11.3 Field Technology Limitations

What III.12 is actually asking

BPELSG Domain III professional activity 12 is identify limitations of technologies for use in survey practice. The knowledge list under Field Operations names B capabilities and limitations of GPS, laser scanning, levels, total stations, and UAS; J error sources (multipath, data input, instrument calibration); and Q geoid models, ellipsoid heights, and orthometric heights. The trap is treating a data collector’s “fixed” flag, a pretty point cloud, or a UAS orthophoto as if the physics went away.

This independent OpenExamPrep section stays in the field: what fails on a California occupation, and what a licensee does instead of hoping the office will filter it later.

The H = h − N reduction trap

NGS hybrid geoid models convert NAD83 ellipsoid heights from GNSS to orthometric heights (elevations). The field equation is:

H = h − N

• h = ellipsoid height (GNSS, relative to the reference ellipsoid in the stated frame)
• N = geoid height (height of the geoid above the ellipsoid) from the geoid model
• H = orthometric height (NAVD88 when GEOID18 is used correctly in CONUS)

GEOID18 is intended for NAD83(2011) epoch 2010.00 and yields orthometric heights consistent with NAVD88 in the conterminous United States. It superseded GEOID12B in CONUS. NGS states GEOID18 does not cover Alaska, Hawaii, Guam, or CNMI—those users stay on GEOID12B. California is CONUS, so GEOID18 is the current hybrid model for that NAD83-to-NAVD88 path.

NGS’s geoid-height service for GEOID18 (model 14) at 38.5816°N, 121.4944°W (downtown Sacramento) returns N = −30.792 m (reported 1-sigma error 0.043 m). At 34.0522°N, 118.2437°W (downtown Los Angeles) it returns N = −35.109 m (error 0.030 m). In much of California N is negative (geoid below the NAD83 ellipsoid), so subtracting N increases the numeric height.

Worked reduction, Sacramento N = −30.792 m, GNSS h = 12.400 m (NAD83(2011) ellipsoid height on a project point):

H = 12.400 − (−30.792) = 43.192 m NAVD88 (geoid model path)

Three field blunders, same numbers:

1. Report h as H. The collector stores 12.400 m. The topo and the sewer invert are 30.8 m too low. That is not a “geoid adjustment later” problem; it is a wrong vertical datum on every shot.
2. Compute h + N because “the geoid is negative so I add.” 12.400 + (−30.792) = −18.392 m. Wrong sign convention relative to H = h − N.
3. Wrong model or wrong ellipsoid. A rover still on GEOID12B, or an ellipsoid height in ITRF/WGS84 fed to a NAD83 GEOID18 grid without transforming h first, applies an N that does not belong to that h. NGS also states GEOID18 estimated uncertainties describe fit to the bench-mark surface used to build the model—not a substitute for occupying valid NAVD88 bench marks when the project needs them.

California crustal motion adds a related limitation: a GEOID18 reduction does not freeze a 1991 epoch into 2026. If the control is CCS83 with a stated epoch, the field reduction must keep realization, epoch, and geoid model in one consistent chain. Mixing a 2010.00 ellipsoid height, a 1991 published ortho, and a leftover GEOID12B file is a knowledge-Q failure even if the rover sat in the open.

GNSS / GPS: multipath and the green “fixed” light

Multipath is energy that reaches the antenna after reflecting from stucco, glass, vehicles, chain-link, water, or a nearby construction trailer. Path length increases; the solution can still report fixed integers. Short occupation is not a cure. Field tells: high PDOP with a “clear” sky plot that ignores a warehouse wall, residuals that jump when a truck parks, SNR that sags on one side of the sky, and a check shot to a known point that misses by hundredths to tenths while the rover insists it is fine.

Mitigation is a method change, not a tighter RMS mask: move the rod, occupy longer and inspect the time series, use a different session, or stop using GNSS and turn the total station or digital level. RTN limitations are in the same family: wrong mount point, latency, a coordinate system in the rover that is not the project CCS83 zone, and a single-base solution with no independent check. Knowledge J’s other two examples still apply: data input (wrong antenna height, 2 m pole entered as 2.00 ft) and instrument calibration (tilt compensator not enabled, or a damaged antenna).

Canopy is not the same as multipath, but they stack. Under oaks, GNSS may not fix at all, or it may fix on a biased set. That is a limitation, not a reason to “hold the rover still until it turns green.”

UAS: canopy voids are missing ground, not a smooth surface

Unmanned aircraft systems (UAS) collect photogrammetry or airborne lidar. Photogrammetry correlates on visible surfaces. Closed oak or redwood canopy becomes the surface in the DSM. Ground-classification software cannot invent ground that was never imaged. Those holes are canopy voids. Painting a DTM across them produces a false ground that will not match a rod shot at the trunk.

Lidar UAS can record some returns through gaps, but still leaves voids where the pulse never reached soil, and vegetation returns that look like ground if the classification is lazy. Other UAS limits the exam expects you to name: wind and flying-height versus ground sample distance, shadows on north-facing slopes, lack of ground control and check shots, and water or uniform pavement that defeats image matching. A flight does not replace a monument search, and it does not replace breakline shots in the voids.

Laser scanning: noise is not a topographic breakline

Terrestrial or mobile laser scanning delivers dense clouds. Density is not accuracy at the feature. Range noise, incidence angle on asphalt, and mixed pixels at a curb lip or building corner smear the discontinuity that a topographic survey must show as a breakline (face of curb, flowline, top of bank, wall, building dripline). Auto-extracted contours from a noisy cloud turn a six-inch curb into a ramp. The limitation is physical: the instrument averaged two surfaces in one pulse, or the pavement return jumped a centimeter between scans.

Field execution: use the cloud as dense evidence, then identify breaklines with coded shots, scans plus photos, or a prism on the actual lip. Do not deliver a mesh and call the edge “good because there are 800 points.” Calibration and data-input errors still apply: wrong scanner height, a dirty window, and a trajectory that was never checked on a known traverse.

Levels and total stations still lose to physics

Digital / optical levels. Collimation error is not canceled unless backsight and foresight distances are balanced. A two-peg test that was last year’s sticker is a knowledge-J calibration miss. Refraction over hot pavement and turning points that punch into soft AC are field error sources, not office mysteries. A GNSS-derived ortho on one bench and a level run on another, without a H = h − N audit, is two vertical datums wearing one job name.

Total stations. Prism constant (0 versus −30 mm) is a data-input classic. Atmospheric ppm from temperature and pressure is a scale error on long EDMs. Dual-axis compensators have a tilt range; on a steep bank they silently quit. EDM to a traffic cone, a chain-link knot, or a wet target is not a prism. Short vises of a construction laser or a passing scanner can look like a return. Calibration: collimation, trunnion axis, and EDM against a known baseline—not “it was fine in March.”

Worked method choice, downtown Los Angeles (N ≈ −35.1 m): a building-face ALTA-style topo. GNSS between towers is a multipath factory—do not hold those ellipsoid heights as NAVD88 even after H = h − N. Scan the facade for dense detail, then still shoot the curb lip and dripline as breaklines because mixed pixels will round them. If the owner wanted a UAS fly-through of the courtyard oaks, budget ground shots in the voids. That set of refusals is III.12 executed, not a sales pitch that every sensor “sees everything.”