Transportation Planning Capacity Building Program

- Peer Exchange Report -

Land Use Models in Transportation Planning

Table of Contents

I.    Summary

II.    Background

III.    Best Practices Shared by Peer Exchange Participants - MPO Perspectives

1. PAG’s Current Forecasting Process
2. The Use of Expert Panels in Land Use Forecasting
3. Three Long-Standing Modeling Programs
   1. A Four Model Integration System (SANDAG - San Diego, CA)
   2. Transition to PLACES & MEPLAN models (SACOG - Sacramento, CA)
   3. MetroScope (PMC - Portland, OR)
4. UrbanSim
   1. UrbanSim Testing (WFRC - Salt="" Lake City, UT)
   2. UrbanSim (PSRC - Seattle, WA)
   3. UrbanSim Implementation (DRCOG - Denver, CO)
5. SAM/LAM modeling practice.
   1. MAG - Phoenix, AZ
   2. MRCOG - Albuquerque, NM

IV.    Lessons Learned

V.     Recommendations

VI.    For More Information:

VII.    Attachments/Links

Appendix A: Peer Exchange Attendees
Appendix B: Peer Exchange Agenda
Appendix C: PAG Problem Statement (Pre-Meeting Handout)
Appendix D: Topics to Be Addressed in Peer Presentations (Pre-Meeting Handout)

I.   Summary

The following report summarizes the results of a Peer Exchange held through the Transportation Planning Capacity Building (TPCB) Program, which is jointly sponsored by the Federal Highway Administration (FHWA) and Federal Transit Administration (FTA).  The Pima Association of Governments (PAG) hosted a one and a half day workshop for eight western metropolitan planning organizations (MPO) to facilitate the exchange of best practices on integrating land use modeling with traditional transportation modeling.  The discussion outcomes are being incorporated into a set of guiding principles for PAG to consider as they select and implement a land use model that best suits the needs of the Tucson region.  

A primary objective of the Tucson Peer Exchange was to facilitate dialogue among technical experts from strong land use modeling programs.  Two critical issues in transportation planning involve creating reliable forecasting of future travel behavior, and understanding the relationships between multi-modal transportation systems and regional development patterns.  The Peer Exchange explored which models adequately address these and other critical transportation issues. 

Dr. Michael Meyer of the Georgia Institute of Technology facilitated the Tucson Peer Exchange.  Participants of the Peer Exchange included land use model experts from the Denver Regional Council of Governments (DRCOG), Maricopa Association of Governments (MAG), Middle Rio Grande Council of Governments (MRGCOG), Portland Metro Council (PMC), Puget Sound Regional Council (PSRC), Sacramento Area Council of Governments (SACOG), San Diego Association of Governments (SANDAG), the Wasatch Front Regional Council (WFRC), FHWA, FTA and the Volpe Center.  The Peer Exchange was held from March 24th to 25th, 2003 in Tucson, AZ.   

II.  Background

The Pima Association of Governments (PAG) serves as the official metropolitan planning organization (MPO) for the Tucson region.  Member agencies of PAG include Tucson, South Tucson, Marana, Oro Valley and Sahuarita, as well as the Tohono O’odham Indian Nation, the Pascua Yaqui Tribe and Pima County.  With a population over 900,000 in an area of 9,200 square miles, Pima County has low population densities that average approximately 2,500 people per square mile.

The local jurisdictions in the PAG region recognize the importance of integrating land use and transportation planning.  The City of Tucson General Plan, enacted by the voters in 2002, includes goals and policies designed to organize land usage in a manner to reduce trips lengths, and to improve the efficiency of the multi-modal transportation system.  Pima County’s Comprehensive Plan also includes goals and regional policies designed to create a more efficient transportation system.  However, in order to fully explore specific strategies designed to accomplish such goals, PAG must improve technical abilities to evaluate the interactions between development patterns and multi-modal transportation systems.

For several years, PAG has been using the Citilabs TP-Plus/Viper transportation model.  This is a fairly robust system capable of producing multiple model runs that assess the performance of various transportation network scenarios.   

However, in determining future travel demand, PAG generates population and employment forecasts using less sophisticated methods.  Today, even though all forecast data are GIS-based, the methods used for distributing future growth are primarily manual and time-consuming operations.  This seriously limits the ability to produce multiple growth forecasts and to assess the transportation impacts of alternative “what if” types of development scenarios.  Furthermore, most assumptions are based on adopted land use plans, with evaluation by local committees and expert opinion.  While these “plancasts” are useful in assessing the potential, cumulative impact of general plans (if implemented as adopted), they lack consideration of other key factors affecting location decisions made by households and businesses.  For instance, there is no consideration of the land market (in terms of price, supply and demand), rent levels, and occupancies.   

With increasing development pressures, a regional population approaching one million, and a continued strain on the Tucson region’s transportation system, PAG recognizes the need for more sophisticated tools to explore the interaction between land use and transportation.  Additionally, provisions within ISTEA and TEA-21 encourage regions to think about the reciprocal relationships between development patterns and multi-modal transportation systems. [1]   Other statutes challenge PAG to consider the associations between adopted land use plans and regional transportation plans. 

Land use model implementation is included in PAG’s current budget and overall work program.  PAG anticipates that developing and implementing a fully integrated and functioning land use and transportation model will be a multi-year effort.  In this first year, PAG is putting forth a concerted effort to evaluate land use models currently used by other MPOs.  The Peer Exchange was an important part of PAG’s initial step in choosing an appropriate land use model for the Tucson region.

[1]See 23 USC, 134(a), 134(f), and 204(e).

III.  Best Practices Shared by Peer Exchange Participants - MPO Perspectives

The initial step in PAG’s multi-year effort to develop, implement and integrate land use models with travel models began with a thorough evaluation of available land use models.  Towards this end, the Peer Exchange explored the following topics:

• Strengths and weaknesses of land use models used by MPOs and DOTs.
• Behavioral accuracy, calibration and credibility of such models.
• Sensitivity/ability to assess land use impacts of policy changes.
• Integration with transportation models.
• Integration with GIS.
• What modelers have learned, and what they would do differently if starting over.
• Data requirements.
• Maintenance costs and staffing levels required.

The Peer Exchange began with an overview of PAG’s current forecasting process. Peer Exchange participants then presented on their land use and travel modeling experiences.  First, a representative from Parsons Brinckerhoff discussed the use of Expert Panels in land use forecasting.  Second, MPOs from the San Diego, Sacramento and Portland regions described their long-standing modeling programs.  Third, MPOs from the Salt="" Lake City, Seattle and Denver regions discussed their experiences with the UrbanSim model.  Finally, MPOs from the Phoenix and Albuquerque regions described their use of the SAM/LAM modeling program.  

A. PAG’s Current Forecasting Process
Andy Gunning, PAG (Tucson, AZ)

PAG’s current forecasting process has enjoyed many successes thus far, yet is insufficient for the changing needs of the Tucson region.  On the one hand, PAG has access to a variety of data through partnerships with local and state agencies (e.g., population committee).  Additionally, PAG relies on data that is readily available for multiple levels of geography, such as census tract, city and county level data.  For anything below the county level, i.e., for traffic analysis zones (TAZ), PAG determines future spatial allocation using GIS data and general plan guidance.  On the other hand, however, piecing together the multiple sources of data is both labor intensive and time consuming.  What results is a single forecast scenario with no consideration of slower or faster growth, or of different spatial distributions of growth. 

A major strength of PAG’s current forecasting process is the cooperative effort to establish a comprehensive geo-spatial database between the eight local agencies.  The database successfully integrates parcel-level GIS layers with orthophotos and terrain data.  Over 400 GIS layers are included in this integrated database, including parcel boundaries and ownership information, street centerlines and networks, TAZ, census geography, and land use information.  The database is accessible through the internet, and has fundamentally changed the way PAG conducts planning, preliminary engineering, and infrastructure management tasks.

Despite this achievement, PAG identified the following characteristics of their current model that are areas for improvement:

• The forecasts based on one scenario.
• The forecasts are blind to socioeconomic factors.
• The “plan-cast” assumes that plans will be implemented the way they were adopted.
• The inconsistent land use data leads to inconsistent methodologies.
• Staff resources for the significant data collection and entry required for GIS are limited.

Gunning concluded the presentation by outlining the following features of a desired “ideal” land use and transportation model: 

• The model operates as a GIS-based information manager.
• The model can assist in small area allocation.
• The model can produce/test multiple scenarios.
• The land use and transportation forecasts have an interactive feedback ability.
• The model takes land markets into consideration.
• The model has applications for local agencies.

B. The Use of Expert Panels in Land Use Forecasting
Sam Seskin, Parsons Brinckerhoff

Expert Panels are one of many methodologies used  to forecast land use impacts of transportation projects.   An Expert Panel is a diverse group of experts who utilize some variation of the Delphi Method to develop land use forecasts.  The Delphi Method is a structured, iterative group process, whereby each member of the selected panelists conducts an independent analysis, then continuously reviews and revises results after multiple rounds of meetings in order to report a single set of panel findings. [2] Seskin advocated that Expert Panels not only provide rigorous analysis while maintaining flexibility and adaptability, but also incorporate the realities of urban development. 

[2] For more information, see The Use of Expert Panels in Analyzing Transportation and Land Use Alternatives, National Cooperative Highway Research Program (NCHRP) Project 8-36, Task 04.

Chart 1: Diagram of the Expert Panel Process
click here for text version

Expert Panels have been utilized in the past on projects such as the WS DOT I-5 and MD DOT I-270 corridors, as well as in Dane County, WI, and Longview, TX.  Seskin provided a step-by-step description of the Expert Panel process by explaining events of the NH DOT I-93 project.  A 14-member panel consisting of environmental, real estate, planning and development professionals, academics, a policy analyst, economist and planning board member, was convened over a nine month period to assess land use impacts of the I-93 project.  Briefing books introduced the study area, regional conditions and trends (i.e., travel time, transit employment, population change, home prices, basic zoning ordinances, land and resource constraints), and no-build/build scenario descriptions.  Each panelist received a briefing book, and conducted an individual analysis using an Excel spreadsheet. After all results were presented, the panel used a blended average allocation ((mean + median)/2) to determine a single report.  For example, because panelist results for population change ranged from 0 to 15%, the report used the blended average of 8-9% for population change.  Stakeholders and NH DOT supported the results and incorporated the Expert Panel analysis into the NH DOT I-93 environmental impact statement (EIS). 

Seskin identified six essential steps for the successful use of Expert Panels:

1.  Know the big picture

• Spell out specific objectives (i.e., is the goal to create a forecast, or to assess land use or economic impacts?).
• Define roles and responsibilities for client agency and stakeholder groups in order to manage the panel well.

2.  Design the Panel Process

• Identify analysis parameters (e.g., the study area, analysis zones).
• Describe the panel’s charge (i.e., the questions to be addressed, how they will be answered).
• Specify the format (i.e., importance of getting feedback, preserving anonymity, attending meetings).

3.  Create the Panel

• Identify experts (e.g., analysts, developers, planners, academics, long-time residents).
• Determine the size of the panel, which depends upon  the type of analysis needing to be done.
• Emphasize the necessity of impartiality and commitment.

4.  Finalize Preparations

5.  Manage the Process

• Respond to unforeseen events (e.g., provide more information, address panelist difficulties).
• End the process (i.e., preserve stability and consensus, emphasize consent versus consensus).

6.  Document the Results

The presentation concluded with an emphasis on three points.  First, a lack of consensus nevertheless provides useful information that might not otherwise have been generated.  Second, details are crucial for the process.  And finally, the client agency should avoid steering the results of the Expert Panel.

C. Three Long-Standing Modeling Programs

1. A Four Model Integration System
Jeff Tayman, SANDAG (San Diego, CA)

The San Diego Association of Governments (SANDAG) oversees an area of 2.7 million acres, half of which are mountains and desert, and a population of three million people.  San Diego experienced an approximate 3% growth rate in the 1980s, and just over 1% in the 1990s.  

SANDAG has been producing long-range forecasts of population, housing, employment, income, and land use for the region since 1971.  The forecasts simulate the potential development patterns in the region given certain public policy changes.  The forecasts have been used to assess growth impacts, project changes in levels of service for public facilities, and determine needs for new or expanded facilities.  Elected officials use the forecasts as a tool to make strategic policy decisions that affect the growth and sustainability of the region.  

SANDAG creates long-range forecasts in two general phases: phase one produces a forecast for the entire San Diego region (Region wide Forecast), and phase two allocates the Region wide Forecast to cities, the unincorporated area, and smaller geographic areas (Cities/County Forecast).  The forecasts are produced collaboratively with local jurisdictions, technical and policy committees, and through extensive public outreach and review. 

SANDAG’s forecasting modeling system integrates four interrelated and integrated models: the demographic and economic forecasting model (DEFM), interregional commuting model (IRCM), cities/county forecast (UDM), and transportation forecasting model (TRANPLAN).

Chart 2: The SANDAG Forecasting Modeling System
click here for text version

The regional forecast provides basic demographic information (i.e., births, deaths, age, sex, ethnicity, migration patterns, labor force) and economic information (employment, output, prices, wages, costs, housing supply/demand, public finance). The IRCM, UDM, and TRANPLAN models incorporate the following land use inputs:

• Planned land use;
• Existing land use;
• Constrained land;
• Redevelopment or infill land; and
• Current development.

The land use and transportation models are linked incrementally and sequentially instead of iterating around a given sequence.  Thus, the output of the allocation model and transportation model forecasts from 2000-2005 is then used to forecast the 2005-2010 output, and so on. 

Chart 3: SANDAG - Linking the Land Use and Transportation Models
click here for text version

Source: Jeff Tayman, SANDAG

The data management and statistical tools underpinning the models are extensive.  The tools include:

• An Enterprise relational database manager;
• Graphical user interfaces;
• Geographic Information System (GIS) for input preparation and output analysis;
• Linear and nonlinear calibration algorithms;
• Spatial interaction gravity models; and
• Probabilistic, n-dimensional controlling routines.

The resource requirements behind the model development and production of forecasts are estimated in the following table:

Table 1: SANDAG Resource Requirements

Source: Jeff Tayman, SANDAG

The one-time IT costs are the initial investment required to develop the relational database.  After the model is in place, the cost of producing a forecast was approximated at $500,000, including data acquisition and labor costs.  Consultants are used primarily for regional modeling; the sub-regional models tend to be customized and require in-house expertise. 

Tayman concluded his presentation by identifying keys to successful implementation and potential challenges to implementation of land use models.  The keys to successful implementation include:

• Provide a tool for effective decision-making.  The effectiveness of the tool is measured by its usefulness to the decision-makers using the forecasts, i.e., elected officials.
• Secure commitment and resources to update and maintain models and databases.  The quality of the models and databases is dependant upon the quality of data input (i.e., accuracy, frequency).
• Establish credibility of models and the process.  The best way to ensure that forecasts will be taken seriously is to ensure that the process behind it is understood and trusted.  Political support is an essential factor impacting the usefulness of forecasts.
• Take the models out of the “black box”.  Technical jargon and complicated terms must be translated into layperson’s language.
• Balance theoretical elegance with practical application. 

Recommendations to address potential challenges include:

• Dedicate sufficient resources (in-house staff, consultants, hardware, software, licensing, data acquisition costs).
• Balance production responsibilities with research and development(R+D).  With a great demand for production of forecasts, little time remains to dedicate to R+D related efforts.
• Improve model accuracy and reasonableness.
• Keep up with demand for increasing spatial and variable detail.
• Turn around model results in a timely fashion.

2. Transition to PLACES³ & MEPLAN models
Gordon Garry, SACOG (Sacramento, CA)

The Sacramento Area Council of Governments (SACOG) underwent a drastic shift in focus from virtually no attention paid to land use to emphasizing the land use-transportation relationship.  A growing quality of life debate prompted a focus on regionalism.   To meet this shift in emphasis, SACOG has implemented new planning and modeling processes with the addition of PLACES³ and MEPLAN.  PLACES³ is a GIS-based planning tool and MEPLAN is primarily an economic and land use model that includes some transportation elements. 

SACOG has attempted to integrate the land use models with the transportation models by connecting them for a base case future forecast.  MEPLAN allocates to the district level, PLACES³ allocates to the parcel level, and the travel information traditionally housed in the SACMET model was converted to MEPLAN.  The parcel level data is then aggregated back to the TAZ level. 

SACOG’s transportation - land use model is a mix of aggregate and simulation processes.  The design is flexible to accommodate changes.  Because of the design of the model, it can be used simultaneously to develop forecasts while the model continues to be improved.  SACOG continues to develop and adjust its current model, primarily after the Oregon DOT’s PECAS example.

Garry identified that the key point to consider when developing an integrated transportation-land use model is to remain flexible.  SACOG  undertook the task of model development in the midst of political change.  Due to the lengthy nature of model development, the political climate will inevitably shift, thereby affecting the level of support through a multi-year effort.

The largest implementation issue SACOG faced during its model development was limited finances.  Significant amounts of money went into a four year data collection effort, which left little to nothing available.  SACOG needed consultants to develop the model, but had limited finances from which to draw.  Another implementation issue was changing priorities, which resulted in both positive and negative consequences.

3. MetroScope
Sonny Conder, Portland Metro Council (PMC)

The Portland Metro Council (PMC) oversees a three county area covering approximately 300,000 acres, with a population of 1.2 million.  The most recent regional econometric model projects a 1.5% annual growth rate, and 1.85% total employment.

The PMC integrated land use and transportation model is called MetroScope.  MetroScope comprises four different models with a GIS-based visualization and accounting system.  The GIS module maintains consistency between the land accounting and model operation. [3]   The four models comprising MetroScope are the transportation model, residential real estate model, non-residential real-estate model and econometric model.

[3] For more information on the GIS module, see Hall, Carol, "Identifying Vacant and Buildable Land." In: Knaap, Gerrit J. (ed.): Land Market Monitoring for Smart Urban Growth. Cambridge, MA: Lincoln Institute of Land Policy, 2001.

Chart 4: What is MetroScope

Source: Sonny Conder, PMC

MetroScope has the ability to test multiple scenarios.  MetroScope can account for impacts of land use policy, such as changes in zoning capacity, changes in the urban growth boundary, implementation or removal of urban renewal areas, and subsidies/penalties for development in specific areas.  MetroScope can also account for impacts of transportation policy, such as transportation project investment by location and mode, transportation pricing options, and Traffic Demand Management options.  Finally, MetroScope can account for changes in the Regional Econometric Model, including changes in personal income, changes in regional growth rates of population and employment, and changes in demographic structure.

PMC coordinates with the local agency to determine the appropriate level of data requirements.  PMC emphasizes local agency scrutiny of model inputs such as acres of vacant land by zone type, acres of infill and redevelopment land and capacity estimates. Data are reviewed at both the model zone level and at the original input level (i.e., the tax lot). In addition, PMC also provides the review data at the TAZ level for jurisdictions accustomed to working with TAZ level data.  Finally, PMC reviews model outputs with local agency technical staff on a five year basis to check on allocation levels, prices, land consumption, densities, traffic levels, etc.

When comparing MetroScope forecasts to the actual outcomes for the period 1970 - 1995, PMC found 75% consistency between the two.  The 25 year ex ante evaluation emphasized MetroScope’s sensitivity to land supply assumptions.  The PMC model has difficulty determining whether vacant land is actually available land supply.  Predicting and modeling land manager behavior is one of the greatest difficulties PMC faces with the model.  

The estimated resources for the MetroScope model are listed in the following table:

Table 2: PMC Resource Requirements

Source: Sonny Conder, PMC

Currently, PMC employs four staff members devoted to updating land development.  Each of the parcels are either built or vacant lands.

Conder identified the following successes in the MetroScope effort:

• Producing model results that have been incorporated in regional policy decisions on UGB expansion and planning policy.
• Gaining regional acceptance to use MetroScope in determining regional allocations for the RTP.
• Actually using the model in an applied setting to do five policy-oriented case studies.
• Communicating to a wide audience that output from integrated models has the advantage of being consistent and complete across the region unlike previously used trend and expert review processes.
• Building the institutional resources necessary to use, interpret, maintain and further develop the integrated model.
• Learning how urban areas function.

Conder also identified the following unsuccessful elements in the MetroScope effort:

• Communicating the “modeling process” including inputs, outputs, assumptions and caveats so that decision makers are intuitively comfortable with the information.
• Packaging the awesome amount of very diverse information available from integrated models in an easily visualized and intuitively explicable manner.
• Being able to use the integrated model to test a wider range of policy options.
• Finding a common ground between the “win-win” political culture and the “cost-benefit tradeoff” culture of public welfare economics.
• Finding a way to make integrated models faster, cheaper and less complex.

Conder concluded his presentation by highlighting a variety of recommendations to consider when developing an integrated land use and transportation model:

• Choose modest, improvements that produce results soon. Fix one piece of the model at a time. Model specifications promising huge jumps in capability may never be implemented.  
• Never discard the current working system until the new working system is fully functioning.
• Avoid committing to a development process that can not be fully understood or controlled.
• Emphasize the inherent “fairness” of an integrated model. All areas are treated equally and consistently.
• Stay flexible and be prepared to improvise. Requests for information and analyses that the given model structure did not anticipate are inevitable.
• Remember that quantitatively insignificant results may be of paramount interest to the audience.  Political significance is qualitative.
• Retain an adequate level of MPO staffing (i.e., available personnel, level of training, continuity). Continuous attention from in-house staff is crucial I moving urban simulation models from specifications to practical use.  Able staff will get you through times of no development money but a lot of development money will not get you through times of no able staff.
• Secure support for a multi-year development process and commitment of in-house resources from the management team.

D. UrbanSim

1. UrbanSim Testing
John Britting, Wasatch Front Regional Council (Salt="" Lake City, UT)

UrbanSim is a simulation model that incorporates the interactions between land use, transportation and public policy in the planning and analysis of urban development.  The model integrates market behavior analysis and the analysis of governmental actions with land policy and infrastructure choices to maximize a realistic prediction of how the urban market will behave.  The UrbanSim model was developed by Paul Waddell, and is hosted by the University of Washington.[4]  

The Wasatch Front Regional Council (WFRC) began its UrbanSim effort in 1998 by initiating a substantial data collection effort.  The development of software and model estimation took the subsequent two years, followed by a period of testing and fixing the model. 

UrbanSim produces household marginals, employment, built environment (i.e., dwelling units, square feet) and land values.  The travel model produces logsums, travel times between the central business district and the airport, and proximity to highways and arterials.

Britting identified the following strengths of the WFRC UrbanSim experience:

• The model is free!
• The University of Washington provides strong support.
• The model is transparent.
• The model attempts to replicate the actions of relevant actors such as the impact of government intervention.
• Producing outputs in UrbanSim is easy with a logit model.
• The internal controls are rigorous.  The allocation procedures are rule-based, i.e., if a job is not listed in the grid of skills, there is no possibility of adding a new job in a land development.

Some of the challenges with UrbanSim include:

• Limitations in data. 
• Calibration/validation, in a traditional sense, is impossible.
• There is no real federal mandate for integrated transportation and land use modeling.
• Randomness exists.
• Support for the model is often superficial, suspicious or skeptical at best.

Peer participants wondered if the credibility of synthetic, “notional” data was questioned.  Britting maintained that the synthesized data is derived by allocating aggregate TAZ level data from travel surveys to households.

[4] For more information, see www.urbansim.org

2. UrbanSim
Larry Blain, Puget Sound Regional Council (Seattle, WA)

The Puget Sound Regional Council (PSRC) oversees four counties and 72 cities in an area of 6,287.8 square miles with a population of 3,275,847 people.  The growth rate is approximately 2% (2.3% high, 1.7% low) and the employment rate is slightly above 2%.  

PSRC has been relying on DRAM/EMPAL, a 20 year old model that has produced forecasts for each decade.  A RFQ for a new land use model was put out in June, 2002 with a memorandum of understanding and initial contract for UrbanSim designated in January, 2003.  Currently, the PSRC is in the process of collecting data to build the UrbanSim model.  The goal is to have operational testing begin in July, 2004.

Blain identified a series of recommendations to consider in developing a land use model:

• Provide a range of years within which the forecast results will reach the project levels.  The model’s numbers may not be accurate for a particular time period due to a boom and bust cycle, but should eventually reach the target. 
• Maintain “transparent” inputs and operation.
• Recognize that there is dynamism over time, not at equilibrium.
• Consider annual time periods.
• Develop parcel level resolution.
• Include a behavioral component.
• Make a discrete choice of locations.
• Use micro-simulation.
• Consider land markets.
• Integrate the land model with the travel model.
• Accommodate user-specified events.
• Incorporate existing performance measures.

Blain concluded his presentation by responding to the specific questions posed by PAG [See Appendix D of this report]:

• One success of PSRC’s modeling effort has been the extensive consultation involved.
• An unsuccessful element of PSRC’s experience has been the analysis of the land use impacts of transportation alternatives.
• One barrier that PSRC needs to overcome in order to achieve more effective implementation of the modeling program is the strength of local independence.
• The modeling capabilities and results have improved decision-making in the region by providing a consistent set of regional forecasts.
• A valuable lesson learned by PSRC was that members should be involved at all stages of model development and implementation.
• The staffing levels required to maintain data and produce model results is intensive.  UrbanSim requires four staff to work with data, four staff to work with GIS, and four staff to work with travel demand modeling.  
• PSRC coordinates with local agencies through a monthly regional technical forum.
• The models are produced at a spatial resolution of 150 meter grid cells.
• The land use and travel models interact through a feedback logsum at five year intervals.
• Economic factors and land markets are considered within the model.
• UrbanSim has the capability to test multiple scenarios.  Eventually, PSRC will integrate a regional econometric model.
• UrbanSim considers market-based redevelopment and infill potential within existing areas.
• UrbanSim can perform corridor level analyses.

3. UrbanSim Implementation
Christine Dumas, Denver Regional Council of Governments (Denver, CO)

The Denver Regional Council of Governments (DRCOG) includes nine counties and approximately 51 municipalities.  DRCOG has an urban growth boundary of 747 square miles, and is responsible for approximately 2.4 residents, 1.4 million employees, and 30,000 - 50,000 jobs/year.

DRCOG has entered a three to four year program to update the regional modeling plan and is moving away from allocation modeling to behavioral modeling using the UrbanSim platform.  Currently, however, DRCOG is refreshing the existing allocation model (traditional input-output model) and is providing forecasts of employment, households and other demographic data at the TAZ level.  DRCOG estimates that full development of the new model will take another couple of years.  DRCOG intends to use GIS as a platform for integration, and is adding GIS staff to help in the effort.

Dumas responded to some of the questions posed by PAG [See Appendix D of this report]:

• DRCOG’s strengths in the model development process thus far include full cooperation from local governments and reasonable results from a utility-based behavioral model.  Cooperation between local agencies and DRCOG has been outstanding, particularly in providing data. The lower end of the model is being built with welfare, medicare, lunch program, energy and low income housing data being shared by various agencies; the upper end of the model is being built with assessor data and an hedonic model.  Additionally, the economic behavior foundation of the model produces more reasonable results than what DRCOG’s current model has produced.
• Conversely, one of DRCOG’s main challenges is moderating the diverse desires of all the local governments.
• With regard to data issues, DRCOG currently has no platform to adequately deal with the extremely large SQL databases.  They need to build a hedonic model to integrate with an age-cohort model (which DRCOG currently possesses) in order to determine the aging of the Denver region’s population for local communities’ policy options model.

E. SAM/LAM modeling practice

1. Rita Walton, Maricopa Association of Governments (Phoenix, AZ)

The Maricopa Association of Governments (MAG) is the MPO for Maricopa County.  MAG utilizes a three-tier modeling process. 

Chart 5: MAG Three-Tier Modeling Process
click here for text version

Source: Rita Walton, MAG

At the top level, the Maricopa County control totals that are derived from the DES demographic model and REMI Economic model.  For the sub-regional model at the second tier of the process, MAG runs DRAM/EMPAL, a spatial distribution model.  And finally, for the small area model at the third tier of the process, MAG iterates with the Sub-area Allocation Model - Information Manager (SAM-IM) and their transportation model. 

The main features of SAM-IM are:

• The control totals consist of DRAM/EMPAL RAZ population and employment.
• GIS is used, with GIS coverage updated for each time period.
• The model is grid-based, currently on a one-acre basis.
• The model can account for each land use separately and in any order.
• Multiple land use categories exist (e.g., mixed use, business park). MAG can identify mixed use of land by percentages of individual categories for every polygon of multiple use land (e.g., 50% high density residential, 40% office, 10 % retail, etc.).
• The information manager can determine secondary allocations within a polygon (e.g., 40% office - 60% retail in a retail polygon).
• The model considers redevelopment.
• The model can develop velocity curves based on an analysis of the last 20 years of development.
• The model incorporates changing densities over time.
• The model acknowledges different factors that influence growth of each land use.
• The model incorporates a land use/transportation feedback loop.
• The model can sum to a TAZ or other level of geography.

2. Chris Blewett, Mid-Region Council of Governments (Albuquerque, NM)

The Mid-Region Council of Governments (MRCOG) covers four counties, and includes 600,000 people.

MRCOG also utilizes the SAM model and possesses strong GIS data. Matching capabilities are addressed at the parcel level.  MRCOG has not been required to get consensus on existing and future land use layers.  Because MRCOG serves two cities and just a few incorporated counties, accessing and updating data for the GIS layers is relatively simple and efficient.  MRCOG generates a new forecast series every three to five years.   

MRCOG receives many kinds of requests, including traffic and future build forecasts. Travel demand forecasts and land use forecasts have been turned around within one day.  MRCOG desires to maintain this level of service rather than engage in a lengthy process.  Blewett maintains that forecasts lose their utility if people stop asking for the information due to a long turn around time.

MRCOG addresses the issue of ethnicity and income by focusing on zones with disproportionate impacts or unfavorable investment impacts, marginal accessibility, etc.  Once the zones have been identified, MRCOG determines the likelihood of the distribution of those characteristics currently found within the zones.   

IV.    Lessons Learned

The concluding portion of the Peer Exchange focused on a discussion of topics identified by PAG as important and relevant to the process of determining an appropriate land use - transportation model.   The following lists the experiences shared by Peer Exchange participants:

Small Area Forecast (at TAZ level)

• In Portland, PMC is currently working on version 8.0 of their forecast due to constant updates.  The further in time they get from the last comprehensive study, the more general and hazy become the rules of moving households and jobs.  Currently, PMC works with approximately 50,000 households and 50,000 jobs.    
• In Seattle, the PSRC does not have detailed data assembled in a way that is useable.  Thus, PSRC disaggregates existing data to the TAZ level.  Well-organized data is crucial.
• In San Diego, SANDAG relates each block to an activity within GIS. 
• In Albuquerque, MRCOG has found that data collection is the most difficult part of the modeling endeavor.  The small area model took 8 months to develop.
• In Phoenix, MAG keeps member agencies involved in reviewing data so that projections and data are considered credible. 

Infill Redevelopment

• In San Diego, SANDAG has found that the local planners know more about residential development and less about employment.  SANDAG has also found that regional growth is due less to domestic migration and more to foreign migration.  Furthermore, many areas experiencing a large population growth do not have much housing.  This indicates that household size is growing differently than regional trends. 
  
  SANDAG has begun to develop an objective set of rules to guide their thinking on where redevelopment might occur.  In order to do so, SANDAG first estimates how much additional capacity existing parcels could hold.  Then, they develop an algorithm that can account for additional activity beyond an additional building. 
• In Portland, PMC incorporates infill and redevelopment into their modeling process with GIS.  Variations exist across the nation on what is considered to be “infill”.  PMC defines the activity as “refill” because of its precise definition of every parcel (i.e., land is either vacant or developed).  Positive or negative refill is possible.  Each tax lot is filtered to determine refill activity.  
  
  PMC makes a concerted effort to gather data from a variety of sources: assessor files, land use map, air photos, building permit and visual audits.  Through sampling, PMC determines what percentage of total built units occur on “developed” land. 
  
  The unit of analysis is based on real estate instead of the number of people in the real estate model.  For the nonresidential model, PMC attempts to count real estate but more often counts employment.  Portland has experienced changes in employment density without any change in real estate development. 
  
  The issue of redevelopment (i.e., losing or gaining capacity) poses definitional challenges to PMC when the land is mixed use.  PMC addresses the challenge by counting capacity twice: once for mixed use and once for residential.
• Denver recently underwent an infill plan.  The models had very little to do with the process, which was very community and policy driven.  A series of neighborhood workshops revealed that neighborhoods desired to increase housing density more than what the city had anticipated.  The MPO supported the infill plan by providing scenario forecasts.
• In Sacramento, SACOG engaged in a “community design program” of infill redevelopment and smart growth development.  GIS has some return on investment calculations.  This allowed SACOG to determine development cost and rent values to apply to infill redevelopment sites. 

Land Market and Interaction with Pricing (State Trust Lands)

• Arizona and New Mexico are in a unique situation regarding the state trust land issue.  When Arizona and New Mexico became states, Congress set aside four sections of each township for state lands.   The Land Department has held onto the land instead of selling them.  The State considers itself as sovereign and not subject to local zoning regulations.
• In Albuquerque, the state of New Mexico owns five large areas.  The state operates like other member agencies in the desire to promote build-out, although the state does not have many resources to invest.  MRCOG treats the state trust lands similarly to any other property within the land use model.  Development gets allocated to the state lands if well located.
• In Phoenix, one third of the land is owned by the Native American community.  The tribal governments tend to exercise strict regulations on who can live on the reservation, but do not have the same restrictions on commercial activity. 
• In San Diego, SANDAG regards casinos as special developments. 

Transportation Accessibility, Implications of Transportation on Urban Form, and Scenario Testing

• Some participants commented that producing multiple alternatives for a particular development impact is often politically difficult to implement. 
• In San Diego, SANDAG has had difficulty getting different stakeholders to agree on varying the model from what is formally “on the books”.  The models often lack credibility and political buy-in because the results of multiple scenarios do not align with various interests.            
• In Portland, PMC can not run robust scenario tests because the politicians do not want to be identified with certain scenarios.
• In Seattle, PSRC runs five or six extreme and disparate scenarios along with each major long range plan update (every six years).  For example, one test will make ferry service extinct.  Another test will have enough roads to have a “c” level of service at all times. And yet another test will build intensive transit systems.  By creating such disparate scenario tests, PSRC established the extreme parameters within which more moderate preferred alternatives can be assessed.
• In San Diego, SANDAG conducts four to five land use scenarios.  The emphasis on smart growth has left other potential urban forms to the wayside.  
• In Sacramento, SACOG runs scenario tests by making a change within a particular neighborhood and assessing the impact on that neighborhood alone.  The rest of the universe is ignored.  Now that a regional model exists, SACOG can reallocate with constrained numbers, or grow as the region grows.
• In Denver, DRCOG runs two extreme cases.  One is an endless sprawl scenario, and the other is a case of extremely high density.

GIS Data Server

• In Seattle, PSRC hired a consultant to relate all the databases so that they refer to the same items. 
• In Portland, PMC tries to maximize the GIS information that the model does not use.   TRANSIM has this capability, and can sample along a block face to generate travel demand.  

Inconsistent Methodologies

• In San Diego, SANDAG compared their GIS data to Census 2000 data and found the Census data to be 90% correct.  With its modeling capability, SANDAG no longer considers the Census data to be sacred and moves factors to the appropriate block when necessary. 
• In Phoenix, MAG moves Census data to the appropriate block within their base model, but never increases or decreases Census totals.

Special build out scenarios and model runs

• In Phoenix, MAG provides a transportation model run for virtually all requestors.  This includes consultants’ requests, as long as they provide the input data, network and other data requirements.  MAG employs one person who is dedicated to responding to all requests for model runs.  MAG provides socioeconomic support to non-member agencies for a fee.
• In San Diego, SANDAG makes the official model runs separately from ad hoc requests. 

Environmental Justice

• One participant noted that the purpose of modeling efforts in the wider issue of environmental justice is simply to provide information to identify where problems currently exist.  The primary concern of MPOs should be to take an inventory of current conditions and potentially disproportionate benefits or negative impacts due to location of transportation projects, rather than attempting to incorporate demographic factors into a long range forecast.
• In Albuquerque, MRCOG deals with environmental justice by making two assumptions.  First, MRCOG assumes that the population will remain where it currently is located, and runs the scenario test of impacts.  Second, MRCOG assesses the 2025 forecast and ascertains where disproportionate impacts result.
• In Portland, the land use model allows for an infinite number of household income, age, class, mode of travel, place of work, etc., and measures the impact on housing prices and transportation costs.  The model allows some assessment of affordable housing and subsidy programs.   
• In Sacramento and Salt="" Lake, SACOG and WFRC do not currently incorporate environmental justice into their land use models.  Instead, SACOG and WFRC assess the income and ethnic characteristics of locations where transportation projects are located informally, but not through a modeling scenario. 
• In Denver, DRCOG assesses accessibility to multi-modal transportation. 
• In Seattle, PSRC does not incorporate environmental justice into their models.  Currently, they do not even compare accessibility with location of pending and current transportation projects.
• In San Diego, demographic factors are built into SANDAG’s model, such as minority TAZ, differential travel times, age/income/ethnic groups, accessibility (distance to bus stop), drive times, etc. 
• In Phoenix, MAG projects income by income category and age of housing.  Although some models can project race, ethnicity, age, etc., these factors are not part of the official regional forecast.

What future activities would you invest in if $100,000 were given to you for your model?

The Peer Exchange Facilitator, Mike Meyer, posed the question of how each MPO would invest money in their current land use and transportation models as a way to stimulate each MPO to identify current priorities.  The MPO responses follow:

• Approximately 10 more runs to test relationships to understand the Portland region even better (PMC).
• Half invested into enhancing geo-database and GIS development, and the other half into land use modeling development (SACOG).
• Speeding up the development process, including more people to conduct database management, data clean-up, and data acquisition at the parcel level (DRCOG).
• Some money invested into tracking employment better, and the rest into programming analysis tools to assess the impacts of MAG’s modeling work for MAG and for MAG member agencies (MAG).
• Half invested into building a strong geospatial database and developing a maintenance protocol, and the rest invested into a method to check if forecasts are reasonable (WFRC).
• Data cleaning and data acquisition to fill in blank fields (PSRC).
• Begin by emphasizing the base capabilities needed, and tweak the model later.  Local agencies in the region appreciate what can be done now with current capabilities (MRCOG).
• Some invested into better linking of inter-regional and sub-regional models to better understand the development process, some into allocation routine improvement, and some into procedures that can streamline simulations (SANDAG).

Peer Exchange participants stated that only a limited number of model-building experts are available today.  The experts identified, include Paul Waddell, Alex Anas, John Abrams, Doug Hunt, Marcel Echenique (MEPLAN), Thomasz de la Barra, Francisco Martinez, David Simmons, and Steve Putnam (DRAM/EMPAL). 

V.    Recommendations

Peer Exchange participants articulated a series of summary observations and recommendations based on the land use-transportation models presented by the representative MPOs.  The following observations were listed as key principles to consider for the successful implementation of a land use- transportation model:

1. Determine the purpose of the model.  Careful consideration of the purpose should drive the particular type of application/model chosen to develop and implement.
2. GIS has provided a platform that allows better data management than before.  The question that remains is how to get to a finer level of detail.
3. Recognize the multiple ways that data can be used.  
4. Successful modeling requires a sufficient level of technical expertise.  Securing sufficient human factors is as important as determining the “right” modeling system. 
5. Flexibility is crucial.  The model will most likely need to be improved.
6. The model should be policy-oriented with the ability to produce various policy scenarios. 
7. Credibility of the model and resultant forecasts is dependent upon transparency and openness throughout the development and implementation process.  Whether qualitative or quantitative in nature, public participation beyond the planner’s and politician’s perspective is the most effective way to gain support early.  Participants noted that whereas public participation would be difficult to incorporate for the official models, scenario modeling would be a more appropriate area for the wider public’s feedback.
8. Remember that the land use-transportation model will usually be incorporated into a travel demand model.
9. The forecasts help provide important justification for various policy choices.  Elected officials and MPO staff need to be continually educated when confronted with various planning choices, and the forecasts provide evidence for a particular action.  The forecasts can help provide credibility, then, for recommended policy actions.

The concluding portion of the Peer Exchange focused on the identification of important recommendations for PAG to consider in the process of determining the appropriate land use - transportation model:

Small Area Forecast (at TAZ level)

• Begin with the GIS system and continue to update GIS to account for the data. 
• When building the model, it is better to attain the smallest level of detail and then aggregate up rather than vice versa.  Although parcel and grid-level detail may not be what is demanded, it is preferable to have that information underlying an aggregated form of data.  Towards that end, tracking the contents of each parcel along the way is much easier than attempting to gather all the information after the fact.
• Develop a model first, no matter how imperfect it is, then refine it later.
• Build strong partnerships with other local agencies because much of the parcel level data belongs to them.  For instance, the county assessor needs to be a strong ally and supporter to the work.

Infill Redevelopment

• Utilize empirical observations in addition to models and simulations to understand the trend of development in the region.   
• Define the terms of the model.  If the item of observation is people and employees, massive changes could occur without any change in the real estate make up of a parcel.  Such changes would not be captured if focusing only on real estate development. 
• Many questions are beyond a model.  The most valuable contribution of a model may be its ability to generate multiple “what-if” scenarios.  Towards this end, build a model that can roughly remain constant with the assumptions and parameters open to change.
• Before diving into a major investment to develop a calibration and market-based land use model, ask, what is PAG’s role?  In the case of Portland, PMC plays a major role in the land development market with urban growth boundaries, etc. However, they are fairly unique relative to other MPOs.  Most MPOs have relatively little control, though they may advise. If what PAG needs is alternatives scenarios, perhaps reconsider the type of investment needed to develop a model.
• Models need to support, not drive, the infill study. Beware of emphasizing a model over community participation; allow the city to be at the forefront.  Infill development will always occur in the context of strong city intervention, such as changing zone ordinances and infrastructure development.    
• Once data is in place, provide quick turn-around results of scenario testing results.  Places GIS Software, which is supported by Parsons Brinckerhoff, is highly recommended. 

GIS data server

• When managing heavy amounts of data, use a relational database.  The current trend is towards ArcGIS.
• Consult with state DOTs to glean from their experience with GIS.  Some state DOTs are probably three to five years ahead of MPOs in their experience with a GIS data server.  In particular, Florida and North Carolina are strong examples.
• Consider how to develop technologies to extract relevant data efficiently, how to best convert data into usable information, and how to visualize the data for easy understanding by the user.

Special build out scenarios and model runs

• Build a model that is not constrained to a market-based model to account for requests which may not be market-based.
• Be careful to special runs from official  or planned runs.  Create a structured  schedule for planned runs.

VI.   For More Information:

VII.   Attachments/Links

Appendix A: Peer Exchange Attendees

Jim Altenstadter, Pima Association of Governments

Natalie Barnes, Pima Association of Governments

Larry Blain, Puget Sound (Seattle) Regional Council

Chris Blewett, Middle Rio Grande (Albuquerque) Council of Governments

J. Bramly, High Indian Standing Council

John Britting, Wasatch Front (Salt="" Lake City) Regional Council

Cherie Campbell, Pima Association of Governments

Sonny Conder, Portland Metro Council

Christine Dumas, Denver Regional Council of Governments

Gordon Garry, Sacramento Area Council of Governments

Andy Gunning, Pima Association of Governments

Bryant Nodine, Town of Oro Valley

Esther Lee, US DOT Volpe NTSC

Curtis Lueck, CLA

Michael Meyer (Facilitator), Georgia Tech University

Gary Oaks, City of Tucson Planning Task Force

Terry Rosapep, FTA

Manny Rosas, Pima Association of Governments

Shalini Sen, Tucson Unified School District

Sam Seskin, Parsons-Brinckerhoff

Bruce Spears, FHWA  

Fred Stevens, Tohono O’odham Indian Nation

Ed Stillings, FHWA  

Kevin Sweeney, Town of Marana

Jeff Tayman, San Diego Association of Governments

Chaz Tompkins, Pima Association of Governments

Jim Veomett, Pima County Planning

Rita Walton, Maricopa Association of Governments

Sandy White, Pima Association of Governments

Appendix B: Peer Exchange Agenda
Land Use Model Peer To Peer Exchange

Appendix C: PAG Problem Statement (Pre-Meeting Handout)

The Pima Association of Governments (PAG) serves as the official metropolitan planning organization for the Tucson region.  The region is about 60 miles north of Mexico, and 100 miles south of Phoenix.  PAG’s local agencies include the incorporated jurisdictions of Tucson, South Tucson, Marana, Oro Valley and Sahuarita, the Tohono O’odham Indian Nation, the Pascua Yaqui Tribe, and Pima County.  The total county population is over 900,000, with over 500,000      seven square miles of vacant land are developed each year.

Pima County is nearly 9,200 square miles, but the modeling area is limited to 4,000 square miles in eastern Pima County and portions of southern Pinal County.  Land ownership in eastern Pima County is mostly non-private.  The State of Arizona owns 33% of the land in eastern portion of the county, followed by private ownership (31%), federal ownership (28%), and tribal ownership (9%).  Much of the State-owned land is held in trust and may be auctioned for sale for future development, generating revenue for state-funded education. 

There have been recent major efforts to improve land use and conservation planning in the area.  Pima County is nearing completion of an ambitious multi-species Sonoran Desert Conservation Plan to ensure compliance with federal Endangered Species Act requirements, and along with its recently adopted Comprehensive Plan, to guide future growth.  The City of Tucson General Plan was ratified by the voters in November 2001, providing an updated vision of future growth in the city.  The other cities and towns have also updated and enacted revised general plans in compliance with the state’s recent Growing Smarter Act requirements. Land use planning and implementation is conducted at the local and county levels by the individual jurisdictions.   

Current forecasting process

PAG coordinates with the local agencies to prepare and spatially distribute population and employment forecasts at different levels of geography.  Population control totals at the county level are generated by the State of Arizona through its Population Technical Advisory Committee.   However, PAG has recently used locally generated, unofficial forecasts as an interim control, absent updated county forecasts produced by the State.  Employment forecast control totals are produced by the University of Arizona’s Economic and Business Research Center. 

Base data includes census 2000 data for population, housing units, household size and occupancy rates, and a current employment database acquired from a national vendor.  Building permit data is collected from the jurisdictions, but only by jurisdiction (not by address or subdivision).  PAG recently subscribed to a local consortium that provides permits data for new housing starts by census tract and subdivision. 

Using the county control totals as a constraint, and GIS and orthophoto imagery for guidance, PAG staff disaggregates population and employment forecasts at the traffic analysis zone level (859 total).  GIS data typically used includes parcel ownership, existing land use (from Assessor records, including vacant land status), general plan coverages for future land use intensity, approved master plans, and conservation land coverages defining environmentally-constrained lands.   PAG works with the local agencies to determine certain assumptions concerning land use intensities and the residential density and mix of development expected within each category.  Assumptions are made and documented regarding the net amount of land available for development once infrastructure and other uses are considered.  Finally, assumptions are made regarding the amount and location of unregulated “wildcat” subdividing occurring in the future, as well as the amount and location of redevelopment activity. 

The result is a single set of forecasts by TAZ (most recently for the year 2030) for the following: total population, occupied housing units, employment by type (6 categories), special traffic generators, school enrollment and income.  Draft data sets are reviewed by the local agencies through a population planning committee until the forecasts are agreed upon by the committee and in conformance with the 2030 county control totals.  

Our most recent forecasting effort may be described as a “bottom-up” allocation based on land suitability at the parcel level, with guidance from adopted land use plans.  The forecasts incorporate the most recent planning assumptions about future growth, and are being used in PAG’s 2030 Regional Transportation Plan currently underway.  This latest forecast process involved a fairly extensive and time consuming data collection effort.  However, it resulted in only one forecast scenario for the transportation plan.  There was no consideration of slower or faster growth rates, or of different development densities or spatial distributions of future growth.  This one set of forecasts provides the basis for determining future travel demand and air quality conformity in the 20 year transportation plan.

Strengths

We feel we have several strengths that enhance our current forecasting process, as limited as it is, and these will assist as we improve our forecasting tools.  They include:

• Comprehensive geo-spatial database.  The region has made a significant effort to collect and maintain a comprehensive parcel-level GIS database, with over 400 distinct data layers, which are now complemented by accurate digital orthophotos.  Pima County has been in the forefront of GIS development since the mid-1980’s with the creation of their countywide electronic parcel base.  All the local agencies have coordinated to create and maintain their GIS data on the same base.  PAG has contributed by coordinating a regional orthophoto program where over 1,500 square miles of high resolution (1/2 foot), accurate imagery and terrain have been collected.  An effort is currently underway to merge the parcel-level GIS data with the orthophoto data to create a comprehensive, geo-spatial database.  This collection of GIS and orthophoto data is readily available to the local agencies, the private sector and the general public via the web through Pima County’s GIS MapGuide site and PAG’s Regional Data Center.  This new enabling tool has fundamentally changed the way local and private sector agencies conduct planning and infrastructure management tasks. 

• Coordination of and support for improved technologies.  The region has made a commitment to improving our technical abilities, and to maintain and build on the geo-spatial data where we have already made a major investment.  There exists significant trust in the local technical community, and effective working relationships and interagency coordination among the regions’ technical leaders. 

• We have only eight agencies to coordinate with in our region (five cities and towns, one county, and two tribal entities), a relatively manageable number.

• Commitment by the local agencies to enhance planning efforts.  This is evident with the county’s draft Sonoran Desert Conservation Plan and the general plans recently adopted by the local agencies under the state’s Growing Smarter Act.  Also, PAG’s current Regional Transportation Plan update involves a major public participation component.  Many participants have expressed the need to better coordinate land use decisions and transportation planning. 

Weaknesses

We also have many weaknesses in our current forecasting process that will need to be addressed as we look for ways to improve.

• Data consistency.  In several key areas, we lack consistent data from one jurisdiction to the next.  This includes land use approvals and committed development, such as approved rezonings and development plans.  Also, building permit data reporting is inconsistent and not at the same resolution for all agencies.  Finally, an existing land use coverage is created primarily from the county tax assessor’s database.  However, converting land use codes from a database for tax assessment purposes, into land use categories for planning purposes, results in significant errors with possibly as high as 20% of all parcels miscoded.

• Data collection and frequency of updates.  Data maintenance, especially for the land use datasets used in forecasting, has been a problem.  Updates requested by PAG have occurred only when needed for the next forecast series, and often two or more years may go by before data sets and forecasts are updated.

• Inconsistent methodology.  Since each jurisdiction maintains different forms of land use data, and at different resolutions, applying a uniform methodology has been a problem.  In our most recent effort, determining future housing unit growth for smaller jurisdictions was easier given their recent completion of general plan updates and with identifiable and active master plans.  For others, especially in the city and the county, there was much less certainty, especially when assuming future residential densities in undeveloped areas and exploring redevelopment potential within the existing urban area.

• Limitations of a single growth scenario/forecast.  The level of effort required to manually produce a set of TAZ forecasts makes it difficult to generate more than one scenario.  The impact of faster or slower growth rates, or alternate distributions of future growth, is not assessed.  Using a manual process, it is difficult to perform even simple adjustments to basic assumptions, such as changes to future household size and occupancy rates.  

• “Plancasts.”  Our forecasts tend to rely heavily on adopted general plans, so they represent a “plancast” which assumes our plans correctly forecast and guide future development (assumes that plans are implemented as adopted).  There is also no true consideration of how the land market reacts to transportation investment and other development trends. There is no way to interrelate land use/population forecasts with travel model forecasts. 

• Inability to consider a full range socioeconomic factors and environmental justice issues.  Our manual process does not lend itself to forecasting population by age, income, transit-dependency, or considering the future distribution of other population groups protected under Title VI.  In our current forecasting process, we assume (probably incorrectly in many areas) the spatial distribution of households by income will remain constant in the future.   

• Simple re-aggregation of the forecasts is difficult.  Re-aggregating the forecast numbers from TAZs to other units of geography, such as census tracts, tends to be cumbersome and time-consuming. 

Future Needs:  Why a model?

PAG’s decision to explore a land use model is based on several issues.  The region continues to face rapid growth in population and travel demand.  We feel better tools are needed to allow planners and elected officials the opportunity to evaluate differences in land use and transportation arising from different policy choices as we attempt to manage this growth and prepare meaningful plans.  Also, provisions within ISTEA and TEA-21 require MPOs to evaluate the relationship between adopted land use plans and the long-range transportation plan.  Ideally this review should include not just the impact of implementing the regional transportation plan, but also the impact of not implementing the plan.  To consider these issues, PAG needs better tools to test different scenarios.

Basic Needs and Desired Functions

Below is a list of features PAG would like to incorporate into a land use modeling structure to meet our needs.  

• It should be GIS-based with the ability to provide output data in a format compatible with PAG’s transportation modeling needs. 



• The model and database structure should serve as an information manager for data collection and coordination, facilitating frequent data updates reflecting new development approvals.  Ideally this would be an accessible database structure where data sets are available to the local agencies for their use.



• Forecast data sets should include, at a minimum, land use by category, housing units, population, employment by type, school enrollment, and income.



• The structure should be capable of preparing forecasts (or aggregating forecasts) for multiple units of geography  (by traffic analysis zone, census tract, municipal boundaries, school districts, water districts, sewer service zones, planning areas, etc.). 



• The model must be relevant to current and future land use and transportation policy issues, and allow for analysis and evaluation of key policy issues such as growth management strategies, conservation planning, infill and redevelopment potential, and environmental justice assessment.  It should enable an evaluation of the consistency between adopted land use plans and the long-range RTP, and must be designed to allow the assessment of multiple land use scenarios and growth assumptions.



• It must be capable of modeling the effects of transportation system investment on development patterns; to evaluate the interactivity between our land use/demographic forecasts and transportation forecasts.  



• It should enable consideration of the land market in determining and allocating future growth.



• Ideally the model would be accessible to PAG’s member agencies, either through a shared license for use by the agencies, or through some other coordinated modeling effort, to evaluate potential impacts of developments of regional significance (of major rezonings, specific plans, plan amendments, and other policy changes).



• The modeling inputs should be transparent and key input assumptions should be clear and subject to review.  Credibility of results should be emphasized, enabled by reasonableness checks, sensitivity testing and validation of the forecasting methodology to maximize understanding and confidence in the model.

Appendix D: Topics To Be Addressed In Peer Presentations

(Pre-Meeting Handout)

1. What has been successful in your modeling efforts, and why?
2. What was not successful?
3. What barriers did (or does) your agency have to overcome to achieve implementation of your modeling program?
4. How have your modeling capabilities and results been used to improve decision-making in your region?
5. What valuable lessons have you learned in your efforts?
6. What data development, maintenance and update requirements does your modeling program require?  And at what resolution are data required?  Is your model GIS-based?
7. What staffing levels and resources are required to maintain data and produce model results?
8. How do you coordinate with local agencies (in terms of data collection, model runs, or other related technical assistance)?
9. At what spatial resolution are model results produced?
10. How do your land use and travel models interact?  Are there multi-modal considerations?
11. How are economic factors and the land market considered in your models?
12. Are you able to test multiple scenarios (test policy changes, impact of proposed land use changes, and proposed transportation investment, for example)?
13. Does your model consider redevelopment and infill potential within existing urban areas?
14. Are you able to perform corridor level analyses?