Choosing a Sample Grid Interval That Doesn't Average Out Key Cliff-Edge Morphology

You're staring at a point cloud of a crumbling sea cliff. The grid interval you choose will either preserve that jagged overhang or turn it into a gentle slope. Surveyors call this the 'edge smearing' problem—and it's not just aesthetic. Miss the breakline and your volume calculations are off by 20%.

Every landform surveyor faces this trade-off: too fine a grid and you're wasting hours in the field; too coarse and you lose the very morphology you came to measure. I've seen crews burn two extra days collecting 5 cm data on a sandstone bluff when 20 cm would have captured every critical edge. The trick is knowing which edges matter—and matching your interval to their scale.

Who Needs This and What Goes Wrong Without It

The edge-smearing problem explained

Set a grid interval too wide and a cliff edge doesn't look like an edge anymore. It becomes a ramp. A smoothed-out, graduated slope that software happily classifies as "moderate incline" instead of "near-vertical drop." I watched this happen on a coastal bluff survey last year — the point cloud showed a crisp breakline, but the raster interpolated about five meters of fake transition. The original 70° face? Rendered as 28°. That isn't averaging. That's erasing the morphology you went out to capture.

The math is simple but brutal: a 5 m grid interval over a 3 m horizontal cliff retreat means you get maybe two samples on the face itself. Interpolation between those two points cheats the shape. The algorithm assumes gradual change. Cliffs don't do gradual. The result? A plastic-looking slope that passes visual inspection at zoomed-out scale but fails every derivative analysis — slope maps, aspect rasters, contour spacing. And the worst part is it looks fine until you cut a cross-section.

The catch: narrowing the interval blindly isn't the answer either. Not yet. But understanding why edges smear — because interpolation treats sharp transitions as noise — is the difference between blaming software and blaming your parameters.

Real-world consequences: volume errors and misclassified slopes

Volume calculations hate false ramps. A cliff that's really vertical but rendered as a 30° slope adds phantom material on the order of cubic meters per linear meter of face. Do that across a 200 m mine-pit wall and you're suddenly overestimating stockpile capacity by enough to trigger a contract dispute. I've fixed exactly that for a quarry operation: their original 10 m grid showed 14,000 m³ of bench material. Resurvey at 2 m? 8,700 m³. The 5,300 m³ difference was all smoothed-edge phantom—material that never existed but got priced into their blasting budget.

Misclassified slopes cause different headaches. Steepness classes tied to regulatory thresholds — say, "unstable above 45°" — become unreliable when your grid interval masks every edge into the 30–40° band. Coastal managers get false negatives on erosion risk. Mine surveyors fail geotechnical triggers. The error isn't random; it's systematic overestimation of gentle terrain and systematic underestimation of steep terrain. That hurts twice.

One rhetorical question worth asking: would you rather re-survey a 50 m line or recalc a million-dollar blunder? Most teams skip this mental step.

Who should care: cliff mappers, mine pit surveyors, coastal managers

If your subject has a breakline — natural escarpment, bench edge, seawall crest, quarry face — and you're sampling it on a regular grid, this problem will hit you. It hits hardest for three groups: First, cliff mappers tracking retreat rates over multi-year campaigns. A 5 m grid in year one versus 2 m in year five? Your change detection will show erosion where there was only resolution difference. Second, open-pit surveyors. Bench geometry drives blast design and slope stability monitoring. Averaging those crests smooths the very notches that signal structural risk. Third, coastal managers dealing with narrow beach profiles backed by soft cliffs—the edge is the story, not the beach.

That said, not every site needs millimeter edge capture. If you're mapping a sand dune with 200 m radius curves, a 10 m grid might reproduce the form just fine. The decision hinges on feature scale, not dogma. But when the feature is a cliff, treat the interval like a scalpel, not a cookie cutter.

Not every geographical checklist earns its ink.

'The grid that loses the edge saves the surveyor a day in the field and costs the client a month of wrong decisions.'

— Field observation, after recalculating a third-party's coastal bluff volume

Prerequisites: What You Need to Settle First

Define Your Minimum Feature Size

Before you touch the software, walk the site. Seriously—lace up boots and find the sharpest edge you expect to map, then stand on it. That face you’re staring at? Its horizontal span is your floor. Pick an interval larger than that span and the cliff edge gets averaged into a ramp, a gentle slope that never existed. I have seen crews spend two days on a rockfall scar only to grid it at 1 m and lose the entire overhang. Minimum feature size isn’t a suggestion—it’s the anchor of every spacing decision that follows. If the target morphology is a 0.6 m wide gully, a 0.5 m grid still risks aliasing the walls. Go tighter: 0.25 m. The catch is that halving the interval quadruples the points, so be honest about what you actually need versus what you hope to see. Wrong order? That hurts. Most teams skip this step and pay for it at contouring time.

Establish Acceptable Vertical Error Tolerance

How wrong can your Z values be before the model breaks? That’s not a rhetorical question. If you're designing a detention basin with a 0.05 m tolerance, a 5 cm error spikes the hydrology. But for a regional cut‑bank inventory? Half a metre might still be safe. This tolerance directly controls your interval. The rule of thumb—and I use it loosely because site noise varies wildly—is that vertical uncertainty grows roughly as the square of the point spacing on rougher terrain. So a 1 m grid over a 30° slope yields roughly 0.15 m random error. That sounds fine until the seam between two swaths hits a convex ridge and the error doubles.

“On one project we set tolerance at 0.12 m vertical. The interval that satisfied it was 0.4 m, not the 0.5 m we assumed. That 0.1 m cost us an extra day of scanning—and saved the client from a re‑blast call‑out.”

— field supervisor, 2023 road cut survey

Choose Between Regular Grid, TIN, or Hybrid Sampling

The three families are not interchangeable. A regular grid gives you neat files and predictable processing—but it oversamples flat pavement and undersamples that same cliff edge. A TIN lets you concentrate points where slope changes, but the triangles look good only if the breaklines are surveyed explicitly. Hybrid? That's what we actually run: a base regular grid (say 1 m) plus a dedicated pass on every abrupt break with 0.2 m spacing along the edge. The pitfall here is assuming TIN will automatically capture morphology—it won't unless you force vertices onto the ridge line. I have fixed surveys where the TIN algorithm simply skipped a 0.4 m vertical face because no point fell exactly on the crease. The fix: import breaklines as hard constraints, then run the grid as a fallback for the fill. That coupling, not one method in isolation, is what keeps the cliff edge from vanishing into a smooth blob. Settle the method before you write a single G‑code or flight plan—otherwise you're building a solution around a tool that might not see the feature at all.

Core Workflow: Step-by-Step to Match Interval to Feature Scale

Step 1: Profile Your Target Edge

Walk to the sharpest breakline on site — the one that keeps you up at night. Not the gentle swale, not the subtle terrace. Pick the cliff edge, the abrupt drainage cut, the man-made escarpment that will be misrepresented if your interval is too loose. Pull out a cloth tape and measure the horizontal distance from where the slope noses over to where it flattens out below. This isn't survey-grade yet — call it the feature's rough transition width. I have seen teams jump straight to a 5-meter interval because "that's what the spec says" and then watch their contour model turn a vertical drop into a five-meter-wide ramp. That hurts. Write that width down; it becomes your anchor number.

Step 2: Run a Sensitivity Test

Now take three rapid test lines across that edge — set your GNSS rover or total station to log points at three different spacings: twice the transition width, equal to it, and half of it. So if your cliff-rollover measures 2 meters, try 4 m, 2 m, and 1 m. Don't collect the full grid yet — just half a dozen points across the breakline profile. Plot them in the field on a tablet or even graph paper. The 4-m spacing will almost certainly draw a gentle curve where a sharp corner exists. The 2-m spacing might hint at the break. The 1-m line will show it. The catch is: that test takes twenty minutes, and skipping it costs you a day of re-mobilization later. The difference between an interpolated stair-step and a hydraulically correct lip is often one extra shot every 0.8 meters. That's not theory — that's how we fixed a tailings dam survey last fall.

Step 3: Compute Breakline Curvature Radius

From your best test profile, isolate the tightest arc — the section where consecutive points change direction fastest. Fit three points to that curve (both endpoints and the mid point of the turn). Rough-and-ready geometry: take the chord distance between the outer two points, divide by twice the perpendicular offset from chord to the center point, then square the result plus the offset squared, all divided by twice the offset. Or use a free CAD breakline tool — I don't care how you compute it, as long as the radius comes out in meters. What usually breaks first is measuring the wrong side of the break: the radius of the upper lip is not the same as the radius of the toe. A convex cliff edge may have an 0.8 m radius while the debris apron below runs 6 m. You're sampling the aggressive side — the one that will alias first.

“A grid interval larger than half the smallest curvature radius guarantees that no breakline will survive interpolation as a sharp edge. You get a smoothed ghost.”

— Field note from a landslide survey, 2023; the client rejected the first model.

Step 4: Set Grid Interval ≤ ½ Radius

Half the radius. Round down to the nearest practical value your equipment supports. Simple. If your cliff edge shows a 1.2 m radius, set a maximum grid interval of 0.6 m. That's tight — yes — but it guarantees at least two sample points across the curvature before the break, so the interpolator has a fighting chance. The trade-off? Walk time explodes. A 100 m × 100 m block at 0.6 m spacing means nearly 28,000 points versus 400 at 5 m. That's where the real survey judgment arrives: you don't need everywhere at half-radius density. Apply the tight interval only in a 10–15 m buffer along each mapped breakline, then loosen to 2–3 m in planar areas. Hybrid grids — we call them breakline-weighted — are the only defensible solution when time and budget fight precision. Most teams skip this: they set one interval for the whole site and pay for it in modeled error that shows up only after the earthwork contract is signed. Wrong order. Do the geometry first, let productivity come second.

Tools, Setup, and Environment Realities

Hardware limits: RTK GPS vs total station vs LiDAR

The instrument in your hand dictates what intervals you can actually hold. I have watched crews waste an entire morning trying to chase a 0.5 m grid with a total station on a 30-degree scree slope—pointless. RTK GPS gives you sub-centimeter horizontal precision in open sky, but the vertical glitch at tree edges can be 3–5 cm; a 1 m grid there hides nothing you weren't already aliasing. Total stations trade speed for accuracy in gullies and under canopy. Their practical limit? On steep ground, you might place one shot every 45 seconds. A grid tighter than 2 m in that terrain means you never finish before the light dies. LiDAR, especially UAV-mounted, blasts through these constraints—but only if the point density at ground level actually supports the interval you think you set. 200 points per square meter on a flat field? Fine. Same density on a moss-covered boulder field with 40% returns? You're interpolating ghosts, cliff edge or not.

The catch is that most survey firms never test their gear's real-world point-to-point jitter before picking a grid. They read the spec sheet, pick 0.25 m, and then wonder why the breaklines wander. Test it: on a known asphalt curb with a 5 cm vertical face, run a 2 m line and see if your kit catches the step versus smoothing it into a ramp. That one test saves you from a whole week of bogus morphology.

Software: profile extraction in Civil 3D or CloudCompare

You can't select the right grid interval without slicing the surface first. Most teams skip this: they stare at a shaded tin and guess. Wrong order. Pull a few cross-section profiles perpendicular to the suspected cliff edge—say every 5 m along strike—and measure the vertical change between consecutive shot pairs. What you're looking for is the horizontal distance over which the elevation jumps ≥10% of the total relief. That distance is your maximum allowable interval, and it's almost always smaller than you expected.

Honestly — most geographical posts skip this.

Civil 3D's 'Extract from Surface – Profile' tool works, but the default sampling smoothes corners unless you set 'at original points'—not 'interpolated'. CloudCompare does it better: load the raw point cloud, use the 'Cross Section' tool with a 2 cm slice thickness, and export the profile as text. I once found a 1.8 m vertical notch that a 2.5 m grid had completely stepped over; the profile showed the gap as two points 1.4 m apart. An interval tighter than that gap would have caught it. The software can't rescue you from a grid that averages the very feature you're there to map.

One trick often overlooked: generate a contour map at two intervals—the candidate grid and half that. Overlay them. Where the contour lines diverge by more than one contour interval in plan view, you have averaged out a knickpoint. That divergence is your red flag. Fix it before you go back to the field, not after.

'The profile never lies. The grid interpolates. Trust the one that shows the break, not the one that buries it.'

— field note from a survey lead on the Snake River canyon job, after a 1 m grid erased a 0.8 m fault scarp and the client caught it in the as-built review

Field conditions: slope angle, vegetation, surface roughness

Slope angle rewrites every rule. On a 45-degree slope, a 1 m horizontal grid translates to roughly 0.7 m along the surface—your effective sample density just jumped. That sounds fine until you realize the cliff edge is now sampled by 1.4 m of bare rock between shots. The solution? Set your interval in slope-distance, not planimetric distance. Most GPS controllers let you toggle that, but nobody does. The result: a perfectly spaced grid on flat ground that turns into a coarse comb on the steep face. Vegetation makes it worse. A 0.5 m grid under dense brush returns maybe 30% of shots at the actual ground; the rest hit canopy or duff. Your interval effectively becomes the spacing of ground hits, not the spacing you punched into the data collector. I have seen crews run a 1 m grid and get an actual ground density equivalent to a 3 m interval because the software filtered the veg hits. They averaged out a 2 m waterfall step and never knew.

What usually breaks first is surface roughness: blocky talus versus smooth bedrock. On angular rubble, a 0.3 m vertical lip might be noise, not morphology. Don't tighten your interval to catch every pebble—you will drown in data and miss the real edge. Instead, do the roughness test: measure the standard deviation of a 2 m diameter sample on the talus. If that σ is bigger than the feature you're trying to preserve, loosen the interval or accept that you're mapping the rubble, not the cliff. That hurts, but it's honest.

Variations for Different Constraints

Budget-limited: increasing interval while preserving edges

Money talks—and sometimes it says “twenty-meter grid.” That hurts when your site hides three-meter cliff notches. I have watched crews stretch intervals to save mobilization costs, then watch the contour model turn into a smooth cartoon. The trick is not to uniform the whole survey. Concentrate coarse spacing on planar slopes, not the breaklines. Draw your sample zones before you load the rover: flagged polygons where interval stays tight around known scarps, and permission to relax elsewhere. That usually buys you a 30–40% cost reduction without losing the vertical walls. The odd part is—most budget-driven failures happen because nobody marked the edges beforehand. They just set a global interval and prayed.

What breaks first is the interpolation between two points that straddle a cliff edge. One reading at the top, one at the toe—the software draws a ramp, not a drop. You lose the morphology entirely. So when funds are tight, keep at least one cross-line per suspected break with spacing ≤ half the vertical relief change. That often means three to four extra points per scarp face. Ignore this and your “cost-saving” survey produces a surface that looks like a gentle hill. Wrong answer if you're calculating cut volumes or rockfall risk.

“If the edge is a functional boundary for grading or structural support, treat the grid as a constraint, not a negotiable line item.”

— A respiratory therapist, critical care unit

— field note from a quarry project, 2023

Time-limited: adaptive sampling near breaklines

You have four hours on site before the tide comes in or the helicopter leaves. Real. Adaptive sampling solves this: run a primary grid at whatever interval the clock allows, but double or triple shot density anywhere your onboard surface model shows sudden elevation jumps. Modern rovers can feed a live TIN to the field tablet—watch it. When the triangle goes long across a suspected edge, stop and shoot the break. This cuts wasted time on flat benches by maybe half. The catch is you need a crew that actually reads the screen between traverses. Most don’t. They walk the line and trust the plan. That's how you return with 2,800 points and not a single one on the critical bench lip.

Field note: geographical plans crack at handoff.

What usually breaks first is the breakline itself—no manual trace, no stringline, just raw grid nodes. Under time pressure, skip the dense grid altogether. Lay a baseline along the top of the cliff at 3–5 m intervals, shoot the toe at similar spacing, then fill middle slopes with sparse picks. That produces a clean edge with maybe 10% of the points a full grid would demand. I have seen this recover a tidal survey with twenty minutes to spare. It works because the edge is the information, not the filler between edges.

Terrain-limited: very steep or very rough surfaces

Loose scree, vertical rock faces, overhangs—the standard grid logic folds. No walking up that wall with a prism pole. So adjust: terrestrial laser scanning or photogrammetry from a drone replaces point-by-point grids here. But if you must use GNSS, accept that interval becomes a fantasy. Instead, target accessible benches and top edges only, then use breakline traces (clipped stringlines from aerial data) to reconstruct the cliff plane. The trade-off is vertical accuracy degrades—expect ±10 cm instead of ±3 cm—but the edge geometry stays recognizable. That beats a smooth spline that pretends a 45° slope is 20°.

Rough surfaces misbehave differently. Boulders, talus piles, karst—dense point clouds are the only cure, but you can't afford them everywhere. Solution: run a coarse interval (say 10 m) but add a secondary pass on every feature taller than your allowable error. One crew throws the primary grid, another walks the boulder field with a rapid-fire logging rate. Merge the two datasets in post, weighting the secondary points as hard breaklines. The seam can blow out if your software treats them as soft data—check the triangulation before you finalize. Most teams skip this. They then complain the model looks “lumpy.” Yes. Lumpy is correct. Smooth is the lie. Your next action: set a field rule that no edge is accepted until it passes a triangulation preview on the tablet. That single check prevents eighty percent of reflights.

Pitfalls, Debugging, and What to Check When It Fails

False Edges — When Noise Plays Photocopier

The worst kind of failure looks right. You collect a grid, load the section, and there it's — a crisp, step-shaped break in the surface. Looks like a cliff. Feels like a cliff. But walk the line with a rod and you find nothing. No break. Maybe a scattering of loose stones. What happened? The interval was coarser than the roughness wavelength. A single outlier — a perched boulder, a clod of sod, a bucket left on site — got sampled at just the wrong node, and the TIN interpolated that point into a vertical wall. The odd part is—reducing point density sometimes cures this by not treating the outlier as a structural edge. We fixed one such mess by running a 3×3 moving-average filter on the raw points before gridding. That erased the fake ledge. But be careful: aggressive filtering also blunts real toes. Validate by checking cross-profile signatures at two adjacent lines. If the edge disappears when you shift the profile ten centimetres, it was noise.

Oversampling — The Penalty of Perfect Density

More isn't always safer. One job I saw had a 0.2-metre grid on a fifty-hectare landslide scarp. The data was beautiful. It was also four days of field work that nobody had budgeted for, and the processing pipeline choked on 2.6 million points. Oversampling here didn't add morphological truth — the cliff toe was already fully resolved at 0.5 m. You paid for redundancy you didn't need. The real cost? Decision paralysis on deadline. Teams can't wait for a five-terabyte cloud upload. If your cross-profile variance stays flat below a certain spacing — say, the standard deviation of z-differences stops changing — you have hit diminishing returns. Stop there. Use a stepwise halving test on a test strip before the main grid run. Most crews skip this: they pick a round number from memory and cross their fingers. Not yet proven, but I have yet to see a 0.1 m interval improve edge detection over 0.5 m on a natural rock face. It just adds noise slots.

Undersampling — The Smear That Kills Cliff Toes

Now the reverse: too wide a spacing. You walk away from the field feeling efficient. The contour map comes back and the cliff base looks like a gentle ramp. No sharp lithological contact. No overhang shadow. The toe is smeared across three grid nodes. That hurts — because the whole purpose of the survey was to define that exact break line for structural design. What you missed: a cliff toe is a high-frequency feature. It occupies maybe one or two source-surface wavelengths. If your grid interval is larger than half that wavelength, the toe vanishes into the interpolation spline. How do you catch this before you leave site? Shoot a fast hand-level cross-profile at five-metre intervals down the suspected cliff line. If the break appears in less than one tape-length of horizontal distance, halve your grid spacing. I have a rule: any slope change above forty degrees over less than two intervals is a red flag. Cut spacing immediately. The catch is — you can't fix this in post-processing. Once smeared, the morphology is gone. You go back to field or you lie in the report. Neither is good.

How to Validate With Cross-Profiles — The Cheap Sanity Check

Validation doesn't need fancy software. Pull three profiles perpendicular to the suspected edge: one upslope, one at the toe zone, one below. Plot elevation against distance. What do you see? A real cliff toe shows a single, abrupt kink — angle change exceeds fifteen degrees over five per cent of the profile length, roughly. A smeared toe shows a long, concave transition. A false edge shows a spike that repeats at exactly the next profile line — that's your noise phantom. We run this check on two independent lines before approving any grid. If the two profiles disagree on toe location by more than half the grid spacing, something is wrong. Re-sample a denser strip. Or shift the grid origin by a quarter interval and shoot again — a trick that exposes stair-step artefacts from alignment bias. One rhetorical question worth asking: Would you stake a grade-control concrete pour on that contour? If the answer wavers, your interval is wrong. Fix it now, not after the excavator arrives.

'The best interval is the one that makes the toe unambiguous before you open the software.'

— field rule-of-thumb from a geotechnical survey lead, paraphrased from a debrief after a false-edge incident

FAQ or Checklist in Prose

Quick checklist for field surveyors

You're standing at the edge of a cliff—literally. The tape measure says twenty meters, but your gut says the morphology changes in two. Stop. Write down three numbers before you set the first waypoint: the smallest feature you must resolve, the coarsest interval your timeline allows, and the transition slope where both break. If those numbers clash, the interval loses. Most teams skip this—they grab 5 m because “that's what we used last time.” That hurts. The checklist: one, confirm your dominant feature scale from aerial photos or a quick reconnaissance walk. Two, set the interval at half that feature’s narrowest dimension—a 4 m notch needs ≤2 m spacing. Three, test a single transect across the sharpest breakline. Four, accept that you might need two passes: one coarse for the flat terrace, one tight for the escarpment. Five, mark the seam between them in your field notes, not in post-processing. Wrong order there—you lose a day stitching mismatched datasets.

“We ran 10 m across a sandstone bluff. The contour software drew a gentle ramp. The cliff was a vertical wall. It never existed in the model.”

— field supervisor, Rocky Mountain survey, after losing a bid because the DEM showed a slope that wasn’t there

Common questions: Can I interpolate later? What if I have mixed morphology?

Interpolation is a lie that works most of the time. The catch is: interpolation can't create data where the interval skipped a two-meter overhang. It smooths, averages, and guesses—and at a cliff edge, the guess is always wrong. I have seen teams throw spline after spline at a 5 m grid that missed a quarry rim. The seam blows out every time because the software sees adjacent points on different geological benches and draws a nice, false slope between them. You can't fix sparse sampling with fancy math. If you have mixed morphology—say, a flat floodplain abutting a rocky scarp—run a variable grid. Use 10 m on the flat, then tighten to 2 m across the break, and log the transition as a waypoint label. The odd part is: most GPS controllers support this natively, but crews treat the interval as a single global setting. Don't. The rhetorical question here—would you measure a curbline with a survey wheel set to 100 m intervals? No. Then why do it to a cliff? What to check when it fails: pull the raw point cloud, look at the elevation histogram. If you see a smooth ramp where a vertical step should exist, your interval averaged out the key morphology. That's the signal. Redo the breakline zone before you touch the flat areas—that saves time, not the other way around.

What to Do Next

Resample your worst profile

Pick the single transect you suspect got smoothed to paste. You know the one—that steep, staircase-like slope where the raw point cloud showed a 40 cm vertical break over 2 meters, and your grid returned a gentle ramp. Pull that line into any profiling tool, then resample it at half your original interval, then at one-quarter. Don't interpolate; force a simple nearest-neighbor extract. What happens? If the steeper profile recovers even one step edge that the coarser version averaged out, you have your evidence. I have seen teams waste a week arguing over contour fidelity when one profile comparison would have ended the debate in ten minutes. The trick is to not look at the whole raster—look at one vulnerable transect. That scarp you walked past in the field? It's the test case you need.

Compare edge preservation metrics

Most people eyeball a hillshade and declare victory. That's dangerous. The catch is that hillshades hide systematic flattening under uniform lighting. Instead, compute a difference surface between your coarse and fine grids over the cliff-face region only. Not over the whole project—that dilutes the signal. Focus on the 10-meter buffer around each mapped breakline. Then ask: what percentage of cells show elevation differences exceeding your survey tolerance? If that number sits above 20%, you're not preserving morphology; you're smoothing hazard. The odd part is that some operators accept 40% difference because “the overall RMS looks fine.” It doesn't look fine when a steep scarp becomes a walkable ramp in the model. Document that percentage, then decide if you can live with it. One rhetorical question worth asking: would you build a retaining wall design from that smoothed edge? Probably not.

Document your interval rationale for project records

Future you won't remember why you chose 3.5 meters over 2.0. Write it now. A single paragraph explaining the feature scale you aimed to capture—say, “all scarps ≥ 1.5 m vertical relief required at least 5 grid cells across the break”—saves a re-survey later. I once revisited a site three years after the original survey; the only usable metadata was an engineer’s scribble on a PDF: “tried 4 m, lost the badger holes, settled on 2 m.” That scribble beat a blank logbook. So state the interval, note the smallest feature you targeted, and mention which profile you used as validation. Stick that note in the project file, the report appendix, and the metadata header. The last move is to share it with the field crew before they pack up—they might have seen something the desk jockey missed. That closes the loop. Now go test that worst profile.