{"id":"4adce59f-144d-41fb-a2de-93601893da64","arxiv_id":"2503.13455","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":3.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"A survey of spherical stochastic geometry for non-terrestrial networks that introduces a new orbital point process (DSBPP) and quantitative case studies on planar approximation and satellite positioning.","lead":"This paper surveys how spherical stochastic geometry is used to model and analyze non-terrestrial networks such as LEO satellite constellations. It organizes the field into model types and adds new technical pieces, including a dual stochastic binomial point process and case studies on planar approximation error.","discovery_kind":"review","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Planar-approximation error thresholds in §III-C rest on one BPP realization and a best-fit plane altitude; no error bars or expectation are provided.","rationale":"The reader's weakest_assumption pinpoints the same fragility: the planar approximation case study uses a single BPP realization and an optimized plane altitude. My independent reading agrees. The survey's taxonomy and the derivations in Section III-B (contact angle, availability, K-availability) are correct, and the DSBPP construction is plausible: isotropy follows from the isotropic orbit-normal distribution and uniform satellite angles on each orbit, although the 'not difficult to prove' step is only sketched. The K-localizability case study in Section V.E.4 also lacks derivations and error bars, but it is presented as an illustrative example rather than the basis for the paper's headline quantitative guidance; the planar approximation study is the explicit 'necessity of spherical modeling' contribution. Therefore the load-bearing concern is the statistical and modeling-choice robustness of Fig. 7. The proposed test resolves it by replacing the single sample with the true expectation and confidence bands, and by comparing against the natural non-optimized altitude. Since the reader's verdict is already CONDITIONAL, and my concern supports that verdict, no change is needed.","tokens_in":17,"tokens_out":5774,"duration_ms":123032,"concrete_test":"Compute the exact expectation E[g(U)] and its variance analytically (or by 10^5 Monte Carlo draws per parameter point) for the mapping in Eqs. (15)-(16), for a grid of θc and h_planar values used in Fig. 7. Then (i) plot the mean relative error with 95% confidence bands for N=100, and (ii) recompute the thresholds 2θc at which mean E = 0.5% for LEO/MEO/HAP, both with h_planar optimized and with the natural value h_planar = R_NTP cos θc. If the mean thresholds differ from 3.6° / 9.9° by more than ~20%, or the confidence band at N=100 is comparable to the separation between the curves, the paper's quantitative guidance does not survive.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The quantitative conclusion of the necessity case study (Section III-C, Fig. 7) is computed from a single generated spherical BPP with N_NTP=100 and from h_planar chosen to minimize E over the interval (R_NTP cos θc, R_NTP). Since E in Eq. (17) is a sample mean of the function g(u) = |d_spherical(u) - d_planar(u)|/d_spherical(u) over i.i.d. uniform u(n) ~ U(0,1), it is a random quantity; with N=100 its fluctuation across realizations is governed by Var(g)/100 and is not negligible, especially for small θc where a few near-rim points dominate the average. The paper reports single-curve thresholds such as E=0.5% at 2θc=3.6° for LEO without confidence intervals or an expectation computation, so the guidance that planar approximation is acceptable for directional beams may not be reproducible. Moreover, excluding the rim-matching altitude h_planar = R_NTP cos θc and instead optimizing h_planar gives the planar model a best-case error; a modeler using a natural fixed altitude would incur larger error, potentially changing the stated central-angle thresholds. Because this case study is one of the paper's original technical contributions, the fragility is load-bearing for its claim of providing quantitative guidance on spherical vs planar modeling.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This manuscript surveys the use of spherical stochastic geometry for modeling and analyzing non-terrestrial networks. It proposes a taxonomy of spherical point process models (non-orbital, stochastic-orbital, fixed-orbital), provides generation algorithms for BPP and for a new dual stochastic binomial point process (DSBPP), derives standard topological results such as the contact angle distribution and availability probability, reviews channel modeling for various NTN links, and discusses advanced topics including routing, security, satellite clusters, energy harvesting, and satellite-enabled positioning. The paper also contains two original quantitative case studies: an analysis of the error introduced by planar approximations of spherical models, and a K-localizability study for DSBPP-modeled MEO satellite constellations.","tokens_in":26,"tokens_out":8330,"duration_ms":136050,"significance":"If the survey's quantitative claims are made robust, this would be a valuable reference for the spherical SG community, providing a structured classification, reproducible algorithms, and guidance on the applicability of planar approximations. The standard derivations in Section III are correct, and the DSBPP is a useful addition to the catalog of orbital point processes, with potential for tractable analysis of isotropic constellations. The positioning case study addresses a genuinely underexplored application. However, the two original case studies currently rest on fragile numerical foundations (single-realization error estimates and an unverified model benchmark), so the paper's contribution is not fully realized in its present form.","major_comments":[{"comment":"The relative error E in Eq. (17) is computed from a single realization of the coupled spherical and planar BPPs with NNTP=100, and the plane altitude h_planar is chosen to minimize E over the interval (RNTP cos θc, RNTP). Since E is a sample mean of the function g(u)=|d_spherical(u)-d_planar(u)|/d_spherical(u) over i.i.d. uniform u(n), E is itself a random variable whose fluctuation across realizations is governed by Var(g)/100; the reported thresholds such as E=0.5% at 2θc=3.6° for LEO are therefore not reproducible without error bars or an expectation computation. Moreover, optimizing h_planar gives the planar model a best-case error that a modeler would not know a priori, and a natural fixed altitude such as h_planar=RNTP cos θc may yield substantially larger errors. The authors should present the expectation of E (analytically or via many Monte Carlo trials) and report its variability, for physically motivated choices of h_planar, before drawing conclusions about when planar modeling is acceptable.","section":"§III-C, Eq. (17), Fig. 7"},{"comment":"The K-localizability case study for DSBPP-modeled MEO networks is presented as an original technical contribution, but no analytical derivation or simulation methodology is given; the text only lists parameters and plots results. In addition, the DSBPP model uses orbit inclinations drawn from the isotropic PDF fθ⊥(θ)=sinθ/2, whereas the actual GPS constellation uses a fixed 55° inclination, so using only the number of satellites per orbit from GPS does not constitute a validation of the DSBPP against a real constellation. The authors should either derive the K-localizability probability (accounting for the dependence of satellites on the same orbit) or clearly state that Fig. 10 is a Monte Carlo estimate with error bars, and they should benchmark the DSBPP against the actual GPS orbital geometry if the comparison to GPS is intended.","section":"§V-E.4, Fig. 10, Eq. (39)"},{"comment":"The paper states that the DSBPP 'is also homogeneous' and then uses independent-point formulas for the DSBPP in the positioning case study (e.g., the product form in Eq. (39) and implicitly the binomial availability counts of Section III-B). However, DSBPP points are not independent: satellites on the same orbit are conditionally dependent given the orbit parameters. The claim of homogeneity (motion-invariance) is not sufficient to justify the application of formulas derived for independent binomial point processes. The authors should clarify the exact invariance properties of the DSBPP and either prove that the relevant marginals coincide with the homogeneous BPP or derive the appropriate correlated expressions for the metrics used.","section":"§II-B.2, §V-E.4"}],"minor_comments":[{"comment":"There is a typo 'θ = 1 − arccos (1 − 2v)' which should read 'θ = arccos (1 − 2v)'.","section":"§II-A.2, Eq. (4)"},{"comment":"The textual description says satellites in each orbit are 'uniformly spaced' (as in a Walker constellation), but the algorithm places them at uniformly random angles; the text should be aligned with the algorithm.","section":"§II-B.2, Algorithm 2"},{"comment":"The definition of K-localizability as a product of marginal probabilities implicitly assumes independence of the K satellites' SINRs; this assumption should be stated explicitly and justified for the shadowed-Rician fading used in the case study.","section":"§V-E.3, Eq. (39)"},{"comment":"The mapping rule for the planar BPP uses ρ(n)=√u(n) RNTP sinθc, which is the inverse-CDF transform for a uniform point on a disk; this is correct, but the text should mention that the spherical and planar BPPs are coupled through the same u(n) so that the error E is not comparing two independent realizations.","section":"§III-C.1"},{"comment":"Table VII lists the additional attenuation ζ = −2 dB, which is a negative value; if ζ is meant to be a loss in dB, the sign convention should be explained.","section":"§V-E.4, Table VII"}],"recommendation":"major_revision","confidential_remarks":"The survey content is solid and offers a useful organization of a fast-growing field, and the standard derivations are correct. The main technical weaknesses are confined to the two original case studies and the claims about the DSBPP's properties; these are fixable within the manuscript's scope, so I recommend major revision rather than rejection. I would encourage the editor to ask the authors to provide either an analytical expectation of the planar-approximation error or Monte Carlo error bars, and to provide a rigorous statement of what 'homogeneous' means for the DSBPP and to validate the positioning case study against real constellation geometry if that comparison is claimed."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper is a solid first survey of spherical stochastic geometry for non-terrestrial networks, and that alone makes it worth having. It organizes the literature into a sensible taxonomy, provides algorithms for the main point-process models, and gives clean derivations of standard results like contact-angle and availability distributions. The DSBPP model is a genuinely new addition: a fixed-count counterpart to the Cox point process, with a simple generation algorithm that practitioners can lift directly. The survey portion is accurate and well-referenced; the authors are major contributors to this subfield, and the self-citations are appropriate.\n\nThe soft spots are in the original quantitative case studies, not the survey. The planar-approximation error analysis in Section III-C computes E from a single spherical BPP realization with N_NTP=100 and chooses h_planar to minimize E over its feasible range. That makes the reported thresholds — E=0.5% at 2θc=3.6° for LEO and so on — best-case numbers for one draw of the random process. No expectation, confidence interval, or repeat-realization check is provided, so a reader following the guidance for when planar modeling is acceptable could easily be misled. This is not a fatal flaw: the qualitative conclusion that HAP is safe to planarize while LEO/MEO need spherical modeling below narrow beamwidths will survive averaging. But the quantitative thresholds should be presented as illustrative, or the authors should add averaging and a fixed-altitude comparison. The K-localizability case study in Section V.E is likewise numerical without derivation or error bars, though it reads more as a plausible demonstration than a hard claim. Minor issues: Algorithm 1 has a typo in step (5) (should be arccos, not 1 − arccos), and Section V.E heading has a misspelling.\n\nOverall, the paper deserves serious review. The survey content is sound and timely, the DSBPP is a useful tool, and the conceptual framing of spherical vs. planar modeling is valuable. The review should push the authors to fix the planar case study's statistical basis, but that is a revision, not a rejection. I'd bring it to our reading group and would cite the survey and DSBPP in my own work.","headline":"Useful first survey of spherical stochastic geometry for NTNs, but the quantitative planar-approximation case study rests on one random realization and a fitted altitude, so its error thresholds shouldn't be taken at face value.","tokens_in":36381,"tokens_out":1228,"would_cite":true,"duration_ms":16287,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60D05","60G55"],"pacs":[],"model":"deepseek-v4-flash","headline":"Spherical stochastic geometry provides an accurate and tractable framework for analyzing non-terrestrial networks, and this survey is the first to organize the field.","keywords":["spherical stochastic geometry","non-terrestrial networks","binomial point process","dual stochastic binomial point process","satellite constellations","performance analysis","planar approximation","K-localizability"],"falsifier":"Regenerate the planar-approximation case study many times: draw independent spherical BPP realizations with $N_{\\text{NTP}}=100$ at HAP, LEO, and MEO altitudes, and for each compute the relative distance error defined by Eq. (17) with $h_{\\text{planar}}$ chosen honestly (for example, the satellite's altitude) rather than error-minimizing. If the mean error at the reported thresholds exceeds 0.5 percent, or if the variance across realizations is large, the paper's rule of thumb fails. A second check is to repeat the MEO K-localizability curve of Fig. 10 with multiple independent DSBPP generations to confirm the 97 percent figure.","tokens_in":35386,"feed_emoji":"🛰️","tokens_out":5791,"duration_ms":53505,"temperature":0.7,"pith_summary":"This survey argues that stochastic geometry on a sphere, rather than on a plane, is the right setting for analyzing non-terrestrial networks, and it organizes the field's models into three families: non-orbital, stochastic-orbital, and fixed-orbital point processes. It adds two original tools: the dual stochastic binomial point process (DSBPP), which gives fixed numbers of orbits and satellites per orbit while preserving homogeneity, and a quantitative case study comparing spherical and planar models for the same constellation. The case study concludes that high-altitude platforms inside their line-of-sight range can be treated as planar with less than 0.5 percent error, while LEO satellites require spherical modeling unless the user's beam is very narrow. A second case study applies the DSBPP to MEO positioning, showing that 24 satellites at 20,000 km give about a 97 percent probability that at least four satellites can participate in localization. If these conclusions hold, researchers gain a structured, tractable toolkit for evaluating coverage, interference, routing, and positioning in satellite networks.","feed_headline":"Satellite models stay spherical: flat maps err fast at LEO","feed_subtitle":"A new survey shows HAPs can stay flat, while LEO and MEO satellites need a sphere beyond a narrow beam.","key_machinery":"The central object is the homogeneous binomial point process on a sphere, where each platform's polar angle has density $\\sin\\theta/2$ and its CDF is the normalized area of a spherical cap, $(1-\\cos\\theta)/2$. This distribution yields the contact-angle CDF, the availability probability, and the $K$-availability probability by binomial counting inside a cap. The paper's new DSBPP extends the same logic to orbits: orbit inclinations drawn from the same sine density, azimuths uniform, and a fixed number of satellites per orbit, mapped by rotation matrices to the sphere. The planar-approximation case study maps each spherical-BPP point to a planar BPP on a disk of radius $R_{\\text{NTP}}\\sin\\theta_c$ with the same azimuth and with radial coordinate $\\sqrt{u}\\,R_{\\text{NTP}}\\sin\\theta_c$, then compares Euclidean distances to a ground user.","core_discovery":"The central claim is that spherical stochastic geometry is both accurate and tractable for NTN analysis, and that the field has matured enough to be surveyed. The paper's original technical contributions are the DSBPP, a stochastic-orbital model with deterministic orbit and per-orbit satellite counts whose generation algorithm is given, and two case studies: a planar-approximation error analysis and a K-localizability analysis of MEO positioning. On the topology side, the paper derives the contact-angle and contact-distance distributions for a homogeneous BPP on a sphere, showing that the probability a user's nearest satellite lies within a central angle $\\theta$ is $1 - ((1+\\cos\\theta)/2)^{N_{\\text{NTP}}}$. On the channel side, it catalogs large-scale fading, small-scale fading, and beam gain models for space/air-to-ground, space-to-air, inter-satellite, and space/air-to-sea links. The paper positions itself as the first survey to cover spherical SG, organizing publications from 2020 to 2024 into four phases and flagging advanced topics such as routing, security, satellite clusters, energy harvesting, and positioning.","pith_inferences":["Beyond the paper, the distance-based error measure likely understates the error in performance metrics like coverage probability, which are nonlinear in distance; a coverage-probability comparison may show planar models failing even earlier than the reported angle thresholds.","Beyond the paper, the same planar-to-spherical mapping could be applied to non-homogeneous constellations such as inclined LEO shells, where latitude-dependent density may make the planar approximation worse or better depending on the inclination.","Beyond the paper, the DSBPP's fixed orbit count makes it natural for studying rare events in small constellations, such as the tail of the localizability distribution, which the Poisson variation in CPP would blur."],"forward_implications":["Model selection becomes a three-tier choice: non-orbital BPP/PPP when only tractability matters, stochastic-orbital DSBPP/CPP when orbital structure matters, and fixed-orbital OGM/PLP when realism matters.","HAPs within line-of-sight range can be modeled as planar BPPs, while LEO satellites cannot unless the receiving beam is narrow enough that the visible spherical cap has central angle below about 3.6 degrees.","The K-availability and K-localizability probabilities give a common language for scenarios requiring multiple simultaneous satellites, such as positioning or joint transmission.","The DSBPP fills the gap between CPP's Poisson randomness and deterministic orbital models, offering fixed counts with homogeneity and tractability."],"supporting_citations":[{"why":"Supplies the nearest-neighbor and contact-distance distributions for a binomial point process on spherical surfaces, which ground several topology derivations in the survey.","marker":"[16]"},{"why":"Provides the stochastic-geometry analysis of LEO satellite communication, including shadowed-Rician fading and gateway relaying, that the channel chapter extends.","marker":"[17]"},{"why":"Establishes that a spherical BPP model gives coverage and rate estimates close to a deterministic Walker constellation, a central accuracy claim.","marker":"[18]"},{"why":"Introduces the Wasserstein distance as a quantitative measure of accuracy for spherical SG models, which the survey uses to guide model selection.","marker":"[56]"},{"why":"Defines the Cox point process for LEO constellations, the basis of the stochastic-orbital family that the new DSBPP is designed to improve.","marker":"[58]"},{"why":"Shows that a CPP-based model yields coverage probability estimates consistent with the Starlink constellation, supporting the accuracy of stochastic-orbital models.","marker":"[69]"},{"why":"Presents a stochastic-geometry routing framework for massive LEO networks, an advanced topic that the survey extends and connects to other use cases.","marker":"[93]"},{"why":"Originates the spherical SG idea by applying it to high-altitude platforms, though it later maps the sphere to a plane; this is the historical starting point.","marker":"[103]"}],"fun_headline_variants":["Sphere beats flat for LEO satellite modeling","Survey: spherical SG sharpens satellite analysis","Flattening the globe skews satellite coverage","Spherical geometry: key to accurate NTN models"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The quantitative thresholds that say when planar modeling is acceptable come from a single randomly generated constellation of 100 points and from choosing the planar altitude that minimizes the error; if those thresholds vary across realizations or parameter choices, the guidance on when to use planar models would change.","fun_headline_variants_meta":{"raw":{"variants":["Sphere beats flat for LEO satellite modeling","Survey: spherical SG sharpens satellite analysis","Flattening the globe skews satellite coverage","Spherical geometry: key to accurate NTN models"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000287,"raw_usage":{"total_tokens":1724,"prompt_tokens":1022,"completion_tokens":702,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":638,"completion_tokens_details":{"reasoning_tokens":644}},"tokens_in":638,"tokens_out":702,"duration_ms":7533,"temperature":1.0,"reasoning_tokens":644,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T20:48:58.626402+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Regenerate the planar-approximation case study many times: draw independent spherical BPP realizations with $N_{\\text{NTP}}=100$ at HAP, LEO, and MEO altitudes, and for each compute the relative distance error defined by Eq. (17) with $h_{\\text{planar}}$ chosen honestly (for example, the satellite's altitude) rather than error-minimizing. If the mean error at the reported thresholds exceeds 0.5 percent, or if the variance across realizations is large, the paper's rule of thumb fails. A second check is to repeat the MEO K-localizability curve of Fig. 10 with multiple independent DSBPP generations to confirm the 97 percent figure.","supporting_citations":[{"cited_title":"A Novel Analytical Model for LEO and MEO Satellite Networks based on Cox Point Processes","cited_arxiv_id":"2212.03549","evidence_quote":"Defines the Cox point process for LEO constellations, the basis of the stochastic-orbital family that the new DSBPP is designed to improve."},{"cited_title":"Spectrum sharing for high altitude platform networks,","cited_arxiv_id":null,"evidence_quote":"Originates the spherical SG idea by applying it to high-altitude platforms, though it later maps the sphere to a plane; this is the historical starting point."}],"review_version":1}