Forest Structure Research Articles

Improving forest age prediction performance using ensemble learning algorithms base on satellite remote sensing data

Forest age plays a crucial role in assessing forest structure, carbon sinks, and other ecological functions. How to estimate forest age by satellite remote sensing data has been a hot research topic. This study focused on the forests of Zhejiang Province, utilizing Landsat 5 as the remote sensing data source to extract distribution information of broadleaf and coniferous forests. Then, the remote sensing feature variables were screened, and the age of broadleaf forests and coniferous forests was estimated by using the multiple linear regression model MLR, the machine learning model (K-nearest neighbor method regression model KNN, support vector regression model SVR), and the ensemble learning model (adaptive boosting model AdaBoost, random forest model RF, and eXtreme gradient boosting XGBoost). After analyzing the forest age estimation results from different models, the best-performing model was selected to create a spatial distribution map of forest age in Zhejiang Province. The study shows that the ensemble machine learning model can better realize the remote sensing inversion of forest age. The optimal model for broadleaf forests is the XGBoost model, with a coefficient of determination R2 of 0.832, a root mean square error (RMSE) of 5.823a and a relative root mean square error (rRMSE) of 21.009%. And the top model for coniferous forests is the RF model, with R2 of 0.800, RMSE of 5.076a and rRMSE of 19.782%. Compared with the MLR model, the best broadleaf and coniferous forest age inversion models improved the R2 by 75.120% and 82.500%, and reduced the RMSE by 52.674% and 47.540%, and reduced the rRMSE by 52.703% and 47.480% respectively. Additionally, the analysis revealed that 50% of the remote sensing feature variables involved in forest age inversion are texture features, indicating that texture is an important feature variable for the construction of forest age remote sensing inversion models.

A Semi-Automatic Approach for Tree Crown Competition Indices Assessment From UAV LiDAR

Understanding the spatial heterogeneity of forest structure is crucial for comprehending ecosystem dynamics and promoting sustainable forest management. Unmanned aerial vehicle (UAV) LiDAR technology provides a promising method to capture detailed three-dimensional (3D) information about forest canopies, aiding in management and silvicultural practices. This study investigates the heterogeneity of forest structure in broadleaf forests using UAV LiDAR data, with a particular focus on tree crown features and their different information content compared to diameters. We explored a non-conventionally used method that emphasizes crown competition by employing a nearest neighbor selection technique based on metrics derived from UAV point cloud profiles at the tree level, rather than traditional DBH (diameter at breast height) spatial arrangement. About 300 vegetation elements within 10 plots collected in a managed Beech forest were used as reference data. We demonstrate that crown-based approaches, which are feasible with UAV LiDAR data at a reasonable cost and time, significantly enhances the understanding of forest heterogeneity, adding new information content for managers. Our findings underscore the utility of UAV LiDAR in characterizing the complexity and variability of forest structure at high resolution, offering valuable insights for carbon accounting and sustainable forest management.

Estimating Brazilian Amazon Canopy Height Using Landsat Reflectance Products in a Random Forest Model with Lidar as Reference Data

Landsat data have been used to derive forest canopy structure, height, and volume using machine learning models, i.e., giving computers the ability to learn from data and make decisions and predictions without being explicitly programmed, with training data provided by ground measurement or airborne lidar. This study explored the potential use of Landsat reflectance and airborne lidar data as training data to estimate canopy heights in the Brazilian Amazon forest and examined the impacts of Landsat reflectance products at different process levels and sample spatial autocorrelation on random forest modeling. Specifically, this study assessed the accuracy of canopy height predictions from random forest regression models impacted by three different Landsat 8 reflectance product inputs (i.e., USGS level 1 top of atmosphere reflectance, USGS level 2 surface reflectance, and NASA nadir bidirectional reflectance distribution function (BRDF) adjusted reflectance (NBAR)), sample sizes, training/test split strategies, and geographic coordinates. In the establishment of random forest regression models, the dependent variable (i.e., the response variable) was the dominant canopy heights at a 90 m resolution derived from airborne lidar data, while the independent variables (i.e., the predictor variables) were the temporal metrics extracted from each Landsat reflectance product. The results indicated that the choice of Landsat reflectance products had an impact on model accuracy, with NBAR data yielding more trustful results than the other products despite having higher RMSE values. Training and test split strategy also affected the derived model accuracy metrics, with the random sample split (randomly distributed training and test samples) showing inflated accuracy compared to the spatial split (training and test samples spatially set apart). Such inflation was induced by the spatial autocorrelation that existed between training and test data in the random split. The inclusion of geographic coordinates as independent variables improved model accuracy in the random split strategy but not in the spatial split, where training and test samples had different geographic coordinate ranges. The study highlighted the importance of data processing levels and the training and test split methods in random forest modeling of canopy height.

The relationship between structure and ecosystem services of forest and grassland based on pattern analysis method: A case study of the Mongolian Plateau

Global warming has led to severe land desertification on the Mongolian plateau. It puts great environmental pressure on vegetation communities. This pressure leads to fragmentation of land use and landscape patterns, thus triggering changes in the spatial distribution patterns of vegetation. The spatial distribution pattern of vegetation is crucial for the performance of its ecosystem services. However, there is not enough research on the relationship between large-scale spatial distribution patterns of vegetation and ecosystem services. Therefore, this study is to construct an ecological spatial network on the Mongolian Plateau based on landscape ecology and complex network theory. Combining pattern analysis methods to analyze the network, we obtained the spatial and temporal trends of forest and grass spatial distribution patterns from 2000 to 2100, and explored the relationship between the topological properties of source patches and ecosystem services in different patterns. It was found that there are four basic patterns of spatial distribution of forest and grass in the Mongolian Plateau. The Core-Linked Ring pattern accounts for 40.74 % and exhibits the highest stability. Under the SSP5-RCP8.5 scenario, source patches are reduced by 22.76 % in 2100. Topological indicators of source patches showed significant correlations with ecosystem services. For example, the CUE of grassland patches in the Centralized Star pattern was positively correlated with betweeness centrality. The most significant improvement in WUE after optimization is 19.90 % compared to pre-optimization. The conclusion of the study shows that the spatial distribution pattern of vegetation can be used to enhance the stability of ecological spatial network and improve ecosystem services at a larger scale. It can provide a certain reference for the study of spatial patterns of vegetation distribution in arid and semi-arid areas.

A dwarf conifer tree from the Triassic of Antarctica: the first fossil evidence of suppressed growth in a favorable climate?

The complexity of fossil forest ecosystems is difficult to reconstruct due to the fragmentary nature of the fossil record. However, detailed morpho-anatomical studies of well-preserved individual fossils can provide key information on tree growth and ecology, including in biomes with no modern analog such as the lush forests that developed in the polar regions during past greenhouse climatic episodes. We describe an unusual-looking stem from Middle Triassic (ca 240 Ma) deposits of Antarctica with over 100 very narrow growth-rings and conspicuous persistent vascular traces through the wood. Sections of the specimen were prepared using the cellulose acetate peel technique to determine its systematic affinities and analyse its growth. The new fossil shows similarities with the form genus Woodworthia and with conifer stems from the Triassic of Antarctica, and is assigned to the conifers. Vascular traces are interpreted as those of small branches retained on the trunk. Growth-ring analyses reveal one of the slowest growth rates reported in the fossil record, with an average of 0.2mm/season. While the tree was growing within the Triassic polar circle, sedimentological data and growth-ring information from other fossil trees, including from the same locality, support the presence of favorable conditions in the region. The specimen is interpreted as a dwarf conifer tree that grew under a generally favorable regional climate but whose growth was suppressed due to stressful local site conditions. This is the first time that a tree with suppressed growth is identified as such in the fossil record, providing new insights on the structure of polar forests under greenhouse climates and, more generally, on the complexity of tree communities in deep time.

Multi-scale lidar measurements suggest miombo woodlands contain substantially more carbon than thought

Miombo woodlands are integral to livelihoods across southern Africa, biodiversity in the region, and the global carbon cycle, making accurate and precise monitoring of their state and change essential. Here, we assembled a terrestrial and airborne lidar dataset covering 50 kha of intact and degraded miombo woodlands, and generated aboveground biomass estimates with low uncertainty via direct 3D measurements of forest structure. We found 1.71 ± 0.09 TgC was stored in aboveground biomass across this landscape, between 1.5 and 2.2 times more than the 0.79–1.14 TgC estimated by conventional methods. This difference is in part owing to the systematic underestimation of large trees by allometry. If these results were extrapolated across Africa’s miombo woodlands, their carbon stock would potentially require an upward revision of approximately 3.7 PgC, implying we currently underestimate their carbon sequestration and emissions potential, and disincentivise their protection and restoration.

Correction: Hankin et al. Second-Entry Burns Reduce Mid-Canopy Fuels and Create Resilient Forest Structure in Yosemite National Park, California. Forests 2022, 13, 1512

In the original publication [...]

The Value of Forests to Pollinating Insects Varies with Forest Structure, Composition, and Age

The Value of Forests to Pollinating Insects Varies with Forest Structure, Composition, and Age

Temporal Variations in Aboveground Biomass, Nutrient Content, and Ecological Stoichiometry in Young and Middle-Aged Stands of Chinese Fir Forests.

Understanding the ecological dynamics of forest ecosystems, particularly the influence of forest age structure on soil carbon (C), nitrogen (N), and phosphorus (P) content, is crucial for effective forest management and conservation. This study aimed to investigate the nutrient storage and ecological stoichiometry across different-aged stands of Chinese fir forests. Soil samples were collected from various depths (0-15 cm, 15-30 cm, and 30-45 cm) across four age groups of Chinese fir forests (8-year-old, 12-year-old, 20-year-old, and 25-year-old) in the Forest Farm, Pingjiang County, China. Soil organic carbon (SOC), total nitrogen (TN), and total phosphorus (TP) were measured, and their stoichiometries were calculated. The results showed that both individual tree biomass and stand biomass, along with SOC, TN, and TP content, increased with stand age, highlighting the significant importance of stand age on biomass production and nutrient accumulation in forests. Specifically, soil C and P contents significantly increased as the forest aged, while variation in N content was relatively minor. Soil C/N and C/P ratios exhibited variation corresponding to forest age, suggesting alterations in the ecological stoichiometry characteristics of the forests over time. These findings are crucial for understanding the dynamics of ecosystem functioning and nutrient cycling within Chinese fir forests and provide a solid scientific basis for the effective management and conservation of these vital forest ecosystems.

Insight into the variation of soil hydraulic properties under beech and spruce forest—A case study in the forest of Tharandt, NE Germany

AbstractBackgroundThe increasing vulnerability of forests in the temperate zone due to climate change has led to modification in the forest structure to secure woody raw materials and ecosystem benefits. Such changes will influence hydrological processes both at the stand and catchment scale. Soil hydraulic properties (SHP) play an important role in assessing the water cycle in these ecosystems. Yet, knowledge regarding the effect of forest‐ and site‐specific conditions on SHP in temperate climates is scarce.AimsThis work addresses this research gap by assessing the variation of SHP under two common European forest stands, Fagus sylvatica and Picea abies (1) with comparable site conditions, and (2) across differing site conditions.MethodsWe determined the soil water retention curve (WRC) and the hydraulic conductivity curve (HCC) in several plots with the bimodal Kosugi–Mualem's hydraulic model. These functions were determined using combined field and laboratory measurements, including hydraulic conductivity and water content from soil samples.Results(1) We observed distinct variations in SHP between beech and spruce forest stands with comparable site conditions; however, no clear pattern in the variation was discernible. (2) A noticeable effect of the site‐specific characteristics on the SHP was detected. Moreover, SHP in each analysed forest type presented individual variations.ConclusionsThis study demonstrates that SHP present a wide range of variations in terms of both forest‐ and site‐specific conditions. Hence, due to its heterogeneity, we emphasise the need for more research to better characterise SHP in temperate zone forests. Moreover, this study underlines the urgent use of a minimum set of parameters in studies when addressing SHP (e.g., tree age, soil texture).

Fishers (Pekania pennanti) are forest structure specialists when resting and generalists when moving: behavior influences resource selection in a northern Rocky Mountain fisher population

BackgroundStudies of animal habitat selection are important to identify and preserve the resources species depend on, yet often little attention is paid to how habitat needs vary depending on behavioral state. Fishers (Pekania pennanti) are known to be dependent on large, mature trees for resting and denning, but less is known about their habitat use when foraging or moving within a home range.MethodsWe used GPS locations collected during the energetically costly pre-denning season from 12 female fishers to determine fisher habitat selection during two critical behavioral activities: foraging (moving) or resting, with a focus on response to forest structure related to past forest management actions since this is a primary driver of fisher habitat configuration. We characterized behavior based on high-resolution GPS and collar accelerometer data and modeled fisher selection for these two behaviors within a home range (third-order selection). Additionally, we investigated whether fisher use of elements of forest structure or other important environmental characteristics changed as their availability changed, i.e., a functional response, for each behavior type.ResultsWe found that fishers exhibited specialist selection when resting and generalist selection when moving, with resting habitat characterized by riparian drainages with dense canopy cover and moving habitat primarily influenced by the presence of mesic montane mixed conifer forest. Fishers were more tolerant of forest openings and other early succession elements when moving than resting.ConclusionsOur results emphasize the importance of considering the differing habitat needs of animals based on their movement behavior when performing habitat selection analyses. We found that resting fishers are more specialist in their habitat needs, while foraging fishers are more generalist and will tolerate greater forest heterogeneity from past disturbance.

Socioeconomic Disparities in Urban Forest Diversity and Structure in Green Areas of Santiago de Chile.

Urban trees enhance biodiversity, provide ecosystem services, and improve quality of life in cities. Despite their benefits, trees are not distributed equitably, and many cities exhibit a "luxury effect". Given the importance of public green space for providing access to urban tree benefits, we investigated the relationship between socioeconomic level and tree diversity and structure in 60 green areas in Santiago de Chile. Species richness and total tree abundance did not significantly vary among socioeconomic levels; however, a differential effect was found according to species origin. Introduced tree species exhibited similar abundance and species richness across socioeconomic levels, but native tree species were more abundant and richer in higher socioeconomic level areas compared to lower ones. Tree cover was higher in the high and medium socioeconomic level areas than in the low socioeconomic level area. A higher average DBH was found in the medium socioeconomic level area, which may be explained by older neighborhoods and a legacy of the luxury effect. Our findings reveal that socioeconomic groups are associated with differences in tree cover, width, and the number of native species in public green areas. Consequently, urban residents have different provisions of ecosystem services and opportunities to interact with natural heritage. Increasing the amount of tree cover and native species available to vulnerable groups will reduce disparities.

Beijing Forest Carbon Storage Potential Capacity

Forests serve as a crucial carbon reservoir. Therefore, optimizing forest carbon storage is a pathway towards achieving carbon neutrality. In this study, the Forest Simulation Optimization System (FSOS) was used to simulate the carbon storage in Beijing forests over 250 years (2018-2268). It was found that under the no management scenario, carbon storage fluctuates with the natural growth and death of trees, with peaks of more than 90 million tons. It proves that forests have a strong capacity of carbon storage. In the management scenario, harvest trees and make them into furniture, total carbon storage is high and maintains a stable level of 108 million tons. This is almost 1.6 times higher than in the no management scenario on average. In addition, the growth rate of carbon storage is fastest in the middle-aged forest and the near-mature forest. Therefore, in order to optimize the carbon sequestration benefits of forests, the forestry sector must pay attention to the age structure of forests in the future. Based on the results of this study, recommendations were made to optimize carbon storage in Beijing forests and to integrate forest managements of Beijing forests into regional economic and environmental planning.

Disturbance and vegetational structure in an urban forest of Indian Siwaliks: an ecological assessment.

Urban forests face multiple human-mediated pressures leading to compromised ecosystem structure and functioning. Therefore, understanding ecosystem structure in response to ongoing pressures is crucial for sustaining ecological integrity and human well-being. We aim to assess the disturbance and its effects on the vegetation structure of urban forests in Chandigarh using a combination of remote sensing techniques and vegetation surveys. The disturbance was evaluated as a change in NDVI (Normalised Difference Vegetation Index) from 2001 to 2021 by applying the BFAST (Breaks For Additive Season and Trend) algorithm to the MODIS satellite imagery data. A vegetation survey was conducted to compare the species composition, taxonomic and phylogenetic diversity as measures of forest vegetational structure. While signals of disturbance were evident, the changes in vegetation structure were not well established from our study. Further, this analysis indicated no significant differences in vegetation composition due to disturbance (F1,12 = 0.91, p = 0.575). However, the phylogenetic diversity was substantially lower for disturbed plots than undisturbed plots, though the taxonomic diversity was similar among the disturbed and undisturbed plots. Our results confirmed that disturbance effects are more prominent on the phylogenetic than taxonomic diversity. These findings can be considered early signals of disturbance and its impact on the vegetation structure of urban forests and contribute to the knowledge base on urban ecosystems. Our study has implications for facilitating evidence-based decision-making and the development of sustainable management strategies for urban forest ecosystems.

Human degradation of tropical moist forests is greater than previously estimated

Tropical forest degradation from selective logging, fire and edge effects is a major driver of carbon and biodiversity loss1–3, with annual rates comparable to those of deforestation4. However, its actual extent and long-term impacts remain uncertain at global tropical scale5. Here we quantify the magnitude and persistence of multiple types of degradation on forest structure by combining satellite remote sensing data on pantropical moist forest cover changes4 with estimates of canopy height and biomass from spaceborne6 light detection and ranging (LiDAR). We estimate that forest height decreases owing to selective logging and fire by 15% and 50%, respectively, with low rates of recovery even after 20 years. Agriculture and road expansion trigger a 20% to 30% reduction in canopy height and biomass at the forest edge, with persistent effects being measurable up to 1.5 km inside the forest. Edge effects encroach on 18% (approximately 206 Mha) of the remaining tropical moist forests, an area more than 200% larger than previously estimated7. Finally, degraded forests with more than 50% canopy loss are significantly more vulnerable to subsequent deforestation. Collectively, our findings call for greater efforts to prevent degradation and protect already degraded forests to meet the conservation pledges made at recent United Nations Climate Change and Biodiversity conferences.

Assessment of Mycological Possibility Using Machine Learning Models for Effective Inclusion in Sustainable Forest Management

The integral role of wild fungi in ecosystems, including provisioning, regulating, cultural, and supporting services, is well recognized. However, quantifying and predicting wild mushroom yields is challenging due to spatial and temporal variability. In Mediterranean forests, climate-change-induced droughts further impact mushroom production. Fungal fruiting is influenced by factors such as climate, soil, topography, and forest structure. This study aims to quantify and predict the mycological potential of Lactarius deliciosus in sustainably managed Mediterranean pine forests using machine learning models. We utilize a long-term dataset of Lactarius deliciosus yields from 17 Pinus pinaster plots in Soria, Spain, integrating forest-derived structural data, NASA Landsat mission vegetation indices, and climatic data. The resulting multisource database facilitates the creation of a two-stage ‘mycological exploitability’ index, crucial for incorporating anticipated mycological production into sustainable forest management, in line with what is usually done for other uses such as timber or game. Various Machine Learning (ML) techniques, such as classification trees, random forest, linear and radial support vector machine, and neural networks, were employed to construct models for classification and prediction. The sample was always divided into training and validation sets (70-30%), while the differences were found in terms of Overall Accuracy (OA). Neural networks, incorporating critical variables like climatic data (precipitation in January and humidity in November), remote sensing indices (Enhanced Vegetation Index, Green Normalization Difference Vegetation Index), and structural forest variables (mean height, site index and basal area), produced the most accurate and unbiased models (OAtraining = 0.8398; OAvalidation = 0.7190). This research emphasizes the importance of considering a diverse array of ecosystem variables for quantifying wild mushroom yields and underscores the pivotal role of Artificial Intelligence (AI) tools and remotely sensed observations in modeling non-wood forest products. Integrating such models into sustainable forest management plans is crucial for recognizing the ecosystem services provided by them.

Bumblebees mediate landscape effects on a forest herb's population genetic structure in European agricultural landscapes.

Spatially isolated plant populations in agricultural landscapes exhibit genetic responses not only to habitat fragmentation per se but also to the composition of the landscape matrix between habitat patches. These responses can only be understood by examining how the landscape matrix influences among-habitat movements of pollinators and seed vectors, which act as genetic linkers among populations. We studied the forest herb Polygonatum multiflorum and its associated pollinator and genetic linker, the bumblebee Bombus pascuorum, in three European agricultural landscapes. We aimed to identify which landscape features affect the movement activity of B. pascuorum between forest patches and to assess the relative importance of these features in explaining the forest herb's population genetic structure. We applied microsatellite markers to estimate the movement activity of the bumblebee as well as the population genetic structure of the forest herb. We modelled the movement activity as a function of various landscape metrics. Those metrics found to explain the movement activity best were then used to explain the population genetic structure of the forest herb. The bumblebee movement activity was affected by the cover of maize fields and semi-natural grasslands on a larger spatial scale and by landscape heterogeneity on a smaller spatial scale. For some measures of the forest herb's population genetic structure, that is, allelic richness, observed heterozygosity and the F-value, the combinations of landscape metrics, which explained the linker movement activity best, yielded lower AICc values than 95% of the models including all possible combinations of landscape metrics. Synthesis: The genetic linker, B. pascuorum, mediates landscape effects on the population genetic structure of the forest herb P. multiflorum. Our study indicates, that the movement of the genetic linker among forest patches, and thus the pollen driven gene flow of the herb, depends on the relative value of floral resources in the specific landscape setting. Noteworthy, the population genetic structure of the long-lived, clonal forest herb species correlated with recent land-use types such as maize, which have been existing for not more than a few decades within these landscapes. This underscores the short time in which land-use changes can influence the evolutionary potential of long-lived wild plants.

Modelling postfire recovery of snow albedo and forest structure to understand drivers of decades of reduced snow water storage and advanced snowmelt timing

AbstractForest fires darken snow albedo and degrade forest structure, ultimately reducing peak snow–water storage, and advancing snowmelt timing for up to 15 years following fire. To date, no volumetric estimates of watershed‐scale postfire effects on snow–water storage and snowmelt timing have been quantified over decades of postfire recovery. Using postfire parameterizations in a spatially‐distributed snow mass and energy balance model, SnowModel, we estimated postfire recovery of forest fire effects on snow–water equivalent (SWE) and snowmelt timing over decades following fire. Using this model, we quantified volumetric recovery of forest fire effects on snow hydrology across a chronosequence of eight sub‐alpine forests burned between 2000 and 2019 in the Triple Divide of western Wyoming. We found that immediately following fire, forest fire effects reduced snow–water storage by 6.8% (SD = 11.2%) and advanced the snow disappearance date by 31 days (SD = 9 days). Across the 15‐year recovery following fire, forest fire effects reduced snow–water storage by 4.5% (SD = 11.4%). Postfire effects on snow hydrology generally recovered over time, but still persisted beyond 15‐years following fire due to the observed postfire shift from forest to open meadow. Estimates of postfire reductions on peak SWE summed over the entire 15‐year postfire recovery period were 18 times greater than the immediate losses in the first winter following fire alone. These lasting effects of forest fires on snow hydrology decades following fire highlight the importance of postfire parameterizations for more accurate watershed‐scale volumetric estimates of forest fire effects on snow–water resources.

Landscape Ecological Structures and Patterns for Green Space Conservation in Forest Monasteries in Northeast Thailand

The green spaces in Wat Pah play an important role in forest conservation in Thailand. This study identified and analyzed the structures and patterns of landscape ecology in Wat Pah Nanachart, Ubon Ratchathani Province, to guide a conceptual framework for green space conservation in forest monasteries. Spatial analysis and modeling using geoinformatics technology were employed to recognize and characterize these landscapes. Information on the green space conservation and management of Wat Pah was also obtained using an in-depth interview and site observation. The results revealed that most of the green space in Wat Pah is forest, characterized by an ecological matrix and a dry evergreen forest. The forest structure can be divided into three canopies, dominated by the Dipterocarpaceae family, which has regenerated into upper and lower canopies. It is an edge matrix and ecological corridor connected to the internal and external green spaces of Wat Pah. It plays an important role in supporting cultural, religious, and aesthetic activities for ordained monks and Buddhists. The spatial landscape model can be divided into three main zones: Thoranisangha, Buddhawas, Sanghawas. Thoranisangha is an open space covered with traditional and artificially planted tree species. It is located in front of the temple and designed for public utilities in the context of managing and conserving the remaining trees. Buddhawas is a semi-open space mostly covered with traditional tree species. It represents Buddhism’s identity and uniqueness and is used for religious ceremonies and dissemination—listening to sermons, meditating, and praying. Meanwhile, Sanghawas is covered with natural forest and contains residences for monks that are designed to be in harmony with the forest ecosystem.

Att-Mask R-CNN: an individual tree crown instance segmentation method based on fused attention mechanism

Tree detection and canopy area measurement are important and difficult tasks in forest inventory, which are important for understanding forest stand structure. This study utilized remotely piloted aircraft (RPA) aerial photography technology to collect remote sensing images of forests in Xiong County, China, creating a dataset comprising 1200 images of six tree species. Based on this dataset, the paper proposes an optimized model, Att-Mask R-CNN, for canopy detection and segmentation. Att-Mask R-CNN outperforms the original models (Mask R-CNN and MS R-CNN) by achieving 65.29% mean average precision for detection, 80.44% mean intersection over union for segmentation, and 90.67% overall recognition rate for the six tree species. In addition, a pixel statistics method based on segmentation masks is introduced for estimating the vertical projected area of individual tree crowns, and comparisons between the measured and predicted vertical projected area of the crowns of six tree species (100 trees of each class) show an overall goodness-of-fit R2 of 85% and a relative root-mean-square error rRMSE of 12.81%. By using remote sensing images from RPAs and optimizing existing deep learning models, the detection and segmentation of individual tree canopies can be achieved, resulting in a more accurate understanding of forest structure, which provides scientific support for forest management and resource monitoring.