Biology subjects

Coomes, D.

Publications and source records attributed to Coomes, D..

15 recordsLinked to original sources

bioRxiv2026

Simplifying complexity: the reliability of lidar-derived forest structural complexity metrics for conservation management

Abstract 1. Forest structural complexity is regarded as a driver of biodiversity. The restructuring of forests to increase structural complexity is now a common objective of forest restoration and conservation projects. Lidar provides an efficient means of monitoring forest structure; however, the reliability of lidar-derived structural complexity metrics remains uncertain. Most metrics require a choice of resolution (such as pixel or voxel size), and whilst metric outcomes are sensitive to this choice, an understanding of how this impacts subsequent ecological interpretations of forest restoration monitoring is missing. 2. We assess how metric resolution choices impact FSC estimates using terrestrial laser scanning data from 32 single-species forest plots in Cairngorms National Park, Scotland. These plots represent four contrasting structural categories: ancient woodland (AW), old plantation (OP), old plantation with restructuring management (OPM) and natural restoration (NR); with a hypothesised order of complexity: AW, OPM, OP, NR (from highest complexity to lowest). We computed 21 FSC metrics and calculated how each metric ranked our forest structural categories across a resolution gradient (0.02-1.5m). We evaluated the stability of metric rankings across resolution values, their ability to discriminate forest categories, and the consistency of ecological interpretation of forest management targets. 3. Nine metrics produced unstable rankings, indicating that ecological conclusions were sensitive to resolution selection. Stable metrics discriminated well between categories with obvious structural differences but failed to discriminate between categories with subtle differences, particularly OP and OPM. 4. AW did not consistently exhibit the highest structural complexity, instead occupying an intermediate position characterised by a combination of canopy occupancy, openness and height inequality, not captured by any single metric. 5. We conclude that methodological choices can influence FSC estimates and ecological conclusions drawn from them. 6. Synthesis and applications: Reliable assessment of forest restoration depends not only on the choice of metric but also on how that metric is parameterised. We recommend transparent reporting and justification of resolution parameter selection together with the use of stable metrics that capture structural characteristics relevant to management objectives being monitored, to improve reliability and comparability of lidar-based forest monitoring.

ecology↗

bioRxiv2026

Liana cutting accelerates tropical forest recovery at a fraction of the cost of tree planting

We urgently need to restore degraded tropical forests to mitigate the climate and biodiversity crises, but how to do so rapidly and cost-effectively remains an open question. Here we provide a long-term, landscape-scale assessment of the effectiveness of enrichment tree planting and liana cutting, the two most common restoration interventions used across many tropical regions. Leveraging one of the worlds largest and longest running forest restoration experiments, we used repeat airborne laser scanning to track the 3D structural recovery of 500 ha of selectively logged rainforest in Borneo. Over an 18-year period, enrichment planting increased mean canopy height by 1.6 m relative to unplanted controls. Remarkably, liana cutting increased canopy height more than four times faster (3.7 m over just 9 years). This recovery was jointly driven by accelerated canopy gap closure, enhanced tree growth, and a 50% reduction in tree mortality. Given that liana cutting is around 10 times cheaper to implement than enrichment planting, our results suggest it provides a cost-effective, scalable solution to accelerate the structural recovery of logged tropical forests.

ecology↗

bioRxiv2026

Geospatial foundation models enable data-efficient tree species mapping in temperate montane forests

Accurate mapping of tree species from satellite data remains challenging in heterogeneous mountain forests due to environmental gradients, mixed stands, limited availability of high-purity training labels, and strong illumination-angle effects. Recent geospatial foundation models offer a new approach by learning generic, cloud-agnostic, information-rich representations from large multi-sensor archives suitable for a range of downstream tasks, but their ecological utility for species-level mapping remains incompletely understood. Here, we evaluate two geospatial foundation-model embeddings, AlphaEarth and Tessera, for tree species classification in the Trentino region of northern Italy, using parcel-level forest inventories as reference data (18 species and species groups). We compare their performance against conventional Sentinel-1+2 satellite composites across a series of controlled experiments examining classification accuracy, label efficiency, classifier complexity, robustness to label impurity, and temporal transferability. Foundation-model embeddings consistently outperform composite-based multispectral satellite baselines (weighted F1 = 0.83 vs. 0.80; macro F1 = 0.55 vs. 0.50), reaching near-asymptotic accuracy with as few as 5% of available training parcels and preserving ecologically meaningful structure aligned with functional and taxonomic groupings. However, realising this advantage requires a nonlinear classifier: a compact neural network provides better results than classic machine learning (i.e. Random Forest) and performs as well as deeper neural networks, while a linear classifier on foundation-model embeddings underperforms a neural network on conventional composites. Ancillary environmental covariates offer no additional classification benefit when added to embedding-based models. Classification accuracy remains robust to moderate levels of label impurity, allowing mixed parcels to be retained in the training dataset without substantial penalties, while training with parcel-level species proportions as soft labels achieves higher peak performance (macro F1 = 0.586 for Tessera, 0.589 for AlphaEarth) and lower Proportion L1 error than hard labels without requiring purity filtering, maximising the value of the full range of input data. However, temporal transfer across years reveals performance degradation, with weighted F1 declining by 9% for Tessera and 15% for AlphaEarth, and disproportionate losses for rare species. Overall, our results show that geospatial foundation models shift a primary bottleneck in species mapping from feature engineering toward the availability, quality, and temporal alignment of ecological reference data, while opening new opportunities for scalable biodiversity monitoring and the analysis of ecological change.

ecology↗

bioRxiv2025

The Global Canopy Atlas: analysis-ready maps of 3D structure for the world's woody ecosystems

Woody canopies regulate exchanges of energy, water and carbon, and their three-dimensional (3D) structure supports much of terrestrial biodiversity. Remote sensing technologies such as airborne laser scanning (ALS) now enable the 3D mapping of entire landscapes. However, we lack the large, harmonized and geographically representative ALS collections needed to build a global picture of woody ecosystem structure. To address this challenge, we developed the Global Canopy Atlas (GCA): 3,458 ALS acquisitions transformed into standardized and analysis-ready maps of canopy height and elevation at 1 m2 resolution. The GCA covers 56,554 km2 across all major biomes. 19% of this area has been scanned multiple times, and 87% of all GCA products are openly available, covering 95% of the total area. To showcase its wide range of applications, we applied the GCA in three case studies. First, we validated three global satellite-derived canopy height maps, finding poor performance at native resolution (1-30 m, R2 < 0.38) and moderate performance at 250 m resolution (R2 < 0.65). Second, analyzing global patterns in canopy gap size frequency we discovered an unexpectedly large variation of power law exponents from branch to stand level ( = 1.52 to 2.38), pointing to a fundamental scale-dependence of forest structure. Third, we developed a framework to standardize forest turnover quantification from multi-source, multi-temporal ALS. In a temperate forest in North America it revealed that 21% of canopy gaps closed within 12 years of opening and would thus be missed by infrequent monitoring. As demonstrated by these case studies, the GCA provides a novel data source for ecologists, foresters, remote sensing scientists and the ecosystem modelling community that substantially advances our ability to understand the structure and dynamics of woody ecosystems at global scales.

ecology↗

bioRxiv2025

A global map of wood density

Wood density influences how quickly woody plants grow, how long they live and how much carbon they store, yet its global variation remains poorly mapped. Here we combined 109,626 wood density measurements from 16,829 species with 300,949 vegetation plots to produce a km-scale map of community-weighted wood density for every woody biome. Our model led to a prediction accuracy 32-51 % higher than previous global products, and a 1.8-3.7-fold wider wood density range (0.28-1.00 g cm-3; global mean: 0.57 g cm-3) than previously assumed. Spatial cross-validation showed low bias ({+/-}2.5 % of the mean), and uncertainties decreased from 20% in poorly sampled drylands and boreal regions to 5% in data-rich temperate forests. Mean annual temperature was the best predictor of community-weighted mean wood density, increasing by 0.01 g cm-3 for every 1{degrees}C change. We deliver a low-bias, high-resolution wood density layer for Earth system models, together with spatially explicit error maps. This study represents a major step forward for carbon accounting and trait-based forecasts of vegetation change.

plant biology↗

bioRxiv2024

The global spectrum of tree crown architecture

Trees can differ enormously in their crown architectural traits, such as the scaling relationships that link their height and crown size to their stem diameter. Yet despite the importance of crown architecture in shaping the structure and function of woody ecosystems, we lack a complete picture of what drives this incredible diversity in crown shapes. Using data from >500,000 globally distributed trees, we explored how climate, disturbance, competition, functional traits, and evolutionary history constrain the height, crown size and shape of the worlds tree species. We found that variation in height scaling relationships was primarily controlled by water availability and light competition. Conversely, crown width was predominantly shaped by exposure to wind and fire, while also covarying with other functional traits related to mechanical stability and photosynthesis. Additionally, several plant lineages had crown architectures that defy their environments, such as the exceedingly slender dipterocarps of Southeast Asia, or the extremely wide crowns of legumes in African savannas. Our study charts the global spectrum of tree crown architectural types. It provides a roadmap for integrating crown architecture with vegetation models and remote sensing observations, so that we may better understand the processes that shape the 3D structure of woody ecosystems.

ecology↗

bioRxiv2024

Harnessing temporal and spectral dimensionality to map and identify species of individual trees in diverse tropical forests

Species-level canopy maps underpin tropical forest biodiversity monitoring, conservation planning, and carbon accounting, yet high species richness and structural complexity make remote classification challenging. Here we evaluate how far a two-step mapping approach can be taken, and where it breaks down, in hyperdiverse moist forest at the Paracou Field Station, French Guiana. First, we delineate individual tree crowns from ten repeat uncrewed aerial vehicle (UAV) RGB surveys with Mask R-CNN, fusing predictions across dates by temporal consensus: mean segmentation F1 rose from 0.68 (single date) to 0.78 (ten dates), covering approximately 86% of test-region canopy area. Second, we classify each crown from a single airborne hyperspectral acquisition (416-2500 nm, 1 m) using machine learning classifiers trained and tested on 3,186 field-verified crowns spanning 169 species. Linear Discriminant Analysis performed best (weighted F1 = 0.75), outperforming more flexible models, but unevenly: across repeated cross-validation (20 x 5-fold), on average 50 species (95% CI: 41-63) attained F1 [≥] 0.7 in a given fold and only 15 did so reliably, with many rare species unclassifiable (macro-average F1 = 0.48). Combining both steps, we estimate approximately 70% of the landscapes canopy area was correctly mapped to species. Band-importance and ablation analyses identified the far-red edge (748-775 nm) as the most informative spectral region. These results advance on studies limited to 20 or fewer species and set out spectral-resolution and training-data requirements for airborne and forthcoming spaceborne imaging spectrometers, while showing that accuracy remains strongly conditioned by training-data availability and the single-site, single-acquisition design.

ecology↗

bioRxiv2024

Sensitivity of estimates of the effectiveness of REDD+ projects to matching specifications and moving from pixels to polygons as the unit of analysis

There is a substantial interest in the potential of carbon credits generated by Reducing Emissions from tropical Deforestation and Degradation (REDD+) and traded on the voluntary carbon market for generating the finance needed to slow forest loss. However, such credits have become marred in controversy. Recent global-scale analysis using a range of methods for estimating the counterfactual rate of deforestation ex post suggest that many REDD+ projects have overestimated their effectiveness at reducing deforestation and consequently issued more credits than can be justified. All such methods include potentially arbitrary choices which can affect the estimate of the treatment effect. In addition, using pixels as the sampling unit, as some of the studies do, can introduce biases. One study which has been widely cited in the debate (Guizar-Coutino et al. 2022) estimated avoided deforestation using statistical matching of pixels and a single set of matching options. We estimate avoided deforestation from the same set of projects using 7-hectare plots rather than pixels to sample deforestation and explore the sensitivity of the results to matching choices (exploring 120 matched sets in total). We filtered the results on three criteria: 1) post-matching covariate balance, 2) proportion of REDD+ samples that were successfully matched, and 3) similarity of trends in deforestation rates prior to REDD+ implementation (parallel trends). While one of the 44 REDD+ projects failed these quality control process, we estimate treatment effects for the remaining 43 projects. There was a substantial correlation between our new estimates and those published in Guizar-Coutino et al. 2022 (0.72 measured in annual percent change, and 0.9 measured in total area change) and our headline estimate of 0.22% per yr (95% CI: 0.13 to 0.34) is essentially unchanged. At a time when confidence in the voluntary carbon markets is low, we hope these results provide reassurance that ex-post counterfactual estimates of avoided deforestation are consistent, helping accelerate their widespread adoption and rebuild trust in nature-based climate solutions.

ecology↗

bioRxiv2022

Tracking canopy gap dynamics across four sites in the Brazilian Amazon

Background information to give context to the studyCanopy gaps are the most evident manifestation of how disturbances disrupt forest landscapes. The size distribution and return frequency of gaps, and subsequent recovery processes, determine whether the old-growth state can be reached. The aim or research questionWe used remote sensing metrics to compare the disturbance regime of four Amazon regions based on the size distribution of gaps, their dynamics and geometric characteristics. A brief summary of the methodology usedWe assessed gap dynamics at four sites in the central, central eastern, southeastern, and northeastern regions of the Brazilian Amazon using repeated airborne laser scanning surveys. We developed a novel analysis to quantify four possible stages of gap dynamics: formation, expansion, persisting and recovering. For that, we overlapped layers of gap locations from two consecutive airborne laser scanning surveys. Key results with some significance measuresThe gap fraction in our study sites varied between 1.26% to 7.84%. All the sites have similar proportion of gaps among size classes. What notably changed between sites was not the gap size-distribution, but the relative importance of stages of gap dynamics. Growing and persisting rates were greatest in the site with the stronger seasonal variation in climate, lower annual precipitation, higher mean wind speed and higher solar radiation. The conclusions, which address the main aimsThe concept of stability reflects the tendency of a system to quickly return to a position of equilibrium when disturbed. We showed that gap dynamics varied among sites, with one example of low recovery rate contrasted to three other sites with faster recovery. Our results support that such as assessing the size distribution of gaps, investigating their return frequency and severity is crucial for understanding forest dynamics at the landscape and regional scales.

ecology↗

bioRxiv2021

Using deep convolutional neural networks to forecast spatial patterns of Amazonian deforestation

AO_SCPLOWBSTRACTC_SCPLOWO_LITropical forests are subject to diverse deforestation pressures but their conservation is essential to achieve global climate goals. Predicting the location of deforestation is challenging due to the complexity of the natural and human systems involved but accurate and timely forecasts could enable effective planning and on-the-ground enforcement practices to curb deforestation rates. New computer vision technologies based on deep learning can be applied to the increasing volume of Earth observation data to generate novel insights and make predictions with unprecedented accuracy. C_LIO_LIHere, we demonstrate the ability of deep convolutional neural networks to learn spatiotemporal patterns of deforestation from a limited set of freely available global data layers, including multispectral satellite imagery, the Hansen maps of historic deforestation (2001-2020) and the ALOS JAXA digital surface model, to forecast future deforestation (2021). We designed four original deep learning model architectures, based on 2D Convolutional Neural Networks (2DCNN), 3D Convolutional Neural Networks (3DCNN), and Long Short-Term Memory (LSTM) Recurrent Neural Networks (RNN) to produce spatial maps that indicate the risk to each forested pixel (~30 m) in the landscape of becoming deforested within the next year. They were trained and tested on data from two ~80,000 km2 tropical forest regions in the Southern Peruvian Amazon. C_LIO_LIWe found that the networks could predict the likely location of future deforestation to a high degree of accuracy. Our best performing model - a 3DCNN - had the highest pixel-wise accuracy (80-90%) when validated on 2020 deforestation based 2014-2019 training. Visual examination of the forecasts indicated that the 3DCNN network could automatically discern the drivers of forest loss from the input data. For example, pixels around new access routes (e.g. roads) were assigned high risk whereas this was not the case for recent, concentrated natural loss events (e.g. remote landslides). C_LIO_LICNNs can harness limited time-series data to predict near-future deforestation patterns, an important step in using the growing volume of satellite remote sensing data to curb global deforestation. The modelling framework can be readily applied to any tropical forest location and used by governments and conservation organisations to prevent deforestation and plan protected areas. C_LI

ecology↗

bioRxiv2021

Why can we detect lianas from space?

Lianas, woody vines acting as structural parasites of trees, have profound effects on the composition and structure of tropical forests, impacting tree growth, mortality, and forest succession. Remote sensing offers a powerful tool for quantifying the scale of liana infestation, provided the availability of robust detection methods. We analyze the consistency and global specificity of spectral signals from liana-infested tree crowns and forest stands, examining the underlying mechanisms. We compiled a database, including leaf reflectance spectra from 5424 leaves, fine-scale airborne reflectance data from 999 liana-infested canopies, and coarse-scale satellite reflectance data covering hectares of liana-infested forest stands. To unravel the mechanisms of the liana spectral signal, we applied mechanistic radiative transfer models across scales, corroborated by field data on liana leaf chemistry and canopy structure. We find a consistent liana spectral signature at canopy and stand scales across sites. This signature mainly arises at the canopy level due to direct effects of leaf angles, resulting in a larger apparent leaf area, and indirect effects from increased light scattering in the NIR and SWIR regions, linked to lianas less costly leaf construction compared to trees. The existence of a consistent global spectral signal for lianas suggests that large-scale quantification of liana infestation is feasible. However, because the traits identified are not exclusive to lianas, accurate large-scale detection requires rigorously validated remote sensing methods. Our models highlight challenges in automated detection, such as potential misidentification due to leaf phenology, tree life-history, topography, and climate, especially where the scale of liana infestation is less than a single remote sensing pixel. The observed cross-site patterns also prompt ecological questions about lianas adaptive similarities across environments, indicating possible convergent evolution due to shared constraints on leaf biochemical and structural traits. Open data statementOf the 17 datasets used, 10 are published and publicly accessible, with links provided in this submission (Appendix S1: Section S1). Upon acceptance, remaining seven datasets will be provided via Smithsonians Dspace. The open-source model code is available as R-package ccrtm (https://cran.r-project.org/web/packages/ccrtm/index.html) and on github (https://github.com/MarcoDVisser/ccrtm). Code will be archived in Zenodo should the manuscript be accepted for publication

plant biology↗

bioRxiv2021

Forest structure and degradation drive canopy gap sizes across the Brazilian Amazon

Canopy gaps are openings in the forest canopy resulting from branch fall and tree mortality events. Light reaches the lower layers of the canopy through these gaps, enabling understory trees to grow and maintaining the high heterogeneity and biodiversity of tropical forests. The size-frequency distribution of canopy gaps follows a power-law distribution, and the slope of this power-law () is a key indicator of forest structure and dynamics. We detected canopy gaps using a unique LiDAR data set consisting of 650 transects randomly distributed across the Brazilian Amazon Forest providing an unprecedented perspective on forest structural variation over 2500 km2 of forest. We then investigated how varied with forest structure, elevation, soil fertility, water deficit, wind gust speed and lightning intensity. We found that human modified forests had more large gaps than intact forests. Within the intact forests we observed a large-scale Northwest to Southeast pattern in (more large gaps in the Southeast), which aligns with recent work on tree mortality rates. The two most important variables in predicting were median canopy height and maximum modeled height, which had opposite effects on the number of large gaps. Forests with higher median canopy height contain fewer large gaps but the presence of very tall trees was associated with more large gaps. Environmental variables were of secondary importance in our model, with larger gaps occurring in drier forests with high soil fertility, wind speed, and lightning intensity. The distribution of large gaps in the forest canopy varies substantially over the Brazilian Amazon as a result of canopy structure and mortality rates. We mapped this variation and found more large gaps in human modified forests, forests on fertile soils and those exposed to higher wind, lightning and drought stress. Increasing extreme weather events due to climate change may therefore increase the number of large gaps in currently intact forests, causing them to resemble human modified forests.

ecology↗

bioRxiv2020

Resource availability and availability and disturbance shape maximum tree height across the Amazon

The factors shaping the distribution of giant tropical trees are poorly understood, despite its importance as a link between evolutionary biology and ecosystem biogeochemistry. The recent discovery of clusters of trees over 80 metres tall in the Guiana Shield region of the Amazon rainforest challenges the current understanding of the factors controlling the growth and survival of giant trees. The new discovery led us to revisit the question: what determines the distribution of the tallest trees of the Amazon? Here, we used high-resolution airborne LiDAR (Light Detection and Ranging) surveys to measure canopy height across 282,750 ha of primary old-growth and secondary forests throughout the entire Brazilian Amazon to investigate the relationship between the occurrence of giant trees and the environmental factors that influence their growth and survival. Our results suggest that the factors controlling where trees grow extremely tall are distinct from those controlling their longevity. Trees grow taller in areas with high soil clay content (> 42%), lower radiation (< 130 clear days per year) and wind speeds, avoiding alluvial areas (elevations higher than 40 m a.s.l), and with an optimal precipitation range of 1,500 to 2,500 mm yr-1. We then used an envelope model to determine the environmental conditions that support the very tallest trees (i.e. over 70 m height). We found that, as opposed to the myriad of interacting factors that control the maximum height at a large scale, wind speed had by far the largest influence on the distribution of these sentinel trees, and explained 67% of the probability of finding trees over 70 m in the Brazilian Amazon forest. The high-resolution pan-Amazon LiDAR data showed that environmental variables that drive growth in height are fundamentally different from environmental variables that support their survival. While precipitation and temperature seem to have lower importance for their survival than expected from previous studies, changes in wind and radiation regimes could reshape our forested biomes. This should be carefully considered by policy-makers when identifying important hotspots for the conservation of biodiversity in the Amazon.

ecology↗

bioRxiv2020

The environmental impacts of palm oil and its alternatives

The destruction of ecosystems for vegetable oil production represents a major cause of global biodiversity loss and greenhouse gas emissions 1. Over the last two decades, oil palm, in particular, has caused societal concern due to its high impacts on biodiverse and carbon-dense tropical rainforests 2-8, leading to calls to source vegetable oils from alternative oil-producing crops. However, given the high yields of oil palm, how does that damage compare with other oil crops that require more land? Here, we estimate the carbon and biodiversity footprints, per unit of oil produced, of the worlds five major vegetable oil crops. We find that oil palm has the lowest carbon loss and species richness loss per-tonne-oil, but has a larger impact on range-restricted species than sunflower and rapeseed. We go on to identify global areas for oil crop expansion that will minimise future carbon and biodiversity impacts, and argue that closing current yield gaps and optimising the location of future growing areas will be much more effective at reducing future environmental impacts of global vegetable oil production than substituting any one crop for another.

ecology↗

bioRxiv2019

The mechanical stability of the worlds tallest broadleaf trees

The factors that limit the maximum height of trees, whether ecophysiological or mechanical, are the subject of longstanding debate. Here we examine the role of mechanical stability in limiting tree height and focus on trees from the tallest tropical forests on Earth, in Sabah, Malaysian Borneo, including the recently discovered tallest tropical tree, a 100.8 m Shorea faguetiana. We use terrestrial laser scans, in situ strain gauge data and finite-element simulations to map the architecture of tall broadleaf trees and monitor their response to wind loading. We demonstrate that a trees risk of breaking due to gravity or self-weight decreases with tree height and is much more strongly affected by tree architecture than by material properties. In contrast, wind damage risk increases with tree height despite the larger diameters of tall trees, resulting in a U-shaped curve of mechanical risk with tree height. The relative rarity of extreme wind speeds in north Borneo may be the reason it is home to the tallest trees in the tropics.

ecology↗