Systematic grid sample points in a polygon
Sample points on a square or triangular grid of a set spacing across a polygon, with a random start from a seed or a start at the corner, and the area each point stands for.
Experimental: not yet fully verified. How results are checked
56 points on a grid, one for every 2,165.06 m².
- Area per point
- Start offset
- Seed
- Plane
- Design
Provenance
- Computed by
- geometry.sampling.systematic-points 1.0.0, core 0.1.0
- Model
- A lattice on the UTM plane from the lower-left corner of the polygon's bounding box (in the grid's turned frame): square, or triangular with rows d·√3/2 apart and every other row shifted d/2; spacing scaled by the zone's point scale at the polygon's center so neighbors are d apart on the ground; a random start moves the lattice by (r₁·d, r₂·row spacing) from two draws (EPA QA/G-5S ch. 7); points kept inside the outline and outside holes
- Accuracy
- Neighbors are the spacing apart on the ground to within the change in UTM scale across the polygon (under 1 mm over 1 km near the center)
- Notes
- 1 shown with the answer
- Cites
- US Environmental Protection Agency, Guidance on Choosing a Sampling Design for Environmental Data Collection (EPA QA/G-5S); Blackman, D., and Vigna, S., ACM Transactions on Mathematical Software, Scrambled linear pseudorandom number generators
Something look off?
How we got thisFormula, worked example, sources, and proof
Model: A lattice on the UTM plane from the lower-left corner of the polygon's bounding box (in the grid's turned frame): square, or triangular with rows d·√3/2 apart and every other row shifted d/2; spacing scaled by the zone's point scale at the polygon's center so neighbors are d apart on the ground; a random start moves the lattice by (r₁·d, r₂·row spacing) from two draws (EPA QA/G-5S ch. 7); points kept inside the outline and outside holes
Show your work
Area per point
d = 50 m= 2,165.06 m²Points inside
rows d·√3/2 apart, every other row shifted d/2d = 50 m on the ground, 49.98 m on the plane= 56
The same steps an agent gets from the MCP server with explain: true.
Accuracy: Neighbors are the spacing apart on the ground to within the change in UTM scale across the polygon (under 1 mm over 1 km near the center)
When to use this: Use this to lay out sample points evenly over a site: soil cores, plots, or check points on a grid. A triangular grid covers ground most evenly; a random start keeps the design a probability sample. The same seed gives the same grid.
Limitations: A grid can line up with patterns in the ground, such as crop rows or drains, and miss or over-sample them; rotate it or use a random design if that is a risk. At most 100,000 points; the polygon may span 500 km and 2,000 corners.
Worked example: A 50 m triangular grid with a random start, seed 42. Source: A 340 m × 333 m field.
You enter
- Pattern
- triangular
- Polygon
- 40, -105 40, -104.996 40.003, -104.996 40.003, -105
- Seed
- 42
- Spacing
- 50 m
You get
- Points placed
- 56
- Area per point
- 2,165.06 m²
- Start offset
- 4.19 m east, 16.41 m north
- Seed
- 42
- Plane
- UTM zone 13N (WGS 84)
- Design
- Systematic triangular grid (EPA QA/G-5S ch. 7)
Review: Not yet independently reviewed by a GIS professional.
Last verified: 2026-10-20, when a maintainer last confirmed this tool's sources at the issuer. See the sources ledger.
Status: version 1.0.0, core 0.1.0. See this tool in the verification report.
Checked against: 21 golden test vectors (download the test vectors, each with its source and tolerance). See how results are checked and every source.
Sources
- Guidance on Choosing a Sampling Design for Environmental Data Collection (EPA QA/G-5S), US Environmental Protection Agency, EPA/240/R-02/005, December 2002. Chapter 7, systematic/grid sampling (p. 63).
- Scrambled linear pseudorandom number generators, Blackman, D., and Vigna, S., ACM Transactions on Mathematical Software, Vol. 47, No. 4. xoshiro256**, seeded by splitmix64 as the authors suggest; doubles from the top 53 bits. Read free.
Experimental means this tool has not yet met the stable bar: at least 20 golden vectors, differential tests, and an independent worked example.