From Footprints to Massing: Engineering a Citywide 3D Model of New York

New York City's citywide massing model, built for the Department of Information Technology and Telecommunications with AppGeo, represents a vast improvement over the initial attempt to create an extruded 3D model of the city. The groundwork involves using planimetric mapping for building footprints, separate tables or rasters for heights, and a digital elevation model for terrain. The model uses US survey feet for horizontal units and feet for vertical units, which can be inconvenient when confronted with files in meters.

Extrusion is a straightforward process involving sampling the ground elevation, adding the height value, and closing the sides. However, defining the meaning of height is crucial, as it could refer to above ground level, sea level, the roof line, or the top of the mechanical penthouse. The choice of height definition needs to be documented, as discrepancies of ten feet can lead to different shadows at different times of the day.

Building heights can come from various sources, including surface models with parapets, bulkheads, and antennas, or attributes reflecting the roof deck. A single tower might have multiple plausible heights, making it difficult to choose the correct one. To address this, it is important to document the source of each height and treat the resulting error as a range rather than a hard fact. This approach acknowledges the uncertainty and helps avoid misleading assumptions.

Dividing the model into smaller tiles allows for more efficient rendering and storage. Using a stable ID for every footprint throughout the different levels of detail ensures consistency when joining datasets. Shadows provide valuable insights into the model, revealing the impact of the terrain and the surrounding environment. Accurate computation of shadows requires a clear sun vector, date, time step, and involvement of neighboring buildings.

Errors in the model are often due to join issues, such as footprints without heights, duplicate IDs, or demolished buildings still present in the data. These errors can make the result look normal, making them particularly insidious. A proposed building may not appear in the footprint file, leading to issues when discussing projects that have not yet been built. In such cases, rough volume estimations can be generated for screening purposes, but they should not be mistaken for actual survey data.

Automated checks help catch most of the major issues, such as heights outside a plausible range, footprint area discrepancies, buildings intersecting terrain, duplicate identifiers, and buildings taller than the tallest recorded tower. Nevertheless, it is essential to supplement automated checks with manual inspections to identify more subtle errors.

Storing height provenance per building from the start, publishing a confidence flag with the geometry, and versioning the height table separately from the footprints can greatly improve the overall quality of the model. Keeping a detailed change log that records what changed and when further enhances the model's reliability and trustworthiness.