A city digital twin is only as trustworthy as the geometry underneath it. Most municipalities hold several overlapping sources: a cadastral parcel layer maintained by the land registry, building footprints from aerial mapping, a terrain model from laser scanning, and a 3D mesh generated from photogrammetry. Each one was produced at a different time, with a different coordinate reference system and a different accuracy target. When they are merged into one scene without checks, the mismatches show up later as buildings that sit half a metre over a property line or roads that appear to float above a junction.
Alignment work starts with agreeing on a single coordinate reference system and a vertical datum. Engineers reproject every layer into that frame, then compare known control points such as surveyed manhole covers, street light bases and building corners. The residual offset at each point is recorded, not just averaged away. A useful habit is to plot these residuals as a map. If they grow steadily toward one edge of the district, the problem is usually a projection parameter or a stale transformation grid. If they cluster around a few isolated spots, the cause is more likely a single misdigitised footprint or a building that was extended after the aerial survey.
Once the baseline is stable, the checks become operational. Cadastral updates arrive every month or quarter, while the 3D mesh may be refreshed only every few years. A parcel that has been subdivided can leave a footprint without a clear owner, and a new building may appear in the mesh before its boundary is registered. A simple rule set helps: flag any footprint that intersects a road polygon, any parcel boundary that cuts through a building by more than a small tolerance, and any mesh object that has no matching record in the asset register. These flags go to the GIS team as a work queue rather than being fixed silently.
The result matters for downstream uses. Flood simulations depend on accurate ground elevation beneath building footprints. Utility planners need to know which side of a boundary a substation sits on. Emergency routing is affected when a fire station's access gate is drawn in the wrong place. In each case, a few centimetres or a few metres of error can change the answer, so the twin should carry metadata about accuracy for every layer, and users should be able to see it when they run an analysis.
For city teams planning a similar project, the practical advice is to budget time for this alignment stage before building dashboards or training models. Choose a small pilot district, document the control points, and agree on tolerances with the departments that will rely on the data. Platforms such as TwinUp's Clavus City handle the layer management and visualisation side, but the quality of the result still depends on the survey discipline and governance behind the geometry.