2005 Progress Report: Predicting the Relative Impacts of Urban Development Policies and On-Road Vehicle Technologies on Air Quality in the United States: Modeling and Analysis of a Case Study in Austin, TexasEPA Grant Number: R831839
Title: Predicting the Relative Impacts of Urban Development Policies and On-Road Vehicle Technologies on Air Quality in the United States: Modeling and Analysis of a Case Study in Austin, Texas
Investigators: McDonald-Buller, Elena , Allen, David T. , Kockelman, Kara , Parmenter, Barbara
Institution: The University of Texas at Austin
EPA Project Officer: Chung, Serena
Project Period: December 20, 2004 through December 19, 2007 (Extended to December 19, 2008)
Project Period Covered by this Report: December 20, 2004 through December 19, 2005
Project Amount: $650,000
RFA: Regional Development, Population Trend, and Technology Change Impacts on Future Air Pollution Emissions (2004) RFA Text | Recipients Lists
Research Category: Global Climate Change , Climate Change , Air
Robust forecasts of land use, emissions, and activity patterns are essential for air quality model predictions.The objectives of this research project are to: (1) apply an integrated transportation-land use model (ITLUM) to investigate the impacts of regional development scenarios and trade policies on the magnitude and spatial distribution of emissions of ozone precursors; (2) compare the air quality impacts of regional development scenarios on predicted ozone concentrations and human exposure patterns using a photochemical grid model; (3) test the hypothesis that predicted human exposure patterns based on ITLUM emission forecasts will differ from those based on the U.S. Environmental Protection Agency (EPA) post-Clean Air Act Amendment emission scenario projections; and (4) test the hypothesis that changes in land use and dry deposition patterns have at least as significant an impact on future air quality as changes in on-road vehicle emission control technologies. ITLUM-based forecasts will be compared with four predetermined metropolitan development scenarios: (1) low-density, segregated-use development based on extensive highway provision; (2) concentrated, contiguous regional growth within 1 mile of transportation corridors; (3) concentrated growth in existing and new communities with distinct boundaries; and (4) high-density development and balanced-use zoning.
The hypotheses described above are being investigated using a case study in the five-county Austin, Texas, Metropolitan Statistical Area (MSA). The Austin MSA, which includes Travis, Williamson, Hays, Bastrop, and Caldwell Counties, is located in Central Texas with a population of 1.25 million people and an economic sector based on computer hardware and semiconductor manufacturing, software development, education, and state government. The Austin area, similar to approximately 30 other urban areas in the United States, prepared an Early Action Compact or voluntary State Implementation Plan (SIP) under the 8 hour National Ambient Air Quality Standards (NAAQS) for ozone. Although the case study focuses on the Austin area, Austin is typical of many urban areas that are or could be facing designation as nonattainment under the 8 hour NAAQS for ozone, and the modeling framework is applicable to other urban areas.
The interdisciplinary research team has focused its efforts to date on three aspects of the project, which are summarized below.
1. Evaluation of the Impacts of Regional Development on Air Quality Because of Changes in Land Covers
This task was a collaborative effort with EPA Project RD831452, Impacts of Climate Change and Land Cover Change on Biogenic Volatile Organic Compounds (BVOCs) Emissions in Texas, also being conducted by the University of Texas at Austin. For Austin, a community-based planning team, Envision Central Texas (ECT), developed four possible land use scenarios (Scenarios A-D) that could result from a doubling of population in Central Texas. These scenarios were combined with Texas vegetation maps to arrive at projected changes in land cover, biogenic emissions, and air pollutant deposition velocities, at a spatial scale of 4 kms. A gridded photochemical model (Comprehensive Air Quality Model with extensions, CAMx) then was used to predict the spatial and temporal patterns of ozone concentrations, based on the revised emissions and deposition velocities. The differences in land cover led to 1 to 5 percent reductions in daily biogenic emissions in the five-county area that includes Austin; these reductions in biogenic emissions, and the corresponding differences in deposition velocities, led to reductions in maximum ozone concentrations. Reductions in daily maximum ozone concentrations, because of increased urbanization, ranged from 0.05 to 1.4 ppb, with values of 0.1 ppb typical for the Austin area. These results are comparable to many commonly employed emission control strategies.
2. Evaluation of the Impacts of Regional Development on Air Quality Because of Changes in Emissions from Anthropogenic Sources, Specifically From On-Road Mobile Sources
On-road mobile source emission inventories were developed for each of the four ECT development scenarios by combining travel demand model (TDM) output for the five-county Austin area link network with emission factors from EPA’s MOBILE6 model. The Envision Central Texas Travel Model (ECTTM) follows the general four-step modeling framework used by most Metropolitan Planning Organizations in the United States, consisting of trip generation, trip distribution, mode choice, and trip assignment, but includes a number of enhancements that make it more sensitive to transportation infrastructure and land use, including a mode choice model that is sensitive to land use along with an auto availability model that is sensitive to residential density and transit service. For each ECT scenario, vehicle miles traveled (VMT) values obtained from the ECTTM were disaggregated by hour of the day and vehicle type. The disaggregated link VMT was matched with the corresponding pollutant-specific MOBILE6 emission type factors tabulated by speed, hour, roadway type, and vehicle type to obtain link-level emissions estimates in grams. The MOBILE6 emission factors were based on a calendar year of 2030 to be consistent with the timeframe for doubling of the region’s population. Although VMT is predicted to continue increasing, emissions of NOx and volatile organic compounds (VOCs) are predicted to decrease through approximately 2025 because of the phase-in of new emission standards with subsequent increases in future years as a result of the effects of continued increases in VMT.
A gridded photochemical model (CAMx) was used to evaluate the changes in predicted 8 hour ozone concentrations because of differences in the on-road mobile source emissions under two ECT regional development scenarios: (1) Scenario A that is based on an extrapolation of recent land development trends and is characterized by a predominance of single-family residential expansion in currently undeveloped areas at the fringe of existing communities, and (2) Scenario D that concentrates growth in existing communities via extensive infill and redevelopment. With all other factors remaining unchanged, changes in on-road mobile source emissions are predicted to decrease daily maximum 8 hour ozone concentrations in the Austin MSA by 2 to 7 ppb. These results are consistent with the projected decreases in NOx and VOC emissions from 2007 to 2030 and are greater than but similar in magnitude and direction to impacts because of changes in land cover described above.
3. Development and Application of an Integrated Transportation-Land Use Model
A. Parcel-Level Models. As part of our integrated land use-transportation modeling effort, parcel-level models of land use were calibrated using 1990 and 1995 geographic information system (GIS)-encoded Austin maps. Parcels evolve in size and shape, not just land use; therefore, parcel subdivision was modeled explicitly, using a binary logit, before modeling land development (via a multinomial logit). A variety of lagged explanatory variables offered insight into land dynamics. These models recognized variables like parcel size and shape, slope, transit and CBD access, distance to nearest highway, and zoning policies, as well as each parcel’s “neighborhood” attributes. The results of the models indicate that neighborhood conditions offer substantial predictive power, but such effects seem inconsequential beyond 2 miles. In addition, the land use models were applied to predict 2005 land uses based on parcels that were undeveloped in the 2000 map. Actual 2005 land use data for a 3 mile by 3 mile study area were collected though onsite surveys. Comparisons of the model-predicted “most likely” land developments to actual 2005 development reveal strengths and limitations of the models. Together with forthcoming models of land use intensity (i.e., population and employment levels, by type), these models provide all key inputs needed for application of standard travel demand models as well as for estimating emissions from area sources for air quality analyses. Moreover, the land use models identify the factors and policy variables that influence land development. Adjustment of these factors and policies could alter the way that lands are utilized, moderating the negative impacts of land development on future traffic conditions and air quality.
B. Microsimulation of Residential Land Use. Residential land use is essential to urban areas, accounting for roughly 60 percent of developed land. Moreover, the emergence of commercial, industrial, office, and civic uses is correlated spatially with residential development. Microeconomic theories tested using disaggregate spatial data offer behavioral foundations for a better understanding of such relationships and human settlement patterns. A household-based residential location choice model was calibrated using recent relocation survey results for the Austin region, and this model was used to simulate equilibrium outcomes of residential land markets, using bid-rent theory. The model considered taste heterogeneity of individual households via random utility maximization and controlled for as well as predicted a variety of variables, including the age of home, value per interior square foot, total interior square footage, household annual income, network commute time, and Euclidean distances to workplace(s) and to the nearest shopping mall. A microsimulation procedure endogenously determined the spatial distribution of households and equilibrium home prices. The results indicate how undeveloped parcels near employment sites enjoy higher purchase prices and the value of travel time affects the spatial pattern of home prices and the demographic distributions.
This microsimulation of residential land use is one component of our future ITLUM, which seeks to include firms, land owners/developers, and the interaction between residential and firm location choices. The systematic and integrated models make a contribution to the behavioral foundation of land development prediction in urban regions. The behavioral foundation ensures optimal allocation of land in the sense that each firm and household chooses a location that most satisfies it, whereas developers/land owners maximize profits/rents. As a result, the models produce more realistic and reliable projections for future land use patterns.
C. Modeling of Land Cover Using Satellite Data. To capture the spatial and temporal effects in land cover evolution, a mixed logit model incorporating spatial and temporal correlations was specified and calibrated. The model is new, recognizing both spatial and temporal effects. Moreover, it undertakes several relatively unusual techniques: Halton sequences are shuffled randomly and Cholesky decomposition of nonpositive definite matrices is achieved. Lagged variables are key to the model’s specification, as is explicit recognition of unequal time intervals between each of the three satellite images. A series of neighborhood variables, computed for each cell in turn, control for population and local development. And uncertainties in predictions are quantified via an entropy statistic. Such statistical specifications and techniques should prove valuable for future studies on land use and land cover, as well as others that must consider spatial and temporal effects in a discrete response setting.
The model results suggest strong agglomeration or clustering effects in development (by similar types of land cover). After controlling for neighboring land covers, future development is most likely in areas offering lower residential densities. Entropy indices indicate greatest land use uncertainty at the “edges” of distinct land cover types. Model predictions may be used as inputs for integrated land use-transport-environment models.
The focus during Year 2 of the project will be on completing the evaluation of the air quality impacts of the ECT development scenarios. The air quality modeling will be expanded to address the relative impacts of development on changes in emissions from other anthropogenic sources (i.e., point, area, nonroad mobile sources) in addition to on-road mobile sources that were explored during 2005. The impacts of different development scenarios on tree cover also will be investigated. Comprehensive changes in anthropogenic and biogenic emissions for the four ECT scenarios will be modeled using CAMx at the conclusion of this phase of the study.
Development of the ITLUM will be expanded in several areas. A GIS data file that combines all protected land, open water, and 100 year floodplains recently has become available to the research group. The land use model and the future land use projections will be revised to accommodate these land development constraints.
Land use intensity models describe the spatial distributions of households and jobs. Seemingly unrelated regression models incorporating spatial weight matrices will be calibrated using zonal data at the level of traffic analysis zones. These intensity models will loop with the land use models (as land use patterns are key explanatory variables to the spatial distribution of households and jobs and vice versa).
Further detailed household types will be considered in the microsimulation of residential land use to provide desired level of detail and realism. A similar microsimulation system will be constructed for industrial land uses. Firm life cycles (i.e., firm formation, firm growth, firm relocation, and firm dissolution) will be modeled using simultaneous equation models and discrete choice models.
Journal Articles on this Report : 1 Displayed | Download in RIS Format
|Other project views:||All 51 publications||14 publications in selected types||All 13 journal articles|
||Wang X, Kockelman K. Tracking land cover change in a mixed logit model:recognizing temporal and spatial effects. Transportation Research Record 2006;1977:112-120.||