【Objective】 Using various, nonlinear machine learning algorithms, different volume models were constructed and compared to provide a theoretical basis for the accurate prediction of the volume of Pinus sylvestris var. mongolica.【Method】 A total of 184 felled Pinus sylvestris var. mongolica trees in the Tuqiang Forestry Bureau of the Greater Khingan Mountains were used to establish a nonlinear binary volume model (NLR). Three optimal machine learning algorithms were obtained using the K-fold cross test and OOB error test, including back propagation neural network (BP), ε-support vector regression (ε-SVR), and random forest (RF). An optimal volume model was obtained by comparing and analyzing the differences between the different models. 【Result】 The results showed that the machine learning algorithm was superior to the traditional binary volume model in the fitting and prediction of standing volume, and the specific order was RF > BP > ε-SVR > NLR. Compared with the traditional model, the R2 of RF increased by 2.00%; the RMSE, RMSE% and MAE decreased by 22.95%, 22.93% and 36.34%, respectively; and the absolute value of MRB was lower than the real value, which proved the superiority of RF in volume prediction. 【Conclusion】 Machine learning algorithms can effectively improve the accuracy at which standing volume can be predicted, providing a new solution for the accurate investigation and management of forest resources.
【Objective】 To accurately predict growth and formulate forest management strategies for Cunninghamia lanceolata in Hunan Province, a mixed-effects individual tree diameter increment model for Cunninghamia lanceolata was developed considering climatic factors. 【Method】 Based on the data of 3 638 observations in 73 plots from the 7th and 8th Chinese National Forest Inventory in Hunan Province, this study used the multiple stepwise regression method to introduce tree size, competition, site conditions, other stand variables, and climate factors as independent variables, and developed and evaluated four different dependent variables: i.e. 5-year diameter increment (D2-D1), the natural logarithm of 5-year diameter increment [ln(D22-D21+1)], the natural logarithm of 5-year squared diameter increment [ln(D22-D21+1)], and 5-year squared diameter increment (D22-D21). An optimal basic model was selected. A linear mixed-effects model with sample plots as random effects was then fitted. In addition, three commonly used variance functions and correlation structures were introduced to remove the heteroscedasticity of the residuals and autocorrelation. Finally, the 10-fold cross-validation method was used to assess predictive ability. 【Result】 Compared with the other three dependent variables, the model performed best with ln(D22-D21+1) as the dependent variable. Therefore, the model in which the dependent variable was ln(D22-D21+1) was selected as the optimal basic model. According to the results of the optimal basic model, the initial diameter, the ratio of the sum of the basal area of trees with diameters larger than the subject tree’s diameter to the initial diameter, stand basal area per hectare, the product of the sine of the slope and the natural logarithm of the altitude, mean annual precipitation, and mean minimum temperature in January significantly affected the increase in the diamteter of Cunninghamia lanceolata. Compared with the optimal basic model, the mixed-effects model showed a significantly improved prediction accuracy. Additionally, the introduction of variance functions and correlation structures also significantly improved the model’s performance, of which the exponent function (exponent) and ARMA(1,1) performed the best. In the 10-fold cross-validation, the mixed-effects model also showed better performance. 【Conclusion】 Climatic factors have a significant effect on the increase of diameter in Cunninghamia lanceolata. Compared with the basic model, the linear mixed-effects model with sample plots as random effects could greatly improve the model’s performance, and we hope that the model could provide support for the scientific management of Cunninghamia lanceolata in Hunan Province.
【Objective】 Structural equation modelling (SEM) was used to determine the effects of climate, soil, and altitude on growth indicators and pathway relationships in Xing’an larch (Larix gmelinii) forests. 【Method】 The annual mean temperature, annual mean precipitation, solar radiation, soil total nitrogen content, soil organic carbon density, and altitude were selected as influencing factors to explore the relationships between aboveground biomass, underground biomass, and tree height and these underlying factors. A structural equation model of climate, soil, and altitude was constructed using AMOS 21.0 software to measure the growth of Larix gmelinii stand. 【Result】 The aboveground and underground biomass of Larix gmelinii first increased and then decreased with an increase in altitude and annual mean precipitation, and the tree height increased with increasing altitude. The aboveground and underground biomass increased with an increase in soil organic carbon density. The total effect coefficient of altitude on the growth of Larix gmelinii was positive (0.200), and the direct effect (0.224) of altitude on the growth of Larix gmelinii was greater than the indirect effect (-0.024). The total effect coefficient of the climatic factors on the growth of Larix gmelinii was negative, at -0.771. The total influence coefficient of soil factors on the growth of Larix gmelinii was -0.216, which means these factors can slightly inhibit the growth of Larix gmelinii. 【Conclusion】 According to the path coefficient of the structural equation model, the absolute value of the total influence coefficient of climate factors was the largest, followed by that of soil and altitude. The static growth of Larix gmelinii forest is mainly restricted by climatic factors, which has guiding significance for predicting and evaluating changes in forest growth at high latitudes under the condition of global climate change.
【Objective】 The aim of this research was to explore the accuracy of extracting the maximum crown radius (CR, RC) at different crown depths of Pinus koraiensis based on Terrestrial Laser Scanning (TLS) point cloud data and developing a crown profile model based on TLS point cloud data. This study provides a practical basis for researching tree crown structures based on the TLS point cloud data. 【Method】 The TLS point cloud data and the branch factor data measured were from 30 analytic Pinus koraiensis. The CR at different heights was extracted by layered projection of the point cloud data and then compared with field-measured data of 30 analytic P. koraiensis trees for precision analysis. Finally, the RC extracted from the point cloud data was used to develop a crown profile model of P. koraiensis. 【Result】The total extraction accuracy of the maximum RC was 86.17%. There were differences in the accuracy of radius extraction for different crown depths. The range of relative crown depths with the best extraction effect was 0.15-1.00, and the accuracy of RC extraction within this range was approximately 90%. The range of relative crown depth with the worst extraction effect was 0-0.15, and the extraction accuracy within this range was 60.27%-75.79%. The three selected equations (mischerlich equation, quadratic parabolic equation, 3-parameter Weibull function) had good fitting effects and the 3-parameter Weibull function was the optimal model. We re-parameterized the optimal model and added DBH and ratio of height to diameter (HD) to the model, and the goodness-of-fit of the model was significantly improved. 【Conclusion】 Based on the TLS point cloud data, the accuracy of the maximum RC extraction using the point cloud layered projection method satisfies the requirements of model development. Therefore, TLS data can be used to approximately replace field-measured data in crown profile model research.
【Objective】 Canopy closure is important for forest management planning. To find methods and models for estimating canopy closure that are less affected by the region, with higher accuracy and better robustness, the study adopted a 4-Scale geometric-optical model to estimate the canopy closure of plantations. 【Method】 Wangyedian Forest Farm in Inner Mongolia and Gaofeng Forest Farm in Guangxi Autonomons Region were selected as the study areas. First, a parameter sensitivity analysis of the 4-Scale model was performed, and the canopy gap fraction Pvg_c (the gap fraction when the canopy was a rigid body) and Pvg (the gap fraction of the gaps in a crown were considered) under different sensitivity parameters were simulated. Then, a one-to-one correspondence database of Pvg_c, Pvg and the sensitivity parameters was established. Second, statistical relationship models between Pvg_c and Pvg and the sensitivity parameters were established based on the database. Then, Pvg_c and Pvg were estimated based on the sensitivity parameters, and the stand canopy closure was estimated. Finally, the canopy closure measured by the transects and the fish eye camera measurement method were used to test the canopy closure based on Pvg_c and Pvg respectively. 【Result】 The accuracy of plantation canopy closure estimated by Pvg_c and Pvg was 88.17% and 92.8%, respectively. Pvg_c had a higher correlation with the number of trees and crown radius. The R2 and RMSE values of the model are 0.814 and 0.043, respectively. The correlation between Pvg and LAI was high, and the R2 and RMSE of the model were 0.795 and 0.040, respectively. 【Conclusion】 Both Pvg_c and Pvg can be used to estimate canopy closure in plantations. Although Pvg has higher accuracy in estimating canopy closure, the estimated canopy closure is not the canopy closure defined in forestry. Canopy closure is defined as the ratio of the vertical projection area of the canopy when it is a rigid body. Therefore, the use of Pvg_c to estimate the canopy closure of plantations was more accurate. To obtain the number of trees and radius of the crown, Pvg_c was used to estimate the canopy closure of the plantation.
