Coupling Spatial Scales Without Double Counting

Why Spatial-Scale Interfaces Need Their Own Accounting

A hybrid model can represent intracellular reactions with cell-specific state variables, cells as agents, and a transported signal as a tissue concentration field. The challenge is not simply to connect these components; it is to ensure that each physical transfer or reaction appears once, in the component where it belongs. If a cell removes signal from the tissue, the cell gains that amount and the tissue loses the same amount. If the cell produces a molecule internally and secretes only part of it, only the secreted amount enters the tissue field.

The earlier material on hybrid PDE-agent models establishes the broad idea of combining cell agents with spatial fields. Here the focus is the accounting at their interfaces: which quantities cross a boundary, how rates are represented on each scale, and how to check that a change in spatial representation does not create or destroy material.

Write Each Balance in Its Own Units

Consider a tissue domain \(\Omega\) with extracellular concentration \(c(\mathbf{x},t)\), measured as amount per unit volume. Let cell \(i\) contain amount \(a_i(t)\) of the same species, and let \(R_i\) be the net rate at which intracellular reactions change that amount. Thus \(R_i\) has units of amount per time; it can be positive or negative. The tissue may also have a local net production rate \(r_{\mathrm{tissue}}\), measured as concentration per time.

For a diffusing tissue species with constant diffusion coefficient \(D\), one consistent continuum balance is

Here \(w_i(\mathbf{x})\) describes where the exchange with cell \(i\) is applied. It is normalized so that \(\int_{\Omega}w_i(\mathbf{x})\,d\mathbf{x}=1\), assuming the cell's exchange is represented within the modeled domain. Since \(w_i\) has units of inverse volume, \(J_iw_i\) has units of amount per volume per time, matching \(\partial c/\partial t\). A positive inward cell flux decreases the tissue concentration; a negative \(J_i\), such as net secretion, increases it.

The corresponding intracellular amount balance is

This equation makes the division of responsibilities explicit. The interface term \(J_i\) transfers material between tissue and cell. The term \(R_i\) accounts for intracellular processes, such as synthesis, conversion, or degradation, that change the amount of the tracked species within the cell. If the tracked species is created by an intracellular reaction and then exported, the export is represented by a negative \(J_i\); the intracellular reaction itself is not also added to the tissue equation.

These balances rely on a few practical assumptions: both components refer to the same chemical species and amount units; the sign convention for \(J_i\) is used consistently; each interface transfer is applied with equal magnitude and opposite sign; and the spatial distribution \(w_i\) is normalized. A chosen membrane transport law can determine \(J_i\), but whatever law is used must return an amount per time in the stated sign convention.

Why Equal-and-Opposite Exchange Conserves Amount

A simple check is to integrate the tissue equation over space. Define the extracellular amount \(A_{\mathrm{e}}(t)=\int_{\Omega}c(\mathbf{x},t)\,d\mathbf{x}\). The divergence theorem and normalization of each \(w_i\) give

The boundary term is positive for net diffusive influx under this convention: the outward diffusive flux is \(-D\nabla c\cdot\mathbf{n}\). Summing the intracellular balances across cells gives \(\sum_i da_i/dt=\sum_i J_i+\sum_i R_i\). When the tissue and cellular equations are added, every interface term cancels:

Thus, interface exchange alone changes neither the combined tissue-plus-cell amount nor its accounting. The remaining terms have identifiable meanings: boundary transport, tissue production or loss, and intracellular production or loss. If an equation for the combined amount still contains \(J_i\), an exchange term has probably been omitted from one component or counted more than once.

Separate Intracellular Production from Export

A frequent source of double counting is to use an intracellular production rate as though it were an extracellular source. Internal synthesis first changes the intracellular amount. Only material that crosses the cell boundary contributes directly to the extracellular field. A simple cell balance can make the distinction visible: synthesis adds to the intracellular pool, intracellular degradation removes from it, and secretion is an outward interface flux.

Distribute Exchange Across Space Without Changing Its Total

A cell-agent model often assigns each cell’s exchange to one or more tissue-grid voxels. Let \(\alpha_{ij}\) be the fraction of cell \(i\)'s exchange assigned to voxel \(j\). These fractions should satisfy \(\alpha_{ij}\geq 0\) and \(\sum_j\alpha_{ij}=1\) for each cell whose exchange lies in the modeled domain. If voxel \(j\) has volume \(V_j\), the cell-exchange contribution to its concentration rate is

The division by voxel volume is essential: the field equation uses concentration per time, whereas each \(J_i\) is amount per time. The fractions allocate the full exchange spatially; they do not multiply the total exchange by a new amount. If a cell spans several voxels, the allocation can reflect overlap or a chosen spatial kernel, but it must still sum to one.

Diagnostics, Alternatives, and Limitations

A coupling scheme should be checked in both amount units and spatial units. Useful diagnostics include the following:

• Check signs. With inward \(J_i>0\), the cell gains \(J_i\) and the field loses \(J_i\). Secretion has \(J_i<0\), so the tissue contribution \(-J_i\) is positive.
• Check units. A cell-level flux is amount per time. Before adding it to a concentration equation, distribute it spatially and divide by the relevant volume, or use a normalized continuous kernel.
• Check normalization. Confirm that each cell’s allocation fractions sum to one, or that its spatial kernel integrates to one. Otherwise, the field receives only part of the flux or an artificial multiple of it.
• Check an amount residual. Over a time step, compare the tissue amount change and the cell amount changes against boundary transport and reaction terms. A residual that grows as the mesh is refined can indicate inconsistent volume scaling or allocation.
• Check for duplicated mechanisms. If an uptake term is already included as cell-to-field exchange, do not also add the same uptake as an independent tissue sink. Likewise, do not add intracellular synthesis to the field unless it is explicitly the amount exported.

There are several ways to represent the interface. A localized agent-to-field scheme assigns each cell’s rate to nearby voxels using normalized fractions or a kernel; it retains spatial heterogeneity but requires careful allocation. A homogenized tissue term replaces many cell-specific exchanges with an averaged source or sink per volume; it can be efficient when local cell variation is not important, but its average rate must still be derived from the total cell exchange divided by tissue volume. A boundary-resolved continuum model represents exchange on cell surfaces and integrates a surface flux; it can express local membrane transport directly, but requires appropriate geometry and numerical boundary treatment. These approaches differ in spatial detail, not in the conservation requirement that transferred amounts balance across the interface.

A conservative interface does not guarantee that the biological model is correct. The transport law may be poorly supported, the chosen spatial kernel may blur a real local gradient, or the assumed tissue volume may omit inaccessible regions. Nor does conservation of one species imply conservation of every chemical species in a reaction network: reactions can consume one species and produce another. Each tracked species needs a consistent balance, with reaction stoichiometry represented in the appropriate component.