Back to the rankings

Follow-up Q21 (theme: flat-vs-round-earth): a low-budget experiment that proves a round earth from first principles

Date: 2026-10-10. Role: STEP F follow-up investigator. Web time: about 2 minutes by the date +%s timer (109 s from the first search to closing the tab; Chrome, one tab, closed). The search stopped early because the question is mostly a design and calculation question; the web was needed only for the objections and for low-budget precedents.

Theme members (map 03, 2026-10-10): Flat-Earth (core, the only member). Files read: theme definition and card (work/themes/entries/flat-vs-round-earth.json); Flat-Earth arbiter.md (sections 1-2 in full, section 9 by grep), primer.md (Corrections section), facts.json (entries for the TFES Electromagnetic Acceleration and Foucault Pendulum pages, Foucault 1851, Wallace 1905), and the deep reads of the TFES Electromagnetic Acceleration page, the TFES Foucault Pendulum page, Foucault 1851 and Hirt et al. 2010.

Tags: [read] project document or its deep read; [web] with URL; [memory] with confidence; [calc] my arithmetic (script in the session scratchpad, numpy; not kept).

The question

Design a step-by-step, low-budget experiment that proves a round earth from first principles, with no belief necessary and round earth as the only possible conclusion. Record every assumption the design makes (for example that the stars are far away). Instead of assuming it, test it within the experiment, or say plainly that it is still an assumption. The orchestrator's first draft (2026-10-10 chat) had: a stars-far test by constant star separations; north-south pole-star elevation and east-west transit timing with a home-made quadrant, cross-staff and watch; Foucault pendulum rates at two latitudes as a check that uses no light; and an assumption register A1-A11. The register itself is not in the files this agent can read, so this note gives its own register (A1-A12 below), keeping A3 = "light travels straight" as in the question.

A position put to the review for A3: light travelling straight is an assumption, a reasonable one, and a laser can prove it. The orchestrator's caveat, agreed for the record: a laser against a taut string or straightedge settles it over tens of metres. Over long paths through air, light bends by a measurable amount (refraction coefficient k about 0.13-0.17 over land, more over water). Over water a laser alone cannot tell the earth's curvature from light bending, which is the TFES Electromagnetic Acceleration case. So the starlight steps rest on A3 beyond the range where it was tested, and the pendulum step does not.

Short answer

The design below (five steps plus a light-free check) can be done for under about 150 US dollars plus one cooperating observer in the other hemisphere. In outline:

• Steps 1-3 measure star separations, pole-star elevations and transit times. They rule out every flat model with near stars and straight light by margins of tens of degrees and tens of percent. Each step also tests its own assumptions where it can.
• Step 4 brings A3 up to the horizontal scale of about 10-20 km at low cost. It does this by measuring the bending of the air (two thermometers at different heights) instead of assuming it. The discriminating number is the offset left after correcting for that bending: 0 on a plane, L^2/8R on a globe.
• No low-budget test brings A3 to the space part of the starlight path. There it stays an assumption. Without it, the starlight steps prove only "a globe, or an untested light-bending law tuned to imitate one".
• The pendulum step (5) needs no long light path. Any flat surface under one uniform rotation gives the same precession rate and sense everywhere. Rates that differ, and reverse sense across the equator, show that the local verticals are not parallel. With a measured ground distance, that gives a radius.
• One assumption is left: that a single uniform rotation sets the pendulum's frame (A10). Without it, the result is "round, or a place-dependent inertia field".
• Strictly, nothing removes every conceivable alternative. What the design reaches is "the only conclusion that needs no separately tuned unknown law for each instrument".

Searches (Chrome, Google; plus curl of pages to work/themes/research/pages/)

Keywords tried: the four Google queries above; TFES page titles Polaris, Southern Celestial Pole, Constellations, Distance to the Stars, Celestial Sphere, Stellar Parallax. Not reached (stopped early, not by the cap): Kamerlingh Onnes 1879 pendulum thesis; Bravais and other 1851-52 latitude series; amateur cross-staff star-separation records; measured k over the Old Bedford River.

What was found

Flat-Earth sources the design must answer

• TFES "Electromagnetic Acceleration" (EA) [read: deep read]. "All light curves upwards over very long distances". The formula is y = (3/4)(beta x^4/c^2)^(1/3), with no value for beta ("attempts will be made to measure this experimentally"). The page says the formula is "not valid for short-range experiments". It also says "the bending of light does not simulate the rate of globe earth curvature. Instead, the bending occurs more gradually over a greater distance." Burden-of-proof argument: "no evidence that light is curving is ... an admission that there is not sufficient evidence that light is straight." Arbiter verdict: Not testable as stated until beta is fixed. Making a sunset needs beta of order c^2/R = 1.4e10 m/s^2 [read: arbiter claim 3].
• TFES "Celestial Sphere" [web: https://wiki.tfes.org/Celestial_Sphere]. It claims EA, unlike "conventional astronomy", "directly provides a mechanism for why lines, containing points at different distances from the observer, appear curved in the sky and for our domed observations." So EA claims the dome-like look of the sky, including star patterns that keep the same angles. It gives no formula for this.
• TFES "Stellar Parallax" [web: https://wiki.tfes.org/Stellar_Parallax]. It quotes N. T. Jones on Tycho catalogue parallaxes: "25%" negative and "29%" positive beyond 20 mas. It concedes that the largest positive value is "on the order of only 0.7 of an arcsecond" and that the effect "is known to be genuine". It reads the negative values as real "outer" stars. [memory, high] Tycho parallaxes have errors of tens of mas, so values near zero scatter to both signs by measurement error. The Hipparcos and Gaia catalogues give mostly positive parallaxes with much smaller errors. That reading is the mainstream one; it is not weighed further here. Even the page's own figures put every star's parallax below 1 arcsec, i.e. the stars are very far compared with any baseline on the ground, provided light travels straight (A3).
• TFES pages Polaris, Southern Celestial Pole, Constellations, Distance to the Stars [web: wiki.tfes.org/{title}]: all empty. The society's wiki has no stated model of pole-star elevation or of the southern sky. The 1904 Blount-Smith two-vortex model is the historical answer [read: arbiter claim 3, verdict Historically superseded but instructive].
• Rowbotham's star distances [read: arbiter section 9.2, from the 1865/1881 star-tube passages]: stars "1,000 to a few thousand miles" away. Rowbotham's sun about 3,000 mi up [read: arbiter claim 3].
• TFES "Foucault Pendulum" [read: deep read]. The motion "has more to do with the initial conditions ... than the supposed rotation of the earth". Museum pendulums are motor-driven. It cites Longden 1919 (reversing the wire reversed the sense), Airy's private "fraud", the elliptical motion seen at Newark in 1851, and Rowbotham's ball-and-socket suspension. It offers Mach's principle as an alternative. It gives no rate and no sin(latitude). It cites the South Pole 2001 page for a rate "twice what is expected".
• Foucault 1851 [read: deep read, PDF p.16]: the plane's angular displacement equals "the angular motion of the Earth in the same time multiplied by the sine of the latitude". At Paris the trace "does not make a full turn in twenty-four hours". He gives no hours figure.
• Wallace 1870 / Oldham 1901 [read: arbiter claim 1 and 9.1]: the equal-height trio over 6 miles. The middle disc was about 5 1/2-6 ft above the line in sketches signed by both referees. The umpire wrote "five feet, more or less". Refraction was not measured. Rowbotham's 1870 rebuttal: the agreed near-telescope disc was missing, so "the money was neither won nor lost".
• Hirt et al. 2010 [read: deep read]: k between -4 and +16 at 1.8 m over grass on sunny days ("mostly between -1 and -3" by day, up to +12 to +16 near sunset). The Gaussian +0.13 "refers to the intermediate atmosphere" (about 20-100 m), and the range is "fairly small ... well above the surface" (+0.10 to +0.20). k from the temperature gradient: k = 503 (p/T^2)(0.0343 + dT/dz).

Low-budget precedents and records found on the web

• ASEE 2021, two-hemisphere small pendulum [web: https://peer.asee.org/36676; PDF saved as pages/asee-36676.pdf]. A student at York College of Pennsylvania (39.94 N) and at a home in San Nicolas de los Arroyos, Argentina (33.36 S) used 3 m pendulums, built "with a low budget". The results were honest and mixed:
• Small balls on a C-clamp suspension precessed too fast. One case was 11.0 deg in 0.5 h, against 4.83 deg predicted. Their sense also reversed during runs.
• "Regarding the trials in Argentina, the results did not match those expected, since the reversal of the precession was observed similarly to the ones conducted in the USA."
• Only the bowling-ball runs on a fixed suspension behaved. They "always precessed in the expected CW direction" whatever the launch direction: 8.09 deg/h against the predicted 9.65 deg/h (16% low).
• No bowling-ball run was made in Argentina. So the low-budget two-hemisphere comparison was not achieved there. It is a design lesson, not a result.
• South Pole 2001 [web: https://www.southpolestation.com/trivia/00s/southpolefoucault.html]. This is the page the TFES Foucault page quotes. Its abstract: the pendulum "was determined to have a period of 24 hours, +/- 50 minutes", measured as "5 deg every 20 minutes as expected", with a 33 m guy wire, a 25 kg bob and no drive. The "twice what is expected" sentence comes from an intermediate attempt that the team then modified. The page also says the first attempt ran backwards and was "modified" after they reasoned which way it should go. A flat-side reader can fairly say that the builders adjusted the pendulum until it gave the expected result. The final measured rate is still the full-rotation rate the sine law predicts at 90 deg, and opposite in sense to northern pendulums.
• Tamarack mine plumb lines [web: https://dsimanek.vialattea.net/hollow/mcnair.htm, reprint of F. W. McNair, Science 15, 20 June 1902]. Wires 4,250-4,440 ft long, about 15 ft apart, were "farther apart at the bottom" in some shafts (+0.11 ft, +0.141 ft). They were closer in others (-0.028 ft). Hollow-earth and concave-earth writers cite the divergence. McNair traced the effect to ventilation air currents: with the air chimney sealed, the divergence fell to +0.018 ft. A globe predicts a convergence of 15 x 4,440 / 20.9e6 = 0.003 ft [calc]. That is about one-fiftieth of the air-current effect. Plumb lines in mines are therefore not a usable low-budget curvature test.
• Rainy Lake experiment, 2018 [web: Bislin page above]. G. Hnatiuk, J. Kozlowski, "Soundly" and W. Bislin ran an advanced Bedford design over about 10 km of frozen lake. It used "Bedford targets and Tangent targets", a theodolite, differential GPS and drone views. They conclude a globe of radius 6,371 km. Their refraction values (k = 0.003 to 0.201 at target adjustment) were derived from the difference between observations and GPS positions, which assumes the GPS geometry. A flat-side reader rejects GPS. The design is the right shape; its refraction step is not independent of the globe.

The design, step by step

Equipment (about 2026 prices): - Quadrant: protractor, plumb bob and sighting tube, about $10. - Cross-staff (Jacob's staff), about $10. - Laser pointer, about $10. - Builder's string, about $5. - Two matched thermistors with radiation shields and a logger, about $40-60. - Pendulum: bowling ball, 3-10 m of steel wire, rigid clamp, Charron ring, burn-thread release, about $60. - A car odometer. - A phone, for time sync by voice call or a radio time signal (DCF77, WWV, MSF). - One cooperating observer in the southern hemisphere for steps 2, 3 and 5.

Step 0. Calibrate the instruments (tests A3 short range, A5)

• Stretch a string taut over 20-50 m and run the laser along it (and along a straightedge). The beam stays on the string to within a millimetre or so: A3 holds at tens of metres [memory, high: standard optics].
• Plumb or level reversal: read an angle, turn the quadrant 180 deg, read again. Half the difference is the instrument's error. This tests A5.
• Calibrate the odometer on a surveyed km (motorway marker posts) or by measuring with a tape (A6).

Step 1. Are the stars far? (tests A1, partly A4; rests on A3)

• With the cross-staff, measure the angle between two stars of a pair: once high (60 deg elevation or more) and once low (25-30 deg) on the same night. Repeat at a second site 500 km or more away, and at the southern partner's site.
• Globe with far stars: the separations stay constant to within the differential refraction. At 25-35 deg elevation that is about 1-2 arcmin [calc, Bennett's refraction formula: 5.4' at 10 deg, 1.7' at 30 deg, 1.0' at 45 deg, 0 at 90 deg], well under the cross-staff's 0.1-0.2 deg.
• Near dome: put the stars on a plane 5,000 km up (Rowbotham's "few thousand miles") [calc]. A pair 10.0 deg apart overhead appears 5.0 deg apart along the line of sight and 7.1 deg across it when seen at 45 deg elevation. At 27 deg elevation the figures are 2.0 and 4.5 deg. Constellations would shrink by 30-80% as they set and would change shape from place to place.
• What it proves: either the stars are far compared with 500-10,000 km, or light bends so as to make a near dome look like a far sphere. The EA Celestial Sphere page claims the second, without a formula. So Step 1 tests A1 only on condition of A3.

Step 2. Pole-star elevation, north-south (A2, A4, A5, A6, A8; rests on A1, A3)

• Measure the elevation of the celestial pole at three or more sites along roughly one meridian, spanning 1,000 km or more (Polaris is offset about 0.7 deg from the pole: average the top and bottom of its daily circle) [memory, high].
• A8 is tested within the step: when the sites lie on one meridian, the same star transits at the same clock time at each.
• Globe: the elevation changes 1 deg per 111.2 km and is linear in distance [calc].
• Flat earth with the pole at height H: elevation = arctan(H/d), where d is the distance to the north-pole point. The best fit to sites at 30 and 45 deg is H = 4,440 km [calc]. It then predicts 53.1 deg where the globe gives 60, and 28.0 deg where the globe gives 15. At the equator it predicts 23.9 deg, where the globe gives 0. At 30 S it predicts that Polaris stands at 18.4 deg; on the globe it is below the horizon.
• The southern partner reports Polaris invisible and a second, southern rotation pole at elevation equal to the latitude, with the sky turning the other way. No single dome over a plane gives that.

Step 3. East-west transit timing (A6, A7, A8; rests on A1, A3)

• Two observers at the same pole-star elevation (that is, the same latitude, A8 tested by Step 2) time the same star crossing their meridians. Time is synced by phone call or radio signal (A7). The odometer gives the east-west distance s (A6).
• Globe: one degree of sky rotation (4 min) per 111.2 cos(lat) km. Flat north-centred map: one degree per (90 - lat) x 1.94 km [calc]. At 45 N the figures are 78.6 vs 87.3 km per degree: 500 km takes 25.4 vs 22.9 min, a difference of 2.5 min that a watch shows easily. At 34 S they are 92.2 vs 240.6 km per degree.
• Step 4 below closes the loop: R from the north-south leg, s/(delta latitude), must equal R from the east-west leg, s/(delta longitude x cos lat). This checks itself.

Step 4. Bring A3 to the 10-20 km horizontal scale (corrected Wallace)

• Over 5-10 km of water or flat fen, put up three to five targets at equal heights above the water. Read their apparent offsets from the line through the near and far marks, with the laser and the cross-staff or a cheap level.
• At the same time, log the air temperature at two heights (for example 1 m and 5 m) at both ends and in the middle. Work out the ray curvature from the air itself: 1/r = 7.9e-5 (p/T^2)(0.0343 + dT/dz) per metre (Hirt eq. 2 with R taken out, 503/6.371e6 = 7.9e-5) [calc]. No earth radius goes into this.
• Prediction: the apparent offset of the middle mark plus L^2/(8r) gives L^2/(8R) on a globe (6.00 ft over 6 miles) and 0 on a plane, whatever the air is doing [calc; same result as primer correction 2].
• Choose sight lines 20 m or more above the surface, overcast or breezy weather and several times of day, where k is steady near +0.1 to +0.2 [read: Hirt]. Near the surface k swings from -3 to +16, and a thermometer pair cannot keep up.
• Precision: getting k to +/-0.1 needs dT/dz to about 0.016 K/m, i.e. 0.06 K over 4 m. Matched, shielded thermistors can do that [calc]. The weak point is how well three stations stand for the whole path.
• What this extends: A3 in the form "light follows the path that the measured air predicts", out to the horizontal scale of the sightings argued over in the Flat-Earth review. It also tests EA's own words: EA bending is "more gradual" than globe curvature, so EA predicts a corrected offset below L^2/(8R). And it closes the TFES point that over water a laser alone cannot separate curvature from bending: the thermometers do the separating.
• What it cannot extend: the starlight path, which is mostly outside the atmosphere. There, a law with beta about c^2/R would imitate a globe exactly in every optical test, horizontal ones included. No low-budget test reaches that scale. A3 stays an assumption for Steps 1-3.

Step 5. The light-free check: Foucault pendulums in both hemispheres (A9, A10, A11)

• Physics [memory, high; Foucault 1851 read]. A free pendulum's plane turns relative to the floor at the rate Omega . n, where Omega is the rotation of the floor relative to the local inertial frame and n is the local upward vertical (the plumb direction).
• Key point [calc, geometry]. Take any flat surface, with all verticals parallel. Under any single uniform rotation, Omega . n is then the same everywhere, in rate and sense. This holds for a rotating earth, a rotating cosmos dragging inertia (the Mach option on the TFES page), or no rotation at all (a stationary universal-acceleration plane, rate zero everywhere).
• Rates that differ from place to place, and above all opposite senses at two sites, mean that the verticals are not parallel. The angle between them is at least |theta1 - theta2|, where cos(theta) = rate / Omega. Omega is 15.04 deg/h, from the sidereal day: the time a star takes to return to the same direction. That uses light only in the weak sense that any bending is the same from night to night.
• Prediction [calc]: 9.66 deg/h clockwise (seen from above) at 39.94 N; 8.27 deg/h anticlockwise at 33.36 S. Thus theta = 50.1 deg and 123.4 deg, so the verticals are at least 73.3 deg apart. The great-circle distance York to San Nicolas is 8,324 km on the globe; the flat north-centred map gives 8,527 km, almost the same, so the distance does not bias the test. That gives R of at most 6,510 km. The globe value is 6,371.
• Measured so far: South Pole 24 h +/- 50 min (that is, 15.0 +/- 0.5 deg/h) [web]; York 8.09 deg/h for 9.65 predicted (bowling ball) [web]. No clean low-budget southern-hemisphere rate was found in this search. The ASEE Argentina runs failed, because their suspension was faulty.
• Controls, from the objections: a bowling ball or a heavier bob; the longest drop available (a stairwell of 10 m or more); a rigid single-point clamp or a Charron ring (stops elliptical drift); burn-thread release (no lateral impulse); no drive (removes the "motorised" objection); launch along north-south and along east-west; repeat with the wire turned 180 deg (Longden's test); at least 1-2 h per run, at least 5 runs, mean and spread reported. A9 is tested if the rate and sense do not depend on launch azimuth or wire orientation. Each run costs only a few dollars, so the controls are within budget.

Step 6. Put it together

• Steps 2, 3 and 5 each give R (or a bound on it) by independent routes: star elevations, star timing, and the turning of the inertial frame relative to plumb lines.
• Step 1 and Step 4 test the light assumptions as far as a low budget reaches.
• A flat model has to reject all of them at once, each with its own ad hoc law: a near dome plus a light law that makes it look far; a pole star that drops below the horizon and a second southern pole; and an inertial frame that turns at different rates, and in opposite senses, at different places on one plane.

Assumption register

Strongest flat-earth objections, step by step

1. Step 1 (stars far): EA "provides a mechanism for why ... [we see] domed observations" [web: TFES Celestial Sphere]. Answer: the page gives no law and no beta, so it predicts no separation. A tuned law that preserves angles exactly is possible in principle, and that is why the design does not stop at Step 1.
2. Step 2 (pole star): perspective (Rowbotham) and EA lowering of receding points. Answer: perspective gives arctan(H/d), which the three-site fit rejects by 7-13 deg [calc]. EA has no beta. Neither model has an account of the southern pole, and the TFES pages on it are empty [web].
3. Step 3 (timing): distrust of GPS and time from official sources. Answer: sync by a voice call between the observers; the error is under 1 s. Odometers do not depend on GPS.
4. Step 4 (horizontal light): "the laser over water shows no curvature" (claims of 16-mile tests in flat-earth groups, seen in search results only, not read) and the Bedford sightings. Answer: low-eye sightings need k of 0.4-0.9, which is unmeasured, so they are "Never properly tested" [read: arbiter]. Step 4 measures k. Second objection: thermometers measure only ordinary refraction, not EA. Answer: then EA's own "more gradual than globe curvature" predicts a measurable shortfall at 10 km.
5. Step 5 (pendulum): initial conditions, motor drives, wire torsion (Longden 1919), "the South Pole one ran twice as fast", "they modified it until it agreed", Mach's principle [read: TFES Foucault deep read; web: South Pole page]. Answers: - The controls in Step 5 address the first three. - The South Pole page's final value is 24 h +/- 50 min. - The Mach option is covered by the parallel-vertical argument. - What stands: the low-budget record so far is poor. ASEE got 16% low in the north and no valid southern runs. A clean two-hemisphere result with controls has not been found in this search and would have to be made.
6. Plumb lines diverge at Tamarack: true as measured. The cause was air currents (McNair 1902) [web]. The design does not use mine plumb lines.

Reasoning and competing readings

• Reading 1 (the design's claim): once the controls are passed, the round earth is the one conclusion that needs no extra tuned law. A flat model has to add three separate laws: a light law (EA with a tuned beta and an angle-preserving form), an inertia law (a frame that turns differently at different places) and a southern-sky law.
• Reading 2 (a strict flat-side reading): a law that can be tuned can imitate anything, so no finite experiment gives "the only possible conclusion". This is logically correct. It applies equally to any physical claim and is a matter of economy of explanation, not of evidence. The design makes the price of the alternative explicit, step by step, which is as far as an experiment can go.
• On A3 specifically: the position that A3 is "an assumption, reasonable and provable by laser" is right at the bench scale. A laser alone does not reach the scale that matters, and over water it confounds curvature with bending, as the caveat agreed. Adding two thermometers turns the laser test into one that separates the two at 10-20 km. Nothing affordable reaches the vertical and space path, so the pendulum branch is what makes the design free of A3.

Corrections to earlier documents

1. Stale copies of the "plane predicts zero whatever k is" error. Primer correction 2 already fixed this for arbiter.md claim 1. The same wording survives in: - review/Flat-Earth/arbiter.md line 59 ("on a plane, zero whatever k is ... no dependence of the offset on measured k at all"); - review/Flat-Earth/physicist.md line 177 ("0 for a plane regardless of k"); - work/themes/entries/flat-vs-round-earth.json, settle.experiment ("plane: zero whatever k is").

It should read: apparent offset = (1 - k) x 6.00 ft on a globe and -k x 6.00 ft on a plane, the same slope with intercepts 6.00 ft apart. So the k-corrected offset is 6.00 ft on a globe and 0 on a plane [calc]. 2. Consequence for Wallace 1870, not stated anywhere. A plane with k of about -1 reproduces Wallace's pattern exactly: the middle disc stands about 6 ft above the line, in the 1:2 pattern. That is light bending away from the ground with radius about R, a temperature gradient of about -0.19 K/m [calc, Hirt eq. 2 at 10 C]. So the claim that "Wallace's design is robust to uniform refraction" holds on the globe side only. Against a plane, the 1870 result excludes everything except k near -1. Hirt measured k "mostly between -1 and -3" in the daytime at 1.8 m over sunlit grass. Over a March canal at 13 ft that value is unusual but was not measured. The arbiter's caveat "refraction was not measured" therefore carries more weight than its wording gives it. The verdict (Tested, not supported, with that caveat) need not change, but the caveat should name k of about -1 as the one condition under which a plane fits. work/rankings/entries/Flat-Earth.json supporting item 1 ("a plane predicts zero") should say "a plane with straight light predicts zero". 3. South Pole pendulum (arbiter claim 2, Differences item 5, "a single station-trivia page, no data or cause"). The cited page's abstract gives a final result: "a period of 24 hours, +/- 50 minutes", 5 deg per 20 min, clockwise rotation of the earth seen from above the South Pole [web: southpolestation.com page above]. The "twice expected" sentence is an intermediate attempt. The TFES Foucault deep read was right that the TFES excerpt omits this. The arbiter's description should be updated: it is an anecdote of an adjusted build whose final reported rate is the predicted one.

Leads for the member project (Flat-Earth), for its next review update (follow-up Q21)

• Supporting information: ASEE 2021 low-budget two-hemisphere pendulum (York 8.09 vs 9.65 deg/h; the Argentina runs invalid) [web: https://peer.asee.org/36676]. The South Pole 2001 final rate, 24 h +/- 50 min [web: southpolestation.com]. The Rainy Lake experiment of 2018, an advanced Bedford design over 10 km with GPS; its k comes from GPS, so it is not independent [web: Bislin].
• Related work: Tamarack mine plumb lines (McNair, Science 1902), cited by hollow- and concave-earth writers; divergence of up to 0.141 ft explained by air currents [web: Simanek reprint].
• Flat-side statements: TFES Stellar Parallax page (Jones; parallaxes all below 0.7 arcsec, negative-parallax reading); TFES Celestial Sphere page (EA explains the "domed" sky); TFES pages Polaris, Southern Celestial Pole, Constellations and Distance to the Stars are empty [web].
• Experiment to add to "What would settle it": the k-corrected equal-height test with thermometers (Step 4) and a controlled two-hemisphere pendulum (Step 5).

Sources

• Flat-Earth review files: arbiter.md (claims 1-3, section 9), primer.md (Corrections), facts.json, deep reads of TFES Electromagnetic Acceleration, TFES Foucault Pendulum, Foucault 1851 (Comptes rendus 32), Hirt et al. 2010, Wallace 1905.
• https://peer.asee.org/36676 (Celario Sedano and Ruzybayev, ASEE 2021)
• https://www.southpolestation.com/trivia/00s/southpolefoucault.html
• https://wiki.tfes.org/Celestial_Sphere ; https://wiki.tfes.org/Stellar_Parallax ; https://wiki.tfes.org/Polaris (empty)
• https://dsimanek.vialattea.net/hollow/mcnair.htm (McNair 1902 reprint)
• https://walter.bislins.ch/bloge/index.asp?page=Proof+of+Earth+Curvature%3A+The+Rainy+Lake+Experiment