Hydraulic stress is a primary driver of tree decline and mortality, yet its spectral detectability remains poorly characterized. In particular, no study has directly quantified hydraulic failure in conifers and linked this physiological collapse to canopy-level spectral responses. Here, we provide the first continuous measurements of hydraulic failure in Norway spruce and evaluate how well UAV-based hyperspectral imagery captures the onset and progression of this decline. The hydraulic stress was caused by attacks from spruce bark beetles (Ips typographus) in 56 trees under a controlled infestation experiment, and we evaluate their detectability against 227 healthy trees using physio-spectral indicators before brood emergence, a critical window for controlling beetle spread. Sap flow measurements revealed a cascading decline in infested trees, with hydraulic stress emerging 3-4 weeks after attack and occasionally advancing to hydraulic failure within 4-10 weeks. Biweekly UAV hyperspectral acquisitions were used to assess spectral sensitivity to hydraulic collapse. Physio-spectral indices from the green-shoulder region (similar to 530 nm), linked to carotenoid dynamics and photosynthetic downregulation, were the most responsive to hydraulic stress, detecting stressed trees up to twice as effectively (e.g. 70% vs. 30%) and 2-6 weeks earlier than chlorophyll-sensitive indices. The spectral trajectory of GSCR1(MS) closely mirrored the magnitude and timing of sap flow decline, and detected hydraulic failure in near-real time. These results provide mechanistic evidence that green-shoulder hyperspectral metrics are sensitive, physiologically grounded indicators of impending mortality. The findings strengthen the remote sensing basis for early detection of hydraulic stress and improve the interpretation of spectral stress signals in mature spruce forests.

This study explores the potential of spatially explicit Harvester Production Reports (HPRs) for automatic annotation of Aerial Laser Scanning (ALS) data at tree-level, enabling accurate tree species classification using Convolutional Neural Networks (CNNs). By integrating HPRs into the modelling process, this approach provides a practical solution for addressing challenges in remote sensing data annotation for forestry applications. The ALS data were acquired in managed Norway spruce-dominated forests in southern Sweden using a dual-wavelength system composed by two monochromatic sensors. Thus, three datasets were produced: the 905 nm miniVUX dataset (similar to 100 points/m(2)), the 1550 nm VUX dataset (similar to 875 points/m(2)), and the dual-wavelength dataset (similar to 975 points/m(2)), the last being a junction of the two first datasets. The automatic annotation was performed by matching tree records in the HPR and ALS data based on spatial proximity and height similarity, with a total of 45,516 HPR-recorded tree positions being linked to ALS-derived segments and assigned species labels based on HPR records. Then, the individual tree-level ALS point clouds were converted into 2D images from multiple viewing angles, with varying image dimensions and pixel sizes to accommodate trees of different sizes. These images served as input for CNN-based classification, enabling species identification across ALS datasets with varying spectral and spatial resolutions. The CNN models were trained and evaluated to classify trees into Norway spruce, Scots pine, Deciduous, and a "Noise" class for segmentation errors. The classification accuracy varied according to the dataset used, with the dual-wavelength dataset achieving the highest macro-F1 score (0.896), followed by the VUX dataset (0.894) and miniVUX dataset (0.835). These findings highlight spatially explicit HPRs as efficient, high-quality reference data for CNN-based tree species classification with minimal annotation effort.

The frequent outbreaks of European spruce bark beetle Ips typographus (L.) have destroyed huge amounts of Norway spruce Picea abies (L.) forests in central and Northern Europe. Identifying the risk factors and estimating the damage level is important for strategic damage control. The risk factors of forest damage by spruce bark beetles have mostly been analyzed on the landscape scale, while the in-stand risk factors have been less investigated. This study aims at exploring the local-scale risk factors in a flat area with spruce-dominated forest in southern Sweden. The investigated factors include four abiotic factors, i.e., soil wetness, solar radiation, slope gradient, and aspect, and three biotic factors, i.e., the number of deciduous trees and trees that died from attacks in previous years that remained ( TreesLeft ) and removed ( TreesRemoved ) from the forest stand. We put up 24 pheromone bags in six stands attacked by bark beetle in the previous years, resulting in different numbers of infested trees in each plot. We explored in which microenvironment a pheromone bag resulted in more colonization, the impact radius of each factor, and the necessary factors for a risk model. The environmental factors were obtained from remote sensing-based products and images. A generalized linear model (GLM) was used with the environmental factors as the explanatory variables and the damage levels as the response variables, i.e., the number of attacked trees for the plot scale, and healthy/infested for the single-tree scale. Using 50 m and 15 m radius of the environmental factors resulted in the best fit for the model at plot and individual tree scales, respectively. At those radii, the damage risk increased both at plot and individual tree level when spruce were surrounded by more deciduous trees, surrounded by dead trees that had been removed from the forest, and spruces located at the north and east slopes (315°–135° of aspect, > 2° slope). Soil wetness, solar radiation, and remaining standing dead trees in the surrounding did not significantly impact the damage level in the microenvironment of the study area. The GLM risk model yielded an overall accuracy of 0.69 in predicting individual trees being infested or healthy. Our efforts to investigate the risk factors provide a context for wall-to-wall mapping in-stand infestation risks, using remote sensing-based data.

The acquisition of high-quality reference data is essential for effectively modelling forest attributes. Incorporating close-range Light Detection and Ranging (LiDAR) systems into the reference data collection stage of remote sensing-based forest inventories can not only increase data collection efficiency but also increase the number of attributes measured with high quality. Therefore, we propose a model-based forest inventory method that uses reference data collected by a car-mounted mobile laser scanning (MLS) system along boreal forest roads. This approach is used for the estimation of diameter at breast height (DBH) and stem volume at the individual tree-level from airborne laser scanning (ALS) data. In addition, we compare the estimates obtained using the proposed method with the ones derived from reference data collected by traditional field inventory of 265 field plots systematically distributed over the study area. The accuracy of the estimates remained comparable regardless of the reference dataset used for estimation of DBH and stem volume. When using the field inventory dataset for model training, the root mean square error (RMSE) of DBH estimates were 4.06 cm (18.8 %) for Norway spruce trees, 6.3 cm (29.6 %) for Scots pine and 8.61 cm (55.9 %) for deciduous trees. Similarly, when evaluating predictions based on the MLS dataset as reference, RMSEs were equal to 3.97 cm (18.4 %) for Norway spruce, 6.12 cm (28.8 %) for Scots pine, and 8.98 cm (58.3 %) for deciduous trees. In general, biases were below 1 cm for most species classes, with the exception of deciduous trees. The accuracy of stem volume also had RMSEs varying across different tree species. For the estimates based on traditional field inventory, the RMSEs were 0.176 m(3) (38.8 %) for Norway spruce, 0.228 m(3) (52.4 %) for Scots pine and 0.246 m(3) (158 %) for deciduous trees. When using the MLS dataset as a reference, the RMSEs were equal to 0.176 m(3) (38.8 %), 0.228 m(3) (52.4 %), and 0.246 m(3) (158 %) for Norway spruce, Scots pine, and deciduous trees, respectively. Car-mounted MLS demonstrated its potential as an efficient alternative for collecting reference data in remote sensing-based forest inventories, which could complement traditional methods.

In recent years, the European spruce bark beetle (Ips typographus, L.) has damaged large amounts of forests in Europe, and detecting infested trees is crucial for damage control and informative decision-making regarding management. This study explores efficient methods of detecting infestations using multispectral drone images, focusing on how using different pixels from the crown segments influences the detection rates. Tree crowns were first segmented using marker-controlled watershed segmentation, and then two pixel-selection strategies were tested, including selecting the pixels closer to the tree tops, and selecting the bright pixels with values higher than certain percentiles of the entire crown segments. Two datasets were used from the same area, including 2021 with an epidemic outbreak and 2023 with an endemic outbreak, to present the potential differences caused by attack intensity. The results showed that, in the early stages (1 - 9 weeks of infestation), using the centermost pixels or the brightest pixels in the tree crowns had higher detectability than using all pixels. Red-edge-based VIs were more sensitive than red-green-based VIs. In the middle stage (10 - 16 weeks of infestation), using pixels from the entire tree crown, including tree tops and the low branches, showed higher detectability than using fewer pixels. For the late stages (after 19 weeks of infestation), using only the center pixel was sufficient, and there were minor differences between different VIs. The results were supported by observations from two datasets from different years, although variations in the detectability between different years and stands were also observed.

Remotely sensed data are frequently used for predicting and mapping ecosystem characteristics, and spatially explicit wall-to-wall information is sometimes proposed as the best possible source of information for decision-making. However, wall-to-wall information typically relies on model-based prediction, and several features of model-based prediction should be understood before extensively relying on this type of information. One such feature is that model-based predictors can be considered both unbiased and biased at the same time, which has important implications in several areas of application. In this discussion paper, we first describe the conventional model-unbiasedness paradigm that underpins most prediction techniques using remotely sensed (or other) auxiliary data. From this point of view, model-based predictors are typically unbiased. Secondly, we show that for specific domains, identified based on their true values, the same model-based predictors can be considered biased, and sometimes severely so.We suggest distinguishing between conventional model-bias, defined in the statistical literature as the difference between the expected value of a predictor and the expected value of the quantity being predicted, and design-bias of model-based estimators, defined as the difference between the expected value of a model-based estimator and the true value of the quantity being predicted. We show that model-based estimators (or predictors) are typically design-biased, and that there is a trend in the design-bias from overestimating small true values to underestimating large true values. Further, we give examples of applications where this is important to acknowledge and to potentially make adjustments to correct for the design-bias trend. We argue that relying entirely on conventional model-unbiasedness may lead to mistakes in several areas of application that use predictions from remotely sensed data.

Remotely sensed data have an important application for estimation of forest variables, e.g. height, volume, and aboveground biomass (AGB). The increased use of remotely sensed data implemented along with model-based inference has shown improved efficiency in prediction and mapping of such forest variables. In this study, plot-level airborne laser scanning data and Swedish National Forest Inventory field reference data were used to predict AGB using generalized linear models (GLMs) assuming Gamma and Tweedie distributions for the field observed AGB. The GLMs were selected considering the convenience of not correcting transformation bias as it is required in other regression models with transformed response variable. To overcome the challenge in providing reliable uncertainty estimates for the estimated forest variable map products at individual pixel-scale, we focused on computing 95% prediction intervals (PIs) for Gamma and Tweedie GLMs with a square root link function. The relative uncertainties were computed as the ratio between the half-width of the PIs and the predicted AGBs. The AGB-airborne laser scanning models were developed with root mean square error values of 22.6 Mgha-1 (26%) and 21.7 Mgha-1 (25%), respectively, for the Gamma and Tweedie GLMs. Two methods were applied to compute PIs for the Gamma GLM, one using the R package 'ciTools' and another derived through asymptotic theory. It was found that the 95% PIs computed using 'ciTools' had the most accurate coverage probability in comparison to the other method. An extended version of these PIs was also utilized for the Tweedie GLMs. The range of PIs associated with the prediction of AGB were narrower for lower predicted AGB values compared with the length of higher predicted AGB values. Comparing the two fitted models, the Gamma GLM showed lower relative uncertainties for the lower range of predicted AGBs, whereas the Tweedie GLM showed lower relative uncertainties for the higher range of predicted AGBs. Overall, the Tweedie GLM provided a better model fit for AGB predictions.

Satellite-based inventories of bark beetle attacks are increasingly used for detecting and monitoring infested forest at the landscape scale. The Normalized Distance Red & SWIR index is one of few indices that have shown higher accuracies than commonly used vegetation indices. In this study, the temporal changes of the distance red swir (ΔDRS) index were analyzed, validated and applied to multi-temporal Sentinel-2 images covering one tile of 110 x 110 km2. The main purpose was to assess the applicability of a new ΔDRS vegetation index to detect spruce forest after bark beetle (Ips typographus) attacks. Harvester data from a private forest company were used to validate the method. The normalized DRS index has previously been developed and tested at test site level, while this study explored and demonstrated the use of ΔDRS in an applied context on a larger scale. Water and chlorophyll induced changes and different disturbances were effectively identified across the landscape. A linear-discriminant analysis was used to classify 274 clusters as attacked and healthy forest, with an overall accuracy of 78%. The largest ΔDRS values in our study (>0.06) corresponded well to clear-cuts, and all 172 clear-cuts were correctly classified. We conclude that the ΔDRS index has a potential to map vegetation changes related to water and chlorophyll changes in the Scandinavian forests and that it can be useful to identify bark beetle-infested forest within 1 year after the attacks and clear-cuts.

Remote sensing (RS) techniques have been used for mapping forest variables, such as stem volume (important for forest management activities associated with timber production), over large areas which can be updated more frequently than with field inventory (FI) data. In this study, wall-to-wall TanDEM-X synthetic aperture radar images were used as auxiliary RS data for model-based prediction of stand-level volumes for two models, trained using volumes computed from FI (A) and airborne laser scanning estimations (B), respectively. The models were validated with harvester data available for independent stands. It was observed that the performance of model B was slightly better compared to model A based on adjusted R 2 and root mean squared error values. Therefore, it can be concluded that a completely RS based approach for prediction and mapping of stand volumes would be as promising as a method based on FI data along with being cost- and labour-efficient.

In sustainable forest resource management, establishing forest conservation areas is important to maintain forest biodiversity. However, assessing the conservation value of forests is challenging because the target areas are often both large and remote. We explored using dense airborne laser scanning (ALS) data to estimate conservation values. Field data were collected at sites in Sweden for standing deadwood (S trees), laying deadwood (L trees), and trees considered particularly important for conserving biodiversity (N trees), and forest conservation values were determined using a commonly employed method in the field. A template matching method was then used to detect L, S and N trees from ALS data. L trees were identified from linear features in the point cloud above the ground, with a 52 % detection error rate. S and N trees were identified from unusually small or large crown diameters, with 71 % and 83 % error rates, respectively. We also tested the relationships between the three types of indicator tree, their summed values and the field inventory-assessed conservation values. Regressions between the assessed conservation values and ALS indicators were most robust comparing the three test sites when using the summed number of L, S, and N trees. A wall-to-wall map covering a 3 km × 4 km area was generated using Kernel density estimation of the summed number of ALS-derived indicators, to represent relative conservation values. The map was validated using 10 1-ha plots, and yielded an R2 value of 0.6 for predicted conservation values at the plot level. We conclude that ALS data can be used to map forest conservation values and inform decisions about which forests should be used for timber production and which should be set aside as conservation areas. The maps could also be used as a data source for habitat analysis.

The estimation of species composition and species-specific stem volumes are critical components of many forest inventories. The use of airborne laser scanning with multiple spectral channels may prove instrumental for the cost-efficient retrieval of these forest variables. In this study, we scanned a boreal forest using two channels: 532 nm (green) and 1064 nm (near infrared). The data was used in a two-step methodology to (1) classify species, and (2) estimate species-specific stem volume at the level of individual tree crowns. The classification of pines, spruces and broadleaves involved linear discriminant analysis (LDA) and resulted in an overall accuracy of 91.1 % at the level of individual trees. For the estimation of stem volume, we employed species-specific k-nearest neighbors models and evaluated the performance at the plot level for 256 field plots located in central Sweden. This resulted in root-mean-square errors (RMSE) of 36 m3/ha (16 %) for total volume, 40 m3/ha (27 %) for pine volume, 32 m3/ha (48 %) for spruce volume, and 13 m3/ha (87 %) for broadleaf volume. We also simulated the use of a monospectral near infrared (NIR) scanner by excluding features based on the green channel. This resulted in lower overall accuracy for the species classification (86.8 %) and an RMSE of 41 m3/ha (18 %) for the estimation of total stem volume. The largest difference when only the NIR channel was used was the difficulty to accurately identify broadleaves and estimate broadleaf stem volume. When excluding the green channel, RMSE for broadleaved volume increased from 13 to 26 m3/ha. The study thus demonstrates the added benefit of the green channel for the estimation of both species composition and species-specific stem volumes. In addition, we investigated how tree height influences the results where shorter trees were found to be more difficult to classify correctly.

Detecting forest insect damage before the visible discoloration (green attacks) using remote sensing data is challenging, but important for damage control. In recent years, the European spruce bark beetle (Ips typographus, L.) has damaged large amounts of forest in Europe, and some studies have been conducted on the early detection of infestations and forest vulnerabilities before attacks. This study assessed the detectability of the green attacks using multispectral drone images and examined the possibility of detecting vulnerable trees before attacks. The study used multispectral drone images covering 24 plots from 6 forest stands in southern Sweden, acquired in May (before attacks), June (green attack), August (green and yellow attack), and October 2021 (red attack). Drone images of individual-tree crowns were segmented and vegetation indices (VIs) were calculated for every single tree. Trees with the same duration of infestation were grouped for the analysis. Random Forest Classification (RF) and linear discriminant analysis (LDA) were used to build and compare models using all bands, sensitive bands, all VIs, and single VIs, respectively. Results were also compared between different ways of dividing training and testing data. When randomly dividing 90% and 10% trees for training and testing, the models could classify vulnerable trees before attacks with low accuracy. However, when training on trees in five stands, no model could predict infestations in the remaining test stand. Similarly, the models could not identify trees infested for fewer than five weeks. We conclude that the detectability of vulnerable trees before attacks and attacked trees with fewer than five weeks of infestation is very low. We noticed a considerable overfitting when using RF with more variables compared to using LDA with single VIs.

In the recent decades, Norway spruce forests (Picea abies Karst.) in Europe have been subject to large-scale tree mortality caused by the spruce bark beetle (Ips typographus L.). The outbreaks were induced by storm-felling events and periods of drought, which are becoming increasingly frequent due to climate change. Because storms and droughts spatially affect forests differently, the infestation patterns and configuration of the bark beetles might differ between storms and drought. In this study, we examined local and landscape factors associated with bark beetle-caused tree mortality after one storm (2005) and one drought-induced spruce bark beetle outbreak (2018), both occurring in southern Sweden. A total of 13,192 infested one-ha pixels after the storm and 6,425 one-ha pixels after the drought (in total 19,617) were compared regarding differences in infestation occurrence and size and associated forest structures and climate between the two different outbreaks, using a generalized linear model (GLM) approach. Based on our findings, we discovered that the allocation of infestation patch sizes (including four classes: 5-10, 11-25, 26-50 and >50 infested trees) for the two outbreaks were quite similar with a large proportion (>0.6) of small groups (& LE;10 trees). However, the outcomes from this study demonstrate that the drivers behind the spatial configuration of bark beetle infestations can differ considerably between outbreaks triggered by storms and droughts, and the main cause seems to be linked to the spatial distribution of susceptible trees. The most consistent differences for both occurrence and infestation size were that storm-induced infestations increased more with spruce volumes and area of protected forests (nature reserves) in the landscape; whereas for the drought-induced infestations, occurrence and size increased more with clear-cuts in the landscape and spruce heights across spatial scales. Soil moisture and mean drought index (SPEI; May-July) were important for both outbreaks, but generally more important for the infestation sizes after droughts than after storms and may involve a time-lagged effect. The reasoning behind the differences between storms and droughts may be that during storm-induced outbreaks, when the wind-felled trees are removed or not suitable anymore, bark beetles need to find specific susceptible standing trees, while after drought all trees are more or less stressed, which results in a selection of large trees in dry and warm landscapes as they have more resources and favorable reproduction conditions. Finally, we show that the previous infestation size influenced the later infestation size negatively within landscapes of 25 ha and this seems to be related to depletion of susceptible host trees. These results are important for the assessment of more specific outbreak predictions, which should be integrated in future risk mapping of bark beetle outbreaks.

Detecting disease- or insect-infested forests as early as possible is a classic application of remote sensing. Under conditions of climate change and global warming, outbreaks of the European spruce bark beetle (Ips typographus, L.) are threatening spruce forests and the related timber industry across Europe, and early detection of infestations is important for damage control. Infested trees without visible discoloration (green attack) have been identified using multispectral images, but how early green attacks can be detected is still unknown. This study aimed to determine when infested trees start to show an abnormal spectral response compared with healthy trees, and to quantify the detectability of infested trees during the infestation process. Pheromone bags were used to attract bark beetles in a controlled experiment, and subsequent infestations were assessed in the field on a weekly basis. In total, 977 trees were monitored, including 208 attacked trees. Multispectral drone images were obtained before and during the insect attacks, representing different periods of infestation. Individual tree crowns (ITC) were delineated by marker-controlled watershed segmentation, and the average reflectance of ITCs was analyzed based on the duration of infestation. The detectability of green attacks and driving factors were examined. We propose new Multiple Ratio Disease-Water Stress Indices (MR-DSWIs) as vegetation indices (VI) for detecting infestations. We defined a VI range of 5-95% as a healthy tree, and a VI value outside that range as an infested tree. Detection rates using multispectral images were always higher than discoloration rates observed in the field, and the newly proposed MR-DSWIs detected more infested trees than the established VIs. Infestations were detectable at 5 and 10 weeks after an attack at a rate of 15% and 90%, respectively, from the multispectral drone images. Weeks 5-10 of infestation therefore represent a suitable period for using the proposed methodology to map infestation at an early stage.

Detecting forest insect damage before the visible discoloration (green attacks) using remote sensing data is challenging, but important for damage control. In recent years, the European spruce bark beetle (Ips typographus, L.) has damaged large amounts of forest in Europe. However, it is still debatable how early the infestations can be detected with remote sensing data. Some studies showed a spectral difference between healthy and green-attacked spruce trees at the plot level, while others showed that spectral differences existed before attacks. Therefore, a hypothesis is proposed that no spectral difference can be identified between green-attacked forests compared to healthy forests if the differences do not exist before the attacks. In this study, we tested this hypothesis using Sentinel-2 and WorldView-3 SWIR images on 24 healthy plots and 24 plots with mild, moderate, and severe attacks. In the results, the severely attacked plots did not show significant spectral differences in the Sentinel-2 images until August, and the sensitivity was found in the blue, red, red-edge, and SWIR band. Only the red band showed a significant difference between the healthy and moderately attacked plots in August, and only the blue, red, and SWIR band showed significant differences in September, October, and November. No significant differences were observed in the WorldView-3 images at the plot or individual tree level. We accepted the hypothesis that green attacks do not show spectral differences with the healthy forests when the differences do not exist before the attacks. We concluded that the SWIR bands were sensitive to attacks in the Sentinel-2 images with 10 m resolution, but not in the WorldView-3 images with 3.7 m resolution. Further studies are needed to explore the methodology of using WorldView-3 SWIR images for the early detection of forest infestation.

The collection of field-reference data is a key task in remote sensing-based forest inventories. However, traditional methods of collection demand extensive personnel resources. Thus, field-reference data collection would benefit from more automated methods. In this study, we proposed a method for individual tree detection (ITD) and stem attribute estimation based on a car-mounted mobile laser scanner (MLS) operating along forest roads. We assessed its performance in six ranges with increasing mean distance from the roadside. We used a Riegl VUX1LR sensor operating with high repetition rate, thus providing detailed cross sections of the stems. The algorithm we propose was designed for this sensor configuration, identifying the cross sections (or arcs) in the point cloud and aggregating those into single trees. Furthermore, we estimated diameter at breast height (DBH), stem profiles, and stem volume for each detected tree. The accuracy of ITD, DBH, and stem volume estimates varied with the trees' distance from the road. In general, the proximity to the sensor of branches 0-10 m from the road caused commission errors in ITD and over estimation of stem attributes in this zone. At 50-60 m from roadside, stems were often occluded by branches, causing omissions and underestimation of stem attributes in this area. ITD's precision and sensitivity varied from 82.8% to 100% and 62.7% to 96.7%, respectively. The RMSE of DBH estimates ranged from 1.81 cm (6.38%) to 4.84 cm (16.9%). Stem volume estimates had RMSEs ranging from 0.0800 m(3) (10.1%) to 0.190 m(3) (25.7%), depending on the distance to the sensor. The average proportion of detected reference volume was highly affected by the performance of ITD in the different zones. This proportion was highest from 0 to 10 m (113%), a zone that concentrated most ITD commission errors, and lowest from 50 to 60 m (66.6%), mostly due to the omission errors in this area. In the other zones, the RMSE ranged from 87.5% to 98.5%. These accuracies are in line with those obtained by other state-of-the-art MLS and terrestrial laser scanner (TLS) methods. The car-mounted MLS system used has the potential to collect data efficiently in large-scale inventories, being able to scan approximately 80 ha of forests per day depending on the survey setup. This data collection method could be used to increase the amount of field-reference data available in remote sensing based forest inventories, improve models for area-based estimations, and support precision forestry development.

This article describes algorithms to extract tree crowns using two-dimensional (2D) and three-dimensional (3D) segmentation. As a first step, a 2D-search detected the tallest trees but was unable to detect trees located below other trees. However, a 3D-search for local maxima of model fits could be used in a second step to detect trees also in lower canopy layers. We compared tree detection results from ALS carried out at 1450 m above ground level (high altitude) and tree detection results from ALS carried out at 150 m above ground level (low altitude). For validation, we used manual measurements of trees in ten large field plots, each with an 80 m diameter, in a hemiboreal forest in Sweden (lat. 58 degrees 28' N, long. 13 degrees 38' E). In order to measure the effect of using algorithms with different computational costs, we validated the tree detection from the 2D segmentation step and compared the results with the 2D segmentation followed by 3D segmentation of the ALS point cloud. When applying 2D segmentation only, the algorithm detected 87% of the trees measured in the field using high-altitude ALS data; the detection rate increased to 91% using low-altitude ALS data. However, when applying 3D segmentation as well, the algorithm detected 92% of the trees measured in the field using high-altitude ALS data; the detection rate increased to 99% using low-altitude ALS data. For all combinations of algorithms and data resolutions, undetected trees accounted for, on average, 0-5% of the total stem volume in the field plots. The 3D tree crown segmentation, which was using crown density models, made it possible to detect a large percentage of trees in multi-layered forests, compared with using only a 2D segmentation method.

Obtaining low vegetation data is important in order to quantify the structural characteristics of a forest. Dense three-dimensional (3D) laser scanning data can provide information on the vertical profile of a forest. However, most studies have focused on the dominant and subdominant layers of the forest, while few studies have tried to delineate the low vegetation. To address this issue, we propose a framework for individual tree crown (ITC) segmentation from laser data that focuses on both overstory and understory trees. The framework includes 1) a new algorithm (SSD) for 3D ITC segmentation of dominant trees, by detecting the symmetrical structure of the trees, and 2) removing points of dominant trees and mean shift clustering of the low vegetation. The framework was tested on a boreal forest in Sweden and the performance was compared 1) between plots with different stem density levels, vertical complexities, and tree species composition, and 2) using airborne laser scanning (ALS) data, terrestrial laser scanning (TLS) data, and merged ALS and TLS data (ALS + TLS data). The proposed framework achieved detection rates of 0.87 (ALS + TLS), 0.86 (TLS), and 0.76 (ALS) when validated with field inventory data (of trees with a diameter at breast height >= 4 cm). When validating the estimated number of understory trees by visual interpretation, the framework achieved 19%, 21%, and 39% root-mean-square error values with ALS + TLS, TLS, and ALS data, respectively. These results show that the SSD algorithm can successfully separate laser points of overstory and understory trees, ensuring the detection and segmentation of low vegetation in forest. The proposed framework can be used with both ALS and TLS data, and achieve ITC segmentation for forests with various structural attributes. The results also illustrate the potential of using ALS data to delineate low vegetation.

Greatly simplified ecosystems are often neglected for biodiversity studies. However, these simplified systems dominate in many regions of the world, and a lack of understanding of what shapes species occurrence in these systems can have consequences for biodiversity and ecosystem services at a massive scale. In Fennoscandia, ~90% of the boreal forest (~21Mha) is structurally simplified with little knowledge of how forest structural elements shape the occurrence and diversity of for example epiphytic lichens in these managed forests. One form of structural simplification is the reduction of the number and frequency of different tree species. As many lichen species have host tree preferences, it is particularly likely that this simplification has a huge effect on the lichen community in managed forests. In a 40-70 years old boreal forest in Sweden, we therefore related the occurrence and richness of all observed epiphytic lichens to the host tree species and beta and gamma lichen diversity at the forest stand level to the stand's tree species composition and stem diameter. Picea abies hosted the highest lichen richness followed by Pinus sylvestris, Quercus robur, Alnus glutinosa, Betula spp., and Populus tremula. However, P. tremula hosted twice as many uncommon species as any of the other tree species. Stand level beta and gamma diversity was twice as high on stands with four compared to one tree species, and was highest when either coniferous or deciduous trees made up 40-50% of the trees. The stem diameter was positively related to lichen richness at the tree and stand level, but negatively to beta diversity. For biodiversity, these findings imply that leaving a few trees of a different species during forest thinning is unlikely as effective as combining life-boat trees for endangered species with an even tree species mixture.

The increased temporal frequency of optical satellite data acquisitions provides a data stream that has the potential to improve land cover mapping, including mapping of tree species. However, for large area operational mapping, partial cloud cover and different image extents can pose challenges. Therefore, methods are needed to assimilate new images in a straightforward way without requiring a total spatial coverage for each new image. This study shows that Bayesian inference applied sequentially has the potential to solve this problem. To test Bayesian inference for tree species classification in the boreo-nemoral zone of southern Sweden, field data from the study area of Remningstorp (58°27′18.35″ N, 13°39′8.03″ E) were used. By updating class likelihood with an increasing number of combined Sentinel-2 images, a higher and more stable cross-validated overall accuracy was achieved. Based on a Mahalanobis distance, 23 images were automatically chosen from the period of 2016 to 2018 (from 142 images total). An overall accuracy of 87% (a Cohen’s kappa of 78.5%) was obtained for four tree species classes: Betula spp., Picea abies, Pinus sylvestris, and Quercus robur. This application of Bayesian inference in a boreo-nemoral forest suggests that it is a practical way to provide a high and stable classification accuracy. The method could be applied where data are not always complete for all areas. Furthermore, the method requires less reference data than if all images were used for classification simultaneously.