Problems of Using Boston's Information Infrastructure to Explore Urban Development Potential

Paul Cote
Geographic Information Systems Specialist
Harvard Graduate School of Design
Prepared for GSD6329 February 26, 2002

This paper has been prepared as part of the research seminar, Real Infrastructure for Virtual Cities. The aim of the seminar is to look at the practical issues of building and using real administrative databases as a means of better understanding three-dimensional urban form. In keeping with the program for this seminar, this investigation begins with a practical problem of urban modeling, investigates applications of information systems toward understanding the problem, and then critically examines the issues of implementing the model in the context of a real design firm or public agency.

Formal Model: Development Potential Expressed through Zoning

The zoning ordinance is a type of urban model. It prescribes the ideal potential of a piece of property. The zoning states what may be built and at what densities. In ordinary development practice, what can be developed as of right refers to the type, amount and configuration of development that can be exploited without first obtaining a variance to the zoning ordinance. A site that is under-built with respect to the ordinance can be said to have development potential. A site which is built in excess of what is allowed is termed non-conforming. With nonconforming uses, there is often less likelihood that the site will be developed, because to do so as of right would mean giving up density.

As a prescriptive model, the zoning ordinance is a perfect representation of the ideals that it is concerned with. The actual city, however, is not a perfect reflection of the zoning ordinance. If we can combine information from the ordinance with information about what is actually built we can derive a new model of development potential and development inertia -- the difference between what is permitted and what is built.

The formal model for describing density in zoning is Floor Area Ratio (FAR.) FAR for a given development site is the ratio of useable floor area to the land area of the site. If there are no other considerations (requirements for open space, etc) a FAR of 1 suggests that a one story building may be built that covers the entire site, or a 2 story building may cover half the site.

The maps below are views of the Use Subdistricts and the FAR as designated in the Boston Zoning Ordiance (This provisional zoning data was provided by the Boston Redeveloment Authority)


Data Model: Vector GIS Layers, and Associations

To build a model of development potential we need to augment our model of what is permissible with information about what is built on sites. Unfortunately, the city does not have a perfect representation for this as they do for zoning. Ideally, a site is a property that is developed as a project with zero or more buildings on it. It seems simple: we need a representation of property parcels, representations of buildings, and a means of associating the buildings with the parcels. This sounds easy -- a perfect application of layers already part of the city's vector GIS:

Layers

  • Building Footprints: obtained from the boston water and sewer commission (BWSC). These are more or less just polygons. This layer will be useful for representing the location and plan of each building.

    Unfortunately, the data as provided by the BWSC are not a perfect representation of buildings for the purposes of our model. As illustrated in the image above, buildings are split along the vestiges of a boundary system used in the compilation of the data in individual map tiles. Also, islands inside buildings, which are evidently light-wells or courtyards are also represented as equal polygons in this layer.

  • Property Parcels: the tax assessor maintains several information elements that we need. First, the geometry of parcels gives us a representation of the shape and extent of building sites (a component of FAR.) The assessor also keeps track of the number of stories for most buildings (the assessor does not collect height information for tax exempt property.) Building heights are important for representing the third dimension of the buildings -- necessary for understanding the current FAR.

Transformations and Associations: Zoning, Sites and Buildings

According to our formal model, sites have a prescribed FAR based on their association with zoning. Buildings have an existing height provided through their association with the assessor's parcel layer Buildings have an as-of-right height given by the association of the area and heigt of a building with with the area of the site and the site's allowable FAR. Building our data model requires that we create associations in our GIS to match the ideal associations in the formal model. Because of the city's data resources are not exactly logical, the layers require a substantial amount of transformation in order to produce a model that performs logically.

Assigning Sites and Buildings: Logically speaking, a building is associated with a parcel. When we examine the city's datalayers, however, we see some exceptions. Many buildings straddle parcels due to imprecise locatation of parcel boundaries. Some buildings occupy several parcels owing either to leasing arrangements or the data creator's inability to distinquish building articulations from aerial photos. Before we can associate site information with buildings, we need to clear up these ambiguities:

  • Associating Buildings with Parcels: First, each building polygon can be transformed into a single centroid point. This point will be associated with its building because of a common building ID field in common in their respective attribute tables. These centroids can be associated with a single parcel using a spatial join. Joining the building polygons with their associated parcel attributes gets us the heights for the buildings.

    As a side-effect of the building-parcel association, some of the fragmentary building polygons (discussed earlier) may be welded together with a dissolve transformation where we have multiple building polygons existing on a single parcel. This dissolve operation gets us a single building area on each parcel. Even unattached buildings sharing single parcel become a single record in the dissolved layer's attribute table.

  • Associating Parcels with Sites: Cases where a single building polygon occupies several parcels have to be cleared up because the area of the "site" must be considered with the volume of buildings in order to determine the actual FAR. To achieve this, we derive centroids for the parcels join these spatially with the buildings, and then dissolve the parcels occupied by a single building into sites.

  • Association of Sites with Zoning: Reducing sites to their centroids, and associating these with the zoning map gives us a prescribed land use and FAR for each site.

The dissolve transformations described above yeild a set of layers that logically represent sites: their land areas, their built areas, and the building heights, and their prescribed zoning unambiguously associated. Because the dissolve function permits arithmetic aggregations of attributes as well as geometric summaries of the spatial entities represented in a layer, we can perform another spatial join between buildings and sites to find the actual FAR.

Technical Implementation of the Model

All of the work of assembling a logically consitent set of layers in a data model sets us up to implement a GIS model to explore the questions about Development Potential that we stipulated in our in our formal model statement. With our new set of layers, we may now make the necessary associations and use the model of sites and buildings to calculate new information about developent potential.

The heart of our model for exploring development potential is a single GIS layer resulting from the associations and transformations described above. Each entity represented in this layer is the built area of a particular site. The attributes stored for each of these entities includes:

Site-Buildings Layer

  • Shape: The built area of the site. This geometry can have multiple island polygons representing separate buildings.
  • Site Land Area: The sum of the land areas for all of the parcels that were aggregated to make the site.
  • Building Height: This is the maximum height (in stories) of any of the parcels that were dissolved for the site. We probably should have used some sort of a weighted average here.
  • Site Prescribed FAR: This is simply the FAR for the zone that falls under the centroid of the site.
  • Site Prescribed Use: From the zoning map.

From the 'given' attributes above, the following new information can be derived with simple arithmetic:

  • Prescribed Height: This attribute is derived as 

       (Prescribed FAR * Site Area) / Built Area

    This figure is expressed as stories.

  • Residual Height: the difference between Building Height and Prescribed Height, if negative, is the amount of excess building height for nonconforming sites. If positive, the residual is an indication of unbuilt as of right potential for the site.

Visualizing the Model Results

Here is a view of a portion of the Site-Buildings layer. We are looking north-east at Hannover Street in Boston's North End. This is the street with many fine italian restaraunts and cafes. The color of the base of the building denotes its prescribed use. If there is a green volume on top, this is the amount that could be added to the building without any variance to the zoning. If the volume is red, this is the amount that would need to be cut off to bring the site into conformance.

If our model is to be believed, we can see that most of the business (pink) buildings along Hannover Street are not built out to their full, as of right, density. There are several large multi-use (yellow) buildings nearby which seem to be nonconforming with densities higher than what is permitted by the zoning. The very tall artifacts are evidently building courtyards that found a one-to-one relationship with a parcel before our sequence of aggregations.

Thinking Critically About our Data Model

Our goal is to use our data model to improve our understanding of development potential. Before we go much further, we should pause to consider, after all of these transformations, whether the data is sufficient, what its errors are likely to be, and how the data model might be improved.

Building Geometry

We are representing all the buildings on a site as a single extruded solid. The useful built area is assumed to be an extrusion of this built area according to the story-height. In reality, buildings on a site are not necessarily the same height. Most buildings are not prisms. Furthermore, in actual far calculations, some aspects of a building's layout are traditionally not included (e.g. parking structures and service areas.) Each of these oversimplications would lead to overepresentation in our estimates for the as-built building volumes. This overage will lead to an under-estimation of the development potential, or the nonconforming excess. Determining the impact of these errors would require more accurate data that we don't have. Nevertheless, it will be an interesting model to look at.

Institutional Considerations

The shortcomings of our model arise from inadequacies of the City of Boston's GIS data model. This exploration required a lot of work which would not have been necessary if the city had considered the logical relationships between buildings, parcels and sites. Because of these inadequacies, this model is impractical except as a one-off research project that is unlikely to be repeated, wich means that this is a virtual city, yet it is not real infrastructure If the city's data were logically planned and maintained, this model and others like it could be a routine view if the data.

As the city continues to plan and implement future information infrastructure, it should try to coordinate information used by various departments. If the next generation of the buildings layer includes better three-dimensional massing models (i.e. wedding-cake models) and programmatic floor-plans, this sort of data view might be easy to create, and actually useful!

[an error occurred while processing this directive]