An algorithm-based approach for tree selection in continuous cover forestry

• Machine learning models were evaluated for predicting biomass in planted forests. • Biomass prediction improved by integrating diverse biotic and abiotic variables. • Key biotic and abiotic drivers of forest biomass were identified. • The support vector machine model achieved very high predictive accuracy (R² = 0.98), outperforming the random forest model. Forest biomass is a critical parameter for assessing the condition of forest ecosystems. It plays a vital role in determining carbon sequestration and serves as an important indicator for evaluating forest health. In this study, we predicted and estimated forest biomass in the Guilan forests of northern Iran using multiple linear regression models, as well as two artificial intelligence models: Random Forest (RF) and Support Vector Machine (SVM). We collected research data through field observations, measurements, and standardized sampling protocols for forest inventory. To ensure the experimental nature of the research was upheld, we followed technical guidelines and applied various statistical techniques. Data collection and sampling occurred in the field, while soil tests were conducted in the laboratory to guarantee data accuracy. We systematically established 32 square fixed-area plots, each covering 0.04 hectares, arranged on a rectangular grid of 50 × 50 meters across the Radar Poshteh section in Siahkal County. In our modeling, we incorporated a variety of biotic and abiotic variables, including tree volume, basal area (BA), and the physical and chemical properties of the soil. The results indicated that the artificial intelligence models predicted forest biomass with greater accuracy and precision than the multiple linear regression model. Among these models, the SVM demonstrated a significantly higher coefficient of determination (R²) value of 0.98, compared to the RF model (R² = 0.49) and the multiple linear regression model (R² = 0.35). Additionally, a sensitivity analysis revealed that the most influential variables affecting forest biomass in the study area were tree volume and basal area. Among the abiotic factors, the percentage of soil organic matter and clay content were also significant. This study highlights that artificial intelligence techniques can greatly enhance the accuracy of biomass estimation.

It is often assumed that more vegetation increases residential landscape satisfaction. We challenge this assumption and conjecture that contested land used, such as plantation forestry, can show negative impacts. To test our hypothesis, we assess in which ways do eucalyptus plantations influence households’ landscape satisfaction in Galicia (NW Spain), a hotspot of ‘eucalyptisation’. We predicate a series of beta-binomial mixed models upon household microdata on residential landscape satisfaction combined with municipal forest inventories. Our models stratify the sample according to quartiles of plantation cover and control for settlement type, age, education and income. Results show a non-linear and context-dependent impact of eucalyptus on landscape satisfaction. Satisfaction declines where eucalyptus is marginal and again where plantations dominate. Interpreted through a Plantationocene lens, these findings indicate that preferences reflect sociodemographic position and historically mediated landscape meanings. Regular monitoring of residential landscape satisfaction would provide planners with citizen-informed evidence on landscape change.

ABSTRACT The uneven quality of Airborne Laser Scanning (ALS) point clouds easily leads to feature loss during the downsampling process of deep learning. This study addresses the problem of non-uniform quality in point clouds by proposing four lightweight, tree-based enhancement strategies: jitter, nearest neighbour interpolation, internal void filling, density, and random. Based on real tree species data, the processing time was recorded and compared with Test-Time Augmentation (TTA) strategies, and systematic experiments were conducted using five deep learning models. Combined with standard testing and cross-validation approaches, the effects of different enhancement methods on distribution restoration capability, tree species classification performance, feature preservation ability, and generalization ability were quantitatively analysed. Among them, nearest neighbour interpolation enhancement performed best, achieving the best F 1 − score of 0.790 with a maximum relative improvement of 29.8%, while preserving tree structural features and showing the highest cross-model consistency. The distorted point clouds generated by random enhancement did not exhibit the expected performance collapse, indicating that the models possess a certain degree of tolerance to structurally heterogeneous point clouds, while further highlighting the importance of input quality for tree point cloud classification. This provides an effective improvement scheme for the lightweight processing of forestry point clouds, noise-resistant modelling of tree point clouds, and forest inventory.

ABSTRACT Cordyla pinnata is a multipurpose forage species that plays important ecological and socio-economic roles in sub-Saharan Africa. We investigated how different land use types affect the species’ population structure and distribution in Senegal. Forest inventories were conducted in 140 plots across four land use types: farmland, fallow, rangeland, and protected area. Diameter at breast height (dbh), crown diameter, height, and regeneration of C. pinnata were measured, and woody species composition was noted. The results showed significant differences in species composition across land use types, with protected areas and rangelands showing the highest species diversity. The density of mature C. pinnata trees was highest in protected areas (393.63 ± 24.21 stems/ha) and lowest in farmlands (140.40 ± 46.70 stems/ha), suggesting that agricultural expansion has a strong negative impact on the species. Diameter class distributions suggested a stable population structure of mature trees, whereas regeneration patterns revealed an unstable state, raising concerns about the species’ future sustainability. Our findings underscore the adverse effects of land use changes on C. pinnata and highlight the need to mitigate human pressure on ecosystems to ensure conservation of this valuable species and the biodiversity it supports.

Carbon stocks and stock changes in harvested wood products (HWPs) are an important part of land sector greenhouse gas (GHG) estimation and reporting. HWPs broadly categorized as products in-use (e.g., solid wood and paper products) and in solid waste disposal sites (SWDS; e.g., landfills), store carbon transferred from harvested trees. In the United States (US), estimates of carbon in HWPs have historically been reported in the US GHG Inventory and included in submissions to the United Nations Framework Convention on Climate Change. These data have been obtained from national and international statistics on production and consumption of forest products and incorporated into a compilation system to estimate carbon in products in-use and in SWDS. In contrast, estimates of carbon in forest ecosystems have been obtained from nationwide forest inventory (NFI) data collected and maintained by the US Forest Service, Forest Inventory and Analysis (FIA) program. Here we describe a case study for the northern Lake States region of the US (Michigan, Minnesota, Wisconsin) where harvest data from the FIA program were integrated into HWP compilation systems. This advance improves consistency and continuity with forest ecosystem from NFI plots with estimates of HWPs. Over the 1900-2024 time period, total estimated net accumulation (i.e., balance of additions from transfers of harvested wood from forest ecosystems and losses from decay of wood harvested in the past) of carbon stored in products in-use was 277.0 ± 17.5 Million Metric Tons (MMT) Carbon (C) and in SWDS was 155.2 ± 9.8 MMT C. We estimate that HWPs from the region represent a carbon sink of 4.9 ± 0.1 MMT C in 2024. These estimates include HWPs produced in the region and exported domestically or internationally, as well as any HWPs produced and retained in the region, but not imports. The proposed methodology enables disaggregation with coarse national and state-level FIA data, and allows for integration of more specific, entity-level data to improve precision and reduce uncertainty in HWPs estimates in the US and improves consistency and continuity with forest ecosystem estimates across spatial and temporal scales.

This exploratory study presents an objective and consistent approach for assessing forest structure across multiple European countries, focusing on the distributions of tree species and tree diameter at breast height (DBH) as assessed by European National Forest Inventories (NFIs) and one landscape inventory. We distinguish six structural classes, ranging from mono-specific plots with a narrow (regular) DBH distribution to multi-species plots with a wide (irregular) DBH distribution. We used tree level data on basal area, species, and diameter, from 18 countries, and harmonised the data as much as possible by adopting a common diameter measurement threshold and by scaling the different plot radii to one ha. Data from 255,418 inventory plots indicate that roughly half of the forests are dominated by a single-species, while the other half consists of multiple-species. According to our approach, the predominant structural type in the surveyed countries is characterized by single-species dominance (56%) and a narrow DBH distribution. The relatively small plot radii across inventories and the diameter threshold of 10 cm also contribute to dominance of this structural type. The single-species regular class was the most prevalent ranging from 35.8% in Switzerland to 79.7% in Spain. The second most important was the multiple-species regular class, present on 37.7% of the forest area. Although the plots are not weighed to the full forest area, these results indicate a regular forest structure on 94% of Europe’s forests. The distribution of forest area per country over the categories varied only moderately. A shortcoming of a groundbased study is the challenge of harmonisation due to the different plot design of NFIs, showing a range in the plot radii from 9 to 25 meters hampering the comparison between countries. The results as presented at 0.2 degrees resolution (approximately 20 x 20 km) provide insight into forest structure in a consistent manner and can be updated in the future based on new releases of forest inventories. Although we did not study the effect of forest management on the current structure, these results are a basis to report temporal and spatial effects of management changes at this semi-high resolution, highly relevant to the EU Nature Restoration Law. We see this spatially explicit result as very promising, with advantages compared to the alternative of highly aggregated international statistics..

A practical workflow for applying validation strategies in tree height\u2013diameter modelling\u2606

Although the reduction of fossil fuel emissions remains of the utmost importance to mitigate climate change, maintaining and enhancing carbon sinks in forests have been widely promoted as nature-based climate solutions1-4. However, disturbances that could result in losses of forest carbon stocks are poorly accounted for when estimating the potential role of forests in climate mitigation5-7. This makes it difficult to appropriately size 'buffer pools': a mechanism designed to compensate for unintended carbon losses in carbon crediting projects8,9. Here we use forest inventory, satellite data, disturbance modelling and machine learning to map reversal (carbon loss) risk in the contiguous United States (CONUS) from natural disturbance. Across CONUS forests, we show that climate change increases the 100-year risk of carbon losses from natural disturbance, particularly in California and the Intermountain West. The current buffer pool of the largest CONUS forest climate mitigation programme is likely too small by an average factor of 6.3, and this could range from 2.2- to 8.0-fold too small when considering uncertainties around future climate scenarios, disturbance severity and other carbon pools. We provide spatially explicit maps of the long-term risks to forest carbon losses from natural disturbances, which highlight that current methodologies used for constructing carbon offset buffer pools require revisions to succeed under climate change.

Mesoherbivores are expanding globally through both native population irruptions and alien introductions, yet their broad-scale ecological impacts remain poorly resolved. We assessed how native white-tailed deer (Odocoileus virginianus) and alien wild pigs (Sus scrofa) influence forest understoreys across the eastern United States by integrating forest inventory and analysis plots, large-scale camera-trap monitoring and environmental data. Deer generally reduced native seedling abundance, although this effect weakened in warmer-wetter or more human-dense environments. Conversely, deer tended to increase invasive plant abundance and richness, consistent with selective browsing on palatable native species combined with the resistance of many common invaders. In contrast, wild pigs typically suppressed invasive plant abundance and richness, while their effects on native seedlings were neutral and strongly context dependent, consistent with their rooting behaviour and broad diet. These contrasting outcomes persisted after accounting for climate, human pressure and forest structure, supporting an important role of mesoherbivores in structuring understorey communities. Collectively, our findings indicate that native and alien mesoherbivores exert divergent and environmentally contingent effects on forest regeneration and plant invasion. Recognizing species identity, functional traits and environmental context will be essential for anticipating mesoherbivore impacts and managing forest biodiversity under global change.

ABSTRACT Remotely sensed data are increasingly used with model-based inference to estimate forest population characteristics (e.g., areal means). However, at-sensor radiance affected by measurement errors has inherent variations over time, space and spectrum, violating the stationary data assumption in regression-based applications. These errors bias relationships with forest variables by attenuating regression coefficients towards zero; heteroscedasticity, often observed in models of forest variables, could aggravate the biased coefficients. We proposed SIMEX-WLS, an errors-in-variables (EIV) modeling approach that corrects coefficient attenuation by incorporating both measurement errors and non-constant residual variances, to rectify the downweighed contribution of remotely sensed data. Validation using Landsat 8 Collection 2 Level-2 data yielded four relevant conclusions: (1) SIMEX-WLS also corrected the attenuated variances of regression coefficients along with the coefficients themselves; (2) in model-based inference, the attenuation correction achieved an 11% reduction in the variance of population mean estimates; (3) pixel-level RMSE may not be an appropriate metric for evaluating the precision of population characteristics; and (4) across samples of varied sizes, attenuation correction consistently improved the estimation precision. These findings explicitly emphasize the significance of incorporating remote sensing measurement errors and validating models for population-level estimation rather than relying solely on pixel-level prediction or mapping.

Revised estimates of forest carbon sequestration reveal the true sink capacity of Japanese forests.

The estimation of total tree height is a fundamental step in forest inventories, being essential for quantifying timber stock and supporting management planning in planted forests. However, direct measurement of this variable requires considerable operational effort and is usually performed only on subsamples, making it necessary to use hypsometric models to estimate the height of remaining trees. In this context, the present study aimed to configure, train, and validate regression models and machine learning algorithms to estimate the total height of individual trees in Eucalyptus spp. plantations located in Lavras do Sul, Rio Grande do Sul, Brazil. Two areas with different ages (12 and 15 years) were analyzed, totaling 395 trees measured for diameter at breast height (DBH) and total height. Logarithmic and N¨aslund models were fitted, along with Random Forest and Artificial Neural Network algorithms, using data partitioning into training (80%) and validation (20%) datasets. Model performance was evaluated using root mean square error (RMSE), percentage bias (BIAS), and coefficient of determination (R²). For the overall dataset, the logarithmic model showed the best predictive performance. In the 12-year-old stand, the N¨aslund model presented better results, whereas in the 15-year-old stand, the Artificial Neural Network achieved superior performance. Overall, all tested techniques demonstrated satisfactory performance for total height estimation, indicating that regression methods remain efficient and operationally simpler, while machine learning algorithms tend to provide advantages for larger and more complex datasets. The integration of traditional statistical approaches and artificial intelligence represents a promising alternative for advancing forest modeling.

ABSTRACT Forestry operations planning has typically relied on forest inventories, which require manual field measurements, a process that is often costly in terms of money and labor. Advancements in remote sensing technologies have demonstrated the potential to enhance both the accuracy and efficiency of forestry operations planning, thereby reducing costs and labor requirements. Integrating these technologies with forestry machinery necessitates automation, particularly in harvesters. The need for an exact three-dimensional description of tree stems has become a critical requirement. However, stem detection and segmentation from LiDAR data have typically required extensive post-processing, which limits their applicability in scenarios where real-time information is critical. This work examines the feasibility of detecting and segmenting tree stems in real-time using LiDAR data obtained through mobile laser scanning (MLS), creating a proof of concept. To achieve this, each frame captured by the sensor is processed independently and transformed into a 2D projection, where a Convolutional Neural Network (CNN) identifies and segments tree stems. This approach eliminates the need for extensive point cloud analysis, enabling faster processing and rapid response in operational environments. Detecting stems through this method offers distinct advantages over point-based methods. To fully validate this proof of concept, further development and integration of the algorithms will be necessary, along with expanding the training dataset to improve model performance. Accurate real-time detection and segmentation of tree stems and logs could be the basis for the automation and robotization of forestry operations, enhancing efficiency and precision in forest management.

Accurate identification of tree species is important for forest management, biodiversity studies, and precision forestry. Near-range LiDAR point clouds provide detailed three-dimensional information about individual trees. However, the complex structure of the point clouds and the unbalanced distribution of species make automatic classification difficult. To address these issues, this study presents a Transformer model with geometric enhancement. The model combines local geometric features and global attention to improve species recognition in forest environments. It uses geometric information with biological meaning, including point cloud normals, local density, vertical structure, and growth direction. A focal loss with class balance is also introduced to reduce the impact of species distributions with long tails. Experiments on the ForSpecial20K dataset show that the proposed method performs better than representative models based on convolution, graph methods, and Transformer architectures. It achieves higher overall accuracy (78.20%), higher mean class accuracy (73.48%), and a higher Macro-F1 score (73.21%). Results from confusion matrices and visual analysis of similar species further verify the effectiveness of the geometric features and the loss design. These results suggest that modeling structural information of forests helps improve robustness and generalization. The proposed method offers a practical solution for tree-level species mapping, fusion of LiDAR data from multiple sources, and fine-scale forest inventory. It also shows the value of combining high-resolution LiDAR data with deep learning for forestry applications.

This study aims to analyze the species diversity of tree vegetation and to estimate aboveground tree biomass, calculate carbon stock and carbon sequestration, and assess the contribution of tree diameter classes within a cluster plot in the dryland forest of Selil Village, Ulilin District, Merauke Regency, South Papua Province. This study employed a quantitative-descriptive approach based on a field survey using the National Forest Inventory (Inventarisasi Hutan Nasional/IHN) 2.0 method. Species diversity was analyzed using the Shannon–Wiener index, biomass was estimated using the Chave allometric equation, while carbon stock and carbon sequestration were calculated based on the IPCC conversion factor. The results showed that 64 individual trees representing 7 species were recorded within the cluster plot, namely kelat (Eugenia densiflora), jale (Casuarina papuana), resak (Vatica papuana), kapur (Dryobalanops aromatica), bintangur (Calophyllum papuanum), wild nutmeg (Myristica sp.), and merawan (Hopea papuana). Species diversity was classified as moderate (1 < 1.477 < 3). The most dominant species was kelat (Eugenia densiflora) with an Important Value Index (IVI) of 100.304, followed by resak (Vatica papuana) with 68.522 and jale (Casuarina papuana) with 49.023. The total above ground tree biomass was 141.228 tons/ha, with the largest contribution coming from kelat (Eugenia densiflora), jale (Casuarina papuana), and resak (Vatica papuana). This biomass produced a carbon stock of 66.377 tons C/ha and carbon sequestration of 243.383 tons CO₂/ha. The diameter class of 40 cm up contributed the greatest share to biomass and carbon, accounting for 70.71% of the total biomass. This indicates that large diameter trees play a major role in climate mitigation, are important as natural carbon sinks, and provide baseline data that can be used to support sustainable forest management and climate change mitigation policies in South Papua.

ABSTRACT Medium- to low-density airborne LiDAR enables cost-efficient, wide-area forest management inventories; however, conventional approaches to evaluating individual tree crowns often fail to identify overcrowded stands in closed-canopy conifer plantations. This study proposes a stand-density-independent framework to assess canopy crowding using two plot-level indicators derived from LiDAR height distributions: mean live crown ratio (CR) and vertical canopy cover (CC). The core of the method is a dynamically defined height threshold that isolates canopy returns while accounting for spatial variability in understory vegetation height. A two-parameter zero-shift Weibull probability density function is then fitted to the extracted canopy points. The resulting Weibull parameters provide robust mean crown base height (CBH) and mean tree height (H), which underpin CR calculation. CC is estimated from the proportion of dynamically defined canopy returns, improving robustness to understory effects relative to fixed-threshold approaches. Field validation in monoculture conifer plantations in Ibaraki and Fukuoka, Japan, yielded CBH biases of 0.17 m and 0.25 m with RMSEs of 1.64 m and 1.94 m (11.2% and 12.3% of field means, respectively), and H biases of 0.10 m and 0.22 m with RMSEs of 1.47 m and 1.56 m (6.7% and 7.8%). CR estimation errors showed no systematic dependence on stand density across a wide density range (250–2800 trees/ha), enabling reliable identification of stands requiring thinning. By combining CR and CC, 20-m resolution canopy crowding condition maps were generated, supporting prioritization of thinning operations. The proposed framework is computationally efficient and potentially applicable to monoculture conifer plantations in different countries.

ABSTRACT Brosimum alicastrum Sw. (Breadnut or Maya nut), of the Moraceae family, is an underutilised tropical food tree with high nutritional, cultural, and ecological importance in Mesoamerica. Despite its potential as a sustainable forest-based food crop, large uncertainty remains regarding its reproductive biology and fruit morphology, contributing to highly variable and unpredictable yields. This study elucidates the reproductive system, pollination strategy, and fruit characteristics of B. alicastrum in two forested areas of the Maya Biosphere Reserve, Guatemala: Uaxactún and Sierra del Lacandón. Using forest inventories, canopy-based floral observations, fruit and seedling assessments, and ethnobotanical interviews with local collectors, to assess site-specific variation in reproductive traits and yield proxies. Results demonstrate clear differences between populations. Uaxactún exhibited a predominantly monoecious system with female and hermaphroditic individuals, where nearly all trees showed signs of reproduction. In contrast, Sierra del Lacandón showed evidence of a dioecious population, with a substantial proportion of non-reproductive (male) trees, explaining lower and less reliable yields. Floral morphology, absence of animal pollinators, exposed stigmas, and abundant pollen strongly indicate wind pollination. Furthermore, the fruit is reclassified from a drupe or berry to a sorosis, a composite infructescence formed from the entire inflorescence. These findings reconcile longstanding contradictions in the literature and highlight the importance of population-specific reproductive strategies for management, conservation, and the future development of B. alicastrum as an opportunity tree crop for food security.

Abstract Synthetic aperture radar remote sensing has long been recognized as a valuable tool for estimating aboveground biomass density in forest ecosystems. In this study, we specifically evaluated the sensitivity of the SAOCOM mission to forest biomass under conditions of high spatial uncertainty in reference data (± 1.6 km). A complex single-look SAOCOM acquisition was processed to obtain a comprehensive set of polarimetric backscatter and decomposition metrics, which were used as predictors in Random Forest regression models calibrated with biomass estimates from the U.S. Forest Inventory and Analysis program. The optimized model achieved a concordance correlation coefficient of 0.50, with a mean square error of 59 Mg ha⁻¹ and a mean absolute error of 47 Mg ha⁻¹ on independent test data. Beyond predictive performance, the results demonstrate that SAOCOM observations retain significant sensitivity to forest structural variability, even when the spatial correspondence between satellite measurements and ground-based biomass references is substantially degraded. These findings highlight SAOCOM’s potential for biomass-oriented applications, even under challenging conditions.

Plant diversity plays a fundamental role in ecosystem functioning and is essential for sustaining ecosystem services. National forest inventories are key instruments for assessing floristic diversity. However, their measurement protocols may introduce bias by omitting smaller individuals because of the stem diameter criterion used or the minimum plant size threshold applied. Such bias is exacerbated in dryland ecosystems where small-statured plants with low-branching stems are particularly abundant. In this study, we evaluated the effects of using basal diameter (BD) instead of diameter at breast height, and of sampling small individuals (BD ≥ 2.5 cm), on the estimation of abundance, alpha and gamma diversity and community composition in different vegetation types in NW Mexico. We found substantial underestimation due to the omission of smaller individuals in xeric shrubland and tropical dry forest, where gamma diversity may be underestimated by up to 209% and 139%, respectively. Broadleaf forest also showed strong underestimation (133%), whereas mixed conifer–broadleaf forests were unaffected. We discuss these differential effects and propose a methodology to attenuate this underestimation and achieve more accurate floristic diversity estimates from national forest inventories in dryland vegetation, which encompasses roughly one-third of the Earth’s surface and more than half of Mexico’s territory.