Zhenhua Hu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Zhenhua Hu has a total of 21 co-authors. Here are the top co-authors by number of shared publications:
Bin Yang
College of Forestry, Shanxi Agriculture University, China
1 shared publication- Bin Yang
College of Forestry, Shanxi Agriculture University, Taigu, 030801, China
1 shared publication - Bingjie Liu
Shanxi Agricultural University
1 shared publication Hanyu Zhang
College of Forestry, Shanxi Agriculture University, China
1 shared publication- Hanyu Zhang
College of Forestry, Shanxi Agriculture University, Taigu 030801, China
1 shared publication - Haodong Liu
China Aero Geophysical Survey and Remote Sensing Center for Natural Resources, Beijing 100083, China
1 shared publication
Latest Publications
- Open Access
- 4 June 2025
- Published by Frontiers Media SA in Frontiers in Forests and Global Change
- Vol. 8, 1431603
Efficient tree species classification using machine and deep learning algorithms based on UAV-LiDAR data in North China
Introduction: The unmanned aerial vehicle -based light detection and ranging (UAV-LiDAR) can quickly acquire the three-dimensional information of large areas of vegetation, and has been widely used in tree species classification.Methods: UAV-LiDAR point clouds of Populus alba, Populus simonii, Pinus sylvestris, and Pinus tabuliformis from 12 sample plots, 2,622 tree in total, were obtained in North China, training and testing sets were constructed through data pre-processing, individual tree segmentation, feature extraction, Non-uniform Grid and Farther Point Sampling (NGFPS), and then four tree species were classified efficiently by two machine learning algorithms and two deep learning algorithms.Results: Results showed that PointMLP achieved the best accuracy for identification of the tree species (overall accuracy = 96.94%), followed by RF (overall accuracy = 95.62%), SVM (overall accuracy = 94.89%) and PointNet++(overall accuracy = 85.65%). In addition, the most suitable number of point cloud sampling of single tree is between 1,024 and 2048 when using the NGFPS method in the two deep learning models. Furthermore, feature value of elev_percentile_99th has an important influence on tree species classification and tree species with similar crown structures may lead to a higher misidentification rate.Discussion: The study underscores the efficiency of PointMLP as a robust and streamlined solution, which offers a novel technological support for tree species classification in forestry resource management.
- 4 June 2025
- Published by Frontiers Media SA in Frontiers in Forests and Global Change
- Vol. 8, 1431603
Open AccessEfficient tree species classification using machine and deep learning algorithms based on UAV-LiDAR data in North China
Introduction: The unmanned aerial vehicle -based light detection and ranging (UAV-LiDAR) can quickly acquire the three-dimensional information of large areas of vegetation, and has been widely used in tree species classification.Methods: UAV-LiDAR point clouds of Populus alba, Populus simonii, Pinus sylvestris, and Pinus tabuliformis from 12 sample plots, 2,622 tree in total, were obtained in North China, training and testing sets were constructed through data pre-processing, individual tree segmentation, feature extraction, Non-uniform Grid and Farther Point Sampling (NGFPS), and then four tree species were classified efficiently by two machine learning algorithms and two deep learning algorithms.Results: Results showed that PointMLP achieved the best accuracy for identification of the tree species (overall accuracy = 96.94%), followed by RF (overall accuracy = 95.62%), SVM (overall accuracy = 94.89%) and PointNet++(overall accuracy = 85.65%). In addition, the most suitable number of point cloud sampling of single tree is between 1,024 and 2048 when using the NGFPS method in the two deep learning models. Furthermore, feature value of elev_percentile_99th has an important influence on tree species classification and tree species with similar crown structures may lead to a higher misidentification rate.Discussion: The study underscores the efficiency of PointMLP as a robust and streamlined solution, which offers a novel technological support for tree species classification in forestry resource management.
- Open Access
Linking Soil Properties and Bacterial Communities with Organic Matter Carbon During Vegetation Succession
Land use change driven by vegetation succession significantly enhances soil carbon storage, yet the microbial mechanisms underlying this process remain poorly understood. This study aims to elucidate the mechanistic linkages between bacterial community dynamics and organic matter carbon stabilization across four vegetation succession stages on the Loess Plateau: abandoned farmland (AF), grassland stage (GS), shrub-land stage (SS), and forest stage (FS). We analyzed soil organic matter carbon (SOM_C) fractions, physicochemical properties, and bacterial communities (16S rRNA sequencing), employing structural equation modeling to quantify causal pathways. The results showed that the content of soil total organic matter carbon (TOM_C), labile organic matter carbon (LOM_C), dissolved organic matter carbon (DOM_C), and microbial biomass carbon (MBC) increased progressively with succession, peaking in the FS, with 23.87 g/kg, 4.13 g/kg, 0.33 mg/kg, and 0.14 mg/kg, respectively. Furthermore, vegetation succession also led to heterogeneity in the bacterial community structure. The number of soil bacterial operational taxonomic units (OTUs) for the four succession stages was 9966, 13,463, 14,122, and 10,413, with the shrub-land stage showcasing the highest OTUs. Nine bacterial taxa were strongly correlated with SOM_C stabilization. Affected by soil bacteria, soil physicochemical properties and litter biomass directly influence SOM_C, with the physicochemical pathway (path coefficient: 0.792, p < 0.001) having a greater impact on organic matter carbon than the litter pathway (path coefficient: 0.221, p < 0.001). This study establishes that vegetation succession enhances SOM_C content not only through increased litter inputs but also by reshaping bacterial communities toward taxa that stabilize carbon via physicochemical interactions.
Open AccessLinking Soil Properties and Bacterial Communities with Organic Matter Carbon During Vegetation Succession
Land use change driven by vegetation succession significantly enhances soil carbon storage, yet the microbial mechanisms underlying this process remain poorly understood. This study aims to elucidate the mechanistic linkages between bacterial community dynamics and organic matter carbon stabilization across four vegetation succession stages on the Loess Plateau: abandoned farmland (AF), grassland stage (GS), shrub-land stage (SS), and forest stage (FS). We analyzed soil organic matter carbon (SOM_C) fractions, physicochemical properties, and bacterial communities (16S rRNA sequencing), employing structural equation modeling to quantify causal pathways. The results showed that the content of soil total organic matter carbon (TOM_C), labile organic matter carbon (LOM_C), dissolved organic matter carbon (DOM_C), and microbial biomass carbon (MBC) increased progressively with succession, peaking in the FS, with 23.87 g/kg, 4.13 g/kg, 0.33 mg/kg, and 0.14 mg/kg, respectively. Furthermore, vegetation succession also led to heterogeneity in the bacterial community structure. The number of soil bacterial operational taxonomic units (OTUs) for the four succession stages was 9966, 13,463, 14,122, and 10,413, with the shrub-land stage showcasing the highest OTUs. Nine bacterial taxa were strongly correlated with SOM_C stabilization. Affected by soil bacteria, soil physicochemical properties and litter biomass directly influence SOM_C, with the physicochemical pathway (path coefficient: 0.792, p < 0.001) having a greater impact on organic matter carbon than the litter pathway (path coefficient: 0.221, p < 0.001). This study establishes that vegetation succession enhances SOM_C content not only through increased litter inputs but also by reshaping bacterial communities toward taxa that stabilize carbon via physicochemical interactions.
- Open Access
The Effect of Soil and Topography Factors on Larix gmelinii var. Principis-rupprechtii Forest Mortality and Capability of Decision Tree Binning Method and Generalized Linear Models in Predicting Tree Mortality
Understanding the factors influencing individual tree mortality is essential for sustainable forest management, particularly for Prince Rupprech’s larch (Larix gmelinii var. Principis-rupprechtii) in North China’s natural forests. This study focused on 20 sample plots (20 × 20 m each) established in Shanxi Province, North China. This study compared three individual tree mortality models—Generalized Linear Model (GLM), Linear Discriminant Analysis (LDA), and Bayesian Generalized Linear Model (Bayesian GLM)—finding that both GLM and Bayesian GLM achieved approximately 0.87 validation accuracy on the test dataset. Due to its simplicity, GLM was selected as the final model. Building on the GLM model, six binning methods were applied to categorize diameter at breast height (DBH): equal frequency binning, equal width binning, cluster-based binning, quantile binning, Chi-square binning, and decision tree binning. Among these, the decision tree binning method achieved the highest performance, with an accuracy of 90.12% and an F1 score of 90.06%, indicating its effectiveness in capturing size-dependent mortality patterns. This approach provides valuable insights into factors affecting mortality and offers practical guidance for managing Larix gmelinii var. Principis-rupprechtii forests in temperate regions.
Open AccessThe Effect of Soil and Topography Factors on Larix gmelinii var. Principis-rupprechtii Forest Mortality and Capability of Decision Tree Binning Method and Generalized Linear Models in Predicting Tree Mortality
Understanding the factors influencing individual tree mortality is essential for sustainable forest management, particularly for Prince Rupprech’s larch (Larix gmelinii var. Principis-rupprechtii) in North China’s natural forests. This study focused on 20 sample plots (20 × 20 m each) established in Shanxi Province, North China. This study compared three individual tree mortality models—Generalized Linear Model (GLM), Linear Discriminant Analysis (LDA), and Bayesian Generalized Linear Model (Bayesian GLM)—finding that both GLM and Bayesian GLM achieved approximately 0.87 validation accuracy on the test dataset. Due to its simplicity, GLM was selected as the final model. Building on the GLM model, six binning methods were applied to categorize diameter at breast height (DBH): equal frequency binning, equal width binning, cluster-based binning, quantile binning, Chi-square binning, and decision tree binning. Among these, the decision tree binning method achieved the highest performance, with an accuracy of 90.12% and an F1 score of 90.06%, indicating its effectiveness in capturing size-dependent mortality patterns. This approach provides valuable insights into factors affecting mortality and offers practical guidance for managing Larix gmelinii var. Principis-rupprechtii forests in temperate regions.