Different heights have different origins
GNSS yields position relative to a reference ellipsoid. A surveyed height above a sea-level-related datum uses a different surface. Geoid separation links those descriptions. Treating both height numbers as interchangeable can introduce a large vertical error.
A local plane is a tool, not a global shape claim
A project can use local east, north and up axes. Curvature error grows with extent, so a survey chooses a projection and applies the corrections needed for its accuracy. Surveyed sight lines also respond to atmospheric refraction.
Independent networks measure different quantities
Ground surveying measures angles and distances. GNSS uses radio timing, satellite laser ranging uses round-trip light time, and VLBI compares arrival times of radio signals at separated stations. Their combination estimates position and orientation without reducing all observations to a photograph.
A bridge made of repeated local spans
The Historic American Engineering Record documents 56-foot spans in the first Lake Pontchartrain Causeway and 84-foot spans in the second. Its 1956 design records ordinary clearance near 16 feet above average water and separate higher navigation passages. Those vertical references and repeated segments are actual project dimensions; the bridge is not one rigid, globally horizontal plank.
For a spherical comparison, a 56-foot segment (17.0688 m) has tangent departure L²/(2R) of about 0.023 mm; the same expression over 23.83 miles is about 115 m. The first describes a local construction scale, while the second describes departure from a single tangent extended across the whole crossing. Neither is a claim that the bridge’s far end sits 115 m below the local water level. A sight-line comparison must also include eye height, clearance profile, water datum and refraction.