Population models can be discrete — tracking individual organisms with integer counts — or continuous — using density functions that evolve by partial differential equations. The continuum limit is standard: as population size grows, the discrete model “should” converge to the continuous one. But the convergence can fail at boundaries, and the failure has been poorly characterized.
Agostinelli et al. (arXiv:2603.15217) use matched asymptotic expansions to systematically derive continuum equations from discrete structured population models, identifying exactly where and how the continuum approximation breaks down.
The breakdown occurs in boundary layers — narrow regions in the structured variable (age, size, maturity) where the discrete model's behavior cannot be captured by any smooth continuum equation. At these boundaries, the discrete model has fundamentally discrete dynamics: individual organisms enter or leave the structured classes in integer steps, and no continuous approximation can resolve this stepping behavior.
The matched asymptotics resolve the ambiguity of truncation order that has plagued discrete-to-continuum derivations. When Taylor-expanding the discrete operators to obtain continuous PDEs, the question of where to truncate — first order, second order, higher? — affects the resulting equation qualitatively. The matched asymptotic approach determines the correct truncation by requiring that the continuum solution matches the discrete solution in the interior, while boundary layers handle the transition regions where matching fails.
The framework applies to structured population models generally — cell cycle models, age-structured demographics, size-structured ecology. In each case, the continuum PDE is valid in the bulk but requires discrete corrections near the boundaries of the structured variable's domain. The organisms at the edges of a population class don't know they're supposed to be smooth.