Geospatial Engineering and Surveying Technology Case Studies
Technical5 Case StudiesUpdated Sep 2026
Geospatial work rarely fails loudly. More often a control network drifts a few centimeters, a coordinate transformation gets applied twice, or a crew skips a check that seemed too minor to matter. The five cases below come from survey, mapping, and positioning teams, and each turns on a decision that looked routine when it was made. Read them with the project records open beside you. The tolerance sheets, field logs, and maintenance history usually tell the story before the people involved do, so start there and then compare your reading with the discussion questions.
Case Study 1
Boresight drift traced to a cracked lidar mount
Halvorsen Mobile Mapping Services, a contractor that operates three vehicle-mounted lidar rigs for municipal asset inventories and corridor surveys.
The situation
Halvorsen runs three vehicle-mounted mapping rigs that collect point clouds for road asset inventories. Each rig carries a lidar unit, a navigation-grade IMU, and a dual-frequency GNSS receiver, all bolted to a welded aluminum crossbar on the roof. During the second month of a county roads contract, the quality group noticed that point clouds from one rig showed a consistent lateral offset of about 4 centimeters on repeated passes over the same bridge deck. The other two rigs were clean. Boresight calibrations had been performed on schedule, and the first check showed nothing unusual. The offset appeared only on passes taken at highway speed and grew on a stretch of rough chip seal.
The challenge
The client wanted a root cause and a corrective design, not another recalibration. The rig in question had been in service for under a year, and the roof crossbar had already been repaired once. Engineering had to decide whether the fault lay in the mount geometry, the vibration environment, the sensor firmware, or the maintenance routine. Each explanation predicted a different pattern in the logs, so the team could not simply pick the most convenient one.
Analysis
Teams grounded in Geospatial Engineering and Surveying Technology start by asking what the mount had to do, not what it looked like. The constraint was that the lidar and IMU must hold their relative pose within a fraction of a degree across the full speed range and temperature swing. The inertial log showed that the lateral error correlated with vertical acceleration peaks above a fixed threshold, and that the peaks clustered on specific road segments. That points to a resonance or compliance problem in the mount, not a software bias. A spectral check confirmed energy near the crossbar's first bending mode, a mode the original design review had never examined.
The maintenance record changed the picture. The crossbar had been welded after a cracked joint and the weld bead was ground flush to save time at service. Nobody had recorded the replacement of the two rubber isolators under the IMU bracket, which the vendor's manual listed on a fixed interval. Tolerance allows drift of no more than 3 centimeters at the 95 percent level, so the rig failed its own acceptance test on the worst segments. Three corrective options were considered: a stiffer gusset-reinforced mount, a relocated IMU on a dedicated sub-frame, or a tuned damper with a revised service interval. Each carried a different cost and residual risk.
Actions taken
- Pull the full vibration history for all three rigs and compare crossbar weld locations against inspection photos before any further calibration is run.
- Replace the cracked crossbar with a single machined bracket and log the isolator interval in the fleet maintenance system so the next service cannot skip it.
- Run a modal test on the retrofitted mount at highway speed and confirm the first bending mode sits well outside the vehicle's excitation band before returning the rig to contract work.
- Defer redesign of the other two rigs until their drift is measured against the 3 centimeter tolerance on the same chip seal stretch.
Results
The retrofitted rig ran a full month of contract passes without a lateral offset above 1 centimeter on the chip seal stretch, well inside the tolerance. The modal test moved the crossbar's first bending mode clear of the vehicle's excitation band. Both remaining rigs were measured, and both showed the same weld pattern, so the fleet-wide fix went ahead. The client accepted the corrected point clouds for the bridge deck without a re-survey, and the isolator interval is now a hard stop in the service system, not a manual note.
Questions for discussion
- Which of the three candidate causes would have been easiest to rule out with the logs the team already had, and what would ruling it out have required?
- How should a contractor weigh a stiffer mount that adds mass against a relocated IMU that changes the calibration procedure?
- What inspection interval would you set for welded structural members on a vehicle that spends most of its days on rough pavement?
Case Study 2
Orthomosaic throughput falls as tile misregistration rises
Brightwater Aerial Mapping, a 40-person production shop that turns drone imagery into orthomosaics and elevation models for civil and agricultural clients.
The situation
Brightwater processes drone imagery in batches. Each flight is ingested, tie points are matched, the block is adjusted against ground control, and the result is cut into tiles for delivery. For most of the year the shop averaged about 14 blocks per week. Over the last quarter throughput dropped to 9 blocks, while the share of tiles rejected at the check climbed from 4 percent to 11 percent. The rejected tiles shared a pattern: edges that did not line up with neighbors, usually near the boundaries between flight blocks. The shop runs two processing servers and one shared storage array, and the production manager, Dana, wanted to know where the time was going before approving a third server.
The challenge
Dana had a purchase request on her desk for a third server, which would cost a month of setup. Two of the bottleneck candidates were cheap to change: how control points were imported and how image metadata was checked. The other two were harder: the storage array's behavior under parallel jobs, and the manual review step every tile passed through. The team had to rank all of them by how much each contributed to lost throughput and rejected tiles, using measurements they could repeat next month.
Analysis
Mapping the last forty blocks showed where hours went. Ingest and tie point matching were stable across blocks. Block adjustment and dense matching varied widely with the number of control points imported per job. The seam mismatch tolerance in the contract is 2 pixels at the ground sample distance, setting the real constraint: any step that shifts a tile past that limit shows up as a defect. Variation came from three sources, image metadata with missing or conflicting altitude values, control points entered in mixed coordinate order by different operators, and a storage array that slowed when more than three jobs wrote at once.
Ranking the sources needed a common scale, so the team counted lost hours and rejected tiles together. Rework from failed seams accounted for most of the rejects, and most of that fell on blocks processed while the storage array was under heavy load. The array had logged elevated write latency during the same weeks, and a firmware update had been deferred twice. The metadata problem was smaller in hours but cheap to fix with a pre-ingest check. Manual review was the slowest step, yet it was also the last defense, so cutting it without first fixing its inputs would have hidden the defects rather than removed them.
Actions taken
- Add a pre-ingest check that holds any image batch with missing or conflicting altitude fields until an operator resolves them, before tie point matching begins.
- Move control point import to a single template with fixed coordinate order, and accept a batch only when the seam check on five test blocks stays under 2 pixels.
- Schedule the storage firmware update during a planned gap and cap parallel write jobs at three until the array's write latency stays flat across a full week of production.
- Keep manual review in place for the first month, but log each rejected tile by cause so that the next priority list is built from counts instead of impressions.
Results
The metadata check caught 14 batches in its first two weeks, and each one carried the altitude problem. Control point imports stopped producing mixed coordinate orders once the template was enforced, and the test blocks held under the 2 pixel limit on every run. After the firmware update and the three-job cap, weekly throughput climbed back to 12 blocks, and the tile rejection rate fell to 5 percent. The third server stayed off the purchase list for the time being.
Questions for discussion
- If the storage firmware had been current from the start, how much of the rejection rate would have remained, and how would you test that without disrupting production?
- Should a shop accept a slower manual review step as a permanent cost when the upstream fixes are only partly known?
- What single measure would you place on the first page of the weekly report so the team does not chase throughput at the expense of seam quality?
Case Study 3
Drone corridor survey near energized lines held for review
Ostrander Utility Surveys, a small firm that performs corridor mapping and pole inventories for two regional power cooperatives.
The situation
Ostrander was asked to map a 22 kilometer transmission corridor with a multirotor drone carrying a lidar payload, followed by a ground survey of the access roads. The corridor crosses a river valley, two county roads, and the edge of a small regional airport's approach path. The client's contract called for delivery in the fourth month, and the field team had already flown a test segment without incident. Before the main flights, the safety lead, Ruth, was asked to sign off on the plan. She found that the test flight log had no entry for the airspace authorization, and the lidar payload's calibration certificate was missing from the project folder.
The challenge
The schedule was tight but not impossible. The question was which hazards were covered by existing controls and which had no evidence behind them. Ruth also had to decide whether the ground crew's work near energized conductors needed a separate permit and a line-clearance check from the cooperative. Three people wanted the flights to go ahead on the existing plan, and two wanted a full stop. She needed a review that could be defended to the client and to the cooperative's own safety staff.
Analysis
The hazards began with fixed constraints. The corridor required a flight altitude that kept the drone clear of the conductor sag envelope with a margin set by the cooperative, which meant the planned altitude band sat close to the minimum clearance at the river crossing. The operating data from the test segment showed wind gusts above the platform's rated limit during two of its six passes, and the telemetry had dropped link for a few seconds near the valley wall. Both pointed to an envelope that had not been checked against the terrain. A review that separates airspace from wind and link data will miss the interaction.
The airspace authorization existed, but it was filed against the wrong block and did not clearly cover the valley segment. The lidar calibration certificate sat on the vendor's portal but had never been uploaded, and the maintenance log for the payload showed a bracket inspection due before the first pass. Under the cooperative's contractor rules, ground crews working within a set distance of energized equipment need a written work plan and a clearance from the operator. None of those were in the folder. Control points were also set late, two of four after flights began, which the mapping standard does not allow.
- 1Airspace authorization refiledMust name the valley block, altitude band, and crossing.
- 2Calibration and bracket recordsCertificate uploaded and bracket inspection signed before any pass.
- 3Wind and link thresholds setPer-segment limits written into the pre-flight checklist and signed.
- 4Crossing work plan approvedCooperative line operator signs clearance; named stop authority on site.
- 5Control points and releaseMissing monuments established, then safety lead releases each flight block.
Actions taken
- Stop all flights until the airspace authorization is refiled against the exact valley block, with the altitude band and the river crossing shown on the approved map.
- Obtain the payload calibration certificate and the bracket inspection record, then hold a pre-flight review that signs off wind limits and link loss thresholds for each segment.
- Write a separate work plan for the ground crew at the crossing, with a clearance from the cooperative's line operator and a named person who can stop work at any time.
- Establish the missing control points before any further flights, and add a documentation gate to the project plan so each flight cannot start without its paperwork attached.
Results
The flights resumed in the fifth week after the authorization was refiled and the wind and link limits were written into the pre-flight checklist. Two of the four control points were established on the first day back, and the remaining two followed within a week. The cooperative's line operator reviewed the crossing plan and asked for one change, a hard stop when gusts approached the rated limit, which the crew adopted. The client accepted the corridor dataset, though the delivery date moved by three weeks. Ruth kept the documentation gate in place for the rest of the contract.
Questions for discussion
- Which of the missing documents should have blocked the first test flight, and which could reasonably have waited for the main survey?
- How much authority should a field safety lead have to stop work when the client's schedule is firm and the cooperative has already approved the plan?
- When a test segment reveals wind or link problems, what evidence would you need before accepting the existing flight envelope as adequate?
Case Study 4
Datum mismatch threatens bridge layout schedule
Calloway Infrastructure Partners, a mid-sized engineering firm delivering highway and bridge projects across three state corridors.
The situation
Calloway was engaged to deliver survey control and as-built geometry for a river bridge replacement. The contract called for a control network of 38 monuments tied to the project's local grid, with the layout package for the piers due before the foundation crew mobilized. The GNSS receivers came on a rental agreement with a fixed return date, and the design team's geometry came from a model that carried its own coordinate reference. The schedule had two weeks of float, which the project manager, Marcus, had already committed to the contractor for pier layout.
The challenge
When the first monument ties were computed, residuals on the far abutment came out near 120 millimeters, well outside the layout tolerance. The design team said the model was correct. The survey team said the field data was correct. Both were partly right, which made the argument slow. Marcus had to decide whether to hold the pier layout, recompute the network under a different datum, or ask the contractor to accept a provisional layout with a known offset.
Analysis
The layout tolerance for the piers was 15 millimeters horizontally, so a residual of 120 millimeters was not a rounding issue. The coordinate transformation between the model's grid and the local datum had been set up with a scale factor that the design team had computed for a different reference point. The operating data made the cause visible. Residuals grew from the origin of the control network outward, which is the signature of a scale error, not a bad monument. GNSS baselines and observation logs showed no receiver fault. The design constraint was clear: one datum, one scale factor, and a documented chain from the model to the field.
Three recovery plans were costed. Holding pier layout while the network was recomputed would consume the two weeks of float and add standby charges for the layout crew, 46,000 dollars a week at current rates. A provisional layout with the known offset would protect the date, but it needs written acceptance from the owner and the contractor, and each pier would later need re-marking. A third option, recomputing with the corrected scale and extending the receiver rental by two weeks, added 9,000 dollars over budget and kept the error off the record. It was the only plan that protected the float without asking anyone to accept an error.
- Week oneDesign team sets the model transformation using a scale factor from another reference point.
- Week twoFirst monument ties show 120 millimeter residuals at the far abutment.
- Early in week threeBoth teams defend their data, and the layout hold begins.
- Mid week threeRecovery options costed; provisional layout offered and declined by owner.
- End of week sixRecomputed network closes within 9 millimeters and pier layout is reissued.
Actions taken
- Recompute the control network under the corrected scale factor and check residuals at all 38 monuments against the 15 millimeter layout tolerance before anything else is released.
- Extend the GNSS receiver rental by two weeks now, while the vendor still has units available, rather than waiting for a return date that would force a second extension.
- Put the corrected datum definition in a single transformation record signed by both the design lead and the survey lead, and attach it to every drawing that carries model coordinates.
- Hold a joint coordination call each morning until the pier layout is reissued, with the contractor's layout foreman present so that the field and office work to one set of numbers.
Results
The recomputed network closed within 9 millimeters at every monument, inside the layout tolerance. The pier layout was reissued in the third week after the datum record was signed, which used up most but not all of the float. The contractor's crew lost one day of standby, and the rental extension came in close to the estimate. The as-built package was delivered on the revised date, and the transformation record was adopted as the project's standard reference for the remaining bridge work. The disagreement between design and survey was settled by a document rather than by argument.
Questions for discussion
- Which of the three recovery options would you have chosen if the owner had refused any provisional layout, and what would the schedule have looked like?
- How should a project record who owns a coordinate transformation when design and survey teams each hold a version of it?
- What early check would have caught the scale factor before the first monument ties were computed?
Case Study 5
Change detection instead of full recapture
Pell Creek County Geographic Information Office, a nine-person team that maintains parcel, road, and orthophoto layers for a rural county of 1,400 square kilometers.
The situation
The office has the county flown every year for an orthophoto layer that feeds assessors, planners, and the road crew. The flight is contracted, processing runs on two local servers, and raw imagery is retained for seven years under records policy. Last year's campaign consumed roughly 31,000 kilowatt-hours of processing energy and filled about 4 terabytes of storage per cycle, and the contractor's invoice was dominated by flight hours. The planning director, Sofia, wanted the cost and carbon footprint of the annual cycle cut without losing the accuracy the assessors rely on for parcel reviews.
The challenge
Two options were on the table. The first kept the annual full recapture but switched to a smaller aircraft and a shorter processing chain. The second compared each new survey against the existing layer, recaptured only parcels where change exceeded a threshold, and cut storage at the same time. Both had to keep horizontal accuracy within the assessor's tolerance of 0.15 meters at the 95 percent level, and neither could strain the staff who maintain the processing servers.
Analysis
Accuracy sets the constraint. The assessor's tolerance applies to parcel corners, not to the county as a whole, so the question is whether change detection can find the parcels where imagery has shifted enough to matter. Two cycles of operating data showed that most of the county had changed by less than the tolerance in a given year, with changes concentrated near new subdivisions and road realignments. Those are the areas where a change mask built from imagery and parcel boundaries is most reliable. A test on one township showed that the mask flagged the right parcels with few misses when the base layer was under two years old.
Lifecycle measures went beyond the flight bill. Processing energy scaled with the number of tiles rebuilt, so a campaign covering a fraction of the county would cut the server hours in proportion, and the two servers would stay on their normal replacement schedule. Maintenance records showed the storage array near capacity for three cycles, and most retained raw imagery was never reopened after review. Retaining only the change masks and the final mosaics, under the records policy, would reduce storage by a large share. The remaining risk was operational: a smaller staff must hand-check the flagged parcels, and that work must fit current staffing.
Actions taken
- Build the change mask from the last two cycles and the parcel layer, then test it against a manually checked sample of 400 parcels in two townships before committing the approach.
- Set the acceptance threshold at a 95 percent detection rate for changed parcels above the 0.15 meter tolerance, and reject any mask that falls short.
- Switch the raw imagery retention to masks and final mosaics, and document the disposal dates in the records schedule so the storage reduction is auditable.
- Keep one full flight every third cycle as a control, so the change method is checked against complete coverage and the staff keep their hands on the whole county.
Results
The pilot flagged about 11 percent of parcels in the test townships as changed, and the manual check confirmed 96 percent of those flags. Processing hours for the cycle dropped by roughly 60 percent, and the storage footprint fell enough that the array no longer ran near capacity. The first cycle under the new approach stayed within the assessor's accuracy tolerance on every sampled corner. Sofia accepted the method with the full-flight control kept in the schedule, and the flight contract was rewritten around smaller, targeted missions.
Questions for discussion
- What would a change mask need to detect reliably before you would trust it on parcels with no recent field verification?
- How should an office weigh a cheaper method that sometimes misses small changes against a full recapture that catches everything but costs more each year?
- Which lifecycle figure would you report first to a county board that cares most about energy and least about accuracy?
Learning Resources
Technical CASE STUDIES
China, India, United States, Indonesia, Pakistan, Nigeria, Brazil, Bangladesh, Russia, Japan, Mexico, Philippines, Vietnam, Egypt, Turkey, Germany, United Kingdom, France, Italy, Canada, South Korea, Australia, Spain.