15.3 Historic Precision, QA, and Technology Limits

Domain IV still has four analysis activities that are not boundary-law ranking. IV.2 is to evaluate accuracies and precisions of historic documents and maps versus measured survey data. IV.3 is to evaluate relevance and spatial relationships of maps and measured survey data (e.g., GIS data, field data, metadata). IV.7 is to perform quality assurance and quality control. IV.9 is to identify limitations of technologies for use in survey practice. Knowledge FF names historical accuracies versus current precision (measuring by chaining, EDM, GPS). Knowledge Y names digital terrain models. Knowledge Z names point clouds (as-built, reduction, management, analysis, extraction, terrain classification). Independent OpenExamPrep teaching here is office analysis. Chapter 11 already covered field sensor limits on the occupation.

The January 2025 test plan and the Prometric candidate information bulletin do not publish a Board numeric accuracy specification, a required PLS closure ratio, or an official exam item count. Do not invent a "BPELSG 0.07 foot" rule. Use the precision the historic method could support, the metadata the GIS actually carries, and the QA the claimed standard (if any) requires.

Historic precision is not a defect in the monument

A 1912 Record of Survey chained 20 chains along a section line and wrote 1,320 ft. A 2026 GNSS inverse of the same recovered monuments is 1,320.18 ft. Knowledge FF asks you to evaluate those two numbers as products of different methods, not to "correct" 1912 to hundredths and call the pipe off.

Chaining—Gunter's chain of 66 feet and 100 links, later steel tapes—carried slope, temperature, sag, keeping the tape on line, and standardization error. Ordinary land-survey tapes are often discussed in textbooks near 1:5,000 to 1:10,000 relative precision. That range is classroom order-of-magnitude teaching. It is not a Board tolerance, not a General Land Office specification you should quote as California law, and not a reason to reject a called-for monument. A chained course written to 0.01 ft on a mid-century plat is often a drafting habit. It does not mean the crew measured to 0.01 ft.

Electronic distance measurement (EDM) and total-station distances add a manufacturer specification of the form ±(a mm + b ppm), atmospheric parts per million, prism constant, and centering. They are typically far tighter than a 1912 chain on the same line. They still do not outrank a called-for monument under the hierarchy in Chapter 12 and Code of Civil Procedure section 2077. Precision of the new measurement is not a hierarchy of evidence.

GPS/GNSS reports a coordinate in a stated realization and epoch. A 0.01 m RMS on the collector is not FGDC network accuracy. Multipath can bias a solution that still displays as fixed. A GNSS-derived CCS83 inverse compared to a chained 1912 line is a method comparison. If the monuments agree within the historic noise, the line is not wrong because GNSS can see 0.18 ft.

Significant figures (knowledge DD) close the loop. Do not average 1,320 ft (1912) with 1,320.18 ft (2026) and record 1,320.09 ft as the record distance. Report both, state the methods, and let the monuments control unless a different legal theory applies.

GIS spatial relationships are metadata problems

A county parcel polygon that sits 8 feet off your real-time kinematic occupied iron is not, by itself, an encroachment. Geographic information system (GIS) data are relevant only after you read metadata: source scale, datum, realization, epoch, accuracy statement, and whether the fabric was adjusted to monuments or to assessor bearings.

A U.S. Geological Survey 1:24,000 quadrangle compiled to National Map Accuracy Standards (NMAS) is commonly cited as 90 percent of well-defined points within 40 feet. Overlaying that raster on a 2-centimeter GNSS survey and measuring a "12-foot GIS encroachment" is a spatial-relationship error. The GIS layer never claimed 0.04 ft. NMAS is a published federal map-accuracy standard. It is not a BPELSG PLS passing score.

Ask, in order:

1. Same datum, realization, and epoch as the field survey? A NAD27 parcel fabric on a NAD83(2011) epoch 2010.00 survey is a transformation problem, not a fence problem.
2. Same projection and units? CCS27 Zone 7 feet versus CCS83 Zone 5 meters is a classic Los Angeles trap from Chapter 6.
3. What accuracy does the metadata actually state? If none, treat the layer as a pointer to records, not as coordinates.
4. Was the layer snapped to unsurveyed centerlines? Topology that looks clean can be geometrically false.

IV.3 is evaluate relevance. A well-documented GIS control layer in CCS83 with FGDC-style metadata may be a search tool. It still yields to field monuments and the record.

QA/QC is a procedure, not a software flag

Quality assurance is the plan made before the crew leaves: weights, independent ties, hold-out checks, datum and epoch notes, and a decision about whether any FGDC or FGCS accuracy will be claimed. Quality control is the evidence that the plan happened: residual review from section 15.1, independent inverse checks, a second occupation, a taped baseline, a level-loop closure, point-cloud registration on known spheres, and a DTM check shot in a canopy void.

When an FGDC or FGCS accuracy is claimed, PRC 8813.2 and 8815.4 already list the written packet: equipment, procedures, closures or residuals, adjustment, and a control diagram. QA that cannot produce that packet should not claim that accuracy. QA that can produce it still should not pretend the Board published a PLS-wide numeric cut in the 2025 test plan.

Hold-out checks belong in the analysis: a distance you measured and did not put in the least-squares run, compared afterward to the adjusted inverse. If the hold-out is 0.08 m and the error ellipse predicted 0.02 m at 95 percent, the model or the weights are lying. That is IV.7 using IV.6. It is also how you catch the common-mode bias section 15.1 described.

Technology limits that survive into the office

Chapter 11 named field limits. Analysis adds failure modes that appear only after import.

Digital terrain models (knowledge Y). A DTM interpolates a surface. Photogrammetry under closed canopy interpolates invented ground. Hydrologic flow paths, pad certifications, and earthwork from that surface are wrong even if the contours look smooth. QA is check shots on actual ground, void polygons, and a refusal to contour across classified vegetation.

Point clouds (knowledge Z). As-built extraction, pipe invert, and curb face are classification and identification problems. Mixed pixels smear a lip. Terrain classification that labels a truck as ground puts a 1.5 m bump in the DTM. Density in points per square meter is not accuracy. Reduction and management include scan-origin metadata, registration residuals, and a coded breakline that a person accepted.

GNSS post-processing will always emit a coordinate. Limitations that belong in IV.9 analysis: wrong antenna model, mixed epochs, GEOID18 on the wrong ellipsoid height, and a real-time network coordinate that was never independently tied when PRC 8813.1(a)(2) required two connections to establish CCS83.

Least-squares software will always emit an ellipse. Limitation: bias from section 15.1, and over-weighting GNSS relative to a taped historic line you should not have adjusted as if it were a 2 mm observation.

The analysis product is a decision: this historic map is usable as a search diagram; this GIS layer is not a boundary; this cloud is usable for planimetric edges after breakline identification; this adjustment is usable after the hold-out check; this technology is the wrong tool for this line. That cluster is IV.2, IV.3, IV.7, and IV.9 together.