Documentary photograph illustrating spreadsheet MRV vs dMRV
← Back to blog

Clipping and Overlap-Checking an AOI the Right Way

Sustainability•30 October 2026•TREEO Indonesia•4 min read

Every carbon project is anchored to a boundary, the area of interest (AOI). It decides which pixels and field points count and which hectares appear in the project document. Errors here are quiet. This article walks through a clip-and-check routine that has a few firm rules: log everything, delete nothing, and check overlap twice.

Principles

- Run the routine end to end and state every assumption in the final report. - Flag overlapping or invalid records. Never delete or modify them. - Run the overlap check on the original data first, then repeat it on the clipped data. Clipping can remove the partner record that made an overlap visible. - Verify numerically (counts, geometry intersections) before claiming anything. - List the output folder before writing. Do not overwrite files that existed before the run.

Step 1: Fix the boundary

- Check the CRS on load. A boundary delivered in WGS84 (EPSG:4326) should be asserted as such. - Repair invalid geometry with a make-valid step, then dissolve the parts into one polygon. Keep a note of what changed. - Report the raster CRS, resolution, bounds, bands, data type and no-data value for any imagery you clip. - If the imagery is in a UTM zone and the clip must be delivered in WGS84, reproject band by band with bilinear resampling and an explicit no-data value of 0.

Step 2: Clean the point tables

Field plots and tree records arrive as tables with longitude and latitude. Before clipping:

- Coerce coordinates to numbers. - Exclude rows with null, non-numeric or zero coordinates, and rows out of range (longitude beyond 180 or latitude beyond 90). - List every excluded row in a QC log. Never drop one silently. - Keep the original row number so any record can be traced back to the source table.

Step 3: Clip

- Clip the raster to the AOI polygon with a mask that crops to the polygon extent and writes no-data outside it. - Clip points by an intersects test, which keeps points that sit exactly on the boundary. - Report counts before and after the clip for each dataset. - Build the footprint of valid raster pixels and count clipped points that fall inside the AOI but outside the raster. - Context tables that are not part of the analysis should not be clipped.

Step 4: The overlap checks

Compare two datasets, for example a baseline list and an intervention list, with five checks.

CheckDefinition
aThe same ID appears in both files
bAn ID is repeated within one file (list every row)
cThe same normalised name appears under different IDs
dA point from each dataset lies within 5 m of the other
eA baseline record shows adopted practices yet also appears in the intervention set

How each one works:

- Name matching. Normalise first: lower-case, replace punctuation with spaces and collapse repeated spaces. Name-only matches are possible duplicates, never merges. Two people can share a name, so refer them to the field team. - Spatial matching. Use a 5 m buffer join in metres, in the local UTM zone, rather than only a nearest-neighbour join. Report within-file cases as well. - Baseline integrity. Check e is a logic conflict. A baseline record with adopted practices that is also counted as intervention should be called out loudly. - Flags on the clipped layers. Compute them against the full original datasets, so a partner outside the AOI still counts. - Flag columns. Keep a column for exact ID overlap, one for the overlap type in priority order (exact ID, spatial duplicate, name only, within-file duplicate), one for the partner scenario, and one listing every type that applies.

Recommendations from these checks are suggestions. For exact-ID or spatial overlap, keep the record in the intervention set only if practices were adopted, and in the baseline only if no adoption is recorded. Otherwise review manually.

Step 5: Polygon overlap between strata

When strata or parcels are polygons, test them in projected metres. One workable rule is to reject the set if any pair of strata intersects by more than 1 square metre, which ignores digitising noise but catches real overlap. A plot-allocation rule can also require that a minimum share of each plot (75 percent in one implementation) lies inside its stratum.

Step 6: Deliver and report

- Write a GeoPackage with one layer per clipped dataset plus the AOI and the overlap records, all in EPSG:4326. - If you also write shapefiles, field names are limited to 10 characters. Truncate them with a unique suffix on collision and record the mapping (layer, original field, shapefile field) in the QC log. - Include the clipped raster, an Excel workbook with the clipped tables, the overlap results, a summary and the QC log. - The report should cover: CRS and raster details, excluded rows, counts before and after, overlap results for original versus clipped data, and assumptions. - Point out data oddities for a human to check, such as intervention rows with no adoption recorded but practices listed.

Checklist

- Make geometry valid and log the changes. - Confirm CRS; measure distances in the local UTM zone. - List excluded rows and out-of-raster points. - Run checks a to e on original data, then on clipped data. - Flag, never delete. - Archive the final boundary and outputs without overwriting earlier files.

For what standards expect of the boundary itself, see project boundary definition, and for how clipped layers feed eligibility work, see forest canopy density eligibility analysis.

Turn climate goals into a verified portfolio

TREEO connects the full carbon cycle — eligibility, simulation, real-time monitoring and registry-ready reporting — combining expert consulting with dMRV technology so the evidence exists before anyone asks for it.

Related Articles