Build rigorous, reproducible
geostatistical pipelines in Python.

Why spatial data needs its own methods

Spatial data violates the i.i.d. assumption. Observations near each other are correlated, scale changes everything, and naive cross-validation silently inflates accuracy. This site documents the methods, code, and design patterns that let you ship spatial models that actually generalise.

Three pillars anchor the site: Core Concepts covers the mathematical foundations — spatial dependence, stationarity, variography, point processes, and weight matrices. Kriging & Interpolation turns sparse point data into uncertainty-aware surfaces using IDW, ordinary kriging, and universal kriging with PyKrige. Python Workflows wires everything together into end-to-end pipelines with GeoPandas, PySAL, scikit-gstat, and Dask — including spatial regression, cross-validation, and memory-efficient processing.

Every page ships copy-ready Python, explicit validation diagnostics, and documented failure modes — the parts that matter in production but rarely appear in tutorials.

The three pillars

Start with whichever fits your current question. Each pillar links to sub-topics and deep-dive articles.

Start here — implementation guides

The most-used walkthroughs. Each one is self-contained: copy the code, run it, ship it.

All implementation articles

Every hands-on guide on the site, organised by pillar.