Reservoir-Flux Network¶
Core Idea¶
A reservoir-flux network is the structural arrangement in which a system is decomposed into a finite set of named reservoirs — stocks, compartments, pools, accounts, levels — connected by fluxes — flows, transfer rates, directed channels — under a conservation closure that holds the total content of the system invariant except across an explicitly declared boundary. Three commitments are jointly load-bearing, and dropping any one collapses the family of reasoning the pattern licenses. Reservoirs individuate the things whose accumulated levels matter. Fluxes specify the directed pathways along which content moves between reservoirs. Conservation asserts that whatever leaves one reservoir arrives somewhere else inside the bounded system, or else crosses a named boundary to the outside, so that the summed contents — corrected for boundary exchanges — do not change under the internal dynamics.
The conservation closure is the discriminating feature. It is what separates the pattern from a bare graph of connected nodes (which carries no conservation) and from a single isolated flow (which carries no graph structure and no stocks). The closure is also what makes the characteristic reasoning catalogue available: mass-balance accounting (inflow equals outflow plus accumulation), steady-state analysis (finding the reservoir levels at which all fluxes balance), residence-time calculation (reservoir size divided by through-flux), time-constant estimation (how long a perturbation takes to relax around the graph), and shock propagation (how a disturbance to one reservoir redistributes through the connected topology). Without conservation the bookkeeping that makes intervention calculable simply disappears.
A fourth structural feature is naming: reservoirs are individuated and labelled rather than treated as anonymous nodes. The decision of where to draw reservoir boundaries — lumping a heterogeneous pool into one reservoir versus splitting it into sub-pools with their own internal flux structure — is itself the major modelling choice, and it determines where every later intervention can be located on the graph. Once the reservoirs are named and the fluxes drawn, the question "where should we act?" becomes a question about a specific edge or node, and the question "is our account complete?" becomes an auditable arithmetic identity.
How would you explain it like I'm…
Buckets and Pipes
Tanks That Never Leak
Stocks, Flows, and Conservation
Structural Signature¶
the set of named reservoirs — the directed fluxes connecting them — the rate laws governing the fluxes — the conservation closure over the bounded total — the declared boundary across which content is non-conserved — the reservoir-individuation (naming) choice
A system is a reservoir-flux network when each of the following holds:
- Named reservoirs. The system is decomposed into a finite set of individuated, labelled stocks — compartments, pools, accounts, levels — whose accumulated contents are the quantities that matter. Naming, not anonymous nodehood, is part of the structure: where reservoir boundaries are drawn is the major modelling choice.
- Directed fluxes. The reservoirs are connected by flows — transfer rates, channels — specifying the directed pathways along which content moves from one reservoir to another.
- Rate laws on the fluxes. Each flux is governed by a law (often first-order in a reservoir's content, possibly nonlinear) that fixes how fast content moves, making the dynamics calculable rather than merely topological.
- A conservation closure. Whatever leaves one reservoir arrives in another inside the bounded system, so the summed contents — corrected for boundary exchange — are invariant under the internal dynamics. This is the discriminating feature: it separates the pattern from a bare graph (no conservation) and from a single isolated flow (no stocks, no graph).
- A declared boundary. An explicit interface separates inside (conserved) from outside; content may cross it as boundary fluxes that are non-conserved with respect to the inside. Misplacing this boundary is the most common and most diagnostic error, since a failed conservation check points straight at it.
Together these license the characteristic reasoning catalogue — mass-balance accounting, steady-state analysis, residence-time calculation, time-constant estimation, and shock propagation — and make "is our account complete?" an auditable arithmetic identity. Drop the conservation closure and the entire calculable bookkeeping disappears.
What It Is Not¶
- Not a bare
network. A network is a graph topology — nodes and ties — that carries no conservation. The reservoir-flux network is the conjunction of graph plus stocks plus a conservation closure, the minimum required for mass-balance reasoning to work. - Not a single isolated flow. A lone directed channel describes one movement but carries no graph structure and no stocks. This prime requires multiple named reservoirs whose summed contents are conserved.
- Not
turnover. Turnover is the rate at which a stock's contents are replaced; this prime is the whole reservoir-and-flux structure under conservation, from which turnover (and residence time) is one derived quantity. - Not
buffering. Buffering is the capacity of one reservoir to absorb fluctuation; the reservoir-flux network is the multi-reservoir conserved system in which buffering is a property of a particular stock. - Not
equilibrium. Equilibrium is a state (flux balance, no net change) that a reservoir-flux network may or may not reach; the prime is the structural arrangement whose dynamics produce equilibrium, oscillation, or runaway depending on topology and rate laws. - Common misclassification. Calling a connected diagram of pools a reservoir-flux network when no total is conserved. Without the closure, the mass-balance bookkeeping that makes intervention calculable simply disappears — it is then a bare graph, not this prime.
Broad Use¶
- Biogeochemistry. The canonical case: the carbon cycle as reservoirs (atmosphere, ocean surface, deep ocean, biosphere, soils, fossil reserves) linked by photosynthesis, respiration, dissolution, and combustion fluxes, with total carbon conserved. The nitrogen, phosphorus, water, and sulfur cycles share the shape.
- Pharmacokinetics. Compartmental PK/PD models represent the body as one to three reservoirs with first-order rate constants between them and to elimination; drug mass is conserved minus elimination.
- Epidemiology. SIR/SEIR models partition a population into Susceptible, Infected, and Recovered reservoirs with infection, recovery, and death fluxes; total population is conserved, and interventions read as perturbations on the flux constants.
- Macroeconomics. Flow-of-funds and sectoral-balance accounting treat households, firms, government, and rest-of-world as reservoirs whose monetary flows conserve by double-entry bookkeeping.
- System dynamics. The entire Forrester stock-and-flow discipline is built around stocks, rates, and the equations connecting them.
- Ecology and hydrology. Trophic energy budgets and watershed water balances are reservoir-level analyses with thermodynamic or hydrological conservation as the closure.
Clarity¶
Naming the reservoir-flux network as its own object separates three things that surface vocabulary persistently conflates. A flow is a single directed channel — useful for describing one movement, useless for accounting. A network is a graph topology — useful for describing connectivity, silent on conservation. A reservoir-flux network is the conjunction with closure: graph plus stocks plus conservation, the minimum required for mass-balance reasoning to work. When someone says "the carbon cycle is just a network," they have dropped the conservation closure that licenses the policy arithmetic; when someone says "pharmacokinetics is just a flow," they have dropped the compartment structure that organises the dose calculation.
The pattern also makes the closure boundary an explicit modelling choice. Every reservoir-flux model must declare what counts as inside (conserved) and what crosses to outside (boundary fluxes, non-conserved with respect to the inside). Misplacing that boundary — treating the atmosphere as closed when emissions cross in, treating the body as closed when elimination crosses out — is at once the most common reservoir-flux error and the most diagnostic, because a failed conservation check points straight at the misplaced boundary.
Manages Complexity¶
The pattern compresses an arbitrarily complicated multi-substance, multi-place, multi-rate system into a short specification: how many reservoirs, what fluxes connect them, what rate laws govern the fluxes, and what total is conserved. Once those are fixed, a wide family of derived quantities becomes computable from the same pipeline regardless of substrate — steady-state levels, residence times, equilibration time constants, sensitivity coefficients, and shock-propagation patterns. The same machinery runs whether the conserved content is carbon, dollars, viral particles, or vehicles.
The structural commitment also disciplines the modelling. Conservation is auditable: if the contents do not add up, then a reservoir is mis-sized, a flux is mis-specified, or a boundary exchange has been overlooked. This is precisely the diagnostic value that double-entry bookkeeping discovered for finance, generalised — an arithmetic identity that can be checked against measurement, with every discrepancy forcing the model back into contact with reality. The discipline turns "the model seems wrong" into a short, finite list of possible structural faults.
Abstract Reasoning¶
The dynamics on a reservoir-flux network admit clean mathematical treatment: a system of coupled rate equations, one per reservoir, in which each reservoir's change equals the sum of its inflows minus the sum of its outflows, with conservation enforced when boundary fluxes vanish. Steady-state levels are the solution of a linear system in the flux coefficients; residence times read off directly; the response to a perturbation is governed by the eigenstructure of the system's Jacobian. The graph's topology constrains the qualitative behaviour: pure feed-forward networks relax monotonically to equilibrium, while networks containing loops can oscillate or settle into multiple stable states depending on flux nonlinearities.
The pattern also licenses a clean abstraction ladder. At one extreme, lumped models with a few reservoirs are analytically tractable and yield order-of-magnitude reasoning; at the other, spatially distributed models written as continuous fields over reservoir position handle fine-grained structure. The choice of granularity is itself a structural trade: adding reservoirs adds parameter burden and data demand, while lumping risks hiding dynamics behind an aggregate stock. A trained reasoner uses the same questions at every rung — what is conserved, what flows, on what timescale, across what boundary — and so can move up and down the ladder without changing vocabulary.
Knowledge Transfer¶
The reservoir-flux pattern transfers with unusual cleanness because its core identities are substrate-blind, and the role mappings are exact. The reservoir maps to a compartment in pharmacokinetics, a population class in epidemiology, a sector in flow-of-funds, a trophic level in ecology, a warehouse in inventory management. The flux maps to a rate constant, an infection rate, a monetary flow, a feeding rate, a shipment. Conservation maps to mass balance, population conservation, double-entry bookkeeping, the first law of thermodynamics, physical inventory continuity. Residence time maps identically across all of them: a reservoir's contents divided by the flux running through it, whether the contents are drug molecules, infected individuals, dollars, or cars.
Because the mappings are exact rather than analogical, closed-form results travel intact. The one- and two-compartment solutions that pharmacology uses for dosing carry over without modification to SIR-family epidemic models, since both are linear reservoir-flux networks. The sectoral-balance identity from macroeconomics ("one sector's surplus is another's deficit") is the same closure that yields trophic energy budgets in ecology. The system-dynamics insight that a stock keeps rising even as its inflow flux levels off — the "bathtub" result — transfers directly to climate reasoning about atmospheric carbon and to inventory reasoning about backorders. Most powerfully, the conservation-failure diagnostic transfers as a single reusable move: across auditing, climate modelling, epidemiology, and supply-chain control, "the contents are not adding up, therefore a reservoir is missing, a flux is unmeasured, or the boundary is drawn wrong" is the same inference with the same three resolution paths. A practitioner who has internalised the pattern in one substrate arrives in another already knowing which questions are answerable, which quantities are computable, and which discrepancies are diagnostic — the transfer cost is close to zero because nothing about the reasoning was ever tied to the original medium.
Examples¶
Formal/abstract¶
The two-compartment pharmacokinetic model is the pattern with every role instantiated and the conservation closure doing visible arithmetic. The named reservoirs are two: a central compartment (blood and well-perfused tissue) and a peripheral compartment (slowly-perfused tissue). The directed fluxes connect them — a rate constant \(k_{12}\) moving drug from central to peripheral, \(k_{21}\) returning it, and an elimination flux \(k_{10}\) carrying drug out of the central compartment across the declared boundary (metabolism and excretion). The rate laws are first-order: each flux equals a rate constant times the source compartment's current content, so the dynamics are the coupled linear system \(\dot{C}_1 = -(k_{12}+k_{10})C_1 + k_{21}C_2\) and \(\dot{C}_2 = k_{12}C_1 - k_{21}C_2\). The conservation closure is exact: total drug mass equals what is in both compartments plus what has been eliminated across the boundary, and that sum equals the dose — an auditable identity. The declared boundary is elimination; treating the body as closed (forgetting \(k_{10}\)) is the diagnostic error, and a conservation check that fails to balance points straight at the missing elimination flux. From this specification the entire derived catalogue follows mechanically: steady-state levels under continuous infusion solve a linear system in the rate constants; the central-compartment residence time is its volume divided by clearance; the biexponential plasma decay (fast distribution phase, slow elimination phase) is read off the eigenstructure of the system's Jacobian; and a dosing perturbation propagates through the two-compartment topology on time constants set by those same eigenvalues. None of this reasoning references chemistry — it is pure reservoir-flux bookkeeping.
Mapped back: Central and peripheral compartments are the named reservoirs, \(k_{12}/k_{21}/k_{10}\) are the directed fluxes under first-order rate laws, dose conservation across the elimination boundary is the conservation closure, and clearance-over-volume is residence time — a reservoir-flux network with the mass-balance identity made numerically explicit.
Applied/industry¶
The SIR epidemic model and macroeconomic flow-of-funds accounting run the identical structure in unrelated substrates, and the closed-form results transfer between them because both are linear (or near-linear) reservoir-flux networks. In SIR, the named reservoirs are three population classes — Susceptible, Infected, Recovered — and the directed fluxes are infection (S to I, governed by a rate law nonlinear in the product \(S \cdot I\)) and recovery (I to R, first-order in \(I\)). The conservation closure is population conservation: \(S + I + R\) is invariant (with births and deaths as declared boundary fluxes if included), so the model's total always balances, and a public-health planner reads interventions directly as perturbations on the flux constants — vaccination drains S, isolation cuts the infection rate, treatment raises the recovery rate. The same pattern governs flow-of-funds: the reservoirs are economic sectors (households, firms, government, rest-of-world), the fluxes are monetary flows between them, and the conservation closure is double-entry bookkeeping, which guarantees the sectoral-balance identity "one sector's surplus is exactly another's deficit" — the same closure that yields trophic energy budgets in ecology. The conservation-failure diagnostic transfers as one reusable move across both: when the contents do not add up — an epidemic count that does not reconcile, a national account that does not balance — a reservoir is mis-sized, a flux is unmeasured, or the boundary is drawn wrong, with the same three resolution paths. An epidemiologist sizing an outbreak and a national-accounts statistician auditing a balance are running one bookkeeping discipline.
Mapped back: S/I/R classes and economic sectors are named reservoirs; infection-and-recovery and inter-sector monetary flows are directed fluxes; population conservation and double-entry bookkeeping are the conservation closures; the "it doesn't add up, so a reservoir/flux/boundary is wrong" check is the shared diagnostic — the same prime in epidemiology and macroeconomics.
Structural Tensions¶
T1 — Conservation Closure versus Open Graph (boundary). The closure is the discriminating feature: it separates a reservoir-flux network from a bare graph (no conservation) and makes mass-balance bookkeeping work. The characteristic failure is misplacing the boundary — treating the atmosphere as closed when emissions cross in, the body as closed when elimination crosses out — so a conserved total is asserted over a system that is actually open. The diagnostic is the conservation check itself: when the contents fail to balance, the fault is a misplaced boundary, a missing reservoir, or an unmeasured flux, and a failed identity points straight at the boundary that was drawn wrong.
T2 — Reservoir Lumping versus Splitting (scalar). Where reservoir boundaries are drawn — lumping a heterogeneous pool into one stock versus splitting it into sub-pools with their own flux structure — is the major modelling choice, and it trades aggregation error against parameter burden. The failure runs both ways: lumping hides internal dynamics behind an aggregate stock (a single "ocean" reservoir masking surface-versus-deep turnover), while over-splitting demands data the system cannot supply. The diagnostic is to ask whether the dynamics of interest live within a candidate reservoir: if a lumped stock has internally heterogeneous residence times, the lumping is hiding exactly the behaviour the model needs, and the granularity must be refined.
T3 — Stock versus Flow (temporal). The "bathtub" result is the prime's signature confusion: a stock keeps rising even as its inflow flux levels off, because the stock integrates the flux over time. The failure is reading a stabilising flow as a stabilising stock — concluding atmospheric carbon is under control because emission rates plateaued, when the stock keeps climbing as long as inflow exceeds outflow. The diagnostic is to distinguish the level from the rate explicitly: a flux returning to a constant does not return the reservoir to its prior level, and any claim of stabilisation must be checked against the accumulation identity, not the flow alone.
T4 — Residence Time versus Absolute Stock Size (scalar / measurement). Residence time — reservoir size divided by through-flux — is a different and often more decision-relevant quantity than the stock's absolute size, yet the two are routinely conflated. The failure is sizing intervention by the stock when the dynamics are set by turnover: a large reservoir with fast through-flux responds quickly to a flux change, while a small one with slow throughput is sluggish, and reasoning from size alone mis-predicts both. The diagnostic is to compute residence time before estimating response speed: how fast a reservoir equilibrates after a perturbation is governed by its turnover, not its magnitude, so the through-flux must be measured alongside the level.
T5 — Feed-Forward versus Looped Topology (coupling). The graph's topology constrains qualitative behaviour: pure feed-forward networks relax monotonically to equilibrium, while networks containing loops can oscillate or settle into multiple stable states under flux nonlinearities. The failure is assuming monotone relaxation on a looped graph — expecting a perturbation to decay smoothly when feedback loops can produce overshoot, oscillation, or a flip to an alternative steady state. The diagnostic is to inspect the flux graph for cycles and nonlinear rate laws before predicting dynamics: where loops and nonlinearity coexist, the eigenstructure can carry complex modes, and monotone-relaxation intuition from feed-forward reasoning fails.
T6 — Conservation Bookkeeping versus Rate-Law Dynamics (scopal). Conservation tells you content is preserved but says nothing about how fast it moves; the rate laws on the fluxes carry the dynamics, and the two are independent commitments. The failure is treating a balanced account as a complete model — a conservation identity that closes perfectly while the rate laws are mis-specified, so steady-state levels and time constants are wrong even though nothing leaks. The diagnostic is to check that the flux rate laws are validated separately from the mass-balance closure: an auditable, balancing account can still mispredict every timescale if the rate constants are wrong, since conservation constrains totals, not speeds.
Structural–Framed Character¶
Reservoir-flux network sits firmly at the structural end of the structural–framed spectrum, consistent with its structural label and aggregate of 0.0. It is a pure formal pattern — a finite set of named reservoirs linked by fluxes under a conservation closure — and every diagnostic reads structural.
No home vocabulary travels with it: the same stocks-plus-fluxes-plus-conservation skeleton is recognised as compartments and exchange in biogeochemistry, compartments and clearance in pharmacokinetics, S/I/R compartments and transition rates in epidemiology, accounts and flows in macroeconomic flow-of-funds, levels and rates in system dynamics, and pools and energy transfers in ecological budgets — each told in its own field's words, with the conservation closure describing the same invariant under all of them (vocab_travels 0). It carries no inherent approval or disapproval: a reservoir or flux is neither good nor bad, only larger or smaller, faster or slower — a value-neutral accounting structure (evaluative_weight 0). Its origin is formal — a directed graph of stocks with a conservation law — statable with no appeal to human institutions (institutional_origin 0). It runs indifferently across physical, biological, and economic substrates — carbon among ocean and atmosphere, a drug among body compartments, money among sectors all instantiate it identically, requiring no human practice to exist (human_practice_bound 0). And invoking it merely recognises a conserved-flow structure already present in the system rather than importing an interpretive frame; the reservoirs and fluxes are there to be individuated whether or not anyone draws the diagram (import_vs_recognize 0). On every criterion it reads structural — one of the catalog's canonical structural primes, with no inherited frame beneath the stock-and-flow skeleton.
Substrate Independence¶
Reservoir–flux network is a maximally substrate-independent prime — composite 5 / 5 on the substrate-independence scale. Its domain breadth is total: the stocks-plus-fluxes-plus-conservation skeleton is recognised, not translated, across biogeochemistry (the carbon, nitrogen, phosphorus, water, and sulfur cycles), pharmacokinetics (compartmental PK/PD models), epidemiology (SIR/SEIR compartments), macroeconomics (flow-of-funds and sectoral-balance accounting), the whole Forrester stock-and-flow discipline of system dynamics, and ecology and hydrology (trophic energy budgets, watershed water balances) — substrates that share no material. Its structural abstraction is complete because the object is a directed graph of stocks linked by fluxes under a conservation law, carrying no domain content; carbon among ocean and atmosphere, a drug among body compartments, and money among sectors are individuated as reservoirs and fluxes by the identical formalism with no human practice required. Its transfer evidence is the strongest kind: the same governing equations carry across — a compartmental rate-constant model is the same mathematics whether it tracks a drug or an isotope, and the conservation closure (mass conserved minus elimination, population conserved, double-entry bookkeeping) is one invariant wearing different domain labels — so a system-dynamics modeller, a pharmacologist, and an epidemiologist write structurally identical equations. Recognised everywhere under one stock-and-flow vocabulary, translated nowhere, and unified by a single conservation law, the composite of 5 is fully earned.
- Composite substrate independence — 5 / 5
- Domain breadth — 5 / 5
- Structural abstraction — 5 / 5
- Transfer evidence — 5 / 5
Relationships to Other Primes¶
Foundational — no parent edges in the catalog.
Children (2) — more specific cases that build on this
-
Conservation Laws decompose Reservoir-Flux Network
The conservation closure (the discriminating feature) is supplied by conservation. The prime CONSUMES conservation as its bookkeeping component.
-
Network decompose Reservoir-Flux Network
A reservoir-flux network IS a graph (network) PLUS stocks PLUS conservation closure. The network/graph topology is one component the prime builds on (the file: 'graph plus stocks plus conservation, the minimum for mass-balance reasoning').
Neighborhood in Abstraction Space¶
Reservoir-Flux Network sits in a moderately populated region (57th percentile for distinctiveness): it has near-neighbors but no dense thicket of synonyms.
Family — Stocks, Flows & Buffering (16 primes)
Nearest neighbors
- Attractor Selection and Basin Control — 0.73
- Source-Sink Dynamics — 0.73
- Turnover — 0.71
- Conservation Laws — 0.70
- Asymmetric Flux — 0.70
Computed from structural-signature embeddings · 2026-06-14
Not to Be Confused With¶
The most important confusion is with the bare network, because a reservoir-flux network is a graph — but a graph with two further commitments that carry nearly all of the prime's reasoning power. A bare network is a topology of nodes and ties: it describes connectivity, betweenness, path length, and reachability, and is silent about anything flowing or accumulating. The reservoir-flux network adds stocks (the nodes are named reservoirs whose accumulated contents matter) and, decisively, a conservation closure (whatever leaves one reservoir arrives in another or crosses a declared boundary, so the summed contents are invariant). The closure is the discriminating feature, and it is exactly what licenses the characteristic catalogue — mass-balance accounting, steady-state analysis, residence-time calculation, shock propagation — none of which a bare graph supports. When someone says "the carbon cycle is just a network," they have dropped the conservation that turns "is our account complete?" into an auditable arithmetic identity. A practitioner who models a conserved system as a bare graph keeps the connectivity and loses the bookkeeping, which is to lose the prime's entire diagnostic value: the ability to localise a missing reservoir, an unmeasured flux, or a misplaced boundary from a failed conservation check.
A second genuine confusion is with turnover, the embedding-nearest neighbour. Turnover is a rate — how quickly a stock's contents are replaced — and its close relative residence time (reservoir size divided by through-flux) is one of the prime's most decision-relevant derived quantities. But turnover is a property computed from a reservoir-flux network, not the network itself. The prime is the whole conserved structure of named reservoirs, directed fluxes, rate laws, and closure, from which turnover and residence time are read off as outputs. The distinction is load-bearing because residence time is routinely conflated with the stock's absolute size: a large reservoir with fast through-flux equilibrates quickly while a small one with slow throughput is sluggish, and reasoning from turnover alone misses the reservoir structure that produces it, while reasoning from the network without computing turnover misses the timescale of response. Treating the prime as turnover collapses the structure into one of its derived rates and forfeits the rest of the catalogue — the steady-state levels, the shock propagation, the topology-dependent qualitative behaviour.
A third confusion is with equilibrium. Equilibrium is a state of a reservoir-flux network — the configuration of reservoir levels at which all fluxes balance and the contents stop changing. The reservoir-flux network is the structure whose dynamics may or may not reach such a state: a pure feed-forward topology relaxes monotonically to equilibrium, but a topology containing loops and nonlinear rate laws can oscillate or settle into multiple stable states, never resting at a single equilibrium. The relationship is structure-to-possible-state. Conflating them produces the prime's signature "bathtub" error in another guise — assuming a balanced or stabilising condition (equilibrium) when the structure's dynamics are still in motion, for instance concluding a stock has stabilised because its inflow flux leveled off, when the stock keeps rising as long as inflow exceeds outflow. The discriminating question is whether the object of interest is a resting configuration (equilibrium) or the conserved structure whose rate laws and topology determine whether any resting configuration is reached at all (this prime).
These distinctions matter because each mis-framing discards a different tool. A network framing keeps connectivity but loses the conservation bookkeeping; a turnover framing keeps one rate but loses the steady-state and shock-propagation catalogue; an equilibrium framing assumes a resting state the dynamics may never reach — whereas the prime's full apparatus (mass-balance audit, residence-time calculation, eigenstructure of the rate equations) follows precisely from holding the reservoirs, fluxes, rate laws, and conservation closure together as one object.
Solution Archetypes¶
No catalogued solution archetypes reference this prime yet.