Why Forests Are Actually Two Forests in One

In a forest, the number of trees of each size follows a pattern so consistent that ecologists have observed it for decades: there are always far more small trees than large ones, and this decline follows a mathematically predictable curve. But why? What mechanisms generate this pattern? A team of physicists and ecologists from the University of Padua has now cracked a piece of this puzzle—and their answer carries implications that extend far beyond forests. Their model reveals that the familiar tree-size distribution is actually two patterns in one, controlled by two distinct forces: competition for resources among seedlings, and competition for light and space among growing trees. More provocatively, they show that disturbances at the edges of a forest—deforestation at a border, a road carved through wilderness—can ripple inward for hundreds of meters, fundamentally reshaping the entire community in ways that have nothing to do with what happens inside the forest's core.

The Science

Christian Grilletta and colleagues at Padua's Laboratory of Interdisciplinary Physics set out to do something that has frustrated ecologists for years: connect the large-scale patterns we observe in forests to the small-scale processes that actually generate them. Trees grow, die, compete for light, produce seeds, and disperse those seeds—sometimes hundreds of meters from the parent. All of these individual actions somehow combine to produce the regular size distributions that have been documented across forests worldwide. But which processes matter most? And how do they interact?

The team built a mathematical model derived from first principles of metabolic scaling—the branch of biology that links an organism's size to its metabolism, growth rate, and resource needs. The core idea is that a tree's metabolic rate scales with its size in a predictable way: a tree that is twice as tall doesn't need twice the energy, but something close to it. This fundamental constraint, they argued, should leave fingerprints on how trees of different sizes are distributed across a forest.

Their model tracks tree density as a function of height, space, and time, incorporating four key processes. First, ontogenetic growth—how fast a tree grows depends on its current size and the resources available to it. Second, light shading—when a tree reaches into the canopy, it blocks light from smaller neighbors, slowing their growth. Third, density-dependent mortality—trees die faster when crowded, whether from direct competition for resources or from being overtopped by neighbors. Fourth, nonlocal recruitment—new seedlings arrive through seed dispersal, which can carry seeds far from the parent tree, creating spatial connections across the forest.

To make the model tractable, the researchers assumed a spatially homogeneous "bulk" region where conditions are the same everywhere—a reasonable approximation for the interior of a large, undisturbed forest. They then solved the resulting equations mathematically to find what the forest looks like when it reaches a steady state, where births balance deaths and the size distribution stabilizes.

What They Found

The stationary solution revealed something elegant: the tree-size distribution isn't a single pattern but two overlapping ones. For small trees, the distribution follows a power law—meaning the number of trees decreases predictably with size—with an exponent that depends on how intensely seedlings compete for resources. For larger trees, a different power law takes over, with an exponent tied to the tree's allometric scaling—essentially, the geometric constraints of how crown size grows with height.

The crossover between these two regimes happens at a characteristic height $h_c$ that depends on three factors: how many seeds are being produced and successfully germinating ($\kappa$), how much resource is available locally ($R_a$), and the metabolic parameters of the trees themselves.

$${\rho }^{\ast }(h)\approx {\left(\frac{h^a}{h_0^a{\rho }^{\ast }(h_0)}+(1+\gamma )\frac{h^{1+2H}}{1-a+2H}\right)}^{-1}$$

where $a=R[{\rho }^{\ast }]/R_a$ is the resource consumption ratio.

Tree Density Declines Across Size Classes

Label	Value
Small trees (h < hc)	1 Relative density
Medium trees (h ≈ hc)	0.2 Relative density
Large trees (h > hc)	0.05 Relative density

Data from Figure 2A parameters: κ=0.001, Ra=220, hc≈4.2, a≈0.5

This dual power-law structure explains a long-standing puzzle in forest ecology. Empirical observations have reported different scaling exponents in different forests—or even in different size ranges within the same forest. The Padua team's model shows this isn't contradictory: both exponents are real, and each dominates under different conditions. When resources are abundant or seed input is high, the resource-competition regime shrinks, and the allometric exponent $-(1+2H)$ dominates throughout. When resources are scarce or recruitment is low, the resource-competition exponent $-a$ extends to larger sizes. In a typical forest with moderate conditions, both regimes are visible, separated by the crossover height.

Crossover Height Reflects Forest Health

Label	Value
Low resources/sparse seeds	101 hc (height units)
Moderate conditions	4.2 hc (height units)
High resources/abundant seeds	0.1 hc (height units)

Data from Figure 2: hc changes by three orders of magnitude across forest conditions

Perhaps more striking is what happens when the spatial assumptions break down. Real forests have edges. A road cut, agricultural expansion, or any boundary disturbance changes the seed rain at the forest's border—seeds that would have been dispersed outward are now lost, while new seeds from outside may not match the local species pool. The researchers modeled this by confining the forest to a finite strip, representing the kind of boundary fragmentation common in logged or fragmented landscapes.

The results were dramatic. Boundary disturbances suppressed the abundance of smaller trees near the edge—not because those trees were dying faster there, but because fewer new seedlings were arriving. This effect propagated inward far beyond the immediate vicinity of the boundary: up to distances comparable to the seed dispersal range, which for many tropical trees can exceed 100 meters. The entire forest's size distribution shifted, with fewer small trees and a proportionally older, larger structure.

The implications deepened when the researchers modeled two competing species with different dispersal ranges. In an undisturbed, spatially homogeneous forest, both species could coexist indefinitely. But the introduction of a boundary created an asymmetry: the species with the longer dispersal range suffered more at the boundary because more of its seeds were lost to the void beyond the forest. Over time, this boundary effect propagated inward, and the long-dispersing species declined—not just at the edge, but throughout the entire forest. The system didn't reach a new equilibrium with both species; it moved toward competitive exclusion, with the more localized species taking over.

The decay was approximately exponential, with a characteristic timescale that diverged as the two species' dispersal ranges became similar. When dispersal ranges were identical, both species were equally affected by the boundary and could coexist. But the slightest mismatch created a winner and a loser, with the disparity growing over time.

Why This Changes Things

For decades, ecologists have debated whether forest size distributions reflect neutral processes—where all trees are essentially equivalent in their competitive ability—or whether they reflect the sorting of species by different traits. The Padua team's model sits in a middle ground: species-neutral in its bulk dynamics, but sensitive to spatial structure and species-specific dispersal. This suggests that the patterns we observe in real forests may tell us less about competition among species than about the spatial architecture of recruitment and resource availability.

The finding that boundary disturbances propagate deeply into forest interiors challenges a common assumption in conservation: that the core of a large forest is insulated from edge effects. The team's analysis shows that this insulation is far weaker than often assumed. When seed dispersal is the only connection between spatial locations—a reasonable description for trees, which cannot move once established—losses at the boundary create a deficit that cannot be filled from within. The entire forest adjusts.

This matters for how we think about forest fragmentation. A road 500 meters wide carved through a forest isn't just a loss of habitat; it's a wound that can affect tree populations for hundreds of meters on either side, for decades or longer. The same logic applies to selective logging, which removes large trees and disrupts seed production without removing the canopy entirely. Even if the forest looks intact from above, its demographic structure may be fundamentally altered.

The connection between size-distribution exponents and forest condition is perhaps the most actionable finding. If the crossover height $h_c$ shifts with resource availability and seed input, then measuring the size distribution in a real forest could serve as a diagnostic tool. A forest with a high crossover—where the resource-competition regime extends to larger sizes—might indicate stressed conditions: low resources, poor recruitment, or both. A forest with a low crossover—where the allometric scaling dominates throughout—might indicate healthy conditions with abundant resources and vigorous seedling establishment.

What's Next

The model makes assumptions that will need testing. The researchers used idealized dispersal kernels and assumed that species are effectively neutral in their growth and competition parameters. Real forests are more heterogeneous: soil varies, topography creates microclimates, and species differ in their shade tolerance, growth rates, and seed biology. Whether the dual power-law structure survives these complexities is an open question.

Empirical work is the obvious next step. Long-term forest plots—some of which have been monitored for decades—could be used to test whether the crossover height in observed size distributions shifts as the researchers predict, when resources become scarcer or when fragmentation introduces boundary effects. Satellite-based remote sensing of forest structure, combined with dispersal data from genetic studies, could test whether the spatial patterns of species decline match the model's predictions.

The broader framework may also apply beyond forests. The researchers suggest the approach could describe any sessile community where growth is size-structured, mortality is density-dependent, and recruitment is nonlocal—coral reefs, grasslands, microbial mats. In each case, the interplay between resource competition, spatial exclusion, and dispersal may generate similar dual scaling regimes. If confirmed, this would constitute a genuinely unifying principle in ecology: not just forests, but any community where size matters.

What the model ultimately offers is a lens. Forests are complex, but their complexity is constrained by deep regularities—in metabolism, in geometry, in the physics of light and resources. By deriving those constraints mathematically, Grilletta and colleagues have shown that the tree-size distribution is not just an observation but a message. The exponents tell us what forces are acting; the crossover tells us where; and the boundary effects tell us what has been lost.