The forest stock volume (FSV) is one of the key indicators in forestry resource assessments on local, regional, and national scales. To date, scaling up in situ plot-scale measurements across landscapes is still a great challenge in the estimation of FSVs. In this study, Sentinel-2 imagery, the Google Earth Engine (GEE) cloud computing platform, three base station joint differential positioning technology (TBSJDPT), and three algorithms were used to build an FSV model for forests located in Hunan Province, southern China. The GEE cloud computing platform was used to extract the imagery variables from the Sentinel-2 imagery pixels. The TBSJDPT was put forward and used to provide high-precision positions of the sample plot data. The random forests (RF), support vector regression (SVR), and multiple linear regression (MLR) algorithms were used to estimate the FSV. For each pixel, 24 variables were extracted from the Sentinel-2 images taken in 2017 and 2018. The RF model performed the best in both the training phase (i.e., R² = 0.91, RMSE = 35.13 m³ ha⁻¹, n = 321) and in the test phase (i.e., R² = 0.58, RMSE = 65.03 m³ ha⁻¹, and n = 138). This model was followed by the SVR model (R² = 0.54, RMSE = 65.60 m³ ha⁻¹, n = 321 in training; R² = 0.54, RMSE = 66.00 m³ ha⁻¹, n = 138 in testing), which was slightly better than the MLR model (R² = 0.38, RMSE = 75.74 m³ ha⁻¹, and n = 321 in training; R² = 0.49, RMSE = 70.22 m³ ha⁻¹, and n = 138 in testing) in both the training phase and test phase. The best predictive band was Red-Edge 1 (B5), which performed well both in the machine learning methods and in the MLR method. The Blue band (B2), Green band (B3), Red band (B4), SWIR2 band (B12), and vegetation indices (TCW, NDVI_B5, and TCB) were used in the machine learning models, and only one vegetation index (MSI) was used in the MLR model. We mapped the FSV distribution in Hunan Province (3.50 × 10⁸ m³) based on the RF model; it reached a total accuracy of 63.87% compared with the official forest report in 2017 (5.48 × 10⁸ m³). The results from this study will help develop and improve satellite-based methods to estimate FSVs on local, regional and national scales.

Accurate estimation of forest height is crucial for the estimation of forest aboveground biomass and monitoring of forest resources. Remote sensing technology makes it achievable to produce high-resolution forest height maps in large geographical areas. In this study, we produced a 25 m spatial resolution wall-to-wall forest height map in Baoding city, north China. We evaluated the effects of three factors on forest height estimation utilizing four types of remote sensing data (Sentinel-1, Sentinel-2, ALOS PALSAR-2, and SRTM DEM) with the National Forest Resources Continuous Inventory (NFCI) data, three feature selection methods (stepwise regression analysis (SR), recursive feature elimination (RFE), and Boruta), and six machine learning algorithms (k-nearest neighbor (k-NN), support vector machine regression (SVR), random forest (RF), gradient boosting decision tree (GBDT), extreme gradient boosting (XGBoost), and categorical boosting (CatBoost)). ANOVA was adopted to quantify the effects of three factors, including data source, feature selection method, and modeling algorithm, on forest height estimation. The results showed that all three factors had a significant influence. The combination of multiple sensor data improved the estimation accuracy. Boruta’s overall performance was better than SR and RFE, and XGBoost outperformed the other five machine learning algorithms. The variables selected based on Boruta, including Sentinel-1, Sentinel-2, and topography metrics, combined with the XGBoost algorithm, provided the optimal model (R² = 0.67, RMSE = 2.2 m). Then, we applied the best model to create the forest height map. There were several discrepancies between the generated forest height map and the existing map product, and the values with large differences between the two maps were mostly distributed in the steep areas with high slope values. Overall, we proposed a methodological framework for quantifying the importance of data source, feature selection method, and machine learning algorithm in forest height estimation, and it was proved to be effective in estimating forest height by using freely accessible multi-source data, advanced feature selection method, and machine learning algorithm.

China’s Conversion of Cropland to Forests Program (CCFP) is the world’s largest afforestation-based Payments for Ecosystem Services (PES) program, having retired and afforested over 24 million ha involving 32 million rural households. Prior research has primarily focused on the CCFP’s rural welfare impacts, with few studies on program-induced environmental improvements, particularly at the household level. In this study, data from a 2010 survey covering 2808 rural households from across China was analyzed using an interval regression model to explain household-reported survival rates of trees planted on program-enrolled cropland. In addition to household-level factors, we explore the influence of local conditions and institutional configurations by exploiting the wide diversity of contexts covered by the data set. We find that households with more available labor and more forestry experience manage trees better, but that higher opportunity costs for both land and labor have the opposite effect. We also find that the local implementation regime- e.g., the degree of prior consultation with participants and regular monitoring - has a strong positive effect on reported survivorship. We suggest that the level of subsidy support to participating households will be key to survivorship of trees in planted CCFP forests for some time to come.

Timely, accurate estimates of forest aboveground carbon density (AGC) are essential for understanding the global carbon cycle and providing crucial reference information for climate-change-related policies. To date, airborne LiDAR has been considered as the most precise remote-sensing-based technology for forest AGC estimation, but it suffers great challenges from various uncertainty sources. Stratified estimation has the potential to reduce the uncertainty and improve the forest AGC estimation. However, the impact of stratification and how to effectively combine stratification and modeling algorithms have not been fully investigated in forest AGC estimation. In this study, we performed a comparative analysis of different stratification approaches (non-stratification, forest type stratification (FTS) and dominant species stratification (DSS)) and different modeling algorithms (stepwise regression, random forest (RF), Cubist, extreme gradient boosting (XGBoost) and categorical boosting (CatBoost)) to identify the optimal stratification approach and modeling algorithm for forest AGC estimation, using airborne LiDAR data. The analysis of variance (ANOVA) was used to quantify and determine the factors that had a significant effect on the estimation accuracy. The results revealed the superiority of stratified estimation models over the unstratified ones, with higher estimation accuracy achieved by the DSS models. Moreover, this improvement was more significant in coniferous species than broadleaf species. The ML algorithms outperformed stepwise regression and the CatBoost models based on DSS provided the highest estimation accuracy (R² = 0.8232, RMSE = 5.2421, RRMSE = 20.5680, MAE = 4.0169 and Bias = 0.4493). The ANOVA of the prediction error indicated that the stratification method was a more important factor than the regression algorithm in forest AGC estimation. This study demonstrated the positive effect of stratification and how the combination of DSS and the CatBoost algorithm can effectively improve the estimation accuracy of forest AGC. Integrating this strategy with national forest inventory could help improve the monitoring of forest carbon stock over large areas.

Background: The carbon pools of forest soils play a vital role in global carbon sequestration and emissions. Forest management can regulate the sequestration and output of forest soil carbon pools to a certain extent; however, the kinetics of the effects of forest density on soil carbon pools require further investigation.Methods: We established sample plots with stand density gradients in three different agedLarix principis-rupprechtiiplantations and quantified the soil respiration, soil organic carbon (SOC), soil dissolved organic carbon (DOC), microbial biomass carbon (MBC), light fraction organic carbon (LFOC), and readily oxidized carbon (ROC).Results and conclusions: During the growth and development of plantations, stand density is an essential factor that impacts soil respiration and its associated elements. Moderate density was observed to promote both the soil and heterotrophic respiration rates and the sequestration of MBC and LFOC, whereas it inhibited the sequestration of ROC. The soil, heterotrophic, and autotrophic respiration rates of older forest stands were relatively rapid, whereas the contents of SOC, MBC, LFOC, DOC, and ROC were higher and more sensitive to changes in stand density. The MBC, LFOC, and ROC in soil labile organic carbon were closely related to both the soil and heterotrophic respiration, but not the SOC. Among them, the LFOC and MBC played the roles of “warehouse” and “tool” and were significantly correlated with soil and heterotrophic respiration. The ROC, as a “raw material”, exhibited a significantly negative correlation with the soil and heterotrophic respiration. When the soil and heterotrophic respiration rates were rapid, the ROC content in the soil maintained the low level of a “dynamically stabilized” state. The stand density regulated heterotrophic respiration by affecting the soil labile organic carbon, which provided an essential path for the stand density to regulate soil respiration.