17 posts
Turn the LOCA group of a real AGS4 file into points in British National Grid, see what happens when the CRS is wrong, summarise each hole's geology from GEOL, check every ground level against Environment Agency LIDAR, and build a QGIS project with labels and a map tip that shows the borehole log.
Build Voronoi (Thiessen) polygons in Python with shapely and GeoPandas, clipped to a real boundary. Compute Thiessen-weighted catchment rainfall for the River Eden from Environment Agency rain gauge readings, check the gauges, compare with the gauge average and IDW, then compare Voronoi 'nearest fire station' areas in Cumbria with drive-time service areas from OpenStreetMap.
Batch geocode all 192 open schools in Sheffield in Python with geopy and OpenStreetMap Nominatim (1 request per second, a custom user agent and an on-disk cache) and with postcodes.io bulk postcode lookups, reverse geocode a sample, and measure every method honestly against the Department for Education's own coordinates. Covers free text against structured queries, failure modes, cross-checking two geocoders, and the licensing of geocoded results.
A MovingPandas tutorial on one day of real AIS vessel tracks in Puget Sound: 277,574 GPS fixes from 703 vessels turned into trajectories, with speed and direction, outlier cleaning, stop detection at ferry terminals, splitting into 148 trips, Douglas-Peucker against time-aware generalisation, a flow map of 319 aggregated flows, and reading and writing GPX.
Cluster GPS and point-of-interest data in Python with DBSCAN and HDBSCAN: use the haversine metric on latitude and longitude or plain metres on a projected grid, choose eps with a k-distance plot, handle variable density with HDBSCAN, draw convex and concave hull polygons, treat noise properly, read silhouette and DBCV with care, and scale up with a BallTree. Uses 3,950 real Overture Maps restaurants, cafés and bars in Glasgow.
Assess a real land cover map in Python: ESA WorldCover 2021 against 8,597 Eurostat LUCAS 2022 field survey points in Ireland. Build the confusion matrix, compute overall, user's and producer's accuracy, see why kappa is discouraged, and use the Olofsson et al. (2014) stratified estimators to get error-adjusted areas with 95% confidence intervals and to plan a new stratified random sample.
Test whether points are clustered with point pattern analysis in Python: a quadrat chi-squared test, the Clark-Evans nearest-neighbour index and Ripley's G, F, K and L functions with Monte Carlo CSR envelopes, edge correction, a global MAD test and a random-labelling null, on 3,292 London plane trees in Camden.
A practical guide to Uber's H3 hexagonal grid in Python with the h3-py v4 API: index points, choose a resolution, fill polygons, count 541,081 Overture places across London, smooth with k-rings, do the same in DuckDB's h3 extension, and compare hexagons with square grids of equal area.
Convert raster to vector in Python with rasterio.features.shapes on 10 m ESA WorldCover for Greater Manchester: polygonize water and built-up land, fix invalid polygons, simplify without gaps, rasterize ONS district boundaries for land-cover shares, then write and validate a Cloud Optimized GeoTIFF with rio-cogeo and measure HTTP range reads.
A practical guide to spatial joins in GeoPandas with real Greater Manchester data: 13,187 NaPTAN bus stops, 1,386 OpenStreetMap schools, 215 ONS wards and 1,702 LSOAs. Covers point in polygon with sjoin, predicates, counts per polygon, how='left'/'inner'/'right', sjoin_nearest with max_distance, clip and overlay, and a timing comparison of a naive loop against the spatial index.
A practical guide to coordinate reference systems in Python with pyproj, GeoPandas and rasterio, tested on real Greater Manchester data. Covers EPSG codes, WGS 84 vs British National Grid (EPSG:27700), the always_xy axis-order trap, OSTN15 vs Helmert datum shifts checked against 13,187 NaPTAN bus stops, measured area and distance errors in the wrong CRS, and reprojecting a GeoDataFrame and a Copernicus DEM raster.
Calculate zonal statistics in Python with exactextract and rasterstats on real data: ONS 2021 LSOA boundaries for Cumbria with GHSL population, the Copernicus DEM and ESA WorldCover. Covers partial-pixel weighting vs pixel centroids, CRS alignment, categorical land-cover fractions, a speed benchmark, GeoPackage export and choropleth maps.
Run a hot spot analysis in Python with PySAL: aggregate 2025 STATS19 road collisions in Birmingham to a hexagon grid, test for clustering with global Moran's I, map Getis-Ord Gi* hot spots and LISA clusters with permutation inference and FDR correction, and see how spatial weights, the areal unit and exposure change the answer.
Pull Bristol's 144,965 buildings and 20,727 places straight from the Overture Maps GeoParquet release with DuckDB, run spatial SQL joins and H3 aggregation, load the result into GeoPandas, write GeoParquet 1.1, and time it against Shapefile, GeoPackage and FlatGeobuf.
Benchmark point-in-polygon joins, buffers, overlays and dissolves on 200,000 points with GeoPandas and Shapely 2, see why spatial indexes give a 100x+ speed-up, and compare Shapefile, GeoPackage, FlatGeobuf and GeoParquet.
Turn GeoDataFrames into shareable Leaflet web maps: a one-line choropleth with .explore(), then a hand-built Folium map with custom styling, tooltips, clustered markers, switchable basemaps and plugins — saved as a single HTML file.
Load Natural Earth vector data with GeoPandas, reproject to an equal-area CRS, calculate population density, join cities to countries, dissolve to continents and draw publication-ready choropleths.