Distribution and altitudinal patterns of carbon and nitrogen storage in various forest ecosystems in the central Yunnan Plateau, China

The carbon (C) pool in forest ecosystems plays a long-term and sustained role in mitigating the impacts of global warming, and the sequestration of C is closely linked to the nitrogen (N) cycle. Accurate estimates C and N storage (SC, SN) of forest can improve our understanding of C and N cycles and help develop sustainable forest management policies in the content of climate change. In this study, the SC and SN of various forest ecosystems dominated respectively by Castanopsis carlesii and Lithocarpus mairei (EB), Pinus yunnanensis (PY), Pinus armandii (PA), Keteleeria evelyniana (KE), and Quercus semecarpifolia (QS) in the central Yunnan Plateau of China, were estimated on the basis of a field inventory to determine the distribution and altitudinal patterns of SC and SN among various forest ecosystems. The results showed that (1) the forest SC ranged from 179.58 ± 20.57 t hm−1 in QS to 365.89 ± 35.03 t hm−1 in EB. Soil, living biomass and litter contributed an average of 64.73%, 31.72% and 2.86% to forest SC, respectively; (2) the forest SN ranged from 4.47 ± 0.94 t ha−1 in PY to 8.91 ± 1.83 t ha−1 in PA. Soil, plants and litter contributed an average of 86.88%, 10.27% and 2.85% to forest SN, respectively; (3) the forest SC and SN decreased apparently with increasing altitude. The result demonstrates that changes in forest types can strongly affect the forest SC and SN. This study provides baseline information for forestland managers regarding forest resource utilization and C management.

C and N are major constituents of plant and soil organic matter and play a fundamental role in nutrient cycling, plant growth, and ecological functions 1,2 . Forest S C is the most important part of the global C pool across various terrestrial ecosystems and plays a long-term and continuous role in mitigating the effects of global warming [3][4][5] . N is a vital and limiting nutrient in forest ecosystems, and C storage is closely linked to the N cycle 6 . Furthermore, N deposition alters S C and S N 7,8 . Consequently, accurate identification of the spatial patterns of forest S C and S N is important for accessing the global C and N pool.
Forest S C is estimated to account for approximately 45% of terrestrial ecosystem S C 9,10 . In forest ecosystems, C is stored in living biomass, litter and soils [11][12][13] . Living biomass has a great capacity to sequester atmospheric C and the aboveground living biomass has been considered as a major C pool 14,15 . Soil is another indispensable component of forest ecosystems and acts as an important C pool in terrestrial ecosystems 16,17 . The amount of C stored in soil is approximately double the amount in the atmosphere 17,18 . Consequently, exploring the distribution patterns of S N in forest ecosystems is essential for understanding the C cycle. Many studies have explored the spatial distribution of S C in forest ecosystems at a landscape scale using remote sensing and statistical methods 4,12,16,19,20 ; however, these estimates are not reliable in hilly terrain, because the mountainous and hilly conditions can increase errors of forest vertical structure measured using remote sensors 13 . Hence, to accurately quantify forest S C at a large scale, it is essential to develop estimates based on ground measurements. Forest inventory data are recognized as one of the most reliable sources of data for global C cycle research 4 .
The amount of C stored in forest vegetation and soil is considered to be the result of a long-term balance between C absorption and release 20,21 . The magnitude of S C and S N in forests depends on stand age, species composition, climate variability, geographical circumstances, management strategy and natural disturbances [22][23][24]  www.nature.com/scientificreports/ distribution patterns of S C and S N also differ among spatial landscape patterns, plant species and plant organs 25 . Mopan Mountain in the central Yunnan Plateau is located in the Yunnan-Guizhou Plateau and the southern margin of the Qinghai-Tibet Plateau. The area belongs to a subtropical mountain climate region 26 , and vegetation patterns shift vertically due to changes in altitude. The main forest vegetation types are subtropical evergreen broad-leaved forest, subtropical mixed coniferous and broad-leaved forests, coniferous forest and alpine forest. In this region, forests cover more than 72.6% of the land area, and they represent the most important forest resources in the central Yunnan Plateau and in Yunnan Province. The main objectives of this study are to (1) assess the spatial variation in forest biomass based on a field inventory; (2) characterize the spatial variation in C and N density and storage in forest ecosystems; and (3) explore the impact of altitude on biomass and S N and S C in Mopan Mountain. This study will provide baseline information for forestland managers regarding forest resource utilization and C and N management. www.nature.com/scientificreports/ Plant census and sampling. In each tree plot, census of plant individuals which diameter at breast height was more than 1 cm was performed. In addition, in each shrub and herb plot, the species name and abundance were recorded 2,17,27 . All plant individuals in each plot were collected with different parts for C and N testing, i.e., trees with roots, trunks, leaves, branches and bark; shrubs with roots, stems and leaves; and herbs with above-and belowground part.

Materials and methods
Litter sampling. Triplicate plots with a size of 1 m × 1 m were established in the tree plots for ground litter sampling 28 . For each of these samples, horizons L, F and H were separated and carefully placed in plastic bags for determining of the dry weight and C and N contents. The L horizon was composed of fresh or slightly discoloured material that was not weak or friable; the F horizon was composed of medium to strongly fragmented material with many mycelia and thin roots; and the H horizon consisted of humified amorphous material.
Mineral soil sampling. Mineral soil samples were collected from each tree plot, with three replicates. Most of the slope gradients of these soil profiles were less than 15°. After removal of the forest floor mass, soil samples were collected from three layers: 0-20 cm, 20-40 cm and 40-60 cm, and the corresponding soil bulk density (BD) of each layer was measured using the cutting-ring method 29 . The soil samples were placed in sacks and air dried for soil C and N testing.
Laboratory analysis. Shrub, herb and ground litter samples were dried to a constant weight at 105 °C and then weighed for biomass estimation. Plant and soil total N concentrations were determined by a continuous flow analytical system (Analytical AA3, SEAL, Germany) with sulfuric acid (H 2 SO 4 ) and hydrogen peroxide (H 2 O 2 ) digestion 30 . The total C concentration was determined by an elemental analyser (Vario TOC cube, Elementar, Germany) 31 .

The estimation of biomass, S C , and S N . The estimation of forest vegetation biomass. Tree biomass
(roots, trunks, leaves, branches and bark biomass) was estimated using allometric equations based on longterm practical measurements of forest vegetation in southwestern China [32][33][34][35][36][37][38] . The shrub and herb biomass was directly expressed as their dry weights. For each forest plot, the total biomass was the sum of the biomass of each vegetation type in the plot.
The estimation of plant, litter and soil S C and S N . The S C (t ha −1 ) and S N (t ha −1 ) of trees, shrubs, and herbs were obtained by multiplying the forest vegetation biomass (t ha −1 ) by the corresponding C and N content coefficient 17,29 .
The litter S C and S N were the sum of the S C and S N of horizons L, F and H. The litter S C and S N storage was calculated by the following formula 17 : where S C (t ha −1 ) and S N (t ha −1 ) are the respective litter C and N storage; TC i and TN i are the C and N (g kg −1 ) contents of horizons L, F and H, respectively; and LB i is the litter biomass (dry litter weight) of horizons L, F and H.
The soil S C and S N were calculated as the sum of the S C and S N of the 0-20 cm, 20-40 cm and 40-60 cm soil layers. The soil S C and S N were calculated using the following formula 21,39 : where S C and S N are soil total C storage (t ha −1 ) and N storage (t ha −1 ), respectively, BD i is the soil BD (g cm -3 ), TC i and TN i are the soil total C and N contents (g kg −1 ), respectively, and D i is the soil layer thickness (cm).
Statistical analysis. Statistical analyses were carried out using the software Statistical Package for the Social Sciences 19 (SPSS 19) and Microsoft Office Excel (version 2013). One-way ANOVA was used to test whether the variations in S C and S N were significantly different among the plant, litter and soil and forest type components. Duncan's shortest range test was used to examine the difference among different forest types at P < 0.05. The relationships between altitude and biomass, S C and S N were examined by linear regression.

Results
Biomass in forest ecosystems. The biomass of the forest ecosystems in the central Yunnan Plateau ranged from 142.36 ± 18.36 to 271.77 ± 34.71 t ha −1 . The biomass of the forest ecosystems was significantly different among the various forests (Table 2). Plant biomass made a significant contribution to ecosystem biomass and accounted for a much higher proportion (more than 90%) than forest litter. Tree biomass was significantly higher than that of shrubs and herbs in PY, PA, KE and EB and accounted for 99.64%, 94.46%, 95.33% and 95.88% of the total plant biomass, respectively. The tree and shrub biomass in QS accounted for a nearly equal proportion of plant biomass at 46.72% and 51.01%, respectively. The biomass of each component of plants and litter is presented in Fig. 1.  . 2 A-B). Generally, the N concentration was classified into three levels by the lines in the figure. The N concentration was highest in the leaves of trees and shrubs and the aboveground parts of herbs, and it ranged from 6.64 ± 2.01 to 21.99 ± 6.66 g·kg −1 . The N concentration in tree branches, shrub stems and the L, F and H litter horizons ranged from 3.86 ± 0.90 to 8.78 ± 1.73 g·kg −1 , and these values were higher than those in the roots and trunks of trees, shrub roots and soil, which had N concentrations lower than 4.89 ± 1.31 g·kg −1 . Significant differences were not observed in the C concentrations in the plant and litter components among different forests, which ranged from 323.21 ± 63.58 to 503.00 ± 97.56 g·kg −1 , and the mean C concentration of the forest vegetation and litter was 425.80 ± 100.34 g·kg −1 . However, the soil C concentrations were significantly lower than those in the plants and litter, i.e., less than 81.08 ± 13.62 g kg −1 , with a mean of 29.74 ± 12.20 g·kg −1 .  www.nature.com/scientificreports/ S C and S N . The ecosystem S C was calculated as the sum of the plant S C , litter S C and soil S C . The ecosystem S C was significantly different among the forests (Table 3 and Fig. 3 A) The plant S C of different forests varied significantly from 62.70 ± 11.33 t hm −1 in QS to 120.35 ± 13.01 t·hm −1 in EB, although the difference between QS and PA was not significant. Tree S C contributed more than 94% to the plant S C in PY, PA, KE and EB; however, the tree S C of QS contributed only 47.11% to the plant S C . The shrub S C of QS accounted for a high proportion of 50.78% of the plant S C , whereas the shrub and herb S C in the other four forests contributed less than 1% to the plant S C. The litter S C varied from a high concentration of 13.20 ± 2.12 t hm −1 in PY to a low concentration of 4.82 ± 0.77 t·hm −1 in QS. Generally, the S C of different layers among the forests decreased in the order of H > F > L, while the litter S C in EB decreased in the order of H > L > F. The highest soil S C was in EB at 240.59 ± 32.90 t hm −1 . The soil S C in KE was 219.21 ± 29.98 t hm −1 , which was significantly lower than that in EB but significantly higher than that in PY and PA, which were 164.42 ± 22.90 t hm −1 and 129.20 ± 17.67 t hm −1 , respectively. The lowest soil S C was in QS at 112.06 ± 15.32 t hm −1 . In KE, the S C at 20-40 cm was higher than that at 0-20 cm and the soil S C decreased with increasing soil depth.
The S N of the forest ecosystems varied significantly among the forests, although significant differences were not found between PY and PA (Table 3 and Fig. 3 B). The Ecosystem S N ranged from 8.91 ± 1.83 t·ha −1 in EB to 4.47 ± 0.94 t·ha −1 in PA, and the S N in KE, QS and PY was 7.13 ± 1.52 t·ha −1 , 6.36 ± 1.19 t ha −1 and 5.14 ± 1.10 t ha −1 , respectively. Soil was the most important contributor to total S N in the forest ecosystems and accounted for  The plant S N differed significantly among the forests with different species and ranged from a high concentration of 1.11 ± 0.33 t·ha −1 in EB to a low concentration of 0.39 ± 0.17 t·ha −1 in PA. The tree S N of PY, PA, KE and EB accounted for more than 85% of the living biomass S N , and the shrubs and herbs contributed less than 15%. However, the shrubs in QS stored more N than trees and the S N of shrubs and trees contributed 24.42% and 70.36% to the living biomass S N , respectively. The estimated mean S N of forest litter was 0.17 ± 0.01 t·ha −1 , and the H layer stored approximately half of the litter N. The soil is a large N pool in forest ecosystems, and in this study, the soil S N accounted for 86.88% on average of Eco S N . More S N was stored in the topsoil (0-20 cm), with a contribution of 53.69%, 45.71%, 45.87%, 39.27% and 42.85% to the total soil (0-60 cm) S N in QS, PY, PA, KE and EB, respectively.

Correlation analysis of biomass, S C and S N in forest ecosystems and altitude. The generalized
linear model illustrates the effects of altitude on biomass, S C and S N of forest ecosystems, which decreased with increasing altitude (Fig. 4A-C). Whether calculated within the same forest or across all forests, significant correlations were found between the altitude and biomass at the P < 0.005 level ( Table 4). The S C of QS, PA, KE and EB was also significantly (P < 0.005) correlated with altitude, but for PY, the correlation between S C and altitude was significant at the P < 0.05 level. In all forests, S C decreased significantly (P < 0.001) with increasing altitude. A significant correlation (P < 0.005) was observed between S N and altitude in QS, PA, KE and EB, and a less significant correlation was observed between S N and altitude (P < 0.01) in PY. However, for all forests, the variation in S N was not significant (P = 0.400) with respect to altitude.

Discussion
Different forest ecosystems have different C sequestration capacities. The total S C values of the forest ecosystems of KE and EB in Mopan Mountain in the central Yunnan Plateau are apparently higher than the average values of forest ecosystems (258.83 t C•ha −1 ) across China 27,40 , whereas the total S C values of the forest ecosystems of QS PA and PY are lower. The results of the present research show that changes in forest types can strongly affect S C and S N values. Generally, broad-leaved species can store more C and N than conifers 30,41 . Although the alpine forest (QS) had the lowest S C , its S N was higher than that in PY and PA. The S C and S N of forests in Mopan Mountain in the central Yunnan Plateau averaged 266.75 ± 26.40 t ha −1 and 6.40 ± 1.32 t ha −1 , respectively . With respect to S C , the living biomass, litter layer and soil accounted for 31.72, 3.55% and 64.73% of the total C storage, respectively. The corresponding S N accounted for 10.27%, 2.85% and 86.88% of the total N storage, respectively. The current and previous results indicate that the soil is the most important component for S C and S N in forest ecosystems 42,43 .
The living biomass of forests is one of the major C and N pools. Quantification of stored C in the living biomass of a forest is necessary for future management 44 . The estimated mean living biomass S C in this study was 84.12 t·ha −1 , which is much higher than the average values of vegetation C storage in Chinese forest ecosystems (57.07 t ha −1 ) 45,46 . This finding is mainly because of the high tree density and low anthropogenic disturbance at the location of Mopan Mountain National Forest Park. The tree growth rate and biomass allocation to different tree parts and varying rates of C sequestration in ecosystem components can affect the rate of C sequestration and longevity of C storage 2,41 . The present study showed that the S C in plants ranged from 62.70 ± 11.33 t·ha −1 in QS to 120.35 ± 13.01 t·ha −1 in EB, which accounted for 31.72% of the total C storage. Among all forests, QS had the lowest S C in living biomass, which was caused by its lower biomass and lower C concentration in living biomass.  www.nature.com/scientificreports/ However, the higher biomass in EB resulted in higher S C of living biomass compared with the other groups in Mopan Mountain in the central Yunnan Plateau. The S N in living biomass varied from a high of 1.11 ± 0.33 t·ha −1 in EB to a low of 0.39 ± 0.17 t·ha −1 in PA, with a mean contribution of 10.27% to total S N .Tree S C and S N accounted for a large proportion of living biomass S C and S N in PY, PA, KE and EB, whereas shrubs contributed more C and N than trees to living biomass S C and S N in QS. The S C and S N of vegetation are mainly determined by the biomass of live vegetation components and C and N contents. Consequently, the interspecific differences in tree biomass caused by inherent variation in growth rates [47][48][49] were the main reasons for the variations in S C and S N allocation among forests. Furthermore, the effect of forest species on the growth and diversity of understorey plant biomass 2,30,50 also resulted in the variation in S C and S N allocation in forest vegetation. Forest litter and its decomposition rate are key factors in nutrient cycling in forest ecosystems 51 , and the current litter S C in the world's forests is estimated at 43 ± 3 Pg·C (5% of total forest C) 52 . In the present study, the estimated mean litter S C and S N in the forests were 8.93 ± 1.44 t ha −1 and 0.17 ± 0.01 t ha −1 , which accounted for 3.55% and 2.85% of the total S C and S N , respectively. The mean litter S C in this study is slightly higher than the mean litter S C in China (8.21 t·ha −1 ) 49 . The study also found that conifer litter stored more C and N than broadleaf litter, and a similar result was found in previous studies 41,53 . The above results occurred mainly because conifer litter is more difficult to decompose than broadleaf litter, resulting in a higher rate of litter accumulation on the forest floor.
The estimated mean soil S C and S N of different forests in this study were 173.70 ± 23.75 t·ha −1 and 5.56 ± 1.08 t·ha −1 , which accounted for 64.73% and 86.88%, respectively, of the total S C and S N . The results showed that soil is the largest C pool in forest ecosystems, similar to a previous study conducted in China 2,30,42 . The mean reported value of soil S C was 193.55 t ha −1 in Chinese forest ecosystems 45,46 , and the soil S N was 6.27 t ha −1 in subtropical forests of China 54   www.nature.com/scientificreports/ than the mean soil S N in China and soil S N in subtropical forests in China. The C stored in soil is significantly influenced by the C inputs (e.g., litter decomposition) and soil organic matter decomposition 55 . Therefore, S C is determined by the balance between the input or output patterns and controlled mainly by tree species under similar environmental conditions 17,41 . There were significant differences in the soil S C and S N at depths of 0-20 cm, 20-40 cm and 40-60 cm among the forests. The topsoil (0-20 cm) in the forests stored 43.38% of the C and 45.48% of the N from 0 to 60 cm. The soil C and N were mainly stored in the topsoil 42,43,56 , which is probably because of the variation in the soil bulk density and concentrations of C and N in soil layers, which are two important determining factors of S C and S N at fixed soil depths 17,57 . Although the soil bulk density decreased with increasing soil depth, the topsoil contained more C and N.  www.nature.com/scientificreports/ increasing altitude, although the S C of PY and the S N of all forest ecosystems in this study were not highly significantly correlated with altitude. Previous reports indicated that the soil S C in forest ecosystems increases with altitude 58,59 and the living biomass and total S C of forest ecosystems decreased significantly with increasing latitude in different regions 60,61 because increasing altitude changed the climate factors (i.e., temperature and precipitation) and resulted in the shifting of vegetation types and a decline in net primary production and litterfall 58,62,63 . The vegetation patterns in the study area shifted vertically due to changes in altitude. With increasing altitude, the forest vegetation types in this area shifted from subtropical evergreen broad-leaved forest, subtropical mixed coniferous and broad-leaved forest, and coniferous forest to alpine forest, and the living biomass of the forests declined significantly. Therefore, the total S C and S N of forest ecosystems exhibited decreasing trends with increasing altitude.
Received: 16 October 2020; Accepted: 4 March 2021 Table 4. The results of the generalized linear model analyses of the effects of altitude on biomass, S C and S N . Three linear models were built: model I: Biomass (t ha −1 ) = C + a × AL (altitude, m); model II: S C (t ha −1 ) = C + a × AL (altitude, m); and model III: S N (t ha −1 ) = C + a × AL (altitude, m), where C is the regression constant and a is the regression coefficient of the given variable.