Location Validation

Assesses geometric quality, commodity presence, and protected area status. This pillar runs 15 geometric checks on every plot polygon, each producing a weighted contribution to the shape validity score.

Geometric checks:

Check	Description
Area validation	Plot area must fall within commodity-specific bounds. Extremely small plots (below detection algorithm resolution) or implausibly large plots are flagged.
Self-intersection	The polygon must not self-intersect. Self-intersecting geometries produce ambiguous area calculations and are automatically repaired where possible (buffer-zero correction).
Duplicate vertices	Consecutive identical or near-identical vertices are detected.
Angle validation (spike detection)	Interior angles are checked for extremely acute angles (near-zero or near-360 degrees) that indicate digitization errors or slivers.
Distance validation	Maximum distance between vertices is checked to detect implausibly elongated shapes.
Precision validation	Coordinates with insufficient decimal places (e.g., integer degrees) are flagged. Minimum precision of approximately 10 m is required.
Boundary segment length	Short boundary segments that suggest over-digitization or artifacts are detected.
Overlap detection	Plots within the same collection are checked for mutual overlap. Significant overlap may indicate duplicate submissions or boundary errors.
Nearest neighbor distance	Distance to the closest other plot in the collection is measured. Extremely close but non-overlapping plots may indicate boundary drawing errors.
Geometry collection processing	Nested geometry collections are flattened and simplified where possible.
Multipolygon flattening	Multipolygon geometries are decomposed into individual polygons for per-plot analysis.
Triangular geometry	Triangular shapes are flagged as potentially artificial or placeholder geometries.
Commodity presence	The plot location is checked against commodity probability maps to verify the declared commodity is plausibly grown in that area.
Protected area overlap (WDPA)	Plot boundaries are checked against the World Database on Protected Areas. Overlap with protected areas triggers a legality risk flag.

Shape validity issue weights:

Each geometric issue contributes to the overall shape validity score with the following weights:

Issue	Weight
Intersects another polygon	0.30
Self-intersection	0.25
Commodity absence	0.25
Distance to nearest neighbor	0.20
Protected area overlap	0.20
Area too small	0.10
Maximum distance	0.10
Maximum angle	0.10
Bad precision	0.05
Short boundary segments	0.05
Triangular geometry	0.05
Duplicate vertices	0.05
Nested geometry collection	0.05
Simplifiable collections	0.05

Plot-level deforestation detection using multiple alert systems:

• GLAD alerts (Global Land Analysis and Discovery)
• TMF alerts (Tropical Moist Forests)
• Hansen Global Forest Change

Deforestation analysis is restricted to natural forest. Managed forest areas are exempted using the following rules:

• GFC (Global Forest Change) subtype 20 areas are excluded (plantation forest)
• Nature Trace Forest Typology (ForTy) planted and plantation forest classes are excluded globally
• Country-level managed forest exemptions apply for Canada and the United States

Only post-EUDR-cutoff deforestation is considered for compliance purposes.

Detects geographic selection bias: whether plots have been drawn to avoid known deforestation hotspots. The system analyzes the vicinity within the collection's extent bounding box and compares deforestation rates inside plots versus the surrounding area.

Risk Level	Vicinity Deforestation Discrepancy	Interpretation
Low	<= 5%	No evidence of selective boundary drawing
Medium	<= 20%	Some indication of avoidance patterns; warrants review
High	> 20%	Strong evidence that plots were drawn to exclude deforestation

Production volume plausibility check. When a production_volume parameter is provided, the system compares the declared volume against yield estimates for the plot area and commodity. This detects volume overstatement -- declaring more production than the land area could plausibly support, which may indicate that production from undisclosed (potentially deforested) land is being attributed to the declared plots.

AI-powered web intelligence assessment using Gemini with Google Search grounding. This is a collection-level check (not per-plot) that searches for:

• Post-2021 deforestation incidents linked to the operator or region
• Land rights disputes or contested tenure
• Environmental violation records
• Sanctions or enforcement actions
• Media reports of illegal activity

A weighted combination of all pillar scores into a single risk assessment. The overall result includes an investigate list identifying the specific issues that require attention.

---

The facility location is checked against commodity probability maps within a configurable radius (default: 5000 m). A facility declaring palm oil in an area with no palm cultivation signal is flagged. The radius accounts for the fact that facilities may be located in towns or industrial zones adjacent to, rather than within, production areas.

Connectivity to transportation infrastructure is assessed using Overture Maps and Google Places data:

• Road access: Distance to the nearest road network segment
• Waterway access: Proximity to navigable waterways (relevant for commodities like palm oil where river transport is common)
• Port access: Distance to the nearest port facility

Facilities with no road access within a reasonable distance are flagged as potentially mislocated.

High-resolution building footprint and industrial facility data from Overture Places is used to verify that structures exist at or near the declared location, within a 100 m radius. Scoring is graduated by facility size:

Detected Feature	Score Contribution
Large industrial building	+0.23
Medium building	+0.12
No buildings detected	No contribution (lowers confidence)

An empty location with no built-up area is flagged as potentially non-existent.

Nearby points of interest (POIs) from Overture Maps and other sources are queried for industrial, agricultural, and commercial features. Matching POI types (e.g., "silo", "mill", "warehouse", "processing plant") increase confidence that the location is a real operational facility.

Latitude constraints are applied per commodity. For example, palm oil facilities at latitudes above 25 N or below 25 S are flagged as geographically implausible.

The facility's surrounding area is screened for deforestation signals:

• Resolution: 30 m
• Radius: 50 km
• Forest type: Natural forest only (managed forest excluded)
• Time period: Post-EUDR cutoff date only

Facilities surrounded by high deforestation activity are flagged for enhanced due diligence.

AI-powered web intelligence assessment (same Gemini + Google Search grounding engine as plot-level legality checks). Searches for regulatory violations, sanctions, and environmental incidents associated with the facility operator and location.