SEMCOG maintains three separate data layers that represent information on built structures in Southeast Michigan: building footprints, building points, and condominium points.

The Building Footprintswere the first dataset collected. These were originally developed by using heads-up digitizing to trace the outlines of buildings in 2010 aerial photography. This process was repeated later using 2015 and 2019 imagery. Due to differences in spatial accuracy between the 2010 and later imagery sources, footprint polygons delineated using 2010 aerial photography may appear slightly shifted compared with imagery that is more recent. This dataset is current as of April 2019, and contains only three attributes: the unique building identifier, median height and the source of the footprint polygon.

The Building Pointsdataset is a point layer representing a single point, at the center of the structure, for each building in Southeast Michigan as of April 2020. For buildings with an associated footprint polygon, the building point represents the centroid of the polygon. For newer structures, a point is placed at the approximate center of the structure as shown in recent imagery, as no footprint polygon yet exists. Since it is the building points that represent our most current universe of buildings in the region, this is the dataset on which we store the associated attributes of a building, mainly location and physical characteristics obtained from local assessing data and other sources.

The Condominium Pointsdataset is also a point layer, with each point representing an attached condominium unit located within a building. Unlike site condominiums, attached condominiums are by definition only a portion of a single building, but as each unit can be individually owned they need to be tracked independently. To ensure an accurate match between condominium points and the building in which they are located, the condominium points are placed on the building centroid, and do not represent the actual physical location of the unit.

Building Attributes

As stated above, SEMCOG maintains building attributes on the building points data layer. The condominium points and building footprints data layers have only a small subset of these attributes. Table B.1 list the current attributes of the buildings dataset:

Table B.1

Buildings Dataset Attributes

1. Attribute only in the Condominium Points layer.

2. Attribute only in the Building Footprints layer.

3. Square footage evenly divisible by 100 is an estimate, based on size and/or type of building, if the true value is unknown.

Building Types

SEMCOG’s building types are based on the 2017 North American Industrial Classification System (NAICS). This system classifies establishments based on economic activity, is widely accepted by the planning community, and used by SEMCOG in developing its long-range socioeconomic forecasts. Each of the 22 nonresidential building types represent one or more entire 3-digit NAICS sector groupings, except for five special types of building that do not fit within this system. These specific types of activity (movie theaters, libraries, dormitories, funeral services, and parking garages) have land use and trip generation characteristics that are different from the other types of activity in their respective NAICS sector and are therefore classified individually. More information on NAICS, including descriptions of the economic activity sectors, can be found at the Census Bureau’s NAICS web site.

There are also four residential building types, which assist us in tracking housing development over time. Residential types are based on a combination of units in structure, and ownership. Single-family homes and site condominiums typically contain only one unit, but there are traditional duplexes where one owner owns both units, that are also classified as single-family. Attached condo buildings contain two or more housing units in a single building, each of which shares one or more walls with an adjoining unit and can be independently owned. Multi-family housing or apartment buildings contain three or more units, all of which are owned by a single entity and intended as rental units.

Figure B.1

Distribution of Buildings by Type

As seen in Figure B.1, 86.9 percent of all buildings are single-family homes, and only 5.8 percent of buildings are of any nonresidential building type.

Mobile homes are a special building type, as we do not receive information on changes to mobile homes in a manufactured housing park through the building permits pipeline. We instead maintain a separate layer of points representing individual mobile homes in parks, updated periodically using imagery to denote if a trailer is still present at each point. Pseudo footprints for these points are generated via code and added to the buildings manually.

There are structures that SEMCOG does not classify as buildings. To be considered a building, a structure must contain either a housing unit or a minimum of 250 square feet of nonresidential workspace. Therefore, garages, boathouses, and other accessory structures are not classified as buildings. A description of SEMCOG building types, including the relevant NAICS codes for the type, can be found in Table B.2:

Table B.2

Building Types by NAICS Sector

The RGB color values match SEMCOG’s typical map symbology for building types.

Relationship between Buildings Layers

County land parcel layers are used to assign the unique APN (Assessor Parcel Number), as well as parcel identifier, to each structure located on a parcel by means of spatial intersection between the parcels and buildings layers. However, there are counties which do not represent parcel splits in their parcel layer when large vacant parcels are developed into single-family subdivisions. In these cases, the polygon geometry of the original parent parcel is cloned once for each new child parcel. A geospatial intersection of buildings to parcels in this situation would lead to each building in the subdivision being assigned to whichever parcel polygon happened to be sitting on top of the stack of cloned parcels. In these cases we use other sources, such as building permits and developer site plans, to identify the APN of each new building, and then assign the parcel identifier by finding the land parcel with that APN.

Figure B.2

Primary Key Relationship of Buildings Layers

Land Parcels	Building Footprints	Building Points	Condominium Points
parcel_id (unique) ----->	parcel_id--------------->	parcel_id --------------->	parcel_id
	building_id (unique) --->	building_id (unique) --->	building_id
APN (unique) ---------->	APN -------------------->	APN -------------------->	APN (unique)
			condo_id (unique)

The APN and parcel identifier are passed to each building on a parcel, with each building assigned a unique building identifier. Since building points are the centroid of building footprints, they share the same building identifier as they reference the same structure. Condominium records are assigned a unique condo identifier, and also the building identifier of the structure they are located within. By this process, each layer has a unique identifier, and can be related to the larger parent geography as needed.

Versions

Dataset Uploaded 05/10/2021 –Added footprint polygons for 5,851 building points that did not yet have a footprint. All building points now have an associated building footprint. Updated addresses for an additional 30 mobile home parks. Removed 640 buildings permitted but not constructed by April 1, 2020.

Dataset Uploaded 01/05/2021 –Added missing buildings between 2,000 and 10,000 square feet in size identified in Nearmap’s building outlines layer. Added mobile home units back into the coverage, and updated addresses for 93 mobile home parks.

Dataset Uploaded 08/18/2020 –Added missing buildings greater than 10,000 square feet identified in Nearmap’s building outlines layer. Updated address and zip code from staff research.

Dataset Uploaded 07/08/2020 –Added buildings constructed from 4/1/2015 through 4/1/2020 to the dataset. Updated missing or default square feet and year built from assessing rolls.

Last Update: 05/10/2021