Why two nearly parallel bearings give an unstable intersection
Compare two rays crossing at 1° and 45°, then perturb one bearing to measure how much the calculated point moves.
10 m between origins · 1° intersection versus 45°On this page
Draw a triangle whose answer can be checked by handRepeat with a stronger crossing angleIndependently check that the computed point lies on both raysDecide whether the geometry supports the field decisionTwo bearings can be entered correctly to the displayed precision and still yield a poorly determined intersection. The problem shows up when the rays cross at a shallow angle. A small turn shifts their crossing a long way along the line, even though both plotted rays look almost unchanged.
Draw a triangle whose answer can be checked by hand
Work in metres on a local plane. Place A at N 0, E 0 and B at N 10, E 0. From A aim due east. From B aim one degree south of east, corresponding to azimuth 91°. The ray from B loses 10 m of northing before it crosses A’s eastward ray. In a right triangle, tangent of the angle below east equals vertical drop divided by horizontal run. Therefore the crossing is at E = 10 / tan 1°, about 572.900 m, N 0.
Change only the second azimuth to 91.1°. The crossing moves to E = 10 / tan 1.1°, about 520.807 m. A 0.1° change moved the intersection about 52.09 m. This example exaggerates a practical sensitivity so the effect is easy to see. It does not imply that a particular instrument normally has a 0.1° error.
Repeat with a stronger crossing angle
Keep the same A and B, but direct B’s ray 45° south of east, azimuth 135°. The intersection is exactly E 10, N 0. Change the azimuth by the same 0.1° to 135.1°. The new easting is about 9.965 m. Here the shift is about 35 mm rather than tens of metres.
| Direction from B | Position along A’s east ray |
|---|---|
| 1.0° below east | 572.900 m |
| 1.1° below east | 520.807 m |
| 45.0° below east | 10.000 m |
| 45.1° below east | 9.965 m |
Both the crossing angle and the distance from the observing stations have changed between those examples. The table illustrates the combined behavior for these fixed station positions. It should not be used as a general precision rule for every pair of stations.
Independently check that the computed point lies on both rays
For the first solution, subtract B’s coordinates from the crossing. The north change is −10 m and the east change is about +572.900 m. The downward angle from east is arctan(10 / 572.900), approximately 1°. Subtract A’s coordinates and the north change is zero, confirming the first ray is due east.
Extend that check to the forward direction. Infinite straight lines may intersect behind one station even when forward rays do not. If the computed point lies west of A, it is inconsistent with a due-east sight from A. A sign error can produce an apparently reasonable line intersection while failing the observed direction at a station.
Decide whether the geometry supports the field decision
Plot the rays and their observing stations at a scale that shows the crossing angle. Recompute using a small positive and negative perturbation consistent with the actual observation procedure. The spread is a sensitivity experiment. A formal coordinate uncertainty needs station uncertainties, observation weights and a suitable network model.
An additional sight from a third station with a different direction can reveal whether the intersection survives independent evidence. A measured distance can provide another check. The Traverse Desk calculator tests closed straight-course observations and cannot assign statistical precision to a pair of rays. Use the inverse directions and perturbation test here when reviewing an intersection produced elsewhere. Record the station pair and both original bearings beside the resulting coordinate so someone reviewing a surprising position can reconstruct why it moved.
For exactly parallel distinct lines the denominator of the intersection calculation is zero and no finite crossing exists. Near-parallel sightings can still produce a valid numeric coordinate. That difference is why a successful computation should be followed by a geometry check.