Manhattan Residential Clustering

Manhattan's residential buildings can be described by characteristics such as construction year, height, and transit access. Eleven building- and street-block-level datasets were combined, merged at the building level using the BIN (Building Identification Number) as the key, so each building can be reasoned about across form, value, and access at once.

The scope is Manhattan's residential stock: Tax Class 1 (mostly residential property of up to three units) and all other primarily residential property outside Class 1.

The dataset was assembled from several public sources and merged at the building level on the BIN. Records provided at the tax-lot level were expanded to individual buildings, and building attributes were joined to street blocks for neighborhood-scale context. Subway travel time was computed separately in Rhino + Grasshopper as a shortest-path analysis to the nearest station.

The eleven variables, their sources, and headline statistics are summarized below.

Each layer is a building-footprint map of one variable, aligned to Manhattan on the basemap below. Switch layers to see how form, value, and access vary across the island.

K-means clustering was applied to the assembled features. Variables were normalized and given custom weights so that the most relevant factors drove each grouping. Four weighting schemes were run, each emphasizing a different lens, and the resulting groups were interpreted qualitatively and given descriptive, context-specific names.

Four clustering lenses each tell a different story about the same building stock. Switch between them to compare how the weighting changes which patterns emerge.