About the project

This Python project performs point interpolation within a given polygon, producing an interpolated raster (GeoTIFF) as output. By default it uses world country boundaries from OpenStreetMap (OSM), but any polygon file can be provided. The point layer is automatically generated to fit within the selected boundaries. Two interpolation backends are supported: scipy.interpolate and matplotlib.tri.

How it works

The pipeline runs in the following steps:

1. Random point generation (generate_random_points)

Given a polygon file (e.g. country boundaries), random points are generated within the bounding box of the polygon and then spatially filtered to keep only those that fall within the polygon. Each point is assigned a random numeric z-value (between 100,000 and 200,000 by default). The result is saved as a GeoPackage (.gpkg).

By default 200 points are generated. This can be overridden with the --points_number / -n option when running the Python script:

python src/run_interpolation_methods.py -l scipy -p /usr/src/app/data/multipolygons.shp -t /usr/src/app/data/random_points.gpkg -m linear -o /usr/src/app/data/linear.tif -n 500

2. Regular grid generation (RasterInterpolator.generate_point_grid)

The spatial extent of the point cloud is used to build a regular x/y grid at a fixed cell resolution (default: 2 CRS units per cell). Each point is also assigned a fresh random z-value at this stage, which is used as the value to interpolate.

3. Interpolation

The scattered point values are interpolated onto the regular grid using one of two backends:

SciPy (scipy.interpolate.griddata)

Uses griddata to fill the grid. Supported methods:

Method	Description
linear	Piecewise linear interpolation (Delaunay triangulation)
cubic	Piecewise cubic interpolation
nearest	Nearest-neighbour interpolation

Matplotlib (matplotlib.tri.Triangulation)

Triangulates the point cloud and evaluates an interpolator on the meshgrid. Supported methods:

Method	Description
linear_tri_interpolator	Linear interpolation on the triangulation
cubic_geom_min_e	Cubic interpolation minimising energy (kind='min_E')
interp_cubic_geom	Cubic interpolation using geometric approach (kind='geom')

4. Raster output (RasterInterpolator.write_rast)

The interpolated grid is written to a single-band, LZW-compressed GeoTIFF with:

• CRS: EPSG:4326 (default)
• Nodata: NaN
• Affine transform centred on pixel centres

Environment

The project runs in a Docker container, with commands specified in the Makefile.

Code is linted using flake8 with --max-line-length=120.
Code formatting is validated using Black.
pre-commit is used to run these checks locally before files are pushed to Git.
The GitHub Actions pipeline also runs these checks.

Running the code

To view the required parameters, run: python run_interpolation_methods.py --help

If the OSM file with polygons is already downloaded, run the following python commands:

python src/run_interpolation_methods.py -l scipy -p /usr/src/app/data/multipolygons.shp -t /usr/src/app/data/random_points.gpkg -m linear  -o /usr/src/app/data/linear.tif

or

python src/run_interpolation_methods.py -l  mpl -p /usr/src/app/data/multipolygons.shp -t /usr/src/app/data/random_points.gpkg -m linear_tri_interpolator -o /usr/src/app/data/linear_mpl.tif

Interpolation method can be: linear, cubic, or nearest for scipy, or linear_tri_interpolator, cubic_geom_min_e, or interp_cubic_geom for mpl.

If the countries boundaries OSM file is not already downloaded, run the following bash commands:

run as: ./src/run_interpolation_methods.sh scipy /usr/src/app/data/multipolygons.shp /usr/src/app/data/random_points.gpkg linear /usr/src/app/linear_scipy.tif /usr/src/app/data/

or

./src/run_interpolation_methods.sh mpl /usr/src/app/data/multipolygons.shp /usr/src/app/data/random_points.gpkg linear_tri_interpolator /usr/src/app/data/linear_mpl.tif /usr/src/app/data/