Individual based model with competition in spatial ecology

We analyze an interacting particle system with a Markov evolution of birth-and-death type. We have shown that a local competition mechanism (realized via a density dependent mortality) leads to a globally regular behavior of the population in course of the stochastic evolution.

💡 Research Summary

The paper presents a rigorous mathematical analysis of an individual‑based model (IBM) for spatial ecology, commonly referred to as the BDLP (Bolker‑Pacala‑Dieckmann‑Law) model, defined on an infinite continuous habitat ℝᵈ. The population consists of motionless plants (or particles) that undergo stochastic birth‑and‑death events. Birth occurs through seed production: each existing individual independently produces seeds at a constant rate κ⁺, and each seed is displaced in space according to a symmetric, integrable dispersion kernel a⁺(·). Death is governed by two components: a constant intrinsic mortality rate m > 0 and a density‑dependent competition mortality. The latter is modeled by a non‑negative competition kernel a⁻(·) which adds an extra death rate proportional to the sum of a⁻ evaluated at the distances to all other individuals. Thus the total death rate for an individual at x is m + ∑_{y∈γ{x}} a⁻(x−y).

The authors formulate the dynamics as a Markov process on the configuration space Γ of locally finite point sets. They introduce the K‑transform (and its dual K*) to pass from probability measures μ on Γ to correlation measures ρ_μ on the space of finite configurations Γ₀, and define correlation functions k^{(n)}t (the n‑point densities) as the Radon‑Nikodym derivatives of ρ{μ_t} with respect to the Poisson reference measure λ_z. Using these tools, they derive an infinite hierarchy of evolution equations for the correlation functions (equation (3.2) in the paper). Each k^{(n)}_t evolves under a linear operator L_i a⁺ (the generator of the pure birth part) together with nonlinear interaction terms that involve the competition kernel a⁻ and lower‑order correlation functions. The hierarchy is formally analogous to the BBGKY hierarchy in statistical physics.

A first major investigation concerns the model without competition (a⁻≡0), which reduces to the continuous contact model. In this case the authors recover known results: the one‑point function grows or decays exponentially depending on whether κ⁺>1 or κ⁺<1, while the critical case κ⁺=1 yields a constant mean density. However, higher‑order correlation functions exhibit factorial growth (∼n! Cⁿ), reflecting strong clustering. This phenomenon is quantified by estimate (3.3) and demonstrates that, without competition, the system tends to form dense clusters.

The core contribution of the paper lies in analyzing the effect of a non‑trivial competition kernel. The authors prove two pivotal results under suitable conditions: (i) if the intrinsic mortality m is sufficiently large and the competition kernel a⁻ is strong enough (in the sense that its L¹‑norm exceeds a certain threshold), then for every n the correlation functions satisfy a sub‑Poissonian bound k^{(n)}_t ≤ Cⁿ n! uniformly in time. This bound implies that clustering is suppressed and the spatial distribution remains more regular than a Poisson process. (ii) Under the same conditions, the hierarchy admits a uniform in time bound, leading to the existence of a unique invariant measure, which is the Dirac measure concentrated on the empty configuration. Consequently, the stochastic dynamics is asymptotically extinct: the population almost surely disappears as t → ∞.

The proofs rely on functional‑analytic techniques developed for configuration spaces. The authors exploit the positivity‑preserving nature of the semigroup generated by L_i a⁺, apply the Minlos lemma to handle multiple integrals over configuration space, and construct appropriate Lyapunov‑type functionals to obtain a priori estimates on the correlation functions. They also discuss the closure problem: truncating the hierarchy at a finite order and approximating higher‑order correlations by products of lower‑order ones. While the closure is not unique, the paper shows that any closure preserving the sub‑Poissonian bound leads to consistent dynamics under the imposed conditions.

In summary, the paper provides a mathematically rigorous demonstration that local competition, modeled via a density‑dependent mortality term, fundamentally alters the long‑term behavior of spatial ecological systems. It prevents the explosive clustering seen in pure birth models, enforces uniform bounds on all correlation functions, and drives the system toward extinction when competition and intrinsic mortality are sufficiently strong. The methodology—combining K‑transform techniques, hierarchical equations, and functional analysis on configuration spaces—offers a robust framework that can be extended to other interacting particle systems in continuous space, such as models of epidemic spread, chemical reactions, or ecological communities with moving individuals.