Human degradation of tropical moist forests is greater than previously estimated

Tropical moist forests (TMFs) provide critical ecosystem services including climate and water regulation, carbon sequestration and biodiversity conservation, yet are rapidly declining. Beyond outright deforestation, degradation from selective logging, fires and edge effects occurs at rates comparable to deforestation and undermines forest function and resilience. Edge effects are structural and functional changes at forest boundaries driven by fragmentation. Degraded forests are more vulnerable to additional disturbances and climate extremes, threatening long-term resilience. Although reducing degradation could substantially cut emissions and enhance sequestration, large uncertainty remains in quantifying degradation’s contribution to carbon fluxes at pantropical scales, limiting REDD+ implementation. Despite advances in remote sensing, a consistent pantropical assessment of how degradation alters forest structure and its persistence has been lacking, partly due to limited forest structure data across the tropics. The deployment of GEDI LiDAR provides new opportunities. This study aims to quantify: (1) the extent of forest degradation in 2022 accounting for edge effects; (2) the impact and persistence (1990–2022) of selective logging, fire and edge effects on canopy structure; (3) recovery rates of structure after degradation and regrowth after deforestation; and (4) the vulnerability of degraded forests to subsequent deforestation.

Prior work shows degradation is a major and underestimated source of greenhouse gas emissions and biodiversity loss, sometimes matching or exceeding deforestation rates. Remote sensing advances have enabled monitoring of carbon fluxes and structural changes from logging and fire, but pantropical, structure-focused assessments have been limited due to sparse structural observations. Fragmentation-induced edge effects were often assumed to penetrate only ~100–120 m; studies reveal stronger and more persistent biomass declines and microclimatic desiccation near edges, variable by forest type and landscape configuration. Previous intact forest landscape mappings (for example, Potapov et al.) and studies on long-term logging and fire impacts provide context, but uncertainties remain in the magnitude, spatial extent, and long-term persistence of degradation impacts and recovery, as well as interactions between degradation, edge proximity, and deforestation risk.

Data integration combined: (a) a wall-to-wall Landsat-based dataset of tropical moist forest change dynamics from the JRC-TMF product (30 m resolution; deforestation, degradation, regrowth from 1990–2022), and (b) spatially discontinuous but structurally explicit GEDI LiDAR observations (2019–2022) of canopy height RH98 (near canopy top), RH50 (median energy height), and aboveground biomass density (AGBD). Intact forests were defined as areas without detected human activity in the past three decades and located ≥3 km from forest/non-forest edges. Analyses were conducted in 1.5° hexagonal grid cells, masking cells with <600 GEDI samples or without significant intact vs non-intact differences (Welch two-sided t-test, P < 0.05). Spatial characterization: GEDI samples were stratified into intact, degraded (logging, fire, natural), edge forests (within 120 m of forest/non-forest boundary), and regrowth on deforested land. Edge-effect scale and magnitude were assessed by measuring RH98 at increasing distances from the forest/non-forest edge separately for the Americas, Africa and Asia. Two metrics quantified edge-effect depth: (1) the distance where RH98 of edge forest equals 95% of intact forest RH98, and (2) the maximum distance with significant RH98 differences relative to intact (one-sided ANOVA with Tukey post hoc). Similar analyses were performed for AGBD. Additional edge effects from degraded patches (forest/degraded forest edges) were assessed by comparing undisturbed forests within 120 m of logged or burned areas against intact forests. Extent mapping: The spatial footprint of edge effects from both deforested land and degraded forests was used to refine intact forest landscape maps (2020), contrasted with Potapov et al. Recovery and persistence: Time-since-disturbance analyses (up to ~30 years) quantified persistence and recovery for RH98, RH50 and AGBD following edge creation, selective logging (including areas logged once over the past 30 years), fire, and regrowth on deforested land. To minimize direct edge-proximity bias in post-disturbance dynamics, responses beyond 120 m from edges were analyzed. Recovery was summarized as percentage of intact forest metrics over time by continent. Deforestation vulnerability: Logistic-style risk analyses related subsequent deforestation probability (2020–2022) to relative canopy heights (RH50, RH98) and distance to edge, examining interactions and continental differences. Statistical tests included Welch t-tests, one-sided ANOVA, Tukey HSD, with spatial standard deviations reported. Sample sizes by class and distance bands were detailed in supplementary figures.

- Canopy structure of intact TMFs varies regionally: mean RH98 higher in Asia (34.4 ± 10.7 m) than Africa (29.3 ± 8.6 m) and the Americas (28.6 ± 7.4 m). AGBD similarly higher in Asia (370.8 ± 205.2 Mg ha⁻¹) than Africa (225.5 ± 110.9) and the Americas (239.5 ± 129.9). - Degraded and edge forests have substantially lower structure: minimum mean differences vs intact of 10 m RH98 and 122 Mg ha⁻¹ AGBD for degraded forests; forests within 120 m of edges average 11 m lower RH98 and 150 Mg ha⁻¹ lower AGBD. Regrowth areas have on average 16 m lower RH98 and AGBD of 80.4 ± 87.3 Mg ha⁻¹. - Edge effects penetrate far into forest interiors: distances to reach 95% of intact RH98 are ~350 m (Americas), 400 m (Africa), 1,500 m (Asia). Significant RH98 differences persist up to 7 km in the Americas and 1.7 km in Africa and Asia. Within the first ~200 m, RH98 is reduced by ~20% relative to intact forests. - Biomass edge-effect scales are large: detectable AGBD effects to ~1,000 m (Americas), ~750 m (Africa), ~1,020 m (Asia), far exceeding the often-assumed 120 m. - Extent of edge-affected area: edge effects encroach on 18% of TMF area in 2022 (~206 Mha), representing a 221% increase over estimates based on a 120 m edge buffer. - Additional edge effects from degraded patches: undisturbed forests within 120 m of logged or burned areas have RH98 reduced by 15% and 22%, respectively (RH50 reduced by 22% and 32%). - Intact forest landscape mapping: refined mapping indicates 426 Mha intact in 2020 versus 502 Mha in Potapov et al.; about 48% of intact TMF falls within protected areas; ~57% of protected TMFs are intact (≈60% in the Americas/Africa, 28% in Asia). - Persistence of edge effects: within 120 m of edges, undisturbed forests average 15% lower RH98, 25% lower RH50, and 30% lower AGBD than intact forests from the first year after edge creation, with no significant recovery over 30 years. - Degraded edges (logging or fire) experience additional structural losses: ~30% further decrease in RH98 (50% RH50, 40% AGBD) with no recovery observed. - Selective logging impacts and recovery: immediate decreases in RH98/AGBD are larger in Asia (20%/50%) than in the Americas/Africa (10%/30%). Within 20 years, RH50 recovers by 25% (Americas), 15% (Africa), 27% (Asia); AGBD recovery averages ~11%; RH98 shows no recovery trend. - Fire impacts and recovery: immediate RH98 declines of 35% (Americas), 40% (Africa), 60% (Asia); AGBD declines of 60% (Americas/Africa) and 80% (Asia). No recovery trend in RH98 or AGBD even 10 years post-fire. - Regrowth trajectories: RH98 reaches ~60% of intact within 10–15 years with low growth rates (0.5% yr⁻¹ Americas, 0.7% Africa, 0.9% Asia). AGBD reaches on average 43% of intact after 20 years (Americas 40%, Africa 33%, Asia 57%). Regrowth slows after ~10 years and is negatively affected by prior land-use intensity and fire legacies (20–75% reductions in regrowth rates in drier regions). - Degradation elevates deforestation risk: degraded forests subsequently deforested (2020–2022) had severe pre-deforestation structural losses (RH50 −60%, RH98 −45%, AGBD −65%), with high spatial variability (±12.8%, ±13.3%, ±14.6%). Deforestation probability reaches 50% when forests have lost ~50% of initial height in the Americas (≈60% in Africa and Asia). Edge proximity interacts with canopy height, amplifying risk within the first kilometre; within 120 m, degradation enhances deforestation risk only in the Americas. - Overall magnitude: degradation effects reduce canopy height and AGBD by 20–80% and extend up to ~1.5 km inside forests, substantially greater than previously estimated.

By integrating GEDI LiDAR structural metrics with three decades of Landsat-based forest change histories, the study directly quantifies how selective logging, fire and fragmentation-driven edge effects alter the vertical structure and biomass of TMFs and how long these impacts persist. The findings resolve key uncertainties by showing that edge effects are more extensive and persistent than assumed, penetrating up to ~1.5 km and enduring without detectable recovery over decades. Logging and fire produce large, long-lasting structural losses, with limited recovery—especially for canopy-top height and biomass—highlighting that degradation’s carbon and biodiversity impacts are not transient at pantropical scales. The strong predictive power of canopy structure (RH50/RH98) and edge proximity for subsequent deforestation risk underscores mechanistic links between degradation, increased accessibility/exposure, microclimatic stress, and land-use conversion. These insights are highly relevant for REDD+ and conservation planning: accounting for degradation and edge-induced losses expands the spatial scope of mitigation needs by >200% relative to narrow edge assumptions, identifies vulnerable forests near edges and heavily degraded stands, and emphasizes protection and restoration strategies that can realistically address century-scale recovery horizons.

The study introduces a pantropical assessment of degradation impacts on forest structure by fusing GEDI LiDAR with long-term optical data. It demonstrates that degradation-induced losses of canopy height and biomass are larger and more widespread than previously thought, with edge effects extending up to ~1.5 km and cumulative impacts from logging and fire persisting with minimal recovery over decades. While secondary forests offset a fraction of deforestation-related carbon losses, full structural recovery after deforestation or degradation likely requires centuries and is hindered by recurrent anthropogenic pressures. Canopy structural metrics combined with disturbance histories can identify forests at highest risk of conversion, guiding targeted monitoring and conservation under UNFCCC and CBD frameworks. Future research should extend temporal baselines beyond 30 years, refine detection of small-scale/low-intensity disturbances, and evaluate how climate extremes, fire regimes and forest composition modulate recovery trajectories and vulnerability.

- Temporal scope: the 30-year analysis window is insufficient to observe full structural recovery in many TMF contexts, leading to conservative recovery inferences. - Sensor sampling: GEDI provides spatially sparse footprints, potentially under-representing fine-scale heterogeneity; masking of hexagons with limited samples may omit some areas. - Detection limits of optical data: Landsat-based products can miss small-scale (<0.09 ha) and low-intensity disturbances, especially near interfaces, biasing degradation attribution; this is partially mitigated by LiDAR but remains a limitation. - Historical data gaps: limited pre-2005 Landsat coverage in parts of Central Africa may underestimate earlier disturbances, affecting comparisons (e.g., intact forest landscape mapping differences with prior studies). - Generalization and variability: recovery rates and edge-effect magnitudes vary with logging intensity, forest type/composition, landscape configuration, and climate; continent-level summaries may mask local variability. - Edge-proximity assumptions: while effects beyond 120 m were analyzed for recovery, significant edge influences can extend much farther, complicating isolation of pure disturbance effects. - Potential confounders: interacting disturbances (e.g., repeated fires, droughts, selective logging practices) and management histories may influence observed structural trajectories and deforestation risk models.