# Finding the Bad Records Across Thousands of Smallholder Plots

Last updated: 2026-10-23

> Find overlapping boundaries, duplicate farmers, impossible locations and made-up GPS across thousands of smallholder plots, then fix them in the field.

Date: October 23, 2026

To find the bad records across thousands of smallholder plots, run five checks on every plot and farmer, as the data comes in: overlapping boundaries, duplicate farmers and farms, missing or impossible locations, GPS that looks made up, and a review loop that sends each problem back to the field officer who can fix it. This guide shows how to run each one with free tools, and how to tell a real problem from noise.

**Where Restoration Scope fits:** we set up checks like these with each project. Checks on a single record run every time it's saved, checks that compare records (overlaps, shared GPS points) run as the data comes in, and a data quality page lists the records that need action beside them on the map. Field officers collect offline in [Restoration Scope Field](https://restorationscope.com/platform/field). [More on where software fits](#software).

The checks matter most for carbon and agroforestry programmes, where one plot mapped twice or a farmer counted twice can reach a verifier, a buyer or an EUDR due diligence statement. A problem you find this week is a farm someone can still visit. The same problem found at verification is only an explanation.

## How do I find overlapping boundaries?

Intersect every plot boundary with every other one, then measure each overlap as a share of the smaller plot. With tens of thousands of plots, use a tool with a spatial index (QGIS, PostGIS or GeoPandas), so each plot is compared only with its neighbours.

Then sort what you find:

- **A thin sliver along a shared edge.** Two neighbours walked the same hedge a few metres apart. Usually drawing noise; our [Verra file checker](https://restorationscope.com/tools/verra-geolocation-file-checker) treats overlaps under 10 square metres as a warning, not a problem.
- **Most of the smaller plot.** The same land mapped twice, often under two farmers or two seasons. Treat it as a possible duplicate (next section).
- **One plot inside another.** A sub-plot, or a family field registered again by a relative. The field officer knows which.

Carbon standards care: Verra's [geolocation file requirements](https://verra.org/documents/geolocation-file-requirements-v5-0/) don't allow overlapping or duplicate project instances. If your boundaries go into a Verra project file, the free [Verra geolocation file checker](https://restorationscope.com/tools/verra-geolocation-file-checker) marks overlaps, duplicate shapes and self-crossing outlines on the map before your VVB sees them.

**An overlap, marked on the map.**

Two farm boundaries overlap, and the shared area is marked so the field officer can see which edge to fix. Illustrative example: the farms and boundaries are made up.

![Illustrative check on a map: two farm boundaries overlap and the shared area is marked](https://restorationscope.com/images/landing/features/real/overlapping-polygons.webp)

## How do I find duplicate farmers and farms?

Duplicates come in two kinds: one farmer registered twice, and one farm registered twice. Look for both.

For farmers, compare in this order, from strongest to weakest evidence:

1. The same national ID number.
2. The same phone number.
3. The same name in the same village, after removing accents and capitals and sorting the words, so "Wanjiru Mary" and "Mary Wanjiru" match.
4. The same photo or signature file attached to two people.

For farms, look for the same boundary or point under two farmer IDs, and for the large overlaps from the last section.

Each match is a question, not a verdict. Spouses share a phone, cousins share a name, and many farmers rightly have more than one plot. When it is a real duplicate, keep one ID, move the other's history onto it and record why. Never delete the second record without a trace: someone will ask where it went.

## How do I find missing or impossible locations?

Start with the records that have no location at all, then test the ones that do:

- **0,0.** A point at latitude 0, longitude 0 sits in the Gulf of Guinea. It's a common placeholder when a phone had no GPS fix, or when a blank cell became zero.
- **Swapped latitude and longitude.** A range check catches this only when the longitude is beyond 90 or -90. In Kenya both numbers fit, so a swapped point still looks valid: it just lands in the wrong place.
- **A lost minus sign.** It moves a point across the equator or the prime meridian, sometimes into a neighbouring country.
- **Outside the country, or the district.** Test every point against the boundary of the country and, better, the district or village the record names. This one test also catches most swaps and lost signs.
- **Too few decimals.** Two decimal places can only place a point within about a kilometre; a spreadsheet that rounded your coordinates has lost the farm.
- **A point where you need a shape.** Under the EUDR, plots over 4 hectares need a polygon, not a point (cattle has its own rule).

To check a column of coordinates, paste it into the free [coordinate converter](https://restorationscope.com/tools/coordinates). It reads each line, says when a pair looks swapped, and puts every point on a map, so a point in the sea or the wrong country stands out. For plot files headed to the EU, the [EUDR plot file checker](https://restorationscope.com/tools/eudr-plot-file-checker) flags plots that look typed latitude first, rounds coordinates to six decimals and flags self-crossing outlines.

## How do I spot GPS that looks made up?

Most bad locations are mistakes. A few are shortcuts: a boundary drawn at a desk, or a farm registered from the office. These signs are worth a question:

- **Identical points on different farms.** Separate readings on separate farms rarely match to six decimals; copied coordinates do. Check the timestamps first: a phone can hand back the same cached location twice.
- **Perfect shapes, or every plot the same size.** A walked boundary has many corners and uneven edges. Squares, circles and a whole register of 1-hectare plots suggest someone drew rough shapes.
- **Walks too fast for the area.** A 2-kilometre boundary recorded in four minutes means 30 km/h. Nobody walks a field edge that fast.
- **Travel between farms that can't happen.** Sort each officer's records by time and divide the distance between farms by the minutes between them.
- **Points clustered around one spot.** Many farms within a few hundred metres of a field office or a meeting place.

None of these is proof. A walk can be finished on a motorbike, and a village meeting can register farms honestly. Ask before you conclude, and look at patterns by officer rather than single records. When a field worker walks a boundary in Restoration Scope Field, the screen shows the current GPS accuracy, and the saved shape leaves out fixes that imply an impossible jump. [How accurate phone GPS is](https://restorationscope.com/blog/phone-gps-accuracy) explains what to expect.

## Who fixes what? The field-officer review loop

A check that only produces a spreadsheet of errors changes nothing. The fixes happen in the field, so send each problem to the person who can resolve it:

1. **Route it.** Each flagged record goes to the field officer who recorded it, or whoever covers that area now, with the check it failed and the plot on a map.
2. **Resolve it.** The officer calls the farmer or visits on their next round, then marks the record: fixed, confirmed as correct with a note, or removed with a reason.
3. **Run the checks again** after each sync, so a fix that broke something else shows up the same week.
4. **Read the patterns.** Count open issues by officer and by check. Many identical points from one person is a training conversation before it is anything else.

Keep the loop weekly during registration and planting, when officers are already near the farms.

**The records to send back.**

Plots not mapped yet, a point used for a plot over 4 hectares and mapped areas that differ from the declared ones, each with the field team's next step, beside those plots on the map. Demo cooperative: producers, plot records, risk cases and field visits are made up, and deforestation statuses, check results and cutoff figures are not computed by Restoration Scope. Plot locations, map layers and imagery are real.

![EUDR demo Data quality page: 5 plots not mapped yet, 1 point used for a plot over 4 ha, 2 mapped areas that differ from the declared area and 70 with no issues, then the plots that need action, each with the field team’s next step, beside those plots on the map](https://restorationscope.com/images/landing/eudr-demo/data-quality.png)

## Where software fits

For a one-off audit, the free tools above and our [restoration data audit skill](https://restorationscope.com/blog/restoration-data-audit-skill) go a long way: give an AI assistant your farmer lists and map files, and it looks for swapped coordinates, points in the sea, duplicate rows and overlapping shapes, and tells you what to fix first.

When registration runs every season across thousands of farms, the checks need to run on every record as it arrives. Restoration Scope does that: field officers collect offline, the checks we set up with your team run as records arrive, and a data quality page shows what needs action on a map and a list. It doesn't decide what a record should say; your field team does. See [Restoration Scope for carbon projects](https://restorationscope.com/solutions/carbon).

More guides: [four tests for restoration data quality](https://restorationscope.com/blog/four-lenses-restoration-data-quality), [getting VM0047 records ready for verification](https://restorationscope.com/blog/vm0047-verification-readiness) and [what geolocation the EUDR requires](https://restorationscope.com/blog/eudr-geolocation).

## Questions about checking smallholder plot data

### How do I find overlapping farm boundaries?

Intersect every boundary with every other one in QGIS, PostGIS or a short script, and measure each overlap as a share of the smaller plot. Thin slivers along a shared edge are usually drawing noise; a large share usually means the same land was mapped twice.

### How do I find duplicate farmers?

Compare national IDs and phone numbers first, then names within the same village after removing accents, capitals and word order. Treat each match as a question for the field officer, not a verdict: relatives often share a phone and a name.

### Why don't range checks catch swapped latitude and longitude?

When both values lie between -90 and 90, as they do in Kenya, a swapped pair still looks like a real coordinate. Check every point against the country, and ideally the district, the record says it is in.

### How can I tell if GPS points were made up?

Look for farms with exactly the same coordinates, coordinates rounded to two or three decimals, perfect squares or circles, every plot the same size, and walks or travel between farms faster than a person can move. Each is a reason to ask, not proof.

### Who should fix the records a check finds?

The field officer who recorded them, or whoever covers that area now. They can call the farmer or visit, then mark the record fixed, confirmed as correct with a note, or removed with a reason.