To compare the impact of spaceborne LiDAR footprint quality under different screening conditions on forest canopy height extraction, providing a reference for the precise estimation of forest structural parameters using spaceborne LiDAR data.

The research was conducted in the Harvard Forest, Worcester County, Massachusetts, USA. GEDI parameters including quality_flag, degrade_flag, sensitivity, solar_elevation, and full_power_beam were used as filtering conditions. Random forest models were built under different conditions to establish canopy height models. By analyzing the variations in model accuracy under different quality filtering conditions, the optimal screening method for spaceborne LiDAR footprint shots was determined.

1) Under single-criterion filtering, the degradation flag yielded the best performance. Compared to the unfiltered data, quality filtering improved the accuracy of forest canopy height retrieval, reducing the RMSE by 6.97%; 2) Multi-criteria combination filtering further enhanced the accuracy. The optimal combination-comprising the quality flag, degradation flag, sensitivity, solar elevation angle, and full-power beam-reduced the RMSE by 20.48% compared to the unfiltered data; 3) Some combined conditions (e.g., solar elevation angle with quality flag) resulted in decreased accuracy, indicating that not all indicators are necessary for filtering; 4) Footprint loss due to filtering mainly occurred in areas with canopy heights between 18 m and 36 m; 5) The combination of quality flag, degradation flag, and sensitivity outperformed the optimal combination in regional interpolation mapping, reducing the RMSE by 4.25% compared to the best combination.

Footprint quality filtering improves the estimation accuracy of forest canopy height, and combined filtering methods are more beneficial than single filtering methods in obtaining high-quality data. However, the screening strategy aimed at single-point accuracy during the modeling phase is not optimal for regional-scale spatial mapping tasks. By selecting appropriate filtering conditions, more accurate forest structural parameters can be obtained, providing reliable technical support for forest resource monitoring and management.

To carry out the research and application study of tree standing volume modeling based on non-destructive data collection by high-precision terrestrial laser scanning using Quercus species in Hunan Province as a pilot project.

143 single trees of Quercus species were scanned by Terrestrial Laser Scanning, the parameters of the trees were extracted from point cloud data, the model that meets the biological characteristics was selected as an alternative, and the Marquardt iterative method and the weighted least squares method were used to fit the model, and the optimal timber standing volume model was selected based on the results of the model fitting, the evaluation of the model and the test of the applicability of the model.

The R² of the traditional one-dimensional and two-dimensional lumber models constructed using point cloud extraction parameters was greater than 0.9, the remaining standard deviation was less than 0.1, the total relative error and average systematic error were less than 3%, and the prediction accuracy was over 95%. The average percent standard error of the two-dimensional lumber model was less than 7%, and that of the one-dimensional lumber model was less than 16%. By analyzing the relationship between the volume of individual tree segments and the overall volume, it is proposed to construct a standing volume model by combining the diameter with the height at the position of 0.276 relative to the height of the tree, and the model constructed has the highest accuracy among all the standing volume models, with an R² of 0.992 and a residual standard deviation of less than 0.03, which is superior to the traditional standing volume model, and enhances the accuracy of the standing volume prediction. And there was no significant difference between the estimated and measured wood volume of the wood standing volume model constructed using Terrestrial Laser Scanning point cloud data.

The forest parameters extracted from the point cloud data have high accuracy, and can replace the measured forest parameters for the related modeling; the accuracy of the constructed timber standing volume model can meet the needs of the daily forestry survey in Hunan Province, and provide a non-destructive technical means for the future modeling; the proposed new parameters can improve the accuracy of the timber standing volume modeling, and provide a new way of thinking for the construction of timber standing volume model.

In response to the current situation where numerous and diverse algorithms are used for forest volume inversion from airborne point cloud data, this study aims to conduct a comparative analysis of different feature selection methods combined with various algorithms to identify the optimal model, providing a reference for airborne LiDAR-based forest volume inversion.

The research was conducted in the Wangyedian forest farm. Plot-level volume was calculated based on field measurement data for individual trees, combined with point cloud height features extracted from airborne point cloud data. Stepwise regression (SR) and the Boruta algorithm were used for feature selection. Six non-parametric models: RF, KRR, XGBoost, KNN, MLP and SVM were constructed. The best model was determined based on accuracy evaluation and residual analysis, leading to the completion of forest volume mapping for the study area.

Compared to the stepwise regression selection method, the Boruta method selects more effective feature variables, making it more suitable for forest volume modeling. The model's average R² improves from 0.73 to 0.76, RMSE decreases from 38.94 to 35.47 m³·hm^-2, rRMSE drops from 29.68 to 26.38 m³·hm^-2, MAE reduces from 20.26% to 18.22%, and SMAPE decreases from 16.26% to 14.31%. XGBoost, combined with Boruta feature selection, provides the optimal model with the highest inversion accuracy, achieving an R² of 0.92, RMSE of 20.02 m³·hm^-2, MAE of 16.62 m³·hm^-2, rRMSE of 10.29%, and SMAPE of 10.58%. Residual analysis shows that the model's residual distribution is reasonable and significantly different from other models.

Airborne LiDAR technology effectively collects 3D forest information and is suitable for forest volume inversion. When combined with Boruta feature selection and the XGBoost non-parametric model, it can efficiently invert the spatial distribution of forest volume.

Normalized difference vegetation index (NDVI) can indicate the growth and coverage of vegetation, and explore the spatio-temporal dynamic change characteristics of long-time series, which is of great significance for clarifying the ecological change of regional vegetation.

Google Earth Engine (GEE) cloud platform was used to obtain the long-term sequence Landsat images of vegetation growing seasons (July-September) and construct NDVI indexes. Through the variation coefficient, Sen+Mann-Kendall trend analysis, future trend change analysis and spatial autocorrelation calculation, the temporal and spatial variation characteristics and spatial pattern of NDVI in Shenmu city from 2000 to 2020 were analyzed, and the driving force analysis was carried out by extracting different aggregation areas combined with the data of cumulative precipitation and mean temperature in vegetation growing seasons.

The NDVI in Shenmu increased significantly from 2000 to 2020, with a growth rate of 1.25%·a^-1. The areas with increased, stable and degraded NDVI accounted for 97.8%, 0.4% and 1.8% of the total area, respectively. The degraded areas were mainly distributed in the urban agglomeration areas. The NDVI variation coefficient was mainly concentrated in the range of 0.3-0.5, and the overall fluctuation was intense. The variation degree showed a spatial pattern of “high in the southeast and low in the northwest”. The average Hurst index of NDVI in the whole region was 0.69, and 93.7% of the area continuously increased significantly, indicating that the vegetation change in the future was good. The global Moran' I index decreased in volatility, and the global spatial autocorrelation of NDVI decreased gradually. The fluctuation of the aggregation area with low-low NDVI decreased, which was significantly negatively correlated with the cumulative precipitation in vegetation growing seasons (P < 0.05). The area of low-high type aggregation area was significantly negatively correlated with the average temperature in vegetation growing seasons (P < 0.01). The fluctuation of high-high type aggregation area increased, which was significantly positively correlated with the cumulative precipitation in vegetation growing seasons (P < 0.01).

In the past 21 years, the overall change of NDVI in Shenmu city was more dramatic, showing an increasing trend, and its spatial distribution tended to be fragmented. Precipitation and warming had a certain role in promoting vegetation restoration, and precipitation had a greater impact on vegetation restoration. In the future, the change of vegetation shows a significant increase in sustainability.

Forest parameters are the basic indicators for estimating forest volume and forest biomass, and the traditional manual survey method is time-consuming and laborious, and it is difficult to adapt to the requirements of digital forest resources monitoring technology under the new situation. Terrestrial LiDAR scanning technology can obtain small-scale and high-resolution internal structure information of forest stands, which provides a new idea for extracting diameter at breast height and tree height under stand environmental conditions. The existing terrestrial LiDAR scanning studies mostly focus on the extraction methods of forest parameters, but pay less attention to the combination of scanning resolution and quality, and the simplification of laser point clouds.

Taking the Chinese fir forest of Lutou Experimental Forest Farm as the research object, this paper designed seven different scanning combinations for the FARO Focus 3D X330 3D laser scanner to scan the sample plot, and proposed the idea of quadrant corner point cloud simplification for parameter extraction and accuracy evaluation, and then explored the influence of different scanning combinations on the accuracy and efficiency of forest diameter at breast height and tree height parameter extraction.

1) When the resolution was 1/2 and the quality was 4X, the accuracy of the diameter at breast height parameter extraction was the highest; and when the resolution was 1/4 and the quality was 4X, the extraction accuracy of tree height parameters was the highest. 2) Under the condition that there was no significant difference in forest extraction parameters, the scanning parameters with a quality of 4X work the most efficiently when the resolution was 1/4. 3) Select scan results with a resolution of 1/4 and a quality of 4X, which took accuracy and efficiency into account, and simplified the quadrant corner point cloud. The simplified point cloud accurately extracted the parameters of the diameter at breast height.

The results of this study have important reference value for the selection of scanning parameters and point cloud simplification methods for forest land with the same or similar geographical conditions and tree species, which can improve the efficiency of the data processing, and also provide a method and technical reference for the field sample survey by terrestrial LiDAR.

In order to quickly and accurately monitor the forest disturbance in the core area of the urban agglomeration, this study uses Landsat time series data to monitor the forest disturbance in Google Earth Engine (GEE) using LandTrend algorithm.

Based on the Landsat time series data of the vegetation growth season from 2000 to 2020, under the threshold restrictions of NDVI, NBR, and TCA indices, the LandTrender algorithm was used to process the time series trajectory to obtain information on the occurrence area and scale of forest disturbance in the core area of Changzhutan urban agglomeration. Combining continuous forest resource inventory data and Google Earth historical images, the accuracy evaluation and validation of forest disturbance monitoring results based on different indices were conducted, and the disturbance information was analyzed using forest cover data from land use data in the study area.

From 2000 to 2020, the total disturbed area of the forest in the core area of Changzhutan urban agglomeration was 264.35 km², with an average annual disturbed area of 13.22 km². Among them, the least disturbed area was 1.91 km² in 2002, and the largest disturbed area was 25.52 km² in 2011. The accuracy of forest disturbance information results under the three index thresholds of NBR, NDVI, and TCA under continuous forest resource inventory data was 90.91%, 81.72%, and 65.08%, respectively; In Google Earth’s historical images, the accuracy of random point distribution was 86.00%, 77.88% and 78.36%, respectively. There were significant disturbances in the forest from 2009 to 2011 to 2013. After processing and analyzing land use data nationwide, it could be seen that the forest area gradually decreased over time over the past 20 years, with the percentage of the total area decreasing from 27.10% to around 22.00%.

NBR is the most suitable forest disturbance index for the core area of Changzhutan urban agglomeration. The results of forest disturbance detection are consistent with the results of field survey and visual interpretation of historical images of Google Earth. The distribution boundary of forest disturbance patches can be fully proposed, and the trend of disturbance area change is the same as that of forest area change in national regional land use data, mainly caused by urbanization process, forest fires, and artificial logging. In the past 20 years, the area of forest disturbance has shown a fluctuating trend, the disturbance area fluctuated greatly in 2001, 2011 and 2013.