Learning paths

Take a live point feed and an administrative boundary layer and produce a defensible coverage report against a policy target: supply per area, walk-distance coverage per unit, the projection choices that make those numbers true, the indexed query that keeps the app responsive, and the tile budget that takes the map offline.

One transport authority's question — is the cycle-hire scheme where people are? — worked from a live docking-station feed to a coverage figure a minister can quote, with the projection mistakes met on the way.

6 steps · ~3 h · 1,110 XP

Work with worldwide vector data in Python without a single projected CRS to hide behind: audit a gazetteer against its own geometry, measure nearest-neighbour distances geodesically, and write clip-and-aggregate scripts in an equal-area projection that a reviewer can re-run.

Natural Earth as a working dataset rather than a backdrop: geodesic distance across a planet, clip-and-credit overlays in an equal-area projection, and the audit that decides whether a reference file can be trusted at all.

4 steps · ~3 h · 940 XP

Take a global boundary layer and make defensible maps and lists from it: quantify projection distortion, shade a rate rather than a count, audit identifiers and geometry before joining on them, and apply an area rule across borders in an equal-area projection.

The countries layer as a working dataset: what a projection does to area, what normalisation does to a choropleth, what generalisation does to a coastline, and the audits that decide whether the layer can be trusted for the job.

5 steps · ~3 h · 1,100 XP

Write the four spatial joins that cover most production PostGIS work — aggregate-per-area, within-a-distance, nearest-neighbour and point-in-polygon — in the index-assisted form, in metres on geography, with NULL-safe logic that keeps the rows a report is about.

Four spatial-SQL patterns on real tables — the outer join that keeps empty rows, the DWithin an index can answer, the KNN that finds the nearest, and a containment audit — each written the way the planner can use.

4 steps · ~2.5 h · 920 XP

Audit a dataset before using it — find the missing, malformed and duplicated keys, measure where attributes disagree with geometry, test one layer's completeness against another, and decide with numbers whether a polygon layer is fit to derive attributes from.

Five audits a professional runs before trusting a layer: identifiers, coordinates against geometry, probable duplicates, completeness against a second layer, and the fitness of a boundary polygon for assignment.

5 steps · ~3 h · 1,120 XP

Turn a live point feed into operational figures and a plan: per-area supply and utilisation in SQL and in a script, walk-distance coverage against a target, an indexed nearest-available query, and a nightly rebalancing plan sized from the data.

One live transport feed worked end to end — supply, aggregates, utilisation, coverage, nearest-available and a rebalancing plan — the way an operations analyst does it every day.

6 steps · ~4.5 h · 1,560 XP

Take delivered rasters through acceptance, summarise a categorical and a continuous band correctly, compute zonal statistics across a CRS boundary without resampling the pixels, turn pixel counts into ground area on a geographic grid, and combine three rasters into one defensible table.

Three delivered rasters — a land-cover map, a July Sentinel-2 scene, an elevation model — worked from acceptance to a programme brief: histograms, zonal statistics, band maths, pixel areas, and the projection rule that keeps them honest.

8 steps · ~6 h · 2,150 XP

Given mixed open-data extracts and a receiving system's rules, you can clean identifiers and attributes, validate geometry and CRS, and deliver a named, typed GeoPackage whose contents and precision survive review.

From messy source attributes to a validated spatial file that another GIS team can load and trust.

7 steps · ~5.5 h · 1,930 XP · Assumes: Basic familiarity with GIS layers, attribute tables and point versus polygon geometry.; Ability to run a short SQL query, Python script or desktop GIS operation with the provided starter contract.

Join and audit country and raster data, choose honest denominators and visual ranges, retain outliers behind a disclosed display cap, and deliver a mapped policy finding with its source limits explicit.

Rates, missingness, raster evidence, outliers and signed change: turn public geospatial data into maps that a policy reader can question and an analyst can reproduce.

6 steps · ~5.5 h · 1,980 XP

Build an auditable response briefing from heterogeneous GIS evidence: reject unsuitable inputs, produce a map of what the data actually measures, retrieve candidate facilities with indexed spatial SQL, retain condition history in PostGIS and explain which access decisions still need direct verification.

A multi-technology response workflow: validate a raster, measure hazard context, screen global event data, query nearby facilities and design a database that keeps condition checks traceable.

7 steps · ~7.5 h · 2,590 XP