Integrated Transportation and Land Use Models (2018)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

26 Chapters 4, 5, and 6 discuss three different types of land use models that have significantly different model structures, data requirements, and analytical capabilities. Sketch planning models (described in this chapter) are commonly used to analyze alternative land use scenarios visually and to quantify the area (in square feet) needed to accommodate certain growth scenarios. Chapter 5 describes microsimulation discrete choice models that commonly are based on random utility theory and tend to enhance the various scenarios that can be analyzed. Chapter 6 describes models that use the spatial input-output model approach. Originally based on input-output mod- els, such models assume an equilibrium among land use distributions, accessibilities, and prices. 4.1 Model Concept Sketch planning models sometimes are called rule-based models. Rather than engaging complex equations, sketch planning models more often use rules (such as “areas near transit stations would be developed before other areas are developed”). Although such rules largely simplify location choice behavior and hide some heterogeneity in preferences, habits, and levels of information, the power of such models is their relative simplicity. Although some micro- simulation discrete choice and spatial input-output models have been accused of being “black boxes” that are impossible to comprehend, sketch planning models tend to have clearly documented rules of development that often are comprehensible even by laypersons. Moreover, such rules can be changed quite easily, allowing users to run scenarios under varying assumptions. Sketch planning models are widely used to analyze the distribution of land uses. Hopkins and Zapata (2007) provide an introduction to this type of model. These models typically run quite fast, which makes them powerful tools for analyzing a wide range of scenarios within a short period. Often, such models have been used in town hall meetings, where participants develop scenarios and the effect can be visualized almost immediately. However, not all sketch planning models run as quickly, because some conduct complex spatial analysis. Sketch planning tools are primarily driven by correlation or univariate elasticities, but tend to be unable to reflect the interconnectiveness among different urban phenomena (Wang et al. 2018). By design, sketch planning models simplify this interdependence in the urban system with the benefit of reduced model complexity and run time. Another distinction is that sketch planning models typically use a land unit, such as a zone or a parcel, as the unit of analysis. This land unit has a certain land use, which may be changed by the model. Behavioral models described in Chapters 5 and 6, by contrast, use households, persons, firms, and/or employees as the unit of analysis and model their behavior, which may lead to land use changes. Numerous sketch planning models exist. Frequently used models include CommunityViz (Orton Family Foundation and Placeways 2010), Envision Tomorrow, I-PLACE3S, UPlan and What If ? (Klosterman 1999). C H A P T E R 4 Sketch Planning Land Use Models

Sketch Planning Land Use Models 27 4.2 Interviews Two agencies were interviewed regarding use of sketch planning models. The Sacramento Area Council of Governments (SACOG) in California used to work with the I-PLACE3S model and is implementing Envision Tomorrow, another sketch planning model that has been under development for more than 10 years. The second agency is Triangle J Council of Governments (TJCOG) in North Carolina, which has implemented and actively uses CommunityViz. The two agencies are comparable in size. SACOG covers 22 cities in 6 counties with a population of 2.4 million while TJCOG stretches over 10 counties with a population of 1.7 million. What If? is one of the most frequently cited sketch planning models. Its developer, Dr. Richard Klosterman, a retired professor from the University of Akron in Ohio, was contacted by the author to learn about successful implementations. According to Dr. Klosterman, the model has been implemented several times by academics, but has not been widely used in practical applications. His paper on What If? (Klosterman 1999) has been cited over 400 times and, as such, has had significant effect on the development of other models, but the effect on practical applications has been limited. A recent implementation by Dr. Chris Pettit, Professor at the University of New South Wales in Sydney, had some effect on urban planning for the city of Perth in Australia (Pettit et al. 2015). Perth’s Department of Planning implemented scenarios and analyzed results provided by the What If ? model. However, the model was not integrated with a transport model and, therefore, is out of scope for this report. 4.3 Model History SACOG In the early 2000s, the SACOG Board was concerned because congestion levels were fore- cast to increase by more than 60%. Initially, alternative transportation investment strategies were tested. This was an important step to show the limits of transportation policies alone. In 2004, SACOG decided to explore different land use scenarios in addition to reducing growth in congestion. This was a difficult step for SACOG—traditionally, the agency had not considered alternative land use scenarios—but it was recognized that the quality of living might deteriorate significantly if growth in congestion were to happen as forecast by the transport model. Previously, forecasts were done manually, making it almost impossible to repeat the forecasting effort and create the same forecast. Implementing a land use model allowed SACOG to forecast land use more systematically and to make this process transparent. In California, the four largest MPOs, including SACOG, were recommended to implement PECAS (compare Chapter 6) as the land use model of choice for large MPOs and the state of California. At SACOG, PECAS was explored briefly. Results appeared difficult to interpret, and it was concluded that SACOG did not have the resources to bring the PECAS model to a level of sophistication that was suitable for their needs. Hence, PECAS was abandoned in an early stage, and a sketch planning model was implemented instead. Initially, a model called PLACE3S was implemented, which later was replaced by an online version of the same model called I-PLACE3S. Once the model was online, it took SACOG a few years to achieve stable model results, but the model was used during that time for forecasting, nevertheless. I-PLACE3S was also briefly used by the San Diego Association of Governments (SANDAG) and some SACOG member cities. The I-PLACE3S model uses indicators to model land use development, such as employment by type, residential use, and density and environmental constraints. These indicators are used to calculate the “Return on Investment,” which equals the developer yield and is used as a prob- ability for land to be developed. In contrast to some other sketch planning models, the focus of the I-PLACE3S model has solely been land use data forecasts and not visualization.

28 Integrated Transportation and Land Use Models SACOG’s I-PLACE3S model was used extensively in about 40 public workshops during the Blueprint Planning Project to discuss how growth could be managed. During these public meet- ings, I-PLACE3S was linked to a simplified transport model that could be run in 45 min. At the beginning of a public meeting, a forecast could be developed with the public, and results were available toward the end of the meeting. This public presence of the model also helped elected officials would embrace the tool. Because land use forecasts were developed in a public and transparent process, the SACOG Board accepted the Blueprint Planning growth strategies quickly. Communities of SACOG voluntarily accepted the plan, commending the transparent development process. Today, many smart growth strategies have been introduced to SACOG’s Metropolitan Transportation Plan, which is released every 4 years. Through the course of several iterations, land use forecasts have moved toward more environmentally sustainable growth scenarios. Even though I-PLACE3S is a powerful tool and has been used to inform decision making, the agency recently stepped away from this model. The original PLACE3S was open source, but the conversion to an online platform (I-PLACE3S) made the code proprietary. It was a concern that model improvements could not be done in house and that SACOG was dependent on a single consultant to maintain the model. SACOG was able to use the model in house, but consultants’ support was needed to maintain the server, to update the ArcGIS version, and to revise parts of the model. For example, when a public health model was added, SACOG relied on the consultant for implementation. Staff of the agency decided that it was unsustainable in the long run to rely on one particular consultant for model updates. In the interim, SACOG moved all data processing steps to ArcGIS, where land use data is processed manually for the most part. This work is done entirely in house with no need for consulting. SACOG staff briefly explored using the UrbanFootprint model, but thought that this model did not perform at the level needed within the available budget. Although UrbanFootprint is open source, it was thought that it would take too much time to set up from scratch. SACOG is preparing to implement Envision Tomorrow (http://envisiontomorrow.org/). It is not clear at the time of the writing of this report whether this model can be implemented com- pletely in house or if support by a consultant is needed. The model allows, however, operation independent of external consultants, which was a high priority for SACOG. TJCOG TJCOG started in 2011 to implement a sketch planning land use model—several years later than SACOG. Similar to the SACOG region, forecasts were done manually, with little trans- parency of the forecast-generation process or consideration of the feasibility of replicating a forecast. Generating alternative scenarios with a manual process tends to rely on broad assertions with limited attention to underlying development foundations. It was challenging to coordi- nate growth assumptions between the two separate MPOs in Raleigh and Durham, which are geographically separated by one of the busiest highways in the region. In contrast to SACOG, TJCOG started with a sketch planning model as the initial land use model. It was decided that a simple tool was needed to reconcile different land use forecasts and make the forecasting effort transparent. This task could be accomplished with a sketch planning model. TJCOG decided to use CommunityViz (http://communityviz.city-explained.com/communityviz/index.html) to generate forecasts for transportation planning. The model was used to develop the land use forecasts for the Long-Range Transportation Plan adopted in 2013 (called the 2040 Plan). Local governments and MPOs provided input data, particularly zoning plans and parcel data. For all 700,000 parcels in the study area, the

Sketch Planning Land Use Models 29 development status [developed, undeveloped, under-developed, redevelopable, or committed] was identified. The preferred forecast was already identified by the end of 2012, roughly 18 months after model implementation started. The model was implemented by the model developer, although the staff of TJCOG coordinated meetings and data collection. An update of the model was largely done in house. The model can be used by MPOs and council members, and it is possible for trained personnel to run the model without consultant support. Summaries for each jurisdiction provide an overview on the development of single-family residential units and multi-family residential units as well as employment by five types. This growth is compared with growth capacities in each jurisdiction. Dot density maps help convey patterns of total employment and total housing. It is also possible to look at grid cell level output, which some local governments do regularly. Today, the model helps build confidence in both model assumptions and model results. The entire model is based on parcel data, so model users know the development capacity in every part of the study area. The model structure makes it easy to add or change assumptions on land uses and densities. For TJCOG, it has been particularly helpful to use a fully developed off-the-shelf product that has credibility and an active base of users. 4.4 Model Implementation and Application Effort SACOG SACOG initially implemented the model for the base year 2002 and has since moved the base year to 2005 and later to 2008. Scenarios are implemented for the future years 2035 and 2050, along with a few interim years for selected scenarios. The California Energy Commission funded the transition from PLACE3S to I-PLACE3S. SACOG has eight staff members who mostly work on land use analyses, which amounts to approximately six full-time equivalent employees. It is anticipated that after moving to Envision Tomorrow, essentially the same level of effort will be required to maintain the model. TJCOG TJCOG has to complete land use modeling work with substantially fewer staff members. One GIS specialist spends about 60% of his working hours on model application, and one more senior person works on CommunityViz approximately 1 day per week. In total, model main- tenance requires about one full-time equivalent employee. Operating a complex model such as CommunityViz with such small staff resources has only been possible because of substantial involvement of local governments, MPO staff, and other partners. This active engagement has the added benefit of providing more eyes to review assumptions and methods and is designed to encourage the model’s use beyond long-range transportation planning Every user of the CommunityViz model has to pay the annual license fee of about $1,000. This also applies to member agencies that would like to run separate scenarios in house. Staff time is a larger expense than this license fee. The base year for the region’s CommunityViz model is 2013, which is also the base year of the region’s transport model. To date, model runs have been done for the horizon year of long-range transportation plans (2045), with interim years being interpolated. A planned improvement is to first allocate to 2025 to better capture development and transportation projects in the pipeline and then a subsequent allocation out to 2045.

30 Integrated Transportation and Land Use Models 4.5 Land Use/Transport Model Integration It is rare for sketch planning models to be tightly integrated with transport models. Although there is no technical reason for this manual coupling, the main purpose of sketch planning models often is large-scale visioning. Tighter integration methods are commonly used in mathematically more advanced models as those described in Chapters 5 and 6, although even some of the advanced models use manual data transfer only. To assess the acres needed to accommodate a certain amount of population growth under various density assumptions does not necessarily need the integration with a transport model. Many sketch planning models, however, are (loosely) connected to transport models to assess the resulting travel demand under various land use scenarios. In most cases, the resulting travel demand is not fed back as accessibility to the land use model. At most, particularly high levels of congestion may lead to adjustments of growth scenarios in certain areas, but this commonly requires the manual intervention of the analyst. Accordingly, neither SACOG nor TJCOG has implemented direct feedback from transport to land use. Land use forecasts of the sketch planning model are used in the transport model, but resulting travel times are not fed back to the sketch planning model to reassess the attractiveness of zones for housing and commercial development. SACOG SACOG occasionally has adjusted the land use forecast manually if congestion in certain areas was too severe. Because I-PLACE3S is an online tool, land use forecast adjustments are fairly easy to implement, once the bottleneck in the transport system has been noticed. The output of I-PLACE3S has been fed to a four-step transport model as well as an activity- based transport model. Today, these land use forecasts are used in every model run of the activity-based model, which is widely considered to be a state-of-the-art travel demand model. Both the sketch planning model and the two transport models use the identical zone system. In addition, I-PLACE3S was linked to an energy consumption model. TJCOG The CommunityViz model of TJCOG is linked to an aggregate four-step transport model. The transport model is a state-of-practice model, in which the more traditional gravity trip distribution model was replaced with a destination choice model. In the future, it is anticipated that the CommunityViz model will also be used by water and sewage providers to estimate demand for water and sewage supply. Although the TJCOG model does not–as is common for sketch planning models–provide direct feedback from the transport system to the land use forecast, some transport aspects are considered in the sketch planning model nevertheless. The land use forecast is based on 15 fac- tors used in the suitability analysis to assess the probability that land is developed. Seven of those factors are related to transport, such as distance to transit or distance to the nearest highway interchange. For this analysis, CommunityViz uses the same networks as the transport models. If a new transit line is built, there is some effect on land use by making zones near transit stops more attractive for populations to locate. More direct feedback to the land use model is not envisioned in the near future. Instead it is argued that land, for example, near Interstate 40 will be developed, even if this interstate is congested. It is argued that developers care more about proximity to Interstate 40 than about the level of congestion. Although this is correct to some degree, proponents of more advanced mathematical models tend to argue that there is a relative difference to develop near a congested highway versus developing at locations with less

Sketch Planning Land Use Models 31 severe congestion. For sketch planning models, however, it seems acceptable and appropriate to simplify this feedback. 4.6 Model Application SACOG The I-PLACE3S generated plausible forecasts for the SACOG region once it was calibrated properly and input data were sufficiently reviewed. No manual adjustments of model output were necessary. Some processing of the sketch planning output data was necessary to convert land use categories into socioeconomic data for the activity-based model. But these data conver- sions are rule-based and fairly simple. For example, the activity-based model requires medical employment as a separate employment category, which is a level of detail that currently can- not be provided by the I-PLACE3S model. A postprocessor separates medical employment and allocates it to zones with a higher probability in the vicinity of hospitals. As another example of conversion, housing units provided by the I-PLACE3S model need to be translated into a syn- thetic population for the activity-based model. In scenarios, different types of land use patterns were tested, including low-density and smart growth development patterns. Initially, low-density growth without redevelopment was used to accommodate forecasting growth, resembling a trend forecast from the 1990s. Then, alternative futures with smart growth strategies were tested, including redevelopment and denser develop- ment, among others. These policies reduced travel demand. Through various model runs, it was attempted to identify the most sustainable development. Based on these scenarios, the following principles for the Blueprint were developed: diverse housing types, higher densities, mixed use, job-housing balance, and quality in design. Because scenarios modeled with I-PLACE3S provided a strategy to reduce growth in conges- tion, the SACOG Board was very interested in model results. There is general support and buy-in from upper management, and the sketch planning model has influenced decision making in the SACOG region. By using the model, forecasts became less arbitrary and less driven by particular interests. Instead, decision making has become data-driven. I-PLACE3S runs almost instantaneously. The time-consuming aspect is the preparation of scenarios. Depending on the complexity of a scenario, it may take a few weeks to prepare input data for a new scenario. TJCOG The TJCOG implementation of CommunityViz starts with parcel-level data, aggregated to 100,000 grid cells for the entire region. Grid cells are variably sized, with 10-acre grid cells in urban areas and larger grid cells in rural areas. Finally, those data are aggregated to Traffic Analysis Zones (TAZ). The system provides considerable flexibility—should the TAZ system be updated, land use forecasts can simply be aggregated to a new zone system. The model for the TJCOG region takes about 3 days to run, once both the parcel-level data and the suitability factors (such as transit lines and roadways) have been identified. This runtime includes some plausibility checks of the results and the creation of summary tables, but excludes the run of the transport model. The most time-consuming step is the aggregation of parcel-level data to 100,000 grid cells, divided into 16 sub-regions. Seven different allocations (two residen- tial and five employment land uses) are made. The subsequent aggregation from grid cells to TAZs takes about an hour. TJCOG uses a standard workstation, where the GIS license is accessed in the cloud—this setup may increase the runtime. Figure 4-1 shows the land suitability analysis for new development in the TJCOG study area.

32 Integrated Transportation and Land Use Models Figure 4-1. Land suitability analysis with CommunityViz for TJCOG (Source: Triangle J Council of Governments).

Sketch Planning Land Use Models 33 For the 2045 Plan update, TJCOG and its MPO partners have started with two realistic scenarios: one is based on the current community plans, and one is termed “AIM High.” The latter is based on selected commitments to more compact, mixed-use development along transit corridors by both local communities and the region’s large anchor institutions, such as large research univer- sities and the Research Triangle Park. The final preferred scenario might combine elements of both or could depart from the initial scenarios based on public engagement. The existing scenarios developed by TJCOG are comparable in terms of distribution of popu- lation and employment, because changes will realistically remain small. Most of the urban areas in 2045 are already developed in 2013, which is why summaries at the regional level look rather similar. The creation of a new scenario is fairly simple in the CommunityViz environment. For exam- ple, for a new fixed guideway investment, a buffer analysis would be done in GIS for all new fixed guideway stations, which would affect the suitability weights calculated in CommunityViz. This rule-based approach is fairly quick to implement. It would also be possible to update every parcel near transit stations manually. That approach would be more precise and probably more realistic, but it would take much more staff time to develop. Nevertheless, since this manual process gives more confidence, TJCOG staff is reviewing manually to see if a different future place type would be more likely for a given parcel after a fixed guideway project was built. No consultant is needed for scenario analysis; everything is done in house by TJCOG staff members. In the future, support from consultants might be needed in order to add features, but minor adjustments of the model can be done in house. Based on the assessment of TJCOG staff members, planning decisions of the MPO policy board probably did not change significantly because CommunityViz was implemented. However, having CommunityViz model output adds confidence to the model results. CommunityViz can help to bolster the argument for or against projects, which can help defend a plan if it is challenged in court. Accordingly, upper management has been supportive of and interested in using CommunityViz for more applications. The use of the model is financially supported by the MPOs and the regional transit agency through the allocation of federal planning funds. TJCOG’s membership provides the required 20% match for federal funds. 4.7 Lessons Learned SACOG SACOG was satisfied with the power and performance of I-PLACE3S. Staff members could run many scenarios in a short time, and results helped to build public consensus on planning alternatives. Although PLACE3S was open source and small model revisions could be done in house, I-PLACE3S is proprietary and its implementation resulted in dependence on the model developer. Due to this dependence, maintenance costs turned out to be higher than anticipated. As a consequence, SACOG staff decided to select tools for which a competition among con- sultants was given. Open-source tools enable staff members to implement some revisions in house or at least review functionality of the model when it is ambiguous (or absent) from the documentation. SACOG recognized that most travel demand models need consulting support periodically to implement updates. The same is likely to be true for land use models. For most agencies, it would be inefficient to train staff members to implement major model updates only required every 10 years or so. But at least minor model updates and certainly model applications should be done in house by staff members. According to SACOG, outside help from consultants or universities should only be required for major model revisions.

34 Integrated Transportation and Land Use Models According to SACOG staff, it was a good choice to have the model development done by a consultant. Later it was important, however, to transfer the model to further in-house develop- ment, with limited support from the consultant for particular improvements. There is a con- tinued engagement with land use planners who provide parcel-based data. They meet three times per year, and SACOG staff members provide training sessions on using the model for local jurisdictions. TJCOG Similar to SACOG, TJCOG has been satisfied with the model performance of their sketch planning model. Agency staff members and management have appreciated the extendibility of the model to incorporate new features. For TJCOG, it was crucial to include stakeholders in model development early on. Focus groups were conducted, each consisting of a dozen developers and representatives of the Urban Land Institute Chapter. These focus groups spent half a day with TJCOG and local government staff members to develop weights that could be used in CommunityViz to calculate suitability indices for land use development. Focus groups developed 15 factors for CommunityViz, and they developed the suitability weights to weigh these factors. Focus group members were asked to do a pairwise comparison of two factors at a time and decide which factor was more relevant for development choices. These values then were aggregated to derive weights for the model. In addition, TJCOG staff members included local planners to decide where new development would likely be allowed to be developed.