The Sun (radius Rs ≈ 696,000 km) and Moon (radius Rm ≈ 1,737 km), separated by d ≈ 149.6 million km, fix the umbra's half-angle and total length:
δ = asin[(R_s − R_m) / d] ≈ 0.266°
L = R_m / tan(δ) ≈ 374,300 km (umbra length from the Moon's center)
If the Moon sits a distance D from Earth's center (356,500–406,700 km over its elliptical orbit), the leftover cone length at the surface is L − D. The umbra's across-track diameter there is:
w₀ = 2 (L − D) tan(δ)
That circle only stays a circle if the shadow axis meets the ground perpendicularly. Away from local solar noon the axis strikes at an incidence angle θ = 90° − γ from the surface normal (γ = Sun's elevation), and simple foreshortening stretches the footprint along the direction of travel:
w_along = w₀ / cos(θ)
The same 1/cos(θ) foreshortening speeds up the point where the cone axis meets the ground, on top of the umbra's intrinsic orbital ground speed v₀ (~0.61 km/s). Earth's own rotation, moving the same direction the shadow travels, subtracts from that:
v_ground = v₀ / cos(θ) − v_eq cos(latitude), v_eq ≈ 0.465 km/s
duration = w_along / v_ground
- Moon distance — closer to perigee leaves more umbra length past the surface, widening w₀; near 374,300 km the leftover (and the path) shrinks to nothing.
- Sun elevation γ — low Sun (dawn/dusk crossings) stretches the footprint into a long ellipse and speeds up its ground track, which is why totality is briefest near sunrise/sunset even though the shadow looks wider.
- Crossing latitude — higher latitude means the ground under the shadow is moving slower with Earth's own spin, changing how much gets subtracted from the ground-track speed.
This is a schematic, first-order model (a real path also curves with Earth's oblateness and the Moon's along-orbit speed), but every term above is the real governing relation used to explain why some eclipse tracks are short, fast slivers and others are slow, wide sweeps.