Equal Proportional Area Criterion (EPAC)

• EPAC is a unifying geometric principle that partitions domains into equal areas, volumes, or analogous measures while respecting constraints like contiguity and alignment.
• It drives applications across choropleth mapping, spherical gridding, shock localization in conservation laws, and power system stability through tailored algorithmic frameworks.
• EPAC implementations range from heuristic greedy methods to optimal dynamic programming, offering precise partitioning and insights into computational trade-offs.

The Equal Proportional Area Criterion (EPAC) is a unifying geometric principle specifying that a domain—be it a set of spatial regions, cells in a table, points on a manifold, or other partitionable structures—should be divided into parts with precisely or nearly equal areas (or volume, weight, or analogous measure), subject to application-specific structural constraints such as contiguity, alignments, or adjacency. EPAC arises in a diverse range of mathematical, visualization, physical, and computational contexts, guiding both division strategies (by explicit constraint) and algorithmic optimization (as a minimization target, sometimes in trade-off with alternative objectives).

1. Formal Statement and Representative Mathematical Formulations

In the general EPAC paradigm, one seeks a partition $\mathcal{D} = \bigsqcup_{i=1}^N R_i$ of a domain $\mathcal{D}$ such that

$$\forall i,\quad \text{Measure}(R_i) = \frac{\text{Measure}(\mathcal{D})}{N}$$

with "Measure" interpreted as geometric area, surface area, integral weight, or analogous quantity dictated by $\mathcal{D}$'s structure. The implementation and optimization of EPAC typically require further problem-specific formalizations.

Choropleth map classification: Given $N$ spatial regions with areas $A_i$ and pre-sorted data values $v_i$, EPAC seeks $K$ contiguous classes $C_c$ (parameterized by break indices $b_c$) to minimize

$\mathcal{D}$0

where $\mathcal{D}$1 is the total mapped area (Abboud et al., 2020).

Partitions of the sphere: For $\mathcal{D}$2, find $\mathcal{D}$3 disjoint regions $\mathcal{D}$4 of exactly $\mathcal{D}$5, where $\mathcal{D}$6 is the $\mathcal{D}$7-sphere's area (Leopardi, 2024, Malkin, 2016). The construction is recursive in dimension, typically via zonal collars and local subdivisions.

Scalar conservation laws: The equal area principle determines shock position $\mathcal{D}$8 so that for the parametric characteristic flow $\mathcal{D}$9, the signed area between overlapping branches in $$\forall i,\quad \text{Measure}(R_i) = \frac{\text{Measure}(\mathcal{D})}{N}$$0 space is zero: $$\forall i,\quad \text{Measure}(R_i) = \frac{\text{Measure}(\mathcal{D})}{N}$$1 with the EPAC location corresponding exactly to satisfaction of the Rankine-Hugoniot condition for conservation law weak solutions (McGregor et al., 2017).

Electric power system stability: EPAC quantifies transient stability in the swing equation by requiring that the area swept by acceleration (during the fault) be less than or equal to the area swept by deceleration (post-fault) as projected onto the $$\forall i,\quad \text{Measure}(R_i) = \frac{\text{Measure}(\mathcal{D})}{N}$$2 (electrical power vs. angle) curve: $$\forall i,\quad \text{Measure}(R_i) = \frac{\text{Measure}(\mathcal{D})}{N}$$3 (Zhao et al., 2024).

2. Algorithmic Frameworks and Computational Methods

EPAC instantiations give rise to algorithmic schemes of varying computational and structural complexity:

Choropleth Classification: Three classes of algorithms are highlighted (Abboud et al., 2020):

Algorithm Type	Complexity	Guarantee
Greedy 1	$$\forall i,\quad \text{Measure}(R_i) = \frac{\text{Measure}(\mathcal{D})}{N}$$4	Simple heuristic, last-chunk bias
Greedy 2	$$\forall i,\quad \text{Measure}(R_i) = \frac{\text{Measure}(\mathcal{D})}{N}$$5	Improved balance, still heuristic
Optimal DP	$$\forall i,\quad \text{Measure}(R_i) = \frac{\text{Measure}(\mathcal{D})}{N}$$6	Minimizes mean area deviation, pruned by monotonicity

The optimal DP applies a recurrence: $$\forall i,\quad \text{Measure}(R_i) = \frac{\text{Measure}(\mathcal{D})}{N}$$7 with tailored candidate pruning based on area monotonicity. For area-count tradeoffs, a weighted error with parameter $$\forall i,\quad \text{Measure}(R_i) = \frac{\text{Measure}(\mathcal{D})}{N}$$8 leads to an $$\forall i,\quad \text{Measure}(R_i) = \frac{\text{Measure}(\mathcal{D})}{N}$$9 DP.

Spherical Partitioning: Both the EQ(d, N) recursive zonal-equal-area partitioning (Leopardi, 2024) and the band-collar method (Malkin, 2016) solve for collar angles $\mathcal{D}$0 to equidistribute area, allocating integer subslices per collar and recursively filling lower-dimensional boundaries. Efficient implementations run in $\mathcal{D}$1 for fixed dimension.

StreamTable Visualization: Given a nonnegative $\mathcal{D}$2 table $\mathcal{D}$3 with weights $\mathcal{D}$4, EPAC constrains rectangle areas ($\mathcal{D}$5) within bands. For fixed permutation and row heights, the minimal excess problem (area of the bounding box minus $\mathcal{D}$6) is solvable in $\mathcal{D}$7 time (Espenant et al., 2021). With variable heights or permutations, feasibility and minimality become NP-hard.

Shock Localization in Conservation Laws: The parametric interpolation framework uses area-conservative polynomial (Hermite or higher) interpolants, with local root-finding (polynomial root in $\mathcal{D}$8) to solve for the shock's EPAC-enforced position on each overturned segment. All integrals in the root equations may be computed analytically or with high-precision quadrature (McGregor et al., 2017).

3. Domain-Specific Applications

EPAC has been applied in:

• Geovisualization: Chiefly in choropleth mapping, where area-uniform color classes correct the visual bias present in quantile or Jenks breaks, especially for imbalanced region size distributions. DP-EPAC correlates with more perceptually even map colorings (Abboud et al., 2020).
• Spherical Gridding: Zonal-equal-area and parallel-band algorithms decompose the $\mathcal{D}$9-sphere for applications in climate sampling, mesh generation, robotics orientation filtering, and numerical integration on manifolds (Leopardi, 2024, Malkin, 2016).
• Visualization: StreamTable constructions combine strict area proportionality with visual flow requirements in table visualizations, balancing adjacency and excess (Espenant et al., 2021).
• Numerical PDEs: For scalar hyperbolic conservation laws, the equal area principle (a form of EPAC) yields mathematically exact tracking of shock positions, equivalent to the Rankine-Hugoniot condition and producing machine-precision solutions in special polynomial-invariant cases (McGregor et al., 2017).
• Engineering Dynamics: In power system transient stability, EPAC directly interprets critical clearing conditions for faults, providing a geometric energy-based threshold for stability under current-limited inverter control (Zhao et al., 2024).

4. Extensions, Trade-offs, and Analytical Properties

EPAC's strict area-equality constraint is sometimes placed in trade-off with competing criteria (e.g., region-count equality in choropleth mapping). The weighted error function

$N$0

enables continuous interpolation between strict area homogeneity ($N$1) and count-uniform quantiles ($N$2), solvable via $N$3 dynamic programming (Abboud et al., 2020).

Analytical guarantees are available in certain geometric contexts. For EQ(d,N) sphere partitions, the maximum cell diameter decays as $N$4, and discrepancy (for quadrature or packing) achieves the same order (Leopardi, 2024). In the parallel-band sphere partition, band boundaries and cell counts are recursively adapted via exact integral constraints, with area errors held to floating-point rounding (Malkin, 2016).

5. Empirical Comparison and User Studies

Quantitative and qualitative assessments underscore EPAC's advantages and possible limitations:

• For world population maps, mean area deviation drops by a factor $N$5 when using optimal EPAC (DP) versus quantile methods, with execution cost negligible even for $N$6 (Abboud et al., 2020).
• Visually, EPAC-based choropleths avoid uneven hue dominance, and user studies show EPAC-preferred maps in aesthetics (15/30 participants ranked EPAC maps as most appealing).
• In transient stability simulations for virtual synchronous generators, EPAC-motivated adaptive current-limiting constrains the acceleration area for improved stability margin, outperforming d- or q-axis-priority laws in direct correspondence with EPAC predictions (Zhao et al., 2024).
• EPAC-based partitions of the sphere provide machine-precision area uniformity and significantly reduced aspect-ratio distortion compared to standard isolatitudinal grids, with band-center latitude errors below $N$7 for $N$8 (Malkin, 2016).

6. Theoretical Connections and Equivalence Conditions

In the analysis of degenerate or singular solutions, such as shock formation in scalar conservation laws, EPAC (equal area principle) is mathematically equivalent to the classical Rankine-Hugoniot jump condition—these are interchangeable as selection criteria for weak solution uniqueness in 1D conservation settings (McGregor et al., 2017). The literature establishes, under suitable convexity and smoothness, the existence, uniqueness, and time-differentiability of the EPAC-located shock.

For partitioning high-dimensional spheres, monotonicity in the sine-power integral ensures, for any integer allocation vector $N$9 summing to $A_i$0, a unique increasing sequence of colatitude boundaries exists that achieves exact area uniformity (Leopardi, 2024). Limitations arise only when the induced structure interferes with other global properties (e.g., for certain polynomial approximation tasks).

7. Limitations, Contingencies, and Open Problems

Algorithmic optimality under EPAC is often constrained by domain structure. In StreamTable layouts, NP-hardness prevails if row permutation or row heights are variable (Espenant et al., 2021). Spherical partitions may degrade analytic properties in specific harmonic reconstruction tasks if the zonal collar geometry aligns poorly with the target basis. In data classification, large region-size disparities may interact with projection artifacts, though EPAC's performance is largely invariant except in pathological cases (Abboud et al., 2020). The general problem of designing efficient heuristics for NP-hard constraints under EPAC remains open in several visualization and partitioning applications.

---

References:

(Abboud et al., 2020, Espenant et al., 2021, McGregor et al., 2017, Zhao et al., 2024, Leopardi, 2024, Malkin, 2016)