A modeling strategy in which a fine-scale or localized model is inserted into a coarser host model with explicit interface provisions to exchange state, fluxes, or boundary data so that the two descriptions interact consistently over the embedding region.

Principle

Preserve consistency and conservation across scales by defining interfaces (boundary conditions, exchange operators, interpolation and restriction) that map variables between the embedded fine model and the coarse host while controlling reflections, overlap, and double-counting.

Demonstration

Embed a high-resolution atmospheric convection model inside a global climate model over a limited region: the nested convection solver receives large-scale winds and thermodynamic tendencies from the host model and returns resolved convective heating and momentum fluxes via a designed coupling operator at each coupling time step.

Misapplication

Inserting a fine-scale model without matching conserved quantities or without compatible temporal synchronization, producing spurious reflections at the interface, energy nonconservation, or inconsistent state updates between host and embedded models.

Consequence

Local detail and fidelity are improved where needed without resolving the whole domain at fine scale; correct implementation yields higher local accuracy and targeted computational effort, at the cost of more complex interfaces and potential stability constraints.

Reversal

Replace the embedded model with a homogenized parameterization or an upscaled closure inside the coarse model so that no distinct fine-scale solver is present and all effects are represented through modified constitutive relations.

Boundary

Applies when localized processes require higher resolution than the host model and when interfaces can be defined; excludes global monolithic multiscale solves that do not segregate an embedded solver, and is not the same as simple grid refinement without explicit signed interface exchange rules.

Semantic Tension

Tension with nested-grid or multi-domain methods that differ in whether the embedded solver is an independent model exchanging data in time (embedded coupling) versus simply refining mesh resolution under the same numerical solver; also competes conceptually with co-simulation where coupling timing and solver independence are primary concerns.

Synthesis

Embedded coupling is a targeted multiscale integration pattern: place a localized high-fidelity model inside a coarser host, design conservative and stable exchange operators, and thereby gain local accuracy while isolating computational complexity to the embedded region.