Associations between stand structure and the predictability of multiple forest carbon pools in the transition zone between the Qinghai–Tibet and Loess Plateaus
- 1Gansu Academy of Forestry
- 2Key Laboratory of System Governance of Mountains, Rivers, Forests, Farmlands, Lakes, Grasslands and Deserts, National Forestry and Grassland Administration
- 3Gansu Xinglong Mountain Forest Ecosystem Positioning Research Station
Abstract
Introduction: Complex mountain forests in the transition zone between the Qinghai–Tibet and Loess Plateaus are characterized by strong topographic heterogeneity and marked stand structural variation. These conditions may lead to differences in the predictability of forest carbon pools. However, the role of stand structure in shaping the predictability of multiple forest carbon pools remains insufficiently understood.
Methods: In this study, we used field survey data from 90 plots (77 forest plots and 13 shrubland plots) collected in 2024-2025 in the Taohe National Nature Reserve, Gansu Province, together with vegetation indices, topographic, climatic, soil, and stand structural variables, to develop random forest models for multiple forest carbon pools. For plot-scale modelling, 77 forest plots were used, and soil carbon was assessed at 0–60 cm depth. We assessed the composition, predictability, and major predictors of biomass, soil, litter, deadwood, and ecosystem carbon densities.
Results: The mean ecosystem carbon density was 306.79 Mg C ha ⁻¹ , with soil carbon and biomass carbon accounting for 71.5% and 24.9%, respectively, suggesting that these two pools comprised the dominant carbon pools. Plot-scale random forest models showed clear differences in predictive performance among carbon pools. Biomass and ecosystem carbon densities were predicted with relatively high accuracy, with cross-validated R2 values of 0.766 and 0.659, respectively. Litter carbon density showed moderate predictability, whereas soil and deadwood carbon densities were poorly predicted. Biomass carbon density was primarily associated with stand volume proxy, tree height, and basal area proxy, whereas ecosystem carbon density was mainly associated with tree height, stand volume proxy, and stand age. Spatial random forest models showed relatively poor predictive performance.
Discussion: These differences in predictive performance indicate that stand structural variables were closely associated with the predictability of biomass and ecosystem carbon densities. Overall, carbon density predictability in complex mountain forests showed clear component dependence. Component-specific modelling of multiple forest carbon pools helps reveal differences in their key predictors, predictive performance, and uncertainties associated with spatial extrapolation, and may provide useful information for forest carbon monitoring, spatial pattern characterization, and carbon management zoning in the Taohe Reserve.
Related
- Community Assembly Mechanisms Facilitate Understanding of Microbial Responses to Habitat Transition along a Spatial Environmental Continuum in Alpine Plateaus
- Aberrant single-subject cortical morphological networks and their associations with multiple symptoms in generalized anxiety disorder
- Genetic mechanisms underlying the associations between brain structure and mental health
- A survey of bacterial communities reveals host-specific associations across multiple Pentatomomorpha species in France
- Multiple energy pathways structure mercury biomagnification in an Amazonian River food web
BibTeX
@article{zong2026associations,
title = {{Associations between stand structure and the predictability of multiple forest carbon pools in the transition zone between the Qinghai–Tibet and Loess Plateaus}},
author = {Wenzhen Zong and Zengcao Ren and Lei Ling and Zhengni Chen},
journal = {Frontiers in Plant Science},
year = {2026},
month = oct,
doi = {10.3389/fpls.2026.1935881},
url = {https://www.frontiersin.org/articles/10.3389/fpls.2026.1935881/full},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Associations between stand structure and the predictability of multiple forest carbon pools in the transition zone between the Qinghai–Tibet and Loess Plateaus
AU - Wenzhen Zong
AU - Zengcao Ren
AU - Lei Ling
AU - Zhengni Chen
PY - 2026
DA - 2026/10/21
JO - Frontiers in Plant Science
DO - 10.3389/fpls.2026.1935881
UR - https://www.frontiersin.org/articles/10.3389/fpls.2026.1935881/full
AB - Introduction: Complex mountain forests in the transition zone between the Qinghai–Tibet and Loess Plateaus are characterized by strong topographic heterogeneity and marked stand structural variation. These conditions may lead to differences in the predictability of forest carbon pools. However, the role of stand structure in shaping the predictability of multiple forest carbon pools remains insufficiently understood. Methods: In this study, we used field survey data from 90 plots (77 forest plots and 13 shrubland plots) collected in 2024-2025 in the Taohe National Nature Reserve, Gansu Province, together with vegetation indices, topographic, climatic, soil, and stand structural variables, to develop random forest models for multiple forest carbon pools. For plot-scale modelling, 77 forest plots were used, and soil carbon was assessed at 0–60 cm depth. We assessed the composition, predictability, and major predictors of biomass, soil, litter, deadwood, and ecosystem carbon densities. Results: The mean ecosystem carbon density was 306.79 Mg C ha ⁻¹ , with soil carbon and biomass carbon accounting for 71.5% and 24.9%, respectively, suggesting that these two pools comprised the dominant carbon pools. Plot-scale random forest models showed clear differences in predictive performance among carbon pools. Biomass and ecosystem carbon densities were predicted with relatively high accuracy, with cross-validated R2 values of 0.766 and 0.659, respectively. Litter carbon density showed moderate predictability, whereas soil and deadwood carbon densities were poorly predicted. Biomass carbon density was primarily associated with stand volume proxy, tree height, and basal area proxy, whereas ecosystem carbon density was mainly associated with tree height, stand volume proxy, and stand age. Spatial random forest models showed relatively poor predictive performance. Discussion: These differences in predictive performance indicate that stand structural variables were closely associated with the predictability of biomass and ecosystem carbon densities. Overall, carbon density predictability in complex mountain forests showed clear component dependence. Component-specific modelling of multiple forest carbon pools helps reveal differences in their key predictors, predictive performance, and uncertainties associated with spatial extrapolation, and may provide useful information for forest carbon monitoring, spatial pattern characterization, and carbon management zoning in the Taohe Reserve.
SN - 1664-462X
ER -