1 WILMINGTON AIR QUALITY STUDY Project Summary and Status Todd Sax Vlad Isakov Planning and Technical Support Division California Air Resources Board Presentation.
WILMINGTON AIR QUALITY STUDYProject Summary and Status Todd SaxVlad IsakovPlanning and Technical Support DivisionCalifornia Air Resources BoardPresentation to Modeling Working GroupMarch 16, 2004OutlineIntroduction and OverviewObjectivesConceptual PlanPreliminary ResultsEmissions InventoryReviewStatus and Preliminary ResultsIndustrial-Commercial FacilitiesNon-Port Mobile Source InventoriesPort Inventories - StatusModel Status and EvaluationOngoing WorkWilmington Air Quality StudyBarrio Logan project - first neighborhood assessment project. Neighborhood scale inventoryApplication of several local-scale and regional modelsWilmington study - next step in neighborhood assessment. Improved local-scale emissions inventory and inventory evaluationLarger modeling domainExpanded model application and evaluationWilmington DomainWilmington modeling sub-domain WAQS ObjectivesGoalsDevelop and evaluate inventory/modeling methods for assessing pollutant impacts at a fine resolutionConduct studies to assess inventory and modeling approaches for statewide assessmentKey QuestionsAre existing emissions inventories adequate for neighborhood assessment? What are the key data gaps?What are key pollutant, source impacts in Wilmington?Which models provide reliable results?How do we integrate model results?EmissionsIndustrial and Commercial Facilities Industrial facilities Non-diesel emissions from marine terminals Gasoline stations Dry cleaners Autobody shops Metal fabricators Magnet Facilities like warehouses and distribution centers that attract diesel on-road sources Dedicated, on-site off-road equipmentOn-Road Sources Automobiles and Heavy duty trucks Freeways, and Ramps Major and Minor ArterialsOther Off-Road Engines Marine, Harbor, and Dockside engines at marine terminals Railroad activityExposure Local scale modeling - ISCST3, AERMOD, CALPUFF, CALINE4 Regional modeling - CALGRID, CMAQ, CAMx Combined results Limited time-activity based exposure modelingHealth Risk OEHHA Guidelines - Inhalation and multipathway risks - Cancer and chronic endpoints - Comparison to health based PM standardsModel EvaluationTracer Study Summer, 2003 Release from elevated stack Toxics Monitoring Long term (one year), one site - >50 pollutants Short term study(12-15 days) - Summer, 2003 - Multiple sites - Estimate diesel PMUncertainty Assessment Gasoline service stations Stationary and Mobile Diesel IC enginesWilmington Neighborhood Assessment - Conceptual PlanInventory Analysis Expand quality assurance Assess contribution of neighborhood sources Evaluate uncertainty OutlineIntroduction and OverviewObjectivesConceptual PlanPreliminary ResultsEmissions InventoryReviewStatus and Preliminary ResultsIndustrial-Commercial FacilitiesNon-Port Mobile Source InventoriesPort Inventories - StatusModel Status and EvaluationOngoing WorkEmissions: Industrial and Commercial Facilities405 facilities-toxics / 259 -criteria170 surveyed facilities(118 neighborhood / 52 CEIDARS)Compiled from multiple inventory databasesEnhanced QA/QCReview by SCAQMD and selected facilitiesOn-Road EmissionsLink-Based Inventory Use Travel Demand Models and EMFAC Marine Terminals and Related Off-RoadPorts of Los Angeles and Long Beach - develop inventories for marine terminals, on-road sources, and related locomotive emissions.Locomotives - develop link and throttle-notch specific inventoriesConstruction - not considered (included in regional modeling).Emissions Inventory ReviewOutlineIntroduction and OverviewObjectivesConceptual PlanPreliminary ResultsEmissions InventoryReviewStatus and Preliminary ResultsIndustrial-Commercial FacilitiesNon-Port Mobile Source InventoriesPort Inventories - StatusModel Status and EvaluationOngoing WorkIndustrial-Commercial FacilitiesDefinitionLarge and small point sources at non-port businessesMethodDevelop facility listMultiple data sources: HRA, AER, CEIDARS, TRI, etc. On-site surveys: verify and augment inventories118 neighborhood sources52 CEIDARS facilitiesChoose best emissions data from hierarchyIf surveyed, include on-site area and mobile emissions categoriesCompile inventoryIndustrial-Commercial FacilitiesHierarchyTable 1Number of facilities in final inventories by data source.Number of FacilitiesData SourceToxics InventoryCriteria InventoryCEIDARS SurveysPrimary Data Source: Health Risk Assessment10Primary Data Source: Annual Emissions Report2828Health Risk Assessments7Air Toxic Inventory Reports (hardcopy files)8Neighborhood Source Surveyswith no additional data115with limited CEIDARS data2with limited AER data2Limited Surveys (AQMD Annual Emission Reports)1213AQMD Annual Emission Reports1998-199911161999-200031492001-20024LAUSD Surveys381ARB Emissions Inventory Database (CEIDARS)Criteria Database74Toxics Database69Both16Energy Commission List of Emergency Generators3132Toxics Release Inventory, Year 20009AQMD Permits - ARB Emission Estimates1642Total405259Industrial-Commercial FacilitiesTable 2 Industrial-Commercial Facility Emission Inventory and Contribution to Potency Scores for Selected Pollutants in the Wilmington Modeling Domainxy.PollutantEmissions (lbs/yr)Percent of Total Cancer ScorePercent of Total Chronic ScoreAmmonia1300000--Industrial-Commercial FacilitiesPreliminary Results: Inventory EvaluationDesigned to test inventory assumptionsWhy evaluate inventories?Existing databases designed for regional-scale analysisInventory update procedures designed and implemented with regional goal in mindBut NAP is local scale, not regional analysisAsking existing databases to do moreNeed to understand strengths and limitationsLearn how to improve and meet modeling needsIndustrial and Commercial FacilitiesDevelopment of a community-specific industrial-commercial facility inventory improved our ability to characterize emissions in WilmingtonWAQS inventory is more recently calculatedToxics Inventory Age65% of records identified by survey; year 2000 or laterCriteria Inventory Age55% or records in local-scale inventory updated by survey (>2000)WAQS is more comprehensive than CEIDARSContains small facilities that are area sources in CEIDARSContains improved stack data in toxics inventory64% of releases are actual data; 36% defaultsOnly 8% of CEIDARS records tied to stacksDuplicate, closed CEIDARS facilities corrected.Industrial-Commercial FacilitiesTotal facility cancer scores differ substantially between inventories.Industrial and Commercial FacilitiesOn a neighborhood scale, diesel PM and CrVI from area-wide sources at facilities are significant80% of diesel PM and 15% of CrVI generated by facilities which are not in CEIDARS as point sources. Other neighborhood sources have minimal impacts, but may be important near receptors.Industrial and Commercial FacilitiesCurrent diesel exhaust particulate inventories representing industrial-commercial facilities need improvement for neighborhood assessmentsOnly ~20% of estimated diesel PM emissions at facilities generated by point sourcesRemaining ~80% generated primarily by off-road sources operating within facilities.Diesel PM from off-road sources is important at larger industrial facilities like petroleum refineriesOff-road diesel PM ~40% of total cancer potency-weighted emissions at refineries.I-C Diesel Exhaust Particulate Inventory75% generated by inventory-reporting facilities in 90744 (Wilmington community)But 23 reporters, ~600 neighborhood sources not surveyed in 90744If extrapolate, inventory doublesImplications of I-C DPMDPM is dominant cancer riskSignificant emissions generated by on-site off-road sourcesPoint source facilities generally do not report on-site mobile source inventoriesHowever, most on-site off-road emissions were generated by facilities subject to other inventory reporting requirementsStatewide inventory based on off-road modelTop-down approach4 km grid cell spatial resolutionIndustrial and Commercial FacilitiesPetroleum Refinery Case StudyMethodEvaluate inventory reports from 6 refineries3 in Wilmington, +1 in SCAQMD, +2 in BAAQMDAnalysis requires process-level inventoriesObtained best toxics data representing each facilityMust be consistently calculated, SCC process codedResult: ability to compare facilities is limited Different process groupings/units between facilitiesWidespread inconsistencies in facility calculationsTop pollutant sources different at different facilitiesNeed to examine other facility categories; results may be consistentExample: BenzeneFacility E: fugitive wastewaterFacilities B and C (AER): oil-water separators. B>C, due to activitySome totals different in AB2588, AERResults consistent for benzene, 1,3-B, H2S, CrVI, CHOH Case Study: Petroleum RefineriesTable 4-4 Top Three Benzene Emissions Sources by Process Hot Spots Data. (#) = Emissions (lbs/yr) by Process. FACILITYRankABCDEF1Fugitives, Not Classified (1000)Process Heaters, Process Gas (700)Process Heaters, Process Gas (300)Floating Roof Tanks (1100)Fugitive, Wastewater (6000)Fugitive Pipeline Valves, etc (1300)2Gasoline Engines (300)Fixed Roof Tanks (300)Fugitives, Not Classified (200)Fugitives Wastewater (400)Fugitive, Pipeline Valves, etc. (3000)Process Gas External Combustion (160)3Floating Roof Tanks (100)Floating Roof Tanks (300)Floating Roof Tanks (60)Fugitives, Not Classified (300)Fugitive, Not Classified (1400)Fugitives, Pump Seals (90)Table 4-5Top Three Benzene Emissions Sources by Process AER Data. (#) = Emissions (lbs/yr) by Process.FACILITYRankABC1Floating Roof Tanks (100)Fixed Roof Tanks (600)Process Heaters Process Gas (300)2Process Heater Process Gas (100)Fugitive Oil/Water Separator (400)Fugitive Oil/Water Separator (60)3Boilers Process Gas (50)Boilers Process Gas (300)Fugitive Valves (40)Case Study: Petroleum RefineriesSubstantial differences between identical facilities, different inventoriesMajor differences in facility-total emissions for high risk pollutants Case Study: Petroleum RefineriesWhen emissions data reported using comparable methods, gain insights.Example: Hexavalent Chromium (CrVI) generated by process-gas fired process heatersOn paper, majority of emissions generated by a few units at few facilitiesOutlineIntroduction and OverviewObjectivesConceptual PlanPreliminary ResultsEmissions InventoryReviewStatus and Preliminary ResultsIndustrial-Commercial FacilitiesNon-Port Mobile Source InventoriesPort Inventories - StatusModel Status and EvaluationOngoing WorkOn-Road Emissions InventoryGoal: develop and evaluate link-specific inventoryDevelop and test approaches for link-specific inventory developmentAssess assumptions in developing a bottom-up inventoryCompare to proposed approach for statewide modelingAssess uncertainty and how to improve calculationsPreliminary ResultsEmissions models need better resolutionEmissions estimates are uncertain due to uncertain activity estimates and uncertain emission factorsEmission models were never intended to provide highly spatially resolved emissions estimatesEMFAC and OFFROAD provide county-total emissions that can be allocated to 4 km grid cellsGreater inventory resolution is required for local-scale modelsAllocating emissions to roadways is uncertain due to county-level assumptionsFleet compositionTravel model limitations: link specific volumes and speedsOperating cycle / trip-based emission factorsMobile Emissions InventoriesLimited test data on diesel PM emissions complicates assessment of diesel PM impacts on a local level.Source test data are extremely limited~200 in-use heavy duty truck source testsNew data on-line with CRC E55-59OutlineIntroduction and OverviewObjectivesConceptual PlanPreliminary ResultsEmissions InventoryReviewStatus and Preliminary ResultsIndustrial-Commercial FacilitiesNon-Port Mobile Source InventoriesPort Inventories - StatusModel Status and EvaluationOngoing WorkEmissions Inventory - PortsPort-wide inventoriesGoal: obtain spatially resolved port-specific inventoriesWork supports WAQS and SSD Port Regulatory ActivitiesWork conducted by Port consultantsContinuous consultation with SSD, PTSDImprove spatial allocation - berth/terminal/rail-link specific Improve inventory assumptions: load, stacks, etc.Improved traffic and idling activity estimates - terminal specificStatus: Draft reports are being reviewed.Commercial marine vessels (POLA)Harborcraft (POLA / SSD)Terminal on-road movement/idling (POLA and POLB) Dockside terminal (POLA and POLB)Locomotives (POLA and POLB)OutlineIntroduction and OverviewObjectivesConceptual PlanPreliminary ResultsEmissions InventoryReviewStatus and Preliminary ResultsModel Status and EvaluationLocal-scale uncertainty analysisTracer study statusOngoing WorkModeling StatusMicroscaleStatus: waiting on port inventoriesRegionalStatus: currently being planned, sensitivity studies in progressModel IntegrationGoal: combine regional and microscale models while minimizing double countingStatus: currently being planned. Model Evaluation - Uncertainty AnalysisGoalUse uncertainty analysis as an objective evaluation procedure to determine the level of confidence we should have in modeling resultsTwo studiesDiesel PM Study in WilmingtonWilmington inventory sensitivity studiesWhat is uncertainty analysis?An analysis method that uses assumptions about the uncertainty in model inputs to assess uncertainty in model output.Model Evaluation - Uncertainty AnalysisWhy Uncertainty AnalysisModels are not realityModel results are a function of assumptionsAssumptions are uncertainWe make best guess estimates to simulate realityThese estimates may be wrongThese estimates are uncertain - we pick a value from a rangeWhat do we hope to learn?How uncertain are our estimates?What are the most uncertain components?How can we reduce uncertainty?Given uncertainty, what are model strengths and limitations?Wilmington Uncertainty Analysis (1)Diesel PM - ZIP 90744Industrial-Commercial facilitiesSurveyed and included in inventoriesExtrapolated, not in I-C inventory directlyOn-Road Major - Freeways, Ramps, Major ArterialsMinor - Minor arterials, Collectors, ConnectorsApproachAssess uncertainty in emissionsRun ISC for Base CaseAssess uncertainty in model results due to meteorology, inventory release characteristics.Develop Monte Carlo meta-model to estimate uncertainty in ISCST3 results(IC, on-road)Wilmington Uncertainty - EmissionsDiesel PM emissions: mobile sourcesMobile source DPM at 4 facilitiesTheoretical link Goal: assess precision, accuracy in emissions, apply to modeling analysisEmissions methodEstimate activity range by on-site surveyQuantify range of emission factors based upon source testsUse Monte Carlo to propagate uncertaintyOrder of magnitude uncertainty in mobile source diesel emissions estimates at facilitiesAssessed on-site on-road and off-road emissionsCase Study: Diesel Exhaust ParticulateCase Study: Diesel Exhaust ParticulateUncertainty is due to emission factorsLimited number of tests, all cycles considered.Order of magnitude uncertainty in on-road diesel emissions estimatesTheoretical link (1-mile, 100 HD, 5 LD, 30 MPH)Bias in Wilmington is likely (volume, fleet, EF)Case Study: Theoretical LinkWilmington Uncertainty Analysis MethodDivide model into componentsEmissions (EMS) Spatial Allocation (SA) Assessed by emissions source category, Moved a set distance to north, south, east, west: IC +/- 25m, ZNS +/- 200m, Major onroad - fixed, Minor roadways +/- 500m. Temporal Allocation (TA) Point sources - base scenario by survey (vary 8, 10, 12, 16, 24 hour day), Roadway sources (Vary temporal allocation +/- 2 hrs)Release parameters (RP) Point sources base case defined by survey, uncertainty using different assumptions: 3 volume scenarios, 3 point source scenarios, Roadways - base case area sources (3 different area source options)Meteorology (MET) Onsite data 2001 (Long Beach cloud data for stability), Assessed Long Beach - 1984-1990, 2001, Ran model, assess percent difference relative to 2001, Developed distribution for interannual variabilityRun Model Assess model differences based on uncertainty in each model componentAssign to distribution (in our case empirical for simplicity)Result - distribution of model results for each model component separatelyModel PropagationAssumes independence between factors in model Spatial allocation, temporal allocation, meteorology, release parameters. Emission rates are independent - unit emission ratesDevelop Monte Carlo propagation model (EMS x C) (SA + TA + RP + MET)Model is iterated for each source contribution to each receptor. ReceptorsChosen to represent different types of sitesUncertainty Analysis: Conceptual ApproachWilmington Uncertainty AnalysisResults: all receptorsReceptor 1: stationary and mobile impactedWilmington Uncertainty AnalysisWilmington Uncertainty AnalysisReceptor 4: residential non-impactedWilmington Uncertainty AnalysisReceptor 6: Wilmington Park ElementaryPreliminary ConclusionsEmissions from on-road sources may be underestimatedUncertainty in emissions appears the dominant sourceLocating emissions in the domain is most importantOnce located, uncertainty in calculations is dominant. No statistical difference between sitesDue to uncertainty in magnitude and location of emissionsModel results should be verified with monitoringConceptual model uncertainty due to model formulation needs to be includedWilmington Sensitivity Studies (2)ObjectiveDemonstrate the effect of different point source emissions inventories on model results using a simplified case study.MethodCompare different level of details in point source emissions inventoryNATA 1996, CEIDARS, WAQSUse NATA 1996 application, ASPEN modeling system for comparison. Assessing Uncertainty due to EmissionsCompare multiple point source inventoriesGoal: estimate uncertainty due to different levels of detail in point source inventories (national, statewide, local-scale)Modeling Domain:Focus on Wilmington sub-domain (10 x 10 km)Outside sources treat as backgroundModel all sources within 50km of domain. Compare with observations: Short term (~18 mo.) toxics monitor in domain.RoadlinksCensustractcentroidsModelreceptorsInventory identified, not in statewideEmissions InventorySources are more precisely located in statewide and local-scale inventoryLarge differences in inventory databasesModel ResultsResults agree with observations Background and mobile source contributions comparable and dominant contributors to riskPoint sources have impact when close to receptorsBenzene resultsHexavalent ChromiumCr (VI)Wilmington modeling domain10km x 10km size (blue box)major road links - black linescensus tracts centroids - black dotsSources of Cr (VI) emissions:Local-Scale (WAQS) - red symbolsStatewide (CEIDARS) - blue trianglesNational (1996 NTI) - yellow trianglesComparison: Local vs. Statewide Inventory Cr (VI) EmissionsX-axis:Statewide:(CEIDARS)Y-axis:Local-scale(Wilmington)Factor of 10Factor of 2Inventory identified, not in statewideEmissions InventorySources are more precisely located in statewide and local-scale inventoryLarge differences in inventory databasesFew sources account for almost all of difference, but moderate differences widespread at many sourcesAssuming 34% CrVI/Cr in NTI is simplified, conservative, has been improved in NTI2001.WAQS has more sources than statewide, but fewer emissionsModel ResultsResults agree with observationsPoint sources have impact when close to receptorsBackground appears consistent, low. CrVI resultsKey QuestionWhat level of resolution in analysis is defensible?What needs to be done to achieve required resolution?Potential AnswerCombination of inventory inputs and model sensitivity. Uncertainty Analysis ReportsCase Study in Regulatory Modeling Applications - Atmospheric Environment, 2003CRC Modeling Conference (2002)AWMA (2003) - Framework for Uncertainty Analysis. AWMA (2004) - DPM Uncertainty AnalysisAWMA (2004) - NATA Conceptual Uncertainty Future WorkOutlineIntroduction and OverviewObjectivesConceptual PlanPreliminary ResultsEmissions InventoryReviewStatus and Preliminary ResultsModel Status and EvaluationLocal-scale uncertainty analysisTracer study statusOngoing WorkTracer ExperimentsWhy tracer studies? existing micro-scale models (such as ISCST3 and AERMOD) have been developed and evaluated using tracer studies not specific to California (Prairie Grass study in Nebraska and Kincaid study in Illinois flat rural conditions, and in Indianapolis, Indiana - urban conditions). What are we doing? We are conducting new tracer studies focused on evaluating micro-scale models (ISCST3 and AERMOD) on both near-field and local scales. Results from these studies will help us understand model performance in California under study conditions and may lead to model refinement in the future.What needs to be done in the future? additional tracer experiments in California are required to analyze the full range of terrain / meteorological conditions in the state.Models need to be evaluated in complex coastal urban conditions common to the Bay area;Models need to be evaluated on-road on a local scale. Databases for model evaluation:Tracer Experiments:1)Near field tracer experiment in San Diego, Barrio Logan2)Field study at Dugway Proving Ground, Utah3)Near field tracer experiment in Riverside - CE-CERT dispersion experiment: trailer wake effect in an urban area4)Tracer experiment in urban areas - ground level release in San Diego, Barrio Logan 5)Tracer experiment in urban areas - elevated level release in Los Angeles, WilmingtonTRACER EXPERIMENTS IN WILMINGTONFigure 1: The growth of the nocturnal thermal urban boundary layer (TIBL)Stable boundary layerUrban boundary layerWilmington Tracer Experiment - P.O.L.A.Wilmington Tracer Experiment - P.O.L.A.Pilot field study - conducted in September, 2003Meteorological measurements: sodars, sonic anemometers, temperature, relative humidity and solar radiation sensorsOne daytime tracer release was performed between 8 a.m. and 2:30 p.m. on September 22nd. Afternoon/evening releases (3 p.m. - 11 p.m.) were performed on September 23rd, 25th and 26th. First three releases - using the 220 tall stack at the LADWP power plant and the fourth - surface release (~ 3 m). Tracer measurements: mobile van monitoring along several downwind transects and bag sampler samplers at three locations Performance audits were conducted on the wind sensors and tracer gas analyzerWilmington Tracer Pilot Study Status Model Evaluation - Tracer StudyReports:Barrio Logan Tracer Study Final Report - under reviewBarrio Logan Tracer Study Results - accepted, Atmospheric EnvironmentWilmington Tracer Pilot Study Status Report2002-2003 Presentations: CRC, EPA, Working GroupNear-Field Modeling for Regulatory Applications - in press, Journal of the Air and Waste Management AssociationOutlineIntroduction and OverviewObjectivesConceptual PlanPreliminary ResultsEmissions InventoryReviewStatus and Preliminary ResultsModel Status and EvaluationLocal-scale uncertainty analysisTracer study statusOngoing WorkModel Evaluation and Planning - RegionalGoal: minimize computing resourcesUseful to determine the impact of grid cell size on model resultsConducting sensitivity 4x4 vs. 12x12. 12x12 results drastically lower (factor 2-10 reduction in toxics)Compared 2x2 MATES-II vs ARB 4x4 regionalResults comparableComparing 6x6, 2x2 to 4x4 results (in progress)We are currently committed to 4x4 for statewide to leverage simultaneous, SIP-related work. Need to determine how many vertical layers are necessary to generate reliable resultsTesting 7 layers vs. standard 17 (in progress)Toxics MonitoringWilmington Toxics MonitoringGoals: evaluate combined microscale and regional modeling results for diesel PM using monitored concentrations of diesel indicatorsTest methodology suggested in ARB meetingsFocus on several sub-areas of Wilmington domainHawaiian Ave School and I110 impactsLong Beach Port and I 710 impactsSan Pedro port impacts and complex meteorologyIntegrate results with other studiesPTSD Freeway, RD RAV4, POLA programStatus: study in progressSchedule and PlanReceive Port inventories - first quarter 04Microscale modeling - summer 04Research to support statewide - ongoingdepends on peer review and working group commentsResearch is focused on answering peer-review questionsPublishing ensures feedback from scientific communityMakes peer-reviewers more confident of resultsProject report - mid-year 05Several reportsPublic consumption - e.g. Wilmington storyTechnical reportdiscussing results from modeling applications and associated researchrecommendations for statewide effortgeared for working group and peer-reviewers