Dryland vegetation organizes into spatial patterns — bands, spots, labyrinths — that have attracted theoretical attention because they look like Turing patterns. The standard model proposes scale-dependent feedbacks: plants facilitate neighbors by concentrating water locally (short-range activation) while depleting water from distant competitors (long-range inhibition). This produces regular patterns through deterministic bifurcation, and the patterns undergo catastrophic collapse when water availability drops below a critical threshold. The framework is elegant, predictive, and testable. It also assumes steady rainfall and continuous vegetation density — assumptions that fail in the actual drylands where these patterns occur.
Gimenez-Romero, Afful, Greenwood, Matias, and Gordillo (arXiv 2602.05583, February 2026) build an individual-based model that replaces both assumptions. Rainfall arrives intermittently — discrete pulses separated by dry intervals, matching the statistical structure of real semiarid precipitation. Vegetation consists of individual plants, not a continuous density field, with demographic stochasticity in germination, growth, and mortality. Each plant has a local Allee effect: its fitness depends on the density of its immediate neighborhood, not on the landscape average.
The patterns that emerge are not Turing patterns. They are irregular clusters — heterogeneous in size, spacing, and shape — that bear little resemblance to the spots and stripes of reaction-diffusion theory but closely resemble aerial photographs of actual dryland vegetation. The irregularity is not noise added to a regular pattern. It is the pattern, produced by the interaction between intermittent rainfall, demographic randomness, and local density dependence.
The key result is about resilience. In the deterministic model, ecosystem collapse is abrupt: the regular pattern destabilizes at a critical aridity threshold and the vegetation vanishes. In the stochastic individual-based model, decline is gradual. Clusters thin, shrink, and occasionally vanish, but the ecosystem does not undergo a sharp transition. And the predictor of whether a cluster survives is not total biomass across the landscape but local density within the cluster. A population can persist at vanishingly small average density across the landscape if local density within surviving clusters stays above the Allee threshold.
This is the spatial Allee effect applied to ecosystem persistence: the relevant density for survival is not the mean but the local maximum. A landscape that looks nearly empty — average vegetation cover approaching zero — can be ecologically stable if the remaining plants are clustered tightly enough that each individual has sufficient neighbors. The global average is misleading because survival is decided locally.
The persistence boundary follows a power law in the product of cluster activation frequency (how often a cluster receives a rainfall pulse) and cluster duration (how long the pulse sustains growth). Reducing rainfall frequency requires increasing rainfall intensity — and vice versa — to maintain the cluster above the Allee threshold. This predicts sharp phase transitions at the cluster level: individual clusters collapse suddenly (local extinction) even as the landscape-level decline is gradual (the rate of cluster loss increases smoothly with aridity).
The implication for monitoring is practical. Landscape-average vegetation indices — the standard remote sensing metrics — miss the relevant dynamics because they average over the spatial structure that determines resilience. A landscape losing vegetation uniformly and a landscape maintaining dense clusters while emptying between them look identical in the average but have different trajectories. The first is declining. The second may be stable, with its resilience concentrated in the spatial structure that the average erases.