Rainforest transformation reallocates energy from green to brown food webs

Losses of biodiversity in terrestrial ecosystems have been documented across continents, biomes, clades and ecosystem compartments¹. Tropical ecosystems are among the most threatened globally, with losses driven primarily by land-use change, such as conversion to commodity crops². Agricultural monocultures drastically reduce plant diversity relative to rainforests, with cascading effects through food webs affecting higher trophic-level invertebrate and vertebrate consumers^2,5,6. Energy acts as a common currency that can constrain species numbers and shape energy pathways through ecological networks, linking biodiversity distribution to energy flux across trophic levels and compartments^9–11. Previous work has shown reduced total energy flux in litter invertebrate communities with biodiversity loss under tropical land-use change⁸, while soil communities may maintain total flux via energetic redistribution across the food web^8,12. Understanding whether land-use change alters the total available energy versus the distribution of energy across trophic levels and compartments is essential for predicting consequences for biodiversity and ecosystem functioning.

The distribution of biomass and energy fluxes in terrestrial ecosystems is organized into ‘green’ (aboveground) and ‘brown’ (belowground) food-web compartments with shared primary producers and mobile predators linking them^1,5. Yet, aboveground and belowground tropical food webs have typically been studied separately, and the cross-compartment distribution of energy across invertebrate–vertebrate webs has not been quantified, obscuring the consequences of rainforest conversion on total animal energy flux and functioning.

Here, the authors quantify energy fluxes across earthworms, birds, and arthropods in soil and canopies of Sumatran rainforests and plantations to characterize energetic structure across compartments. They hypothesize that keystone animal groups differ between rainforest and plantations and that land use reallocates energy by: (1) shifting energy towards aboveground in plantations; (2) reducing energy to higher trophic levels in plantations; and (3) increasing reliance on living plants at the food-web base in plantations due to reduced litter and predation pressure.

The study builds on work demonstrating widespread biodiversity declines from tropical land-use change across taxa and trophic levels^{1,2,5,6,30–33}. Energy flux is established as a key link between biodiversity and ecosystem functioning¹¹, with evidence that biodiversity loss in litter communities reduces total energy flux⁸, whereas soil communities may redistribute energy rather than reduce it^8,12. Aboveground–belowground linkages via plants and generalist predators are well documented^14,17, but tropical studies often treat compartments independently. Prior research indicates that conversion to plantations can reduce arthropod biomass and alter soil microbial communities, shifting from fungal- to bacterial-dominated processing and affecting soil carbon storage^{6,15,29,36,47}. The ‘green–brown imbalance’ hypothesis suggests higher resistance of belowground food webs to perturbation due to fewer specialized links¹³. Together, these studies motivate an integrated, multitrophic, cross-compartment energy-flux perspective to interpret biodiversity-functioning changes under tropical land-use transformations.

Study design encompassed 32 sites in Sumatra, Indonesia, spanning primary rainforest and three plantation systems: jungle rubber (selectively logged rainforest with planted rubber), rubber monoculture, and oil palm monoculture. Abundance and biomass were estimated for: (a) canopy arthropods via insecticide fogging; (b) birds via automated audio recorders and point counts; and (c) soil arthropods and earthworms via high-gradient heat extraction from soil/litter cores. Body mass and biomass data were linked to literature-based traits and feeding preferences to assign taxa to 62 trophic guilds and reconstruct site-specific food-web topologies.

Steady-state food-web modeling was applied, assuming energetic demands (metabolism, assimilation losses, consumption by predators) of each guild are balanced by energy uptake from lower trophic levels. Metabolic rates per unit biomass were predicted from body masses using metabolic regressions and multiplied by observed biomasses to compute energy fluxes. Fluxes quantified consumption of basal resources (living plants, leaf litter, fungi, bacteria, soil organic matter) and animal prey across aboveground and belowground compartments.

Sensitivity analyses included: (1) testing canopy fogging undersampling related to canopy height; the most severe correction increased estimated aboveground flux but did not eliminate the large aboveground–belowground difference; and (2) food-web reconstruction uncertainty tests, showing feeding specialization/omnivory affected absolute values but not conclusions. Vertebrate sampling did not include amphibians, reptiles, bats, and other mammals, but their omission is unlikely to overturn the >10-fold flux differences. A validation survey at the same sites (except jungle rubber) 4 years later (2016–2017) independently re-estimated biomass and fluxes, corroborating key patterns (belowground dominance, canopy decline, energy reallocation, trophic-function shifts).

• Energetic dominance of belowground in rainforest: Aboveground total energy flux (canopy arthropods + birds) was 21.6 ± 9.7 mW m⁻² with fresh animal biomass 0.8 ± 0.6 g m⁻³; belowground total energy flux (litter + soil arthropods + earthworms) was 295.8 ± 125.5 mW m⁻² with biomass 9.5 ± 7.1 g m⁻².
• Canopies dominated by arthropods: Canopy arthropods accounted for 18.0 ± 9.7 mW m⁻²; birds contributed 1.6 ± 1.9 mW m⁻².
• Total system energy flux similar or higher in plantations: Estimated totals (mean ± s.d.) were rainforest 317 ± 129, jungle rubber 494 ± 325, rubber 310 ± 168, oil palm 314 ± 179 mW m⁻² (n = 8 sites per system).
• Keystone shift toward earthworms in plantations: Earthworms channelled ~13% of energy in rainforest (29.4 ± 37.1 mW m⁻²) versus 60–79% across plantations (χ² = 50.1, P < 0.0001). This increase mirrored declines in soil arthropod-associated fluxes.
• Aboveground-to-belowground reallocation with land use: Belowground flux exceeded aboveground by ~14-fold in rainforest, increasing to ~30-fold (jungle rubber), ~55-fold (rubber), and ~68-fold (oil palm). Aboveground total energy flux declined by −75% to −79% in monocultures versus rainforest (up to −92% with conservative canopy undersampling correction), while belowground flux changed little.
• Reduced food-web complexity: Trophic interactions decreased by 13% to 37% across compartments and systems, with soil communities relying on ~21% fewer interactions yet processing similar energy amounts as rainforest soils.
• Predation decline: Predation-to-basal consumption ratio decreased by 18% aboveground and up to 90% belowground in jungle rubber and oil palm relative to rainforest. Rubber monocultures showed an 11% increase aboveground, driven by high dipteran biomass linked to water in sap collection buckets. In oil palm, predation-to-herbivory ratios were lower than rainforest (birds 0.37 ± 0.16 vs 0.64 ± 0.29; canopy arthropods 0.28 ± 0.05 vs 0.34 ± 0.05; soil arthropods 1.14 ± 0.63 vs 1.95 ± 0.74).
• Faster, bacteria-dominated energy channels in plantations: Bacteria/fungi energy flux ratio increased 3.2–4.4-fold overall in plantations; increases driven largely by abundant earthworms assimilating bacterial carbon from older soil organic matter, with additional increases in soil arthropods in oil palm.
• Shift in carbon balance: The ratio of faeces production (unassimilated food) to soil organic matter consumption decreased by >75%, from 27.6 ± 29.6 (rainforest) to 3.8 ± 2.9 (jungle rubber), 6.2 ± 10.4 (rubber), and 2.3 ± 0.3 (oil palm), consistent with accelerated turnover and soil carbon stock depletion.
• Validation survey confirmed: Belowground energetic dominance, canopy flux decline, energy reallocation to belowground, increased bacteria-to-fungi flux ratio, and reduced faeces/soil feeding ratio were replicated, though not the general loss of trophic links.

The study demonstrates that rainforest conversion restructures energetic pathways across aboveground and belowground compartments without necessarily reducing total animal energy flux. Contrary to the initial hypothesis of increased aboveground allocation in plantations, land-use change caused a pronounced relative shift toward belowground energy processing, largely mediated by earthworms. Aboveground communities experienced substantial energy flux declines, consistent with rapid responses to vegetation simplification, while belowground webs showed inertia, potentially due to legacy soil organic matter and the dominance of invasive earthworms.

Energetic redistribution also altered trophic structure, with marked declines in predation relative to basal consumption, especially in jungle rubber and oil palm, implying weakened natural pest control and contributing to pest outbreak risk. Rubber’s idiosyncratic increase in aboveground predation highlights that crop-specific management (e.g., water availability in sap buckets) can shape trophic functions. Basal resource use shifted from slower fungal to faster bacterial pathways, indicating accelerated carbon turnover and a move from faeces production toward direct consumption of soil organic matter. These changes align with observed soil carbon depletion in plantations and suggest increased system susceptibility to perturbations due to stronger, faster interactions.

Collectively, findings address the research questions by identifying different keystone groups across land uses (arthropods in rainforest vs earthworms in plantations), documenting reduced energy to higher trophic levels, and revealing a shift in basal resource use. The results underscore the need for integrated management strategies that rebalance energy allocation and support multitrophic biodiversity and ecosystem services in tropical landscapes.

This work provides an integrated energetic characterization of tropical rainforest and plantation food webs across aboveground and belowground compartments. Key conclusions are: (1) rainforest animal communities are energetically dominated by arthropods belowground; (2) rainforest transformation disproportionately reduces aboveground energy flux, while belowground energy in plantations is reallocated from diverse arthropod communities to invasive earthworms; (3) land-use change generally lowers predation relative to primary consumption (except for higher predation in rubber canopies), affecting natural pest control; and (4) belowground food webs in plantations increasingly rely on faster, bacterial energy channels and shift carbon balance from faeces production toward soil organic matter consumption, consistent with soil carbon stock depletion.

Future research should deploy dynamic, ecosystem-level models and targeted experiments to quantify animal-mediated effects on carbon cycling over time, evaluate the long-term trajectories across plantation cycles (including replanting), and test restoration and management interventions (e.g., mulching, reduced herbicide use, increased canopy complexity, biodiversity enrichment) for their capacity to restore energetic balance and multitrophic functionality.

• Food-web reconstruction uncertainty: Assignment of trophic guilds and interactions influences absolute flux estimates; sensitivity analyses identified feeding specialization/omnivory as the main source of variation but did not alter conclusions.
• Aboveground sampling biases: Potential undersampling of canopy invertebrates by fogging and omission of certain vertebrate predators (amphibians, reptiles, bats, other mammals) could bias absolute aboveground flux estimates; however, corrections do not close the >10-fold above–below difference.
• System age and nonequilibrium: Plantation sites were 14–18 years old and may not represent steady states; validation 4 years later confirmed key patterns but suggested possible partial recovery of trophic links with age.
• Generalizability: Findings pertain to Sumatran systems and specific plantation types; broader applicability requires cross-regional replication and long-term studies.