geointerpo

Spatial interpolation for Python — 18 methods, live data APIs, and boundary clipping.
Drop in point data, get a smooth interpolated raster out.

Fetch live weather, air quality, or precipitation data from Meteostat, OpenAQ, Open-Meteo, NASA POWER, and ERA5.
Define your study area by place name, polygon file, or bounding box — boundaries are resolved automatically.
Export to GeoTIFF or NetCDF, run spatial cross-validation, compare methods side by side, and visualize interactively.

📖 Documentation · Install · Quickstart · Methods · Examples

Ordinary Kriging · Natural Neighbor · Gaussian Process — same 60 stations, Alberta, Canada

Install

pip install "geointerpo[full]"

Quickstart

from geointerpo import Pipeline

result = Pipeline(
    data="stations.csv",               # CSV, GeoDataFrame, or live API
    boundary="Calgary, Alberta",       # place name, bbox, or polygon file
    method=["idw", "kriging", "spline"],
).run()

result.plot()            # side-by-side comparison
result.plot_interactive()# zoomable plotly/leafmap map
result.metrics_table()   # cross-validation RMSE / r
result.best_method()     # best CV score
result.save("outputs/")  # GeoTIFF + PNG + CSV

Methods

geointerpo covers 18 canonical methods across deterministic, geostatistical, and ML workflows. All share the same interface — swap method= to compare.

Distance-based

The fastest methods — no statistical assumptions, exact at data points. Ideal as a quick baseline or when data is dense and evenly distributed.

idw · nearest · linear · cubic

Spline & Trend

Fit smooth continuous surfaces. Splines minimise curvature; RBF offers eight kernel choices; Trend fits a global polynomial for large-scale patterns.

spline · spline_tension · rbf · trend

Geostatistical

Account for spatial autocorrelation via a variogram model. Produce statistically optimal, unbiased estimates. Natural Neighbor uses Voronoi area-stealing weights — smooth and exact at data locations.

kriging (Ordinary) · uk (Universal) · natural_neighbor

Machine Learning

Capture non-linear spatial patterns. GP returns a full uncertainty surface alongside the mean prediction. Regression Kriging combines an ML trend with Kriging of the residuals.

gp (Gaussian Process) · rf (Random Forest) · gbm (Gradient Boosting) · rk (Regression Kriging)

References

• ArcGIS Pro — Interpolation Tools Overview
• 3 Best Methods for Spatial Interpolation — Towards Data Science
• GeoStat-Framework/PyKrige — Ordinary, Universal, and Regression Kriging
• GeoStat-Framework/GSTools — Covariance models, variograms, random fields
• mmaelicke/scikit-gstat — Variogram estimation and ordinary kriging
• DataverseLabs/pyinterpolate — IDW, kriging, Poisson kriging
• fatiando/verde — Machine-learning-style spatial gridding
• GeostatsGuy/GeostatsPy — GSLIB-based geostatistics