A challenging future faces many of our post-industrial cities. One of these challenges concerns the ability to successfully address their deteriorating housing infrastructure. The City of Philadelphia faces just such an issue today. With approximately 8,000 buildings that are imminently dangerous and an additional 60,000 vacant lots and buildings, the City has turned to GIS technology to assist its Neighborhood Transformation Initiative Project to schedule more than 14,000 demolitions over the next 5 years. A Decision Support Model employs raster modeling to develop and assign priority to decision-making criteria. The model then generates a demolition schedule-informing output for this massive undertaking.
Philadelphia, like many other old cities, faces the challenges of depopulation, economic decline, and deteriorating housing infrastructure. In May 2000, Mayor John F. Street launched the Neighborhood Transformation Initiative (NTI) to address blight caused by dangerous buildings, abandoned property, and vacant land and to rebuild and revitalize neighborhoods. Neighborhood transformation goals include eradication of blight; citywide community planning; abatement of public nuisances; improved ability to acquire and redevelop vacant property; more efficient use of public resources; and increased private investment in neighborhoods. [1] “Demolition and Encapsulation of Vacant and Deteriorating Buildings” project involved by the authors was started in October 2001 as the fist phase towards these goals (see end notes). In May 2002, Mayor Street signed into law a bond ordinance that authorizes the City to issue bonds worth approximately $295 million to pay for blight elimination, redevelopment, land assembly, housing investment and neighborhood preservation activities. With NTI bond proceeds, over the course of the next three to five years, the City will demolish 14,000 dangerous buildings to protect the health and safety of the public and facilitate large-scale land assembly for redevelopment. It will also preserve 2,500 less deteriorated structures for renovation and reuse, and provide home repair grants and loans to homeowners. [2]
In January 2001, inspectors from Licenses and Inspections department (L&I) and the Philadelphia Housing Authority (PHA) completed a comprehensive survey of vacant properties throughout the City. The Survey identified a total of 26,115 vacant residential buildings three fourth or 8,500 of which were determined to be dangerous or imminently dangerous and in need of demolition. In addition, inspectors identified 2,950 vacant commercial buildings and 30,729 vacant lots. [1]
How can such a large-scale urban demolition project be done? How to prioritize and schedule this massive demotion process? Facing the nature of this project, complex decision making process and inevitably multiple points of view (most with political influences), the project team used GIS, not merely as a useful tool for displaying and mapping, but as a foundation for a Decision Support System (DSS) to assist the city to go through this enormous endeavors.
Listed below are a number of steps or tasks involved with GIS.
Task 1 Identifying Preliminary Bid Package
1.1 Print maps (24x30) of administration-defined scenarios for Council District coming to GIS Decision Support Model (DSM) meeting
1.2 Conduct Decision Support Model meeting with Council District representatives, develop and run new scenarios and delineate multiple preliminary bid package areas
1.3 Print maps (24x30) of preliminary bid package areas
1.4 Generate and print tabular reports for each preliminary bid package area that list addresses of dangerous buildings, vacant lots, lots with vacant buildings on them
1.5 Council District chooses preliminary bid package area and/or runs new scenario through DSM to identify preliminary bid package area
Task 2 Initial Field Inspections
2.1 Print one large context (34x44) and small (8.5x11 for each block) existing condition maps for bid package area field inspection
2.2 Field inspection of bid package area and record errors in parcel delineation, building condition, vacancy, ownership map data
2.3 Photo-documentation and block image compilation for each block with bid package
2.4 Edit and update parcel, building condition and vacancy databases
2.5 Generate field verified maps (8.5x11)
2.6 Generate address lists of potential demolition sites within bid package area
Task 3 Identify Bid Package Candidates and Assign Treatment
3.1 Import digital block photos and link to parcel database
3.2 Decide building treatment and determine demolitions/encapsulations within bid package using GIS in the community meeting.
3.3 Record preliminary treatment for buildings as assigned during the meeting
3.4 Determine final treatment for buildings, incorporate treatment listings into Treatment database
3.5 Generate reports listing for bid package acquisitions, encapsulations, demolitions, side yard acquisitions, in alphabetical order
3.6 Create maps (8.5x11) displaying final building treatments within bid package
This paper will discuss how we use GIS to develop a GIS Decision Support Model (DSM) and Decision Support System (DSS) for targeting the priority areas for demolition.
The DSS uses the city’s existing digital data, evaluates both demolition/stabilization needs and redevelopment/rehabilitation opportunities, speeds the decision-making process and produces maps of potential properties. The DSS permits decision-makers to weigh various criteria so that different criteria may be applied to different neighborhoods and different scenarios can be compared and evaluated.The GIS Decision Support Model employs Raster Modeling techniques. Basically it is an overlay of different data layers with corresponding weights. The data layers or variables used in the model are those factors that might influence the decision making of demolition and redevelopment and that are readily available or can be derived from the existing data sources. We consider only those variables that are available because this is a practical solution rather than a theoretical research of an old city problem. Other variables that might also be important and available but probably not significant for the city may also be excluded, e.g., the public transportation network is excluded because the city is covered extensively by the public transit. Overall, we have chosen ten variables or criteria and their causal relationship with the demolition priority are illustrated in the following table.
| Criteria | Causal Relationship and Demolition Prioritization | |
|---|---|---|
| 1 | Condition of Vacant Structures | The greater the density of dangerous and imminently dangerous structures, the greater the need for demolition. Higher density = higher priority, Lower density = lower priority |
| 2 | Number of Vacant Structures/Block | The greater the density of vacant sites, the greater the need for demolition. Higher density = higher priority, Lower density = lower priority |
| 3 | Social/Cultural/Economic Anchors | The closer a vacant site is to a community anchor, the easier it will be to redevelop. Close to anchor = higher priority, Farther from anchor = lower priority |
| 4 | Ownership Characteristics | The greater the neighborhood stability, as a function of owner-occupied structures and ownership longevity, the easier it will be to redevelop. Higher stability = higher priority, Lower stability = lower priority |
| 5 | Parcel Characteristics | The smaller the contiguous vacant area and the narrower the cartway, the more difficult it will be to redevelop. Optimal layout = higher priority, Poor layout = lower priority |
| 6 | Elementary Schools | The closer the site is to an elementary school with high achievement, the easier it will be to redevelop. Closer to strong elementary school= higher priority, Farther from strong elementary school= lower priority |
| 7 | Home Sales | The greater the value of neighborhood home sales is above the regional average, the easier it will be to redevelop. Higher value than regional median = higher priority, Lower value than regional median = lower priority |
| 8 | Population Change | The greater the negative population change, the easier it will be for demolition. Greater depopulation = higher priority, Lower depopulation = lower priority |
| 9 | Proximity to noxious landuse | The closer a vacant site is to a noxious landuse, the more difficult it will be to redevelop. Farther from noxious land use = higher priority, Closer to noxious land use = lower priority |
| 10 | Market Quality | The closer the site is to an existing or proposed reinvestment project, the easier it will be to redevelop. Higher market quality = higher priority, Lower market quality = lower priority |
The importance of each variable can vary for different objectives and from the points of view of different people. That is why we introduced weights in the model. Figure 1 illustrates the model.
The city of Philadelphia has maintained a wealth of digital data, we have obtained from the Mayer’s Office of Information Services digital data layers such as parcels, census tract, census block groups, streets, water body, schools. Dangerous buildings and vacant structure data is maintained by the City of Philadelphia Licenses and Inspections (L&I) Department in Microsoft Access. Bureau of Revenue and Taxes (BRT) maintain a database of public owned buildings in tabular format as well. There are four public entities that might own a building, the City, Philadelphia Housing Authority (PHA), Philadelphia Industrial Development Council (PIDC), Redevelopment Authority of the City of Philadelphia (RDA). Housing Market data was obtained from Redevelopment Fund. The first task undertaken is to geocode the address list of dangerous and vacant buildings and public owned properties. We would first geocode the addresses to parcel layer so that each address will be exactly geocoded to the centroid of its parcel. Since parcel data is unlikely or practically impossible to be 100 percent accurate and consequently some addresses may not be geocoded successfully. We have an alternative to geocode addresses to street layer. This type of geocoding is not quite accurate since an address is geocoded proportional to the potential range of all addresses along the street and that range is often not accurately reflect the reality. For example, a street with address range of 100 to 199 on one side may only has five houses 100, 102, 104, 120, and 122 with address number 100 and 122 be the end buildings. The geocoding, however, will locate all houses towards the 100 end of the streets. Even though this type of geocoding is not quite accurate, at large scale, e.g., at councilmanic district or city level, this inaccuracy has minimal impact on the overall results.
The geocoding of these addresses is very successful. For example, for dangerous buildings, there are 21,451 Records in Dangerous Building Database, 12,851 have already been demolished; 811 have already been awarded for demolition; 7,781 dangerous buildings are pending demolition. For these 7781 Dangerous Buildings, 6767 or 87% were able to be geocoded to parcel layer, 946 of 1014 of those couldn’t be geocoded to parcel layer (or 12% of total) were geocoded to street layer, only 68 or 1% were not geocoded to either data layer. For vacant structures, there are 59,796 vacant land lots or structures in the database, 52,837 or 88.4% were geocoded to parcel layer; 6,584 of 6959 (or 11% of total) were geocoded to street layer. There are only 375 (or 0.6%) unmatched records.
To use GIS DSM, we need to convert vector data to raster surfaces. There are a number of ways to do this. For example, the value of a cell in a raster surface can be derived from the attribute value of the corresponding feature that is located in that cell, or from distance to a particular feature. There are also a lot of ways to reclassify or process a raster surface into new ones. For example, arithmetic and algebraic functions can be applied to a raster data layer to produce new layers. In this particular project, we have used several ways to generate raster surfaces for ten criteria, such as converting from attributes of the corresponding vector features (for ownership characteristics, home sales, household changes, market quality); or from attribute values of the closest feature (cartway); or from density (dangerous buildings, vacant structures); or distance (elementary schools, proximity to industrial land); or a combination of distance and size (for anchors). All raster surfaces are reclassified to values of 0 to 10 so that they can be combined regardless of different units in the original data set.
A GIS Decision Support System is developed to assist decision makers to use the DSM. It allows decision makers to assign weights to each variable. The system will then run the model and display the results within 2 to 3 minutes. Esri’s ArcGIS, ArcObjects, and Microsoft Visual Basic are used to develop this system. After a decision maker assign weights through a simple and easy to use Graphic User Interface (see figure 2), the system runs the following processes behind scene:
The building list can then be imported to a database to generate demolition schedule.
The resulting raster surface has a 12 by 12 feet resolution. The whole city is covered by 7643 by 8358 or 63,880,194 grid cells. This resolution is able to differentiate each parcel. Decision makers are able to look at the map that shows detailed information such as parcel outline, condition (dangerous or unsafe), vacancy (whether it is a vacant lot, a vacant residential building, or a vacant commercial building) and whether or not it's a city-owned property, etc. They also have the opportunities of looking at the digital orthophoto of these properties.The system works very well. In the meetings with city council members and other interested groups to delineate preliminary bid packages, we have asked them who are knowledgeable of the local issues to think about the real problems in their council districts; what neighborhood they think should be the priority areas for demolition; what objectives or goals are desired for these communities after demolition; and how these objectives are influenced by ten criteria in the DSM. With these things in mind, they can easily understand and differentiate ten variables and weigh them according to their importance. For example, for a deteriorating neighborhood that you want to assemble more land for redevelopment, you would put more weight on the criteria of dangerous buildings and vacant structures; to identify those neighborhood that you want to preserve, you may want to put more weight on ownership characteristics, social, economic, and cultural anchors, elementary schools, etc. Amazingly, every time, the model pick up the communities they think should be the priority areas, which validate their intuitions and also validate our models.
In the demolition, in addition to 8,500 dangerous buildings, we may want to demolish also vacant structures that are adjacent to these dangerous buildings because large scale “string demolitions” (clearing sizable parcels of land by removing large numbers of adjoining properties) is more economical or practical.
Those adjacent buildings can be properties that are immediately next to a dangerous building, or separated by a narrow alley or driveway, or separated by other dangerous buildings or vacant structures. How can we automate this process of incorporating vacant structures in the final output of demolition list? Again we turned to GIS. GIS allows us to analyze the continuity and connectivity of features. Here is how we resolved this problem. The key is to generate a dangerous building to dangerous building or vacant structure index table.Once we have this index table and the address list from the DSM output, we use a database to sort buildings by score, join dangerous buildings to be demolished with adjacent dangerous buildings or vacant structures, remove duplicated records, and export final address list. Figure 3 illustrates the diagram of this modeling procedure.
We have also experimented automatic generation of bid packages directly from DSM output. This comes when there is interest to automatically generate bid packages for all 10 council districts in the city, say, 5 top priority bid packages for each council district with each bid package has approximately 80 dangerous buildings. We use GIS to resolve this problem as well. The final scores obtained from DSM for each dangerous building can be aggregated into census block groups, so that each block group has an average score value and the number of dangerous buildings in it. Then we sort these block groups by council district and the score, starting from the highest score value, we aggregate continuously with the block group that has the next highest value until the limit of 80 buildings is reached. We then begin with another bid package. In this way, we automatically generate all 50 bid packages in a few minutes. The same work without a GIS would takes days or weeks.
Using GIS and GIS DSM, we have successfully met the needs of prioritizing a massive demolition project. GIS allows us to do the planning job faster, reliable, more cost effectively, and comprehensible by local government agencies, legislative bodies, and special interest groups. The same concept we implemented in Philadelphia can be applied to other cities with the similar problems. It can also be employed in other planning projects that involve complex decision making, different objectives, and points of view.
The authors want to thank Office of Neighborhood Transformation Initiative, Mayor's Office of Information Services, Philadelphia City Planning Commission, City of Philadelphia Licenses and Inspections Department, Redevelopment Fund for providing valuable digital data.
Demolition: Demolition is the complete removal of a dangerous or vacant property and the disposal of the materials with which the property was built. L & I demolishes an average of 1,400 dangerous residential buildings per year. During fiscal years 1990 through 2000, the City demolished 14,185 abandoned and imminently dangerous structures throughout the City. [1]
Encapsulation: Encapsulation is the total sealing of vacant properties that are not in danger of collapsing. These properties are protected from water damage, cleaned and sealed. It may also include some stabilization treatments such as roof replacement and repair of drainage systems.
Bid Package: A bid package is a designated area for demolition (usually has about 150 buildings to be demolished) and open for public bid for contractors to actually do the physical demolition.
[1] CITY OF PHILADELPHIA, CAPITAL PROGRAM OFFICE, PROJECT # 07-01-4371-99, REQUEST FOR PROPOSAL FOR PROGRAM MANAGEMENT SERVICES FOR DEMOLITION AND ENCAPSULATION OF VACANT AND DETERIORATING BUILDINGS, March 14, 2001, accessed from City of Philadelphia Neighborhood Transformation Initiative (NTI) web site http://www.phila.gov/mayor/jfs/mayorsnti/pdfs/NTI_RFP.pdf on 6/20/2002.
[2] MAYOR'S OFFICE OF COMMUNICATIONS, News Release, “Official Signing of NTI Legislation”, May 13, 2002, accessed from City of Philadelphia Neighborhood Transformation Initiative (NTI) web site http://www.phila.gov/mayor/jfs/mayorsnti/news/releases/releases_0.html, on 6/20/2002.