Pith. sign in

REVIEW 3 major objections 4 minor 15 references

Minimum-distortion continuous cartograms by numerically optimized meshes

T0 review · 3 major / 4 minor · reviewed 2026-08-12 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read This paper claims that continuous cartograms can be built by numerically optimizing a triangle mesh to minimize a hand-tuned distortion cost, reaching median relative area errors below $10^{-10}$ while preserving region shapes far better…

desk verdict A well-engineered mesh-optimization cartogram method with genuinely new on-sphere and hybrid variants, but the headline shape-preservation claim is not quantitatively tested and the omitted H definition blocks full reproduction. read the letter →

arxiv 2411.17129 v1 pith:5PTK2FPF submitted 2024-11-26 cs.CG cs.NAmath.NA

Robert C. Sargent This is my paper · ORCID
classification cs.CGcs.NAmath.NA MSC 68U0565K10
keywords continuouscartogramsmeshoptimizationcartographicerrorshapedistortionmapprojectionsnumericalsphereandhybridLiquidEarthprojection
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper tries to establish that continuous (contiguous) cartograms can be produced by numerical optimization of a triangular mesh, rather than by diffusion or rubber-sheet deformation, and that this yields far better preservation of region shapes at essentially any desired level of area accuracy. It develops three variants--plane, sphere, and hybrid--where the hybrid optimizes on the sphere while anticipating the distortion of a chosen final projection. On world population maps, all three variants drive median relative cartographic error below $10^{-10}$ after ten optimization stages. The same hybrid technique is then applied outside cartograms to create a low-distortion map projection released as Liquid Earth.

What carries the argument

The load-bearing object is a triangle mesh with precomputed area portions $\psi_{R,T}$: because each initial triangle maps affinely to its transformed counterpart, the transformed area of every region is a linear combination of triangle determinants, so cartographic error and its gradient are easy to compute. Distortion at each triangle is measured by the $2\times2$ matrix $K$ of the affine map; shape distortion uses $\|K\|_F^2/\det K - 2$, and scale distortion uses $\det K/s + s/\det K - 2$, both infinite when the triangle flips. For sphere and hybrid variants, the transformed triangle is projected onto the tangent plane at its midpoint before measuring distortion, and the hybrid variant premultiplies $K$ by the Jacobian of the target projection, blurred near interruptions, with a basis-rotation correction derived via geodesic curvature. This machinery keeps the whole cost function continuously differentiable, which is what lets the optimizer navigate to a local minimum.

What would settle it

Construct a map with two small regions lying entirely inside a single mesh triangle, or a region smaller than the mesh resolution, and run the ten-stage optimizer: if relative area error fails to approach $10^{-10}$ because the target areas are unattainable without splitting the triangle, the subdivision assumption is shown to be load-bearing.

Watch

Extended reading notes

Core claim

The central claim is that an accurate, low-distortion cartogram can be found by fixing an initial mesh on the globe, radially projecting borders onto it, and then moving mesh vertices to minimize a differentiable cost function $C(V) = W_{\mathrm{error}} E(V) + W_{\mathrm{dist}} D(V)$. The error term $E$ sums squared area mismatches weighted per region; the distortion term $D$ sums per-triangle scale and shape distortion weighted by initial area and by land/water and density factors. Shape distortion is defined as $\sigma_1/\sigma_2 + \sigma_2/\sigma_1 - 2$ in terms of the singular values of the local affine map, and scale distortion analogously against a desired scale. Because the cost function is differentiable and topology violations are assigned infinite cost, L-BFGS with backtracking line search drives the error to machine-precision levels, with the sphere and hybrid variants additionally guaranteeing no flipped triangles. The paper reports median relative errors below $10^{-10}$ for all three variants after ten stages and claims a significant shape-preservation improvement over the diffusion and rubber-sheet baselines.

Load-bearing premise

The paper assumes that the heuristic initial mesh subdivision--each region covered by at least four triangles and each triangle at most $1/2048$ of the sphere's area--is fine enough for an accurate cartogram to exist; the paper itself notes that no sufficient condition for area accuracy is known.

Editorial extensions

If this is right

  • Population cartograms can be generated with median relative area error below $10^{-10}$ while keeping region shapes close to the original map.
  • Sphere and hybrid variants rule out flipped triangles by construction, so topological problems are avoided before any planar projection.
  • Because the hybrid cost anticipates the projection, the final planar map keeps low distortion and a regular boundary at once.
  • The same optimization can produce new map projections, as in the Liquid Earth projection, where land is weighted far above water.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Beyond the paper, the differentiable-cost recipe should transfer to adaptively refined meshes, offering a route toward the sufficient-condition question the paper leaves open.
  • A testable extension is to run the hybrid optimizer with target projections other than Mollweide and Equal Earth; the framework predicts each projection's interruption distortion will be absorbed with different trade-offs between shape and boundary regularity.
  • The reported runtime is dominated by thousands of gradient steps, so warm-starting later stages from coarser cartograms is a natural speedup the paper does not test.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 4 minor

Summary. The manuscript develops a mesh-based numerical optimization method for continuous cartograms. It defines a cost function combining per-region area error with per-triangle scale and shape distortion, and derives gradients for three variants: a plane cartogram, a sphere cartogram optimized on S^2 before equal-area projection, and a hybrid cartogram that optimizes on the sphere while anticipating distortion from a target projection. The same hybrid machinery is used to construct the Liquid Earth projection. The paper reports very low relative area errors and multihour runtimes in Section 6, and claims in the abstract and Section 7 that the method significantly improves shape preservation over existing automated methods.

Significance. If substantiated, the main contribution is a practical framework for distortion-aware cartogram generation that avoids projection-induced distortion by optimizing on the sphere, plus a reusable technique for projection design. The gradient derivations, especially the tangent-plane handling in Section 4 and the rotation correction in Section 5, are careful and presented in unusual detail, and the availability of source code is a strength. However, the paper's headline comparative claim about shape preservation is not tested by the reported experiments, and one central definition (the blurred Jacobian H) is omitted, so the current version does not yet support the abstract's strongest statement.

major comments (3)
  1. [Abstract; §6; §7] The central comparative claim—'significant improvement in preservation of region shapes compared to existing automated methods' (Abstract, restated in §7)—is not supported by quantitative evidence in the manuscript. Section 6 reports only relative cartographic error and runtime, both of which concern area accuracy, not shape preservation. The single illustrative comparison in Figure 2 does not constitute a test. To make the claim load-bearing, please add an independent shape-preservation metric (e.g., boundary displacement, angular deformation, or a commonly used cartogram quality measure from [4]) and report it for the mesh method and for the Gastner–Newman and rubber-sheet baselines on the same data.
  2. [§5] Section 5 introduces a function H : S^2 -> R^{2x2} that blurs the projection Jacobian near the antimeridian and states 'We omit the specific definition of H.' Because H enters the definition of eK and its derivative, the hybrid cartogram and the Liquid Earth projection cannot be reproduced from the paper alone. Please include an explicit definition of H (or complete pseudocode) and state where it is evaluated; the footnote's regularity condition is not enough for implementation.
  3. [§6, error tables] The relative cartographic errors in the Section 6 tables for the sphere and hybrid cartograms are computed from the transformed mesh without considering the projection to the sphere, and the paper estimates that this projection introduces relative error up to about 0.003. Reporting median errors below 10^-10 without incorporating or separately bounding this projection error overstates the actual area accuracy of the final cartograms. Please report total error that includes the projection term, or add a clear bound such as 'median error <= max(10^-10, 0.003)' with a derivation.
minor comments (4)
  1. [§3] The 'blurred intended scale' for water-only triangles is mentioned but not defined; the source code is the only place a reader can find it. Please state the blur kernel or at least cite the code file or function that implements it.
  2. [References] Reference [14] appears in the bibliography but is not cited in the text; please either cite it where relevant or remove it.
  3. [§6] The paper states that the last few stages are unnecessary for the sphere and hybrid cartograms because of the projection-induced error, but the tables still show ten stages; clarifying which stages are actually needed for the reported accuracy would help readers reproduce the method efficiently.
  4. [§4] There is a typo in Section 4: 'the second basis vector points points towards the North Pole' has a duplicated word.

Circularity Check

1 steps flagged · score 2.0 of 10

No load-bearing circularity: the only self-referential element is that 'minimum distortion' is the optimizer's own objective; the comparative shape-preservation claim is under-evidenced but independent.

  1. self definitional [Section 1 (Introduction); Section 2 (Algorithm overview), C(V) = W_error E(V) + W_dist D(V)]
    "We define a cost function whose inputs are the vertices of the transformed mesh. The cost function quantifies the total cartographic error (differences between the areas of regions and their desired values) and the total distortion. A local minimum of this function is an accurate cartogram with minimal distortion."

    'Minimal distortion' is defined as the minimum of the paper's hand-defined D(V), and the algorithm returns a local minimum of C(V) = W_error E(V) + W_dist D(V). The accuracy tables report the same relative errors that enter E(V), so the numerical results demonstrate optimizer convergence rather than an independent test of distortion or shape quality. This is a mild tautology, not a hidden fit: no parameter is fitted to an external dataset, and the transformation is derived from an explicit geometric objective.

full rationale

The derivation is otherwise self-contained. The mesh transformation is obtained by L-BFGS minimization of a fully specified cost function; the reported area errors are direct evaluations of the optimized objective, and the gradient computation is explicit. The claim of significant shape preservation over existing methods is not circular but is under-evidenced: Section 6 contains only area error and runtime tables, with no quantitative shape metric or comparison to Gastner-Newman or rubber-sheet baselines, so this is a validation gap rather than a circular derivation. The paper explicitly admits in Section 7 that no sufficient condition for area accuracy is known and that mesh subdivision is heuristic, which mitigates hidden-assumption concerns. The only self-citation ([12], Liquid Earth) is a released artifact of the same method and is not load-bearing for the cartogram derivation.

Assumptions & free parameters 9 free parameters · 6 assumptions · 0 invented entities

The algorithm depends on a set of hand-chosen weights and an unspecified smoothing function H, but these are design choices rather than fitted parameters. The main structural assumption is that the heuristic mesh resolution suffices for accurate cartograms, which the paper flags as open.

free parameters (9)
  • W_dist schedule = stage 1: 0.1, multiplied by 0.1 per stage
    Controls tradeoff between error and distortion in the cost function C(V) = W_error E(V) + W_dist D(V); chosen by hand (Section 2).
  • alpha_shape = 0.5
    Weight on shape distortion in D(V); chosen by hand (Section 3).
  • alpha_scale = 0.2
    Weight on scale distortion; chosen as compromise between runtime and aesthetics (Section 3).
  • alpha_land = 0.1 for water-only triangles, 1 otherwise
    Reduces distortion weight on water (Section 3).
  • alpha_density coefficients = 0.2 + 0.8 s_T
    Shape of density weighting; chosen by hand (Section 3).
  • w_boundary = 1e-6
    Small weight on pole self-intersection penalty (Section 3).
  • Armijo constant c = 0.1
    Line search parameter (Section 4).
  • Stopping threshold gamma = 0.01, times 0.1 per stage
    Convergence criterion (Section 6).
  • Liquid Earth weights = land weight 100x water; scale weight ramps 0.1x to 100x
    Design choices for the projection (Section 5).
assumptions (6)
  • standard math L-BFGS with Armijo backtracking converges to a local minimum of the smooth cost function.
    Used throughout for optimization (Sections 2 and 4).
  • domain assumption Affine maps between triangles scale region areas proportionally, so per-triangle area fractions psi_R,T are vertex-independent.
    Basis of the area computation in Section 3.
  • domain assumption Radially projecting region borders onto the initial mesh introduces no cartographic error because both area calculation and final map use the projected borders.
    Section 2.
  • domain assumption The target projection omega is equal-area, so its Jacobian has determinant 1.
    Section 5.
  • ad hoc to paper A smoothly blurred version H of the projection Jacobian exists near the antimeridian, with the regularity described in the footnote.
    Section 5; the paper explicitly omits the specific definition of H.
  • ad hoc to paper The chosen mesh subdivision (at least four triangles per region, max triangle area 1/2048 of sphere) is sufficient for an accurate cartogram to exist.
    Section 2; the paper states it is unclear how much subdivision guarantees accuracy (Section 7).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Minimum-distortion continuous cartograms by numerically optimized meshes." pith.science (2026). https://pith.science/paper/5PTK2FPF

@misc{pith2026241117129,
  author       = {Pith},
  title        = {Pith review of: Minimum-distortion continuous cartograms by numerically optimized meshes},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5PTK2FPF}},
  note         = {Machine review of arXiv:2411.17129}
}
read the original abstract

We present an algorithm for creating contiguous cartograms using meshes. We use numerical optimization to minimize cartographic error and distortion by transforming the mesh vertices. The vertices can either be optimized in the plane or optimized on the unit sphere and subsequently projected to the plane. We also present a hybrid "best of both worlds" method, where the vertices are optimized on the sphere while anticipating the distortion caused by the final projection to the plane. We show a significant improvement in the preservation of region shapes compared to existing automated methods. Outside the realm of cartograms, we apply this hybrid technique to optimized map projections, creating the Liquid Earth projection.

Figures

Figures reproduced from arXiv: 2411.17129 by the authors.

Figure 1
Figure 1. Part of the initial mesh and transformed mesh. Notice how the movement [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Comparison of shape preservation between the rubber-sheet method (left) [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. The initial mesh before and after subdivision. [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: A plane cartogram. The second version is the sphere cartogram (fig. 5). For this cartogram, the cost function takes in the positions of vertices on the unit sphere. Minimizing 4 [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]
Figure 5
Figure 5. Figure 5: A sphere cartogram shown in orthographic view and projected using the [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]
Figure 6
Figure 6. Figure 6: A hybrid cartogram whose target projection is Mollweide. [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]
Figure 7
Figure 7. Figure 7: The Liquid Earth projection, an approximately equal-area map projection [PITH_FULL_IMAGE:figures/full_fig_p006_7.png]
Figure 8
Figure 8. Figure 8: A large self-intersection caused by Antarctica contracting. [PITH_FULL_IMAGE:figures/full_fig_p015_8.png]
Figure 9
Figure 9. Figure 9: The sets of boundary vertices Q0, Q1, Q2, Q3 and the vertices p north and p south. Mesh resolution reduced for clarity. This addition does not prevent all self-intersections in the final map, but it avoids a common and severe case. The sphere and hybrid cartograms do g…
Figure 10
Figure 10. Figure 10: How normalizing vertices changes the cost reduction. The shading repre [PITH_FULL_IMAGE:figures/full_fig_p019_10.png]
Figure 11
Figure 11. Figure 11: The effect on the local basis caused by moving east in the Northern Hemi [PITH_FULL_IMAGE:figures/full_fig_p021_11.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

15 extracted references · 6 canonical work pages

  1. [4]

    G., Veeramoni, S

    Jawaherul Alam, M., Kobourov, S. G., Veeramoni, S. (2015). Quantita- tive Measures for Cartogram Generation Techniques. Computer Graphics Forum. https://doi.org/10.1111/cgf.12647

    doi:10.1111/cgf.12647 2015

  2. [1]

    Armijo, L. (1966). Minimization of functions having Lipschitz continuous first partial derivatives. Pacific Journal of Mathematics , 16(1), 1–3. https: //doi.org/10.2140/pjm.1966.16.1

    doi:10.2140/pjm.1966.16.1 1966

  3. [2]

    A., Chrisman, N

    Dougenik, J. A., Chrisman, N. R., Niemeyer, D. R. (1985). An algorithm to construct continuous area cartograms. The Professional Geographer . https://doi.org/10.1111/j.0033-0124.1985.00075.x

    work page arXiv 1985

  4. [3]

    Gastner, M., Newman, M. E. J. (2004). Diffusion-based method for pro- ducing density-equalizing maps. Proceedings of the National Academy of Sciences. https://doi.org/10.1073/pnas.0400280101

    doi:10.1073/pnas.0400280101 2004

  5. [5]

    Kronenfeld, B. J. (2017). Manual construction of continuous cartograms through mesh transformation. Cartography and Geographic Informa- tion Science , 45(1), 76–94. https://doi.org/10.1080/15230406.2016. 1270775

    arXiv 2017

  6. [6]

    Kunimune, J. H. (2020). Minimum-error world map projections defined by polydimensional meshes. International Journal of Cartography . https: //doi.org/10.1080/23729333.2020.1824174

    work page arXiv 2020

  7. [7]

    Kunimune, J. H. (2023). Introducing the Elastic pro- jections. Web. https://kunimune.blog/2023/12/29/ introducing-the-elastic-projections/ 26

    work page 2023

  8. [8]

    Li, Z., Aryana, S. (2017). Diffusion-based cartogram on spheres. Cartog- raphy and Geographic Information Science , 45:5, 464–475. https://doi. org/10.1080/15230406.2017.1408033

    arXiv 2017

Show all 15 references
  1. [9]

    Loncaric, M. (2024). Map Projections 2: Solving Numerically. Web. https: //graphallthethings.com/posts/map-projections-2

    2024

  2. [10]

    Nocedal, J. (1980). Updating quasi-Newton matrices with limited stor- age. Mathematics of Computation , 35(151), 773–782. https://doi.org/ 10.1090/S0025-5718-1980-0572855-7

    1980 doi

  3. [11]

    Nusrat, S., Kobourov, S. (2016). The state of the art in cartograms. Com- puter Graphics Forum. https://doi.org/10.1111/cgf.12932

    2016 doi

  4. [12]

    Sargent, R. C. (2024). Introducing the Liquid Earth projection. Web. https://rsargentmath.github.io/posts/liquid_earth/

    2024

  5. [13]

    Sun, S. (2013). A fast, free-form rubber-sheet algorithm for contiguous area cartograms. International Journal of Geographic Information Science , 27 (3): 567–93. https://doi.org/10.1080/13658816.2012.709247

    2013

  6. [14]

    Sun, S. (2013). An optimized rubber-sheet algorithm for continuous area cartograms. The Professional Geographer . https://doi.org/10.1080/ 00330124.2011.639613

    2013

  7. [15]

    Sun, S. (2015). Carto3F program: A fast, free-form algorithm implemen- tation for area cartograms. Web. http://www.sunsp.net/download.html 27

    2015

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.