Aim: Motivated by the need to enhance the accuracy of forest aboveground carbon storage (ACS) assessments, this study aimed to explore the effectiveness of different machine learning models in predicting ACS across various subtropical forest types in southern China. Location: The study was conducted in southern China, focusing on different types of subtropical forests. This region harbors several types of subtropical forests, which are rarely found at similar latitudes in the world. Methods: Variance inflation factor was employed to screen independent variables, resulting in the selection of 13 significant predictors. Four machine learning models—support vector machine (SVM), random forest (RF), multi-layer perceptron (MLP), and extreme gradient boosting (XGB)—were constructed to estimate carbon storage. Model performance was evaluated using root mean square error, coefficient of determination (R 2), and mean absolute error. The model with the best generalization ability was selected to calculate SHAP values for each predictor. Results: The XGB model demonstrated superior performance across all forest types, with R 2 values ranging from 0.898 to 0.974. In mountainous evergreen broad-leaved forests, the prediction accuracy followed the order of XGB > MLP > SVM > RF. In valley rainforests, MLP showed the highest R 2 value, but with higher MAE and RMSE, making it the second-best choice. The RF model performed moderately, while the SVM model showed the poorest performance. The SHAP values indicated that maximum diameter at breast height, slope, mean DBH, species evenness, altitude, and maximum tree height had significant effects on ACS. Conclusions: XGB model exhibits the best prediction performance and strongest adaptability for estimating ACS in subtropical southern China forests.Machine learning methods provide valuable references for predicting and assessing ACS in different types of zonal forests.

Phylogenetic trees have been extensively used in community ecology. However, how the phylogenetic reconstruction affects ecological inferences is poorly understood. In this study, we reconstructed three different types of phylogenetic trees (a synthetic-tree generated using VPhylomaker, a barcode-tree generated using rbcL+matK+trnH-psbA and a genome-tree generated from plastid genomes) that represented an increasing level of phylogenetic resolution among 580 woody plant species from six dynamic plots in subtropical evergreen broadleaved forests of China. We then evaluated the performance of each phylogeny in estimations of community phylogenetic structure, turnover and phylogenetic signal in functional traits. As expected, the genome-tree was most resolved and most supported for relationships among species. For local phylogenetic structure, the three trees showed consistent results with Faith’s PD and MPD; however, only the synthetic-tree produced significant clustering patterns using MNTD for some plots. For phylogenetic turnover, contrasting results between the molecular trees and the synthetic-tree occurred only with nearest neighbor distance. The barcode-tree agreed more with the genome-tree than the synthetic-tree for both phylogenetic structure and turnover. For functional traits, both the barcode-tree and genome-tree detected phylogenetic signal in maximum height, but only the genome-tree detected signal in leaf width. This is the first study that uses plastid genomes in large-scale community phylogenetics. Our results highlight the outperformance of genome-trees over barcode-trees and synthetic-trees for the analyses studied here. Our results also point to the possibility of Type I and II errors in estimation of phylogenetic structure and turnover and detection of phylogenetic signal when using synthetic-trees.

Light-regime variability is an important environmental factor which shapes a forest community. So far, none focused on the phylogenetic pattern of plant light receptors, which reflects how genes’ evolution influences the coexistence of species in a community. In this study, we analyzed community phylogenetic structure of the south subtropical forest by sequences of plant blue light receptor cryptochrome (CRY) and compared the results of DNA barcodes. Patterns of community assembly was estimated by net relatedness index (NRI) and nearest taxon index (NTI). We found that CRY showed quite different phylogenetic structure as compared to DNA barcoding results, all habitats displayed consistent phylogenetic structure patterns, suggesting a convergent evolution of light sensing system of plant in local adaptation. Also, both NRI and NTI values increased through the time, indicating that the phylogenetic structure of tree community became more overdispersion as succession proceeds; phylogenetic closely-related species tended to co-occur and environmental filtering played a more important role in the community assembly. Furthermore, phylogenetic patterns were more clustering in upper canopy layers, and NTI values of all canopy layers were above zero, suggesting that phylogenetically related species tended to coexist and adapted to similar light conditions.

Serious disconnection among researches on individual physiological adaption, population genetic diversity and spatiotemporal demography has obstructed us in the knowledge of plant population ecology. Here we develop an approach to integrate those three research aspects by taking advantage of recent knowledge about DNA methylation and multivariate analysis. We show that by using various epigenetic parameters corresponding to individual physiological metabolic reprogramming potential, gene expression repression degree and physiological reaction characteristics, the contribution of various biotic and abiotic factors to an individual state and population structure can be quantified. Furthermore, population dynamics can be narrowly estimated by analysing DNA methylation of populations at different developmental stages. This study demonstrates an approach for the overall analysis of plant populations and exploring the spatiotemporal dynamics and developmental mechanisms of a population using Castanopsis chinensis as the model species.