The area gap is physics, not people.
Contract-farming programmes issue inputs, advance credit and settle on plot acreage. One boundary walk on an ordinary Android phone misses a half-acre plot by ±12.6%, so a large share of acreage disputes are the measurement arguing with itself, not the farmer or the technician.
Traverse reconciles the six lifecycle visits technicians already make, and the edges each plot shares with its neighbours, into one acreage with a stated error band: ±4.3% on the handsets already in the field, ±2.2% with a dual-frequency refresh. Offline-first, with no external GPS.
One half-acre boundary, Rs 60,000 an acre
Value in the error band on that one plot falls from Rs 3,777 to Rs 645, a 5.9x reduction with no external hardware.
The executive answer, on a 10,000 acre book
ModelledOn the reference book’s plot-size mix (median 1.59 acres), Rs 60,000 an acre and Rs 1,500 a re-survey. Every input is yours to change below.
What the error band costs, and whether a handset refresh pays.
Set your acres, plot size, tolerance and re-survey cost. The engine prices the error band three ways: today, with the six visits reconciled on current handsets, and with dual-frequency handsets. Most of the gain needs no hardware at all; the refresh earns its keep on small plots and tight tolerances.
Starts at the reference book's own plot mix. Smaller plots carry more error per acre.
How far measured area may sit from contract before it is disputed.
Today: one walk per settlement
Traverse, current handsets
Traverse, dual-frequency handsets
Reconciliation clears 87.8% of today's breaching plots with no new hardware, and the handset refresh pays back in 7.7 months on re-surveys alone.
ModelledArea error comes from the measurement engine at each plot’s size; breaches are the two-sided chance that measurement alone, with the contract exactly right, lands outside tolerance. Payback counts re-survey cost only: settlement value moved out of the error band is not counted and no software price is assumed. 2 seasons a year, 6 captures a season, 175 farmers per technician.
Split the settlement gap and neither half is carelessness.
Settlement variance is contracted area against measured area. It has two components: an area written from an eye estimate at onboarding, and a single walk at the handset's horizontal error. Both are properties of the method, and both shrink when the method changes rather than when the technician tries harder.
median 11.3%, 90th percentile 26.3%
written once at onboarding, then never revisited
one walk at the handset's error, 51% of the variance
In this reference book 31% of plots settle more than 15% from contract and 12% more than 25%. Household-survey research finds the same pattern in the field: farmer-reported areas diverge systematically from GPS-measured ones, most on the smallest plots.
Reference bookReference book of 157 settled plots, with the eye-estimate and handset error stated on the Architecture tab. A pilot measures this split on the operator’s own settlements in the first season.
Why a slower, more careful walk does not help
Phone GNSS error drifts, it does not flicker. It stays correlated over about 7 m of walking, so two fixes 2.5 m apart carry very nearly the same error. Logging more of them adds points, not information.
So the mechanism has to be more separate looks at the boundary, on different days and from different sides: six stages, and a neighbour walking the other face of the same bund.
And why the smallest plots are the worst
Area error scales with the perimeter and the area does not, so error as a percentage runs roughly as one over the square root of the plot. The plots that matter most to a smallholder are the ones the method serves worst.
One walk, 5.0 m handset error. The median plot in this book is 1.59 acres; 69 of 179 sit under half a hectare.
Change the plot, the handset and the method. Watch the error band move.
Every output below is recomputed on each change: the error, the polygon, the band around it, the simulated walks and the rupees. Start where most operators are today, one walk on an L1 handset, then turn the mechanism on.
0.5 to 10 acres is the real range in this belt.
Published range 4.5 to 8 m for a single-frequency phone.
App logs a track while the technician walks the bund.
One look at the boundary. Today's model.
A shared edge is walked from both sides, so it is observed twice.
± 0.063 ac on 0.50 ac
at Rs 60,000/ac
one walk today: Rs 12.5 L
Simulated1 seeded GNSS walk of this boundary, base seed 4211, sigma 5.00 m, error decorrelating over 7 m along the track. No device was used and no real plot is represented.
Same technician, same boundary, same care. The spread between these readings is the problem, and it is why a single capture cannot be the settlement number.
0.5 acre plot, sigma 5.0 m, 72 fixes at 2.5 m spacing, decorrelating over 7 m.
6 independent looks at one boundary on 6 different days. Random error falls as 1/sqrt(N), so 2.45x. Assumes no shared path offset.
Sigma 5.0 m to 2.50 m. Area error is linear in sigma. Handset refresh, no external hardware, no new training.
60% of the perimeter is shared with a neighbour and therefore observed twice. Variance halves on that portion: sqrt(1 - f/2).
The four rungs above scale sigma linearly, which is the first-order result. This row re-integrates the covariance at the sigma and method you actually selected, so it sits a little off the rung above it. Where they disagree, this row is the accurate one.
Two models, and where they part
These two do not agree on the level and should not be presented as if they did. The generator observes 4 vertices independently; the engine integrates a correlated error along the whole walked track, which is the more faithful model and the one every figure on this page uses. On the same statistic they sit about 20% apart. What survives both models is the reduction factor, and that is the claim the product makes: the mechanism is worth about 5.9 to 6.3 times, whichever noise model you believe. The closed form is also checked against a Monte Carlo of its own covariance in the engine’s automated tests, which is the like-for-like comparison.
Why a longer walk does not fix it
Receiver error drifts, it does not flicker. Two fixes 2.5 m apart are almost the same error, so logging more of them buys almost nothing, and standing still for 20 s buys a fraction of what the square root of the sample count would promise. The way out is more independent looks at the boundary, not a longer walk. That is the whole design.
Walked track, one fix every 2.5 m, error decorrelating over 7 m.
Every rung assumes the captures are of the same physical boundary. A plot that genuinely changed, for example a sub-let strip, is a data event, not a measurement error, and the anomaly rules must separate the two.
Book roll-up: each of the 179 plots is modelled at its own acreage, rounded to the nearest quarter acre, at its own measured shared-boundary share and at its own crop gross value. Disputes settle plot by plot, so they add rather than cancel.
At this plot size, free imagery tells time, not boundaries.
Checking the walk against 10 m imagery does not settle a half-acre dispute: one pixel straddles two neighbours' contracts. Sub-metre imagery and the state cadastral layer are useful references, but neither is a settlement-grade measurement of this season's cultivated boundary. What free imagery does well is the temporal question: did the crop emerge, die back or recover between visits.
ModelledPixels classified at render time against this plot’s actual geometry and its 4 abutting neighbours, using a stated rule: a pixel counts as pure only when its centre and all four corners fall inside one holding. 99.9% of this boundary is shared with a neighbour, so nearly every edge pixel is arbitrating two contracts at once.
At 10 m this plot carries a time series (16 clean pixels) but not a boundary (11.4% area error, 36% of the area in mixed pixels).
The imagery is worse than the walk it would be checking. Using it as the arbiter of a boundary dispute does not settle disputes, it manufactures them.
Canopy cover over the season, from the lifecycle model. Did the crop emerge, did it die back, did it recover: every one of those is a question about a time series, and a time series needs a handful of clean pixels and no boundary at all. Counted on this plot’s real geometry there are 13 of them at 10 m, against 16 from the closed form for a square of the same acreage, refreshed every 5 to 10 days. Either way it is plenty. That is the job satellite should be given.
Whatever a delineation recovers cannot be larger than the plot's pure-pixel core, because every boundary pixel mixes this holding with the next one. A recovered polygon that has lost more than a quarter of the field's area cannot be used to settle that field's area, so retention below 0.75 counts as not separable.
Reference bookCounted over all 69 sub-0.5 hectare holdings in the reference book. A plot counts as separable when its pure-pixel core keeps at least 75% of its area. The pixel grid is aligned to the plot frame, which flatters the coarse case, because a real orbit grid is not.
On fields under 0.5 hectare, moving from 10 m to 3 m imagery raises panoptic quality from 5.8 to 15.7 out of 100, and across all field sizes it cuts matched-boundary error from 18.6 m to 7.4 m. Even at 3 m, that boundary error is no better than one phone walk, and several times worse than six reconciled ones.
Source: Corley, Robinson, Marcus and Kerner (2026), Fields of the Planet: Field Boundary Mapping Beyond 10m.
This bears on BOUNDARY work only. Temporal signals - did the crop emerge, did it die back, did it recover - need no boundary resolution, and remain the right job for 10 m Sentinel-class imagery at this plot size.
The hardware answer is a handset, not a receiver.
Dual-frequency L1 + L5 phones cut standalone horizontal error substantially on average, though units vary; the engine plans on 5.0 m falling to 2.5 m and a pilot measures each model. Because area error is linear in that error, the upgrade is a device refresh the operator already budgets for: no second gadget, no new training, the same walk.
| Handset | Bands | Sigma | Area error | At risk / plot | Unit cost |
|---|---|---|---|---|---|
| Entry Android (2020 class)Android 11 · Ramulu M. · no raw GNSS API | L1GPS, GLONASS | 5.6 m | 14.4% | Rs 4,326 | Rs 8,500 |
| Mid Android (2021 class)Android 12 · Srikanth B. | L1GPS, GLONASS, NavIC | 4.9 m | 12.3% | Rs 3,689 | Rs 10,500 |
| Mid Android (2023 class)Android 14 · Anusha P. | L1 + L5GPS, Galileo, NavIC | 2.8 m | 6.6% | Rs 1,981 | Rs 13,900 |
| Mid Android (2024 class)Android 15 · refresh candidate, not yet issued to any technician | L1 + L5GPS, Galileo, BeiDou, NavIC | 2.5 m | 5.9% | Rs 1,757 | Rs 15,200 |
On a 0.5 acre plot the oldest handset in this fleet carries 14.4% of area error and the newest dual-frequency one carries 5.9%, for a difference of Rs 2,569 of arguable value on that single plot, every time it is settled. Area error is very nearly linear in sigma, so almost the whole gain is the second civil signal resolving ionospheric delay and rejecting multipath. The technician walks the same boundary the same way, on a phone that costs Rs 6,700 more than the one being replaced.
ModelledError recomputed per handset at the selected plot size by the same covariance sum behind the model above. Handset sigma, cost and fleet status are attributes of the synthetic device table, not estimates made here.
This is the device rung on its own: one walk, no reconciliation, no shared-edge constraint, so none of it is double counted against the mechanism above. And it is contract value sitting inside the error band, which is what a settlement argument is about. It is not a profit line.
Measured across the synthetic captures
Reference bookMean absolute area error of every capture in the synthetic reference book, against the boundary the generator drew. Higher than the closed form because these captures also carry canopy attenuation and village sky obstruction, which the idealised model above leaves out.
Why not an external receiver
A bluetooth GNSS receiver is a second device to charge, pair, carry, justify and lose, on a caseload of 150 to 200 farmers per technician across villages. It needs its own training step and its own support path, and the training budget here is the binding constraint, not the hardware budget. A handset refresh is a line item the operator already runs on a cycle, costs nothing in training because the walk does not change, and it upgrades the whole app at the same time.
That is why DeviceClass rides on every capture in this design rather than sitting in an asset register. A capture taken on an L1-only handset is a different measurement from one taken on L1+L5, and every tolerance, confidence band and settlement check downstream has to know which it is holding.
The measurement and evidence layer under every acre a programme pays on.
A plot never stores a single boundary. It stores what the contract says and every capture, each with its own error budget, handset class and evidence. The lifecycle walks and the shared edges a farmer club already has in the ground are reconciled into one boundary with a confidence interval, so settlement runs against a number that knows how certain it is, and a dispute ends with an evidence pack the farmer and the auditor can both read.
The phone assumes no network. Two technicians on the same plot and stage produce a conflict with both sides preserved, never a silent overwrite. A position the handset reports as mocked never enters a settlement.
Where it sits
Traverse is built as a module, not another farm-management suite. It would sit beside field-operations platforms such as Cropin or SourceTrace, which run the farmer relationship, and beside India’s AgriStack farmer registry and cadastral maps, which say who holds the land. It adds the one thing they do not: a settlement-grade acreage with a stated error band, exportable as standard polygons for traceability and deforestation-regulation filings.
Cropin, SourceTrace and AgriStack are named only to show where Traverse would fit. Traverse is an independent product concept and is not affiliated with, or endorsed by, any of them.
What is cited, what is modelled, and what you set.
Cited
- Corley, Robinson, Marcus and Kerner (2026), Fields of the Planet: Field Boundary Mapping Beyond 10m, arXiv 2607.04449arxiv.org
On fields under 0.5 hectare, moving from 10 m to 3 m imagery raises panoptic quality from 5.8 to 15.7 (out of 100); across all field sizes it cuts matched-boundary error from 18.6 m to 7.4 m. Independent benchmark. The pixel arithmetic on this plot set reaches the same conclusion from geometry alone.
- Sensors 22(24):9879 (2022), dual-frequency smartphone positioningdoi.org
Single-frequency L1 phones fix to roughly 5 to 8 m. Dual-frequency L1+L5 phones do markedly better on average, but units vary widely, and the 1.75 m figure often quoted is L5 differential GNSS with corrections, not a standalone fix. Planning value 2.5 m per axis for a standalone dual-frequency fix, set above the published best case on purpose. A pilot replaces it with each handset's measured sigma.
- EUSPA (2018), World's first dual-frequency GNSS smartphone hits the marketeuspa.europa.eu
Dual-frequency L1+L5 GNSS entered commercial phones in 2018 and is now common on mid-range Android. Establishes when the capability entered phones. Which mid-range models carry it is checked per model at procurement.
- Carletto, Gourlay and Winters (World Bank LSMS), From Guesstimates to GPStimates: Land Area Measurement and Implications for Agricultural Analysisdocuments1.worldbank.org
Farmer-reported plot areas diverge systematically from GPS-measured areas, with the largest relative errors on the smallest plots. The size of the gap on a given programme is an operator input: this reference book reproduces a comparable spread from its own eye-estimated contracts.
- Copernicus SentiWiki, Sentinel-2 missionsentiwiki.copernicus.eu
Four bands at 10 m; five-day revisit for the two-satellite constellation.
- Persello, Tolpekin, Bergado and de By (2019), Delineation of agricultural fields in smallholder farms, Remote Sensing of Environment 231doi.org
Smallholder field delineation is benchmarked against very-high-resolution imagery because small, irregular parcels have vague boundaries at coarser scales.
- Observation Quality Assessment and Performance of GNSS Standalone Positioning with Code Pseudoranges of Dual-Frequency Android Smartphones, Sensors 21(6), 2021doi.org
L5 and E5a code observations are measurably less noisy than L1 and E1 on the same handset: the physical reason the second signal helps.
Operator inputs
This book's crops run Rs 38,000 to Rs 1,47,600.
A settlement-grade standard; set your own above.
Technician visit, supervisor time and the delay.
Single and dual frequency. A pilot replaces both with measured values.
Onboarding, sowing, germination, pest and disease, pre-harvest, post-harvest.
The one fitted parameter; the first thing a pilot measures.
Modelled, on a reference book
Error percentages, intervals and rupee figures are computed by the measurement engine on every render. The book behind the aggregates is a synthetic reference programme of 160 farmers, 179 plots and 954 captures in 8 clubs, with simulated GNSS walks. No real farm, farmer or settlement is represented. A field pilot replaces the handset error and the decorrelation length with values measured on the operator’s own phones.
Reference book generated 2026-03-14. Separately from the measurement band, it carries about Rs 8.58 L attached to 65 planted anomalies across 24 disputed plots, which the Operations tab works through.