Article https://doi.org/10.1038/s41467-023-44004-5 Edge effects on tree architecture exacerbate biomass loss of fragmented Amazonian forests Received: 24 April 2023 Matheus Henrique Nunes 1,2 , Marcel Caritá Vaz3, José Luís Campana Camargo 4,5, William F. Laurance 6, Ana de Andrade 5 , Accepted: 27 November 2023 Alberto Vicentini5,7, Susan Laurance 6, Pasi Raumonen 8, Toby Jackson 9 , Gabriela Zuquim10, Jin Wu 11, Josep Peñuelas 12,13, Jérôme Chave 14 & Eduardo Eiji Maeda 1,15 Check for updates 1234567890():,; 1234567890():,; Habitat fragmentation could potentially affect tree architecture and allometry. Here, we use ground surveys of terrestrial LiDAR in Central Amazonia to explore the influence of forest edge effects on tree architecture and allometry, as well as forest biomass, 40 years after fragmentation. We find that young trees colonising the forest fragments have thicker branches and architectural traits that optimise for light capture, which result in 50% more woody volume than their counterparts of similar stem size and height in the forest interior. However, we observe a disproportionately lower height in some large trees, leading to a 30% decline in their woody volume. Despite the substantial wood production of colonising trees, the lower height of some large trees has resulted in a net loss of 6.0 Mg ha−1 of aboveground biomass – representing 2.3% of the aboveground biomass of edge forests. Our findings indicate a strong influence of edge effects on tree architecture and allometry, and uncover an overlooked factor that likely exacerbates carbon losses in frag- mented forests. The three-dimensional form of trees, or tree architecture, reflects the transpiration and maximise light capture3–8. In Amazonian forests, allocation of photosynthetically fixed carbon within the plants. Tree trees vary greatly in size and architecture across species, as a result of architecture can be considered a by-product of environmental pres- evolutionary processes over millions of years9–11. Size and architecture sures on plant growth, reproduction and survival1,2. Fine adjustments also vary considerably across individuals due to short- to mid-term of the aboveground architecture of trees can minimise competition acclimation and adaptation to changing environmental conditions, from neighbouring trees, improve hydraulic conductance, limit including canopy gaps caused by the mortality of large trees and forest 1 Department of Geosciences and Geography, University of Helsinki, Helsinki, Finland. 2Department of Geographical Sciences, University of Maryland, College Park, MD, USA. 3Institute for Environmental Science and Sustainabilty, Wilkes University, Wilkes-Barre, PA, USA. 4Ecology Graduate Program, National Institute for Amazonian Research, (INPA), Manaus, Brazil. 5Biological Dynamics of Forest Fragments Project (BDFFP) at National Institute for Amazonian Research (INPA), Manaus, Brazil. 6Centre for Tropical Environmental and Sustainability Science, College of Science and Engineering, James Cook University, Cairns, Queensland, Australia. 7Coordenação de Pesquisas em Ecologia, Instituto Nacional de Pesquisas da Amazônia (INPA), Manaus, AM, Brasil. 8Computing Sciences, Tampere University, Tampere, Finland. 9Plant Sciences and Conservation Research Institute, University of Cambridge, Cambridge, United Kingdom. 10 Amazon Research Team, Department of Biology, University of Turku, Turku, Finland. 11School of Biological Sciences and Institute for Climate and Carbon Neutrality, The University of Hong Kong, Hong Kong, China. 12CREAF, Cerdanyola del Vallès, Barcelona, Catalonia, Spain. 13CSIC, Global Ecology Unit CREAF- CSIC-UAB, Bellaterra, Barcelona, Catalonia, Spain. 14Laboratoire Evolution et Diversité Biologique, CNRS, UPS, IRD, Université Paul Sabatier, Toulouse, France. 15 Finnish Meteorological Institute, FMI, Helsinki, Finland. e-mail: mhnunes@umd.edu; eduardo.maeda@helsinki.fi Nature Communications | (2023)14:8129 1 Article https://doi.org/10.1038/s41467-023-44004-5 blowdowns6,12–15,. The architectural traits of Amazonian trees control additional losses in AGB (i.e. thinner and shorter branches, crown CO2 loss from stem and branch respiration, hydraulic safety and effi- damages) or offset this biomass loss in fragmented forests. Terrestrial ciency, light capture and mechanical stability, which together mod- laser scanning (TLS) or “terrestrial LiDAR” surveys can be of particular ulate biomass allocation and carbon storage16–19. Changes in tree importance to reduce uncertainties in tree volume estimates by con- architecture could therefore reveal biome-wide impacts on carbon sidering the geometry and shape of trees, and without the difficulties cycling, with regional and global influences on vegetation associated with traditional destructive methods of tree feedbacks1,9,20,21. measurement34. Furthermore, TLS surveys offer new perspectives into The architecture of Amazonian trees could be affected by dis- the three-dimensional (3D) structure of trees (Fig. 1), including fine- turbances arising from forest fragmentation. The edges of forest frag- scale architectural traits2,35 and accurate tree allometry estimates36. ments tend to have greater light availability due to the mortality of large Understanding how forest edge trees adjust their architecture and trees and lateral light penetrating from the edges22–24. This may induce allometry can help us predict how plants acclimate to and survive changes in tree architecture to optimise the capture and use of light environmental changes, and their impacts on biogeochemical fluxes under these new circumstances, including higher vertical and horizontal and on the terrestrial carbon cycle. crown growth that modify branching patterns and crown shape17,25,26. Here, we tested the hypotheses that: (1) both pre-existing trees Higher temperatures and lower water availability in forest edges established before forest fragmentation and trees that colonised the increase the evaporative demand of the vegetation27,28, and trees can forest fragments had their architectural traits and allometry affected shorten the distances for transporting water and nutrients to minimise by forest edges, given the higher light availability24, higher wind hydraulic conductance7. High wind turbulence near the fragment edges turbulence29, the hotter and drier conditions near these forest edges28 may kill highly asymmetrical trees that deviate their centre of gravity and the high mortality of large trees that may damage their substantially from their stems29,30. On the other hand, highly symme- neighbours;23 and (2) the AGB of fragmented forests is impacted by trical trees can have greater mechanical stability but a limited ability to edge effects on tree allometry, with potentially significant biome-wide avoid competition for light with neighbouring trees that can fit their implications. To address these questions, we investigate the long-term crowns into irregular spaces. To complicate matters further, the high edge effects on tree architecture and tree allometry 40 years after mortality of large individuals can damage neighbouring trees with sup- forest fragmentation, and how changes in tree allometry impact the pressed aboveground biomass allocation, which can have large effects AGB of fragmented Amazonian forests. To quantify the edge effects on on their architecture and size31. However, large uncertainties remain tree architecture, we conducted fieldwork in the Biological Dynamics regarding the effects of forest fragmentation on tree architecture, par- of Forest Fragments Project (BDFFP) in Central Amazonia, the world’s ticularly because (i) tree architecture varies considerably across life longest‐running experimental study of habitat fragmentation37. We stages32 (ii) multiple architectural attributes interact to affect the struc- used TLS surveys to measure the 3D structure of trees and develop tural complexity of individual trees2, and (iii) the responses of trees to allometric models for edge versus interior trees. We then used more forest fragmentation vary enormously within and between species33. than 12,000 DBH measurements to quantify the impacts of Edge effects on tree architecture could affect allometric models fragmentation-related allometric changes on the forest AGB. We find that predict the aboveground biomass (AGB) of fragmented Amazo- that edge effects impact the allometry of trees by altering the archi- nian forests as a function of more easily measurable properties of stem tecture of both colonising and surviving trees, and these effects diameter (DBH, diameter at breast height) and tree height34. Long-term exacerbate the biomass loss of fragmented Amazonian forests. tree measurements have shown that forest fragments in Central Amazonia experience a dramatic loss of aboveground tree biomass Results caused by the mortality of large trees that is not offset by the growth Edge effects on architectural traits and recruitment of new trees22,23. However, differences in tree allo- We used tree height to determine whether trees had been established metry caused by edge effects on tree architecture could either lead to before or after forest fragmentation. Trees above 20 m in height have 34 Tree height (m) Forest edges 0 0 Distance from edges (m) 500 Fig. 1 | Point cloud of trees growing in fragmented forests in Central Amazonia, tree asymmetry and vertical distribution of branches. Here, we investigate how the obtained using high-resolution terrestrial laser scanning (TLS). The quantifi- architecture of surviving and young colonising trees change with proximity to cation of architectural attributes of tropical trees is challenging and has been lar- forest edges. We used TLS data that resulted in a point spacing of 1.4 cm at a 20 m gely overlooked10. TLS offers new perspectives into the three-dimensional (3D) distance from the scanner. structure of trees, including descriptions of fine-scale architectural traits such as Nature Communications | (2023)14:8129 2 Article https://doi.org/10.1038/s41467-023-44004-5 a Thinner branches b Thinner trunk 3 350 3 140 Branch surface area / volume m2 m Trunk surface area / volume m2 m 300 120 100 250 80 200 60 150 40 100 20 50 5 10 15 20 25 30 35 5 10 15 20 25 30 35 c Umbrella-shaped crowns d Shift in wood allocation relative to maximise light capture 20 0.75 0.70 15 0.65 Path fr 10 0.60 0.55 5 0.50 5 10 15 20 25 30 35 5 10 15 20 25 30 35 e Vertical crown size per f Horizontal crown size per unit tree height 0.7 unit DBH 0.8 1 1 Relative crown width m1 cm Relative crown depth m1 m 0.6 0.7 0.5 0.6 0.4 0.3 0.5 0.2 0.4 5 10 15 20 25 30 35 5 10 15 20 25 30 35 Tree height (m) Tree height (m) Fig. 2 | Architectural traits acquired using a Terrestrial Laser Scanner (TLS) in width (m cm−1) of trees in forest edges (orange) and forest interior (green). Each line Central Amazonia varied with tree height and were affected by edge effects. corresponds to the model prediction obtained by fitting 200 randomised permu- Nonlinear mixed-effects models were used to predict a surface area per unit tations of subsets split into 80/20 for calibration and validation, respectively. Plot volume (m2 m−3) of branches, b surface area per unit volume (m2 m−3) of trunks, titles indicate the ecological meaning of high trait values. TLS-based trees are c path fraction, d asymmetry, e relative crown depth (m m−1), and f relative crown shown for a visual representation of the variation between low and high trait values. been established before the forest fragmentation in 1980 and thus (CI95%: 0.45–0.65 m m−1 in the edges vs. 0.51–0.56 m m−1 in the have survived the 40-year-old edge effects. Two-thirds of trees below interior) of tall trees in the edges were not distinct from those of 20 m in height were post-fragmentation recruits, which demonstrates interior forests, indicating that crown dimensions in the edges were that a large proportion of small trees have already been exposed to proportional to trunk size and tree height compared to trees in the edge effects during their juvenile phase (Supplementary Fig. 3). Edge interior. effects on architectural traits varied in extent; differences in relative Our results also demonstrate that short trees colonising the crown width and relative crown depth were most pronounced within forest edges had thicker branches and trunks, owing to reduced 10 m from the edges, ~20 m for path fraction, ~40 m for trunk and branch surface area per unit volume (CI95%: 170 − 200 m2 m−3 in the branch surface area unit volume and 55 m for asymmetry (Supple- edges vs. 250–400 m2 m−3 in the interior) and trunk surface area per mentary Fig. 4). unit volume (CI95%: 74–77 m2 m−3 in the edges vs. 80–100 m2 m−3 in Our models demonstrated that edge effects affected archi- the interior), trees were more asymmetrical (CI95%: 3.0–3.2 in the tectural traits, but these effects were dependent on when plants edges vs. 2.0–2.5 in the interior) and had higher path fraction (CI95%: were established in the forest fragments (Fig. 2). The surviving tall 0.59–0.63 in the edges vs. 0.53–0.57 in the interior). The crowns of trees in the edges had higher surface area per unit volume of trunks small trees in the edges had larger relative depth most possibly (CI95%: 24−26 m2 m−3 in the edge vs. CI95%: 14–16 m2 m−3 in the owing to multi-stemmed trees that colonised the forest edges interior), which demonstrates that edge effects led to thinner (CI95%: 0.60–0.75 m m−1 in the edges vs. 0.50–0.55 m m−1 in the trunks (as thinner objects have a higher area per unit volume). Tall interior), but smaller relative width given the large trunk sizes of trees were more symmetrical in the edges than in the interior for- short trees in the forest edges (CI95%: 0.41–0.50 m cm−1 in the edges ests (CI95%: 14–18 in the edge vs. 11–13 in the interior) and had a vs. 0.51–0.56 m cm−1 in the interior). reduced path fraction (CI95%: 0.63 − 0.67 in the edge vs. 0.69–0.73 in The within-plot variability - seen as local effects arising from edge the interior). Relative crown width (CI95%: 0.20–0.29 m cm−1 in the effects, as well as species-specific and ontogenetic influences edges vs. 0.25–0.27 m cm−1 in the interior) and relative crown depth - accounted for most of the trait variability (Supplementary Figure 5). Nature Communications | (2023)14:8129 3 Article https://doi.org/10.1038/s41467-023-44004-5 a Edge effects on wood allocation to b Edge effects on wood allocation to trunks and branches trunks and branches + on tree height R2interior = 0.90 R2interior = 0.88 2 2 R2edge= 0.89 R2edge= 0.87 Log of aboveground woody v olume m3 1 1 0 0 0 2 Log of D BH 2 H m3 Log of D BH m Interior Edge Fig. 3 | Allometric models for predicting the woody volume of Amazonian landscape (position of fragment within the landscape and fragment size) and trees. DBH is the diameter at breast height measured at 1.3 m above the ground in region within Central Amazonia were included as random variables. Points repre- cm, and H is the total tree height in metres. Tree measurements were acquired sent the observed values and each line corresponds to the model prediction using a Terrestrial Laser Scanner (TLS) within the Biological Dynamics of Forest obtained by fitting 200 randomised permutations of subsets split into 80/20 for Fragments Project (BDFFP) in Central Amazonia. Plot identity nested within calibration and validation, respectively. The within-plot variability also includes analytical errors, such as those (Supplementary Fig. 10), which may have negatively affected the arising from measurement, co-registration, tree extraction and QSM – woody volume predictions when using DBH only. although we were unable to quantify them. To illustrate these allometric effects, a surviving tree with 33 m Architectural traits, except relative crown depth, were correlated height and 70 cm DBH is predicted to have a woody volume of 7.7 m3 in with the woody volume of individual trees (Supplementary Figure 6). interior forests and 7.4 m3 in the forest edges, a non-significant varia- In particular, one component of variation was tightly linked to varia- tion in the woody volume of surviving trees related to edge effects on tion in woody volume, and explained nearly half of all the changes in architectural traits (CI95%: 7.3–8.0 in the interior vs. 7.0–8.0 in the architectural traits (Supplementary methods 5; Supplementary Fig- edges). A colonising tree with 10 m height and 10 cm DBH is predicted ure 7). The component shows that trees with thicker branches and to have a woody volume of 0.12 m3 in the interior and 0.18 m3 in edges, trunks had higher woody volume, but they tended to be more asym- an increase of 50% in the woody volume caused by edge effects. metrical and had a higher path fraction. However, when woody volume is predicted as a function of DBH only, a 70 cm DBH tree in the interior is predicted to have a woody volume of Edge effects on tree allometry 8.14 m3, but edge effects reduce a surviving 70 cm DBH tree to a We developed allometric equations for trees in the forest interior and volume of 6.27 m3, which corresponds to a 30% decline in its woody forest edges that predict the aboveground woody volume as a function volume. This also demonstrates the ability of the allometric model to of stem size and tree height (DBH2 x H) (Eq. 1 and 2) or as a function of capture variation in tree height caused by edge effects. DBH only (Eq. 3 and 4) for trees in the forest interior and forest edges. Effects of fine-scale architectural variation on allometric relationships ln ðWoody volumeinterior Þ =  0:21 + 0:81 ln ðDBH2 Hinterior Þ + εi ð1Þ (Eq. 1 and 2) were pronounced within the first 76 m from the forest edges, whereas edge effects on Eq. 3 and 4 with woody volume as a function of DBH only were most pronounced within the first 55 m from ln ðWoody volumeedge Þ =  0:13 + 0:72 ln ðDBH2 Hedge Þ 2 ð2Þ the forest fragment margins (Supplementary Fig. 8a, b). While equa- + 0:016 ln ðDBH2 Hedge Þ + εi tions 1 and 2 reflect differences in allocation patterns at the trunk and branch levels caused by fine-scale architectural variation (i.e. thinning or thickening of branches and trunks, branch loss and changes in ln ðWoody volumeinterior Þ = 2:80 + 1:97 ln ðDBHinterior Þ + εi ð3Þ branch length), equations 3 and 4 also capture edge effects on tree height (i.e. height growth or height reduction from collateral damages ln ðWoody volumeedge Þ = 2:45 + 1:72 ln ðDBHedge Þ + εi ð4Þ to living trees). We observed that for a given DBH and height, the surviving tall trees in the edges had similar aboveground woody volume as their Edge effects on AGB estimates across larger-spatial scales counterparts in the forest interior, while short colonising trees in the We used allometric models to predict woody volume of trees in edge edges had larger woody volume than those in the forest interior and interior forests, enabling us to estimate aboveground biomass (Fig. 3a). These allometric relationships were also valid for the com- (AGB) across larger spatial scales. Linear mixed models applied to data partmentalised volume in trunks and branches (Supplementary Fig. 9a, from 44 1-ha permanent plots revealed a statistically significant b). However, predictions of woody volume as a function of DBH only reduction in AGB of 24.7 Mg ha−1 due to edge effects (t =− 3.1; P- revealed a reduced aboveground woody volume of the surviving trees value = 0.003). This reduction accounted for nearly 10% of the AGB of in the edges (Fig. 3b). We found that a considerable fraction of trees structurally intact forests (282.2 ± 15.3 Mg ha−1) and comprised two near the edges had a disproportionately lower height for a given DBH components: first, there was an 18.7 Mg ha−1 decline in AGB due to edge Nature Communications | (2023)14:8129 4 Article https://doi.org/10.1038/s41467-023-44004-5 300 architecture. 3D measurements of tree size and shape provide a per- AGB loss caused by edge effects on forest structure spective on forest structure and its spatial variability that is difficult to 290 AGB loss caused by edge effects on tree allometry achieve with destructive methods of tree measurement2,36. Forest fragmentation had strong edge effects on architectural 1 traits relative to trees in structurally intact forests. However, these Predicted AGB Mg ha 280 effects were primarily dependent on which stage of their lives trees 18.7 Mg ha 1 started experiencing edge effects. Tall trees above 20 m in height were 270 likely established before forest fragmentation in 1980. These surviving trees were able to allocate a similar amount of wood to branches and 260 6.0 Mg ha 1 trunks in comparison to trees in interior forests (Supplementary Fig- ure 9). The forest edges of our study have temperatures 5 °C hotter and lower water availability than the interior, which leads to higher branch 250 loss near the edges28. However, we found that surviving trees have the ability to maintain their woody volume, possibly by producing new 240 Not considering edge Considering edge effects branches (resprouting), which has positive influences on tree growth effects on tree allometry on tree allometry and survival38. The tall trees in the edges had lower path fraction and Interior Edge were more symmetrical than trees in interior forests. In fact, trees n = 16 ha n = 28 ha under stress can lose terminal branches39, with the remaining branches Fig. 4 | Predicted aboveground biomass (AGB) for edge (N = 28 ha) versus shorter in path length to increase hydraulic safety by decreasing the interior (N = 16 ha) forest plots within the Biological Dynamics of Forest length that water and nutrients need to travel for photosynthesis7,40. Fragments Project (BDFFP) in Central Amazonia, the world’s longest-running The observed higher symmetry of surviving trees near the edges can experimental study of habitat fragmentation. Data are presented as mean pre- increase their mechanical stability to high air turbulence caused by dicted AGB values (black points) using linear mixed modelling and error bars strong winds in the forest edges29,30,41,42. These results suggest that denote 95% confidence intervals. 44 1-ha plots used for AGB predictions contained surviving trees had mechanisms of acclimation to stressful conditions tree measurements on more than 12,000 individual stems ≥10 cm diameter across near the forest edges - in particular higher temperatures, lower water 1026 tree species. The shaded dark grey area corresponds to AGB loss after 40 availability, stronger winds, with minimal effects on allocation of bio- years of fragmentation caused by changes in forest structure owing to edge effects mass to the aboveground compartments. However, a complementary on tree mortality, growth and recruitment. The red area corresponds to AGB loss explanation could be that edge effects selected for individuals or caused by edge effects on tree allometry, calculated by comparing AGB estimates species that were more adapted to the edge conditions. Indeed, forest using an allometric model that considered edge effects on tree allometry with an fragmentation has led to a high mortality rate of tall trees23, and the allometric model developed for interior forests. surviving trees may be those with traits that provide better fitness to the micro-environmental conditions near the forest edges. Better understanding the role of species and their intra-specific controls effects on forest structure—caused by fragmentation-related variation would help us predict variation in tree architecture with forest frag- in tree size, tree density and species composition—within 100 m from mentation in the Amazon. the forest margins; second, there was a 6.0 Mg ha−1 decline in AGB The new recruits that colonised the edges of the forest fragments caused by edge effects on tree allometry within 55 m from the forest were adapted to maximise light capture given their higher woody margins. These distance thresholds were chosen based on previous volume in the branches and trunks, higher asymmetry and higher path studies indicating stronger edge effects within 100 m from forest fraction. The forest edges had multi-stemmed trees with deeper edges on forest structure, and our own study indicating stronger edge crowns because of the high light availability near the edges17. effects on tree allometry within 55 m from the forest edges (see Short trees near the edges were more asymmetrical, which may help Methods for detailed explanations). We visually represented the pre- them capture more light by shifting their biomass towards the forest dicted reduction in AGB caused by edge effects on both forest struc- edges43. However, this may lead to a lower mechanical stability42, and ture and tree allometry, comparing them to control plots in interior trees must increase the volume of trunks prior to growing asymme- forests. Notably, tree allometry alone contributed to one-third of the trical crowns to avoid damage44, a strong predictor of tree mortality in total AGB decline resulting from edge effects (Fig. 4). fragmented Amazonian forests19. These colonising trees also produced numerous branches of high path lengths, which may lead to lower Discussion hydraulic efficiency, as these trees have longer paths to transport High density terrestrial LiDAR combined with long-term tree mea- water and nutrients for photosynthesis7. These architectural changes, surements across Central Amazonia provided a fresh perspective on alongside a large production of new leaves throughout the the architecture and allometry of trees and their associated impacts on year observed in the same forest edges28, may explain the increased the aboveground biomass of forest fragments. We found that edge growth rates of understory tree species in the forest edges of the effects on tree architecture impacted allometric relationships for BDFFP project in comparison to the same tree species growing in the predicting woody volume as a function of stem size, and despite the understory of interior forests45, despite the hot temperatures and large biomass allocation in the aboveground compartments of young lower water availability in the forest edges28. These results illustrate the trees, the lower height of tall surviving trees near the edges led to a ability of colonising trees in forest fragments to capture light and decline in the aboveground biomass of forest fragments. We estimated grow, which demonstrates the importance of protecting forest frag- an overall 10.0% reduction in AGB (24.7 Mg ha−1) owing to edge effects ments for carbon cycling. However, will these colonising trees be able on the forest structure (caused by mortality, growth and recruitment to grow tall and have high rates of survival in the future? These traits of trees, 18.7 Mg ha−1) and tree allometry (6.0 Mg ha−1). These numbers that maximise construction costs (i.e. high asymmetry and high path indicate that altered tree allometry alone accounted for one-third of all fraction) in the light-rich environments near the edges may also lead to biomass losses caused by edge effects on these forests in Central higher risks of mechanical damage and hydraulic failure. Additionally, Amazonia. These findings demonstrate the value of terrestrial LiDAR as trees grow taller, increasing leaf water stress due to gravity and path surveys to challenge some of the assumptions about how much carbon length resistance can limit leaf expansion and photosynthesis, and trees store, and to allow the detection of fine-scale changes in tree consequently limit further height growth46. Long-term monitoring of Nature Communications | (2023)14:8129 5 Article https://doi.org/10.1038/s41467-023-44004-5 tree architecture in controlled fragmentation experiments, including topography, can shape the species composition and forest structure, the BDFFP, is crucial to predict how these understory trees may survive and thus influence the recovery of fragmented forests. in fragmented forests when they grow older and, perhaps, taller. Edge effects on tree allometry further intensified AGB loss, con- Allometric models used to predict woody volume based on tributing to an additional 6.0 Mg ha−1 AGB reduction in Central Ama- stem diameter and tree height were affected by edge effects. zonian forests. This accounted for one-third of all AGB losses (24.7 Mg Short trees colonising the edges exhibited woody volume up to 50% ha−1) in these forests 40 years of forest fragmentation—a value pre- higher than trees with similar diameter at breast height (DBH) and viously unquantified in the literature. These values, however, could be height in the forest interior. However, although surviving tall trees conservative, considering that edge effects on tree allometry may for a given height and DBH were able to maintain their woody penetrate farther into the forest interior ( > 55 m) across other sites in volume under edge effects, some tall trees had a disproportionately Amazonia and lead to more significant negative impacts on forest lower height for a given DBH (Supplementary Figure 10). This AGB60. Given that fragment edges cover a total area of 176,555 km2 of resulted in a 30% decline in the woody volume of large trees, pos- Amazonian forests58, our findings of edge effects on tree architecture sibly caused by. the higher microclimatic stress, increased wind and allometry could translate into a substantial component of speeds, proliferating lianas, and the mortality of large trees and fragmentation-related carbon losses. branch loss23,28,47 that may have led to collateral damage to neigh- The decline in AGB near the edges presents an apparent contra- bouring trees, resulting in reduced heights near the forest diction with the higher Plant Area Index (PAI) found in these forest edges31,48,49. Alternatively, trees in the edges may have grown less in edges24. Specifically, Maeda and colleagues (2022) found higher PAI height, or edge effects may have favoured shorter individuals or values, a combination of leaf and wood surface areas, near the forest species acclimated to edge conditions, as the height-to-diameter edges due to a large increase in the density of small trees. This increase ratio is a strong determinant of mechanical safety50. Consequently, compensated for the reduction of upper canopy PAI caused by the a potential combination of stem breakage from damages caused by mortality and damage of tall trees. This contradiction can be explained neighbouring trees and selective forces favouring shorter trees for a by the disproportionate contribution of tall trees to the aboveground given DBH led to reduced woody volume in large trees. This aligns forest biomass63,64. Furthermore, the edges of these fragments are with the significant negative effects of human-induced disturbances dominated by pioneer species with acquisitive traits65, thriving under on tree biomass31. the light-rich environment owing to lateral light penetration and gap Variation in the woody volume of trees, coupled with variation in formation associated with the mortality of large trees66. Traits of pio- the wood density caused by shifts in species composition, determine neer species include a larger volume per mass unit (low wood density) the spatial patterns of AGB, which vary considerably across undis- and large foliar area67, which together increase the PAI of edge forests turbed Amazonian forests. Intact forests at a site about 15 km south of without a concomitant increase in AGB68. This highlights the challenge our study area, in similar lowland forests, were found to have mean faced by passive sensors onboard satellite platforms to capture a AGB values of 325.5 ± CI95% 13.6 Mg ha−1 in comparison to our estimates potential reduction in AGB in fragmented forests with increasing PAI. of 282.2 ± 15.3 Mg ha-151. This demonstrates that our estimates of AGB Despite our efforts to understand architectural variation in Cen- are conservative given their lower values relative to other estimates for tral Amazonian forests with edge effects, the mechanisms allowing the same region. Indeed, AGB estimates can vary significantly among species to change their architecture remain elusive. The within-plot equations, as the choice of allometric models and measurement variability, that can be seen as local effects arising from edge effects, as uncertainty can lead to uncertainties of up to 31% and 16%, respec- well as species-specific and ontogenetic influences, accounted for tively, in the AGB estimation of Amazonian forests52. Furthermore, most of the architectural variability. These results may help elucidate local environmental conditions that vary at small spatial scales, such as the generality of our findings in architectural variation across Central soil fertility, can account for a third of the variation in AGB in terra- Amazonia, although large-scale Amazonian gradients in topography, firme Amazonian forests53. edaphic properties and climate remain to be tested. Moreover, a shift Forests in edges tend to experience a dramatic decrease in AGB in species composition caused by edge effects may have partially caused by changes in forest structure, and are often structurally similar contributed to our observed changes in tree architecture65. We pro- to secondary forests54. We observed a reduction of 18.7 Mg ha−1 (6.6%) pose that future research on tree architecture should continue to in the AGB of forest fragments owing to changes in forest structure unravel the interactions of the environment with functional diversity within the first 100 m from the forest edges. This reduction resulted within species. TLS-based data, combined with molecular, genetic and from combined edge effects on tree mortality, growth and recruitment physiological processes regulating tree architecture, can help resolve that influence tree size, tree density and species composition. Forest debates concerning the mechanisms by which trees change their fragmentation increases the mortality of large trees22,55,56, but can also architecture1,69, which may enhance predictions of plant responses to lead to higher growth rates of large trees and the recruitment of new global changes. Understanding these mechanisms is crucial for pre- individuals57. However, large tree mortality represents a significant dicting the ability of plants to survive and grow with increasing forest proportion of biomass loss, contributing to the high carbon emissions fragmentation, and their contribution to the terrestrial carbon sink of fragmented ecosystems22,58,59. The observed reduction in AGB in our dynamics. study area near the edges was lower than anticipated, potentially due to the controlled conditions in the BDFFP that minimise additional Methods anthropogenic influences such as illegal logging, hunting, fire pene- Study site and sampling design tration and pollution60. Indeed, edge effects on forest structure in The study was conducted in Central Amazonian forests (2°20 30 ′S, 60° Amazonian forests can extend up to 1000 m, as fragments are more 05 37 W) near Manaus, Brazil, in the reserves of the Biological susceptible to recurring disturbances from fires and illegal logging, Dynamics of Forest Fragments Project (BDFFP), the world’s longest‐ and sensitive to the land use in the matrix surrounding the forest running experimental study of habitat fragmentation37. The region has fragment61. Moreover, edge effects can be stronger in the initial dec- seen notable carbon and biodiversity losses due to forest fragmenta- ades post-fragmentation, with up to 36% of the forest biomass lost in tion effects58,65. The BDFFP sites consist of forest fragments that were the first 10 to 17 years after fragmentation, potentially followed by a initially isolated in 1980-1983 following the conversion of the sur- recovery in the subsequent decades22,62. Long-term monitoring of edge rounding mature forest into cattle pastures. The project was specifi- and interior forest plots is crucial to investigate edge effects on AGB cally designed to investigate the relationship between the size of a dynamics, as confounding environmental factors, such as soil and forest fragment and its stability and functioning. The experimental Nature Communications | (2023)14:8129 6 Article https://doi.org/10.1038/s41467-023-44004-5 Forest fragments 02O 15 00” S Dimona 02O 20 00” S Porto Alegre Colosso 0 3000 m Florestal 02O 25 00” S 02O 30 00” S 60O 05 00” S 60O 00 00” S 59O 55 00” S 59O 50 00” S 59O 45 00” S 59O 40 00” S Fig. 5 | The Biological Dynamics of Forest Fragments Project (BDFFP), the forest into cattle pastures. The forest fragments are surrounded by a 100 m matrix, world’s longest‐running experimental study of habitat fragmentation, is regularly cleaned by cutting the regrowth vegetation to keep the forest fragment located in Central Amazonia. The BDFFP sites are composed of forest fragments isolated. originally isolated in 1980−1983 after the conversion of the surrounding mature design of BDFFP is based on comparisons of a replicated series of The scans covered three transects of 100 × 10 m and three transects of forest fragments or reserves of different sizes before and after their 50 × 10 m near the fragment edges and perpendicular to the forest isolation from continuous forests70. Currently, the matrix surrounding fragment margins (Fig. S1). We also included one transect of 30 × 10 m the forest fragments is dominated by secondary growth forests, but a length in the centre of the 100-ha forest fragment, which lies 500 m 100 m strip surrounding the forest fragments is cleared regularly by from any fragment margin to ensure sampling of forest interior in the cutting vegetation regrowth to maintain their isolation (Fig. 5). As an absence of edge effects on the tree architecture. experiment that minimises additional anthropogenic influences such To ensure a full 3D representation of the upper canopy (maximum as illegal logging, hunting, fire penetration and pollution, the project canopy height = 36 m), each transect consisted of three scan lines offers insights into ecological and environmental changes in frag- parallel to each other with scans spaced at 5 m intervals within and mented forests. between lines (Supplementary Fig. 1). The distance between scanning For this study, we investigate how the architecture of trees varies positions was smaller than the 10–40 m usually applied in previous with distance from the forest edge in four forest fragments across two studies to minimise data uncertainties due to occlusion in dense tro- regions in Central Amazonia (Dimona and Colosso; two of 1 ha, one of pical forests and maximise data acquisition in the upper canopy70. 10 ha, and one of 100 ha reserves). The maximum distance from forest Given that the RIEGL VZ-400i has a zenith angle range of 30–130°, an edges towards the forest interior varies between fragments, with 50 m additional scan was acquired at each sampling location with the for 1 ha fragments, 100 m for the 10-ha fragment and 500 m for 100 ha scanner tilted at 90° from the vertical position. A total of 1,188 scans the fragment. Compared to undisturbed forests in the fragment across all transects resulted in a complete sampling of the full hemi- interior, the forest near the edges is dominated by a higher density of sphere in each scan location (Supplementary Figure 1). All scans were early successional, fast-growing species because of elevated tree later co-registered into a single point cloud per transect using the mortality and seed dispersal from degraded neighbouring habitats37,65. RiSCAN PRO software version 2.9. Given that the RIEGL VZ-400i uses The edges of these fragments can have up to 5 °C higher temperatures onboard sensor data with an algorithm to align scans without the use than the forest interior during the dry season28. During periods of of reflectors, automatic registration was done before a final adjust- lower precipitation in the region, the edges of these fragments have ment of scans. More details on how the data was used to investigate lower soil moisture, which creates periods of higher evaporative edge effects on structure and dynamics can be found in Maeda and demand for tall trees in the edges28. The fragment edges have a sparser colleagues24 and Nunes and colleagues28, respectively. upper canopy marked by lower plant area index (PAI) above 15 m in vertical height, which creates a light-rich environment and higher light Individual tree segmentation and quantitative structural availability in the understory of forest edges24. modelling Individual tree segmentation used an automatic approach followed by Terrestrial laser scanning: data acquisition and pre-processing manual corrections. The automatic segmentation of individual trees The three-dimensional structure of trees was assessed based on point- was based on the shortest paths method by Raumonen and colleagues71. cloud data acquired using a state-of-the-art terrestrial laser scanning In this method, the shortest paths from the points to the base layer (TLS) system. The TLS data were acquired using a RIEGL VZ-400i sys- above the ground level are determined with restrictions on horizontal tem in April 2019 within the BDFFP permanent forest plots. We used a directions allowing better dealing with occluded regions and holding scan resolution of 40 mdeg in both azimuth and zenith directions, back paths going sideways into neighbouring trees. After the shortest which results in a point spacing of 34 mm at 50 m distance from the paths were determined, the trees are defined as all points connected via scanner. The laser pulse repetition rate used was 600 kHz, allowing a shortest paths to the same stem section close to ground. The stem measurement range of up to 350 m and up to eight returns per pulse. sections are iteratively defined around the base layer points that are Nature Communications | (2023)14:8129 7 Article https://doi.org/10.1038/s41467-023-44004-5 Table 1 | Traits, structural characteristics of trees with high trait values and their ecological meaning. This table summarises the information presented in more detail in the section “Tree architectural traits” of the Supplementary Methods 2 Trait Structural characteristics Ecological interpretation Branch/trunk surface area divided Trunk and branch thickness High trait values indicate a higher proportion of metabolically inactive wood6. by branch/trunk woody volume This is linked to lower support of aerial structures for light capture and pro- ductivity. On the other hand, higher values indicate higher interaction with the atmosphere, including higher respiration rates of the living tissues5. Relative crown width Horizontal crown size relative to DBH High trait values indicate trees that develop wide crowns, usually linked to limited access to light77,86, although it can be limited by competition from neighbours87. Shade-tolerant trees can also expand their crowns to maintain minimal leaf overlap for light capture and withstand falling debris88. Relative crown depth Vertical crown size relative to tree height, In high-light environments trees invest in deep crowns to better compete with often caused by multi-stemming neighbours17. Multi-stemming also leads to large relative crown depths, as our algorithm also considers the crown as of the upper segments from any multi- stemming point. Path fraction Umbrella-shaped tree crown Longer paths of trunks, branches and twigs prioritise sun exposure and light capture. This leads to longer water and nutrient transport distances that require high costs related to construction tissue7. High path fraction values represent umbrella-shaped crowns that prioritise sun exposure and light capture but are structurally expensive to build and are hydraulically less efficient10. Asymmetry Wood allocation in branches and trunks shif- Tree asymmetry is a result of competitive pressure from neighbouring trees78. ted relative to the trunk’s vertical axis Asymmetrical crowns can maximise capture of solar radiation by shifting their trunks and branches towards canopy gaps or away from their neighbours to avoid competition43. However, asymmetrical trees have lower mechanical stability and are more vulnerable to winds30, which is a major cause of mor- tality of fragmented Amazonian forests19. ranked by the number of shortest paths connected to the point. It trees that varied in DBH (2.4−72.3 cm), tree height (5.0−35.8 m) and started from the point with most paths, the stem section around was volume (0.01−13.0 m3) (Supplementary Fig. 2). first defined, and proceeded with the next highest ranked point not yet assigned to a tree. Considering a reverse J-shaped distribution of tree Tree architectural traits size of Amazonian forests - with a decreasing density of individuals with In this study, we propose that surface area per unit volume (m2 m−3) increasing diameter at breast height (DBH)72, we started the extraction of branches and trunk, asymmetry [-], path fraction [-] and relative of trees from the largest to smaller individuals within each transect to crown dimensions, including relative crown depth (m m−1) and ensure that most large trees within the study area were included. relative crown width (m cm−1), are traits that may be sensitive to However, some limits were imposed to the automatic approach, forest fragmentation, as they have been linked to maintenance costs including (1) the dense canopy with trees occupying different strata, (2) for respiration, mechanical stability, hydraulic conductance and the high diversity and variability in tree architecture, (3) the high light capture. Retrieval of architectural traits are described in detail abundance of lianas and (4) the noise in the data produced by wind in Supplementary Methods 2. The ecological meaning of each flapping. Manual corrections to remove lianas and to identify extra- architectural trait is synthesised in Table 1, and in more detail in neous and missing branches of trees represented a substantial fraction Supplementary Methods 2. of the processing time. Trees that were severely damaged or dead, as well as trees whose crowns could not be clearly discerned from those of Woody volume and allometric relationships lianas or neighbouring trees were excluded from our analysis. The accuracy of allometric models to predict aboveground biomass A total of 315 individual trees were segmented and extracted from is crucial for terrestrial carbon stock mapping. Total tree height co- the original TLS point clouds. A quantitative structure model (QSM) of varies with bioclimatic stress that depends on temperature and the woody structure, proposed by Raumonen and colleagues73, quan- precipitation variability76,77, but little is known how forest frag- titatively describes the topological, geometric and volumetric prop- mentation may affect these relationships. Locally derived DBH- erties of trees (Supplementary Fig. 1). A QSM consists of a hierarchical height relationships improve existing allometric equations77. Dif- collection of building blocks—usually geometric primitives such as ferences in wood allocation in branches caused by fragmentation- cylinders and cones—which are fitted to the point clouds to locally related environmental changes may lead to differences in equations approximate the woody parts. The use of circular cylinders in our that predict woody volume as a function of DBH and total tree study is a robust and accurate approach to estimate diameters, height78. In this paper, we investigated the effects of fragmentation lengths, directions, angles and volumes74. Fitting cylinders to smaller on allometric relationships between DBH2 H and woody volume, and branches, however, can lead to significant overestimation of the dia- also on allometric relationships between DBH and total woody meters and volumes stemming from the uncertainties of LiDAR volume, and discussed the implications of potential effects on the measurements75. To reduce these uncertainties related to branch size, mapping of terrestrial carbon in fragmented forests. Woody we followed Jackson and colleagues42 and trimmed off branches with volume, DBH and tree height were derived from the tree’s QSMs. diameters <2 cm from the QSMs. This procedure was reported to not The QSM method can provide accurate woody volume estimates affect the estimate of architectural traits across tropical forests42. with no systematic bias regardless of the tree structural character- QSMs provided estimates of branch size distribution, parent-child istics across tropical forests in Cameroon, Peru, Indonesia, and relations of the branches, diameters, lengths, angles, directions, and Guyana79,80. Therefore, nondestructive estimates of woody volume volumes, which are key properties that can be used to estimate from TLS can be a replacement for traditional sampling methods, architectural traits10. Details of parameters for QSM reconstruction can and for updating allometry and reducing uncertainties in landscape- be found in Supplementary Methods 1. The QSM database included level biomass estimates81. Nature Communications | (2023)14:8129 8 Article https://doi.org/10.1038/s41467-023-44004-5 Tree height to classify trees by time of establishment before or where β0 to β3 are the model parameters, μi is the random intercept for after forest fragmentation the nested effects of region i, fragment size i and plot identity i, and εi is Forest fragmentation may affect each architectural trait differently the normally distributed residual error. depending on the ontogenetic stage and tree size. As the fragment We then tested the influence of distance to edges on each archi- isolation occurred in 1980, the tall trees of our dataset are individuals tectural trait with distances to edges varying between 1 and 100 m, that likely have survived the fragmentation effects, and differences in including tree woody volume. We then tested the influence of distance their architecture with forest fragmentation may reflect the ability of to edges, varying from 1 to 100 m, on each architectural trait; we adult plants to acclimate to edge effects. On the other hand, short trees determined the edge effects extent for each architectural trait based can be included in three groups: (1) ~2/3 of these trees are less than 30 on the maximum absolute t-value of the term “edge effects” of the years old recruits that have colonised the area after the establishment model (Supplementary Method 4; Supplementary Fig. 4). The edge of the forest fragments; (2) surviving trees from understory species at effects extent was then used to categorise our analysis into edge versus varying ontogenetic stages; and (3) trees from slow-growing species, interior trees during all trait analyses. including upper canopy species at varying ontogenetic stages. Trees in We also examined the explained variance by the random variables group 1 have already been exposed to edge effects during their juve- of Eq.5 to investigate the spatial variability of architectural traits arising nile phase. Conversely, changes in the architecture of trees in groups 2 from region, landscape and plot (Supplementary Fig. 5). The LME and 3 reflect acclimation of short-stature adult individuals to edge models were fitted using the lme function in the “nlme” R package. effects that, analogous to the tall surviving trees, change their archi- tecture during the adult phase. Statistical modelling to predict edge effects on Long-term measurements from the BDFFP project since 1980 architectural traits revealed that, among all the trees that have been tagged and mon- We used nonlinear mixed-effects models of architectural traits of trees itored, all trees > 20 m in height have been growing in the BFFDP forest in forest edges and trees in forest interior as a function of tree height fragments before 1990 (Supplementary Methods 3; Supplementary (Equation 8). Allometric relationships between woody volume and DBH2 Fig. 3), which indicates that trees > 20 m in height are those that have H or between woody volume and DBH were fitted using linear mixed survived the fragmentation effects. Therefore, any change in the models after log-transforming the dependent and independent architecture of these tall trees must have occurred during their adult variables84 (Equations 9 and 10). We tested whether a quadratic term phase. Our data also indicated that ~ 33% of trees <20 m in height have should be included to increase goodness-of-fitness of the allometric also been growing in these forests since before 1990. This suggests models by comparing each model’s AIC (Supplementary Table 1). For all that 66% of the small trees <20 m in height are new, young recruits (i.e., the models, plot identity nested within landscape (position of fragment trees that have colonised these fragments within the last 30 years) that within the landscape and fragment size) and region within Central were exposed to edge effects already in their juvenile phase. We thus Amazonia were included as random variables, allowing us to include any used 20 m tree height as a threshold separating colonising short trees idiosyncratic differences between plots, fragments and regions. The ( < 20 m) and surviving adult trees ( > 20 m). models were fitted using the nlme function for the nonlinear mixed- effects model and lme function for the linear mixed-effects model in the Determining edge effects extent on architectural traits “nlme” R package. Performance of the final models was also evaluated Long-term studies from the BDFFP have demonstrated that edge effects using an 80/20 split of the data for calibration and validation, respec- vary widely in extent, with biophysical changes that can penetrate from tively, over 200 randomised permutations of the dataset. These ana- 20 m to 300 m37. To estimate the extent of edge effects on each lyses generated a distribution of model coefficients and allowed an architectural trait, we used mixed linear models (LME, Eq. 1) that con- assessment of model stability and uncertainty of predictions. tained a variable representing the plot category of location near an edge or in the forest fragment interior (edge effects), following Nunes and β1 Architectural trait = β0 Tree Height + μi + εi ð8Þ colleagues82. The model also included a variable that represented tree height (H), considering that architectural traits co-vary with tree size and ontogenetic stage10,83. Edge effects and tree height were treated as 2 ð9Þ lnðWoody volumeÞ = β0 + β1 lnðDBH2 HÞ + β2 lnðDBH2 HÞ + μi + εi additive terms to examine the significance of fragmentation and tree height on the variation of architectural traits. We also included an interaction term edge effects × tree height, as fragmentation may have lnðWoody volumeÞ = β0 + β lnðDBHÞ + β2 lnðDBHÞ2 + μi + εi ð10Þ different effects on trees of different heights (Eq. 5). To examine the influences of distance from edges on the allometric relationships between woody volume and DBH2 H or woody volume and DBH, we where β0 to β2 are the model parameters, μi is the random intercept for tested how the log transformed variables interacted with edge effects the nested effects of region i, fragment size i and plot identity i, and εi is (Eq. 6 and Eq. 7). Nested effects of forest site (Colosso versus Dimona the normally distributed residual error. Let DBH in cm, H in m and sites), fragment size (1, 10 or 100 ha) and plot (or transect) identity were woody volume in m3. treated as random variables (μ), allowing us to account for the nested spatial variation in architectural traits and to include any idiosyncratic AGB calculation differences between forest site, fragment size and micro-environmental Permanent plots within the BDFFP were distributed across a large area variation (i.e., topography, soil) between plots. of 1000 km2 in Central Amazonia (Fig. 6). We used 44 permanent plots of 1 ha each in edge (28 ha) and interior forests (16 ha) to compile Architectural trait = β0 + β1 × ðedge effectsÞ + β2 × H + β3 aboveground biomass (AGB) using data from Dimona, Colosso, Flor- ð5Þ estal and Porto Alegre. The project was designed specifically to × ðedge effectsÞ × H + μi + εi investigate the relationship between the size of a forest fragment and its stability and functioning. The experimental design of BDFFP is lnðwoody volumeÞ = β0 + β1 × ðedge effectsÞ × lnðDBH2 HÞ + μi + εi ð6Þ based on comparisons of a replicated series of forest fragments or reserves of different sizes before and after they were isolated from continuous forests. We then examined how edge effects on tree allo- lnðwoody volumeÞ = β0 + β1 × ðedge effectsÞ × lnðDBHÞ + μi + εi ð7Þ metry control AGB relative to edge effects on forest structure (for Nature Communications | (2023)14:8129 9 Article https://doi.org/10.1038/s41467-023-44004-5 Florestal Dimona Edge 1-ha permanent plots Interior 1-ha permanent plots TLS transects Colosso Porto Alegre Fig. 6 | Allometric models for woody volume estimates were developed from Florestal, Dimona, Colosso and Porto Alegre were used to estimate the above- tree measurements using a Terrestrial Laser Scanner (TLS) within the Biolo- ground biomass (AGB) in edge (yellow plots) and interior (white plots) forest plots. gical Dynamics of Forest Fragments Project (BDFFP) in Central Amazonia. Six Our findings demonstrate that edge effects on tree allometry penetrated 55 m from transects (five of 100 × 10 m at the edges and one of 30 × 10 m in the forest interior), the forest edges. Comparisons of AGB values using edge- versus interior-specific denoted as black lines in the Colosso and Dimona sites, allowed us obtain high allometric equations were made to predict the influences of tree allometry on the quality point clouds to segment and extract 315 trees. Tree measurements of stem forest AGB. size combined with species identification across 44 1-ha permanent plots within example, edges effects on tree density and distribution of tree size 100 m from the forest edges within the BDFFP22,55. Forest site (Colosso, caused by variations in tree mortality, growth, and recruitment). Dimona, Florestal e Porto Alegre) was considered a random variable, We used diameter at breast height (DBH), defined as 1.3 m from accommodating any unique differences between forest sites that the base of the stem, from 12,112 tree stems. For buttressed stems or could impact species composition and forest structure, such as soil other deformities, the point of measurement is raised above the and topography. Differences in predicted AGB values between edge deformity. All trees were distributed in 66 families, 254 genera, and and interior forests using allometric equation 3 (for interior forests) 1026 species. Plants were identified at the species level, or at least at represented variations in forest structure, subsequently affecting for- the genus level. We then converted diameter measurements to woody est AGB. Comparing predicted AGB values in forest edge and interior volume estimates using the allometric equations 3 (interior forests) forests using their respective allometric models (Equation 3 for inter- and 4 (edge forests), which predict woody volume as a function of DBH ior forests and Equation 4 for edge forests) enabled us to capture AGB only. These models capture variation in woody volume related to tree variations caused by edge effects on both forest structure and tree height (i.e. tree damage and breakage) and fine-scale architectural allometry. We calculated 95% confidence intervals based on the variation (i.e. branch production, branch loss, branch and trunk uncertainty in model parameters. The linear mixed effects (LME) thickening etc). To convert woody volume values to tree-level AGB, we models were fitted using the lme function in the “nlme” R package. used wood density values extracted from a global wood density database (https://datadryad.org/stash/dataset/doi:10.5061/dryad. Reporting summary 234)85, using the BIOMASS package in R. In cases where a stem was Further information on research design is available in the Nature unidentified or where no taxon-specific wood density data were Portfolio Reporting Summary linked to this article. available, we applied the appropriate genus or family-specific wood density values. Aboveground biomass for each 1-ha plot was estimated Data availability as the sum of the tree-level AGB. The architectural traits data generated in this study have been We employed linear mixed modeling to analyze AGB values, deposited in the national Finnish Fairdata services database under treating a categorical variable as a fixed effect indicating whether plots accession code https://etsin.fairdata.fi/dataset/c71785e6-58ab-4316- were near an edge ( < 100 m from the forest margins) or in the interior a0fc-28236c76bbe7. The point clouds of individual trees can be ( > 100 m from the forest margins). This approach allowed us to esti- available by contacting the corresponding authors upon reasonable mate AGB variation resulting from edge effects. We chose a 100 m request within 10 days. Measurements of all trees and their corre- threshold to differentiate between edge and interior forest plots due sponding species identification at the plot level are products of the to the higher mortality and turnover rates observed within the first long-term Biological Dynamics of Forest Fragments Project. The Nature Communications | (2023)14:8129 10 Article https://doi.org/10.1038/s41467-023-44004-5 interested party should contact William Laurance (bill.- 21. Valladares, F. & Niinemets, Ü. The architecture of plant crowns: laurance@jcu.edu.au), Ana Andrade (asegalin@gmail.com) and From design rules to light capture and performance. in Functional Alberto Vicentini (vicentini.beto@gmail.com) for potential access to Plant Ecology 101–150 (CRC Press, 2007). https://doi.org/10.1201/ these data. 9781420007626-4 22. Laurance, W. F. et al. Biomass collapse in amazonian forest frag- References ments. Science 278, 1117–1118 (1997). 1. West, G. B., Enquist, B. J. & Brown, J. H. A general quantitative theory 23. Laurance, W. F., Delamônica, P., Laurance, S. G., Vasconcelos, H. L. of forest structure and dynamics. Proc. Natl Acad. Sci. USA 106, & Lovejoy, T. E. Rainforest fragmentation kills big trees. Nature 404, 7040–7045 (2009). 836 (2000). 2. Lines, E. R., Fischer, F. J., Owen, H. J. F. & Jucker, T. The shape of 24. Maeda, E. E. et al. Shifts in structural diversity of Amazonian forest trees: reimagining forest ecology in three dimensions with remote edges detected using terrestrial laser scanning. Remote Sens. sensing. J. Ecol. 110, 1730–1745 (2022). Environ. 271, 112895 (2022). 3. Sheil, D., Salim, A., Chave, J., Vanclay, J. & Hawthorne, W. D. 25. Poorter, L., Bongers, F., Sterck, F. J. & Wöll, H. Architecture of 53 Illumination-size relationships of 109 coexisting tropical forest tree Rain Forest tree species differing in adult stature and shade toler- species. J. Ecol. 94, 494–507 (2006). ance. Ecology 84, 602–608 (2003). 4. Robertson, A. L. et al. Stem respiration in tropical forests along an 26. Poorter, L., Bongers, L. & Bongers, F. Architecture of 54 moist-forest elevation gradient in the Amazon and Andes. Glob. Chang. Biol. 16, tree species: traits, trade-offs, and functional groups. Ecology 87, 3193–3204 (2010). 1289–1301 (2006). 5. Trumbore, SusanE. Angert, Alon, Kunert, Norbert, Muhr, Jan & 27. Camargo, J. L. C. & Kapos, V. Complex edge effects on soil moisture Chambers, JeffreyQ. What’s the flux? Unraveling how CO2 fluxes and microclimate in central Amazonian forest. J. Trop. Ecol. 11, from trees reflect underlying physiological processes. N. Phytol. 205–221 (1995). 197, 353–355 (2013). 28. Nunes, M. H. et al. Forest fragmentation impacts the seasonality of 6. Reich, P. B. et al. Temperature drives global patterns in forest bio- Amazonian evergreen canopies. Nat. Commun. 13, 917 (2022). mass distribution in leaves, stems, and roots. Proc. Natl Acad. Sci. 29. Laurance, W. F. & Curran, T. J. Impacts of wind disturbance on USA 111, 13721–13726 (2014). fragmented tropical forests: a review and synthesis. Austral Ecol. 7. Smith, D. D. et al. Deviation from symmetrically self-similar 33, 399–408 (2008). branching in trees predicts altered hydraulics, mechanics, 30. Jackson, T. et al. A new architectural perspective on wind damage in light interception and metabolic scaling. N. Phytol. 201, a natural forest. Front. For. Glob. Chang. 1, (2019). 217–229 (2014). 31. Zuleta, D. et al. Damage to living trees contributes to almost half of 8. Owen, H. J. F., Flynn, W. R. M. & Lines, E. R. Competitive drivers of the biomass losses in tropical forests. Glob. Chang. Biol. https://doi. interspecific deviations of crown morphology from theoretical org/10.1111/gcb.16687 (2023) predictions measured with Terrestrial Laser Scanning. J. Ecol. 109, 32. Poorter, L. Are species adapted to their regeneration niche, adult 2612–2628 (2021). niche, or both? Am. Nat. 169, 433–442 (2007). 9. Goodman, R. C., Phillips, O. L. & Baker, T. R. The importance of 33. Coelho de Souza, F. et al. Evolutionary heritage influences Amazon crown dimensions to improve tropical tree biomass estimates. Ecol. tree ecology. Proc. Biol. Sci. 283, (2016). Appl. 24, 680–698 (2014). 34. Calders, K. et al. Laser scanning reveals potential underestimation 10. Malhi, Y. et al. New perspectives on the ecology of tree structure of biomass carbon in temperate forest. Ecol. Sol. Evidence 3, (2022). and tree communities through terrestrial laser scanning. Interface 35. Disney, M. Terrestrial LiDAR: a three-dimensional revolution in how Focus 8, 20170052 (2018). we look at trees. N. Phytol. 222, 1736–1741 (2019). 11. Burt, A. et al. New insights into large tropical tree mass and struc- 36. Stovall, A. E. L., Vorster, A., Anderson, R. & Evangelista, P. Devel- ture from direct harvest and terrestrial lidar. R. Soc. Open Sci. 8, oping nondestructive species‐specific tree allometry with terres- 201458 (2021). trial laser scanning. Methods Ecol. Evol. https://doi.org/10.1111/ 12. Sussex, I. M. & Kerk, N. M. The evolution of plant architecture. Curr. 2041-210x.14027 (2022) Opin. Plant Biol. 4, 33–37 (2001). 37. Laurance, W. F. et al. An Amazonian rainforest and its fragments as a 13. Negrón-Juárez, R. I. et al. Widespread Amazon forest tree mortality laboratory of global change. Biol. Rev. Camb. Philos. Soc. 93, from a single cross-basin squall line event. Geophys. Res. Lett. 223–247 (2018). 37, (2010). 38. Paciorek, C. J., Condit, R., Hubbell, S. P. & Foster, R. B. The demo- 14. Hollender, C. A. & Dardick, C. Molecular basis of angiosperm tree graphics of resprouting in tree and shrub species of a moist tropical architecture. N. Phytol. 206, 541–556 (2015). forest. J. Ecol. 88, 765–777 (2000). 15. Reis, C. R. et al. Forest disturbance and growth processes are 39. McDowell, N. et al. Mechanisms of plant survival and mortality reflected in the geographical distribution of large canopy gaps during drought: why do some plants survive while others succumb across the Brazilian Amazon. J. Ecol. 110, 2971–2983 (2022). to drought? N. Phytol. 178, 719–739 (2008). 16. Chambers, J. Q. et al. Respiration from a tropical forest ecosystem: 40. Olson, M. E. et al. Plant height and hydraulic vulnerability to drought partitioning of sources and low carbon use efficiency. Ecol. Appl. and cold. Proc. Natl Acad. Sci. USA 115, 7551–7556 (2018). 14, 72–88 (2004). 41. Vovides, A. G. et al. Change in drivers of mangrove crown dis- 17. Laurans, M. & Vincent, G. Are inter- and intraspecific variations of placement along a salinity stress gradient. Funct. Ecol. 32, sapling crown traits consistent with a strategy promoting light 2753–2765 (2018). capture in tropical moist forest? Ann. Bot. 118, 983–996 (2016). 42. Jackson, T. et al. An architectural understanding of natural sway 18. Ribeiro, G. H. P. M. et al. Mechanical vulnerability and resistance to frequencies in trees. J. R. Soc. Interface 16, 20190116 (2019). snapping and uprooting for Central Amazon tree species. Ecol. 43. Olivier, M.-D., Robert, S. & Fournier, R. A. Response of sugar maple Manag. 380, 1–10 (2016). (Acer saccharum, Marsh.) tree crown structure to competition in 19. Reis, S. M. et al. Climate and crown damage drive tree mortality in pure versus mixed stands. Ecol. Manag. 374, 20–32 (2016). southern Amazonian edge forests. J. Ecol. 110, 876–888 (2022). 44. Kunz, M. et al. Neighbour species richness and local structural 20. Trumbore, S. Carbon respired by terrestrial ecosystems - recent variability modulate aboveground allocation patterns and crown progress and challenges. Glob. Chang. Biol. 12, 141–153 (2006). morphology of individual trees. Ecol. Lett. 22, 2130–2140 (2019). Nature Communications | (2023)14:8129 11 Article https://doi.org/10.1038/s41467-023-44004-5 45. Albiero-Júnior, A., Venegas-González, A., Camargo, J. L. C., Roig, F. 68. Savage, V. M. et al. Hydraulic trade-offs and space filling enable A. & Tomazello-Filho, M. Amazon forest fragmentation and edge better predictions of vascular structure and function in plants. Proc. effects temporarily favored understory and midstory tree growth. Natl Acad. Sci. USA 107, 22722–22727 (2010). Trees (2021) https://doi.org/10.1007/s00468-021-02172−1. 69. Bierregaard, R. O., Lovejoy, T. E. & Kapos, V. Angelo Augusto dos 46. Koch, G. W., Sillett, S. C., Jennings, G. M. & Davis, S. D. The limits to Santos & Hutchings, R. W. The Biological Dynamics of Tropical tree height. Nature 428, 851–854 (2004). Rainforest Fragments. Bioscience 42, 859–866 (1992). 47. D’Angelo, S. A., Andrade, A. C. S., Laurance, S. G., Laurance, W. F. & 70. Wilkes, P. et al. Data acquisition considerations for Terrestrial Laser Mesquita, R. C. G. Inferred causes of tree mortality in fragmented Scanning of forest plots. Remote Sens. Environ. 196, 140–153 (2017). and intact Amazonian forests. J. Trop. Ecol. 20, 243–246 71. Raumonen, P., Brede, B., Lau, A. & Bartholomeus, H. A Shortest Path (2004). Based Tree Isolation Method for UAV LiDAR Data. in 2021 IEEE Inter- 48. Chambers, J. Q., Santos dos, J., Ribeiro, R. J. & Higuchi, N. Tree national Geoscience and Remote Sensing Symposium IGARSS damage, allometric relationships, and above-ground net primary 724–727 https://doi.org/10.1109/IGARSS47720.2021.9554346 (2021). production in central Amazon forest. Ecol. Manag. 152, 72. Vieira, S. et al. Forest structure and carbon dynamics in Amazonian 73–84 (2001). tropical rain forests. Oecologia 140, 468–479 (2004). 49. Leitold, V. et al. El Niño drought increased canopy turnover in 73. Raumonen, P. et al. Fast automatic precision tree models from Amazon forests. N. Phytol. 219, 959–971 (2018). terrestrial laser scanner data. Remote Sens. 5, 491–520 (2013). 50. van Gelder, H. A., Poorter, L. & Sterck, F. J. Wood mechanics, allo- 74. Åkerblom, M., Raumonen, P., Kaasalainen, M. & Casella, E. Analysis metry, and life-history variation in a tropical rain forest tree com- of geometric primitives in quantitative structure models of tree munity. N. Phytol. 171, 367–378 (2006). stems. Remote Sens. 7, 4581–4603 (2015). 51. Nascimento, H. E. M. & Laurance, W. F. Total aboveground biomass 75. Demol, M. et al. Volumetric overestimation of small branches in 3D in central Amazonian rainforests: a landscape-scale study. Ecol. reconstructions of Fraxinus excelsior. Silva Fenn. 56, (2022). Manag. 168, 311–321 (2002). 76. Lines, E. R., Zavala, M. A., Purves, D. W. & Coomes, D. A. Predictable 52. Chave, J. et al. Error propagation and scaling for tropical forest changes in aboveground allometry of trees along gradients of biomass estimates. Philos. Trans. R. Soc. Lond. B Biol. Sci. 359, temperature, aridity and competition. Glob. Ecol. Biogeogr. 21, 409–420 (2004). 1017–1028 (2012). 53. Laurance, W. F. et al. Relationship between soils and Amazon forest 77. Chave, J. et al. Improved allometric models to estimate the biomass: a landscape-scale study. Ecol. Manag. 118, 127–138 (1999). aboveground biomass of tropical trees. Glob. Chang. Biol. 20, 54. Berenguer, E. et al. A large-scale field assessment of carbon stocks 3177–3190 (2014). in human-modified tropical forests. Glob. Chang. Biol. 20, 78. Martin-Ducup, O. et al. Terrestrial laser scanning reveals con- 3713–3726 (2014). vergence of tree architecture with increasingly dominant crown 55. Laurance, W. F., Ferreira, L. V., Rankin-de Merona, J. M. & Laurance, canopy position. Funct. Ecol. 34, 2442–2452 (2020). S. G. Rain Forest fragmentation and the dynamics of amazonian tree 79. Takoudjou, S. M. et al. Using terrestrial laser scanning data to esti- communities. Ecology 79, 2032–2040 (1998). mate large tropical trees biomass and calibrate allometric models: 56. Laurance, W. F., Laurance, S. G. & Delamonica, P. Tropical forest A comparison with traditional destructive approach. Methods Ecol. fragmentation and greenhouse gas emissions. Ecol. Manag. 110, Evolution 9, 799–1156 (2018). 173–180 (1998). 80. Gonzalez de Tanago, J. et al. Estimation of above‐ground biomass of 57. Nascimento, H. E. M. & Laurance, W. F. Biomass dynamics in ama- large tropical trees with terrestrial LiDAR. Methods Ecol. Evol. 9, zonian forest fragments. Ecol. Appl. 14, 127–138 (2004). 223–234 (2018). 58. Silva Junior, C. H. L. et al. Persistent collapse of biomass in Ama- 81. Stovall, A. E. L., Anderson-Teixeira, K. J. & Shugart, H. H. Assessing zonian forest edges following deforestation leads to unaccounted terrestrial laser scanning for developing non-destructive biomass carbon losses. Sci Adv 6, (2020). allometry. For. Ecol. Manage. 427, 217–229 (2018). 59. Qin, Y. et al. Carbon loss from forest degradation exceeds that from 82. Nunes, M. H. et al. Recovery of logged forest fragments in a human- deforestation in the Brazilian Amazon. Nat. Clim. Chang. 11, modified tropical landscape during the 2015−16 El Niño. Nat. 442–448 (2021). Commun. 12, 1526 (2021). 60. Laurance, W. F. Theory meets reality: How habitat fragmentation 83. Verbeeck, H. et al. Time for a plant structural economics spectrum. research has transcended island biogeographic theory. Biol. Con- Front. For. Glob. Chang. 2, (2019). serv. 141, 1731–1744 (2008). 84. Chave, J. et al. Tree allometry and improved estimation of carbon 61. Broadbent, E. N. et al. Forest fragmentation and edge effects from stocks and balance in tropical forests. Oecologia 145, 87–99 (2005). deforestation and selective logging in the Brazilian Amazon. Biol. 85. Zanne, A. E. et al. Towards a worldwide wood economics spectrum. Conserv. 141, 1745–1757 (2008). https://doi.org/10.5061/dryad.234 (2009) 62. Almeida, D. R. A. et al. Persistent effects of fragmentation on tro- 86. Iida, Y. et al. Linking size-dependent growth and mortality with pical rainforest canopy structure after 20 yr of isolation. Ecol. Appl. architectural traits across 145 co-occurring tropical tree species. 29, e01952 (2019). Ecology 95, 353–363 (2014). 63. Fauset, S. et al. Hyperdominance in Amazonian forest carbon 87. Valladares, F., Skillman, J. B. & Pearcy, R. W. Convergence in light cycling. Nat. Commun. 6, 6857 (2015). capture efficiencies among tropical forest understory plants with 64. Gora, E. M. & Esquivel-Muelbert, A. Implications of size-dependent contrasting crown architectures: a case of morphological com- tree mortality for tropical forest carbon dynamics. Nat. Plants 7, pensation. Am. J. Bot. 89, 1275–1284 (2002). 384–391 (2021). 88. Seidel, D., Leuschner, C., Müller, A. & Krause, B. Crown plasticity in 65. Laurance, W. F. et al. Rain forest fragmentation and the proliferation mixed forests—Quantifying asymmetry as a measure of competi- of successional trees. Ecology 87, 469–482 (2006). tion using terrestrial laser scanning. Ecol. Manag. 261, 66. Lohbeck, M. et al. Successional changes in functional composition 2123–2132 (2011). contrast for dry and wet tropical forest. Ecology 94, 1211–1216 (2013). Acknowledgements 67. Malhado, A. C. M. et al. Spatial trends in leaf size of Amazonian This study was funded by the Academy of Finland (decision numbers rainforest trees. Biogeosciences 6, 1563–1576 (2009). 318252, 319905, and 345472 to Eduardo Eiji Maeda). This publication is Nature Communications | (2023)14:8129 12 Article https://doi.org/10.1038/s41467-023-44004-5 number XXX of the Technical Series of the Biological Dynamics of Forest Correspondence and requests for materials should be addressed to Fragment Forest (BDFFP—INPA/STRI). We thank the Biological Dynamics Matheus Henrique Nunes or Eduardo Eiji Maeda. of Forest Fragments Project for the great logistical support in the field and the research assistance. Jérôme Chave acknowledges “Investisse- Peer review information Nature Communications thanks Christopher ment d’Avenir” grants (CEBA, ref. ANR-10-LABX-25-01; TULIP, ref. ANR- Hakkenberg, and the other, anonymous, reviewer(s) for their contribu- 10-LABX-0041). Jin Wu was supported by the National Natural Science tion to the peer review of this work. Foundation of China (#31922090), the Innovation and Technology Fund (funding support to State Key Laboratories in Hong Kong of Agrobio- Reprints and permissions information is available at technology) of the HKSAR, China, and the Hung Hing Ying Physical http://www.nature.com/reprints Sciences Research Fund. Josep Peñuelas was supported by the TED2021-132627B-I00 grant, funded by MCIN and the European Union Publisher’s note Springer Nature remains neutral with regard to jur- NextGeneration EU/PRTR, and the CIVP20A6621 grant funded by the isdictional claims in published maps and institutional affiliations. Fundación Ramón Areces. Gabriela Zuquim acknowledges the Academy of Finland (grant number 344733) and Danish Council for Independent Open Access This article is licensed under a Creative Commons Research—Natural Sciences (grant number 9040-00136B). Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as Author contributions long as you give appropriate credit to the original author(s) and the M.H.N. and E.E.M. conceived the study. M.H.N. and E.E.M. led TLS data source, provide a link to the Creative Commons license, and indicate if collection in the field. J.L.C.C., W.F.L., A.A., A.V., and S.L. collected tree changes were made. The images or other third party material in this measurement data and identified tree specimens. M.H.N., P.R., and article are included in the article’s Creative Commons license, unless E.E.M. processed the LiDAR data. M.H.N. performed data analyses and indicated otherwise in a credit line to the material. If material is not wrote the manuscript. M.H.N., M.C.V, J.L.C.C., W.F.L., P.R., T.J., G.Z., J.W., included in the article’s Creative Commons license and your intended J.P., J.C. and E.E.M. contributed to the revision of the paper. use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright Competing interests holder. To view a copy of this license, visit http://creativecommons.org/ The authors declare no competing interests. licenses/by/4.0/. Additional information This is a U.S. Government work and not under copyright protection in the Supplementary information The online version contains US; foreign copyright protection may apply 2023 supplementary material available at https://doi.org/10.1038/s41467-023-44004-5. Nature Communications | (2023)14:8129 13