Capture Basin: Theory, Modeling, Applications

• Capture Basin is a term representing regions defined by geometric or dynamic criteria where states, fluids, or energy are captured or redirected.
• It spans applications in hydrology, dynamical systems, optimization, and quantum systems, with each field using domain-specific models and terminology.
• The concept emphasizes the need to specify dynamics, time horizons, and definitions to enable precise cross-domain comparisons and practical implementations.

“Capture basin” is not a single standardized term across the arXiv literature. In different research programs it denotes, or is closely approximated by, a drainage catchment assigned to an outlet, a finite-time return set inside a basin of attraction, a Lagrangian region whose trajectories are eventually intercepted by a sink, a solver-induced attraction region in nonconvex optimization, a schedule-dependent occupation region in quantum energy landscapes, or a topographically confined volume that traps air, seismic energy, or gravitationally bound particles (Oliveira et al., 2018, Schultz et al., 2017, Smith et al., 2018, Lauga et al., 18 May 2026, Wirgin, 12 Jul 2025). This suggests that the common structure is not a single formal definition but a recurring geometric idea: a basin-like region that organizes which states, parcels, waves, or trajectories are received, retained, or redirected by a target configuration or enclosing domain.

1. Formal meanings and terminological range

Several of the cited papers explicitly note that they do not use the exact term “capture basin,” even when they study a closely related object. In dynamical systems, the primary objects are the basin of attraction $B$, basin stability, and finite-time basin stability rather than a viability-theoretic capture basin (Schultz et al., 2017). In chaotic advection, the operative term is the captured region $C(m)$ of a sink (Smith et al., 2018). In global optimization, the explicit object is the solver-induced attraction set $\mathrm{Att}(M)$ of a local minimizer component $M$ (Lauga et al., 18 May 2026). In hydrology and geomorphology, the standard term is drainage basin or watershed rather than capture basin (Oliveira et al., 2018).

The formal objects most closely associated with “capture basin” in these papers are summarized below.

Domain	Term used in the paper	Formal object
Dynamical systems	Finite-time return/capture set	$\mathcal{C}_S(T):=\{x\in X:x\in B,\ V_S(x)\le T\}$
Flow capture	Captured region	$C(m)=\bigcup_{n=0}^{\infty}T^nS(m)$
Global optimization	Attraction set	$\mathrm{Att}(M):=\{y\in\mathbb{R}^d:T_f(y)\in M\}$
Quantum annealing	Basin via steepest descent	Pre-image of a local minimum under deterministic single-spin-flip descent
Hydrology	Drainage basin	Set of sites whose invasion-percolation cluster first reaches sink $S_k$

A recurrent misconception is that all “basin” uses are interchangeable. The literature does not support that equivalence. Some basins are deterministic sets under autonomous flow, some are probability-weighted subsets under perturbation laws, some are transport-capture regions defined by sinks, and some are physical containers whose geometry traps waves or dense fluids. A plausible implication is that cross-domain comparison is most precise when the admissible dynamics and the target object are made explicit.

2. Hydrologic, geomorphic, and engineered catchments

In hydrology, the closest classical meaning of capture basin is the drainage basin or catchment assigned to an outlet. “A universal approach for drainage basins” defines a drainage basin as all land cells topographically assigned to a common outlet/sink and extends the Invasion Percolation-Based Algorithm to delineate multiple basins at once from a gridded height field $L_x\times L_y$ with site heights $h_i$ and threshold $C(m)$0 in the planetary applications (Oliveira et al., 2018). In that framework, each above-threshold site is assigned to the sink $C(m)$1 first reached by its invasion-percolation cluster; watersheds are the interfaces between differently labeled sink domains, and anti-basins are obtained by applying the same construction to the inverted landscape. The paper further proposes a geometric relation for Hack’s law, $C(m)$2, and measures $C(m)$3 for Earth, close to the estimate $C(m)$4 (Oliveira et al., 2018).

A more explicitly computational formulation appears in “Development of a data model to facilitate rapid Watershed Delineation,” where a watershed system $C(m)$5 is represented as a directed graph $C(m)$6 whose vertices are stream reaches and whose local catchments are polygons $C(m)$7 attached to each reach (Haag et al., 2016). The total upstream watershed of a selected reach $C(m)$8 is $C(m)$9, assembled from upstream local polygons after Modified Nested Set labeling with discovery time $\mathrm{Att}(M)$0, finish time $\mathrm{Att}(M)$1, and root distance $\mathrm{Att}(M)$2. Log Reduced Graphs and Stitching Watershed then reduce query cost by storing aggregated basin fragments across logarithmically reduced graph levels. On the $\mathrm{Att}(M)$3 Delaware River Watershed, the proposed technique is reported to provide a 99–98% reduction in preprocessing, 96–80% reduction in query complexity, and a 76% reduction in storage costs relative to hypothetical alternatives (Haag et al., 2016).

In urban water engineering, the same basin logic becomes operational storage-and-release control. “Predictive Real-Time Control Optimization of a Stormwater Management System” treats a stormwater pond as a controlled detention/treatment basin with inflow $\mathrm{Att}(M)$4, outflow $\mathrm{Att}(M)$5, water depth $\mathrm{Att}(M)$6, area $\mathrm{Att}(M)$7, maximum depth $\mathrm{Att}(M)$8, and maximum outflow $\mathrm{Att}(M)$9 (Shishegar et al., 2019). During wet periods the control problem is posed as

$M$0

subject to storage and continuity constraints, including

$M$1

During dry periods, generalized rules hold water to promote sedimentation, with a target retention time of $M$2 before the next rainfall when feasible (Shishegar et al., 2019). “Open storm” generalizes this into an open-source sensing-and-control stack with a PSOC5-LP node, cellular communications, cloud services, and actuators such as butterfly valves and gate valves; in Ann Arbor, retrofitted ponds used two $M$3 gate valves and one $M$4 butterfly valve to modulate discharges, increase residence time by nearly $M$5 hours, and reduce a watershed outlet peak that would have been nearly $M$6 to a measured $M$7 (Bartos et al., 2017).

Catchment-scale pollution work further shows that a basin may function simultaneously as a receiving, mixing, and redistribution structure. In the Lower Llobregat River basin, an investigative monitoring design across $M$8 surface-water sites and $M$9 groundwater locations combined pesticide occurrence and isotopes $\mathcal{C}_S(T):=\{x\in X:x\in B,\ V_S(x)\le T\}$0-NO$\mathcal{C}_S(T):=\{x\in X:x\in B,\ V_S(x)\le T\}$1, $\mathcal{C}_S(T):=\{x\in X:x\in B,\ V_S(x)\le T\}$2-NO$\mathcal{C}_S(T):=\{x\in X:x\in B,\ V_S(x)\le T\}$3, $\mathcal{C}_S(T):=\{x\in X:x\in B,\ V_S(x)\le T\}$4-NH$\mathcal{C}_S(T):=\{x\in X:x\in B,\ V_S(x)\le T\}$5, and $\mathcal{C}_S(T):=\{x\in X:x\in B,\ V_S(x)\le T\}$6 to discriminate sources (Postigo et al., 2020). Surface-water total pesticide concentrations were one order of magnitude higher than in groundwater, and stable isotopes indicated an organic origin of nitrate in surface waters and most groundwater samples, with urban/industrial activity identified as the main pressure on surface-water quality and, to a large extent, also on groundwater quality (Postigo et al., 2020). This suggests that, in managed catchments, a basin is not merely a topographic partition but also a coupled infrastructure–hydrology domain in which source signatures are redistributed across channels, aquifers, and reuse pathways.

3. Dynamical-systems basins, finite-time return, and probabilistic recapture

In nonlinear dynamics, the closest formal analogue to a capture basin is a return set inside a basin of attraction. “Bounding the first exit from the basin” studies an autonomous ODE

$\mathcal{C}_S(T):=\{x\in X:x\in B,\ V_S(x)\le T\}$7

with stable fixed point $\mathcal{C}_S(T):=\{x\in X:x\in B,\ V_S(x)\le T\}$8 and basin of attraction $\mathcal{C}_S(T):=\{x\in X:x\in B,\ V_S(x)\le T\}$9, then adds jump-like perturbations at times $C(m)=\bigcup_{n=0}^{\infty}T^nS(m)$0 with magnitudes $C(m)=\bigcup_{n=0}^{\infty}T^nS(m)$1 drawn from a density $C(m)=\bigcup_{n=0}^{\infty}T^nS(m)$2 (Schultz et al., 2017). Basin stability is defined as

$C(m)=\bigcup_{n=0}^{\infty}T^nS(m)$3

and finite-time basin stability as

$C(m)=\bigcup_{n=0}^{\infty}T^nS(m)$4

where $C(m)=\bigcup_{n=0}^{\infty}T^nS(m)$5 is the return time to a transverse surface $C(m)=\bigcup_{n=0}^{\infty}T^nS(m)$6. The implicit finite-time return/capture set is

$C(m)=\bigcup_{n=0}^{\infty}T^nS(m)$7

so that $C(m)=\bigcup_{n=0}^{\infty}T^nS(m)$8 is the probability mass of that set under $C(m)=\bigcup_{n=0}^{\infty}T^nS(m)$9 (Schultz et al., 2017). The paper then introduces the independence time

$\mathrm{Att}(M):=\{y\in\mathbb{R}^d:T_f(y)\in M\}$0

and proves the lower bound

$\mathrm{Att}(M):=\{y\in\mathbb{R}^d:T_f(y)\in M\}$1

when jumps are separated by more than $\mathrm{Att}(M):=\{y\in\mathbb{R}^d:T_f(y)\in M\}$2 (Schultz et al., 2017). The important distinction is between eventual return and return within a prescribed time horizon.

“Potentials and Limits to Basin Stability Estimation” addresses the numerical side of this problem for systems with difficult basin geometry (Schultz et al., 2016). Basin stability is estimated by Monte Carlo as the probability mass of a basin of attraction under a perturbation measure, but the paper shows that the reliability of such estimates depends strongly on the geometry of the basin. For fractal boundaries, the uncertainty fraction scales as

$\mathrm{Att}(M):=\{y\in\mathbb{R}^d:T_f(y)\in M\}$3

yet basin-stability estimation remains meaningful. For riddled or intermingled basins, finite precision can make rounding error comparable to or larger than the sampling error, so the method “reaches its limits for riddled basins with holes” (Schultz et al., 2016). This corrects a common overstatement: probabilistic basin size can remain numerically robust even when pointwise membership is unstable, but that robustness is not generic.

A plausible synthesis is that the dynamical-systems notion of capture basin becomes operational only after specifying three ingredients: the target set, the admissible dynamics between perturbations, and the time horizon on which recapture matters.

4. Lagrangian capture in heterogeneous flows

In transport theory, a capture basin is the set of initial positions whose trajectories are eventually intercepted by a sink. “Chaos and the Flow Capture Problem: Polluting is Easy, Cleaning is Hard” studies passive pollutant removal by localized perfect sinks in a prescribed two-dimensional flow, neglecting diffusion and using particle advection

$\mathrm{Att}(M):=\{y\in\mathbb{R}^d:T_f(y)\in M\}$4

as the governing dynamics (Smith et al., 2018). For a sink at location $\mathrm{Att}(M):=\{y\in\mathbb{R}^d:T_f(y)\in M\}$5, the paper’s central object is the captured region

$\mathrm{Att}(M):=\{y\in\mathbb{R}^d:T_f(y)\in M\}$6

where $\mathrm{Att}(M):=\{y\in\mathbb{R}^d:T_f(y)\in M\}$7 is the one-period flow map and $\mathrm{Att}(M):=\{y\in\mathbb{R}^d:T_f(y)\in M\}$8 is the set of streak trajectories emanating from the sink over one forcing period (Smith et al., 2018). This is the paper’s explicit asymptotic capture set.

The geometry of $\mathrm{Att}(M):=\{y\in\mathbb{R}^d:T_f(y)\in M\}$9 is determined by chaotic seas, coherent islands, invariant tori, and streak-surface structure rather than Euclidean distance from the sink. If $S_k$0 intersects a chaotic region, then $S_k$1 contains the entire chaotic region except a set of measure zero; if it intersects only certain invariant tori, then only those coherent regions are captured (Smith et al., 2018). With multiple sinks, the total capture basin is the union $S_k$2, and overlap between individual basins explains the reported diminishing returns in long-time efficiency. The paper distinguishes asymptotic capture-basin size from finite-time capture performance by introducing

$S_k$3

for long-time efficiency and

$S_k$4

for rate efficiency, where $S_k$5 is the time to reach a target residual $S_k$6 (Smith et al., 2018).

This literature corrects another common simplification: in heterogeneous or chaotic flow, successful capture is not primarily a matter of placing sinks near pollutant in a geometric sense, but of placing them on transport structures whose forward images generate large captured regions.

5. Optimization, variational landscapes, and quantum basin occupation

In nonconvex optimization, basin notions become algorithm-dependent. “Proximal basin hopping: global optimization with guarantees” defines local-minimizer components $S_k$7 and a deterministic local solver $S_k$8, then introduces the solver-induced attraction set

$S_k$9

as the explicit basin of attraction under that solver (Lauga et al., 18 May 2026). The key PBH potential is

$L_x\times L_y$0

and the proximal selection rule is

$L_x\times L_y$1

The method therefore does not redefine basins geometrically; it reweights them by objective value and distance-to-basin. A central theorem shows that for each $L_x\times L_y$2 there exists $L_x\times L_y$3 such that the unique global minimizer $L_x\times L_y$4 belongs to $L_x\times L_y$5 for every $L_x\times L_y$6, and the small-noise large-deviation relation

$L_x\times L_y$7

links Gaussian perturbation mass to the basin-selection potential (Lauga et al., 18 May 2026).

In variational quantum eigensolvers, the same language is used for initialization rather than global search. “Symmetry-Protected Basin Localization in Variational Quantum Eigensolvers” argues that VQE often fails before optimization begins because strong correlation splits the energy landscape into competing basins and the initial parameter vector selects a non-ground-state basin (Wang, 11 May 2026). The proposed geometry-conditioned equivariant preconditioner

$L_x\times L_y$8

is constrained by

$L_x\times L_y$9

and aims to place $h_i$0 directly inside the correlated ground-state basin (Wang, 11 May 2026). Within a strongly convex basin $h_i$1,

$h_i$2

the paper shows that gradient statistics become curvature-controlled rather than concentration-controlled. On six stretched molecules, the preconditioner reduced Hartree–Fock initialization errors by factors of $h_i$3–$h_i$4 (Wang, 11 May 2026).

“Schedule-dependent basin occupation in a programmable quantum annealer” gives perhaps the most literal modern use of basin occupation in quantum hardware (Lozano, 17 May 2026). On a mixed-frustration 12-qubit Ising instance, a basin is defined via steepest descent as the pre-image of each local minimum under deterministic single-spin-flip descent, computed exhaustively over all $h_i$5 classical configurations (Lozano, 17 May 2026). The paper shows that schedule shape modulates basin occupation on six of the thirteen multi-basin-in-readout instances, with dominant-configuration shifts of up to 38 percentage points, including changes of the dominant configuration. It also revises an earlier two-pause-enhancement claim, arguing that reverse-anneal schedules are instance-specific basin-occupation probes rather than universal enhancement knobs (Lozano, 17 May 2026). A common misconception in this area is that one schedule universally improves basin exploration; the paper explicitly rejects that interpretation.

6. Physical basins that trap air, waves, or particles

In atmospheric boundary-layer dynamics, basin capture can be realized by a receiving cold-air volume. “Observations and numerical simulations of a valley-exit wind in the Alpine Bolzano basin” analyzes nocturnal drainage flow from the Isarco Valley into the wider Bolzano basin and uses the setting as a natural analogue for capture-basin behavior (Gucci et al., 9 Mar 2026). The basin either behaves as a blocked receiving volume when it contains a strong cold-air pool or allows the tributary outflow to descend to the basin floor when stratification is weak. In Episode 1, a basin cold-air pool with observed depth about $h_i$6 AGL forced the valley-exit wind upward, leaving calm lower levels; in Episode 2, weak stratification allowed the outflow to descend and produce surface easterlies of about $h_i$7 at Bolzano (Gucci et al., 9 Mar 2026). The operational criterion proposed by that case study is the relative thermal and density structure of tributary outflow and resident basin air, not merely the existence of a valley-exit jet.

In seismology, a sedimentary basin captures energy through resonance and reverberation. “Basin Effects in Strong Ground Motion” describes Kathmandu basin response during the 2015 $h_i$8 Gorkha earthquake as entrapment and reverberation of waves in soft sedimentary deposits over a concave basement, with basin-edge diffraction and energy focusing generating large-amplitude surface waves (Ayoubi et al., 2018). An idealized 2D elastic plane-strain model with a simplified layered velocity structure captures the low-frequency reverberatory pulse across the basin reasonably well, whereas a 1D nonlinear shallow-soil model improves the higher-frequency band (Ayoubi et al., 2018). “Resonant Seismic Motion Within a Sedimentary Basin Revisited” makes the trapping mechanism explicit for SH waves in a soft semicircular basin by deriving a conservation relation

$h_i$9

where $C(m)$00 is volumic material-damping loss and $C(m)$01 is radiation-damping loss (Wirgin, 12 Jul 2025). Resonant frequencies are determined by zeros of the partial-wave denominator $C(m)$02, and resonant displacement capture corresponds to peaks in $C(m)$03 and troughs in $C(m)$04. The paper reports, for example, a first resonance at $C(m)$05, where $C(m)$06, compared with $C(m)$07 at a non-resonant $C(m)$08 (Wirgin, 12 Jul 2025).

In celestial mechanics, the “solar basin” is a bound phase-space reservoir rather than a spatial cavity. “Orbital Dynamics of the Solar Basin” studies weakly interacting particles on bound heliocentric orbits produced in the Sun or, equivalently by time reversal, gravitationally captured from the Galactic halo (Giovanetti et al., 2024). The practical density relation is

$C(m)$09

and at Earth the effective basin lifetime is

$C(m)$10

The paper also predicts annual and semi-annual modulation of the basin density at Earth at the fractional level of $C(m)$11 and $C(m)$12, respectively (Giovanetti et al., 2024). Here “capture” means long-lived occupation of gravitationally bound orbits shaped by secular perturbations, Kozai-like oscillations, diffusion in semi-major axis, and eventual Jupiter-driven ejection.

7. Comparative interpretation and scope

Across the cited literature, the word “capture” ranges from exact target-reaching sets to probabilistic return measures, transport interception regions, algorithm-induced attraction domains, and physically trapped reservoirs of air, wave energy, or particles. The shared motif is a receiving region whose geometry, dynamics, or control law determines whether trajectories remain excluded, are redirected, or are retained. The differences are equally important. Hydrologic basins are outlet-partition sets; dynamical-systems basins are asymptotic or finite-time return domains; flow-capture basins are forward Lagrangian interception sets; optimization basins depend on the local solver; quantum basin occupation is schedule- and instance-dependent; atmospheric and seismic basins are physical enclosures whose density or impedance structure produces blocking, reverberation, or trapping (Oliveira et al., 2018, Schultz et al., 2017, Smith et al., 2018, Lauga et al., 18 May 2026, Gucci et al., 9 Mar 2026).

This suggests that “capture basin” functions best as a family resemblance term rather than a universal formalism. Precision therefore requires stating, for each domain, the underlying state space, the dynamics or transport law, the target or receiving set, the admissible time horizon, and whether capture is defined deterministically, probabilistically, or operationally.