A Few Trees Hold Most of the Carbon: Why Remnants and Giants Matter in Cocoa Agroforestry

Walk through a mature cocoa farm in Côte d'Ivoire, the Alto Beni, or southern Bahia and you will see the same pattern: dozens of cocoa trees, a scatter of planted shade, and, here and there, one or two old giants looming over the canopy. Those giants usually hold more carbon than everything else on the plot combined. That simple fact has outsized consequences for how farms should be designed, how supply chains should report their footprint, and how carbon projects should be credited.

Remnant and large-diameter trees are often the easiest thing to miss in an inventory, the cheapest to remove during replanting, and the hardest to replace. The science is now unambiguous. The managerial, commercial, and regulatory response is still catching up. The gap between what the ecology tells us and what corporate MRV systems currently measure is where most of the risk (and the opportunity) sits.

The arithmetic of large trees

The disproportionate importance of large trees is not a rhetorical device; it is a well-characterised empirical pattern. In a global synthesis of 48 large forest plots containing more than 5.6 million stems, Lutz and colleagues found that the largest one percent of trees held roughly half of aboveground live biomass, and trees at or above 60 cm DBH accounted for 41 percent1. The same heavy-tailed signature recurs across the moist lowland tropics, where Slik and colleagues showed that a small number of large trees drives most of the plot-to-plot variation in aboveground biomass2, and Meakem and colleagues found that stems at or above 50 cm DBH held 59 percent of live aboveground biomass and drove 45 percent of woody productivity in moist tropical forest3. The largest individuals also carry most of the predictive weight: Bastin and colleagues found that in Central African forests the largest five percent or so of stems predict whole-plot aboveground biomass to within about 14 percent4, and Ali and colleagues framed the wider pattern as the “big-sized trees hypothesis,” showing that the structural attributes of the top one percent of trees explain 55 to 70 percent of variation in aboveground biomass, dwarfing the effect of species richness5.

Two mechanisms drive this. First, allometry: aboveground biomass scales roughly with DBH raised to a power of about 2.4 in the pantropical model of Chave et al., so a 100 cm tree holds on the order of 30 times more carbon than a 30 cm tree of similar wood density6. Second, large trees do not stop growing. Stephenson and colleagues, analysing 403 temperate and tropical species, showed that absolute mass growth rate increases continuously with tree size7. A single big tree can add the same amount of carbon to the forest in a year as is contained in an entire mid-sized tree. Large old trees are not senescent reservoirs; they are active sinks.

Why remnant trees carry cocoa farm carbon

These patterns translate directly to cocoa agroforestry. Across Central America, Somarriba and colleagues measured about 49 tonnes of aboveground carbon per hectare across 229 plots, of which timber and fruit trees (not cocoa) stored 65 percent8. In Ghana, Dawoe et al. found that shade trees held more carbon than cocoa itself, and Wade and colleagues estimated that shade and other non-cocoa trees hold roughly three-quarters or more of vegetation carbon in cocoa-growing landscapes9. Asigbaase et al. showed that organic cocoa farms with higher shade-tree density held nearly twice the aboveground biomass carbon of conventional ones, and the tree cover lifts the soil pool as well: in a multi-site survey, Cardinael and colleagues measured significantly higher soil organic carbon stocks under agroforestry than under matched treeless plots10. In the cabrucas of southern Bahia, Schroth and colleagues estimated 87 tonnes C per hectare in traditional agroforests, about half the stock of adjacent old-growth forest but nearly twice the stock of intensified cocoa systems11. Within these averages, a handful of individual trees typically dominate.

The most direct evidence comes from a 2025 study by Konan and colleagues, who inventoried 11,568 trees across 150 cocoa fields in Côte d'Ivoire12. They disaggregated carbon by tree origin, remnant, spontaneous, and planted, and found median carbon stocks of 6.33, 2.06, and 1.53 tonnes C per hectare respectively. Put differently: at the median, remnant trees carry about four times the carbon of planted shade trees per hectare, from a much smaller number of individuals. Once trees are older than about seven years, spontaneous trees accumulate biomass at roughly 11 kg per year against just 4 kg per year for planted ones, probably because remnants and wildlings tend to be forest species with higher wood density and more advantageous positions.

The belowground half of the story

Focusing only on aboveground biomass understates the role of large trees further. Root-to-shoot ratios of tropical broadleaf species run from roughly 0.20 to 0.25, though the ratio is not constant: analysing individual trees worldwide, Ledo and colleagues found that root-to-shoot declines as trees grow larger, so a giant's belowground fraction shrinks in proportion even as its absolute coarse-root mass keeps climbing13. Borden and colleagues measured about 0.23 for cocoa itself in Ghana, with cocoa roots contributing 5 to 6 tonnes C per hectare. Zekeng and colleagues found roots accounted for about 22 percent of total carbon stock in Cameroonian cocoa agroforests. Large trees therefore bring proportionally large coarse-root systems that extend well beyond the trunk, not just additional biomass carbon but also hydraulic redistribution, deep nutrient cycling, and long-lived inputs to the soil.

The soil organic carbon story is more contingent. Silue and colleagues, using a 20-year cacao-Albizia/cacao-Acacia/full-sun trial in Côte d'Ivoire, showed that shade-tree species choice determines whether soil carbon rises or falls14. Albizia lebbeck, with low-C:N litter, increased SOC down to 60 cm depth by 11 percent, while Acacia mangium reduced SOC in some horizons but generated more biomass overall. Large remnant forest trees with high biomass and continuous, diverse litter inputs tend to be SOC-positive, though replicated long-term data remain scarce.

Why large trees are also the most vulnerable

Large trees are, at the same time, the most climatically exposed element of an agroforestry farm. Bennett and colleagues showed that during drought, large trees suffer disproportionately higher mortality than small ones15. Reviewing the drivers of tree death in moist tropical forests, McDowell and colleagues tied that elevated mortality to hydraulic failure and carbon starvation, mechanisms that fall hardest on tall, hydraulically stressed individuals16. The exposure compounds a slower loss: Lindenmayer and colleagues documented a global decline in large old trees across biomes17, and a later synthesis by Lindenmayer and Laurance set out how hard these individuals are to conserve and how many decades they take to replace18. On cocoa farms, pressures are amplified: isolation reduces crown support, higher radiation and wind load stress oversized boles, and farmer incentives often favour removing a shade tree that competes with cocoa or threatens workers. The same trees that dominate farm carbon are the hardest to replace and the most likely to be lost in the next drought year or replanting cycle, a double asymmetry that any serious climate strategy must reckon with.

What this means for farm design and supply chains

Three design rules follow for companies sourcing cocoa and for their field partners.

The symmetric lesson deserves its own emphasis. Remnants are finite, but nothing prevents a project from planting them. Four to six carefully chosen long-lived emergents per hectare, such as Milicia excelsa, Terminalia ivorensis, Entandrophragma spp., Khaya anthotheca, and Cedrela odorata, planted from year one under a non-felling covenant and released from cocoa competition as they mature, will deliver the same heavy-tailed carbon structure that existing remnants provide today. The per-tree carbon outcome at maturity is one to two orders of magnitude greater than conventional shade planting; the managerial shift is modest. A cocoa agroforest established in 2026 with a handful of contractually protected future-remnants will carry carbon like a traditional cabruca by 2060. Because tree density matters as much as size, a thoughtful mix of moderate shade density and a small number of large, protected trees outperforms either a pure high-density shade design or a plantation with a single emergent per block.

Every renovation and replanting protocol should include a mandatory remnant-tree retention rule, a ban on clear-felling pre-existing shade during block rehabilitation, and a planted-successor target that is tracked and audited tree-by-tree. Abou Rajab et al. showed in Sulawesi that diverse shade can quintuple biomass carbon without penalising cocoa yield; the “yield gap” argument is weaker than commonly assumed when compared over full rotation.

Big trees and the MRV problem

The same statistical distribution that makes large trees ecologically dominant makes them statistically awkward. Because aboveground biomass in agroforestry systems is heavy-tailed, small plots systematically under-sample large trees and therefore under- or over-estimate total carbon with very wide confidence intervals. The 20 × 20 m or 25 × 25 m plot commonly used in cocoa inventories is usually too small; nested designs that fully inventory trees ≥ 30 cm DBH in a larger plot (for instance, 50 × 50 m), aligned with the IPCC 2019 Refinement and the approach used by Somarriba and colleagues, materially reduce bias.

Allometric choice matters equally. The pantropical Chave et al. model uses DBH, wood density, and either tree height or a climate-based E-factor; it remains the default for most voluntary-market agroforestry methodologies. But its calibration data included relatively few trees above 100 cm DBH, and terrestrial LiDAR work has shown that allometric bias increases systematically with tree size, with errors of 15 to 35 percent for the largest individuals. For farms with a few giants, this is the single largest source of MRV uncertainty and should trigger site-specific calibration or LiDAR verification rather than reliance on defaults.

Baseline underestimation of remnant trees is also a live crediting risk. If a project's baseline assumes low-carbon cocoa monoculture and credits shade planting as the “intervention,” but the starting landscape actually contains retained remnants that dominate stock, the additionality claim is overstated. Conversely, conservation interventions that preserve existing remnants, often not creditable under current afforestation/reforestation methodologies, may deliver more real carbon than eligible planting projects. The GHG Protocol Land Sector and Removals Guidance and the SBTi FLAG framework require corporate actors to account for both land-use-change emissions and removals with traceable, site-specific data. The EU Deforestation Regulation tightens the definitional stakes further, since whether a shaded cocoa farm is classified as “forest” or “agricultural use” depends partly on the presence of large trees and canopy structure.

What this means, in practice

For agronomists and field teams: survey and tag every tree ≥ 30 cm DBH on focal plots before any rehabilitation decision is taken, and treat retention of trees ≥ 60 cm DBH as a non-negotiable design constraint. For buyers and brands: align supplier contracts with remnant-tree protection clauses, not only planting targets, and insist on MRV plot designs large enough to capture the heavy tail. For carbon project developers: adopt stratified sampling that over-weights large-tree strata, use site-calibrated allometrics or terrestrial LiDAR on a sub-sample of trees above 70 cm DBH, and report uncertainty explicitly rather than collapsing to a single mean. For policy actors: reflect the ecological reality that the marginal carbon value of an existing large tree typically exceeds that of a planted seedling by one to two orders of magnitude over the next decade, and design incentives accordingly.

Key takeaways