mean relative error (MRE), root mean square error (RMSE) and coeffi cient of determination (R2 ).

wherey iis the observed value,^yiis the predicted value, -y is the mean value of the observed value, andnis the number of the sample plots.

The independent sample data not used in the modeling were used to test the model, and the prediction performance of the global site index models was evaluated. Equation 8(Jiang and Li 2014) was used to calculate the random parameter value in the LMM models:

Results

Variation in site index across diff erent climatic regions

The eff ect of climatic region on the site index of Chinese f ir was signif icant. The site index in Sichuan province was signif icantly lower than that in Fujian and Guangxi provinces (Fig. 2).

Fig. 2 Variation in Chinese f ir site index among diff erent climatic regions. Data represent mean ± standard error (SE) of the mean(SEM). Diff erent lowercase letters indicate signif icant diff erences( P ＜ 0.05)

Correlation of Chinese f ir site index with soil and climatic variables

Soil variables

The correlations between soil factors at the f ive soil depths and site index in diff erent climatic regions are shown in Fig. 3. The correlation between site index and a given soil factor varied with the climatic region and soil depth, indicating that soil factors in diff erent regions and at diff erent depths had diff erent eff ects on the site index. Fujian site index of Chinese fir was strongly negatively correlated with the total P at all soil depths and was strongly positively correlated with available P in the top three soil layers (0-20, ＞ 20-40, ＞ 40-60 cm) (P＜ 0.05). The correlation between Fujian site index and available P decreased with the increase in soil depth; consequently, the Fujian site index was negatively correlated with available P in the last two soil layers (＞ 60-80 cm, ＞ 80-100 cm). The strength of the correlation between Fujian site index and available K increased with the increase in soil depth,reaching statistical signif icance in the last three soil layers(＞ 40-60, ＞ 60-80, ＞ 80-100 cm) (P＜ 0.05). A signif icant positive correlation was detected between Fujian site index and soil bulk density, although the strength of the correlation decreased with the increase in soil depth. The opposite trend was observed for N/P ratio (i.e., the deeper the soil layer, the stronger the correlation), and a signif icant correlation was observed in the lowest soil layer (＞ 80-100 cm). Soil pH,total K and C/P ratio at all soil depths were positively correlated with the Fujian site index, although the correlations were not statistically signif icant. Additionally, the Fujian site index was negatively correlated with the soil water content at all soil depths; however, Fujian site index was not closely correlated with the soil organic matter content, total N and alkali-hydrolyzable N, and there was no obvious soil layer eff ect.

The Guangxi site index, however, was positively correlated with total P and water content of all soil layers and negatively correlated with available P, C/P ratio and N/P ratio of all soil layers, and its correlation with available P was statistically signif icant (P＜ 0.05). Moreover, the Guangxi site index was signif icantly positively correlated with total K at all soil depths, consistent with the Fujian site index.In addition, the organic matter content, C/N ratio and total N content in all soil layers (except the total N content of the ＞ 80-100-cm soil layer) were negatively correlated with the Guangxi site index. However, soil pH, alkali-hydrolyzable N and bulk density of diff erent soil layers showed weak or nonsignif icant correlations with the Guangxi site index,and these correlations were not verif ied with the increase in soil depth.

Fig. 3 Plots showing the correlation coeffi cient between site index and soil variables at f ive depths in three diff erent climatic regions. Signif icant diff erences: * P ＜ 0.05 and ** P ＜ 0.01 levels

Compared with Fujian and Guangxi site indices, the Sichuan site index showed a negative correlation with total K at all soil depths. A signif icant positive response of the Sichuan site index to increasing available K of all soil layers, according to this result, the three climatic regions were consistent. Furthermore, available P showed a signif icant positive relationship with the Sichuan site index at all soil depths, except at 0-20 cm. The correlations between Sichuan site index and other soil factors were weak or not signif icant,and the correlations were not verif ied with the increase in soil depth.

Fig. 4 Correlation coeffi cient between site index and climatic variables. Signif icance: ** P ＜ 0.01 level

The results showed that soil variables closely related to the site index were not consistent due to the diff erences in climatic conditions among diff erent geographical locations and the uneven distribution of nutrients in diff erent soil prof iles. Additionally, the correlation between a given soil variable and site index varied across the three climatic regions.

Climatic variables

The correlation between the site index of all plots and climatic and elevation factors is shown in Fig. 4. The site index showed a signif icant correlation with elevation (Elev), MAP,AHM, Tmax_JJA, PPT_MAM and Tmax07. Additionally,the site index was signif icantly positively correlated with MAP, Tmax_JJA, PPT_MAM and Tmax07 and signif icantly negatively correlated with Elev and AHM. However, there was no signif icant correlation between site index and various thermal variables, such as MAT, DD_0 and DD5.

Linear mixed eff ects model

Regional model based on soil factors

Based on the stepwise regression analysis, the dominant soil variables inf luencing site index varied among the diff erent climatic regions and soil depths. All these selected variables were signif icant (P＜ 0.05) in the site index models of these three climatic regions. The f inal base models for each climatic region are:

Table 3 Regional models based on soil factors and climatic model using MAP

In Fujian, three variables were selected in the f inal model,including available P in the 0-20-cm soil layer, total N in the ＞ 80-100-cm soil layer and bulk density in the ＞ 60-80-cm soil layer, as these were the most important soil factors that signif icantly aff ected the site index; the site index values increased with the increase in available P, total N and bulk density in the specif ied soil layers. In Guangxi,available P in ＞ 40-60-cm soil layer, total N in 0-20-cm soil layer and total K in ＞ 40-60-cm soil layer were the most important factors controlling the growth of Chinese f ir. The Guangxi site index decreased with the increase in available P in (＞ 40-60 cm) and total N (0-20 cm), and increased linearly with the increase in total K (＞ 40-60 cm). In Sichuan,available P in ＞ 40-60-cm and ＞ 20-40-cm soil layers had signif icant eff ects on site index.

The f itting results of the linear mixed models for the three climatic regions showed that MAE was below 0.7, MRE was less than 0.05, RMSE was less than 0.7, andR2 ranged from 0.86 to 0.97. These values from the planting density-level random eff ects model indicated that the models of the three climatic regions f it the data well and better simulated the variation in site index in the local area.

Eff ect of climatic factors on site index

Stepwise regression analysis showed that MAP was the main climatic factor that signif icantly inf luenced the site index.The f inal base model is:

whereSICrepresents the site index estimated by the climatic variable;μ04is the random-eff ect parameter of the region,andμ04～N(0,σC2).

Although the correlation analysis showed that six climatic factors had signif icant correlations with the site index, MAP was the only climatic variable output atP＜ 0.05 when considering the multicollinearity of various climatic variables and the signif icance test of regression coeffi cient.

Relationship of the site index with dominant soil factors in the whole study area

The site index changed gradually with the increase in available P content within the 0-2 mg kg -1 range but increased rapidly within the 2-4 mg kg -1 range (Fig. 5), probably because the available P content in Guangxi and Sichuan was low (mostly within the 0-2 mg kg -1 range), and the range of the site index was also small; therefore, the change in site index within this range was relatively smooth. However,the content of available P in Fujian generally ranged from 2-4 mg kg -1 , and the range of site index in Fujian was large.Therefore, the site index increased rapidly with the increase in available P content within the range of 2-4 mg kg -1 range.The total N content of the 0-20-cm soil layer was higher than that of the soil layer of ＞ 80-100-cm soil layer. The content of total N in the 0-20-cm soil layer mainly ranged from 0.7 to 1.5 g kg-1; however, the site index increased with the increase in total N content within the 0.7-1.3 g kg-1range, and decreased slightly within the 1.3-1.5 g kg-1range (Fig. 5), which may be due to the fact that the total N content of the 0-20-cm soil layer in Sichuan was relatively low (0.7-1.3 g kg-1) and was positively correlated with the site index. However, the total N content of the 0-20-cm soil layer in Fujian and Guangxi was relatively high and was negatively correlated with the site index within the 1.3-1.5 g kg-1range. Additionally, the site index changed gradually with the increase in the total N content of the soil layer of ＞ 80-100 cm within the 0-0.4 g kg-1range but increased rapidly within the 0.5-1.0 g kg-1range (Fig. 5). The site index increased slightly with the increase in soil bulk density in the ＞ 60-80-cm soil layer within the 1.2-1.4 g cm-3range but decreased signif icantly with further increase in soil bulk density from 1.4-1.7 g cm-3. However, the data range of bulk density in the 0-20-cm soil layer was relatively small,and changes in bulk density in the 0-20-cm soil layer had less eff ect on changes in the site index (Fig. 5). Total K in the ＞ 40-60-cm and 0-20-cm soil layers within the 0-3 g kg-1range showed no clear correlation with site index in the whole region. And site index increased with the increase in total K content within the 10-30 g kg-1range (Fig. 5).However, both had less eff ect on changes in the site index across the whole study area, mainly because of the diff erent ranges of total K content and diff erent correlation between site index and total K in the three climatic regions.

Fig. 5 Correlation of site index with available P (AP), total N (TN),bulk density (BD) and total K (TK) in the whole study area. Numbers 1-5 in variable names represent the f ive soil layers: 1, 0-20 cm;2, ＞ 20-40 cm; 3, ＞ 40-60 cm; 4, ＞ 60-80 cm; 5, ＞ 80-100 cm. y is the dependent variable (site index), x is the independent variable (AP,TN, BD, TK)

From Sichuan, to Guangxi, to Fujian, the available P and total N content increased gradually, consistent with the actual stand productivity. For the whole study area,the strength of the correlation between site index and total N increased with the increase in soil depth, while the correlation between site index and available P decreased with soil depth. Meanwhile, the f itting results in Fig. 5 also showed that the available P in the 0-20-cm soil layer and total N in the ＞ 80-100-cm soil layer were the most indicative soil factor.

Development and evaluation of global site index models

The global site index model 1 for the whole study area was established using climatic region as a dummy variable and the most important soil variables of the three study sites as predictors. In addition, the global site index model 2 using climatic variable and the most indicative soil variables of the whole study area also developed. The f inal base model of the global site index model 1 and the global site index model 2 were listed as follows:

Table 4 Global site index model 1 based on soil variables

Considering the random eff ects of region on site index,we added the random parameters of region to the intercept of global site index model 1 and global site index model 2.Planting density was used as the group variable for the random eff ect in global site index model 2. The LMM models are given in Tables 4 and 5 as:

Table 5 Global site index model 2 based on soil and climatic variables

whereSI1andSI2represent the site index estimated by global site index model 1 and global site index model 2, respectively;μ05andμ06are the random-eff ect parameters of the region; andμ05～N(0,σmodel12),μ06～N(0,σmodel22). According to the group variable planting density, the covariance of random eff ect generated by region was divided into f ive matrices: (1) when the planting density is A, the parameter of the covariance matrix of random eff ect caused by region isσmodel212; (2) when the planting density is B, the parameter of the covariance matrix of random eff ect caused by region isσmodel 22 2 ; (3) when the planting density is C, the parameter of the covariance matrix of random eff ect caused by region isσmodel232; (4) when the planting density is D, the parameter of the covariance matrix of random eff ect caused by region isσmodel242; and (5) when the planting density is

The f itting performance of the global site index model 1 showed that MAE, MRE, and RMSE were all smaller than that of global site index model 2 andR2 was greater than that of global site index model 2, which indicated that the f ittingaccuracy of global site index model 1 was higher than that of global site index model 2 (Tables 4 and 5). The predictive ability of the model was also tested using 15 groups of independent validation data, and the results showed that theR2 for global site index model 1 was higher than that for global site index model 2, and the MAE, MRE, and RMSE were all smaller than for global site index model 2, indicating that the prediction accuracy of global site index model 1 was also higher than that of global site index model 2 (Table 6).

Table 6 Predictive test of global site index models using independent validation data

Discussion

Eff ects of soil factors at diff erent depths on site index

Eff ects of total P, total K, available K, bulk density and water content on site index

The relationship between site index and soil factors has been widely discussed. Because of diff erences in ecosystem types and tree species in the study area, the response of site index to environmental factors varied with the climatic region.Farrelly et al. ( 2011a, b) explored the relationship between site index of Sitka spruce in Ireland and various soil physical and chemical properties, and developed a site index model,and showed a signif icant negative correlation between site index vs. soil water content, organic carbon, available K and P. Here, we found that the soil variables that signif icantly aff ected the site index varied among climatic regions, and those that aff ected the site index within a certain climatic region showed diff erent trends at diff erent depths.

A signif icant negative correlation was detected between Fujian site index and total P in all soil layers. However, in Guangxi and Sichuan provinces, the correlation between site index and total P was small or not signif icant. With the increase in soil depth, the positive correlation between Fujian site index and available K became stronger, whereas the positive correlation between site index and soil bulk density weakened. By contrast, site index showed no signif icant correlation with available K and bulk density in both Guangxi and Sichuan, and the soil layer eff ect was not obvious. In addition, a signif icant positive correlation was detected between Guangxi site index and total K in all soil layers. However, the positive correlation between the Fujian site index and total K decreased with increasing soil depth,indicating that the response of site index to total K in deeper soil layers was poor in Fujian Province. Moreover, a small negative correlation was evident between the site index and total K in all soil layers in Sichuan.

The site index in all three climatic regions showed a small correlation with pH and alkali-hydrolyzable N in each soil layer, indicating that these variables did not have a major impact on the site index. Furthermore, we found that the soil

water contents were relatively high in Fujian (19%-34%) and Sichuan (16%-40%), and were negatively correlated with the site index. However, the soil water content in Guangxi was relatively low (11%-18%), and the water content of each soil layer had a strong positive correlation with the Guangxi site index. These results suggest that excess soil water content could hinder the improvement of forest productivity. The water-holding capacity and water availability of the soil play a key role in the productivity of forest plantations (Besson et al. 2014). A reduction in soil water accessibility decreases forest productivity (Paulo et al. 2015). Because neither excess nor def icient soil water content are conducive to the growth of the stand, it is necessary to regulate the relationship between soil water content and soil physical properties reasonably (Farrelly et al. 2011a, b).

Eff ects of available P on site index

P and K are highly valued elements for plant growth (Zeng et al. 2019; Bai et al. 2020). In this study, available P and total K were the limiting factors aff ecting the growth of Chinese f ir. The results of site index LMM models in three climatic regions showed that available P was an important predictor of site index. However, the eff ect of available P in diff erent soil layers on site index varied among the three climatic regions. For example, the available P in 0-20 and 40-60-cm soil layers aff ected the site index in Fujian and Guangxi, respectively, while the site index in Sichuan was mainly aff ected by the available P in 20-40 and 40-60-cm soil layers. Additionally, the correlation between site index and available P was diff erent in diff erent climatic regions.For example, a signif icant positive correlation was observed between site index and available P in Fujian and Sichuan,but a signif icant negative correlation was detected between site index and available P in Guangxi. Previously, several studies have shown that P is the main limiting factor for tree growth and that increasing the application of P fertilizer can eff ectively promote stand growth and improve site productivity (Ma et al. 2015; Shang et al. 2020). However,in this study, the site index of Chinese f ir did not show a linear positive correlation with the increase in available P in Guangxi; instead, the Guangxi site index decreased with the increase in available P. A similar result was also obtained by Kayahara et al. ( 1995) and Chen et al. ( 1998). Kayahara et al. ( 1995) explored the relationship between site index and various soil nutrient factors and pointed out that there was no linear positive correlation between site index and soil nutrient availability. In other words, the highly productive sites were not always nutrient-rich; as the author explained, once a tree species reaches the early nutritional suffi ciency point,it is not nutrient-limited as measured by the tests, but it may be limited by one or more of the several other soil physical,chemical and biological factors. Therefore, P accumulates because of excessive consumption in poor site conditions.However, site productivity decreases due to limitations in other nutrient elements, which increases the P content in regions with a low site index (Kayahara et al. 1995). While some factors cause P def iciency in soil, plants can alleviate P def iciency by stimulating a series of other pathways of P acquisition, such as increasing P mineralization by changing microbial communities (Deng 2016). In addition, the subtropical region of southern China is severely aff ected by N deposition. The intensif ication of N deposition can aff ect the soil nutrient cycle and increase P def iciency in the subtropical region (Xie et al. 2020). Moreover, the increase in N deposition can also reduce the f ine root biomass of plants, thus reducing the uptake of available P by plants and increasing the accumulation of available P in the soil (Mao et al. 2018; Xie et al. 2020). However, this f inding needs further verif ication.

In Guangxi, which is located in the southern subtropical region, Chinese f ir plantation grows on marginal land, with low available P content. Forest growth in Guangxi may be limited by a simultaneous def iciency in several other soil nutrient. However, because of the combination of N, K and various biophysical factors, the stand growth is restricted,which results in low site productivity and showing a negative correlation between site index and available P. Some studies showed that P addition might increase biological N f ixation in subtropical forest ecosystems (Zheng et al. 2015;Wang et al. 2017a, b). Therefore, more P fertilizer should be applied to Chinese f ir plantations in Guangxi, and measures should be taken to promote the absorption of available P by Chinese f ir.

Eff ects of organic matter, total N and C/N ratio on site index

Organic matter and total N are the basis of soil fertility and also have a major impact on stand growth (Li et al. 2014).However, in this study, we found that the organic matter content in each soil layer in the three climatic regions had little eff ect on site index, except the organic matter content of the 0-20 and ＞ 60-80-cm soil layers in Guangxi, which showed a signif icant correlation with site index. Farrelly et al. ( 2011a, b) pointed out that soil total N had no signif icant eff ect on the site index of Sitka spruce. In the current study, we found no signif icant correlation between site index of Chinese f ir and total N content in any soil layer in Fujian or Guangxi, although total N signif icantly aff ected site index as a result of other signif icant soil physical and chemical properties. Thus, with other soil nutrients, total N plays a signif icant role in the growth of Chinese f ir.

Studies have shown that N deposition and forest litter input have the greatest impact on surface soil nutrients, especially on the increase of soil surface N content, which may lead to the instability of the N content on the topsoil (Fan et al. 2008; Guo et al. 2014). However, N elements are transported downward into deeper soil layers by leaching and other means. In deeper soil layers, N levels are less aff ected by climatic and biological factors than in the topsoil and thus more stable (Yuan et al. 2007). In this study, the strength of the correlation between the Fujian site index and total N increased with the increase in soil depth, which indicated that the Fujian site index had a better response to the total N content in the deeper soil layer. On the contrary, the strength of the correlation between Guangxi site index and total N decreased with the increase in soil depth, indicating that the Guangxi site index had a better response to the total N content in topsoil. Due to the diff erent climatic and soil characteristics, the N deposition and the degree of forest litter decomposition diff er in diff erent regions. These factors will aff ect the N cycling in soil, resulting in diff erent responses of soil N components in diff erent regions and diff erent soil layers to the changes in soil N input (Ma et al. 2013; Lin et al. 2016). However, the relationship between forest growth and N components in diff erent soil layers needs to be further studied.

Many studies show that soil C/N ratio is reportedly an important index of N-use effi ciency and a signif icant predictor of site index. For example, Seynave et al. ( 2005) showed that the growth of Norway spruce was aff ected by soil pH and N availability; low productivity was found in sites with high pH and high C/N ratio, and pH and C/N ratio were signif icant predictors of site index only when they present together. However, Bergès et al. ( 2005) suggested that C/N ratio had no eff ect on the site index and concluded that C/N ratio was not an accurate indicator of N supply. In the current study, we found that the C/N ratio of each soil layer in the three climatic regions had no obvious eff ect on site index,and the correlation was neither close nor signif icant.

The site index of Chinese f ir showed diff erent responses to soil physical and chemical properties in diff erent climatic regions, and the eff ect of a given soil variable on site index varied with the soil depth, probably because climatic factors such as elevation and slope position and aspect diff er among diff erent climatic regions, resulting in some variation in hydrothermal conditions in each plot. These diff erences aff ect the spatial distribution and transformation process of various soil factors, resulting in an uneven distribution of soil nutrients in diff erent soil prof iles, aff ecting the absorption of available soil nutrients by plant roots, and thus aff ecting tree growth (Zhang et al. 2015).

Eff ects of climatic variables on site index

Our results showed that the site index of Chinese f ir varied signif icantly among diff erent climatic regions. MAP was the most important climatic factor responsible for the variation in site index among diff erent regions. In the whole study area, MAP showed a signif icant positive linear correlation with site index, consistent with the f indings of Menéndez-Miguélez et al. ( 2015) and Gülsoy and ?inar ( 2019), who showed that precipitation was strongly correlated with tree height. However, Monserud et al. ( 2006) established a site index model with various climatic variables as predictors and showed that growing degree days ＞ 5 °C (GDD5), the Julian date when GDD5 reaches 100 (D100), and July mean temperature (MTWM) had the strongest predictive ability,while precipitation was not strongly correlated with site index. Nevertheless, our result showed that the site index was not closely correlated with various thermal conditions(DD_0, DD5, MAT), indicating that the response of Chinese f ir plantations to these heat-related factors was low in this study area, and it was diffi cult to improve the prediction ability of the site index model after including these temperature variables in the model.

Global site index models

The results showed that the LMM model had some advantages after considering the random eff ects of region or planting density on site index. And the global site index models for the whole study area showed a good f it.R2 of the global site index model for the whole study area using climatic region as a dummy variable and random parameter was larger than that of the single climatic region, indicating that the f itting performance of the LMM model will not decrease with the expansion of the spatial scale. In addition, the evaluation indices ofR2 , MAE, MRE, and RMSE, showed that the prediction accuracy of global site index model using the climatic region as dummy variable (Table 6, Eq. ( 19)) was higher than that of global site index model using the climatic variable and the most indicative soil variables of the whole study area (Table 6, Eq. 20). The main reason may be that Eq. 19 took into account the diff erences among diff erent climatic regions, and it is more effi cient in predicting site index among these three climatic regions. Unfortunately, Eq. 19 may only be applicable to these three climatic regions, which limits the applicability of this model in other regions. However, Eq. 20 is more likely to be universal and applicable.Our study showed that the global site index model is applicable across diverse regions and that the site index prediction quality will not decrease with the increase in spatial scale.This result is consistent with the f indings of Bergès et al.( 2005), who showed that increasing the spatial scale did not decrease the prediction quality of site index. However, a few studies had verif ied a reduction in the prediction quality of site index with the expansion of study area (Chen et al. 2002;Aertsen et al. 2012).

Conclusions

In this study of the relationship of Chinese f ir site index with climatic and soil factors at three climatic regions in southern China, the site index was modeled in relation to the dominant soil and climatic factors in each local area or in the whole study area. Results showed that the dominant soil factors with the strongest predictive ability to the variation of site index varied with the climatic region and soil depth. Available P, total N, bulk density and total K were good predictors of site index in three climatic regions. Total N signif icantly aff ected site index in conjunction with other signif icant soil factors. Linear mixed eff ects SI models built using these dominant soil factors in the three climatic regions f it well,andR2 was in the range of 0.86 to 0.97. In addition, MAP was the climatic factor responsible for the variation in site index among diff erent regions. The global site index model for the whole study area using climatic region as a dummy variable and random parameters and the most important soil factors of the three climatic regions as predictors improved the f it and prediction accuracy of the site index model. This model eff ectively resolved the impact of diff erent site types on the prediction of site index of Chinese f ir plantations, thus improving its applicability to diff erent regions. These results will aid in evaluating site quality of Chinese f ir plantations and selecting appropriate sites for plantations in southern China as the climatic changes.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing,adaptation, distribution and reproduction in any medium or format,as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons.org/ licen ses/ by/4. 0/ .