Urban Morphology & Figure-Ground at Scale

A figure-ground map fills the buildings solid black and leaves everything else white. Strip away colour, labels and land use, and what remains is pure form - the relationship between built mass (the figure) and open space (the ground). At a glance it tells you density, continuity, the width and rhythm of streets, and whether public space is shaped and contained or merely left over.

The trick that made it a design instrument is the Nolli map (Rome, 1748): invert it, so the public void - streets, squares, and even the accessible interiors of churches - becomes the solid figure. Suddenly you read the city as a system of connected rooms, which is how a pedestrian actually experiences it. A GIS makes both versions trivial from a single building-footprint layer, at any scale from a lane to a metropolis.

Morphological analysis - the Conzenian tradition of urban geography - reads the city as a hierarchy of nested elements, each shaping the next. The street network divides land into blocks (the island of land bounded by streets). Each block is subdivided into plots (the individual land parcels, the unit of ownership and development). Each plot carries its building(s). The way these four fit together - the town plan in Conzen's terms - is remarkably persistent: streets and plots outlive the buildings on them by centuries, which is why a medieval plot pattern can still shape a modern high street.

For a designer this hierarchy is practical. It tells you the unit of change: a fine-plot street changes one building at a time and stays diverse; a superblock changes only wholesale, which is why superblocks so often flip from empty to mega-project with nothing in between.

Grain is the size and repetition of the morphological units - fine grain means many small plots and buildings, coarse grain means few large ones. It is the single most useful word in morphology, and a GIS lets you replace the adjective with figures.

From a building-and-plot layer you can compute, per block or per hectare: the building coverage ratio (footprint area / plot or block area - how much of the ground is built on); the median plot size and its spread; the number of buildings per hectare; and the frontage grain (buildings or plots per 100 m of street). Add height (from a DEM difference, LiDAR, or attributes) and you get floor space index / FSI and a fuller three-dimensional picture. Cluster blocks by these numbers and distinct typologies fall out - bazaar street, gated enclave, industrial shed, villa plot - each a recognisable, measurable signature rather than a vibe.

The bottleneck used to be data; it no longer is. Open, machine-learning-derived building footprints now cover most of India: Microsoft's GlobalMLBuildingFootprints and Google's Open Buildings (both free, with Overture Maps packaging building data too), plus OpenStreetMap where mappers are active. Any of these gives you a footprint layer to build figure-grounds and grain metrics at city scale in an afternoon.

Two honest cautions. First, quality varies: ML footprints can merge adjacent buildings, miss small or tin-roofed structures, and lag new construction - so ground-truth before you trust block-level counts. Second, the scale trap, formally the modifiable areal unit problem (MAUP): the grain numbers you get depend on the unit you measure over - per block, per hectare, per ward - and can change, even reverse, if you change that unit. Fix your unit deliberately and report it; a coverage ratio without its denominator is meaningless.