Approaches and Limitations in Assessing Public Health Impacts of Transportation Infrastructure Chad Bailey US EPA OTAQ January 24, 2008 UC Davis Overview • Relevant policy context • Health and transportation • General approaches taken • Review of common analysis methods used in analysis of transportation projects • Proposals for improvement Policy Context • National Environmental Policy Act – Requires documentation and consideration of impacts of major Federal activities, including infrastructure decisions – Wave of EIS documents addressing near- road issues in context of toxics • NAAQS Attainment Demonstration – Areas with localized hotspots – Benefits and costs of local controls Demand for Where health fits… time Public Health Air pollution Demand for Obesity goods Personal Gas activity SOV tax Transit Land Use Rail, ships, Highways Petroleum planes Water quality Trucks Biofuels Puller IM, rail, air Invasive Vehicle terminals species Demand for Infrastructure real estate Impact Traditional NEPA Focus Demand for time Public Health Air pollution Demand for Obesity goods Personal Gas activity SOV tax Transit Land Use Rail, ships, Highways Petroleum planes Water quality Trucks Biofuels Puller IM, rail, air Invasive Vehicle terminals species Demand for Infrastructure real estate Impact Spectrum of Approaches (What I‟ve heard out there) Qualitative discussion Emissions (Regional/ Corridor) Emissions (Link-level) Dispersion Quantitative risk characterization Spectrum of Approaches (What I‟ve heard out there) Qualitative discussion Emissions (Regional/ Corridor) Emissions (Link-level) Dispersion Quantitative risk Description of health concerns, characterization emission sources, population impacts, weight of evidence, and uncertainties Spectrum of Approaches (What I‟ve heard out there) Qualitative discussion Emissions (Regional/ Corridor) Emissions (Link-level) Dispersion Quantitative risk Emissions inventory for characterization large areas without local detail. Spectrum of Approaches (What I‟ve heard out there) Qualitative discussion Emissions (Regional/ Corridor) Emissions (Link-level) Dispersion Quantitative risk characterization Quantifies and presents emissions on each road link affected by a project Spectrum of Approaches (What I‟ve heard out there) Qualitative discussion Emissions (Regional/ Corridor) Emissions (Link-level) Dispersion Quantitative risk characterization Estimates ambient concentration of relevant species and compares with some threshold of concern (e.g. NAAQS) Spectrum of Approaches (What I‟ve heard out there) Qualitative discussion Emissions (Regional/ Corridor) Emissions (Link-level) Dispersion Quantitative risk characterization Predicts cancer and non-cancer risks under EPA toxics framework (URE/RfC) Qualitative Approaches • Done well • Done poorly – Objectively describes – “Cherry picks” information relevant issues to support desired outcome – Makes statements in – Assertions made without keeping with current regard to professional professional norms within practices and best relevant fields practices – Addresses areas where – Uses laundry list of quantifiable impacts are not uncertainties to cast doubt possible on health concerns – Describes consequences – Treats quantifiable of being unable to quantify information as the only impacts useful material P.S. As currently implemented, no discernable impact on decision making! Example • “The tools to predict how MSATs disperse are also limited. The EPA's current regulatory models, CALINE3 and CAL3QHC, were developed and validated more than a decade ago for the purpose of predicting episodic concentrations of carbon monoxide (CO) to determine compliance with the NAAQS. The CALINE4 model used in California is an improvement on the CALINE3-based EPA models, but like them was built primarily for CO analysis, has not been specifically validated for use with other materials such as MSATs.” – CalTrans District 7 I-5 Improvement FEIS Emissions-only Analyses • Done well • Done poorly – Characterizes trends in – Selects only “hotbutton” emissions from all relevant pollutants processes (tailpipe, evap, road – Limits analysis to very small dust, tire/brake wear) number of factors without – Reveals influence of regard to influence on results technology, traffic flow and – Limits trends analyses to composition, land use choices “official” land use future – Presents scenarios reflecting – Aggregates emissions inputs different assumptions and and reporting to larger spatial inputs (e.g. speed model) domains without information – Identifies alternatives that on proximity to sensitive bring emissions closer to receptors populations – Dismisses local emissions as – Presents local-regional a trivial % of regional inventory tradeoffs (e.g. IM freight) Example: Atlanta‟s “Northwest Corridor” Toxics Paradigm Approaches • Done well • Done Poorly – Quantifies locations of – Suggests excessive highest population precision exposure • Lung cancer – Provides objective – Reduces all metrics for comparison information to “a between scenarios number” – Allows transparent – Downplays discussion of unquantifiable risks acceptable risk – Non-cancer outcomes treated under threshold paradigm Example: Diesel PM at Truck Stops Best AERMOD Run with Background Excluded 12 10 16 2 8 10 8 4 6 22 10 12 14 24 16 18 20 22 14 Hartley et al., 2006 The typical tools of the trade… • Emission factors – MOBILE6.2 or CT-EMFAC • Dispersion model – CALINE3/4, CAL3QHC, or AERMOD • Simplified exposure assumptions – Residential outdoor receptors, no indoor sources – Some “sensitive” receptors (e.g. schools) – Time-activity models rarely used for projects • Dose-response estimates or NAAQS – Cancer IUR and non-cancer RfC/RfD from IRIS or OEHHA – CO, PM10, PM2.5, NO2 NAAQS Missing in Most EIS Analyses • An integrative approach 1. Objective characterization of qualitative information (e.g. hazard ID) 2. Identification of influential assumptions and variables (e.g. “upstream” drivers) 3. Use of ranges of impact for uncertain quantitative information 4. Verification of coherence with best practices in AQ, risk assessment, and public health decision-making 5. Discussion of literature of mitigation Missing (1): Objective characterization of qualitative information • FHWA (2006): “Some recent studies have reported that proximity to roadways is related to adverse health outcomes -- particularly respiratory problems. Much of this research is not specific to MSATs, instead surveying the full spectrum of both criteria and other pollutants. The FHWA cannot evaluate the validity of these studies, but more importantly, they do not provide information that would be useful to alleviate the uncertainties listed above and enable us to perform a more comprehensive evaluation of the health impacts specific to this project.” • American Academy of Pediatrics (2004): “Siting of school and child care facilities should include consideration of proximity to roads with heavy traffic and other sources of air pollution. New schools should be located to avoid „hot spots‟ of localized pollution.” Recent Literature Reviews • Exposure: – Zhou and Levy (2007) BMC Public Health: “our findings emphasize that policymakers should be able to develop reasonable estimates of the „zone of influence‟ of mobile sources, provided that they can clarify the pollutant of concern, the general site characteristics, and the underlying definition of spatial extent that they wish to utilize” • Health – Salam et al. (2008) Curr Opin Pulm Med: • “There is consistent evidence that living near traffic sources is associated with asthma occurrence and exacerbations.” – Samet (2007) Inhal Tox: • “To date, the findings indicate associations of traffic indicators with increased risk for multiple adverse health effects including asthma and allergic diseases, cardiac effects, respiratory symptoms, reduced lung function growth, adverse reproductive outcomes, premature mortality, and lung cancer.” – Adar et al. (2007) Inhal Tox: • “In summary, we found consistent evidence from a variety of study designs that links traffic-related pollution with adverse cardiovascular health outcomes.” Observation • Transportation agencies use – decision criteria more tolerant of Type II (false negative) than Type I (false positive) errors in regards to public health – use information quality criteria more stringent than journal editors and reviewers • Many agencies and parties – Focus on “high profile” pollutants, such as diesel PM and air toxics – Implicitly tie health to technology, which may/not be accurate assumption… Defining Environmental Progress 2003 EPA Emission Trends Report Evidence that reducing tailpipe PM helps… Li et al, 2003 …but not 100% • Lwebuga-Mukasa et al. (2004) – 10 year retrospective study of residents along Peace Brdige border crossing corridor in Buffalo, NY area – Highest astma prevalance and health care utilization rates in ZIP codes closest to bridge complex – From 1991-1996, overall national asthma hospitalization declined, while it increased by a statistically significant amount during the same period in the study area – Authors attribute to NAFTA Missing (2): “Upstream” influences on air quality projections • Gaps in regional travel models (ref: TRB Report 288) – Very poor ability to predict goods movement or travel time reliability – Large traffic analysis zones limit analysis of non-motorized travel – Generally fixed land use scenarios • “Forecast by negotiation” in around 1/3 of MPOs • Assume that local zoning/land use plans are static – Limitations prevent “out of box” thinking about solutions: • Behavioral strategies • Congestion pricing • Intermodal vs. highway freight trade-offs • Farmer‟s markets vs. international food trade • Urban bike rental fleets • Private transit markets, etc. Addressing Regional Travel Model Limitations at the Project Level • Make influential assumptions known – Projections based on static inputs are only one scenario of the future • Explicitly outline which decisions affect projections (in and out of project manager control) – Local land use authority aims? – Multimodal availability? An Example of Land Use-Highway Interaction • Wasatch Front Regional Council model – Baseline land use assumptions compared with UrbanSim land use simulator • Studied effect of land use changes on congestion estimates • Large differences found in areas of high low-density growth Waddell et al. (2003) Missing (3): Expressing Uncertain Information as Ranges • Not every quantity can be characterized with precision, but that doesn‟t imply irrelevance • Important to objectively describe ranges of data, particularly when diverse sets of inputs are possible (controversy) Addressing Uncertain Health Information • Ballpark magnitudes of health effects – “What if” these effects are real? – “What if” attributable to toxics, PM, or road dust? – How do alternatives vary in their potential impacts? • Approaches addressing uncertainty – Decision tree (branches assigned probability that health result is true, based on expert judgment)? – Insurance simulation (what would zero-profit life and health insurers charge for standard coverage based on strength and likelihood of causal association)? – Advanced simulation (e.g. prospective life table) Example Decision Tree to Show Ranges (Way too simple, also should apply to mitigation, costs) For two project alternatives. Given: •50,000 person study area •3% more people “exposed” by living within 100 m of freeway or 50 m of major urban road for build Hoek vs. no build p •Relative risk of exposure = 1.41 (Hoek), 1.18 Mortality q (Finkelstein), 0 (Reject studies as false) Finkelstein Effect Excess Mortality = 1-p-q Annual Mortality Rate * Excess Risk from False Exposure * Total Population * Fraction of Population Exposed •(816/100K)(1.41-1)(50,000)(3/100) = 5.02 •(816/100K)(1.18-1)(50,000)(3/100)=2.20 •(816/100K)(1-1)(50,000)(3/100)= 0 E(Excess Mortality) = p + 0.44q +0(1-p-q) Assume p = 0.25, q = 0.25, E = 1.8 Missing (4): Coherence with Professional Practice • Agencies and advocates often use methods that favor their desired outcomes – “Summary of Existing Credible Scientific Evidence Relevant to Evaluating the Impacts of MSATs” (FHWA Interim Guidance on Air Toxics) – Sierra Club‟s 2004 Highway Health Hazards • Rarely references totality of relevant literature • Commonalities – Lack of public process or peer review – Continual “reinvention” of the wheel Example • “But, although transportation and air quality models are constantly being tested and improved, models to calculate the dispersion of PM2.5 and air toxics, and the resulting concentrations at any given point, have not been adopted for regulatory use.” – Michigan DOT, DEIS for Detroit Intermodal Freight Terminal Improving Data Quality • Quality assurance project plans (QAPP) – Ensures relevant data and assumptions are made explicit – Ensures that standards for information acceptability are documented – Places all contributors to analytic outcomes on level playing field (substantive, procedural) • Peer review • All Federal (and some state) agencies are subject to information quality requirements Missing (5): Mitigation • Generally limited to construction: – Low-sulfur diesel in diesel equipment – Idling restriction – Retrofit in some cases • No discussion of novel mitigation options – Windbreaks (e.g. barriers, vegetation) – HVAC intake location – Longer-wearing concrete – Truck routes – Weight and length limits – The list goes on… Conclusions • Commonly employed methods for analyzing impacts of transportation projects provide little public health information • Limitations include lack of objective qualitative information, “upstream” analytical problems, error preference bias, and lack of adherence to contemporary standards of practice • Numerous well-accepted methods exist for dealing with uncertain health information • Data quality objectives can help! Moving toward quality!