The tree canopy represents a fundamental element of tree-related information. However, achieving precise canopy information from remote sensing images remains a significant challenge due to varying canopy sizes, mutual overlap, and diverse woodland environments. This study aims to leverage high -resolution Chinese fir images captured by an unmanned aerial vehicle (UAV) from a state forest farm in Jiangle County, Fujian Province, China. The images are integrated with the Mask R-CNN model to autonomously extract attributes at both individual and stand levels, facilitating precise forest mapping at the level of individual trees. The fusion entails a corresponding band saturation-weighted approach between RGB images and thermally-enhanced Canopy Height Model (CHM) images. The fusion threshold is set equal to the weight assigned to the CHM, and the weight of the RGB is computed as 1 minus the fusion threshold, ranging from 0 to 1 in intervals of 0.1. The dataset is then trained using three instance segmentation models: feature extraction networks based on ResNet50, ResNet101, and ResNeXt101, respectively. An instance merging approach based on the canopy crossoccupancy ratio is introduced, to enhance the accuracy of individual tree and stand-level attributes extraction, as well as forest mapping for individual tree canopies by using extensive stand images. This study focuses on evaluating the performance of instance segmentation models using two critical metrics: Bounding Box Average Precision (Box-AP) and Segmentation Average Precision (Segm-AP). The results highlighted that the Mask R-CNN model integrated with ResNeXt101, coupled with sample fusion using a threshold of 0.1, demonstrated exceptional performance. The accuracy of segmentation is acceptable, with Box-AP at 51.697 % and Segm-AP at 54.946 %. The extractions of individual level attributes, crown area, north - south crown width, and east - west crown width yielded R 2 of 0.933, 0.871, and 0.877, respectively. As for stand-level attributes, canopy density, and individual population extraction resulted in R 2 values of 0.901 and 0.912, respectively. Relative to the original images (with a fusion threshold of 0), segmentation accuracy was improved for all combinations, with the optimal configuration showing a 0.019 increase in R 2 for canopy density and a 0.014 increase in R 2 for individual population extraction. Furthermore, Box-AP and Segm-AP exhibit enhancements of 10.795 % and 10.746 %, respectively. The instances-merging method enhances the extraction accuracy of individual population by 5%. This study underscores the precision and efficacy of the Mask R-CNN instance-based segmentation model and fusion strategy, providing a robust support for the integration of deep learning in canopy extraction research. Its implications are of great significance for large-scale forestry surveys and the advancement of precision forestry, highlighting the substantial potential.

[目的]探索树高-胸径模型构建新方法,将分位数回归与非线性混合效应法相结合应用于树高-胸径模型构建,以此提高模型的拟合精度.[方法]利用2018年福建省将乐国有林场30 m×30 m固定样地1306株杉木的实测树高、胸径数据,从4个树高-胸径模型中筛选拟合效果最好的为基础模型,基于基础模型分别采用非线性混合效应、分位数回归以及非线性分位数混合效应构建树高-胸径模型.采用评价指标均方根误差(RMSE)、调整后决定系数(R2adj)和均方差(MSE),对各模型的拟合结果进行评价比较,采用赤池信息准则(AIC)、贝叶斯信息准则(BIC)以及对数似然函数值(Loglik)比较各最优模型的拟合精度和预测精度.[结果]根据评价指标对比显示,Logistic模型为基础模型.非线性混合效应模型的拟合效果最优(AIC为3953.986,BIC为3988.199,Loglik为-1969.993),非线性分位数混合效应模型(AIC为3979.418,BIC为4028.293,Loglik为-1979.709)次之.模型拟合效果排序为非线性混合效应模型>非线性分位数混合效应模型>基础模型>分位数回归模型.比较各模型的残差图可知各模型均不存在异方差现象,预测效果排序为非线性混合效应模型>非线性分位数混合效应模型>基础模型>分位数回归模型.[结论]本研究将分位数回归与非线性混合效应法相结合,该方法对分组数据结构中不同分位点个体间的差异与关联做出解释,提高了模型的稳定性以及拟合精度,将该方法应用到树高-胸径关系的研究上是一个可行的思路,为构建树高-胸径模型提供新方法.

Stand structural complexity and growth partitioning between different tree sizes influence trees competition and resource use efficiency, and thus are closely linked to stand productivity. The main aim of this study was to examine how the stand structure complexity (Gini coefficient Gini) and size-related partitioning of stand growth (growth dominance coefficient GDC and size-growth relationship SGR) temporally changed after different thinning intensities in Chinese fir (Cunninghamia lanceolata) forests, Southeastern China. We analyzed the data of 6 repeated measurements over a nearly 10-year period from a forest thinning experiment with progressively increasing intensity (0, 20%, 25%, 33% and 50% reduction of stand density by selection logging of releasing target trees) and assessed the effects of different thinning intensities using generalized mixed effect model. The thinning intensity of 0 was taken as a control group to identify the stand changes in unthinned stands over time. Size heterogeneity and the relative contribution of larger trees to total stand growth tended to increase as forest grew without thinning management. While the increases in Gini and SGR (excluding between 0 and 0.5) overall negatively influenced the average annual basal area increment of the remaining trees (BAI). Gini in thinned stands decreased during the short term (two years or more) after thinning and the heavy thinning had lower Gini than other treatments several years (eight years or less) after thinning. GDC reduced greater with increasing thinning intensity, and thinning had a significant negative effect on Gini and SGR and a positive effect on BAI. The thinning impacts on BAI increased with the thinning intensity, while the effect magnitudes of different thinning intensities on Gini and SGR were: T50 < T20 < T25 < T33. Overall, thinning increased the stand growth partitioning of small trees, contributing to a uniform stand structure. However, the changes in stand structural complexity and the partitioning of stand growth were related to both thinning intensity and post-thinning time, particularly to thinning intensity. The results provided some implications for designing the intensity and timing of thinning, as well as for determining the size classes of removed trees for different management goals.

Tree growth is driven by various factors, including site climate, stand and tree variables. A thorough understanding of the respective roles of the main drivers is imperative to project or promote forest productivity in future scenarios. However, the complex relationships among the drivers and tree growth remained insufficiently understood. Using the structural equation modeling (SEM) approach, we examined the multiple and interactive relationships between annual tree growth and climate, stand- (site quality and competition) and tree-level (tree age and tree size) variables based on data from Chinese fir (Cunninghamia lanceolata) plantations, Southeastern China. To simplify the SEM structures, randomforest (RF) was employed initially to screen for factors with greater impacts on tree growth. SEMs showed that annual tree growth largely depended on climate and site quality, while competition had a smaller relative effect of less than 8%. Climate was the primary driver. Different climate factors influenced tree growth directly, and also indirectly via tree size and site quality. Tree growth increments decreased with the increase of maximum daily temperature (especially in current-year August), while increased with relative humidity (especially in previous-year October) and potential evapotranspiration (especially in current-year May). Site index (SI) and tree size were both of higher importance in predicting the annual tree increments and had a positive impact on tree growth, contributing about 1/3 and 15 % to the total effect, respectively. In contrast, both the direct and indirect effects of stand competition were negative. Basal area of larger trees (BAL) appeared to better capture competition than stand density. This study suggested that SEM combined with RF provided a promising way to understand the complexity of tree growth. Larger trees in forests with higher site quality and lower density grow faster, however, their annual increments seem to decrease with increasing temperature.