Orbit Dissimilarity Criterion (D-Criterion)

• Orbit Dissimilarity Criterion (D-criterion) is a scalar metric that reduces multivariate orbital comparisons to a decision variable using parameters like q, e, i, omega, and Omega.
• It combines a D-function with a decision threshold to operationally identify candidate meteor streams, showers, or genetic associations in orbit databases.
• The approach emphasizes statistical validation and acknowledges limitations from chaotic dynamics, observational uncertainties, and multiple-comparison effects.

Searching arXiv for the specified papers and closely related work on orbit dissimilarity criteria in meteor science. The Orbit Dissimilarity Criterion, commonly called the $D$-criterion, is an operational framework for quantifying resemblance between two heliocentric or meteoroid orbits and deciding whether that resemblance is sufficiently small to justify treating the pair as part of a candidate stream, shower, or genetic association. In meteor science, the term is best reserved for the combination of a $D$-function—a scalar orbit-dissimilarity function—and a decision threshold $D_{\rm th}$; two objects are considered similar only when the computed value falls below that threshold (Courtot et al., 25 Jul 2025). The modern literature presents the $D$-criterion simultaneously as an indispensable first-pass clustering device and as a source of recurrent overinterpretation: low $D$-values encode present-day orbital resemblance, but do not by themselves establish common origin, especially in near-Earth phase space where chaos, observational error, and multiple-comparison effects strongly limit inference (Shober et al., 2024).

1. Definition and conceptual scope

A $D$-criterion reduces a multivariate orbital-comparison problem to a single scalar decision variable. In the classical orbital-element formulation, the comparison is performed using combinations of perihelion distance $q$, eccentricity $e$, inclination $\iota$ or $i$, argument of perihelion $D$0, and longitude of ascending node $D$1 (Shober et al., 2024). In the formal usage emphasized by the comparative review, the distinction between a $D$2-function and a $D$3-criterion is essential: the function returns a scalar dissimilarity value $D$4, whereas the criterion adds the rule

$D$5

thereby converting similarity into an operational yes/no relation for clustering and stream search (Courtot et al., 25 Jul 2025).

The physical motivation is straightforward. If meteoroids were released from a common parent body, they should initially populate nearby regions of orbital-parameter space, so a genuine meteor stream or freshly disrupted near-Earth object (NEO) family is expected to contain many low-$D$6 pairings. Historically, this logic made $D$7-criteria central to two related tasks: meteor stream identification and parent-body association (Shober et al., 2024). The first seeks statistically coherent orbital populations against the sporadic background; the second attempts to link showers, fireballs, meteorite falls, or asteroids to a common progenitor.

The principal limitation identified in recent work is that orbital similarity is a proxy, not a proof. A low $D$8-value measures closeness in a chosen coordinate representation at the present epoch. It does not directly encode dynamical history, fragmentation chronology, or the probability that the resemblance could have arisen accidentally in a large catalog (Shober et al., 2024). This has led to a broad methodological reframing: $D$9-criteria are best understood as candidate-generation tools whose scientific validity depends on subsequent statistical and dynamical testing (Courtot et al., 25 Jul 2025).

2. Principal formulations and their variables

The modern review literature organizes $D_{\rm th}$0-functions into several families: classical orbital-element criteria, geocentric-observable criteria, vectorial criteria, and mathematically motivated proper-distance constructions (Courtot et al., 25 Jul 2025). Among the classical orbital-element family, three functions dominate both historical practice and recent critique: $D_{\rm th}$1, $D_{\rm th}$2 or $D_{\rm th}$3, and $D_{\rm th}$4.

The focused near-Earth critique explicitly analyzes $D_{\rm th}$5, $D_{\rm th}$6, and $D_{\rm th}$7 (Shober et al., 2024). $D_{\rm th}$8 is the Southworth–Hawkins criterion; $D_{\rm th}$9 is attributed to Drummond; and $D$0 is due to Jopek. These criteria all operate in orbital-element space but differ in normalization and weighting. The review emphasizes that $D$1 is historically foundational, $D$2 was introduced to address perceived scaling and dimensional issues in $D$3, and $D$4 was designed as a hybrid intended to combine the advantages of the earlier two (Courtot et al., 25 Jul 2025).

The same review also surveys other formulations. Valsecchi et al.’s $D$5 and $D$6 compare meteors using geocentric observables such as solar longitude and the unperturbed geocentric velocity vector rather than classical orbital elements, motivated by the fact that meteors are observed at Earth and that derived orbital elements can propagate measurement uncertainties unfavorably (Courtot et al., 25 Jul 2025). Rudawska et al.’s $D$7 is similarly geocentric in spirit, while Neslušan’s $D$8 uses the angular momentum vector alone. Jopek et al.’s $D$9 employs orbital energy together with eccentricity and angular-momentum vectors, with weights tied to standard deviations of known meteor-shower orbital elements. Jenniskens introduced $D$0 and $D$1, built from constrained combinations of orbital elements and the Tisserand parameter respectively (Courtot et al., 25 Jul 2025). Kholshevnikov et al.’s true-distance functions, including $D$2, are discussed because they address the failure of older $D$3-functions to satisfy the triangle inequality (Courtot et al., 25 Jul 2025).

A concise comparison of the criteria explicitly discussed in the two recent arXiv papers is given below.

Criterion	Variables emphasized	Role in recent discussion
$D$4	$D$5	Historically dominant; widely used, heavily criticized (Courtot et al., 25 Jul 2025)
$D$6 / $D$7	normalized $D$8, angular separations	Dimensionless reformulation; less favored in recent review (Courtot et al., 25 Jul 2025)
$D$9	$D$0	Hybrid classical criterion; comparatively robust within the classical family (Courtot et al., 25 Jul 2025)
$D$1, $D$2	geocentric $D$3	Conceptually attractive because they use observed/geocentric quantities (Courtot et al., 25 Jul 2025)
$D$4	$D$5	Vectorial criterion with data-based weights; comparatively well tested (Courtot et al., 25 Jul 2025)

The general conclusion is not that one criterion is universally best, but that different formulations encode different assumptions about which orbital or observational differences matter most. This suggests that the choice of criterion is inseparable from the intended dataset, clustering algorithm, and validation protocol (Courtot et al., 25 Jul 2025).

3. Mathematical structure and thresholding

The recent literature distinguishes between the algebraic form of a $D$6-function and the statistical meaning of any chosen threshold. For the classical case, the near-Earth critique does not reproduce explicit formulas for $D$7, $D$8, or $D$9, but states that they “consider different combinations of orbital parameters such as perihelion distance ($q$0), eccentricity ($q$1), inclination ($q$2), argument of perihelion ($q$3), and longitude of ascending node ($q$4)” (Shober et al., 2024). By contrast, the later review presents explicit formulae for many historical criteria, including $q$5, $q$6, and $q$7, while also noting transcription corruption in some expressions (Courtot et al., 25 Jul 2025).

A threshold turns the scalar value into an operational criterion. In practice, threshold choice is one of the most consequential and least standardized steps in meteor-shower identification. Many studies inherit values from prior literature, but the comparative review argues that such portability is often poorly justified because threshold effectiveness depends on database size, sporadic background, the $q$8-function itself, and the clustering algorithm (Courtot et al., 25 Jul 2025). The review therefore recommends threshold calibration against random or background-like datasets, highlighting the approach of Jopek et al. in which random samples are generated and the threshold is selected to meet a target reliability level (Courtot et al., 25 Jul 2025). It also discusses the break-point method of Neslušan, in which a change of slope in the cumulative number of meteors as a function of $q$9 is used to separate putative stream and sporadic components (Courtot et al., 25 Jul 2025).

The critique of near-Earth applications reinforces why thresholding cannot be treated as a mere convention. With a dataset of size $e$0, the number of unique orbit pairs is

$e$1

so the number of opportunities for accidental near-neighbors grows quadratically with catalog size (Shober et al., 2024). This produces a multiple-comparisons problem: the chance of finding at least one very small $e$2-value somewhere in a large search set can be substantial even if no physical linkage exists. The paper explicitly connects this to a “birthday paradox” logic, arguing that the relevant null hypothesis is not whether one specified pair is unlikely, but whether any pair in the searched populations would appear similarly close by chance (Shober et al., 2024).

A further mathematical issue concerns the status of many classical $e$3-functions as “distance-like” rather than true distances. The review notes that $e$4 and $e$5 do not satisfy the triangle inequality, which makes the ordinary geometric language of distance potentially misleading (Courtot et al., 25 Jul 2025). This does not prevent their practical use, but it undercuts any overly literal interpretation of their values as metrics in a strict mathematical sense.

4. Use in meteor-shower searches and clustering workflows

The standard workflow described in the review begins with pairwise computation of orbit dissimilarities across a meteor database, or between observed meteors and a reference orbit for membership assignment (Courtot et al., 25 Jul 2025). Thresholding then induces an adjacency structure from which larger groups are assembled by a clustering or linking algorithm. The paper stresses that this algorithmic layer is not secondary: it changes which groups appear, how elongated they are, and how susceptible the method is to false positives (Courtot et al., 25 Jul 2025).

Historically, single-neighbour or single-linkage approaches have been common. In such methods, any pair with $e$6 is linked, and connected components become candidate groups. The principal defect is the long-chain effect, in which a chain of locally similar meteors spans a broad region so that distant endpoints need not be truly similar (Courtot et al., 25 Jul 2025). Alternative workflows include Sekanina’s iterative mean-orbit method, in which membership is repeatedly refined around a weighted mean orbit, and multi-criterion or two-stage schemes in which one $e$7-function is used to seed associations and another to regroup them (Courtot et al., 25 Jul 2025).

Density-based clustering has received increasing emphasis. The review identifies DBSCAN as a major improvement over unconstrained chaining because it requires both a radius parameter $e$8 and a minimum-neighbor count $e$9, thereby forcing clusters to possess a denser core (Courtot et al., 25 Jul 2025). When $\iota$0, DBSCAN reduces to single-linkage; when $\iota$1, fragile chains and isolated seed points are suppressed. The paper also notes the appearance of HDBSCAN as a hierarchical extension that can separate dense cores from sparse filaments without fixing a single $\iota$2 value (Courtot et al., 25 Jul 2025).

The near-Earth critique applies precisely this broader workflow to significance testing rather than mere group extraction. It uses Kernel Density Estimation (KDE) to model the orbital background, generates synthetic samples via

$\iota$3

computes $\iota$4-value distributions in Monte Carlo realizations, and compares real cumulative counts of low-$\iota$5 pairs to the Monte Carlo mean with a $\iota$6-$\iota$7 envelope (Shober et al., 2024). For cluster extraction, it uses DBSCAN with

$\iota$8

with minimum points set to 3 for fictitious-stream decoherence analysis, while NEO clustering with $\iota$9 uses $i$0 and requires a core point to have at least two connections (Shober et al., 2024). This operationalizes a key methodological shift: $i$1-criteria are embedded in a full inferential workflow that includes null-model construction, Monte Carlo significance testing, and density-based clustering rather than simple thresholding alone.

The review’s recommended terminology follows from this workflow. A set of objects found by a $i$2-criterion is properly called a meteor group; only after statistical and dynamical validation should it be regarded as a meteor shower (Courtot et al., 25 Jul 2025). This linguistic distinction is more than editorial. It encodes the recognition that algorithmically recovered similarity groups are provisional structures whose physical meaning remains to be demonstrated.

5. Statistical interpretation, chance association, and limits of inference

The most important recent development is the explicit demonstration that low $i$3-values often arise at rates fully compatible with chance once realistic background distributions and large catalog sizes are taken into account. The near-Earth study examines meteorite falls, fireballs, FRIPON and EFN datasets, USG/CNEOS impactors, and a large NEO sample, asking how many low-$i$4 pairs would appear even in random populations with the same broad orbital distribution (Shober et al., 2024). Its null model is built with KDE in orbital-element space, followed by repeated Monte Carlo sampling of synthetic catalogs of the same size as the observed samples (Shober et al., 2024).

For meteorite–meteorite associations, the observed similarities under $i$5, $i$6, and $i$7 all remain within the $i$8-$i$9 region expected from random association; accordingly, the dataset shows no statistically significant stream structure (Shober et al., 2024). The paper reexamines the well-known Příbram–Neuschwanstein pair and concludes that even though the pair has an unusually small $D$00, the probability of obtaining such an association in a random sample remains non-negligible. In its KDE model for 350 probable meteorite-dropping fireballs, drawing 481 random orbits gives an estimated chance of about $D$01, which the authors regard as insufficient for a robust stream claim, especially once database growth and multiple testing are considered (Shober et al., 2024).

The same conclusion extends to meteorite–NEO and USG/CNEOS–NEO comparisons. For 50 meteorite falls versus 35,012 NEOs, and for 310 USG/CNEOS impact events versus the same NEO sample, the number of low-$D$02 pairs is consistent with the random-association prediction for all three criteria analyzed (Shober et al., 2024). The authors therefore conclude that there is no statistically significant evidence for meteorite–NEO or impactor–NEO associations on the basis of orbital similarity alone (Shober et al., 2024).

Within the impacting population, 616 possible $D$03 g meteorite-dropping fireballs observed by FRIPON/EFN/GFO also show low-$D$04 counts fully consistent with random association. In an $D$05-versus-$D$06 binned analysis restricted to pairs with $D$07, no bin exceeds the $D$08-$D$09 threshold (Shober et al., 2024). This is significant because it directly tests the common intuition that a low $D$10 should be more meaningful in rare orbital regions; in the analyzed impact datasets, the anticipated gain in discriminating power does not materialize.

The one clear positive result is within the NEO population itself. There, the cumulative $D$11-value distribution departs from the random KDE model at very small $D$12, with an evident kink around

$D$13

and significance emerging around $D$14 (Shober et al., 2024). Using DBSCAN with $D$15 and $D$16, the authors identify 12 statistically significant clusters, including the fragment complex of comet 73P/Schwassmann–Wachmann (Shober et al., 2024). This contrast between null results in impact datasets and positive results in the NEO population is attributed to the much larger NEO database and the lower quality of fireball-derived orbits, especially USG/CNEOS measurements (Shober et al., 2024).

Taken together, these findings recast the $D$17-criterion as a statistical observable whose meaning depends on the background model and search multiplicity. A small $D$18-value may indicate a real stream or fragment family, but absent significance testing it is equally compatible with the combinatorics of large, structured catalogs (Shober et al., 2024, Courtot et al., 25 Jul 2025).

6. Dynamical constraints: chaos, decoherence, and physical interpretation

The modern critique of $D$19-criteria is not only statistical but dynamical. In near-Earth space, orbital evolution is chaotic, so present-day similarity has limited memory. The near-Earth study maps Lyapunov characteristic lifetimes across $D$20 space and finds that many Earth-crossing or Earth-encountering trajectories have characteristic times of only 60–200 years over large regions, with Earth-encountering zones typically showing $D$21 yr for $D$22, $D$23 yr for $D$24, and $D$25 yr for $D$26 (Shober et al., 2024). The Lyapunov characteristic lifetime is described as

$D$27

so the local inverse Lyapunov exponent sets the timescale over which nearby trajectories separate exponentially (Shober et al., 2024).

Although collective streams can remain recognizable longer than individual trajectory similarity, they too lose coherence on finite timescales. In the fictitious-stream integrations, decoherence is defined to occur when the largest identified cluster contains at most $D$28 of the original stream members,

$D$29

equivalently meaning that at least $D$30 of the original coherent membership has been lost (Shober et al., 2024). The resulting decoherence lifetimes are generally of order $D$31–$D$32 yr (Shober et al., 2024). The review echoes this scale, stating that Earth-crossing streams have decoherence lifetimes on the order of tens to hundreds of thousands of years, and $D$33–$D$34 kyr for essentially all asteroidal debris visible as fireballs (Courtot et al., 25 Jul 2025).

These timescales impose a strong interpretive constraint. Many meteorites have cosmic-ray exposure ages far longer than the interval over which a near-Earth stream would remain coherently identifiable in present-day $D$35-space (Shober et al., 2024). The same paper notes Earth impact timescales for suitable meteoroid orbits of roughly $D$36–$D$37 yr in inverse-frequency terms, together with near-Earth tidal disruptions occurring roughly once every $D$38 yr (Shober et al., 2024). This combination implies that genuine genetic relationships may exist physically but no longer be visible as low-$D$39 associations at the present epoch. A plausible implication is that orbital similarity is biased toward detecting recent fragmentation, whereas older but real relationships are dynamically erased.

The critique also emphasizes that local dynamics vary strongly across orbital space: Earth-crossing orbits decohere faster, resonant structure matters, planetary encounters accelerate divergence, and inclination changes the geometry of instability (Shober et al., 2024). Because of this heterogeneity, a fixed metric with fixed weights on $D$40 cannot be dynamically optimal everywhere. The authors therefore describe existing $D$41-criteria as simplistic in the sense that they do not encode the true location-dependent evolution of streams (Shober et al., 2024). The review converges on the same conclusion by insisting that post-search orbital-dynamics checks are indispensable and that no perfect orbit-similarity method exists (Courtot et al., 25 Jul 2025).

7. Critiques, controversies, and current best practice

Several recurring criticisms attach to the classical $D$42-criteria. For $D$43, the review enumerates unit and scaling problems, criticism of chord-based angular treatment, sensitivity to $D$44, overparameterization relative to the constrained meteor-shower problem, dependence on database error structure, and failure of the triangle inequality (Courtot et al., 25 Jul 2025). For $D$45, it notes the attempt to normalize terms and make them dimensionless, but also cites criticisms concerning weighting imbalance at low eccentricity, reduced accuracy at small perihelion distance, lack of physical definition, and the persistence of general orbital-element weaknesses (Courtot et al., 25 Jul 2025). $D$46 is presented as holding up somewhat better, though it too inherits the broader limitations of orbital-element approaches (Courtot et al., 25 Jul 2025).

A major controversy concerns the widespread habit of treating low $D$47-values as if they directly established physical linkage. The near-Earth critique identifies this as the central misuse of the method, especially in claims of specific meteorite–NEO or fireball–asteroid parentage (Shober et al., 2024). It also criticizes inappropriate null-hypothesis framing, such as restricting the comparison set in a way that inflates significance or ignoring the full multiplicity of possible pairings. The discussion of the Chelyabinsk–1999 NC43 claim is exemplary: the significance was said to be overstated because the comparison set was artificially limited and because the many-comparisons problem was mishandled (Shober et al., 2024).

Observational uncertainty is a further source of controversy. The near-Earth study stresses that fireball-derived orbits have non-negligible uncertainties and that USG/CNEOS bolide orbits can have velocity errors exceeding $D$48 or radiant errors up to $D$49 (Shober et al., 2024). Under such conditions, the use of orbital-element $D$50-criteria can generate apparently precise similarity values from intrinsically degraded orbit determinations. This suggests that geocentric formulations such as Valsecchi’s criteria may be conceptually attractive not merely for elegance but because they may align more directly with what is actually measured (Courtot et al., 25 Jul 2025).

Current best practice, as it emerges from the two papers, is notably conservative. $D$51-criteria should be used as exploratory tools, candidate-selection metrics, and stream-finding statistics for populations rather than as standalone proof of a specific parent-body relationship (Shober et al., 2024). Thresholds should be calibrated against the actual sporadic background rather than copied uncritically from older studies (Courtot et al., 25 Jul 2025). Clustering should avoid unconstrained chaining when possible, with DBSCAN and related density-based methods offering a more robust alternative, albeit with their own parameter-selection issues (Courtot et al., 25 Jul 2025). Most importantly, any claimed group should undergo both statistical validation against random-group formation and dynamical validation for common-origin plausibility (Courtot et al., 25 Jul 2025).

The resulting methodological doctrine is narrow but clear. A $D$52-criterion can reveal a meteor group; it cannot by itself establish a meteor shower or a definitive genetic association (Courtot et al., 25 Jul 2025). In near-Earth applications, population-level excesses of very small $D$53-values remain meaningful, as shown by the recovery of a small number of statistically significant NEO clusters, including the 73P fragment complex (Shober et al., 2024). By contrast, specific pair associations among meteorites, fireballs, and impactors are generally too fragile—statistically and dynamically—to survive modern scrutiny on orbital similarity alone (Shober et al., 2024).