C OUPLING C HEMICAL T RANSPORT M ODEL S OURCE A TTRIBUTIONS WITH P OSITIVE M ATRIX F ACTORIZATION : A PPLICATION TO TWO IMPROVE SITES IMPACTED BY WILDFIRES.

Slides:



Advertisements
Similar presentations
Source Apportionment of PM 2.5 in the Southeastern US Sangil Lee 1, Yongtao Hu 1, Michael Chang 2, Karsten Baumann 2, Armistead (Ted) Russell 1 1 School.
Advertisements

1 Policies for Addressing PM2.5 Precursor Emissions Rich Damberg EPA Office of Air Quality Planning and Standards June 20, 2007.
Critical Review and Meta-analysis of ambient particulate matter source apportionment using receptor models in Europe C.A. Belis, F. Karagulian, B.R. Larsen,
Sources of PM 2.5 Carbon in the SE U.S. RPO National Work Group Meeting December 3-4, 2002.
FIRE AND BIOFUEL CONTRIBUTIONS TO ANNUAL MEAN AEROSOL MASS CONCENTRATIONS IN THE UNITED STATES ROKJIN J. PARK, DANIEL J. JACOB, JENNIFER A. LOGAN AGU FALL.
Using field campaigns results to reduce uncertainties in inventories Wenche Aas, Knut Breivik and Karl Espen Yttri And material from: Eiko Nemitz (CEH,
Fire Modeling issues: fire effects on regional air quality under a changing climate Douglas G. Fox
1 Recent PM 2.5 Trends in Georgia André J. Butler Mercer University EVE 290L 14 April, 2008.
RECEPTOR MODELLING OF UK ATMOSPHERIC AEROSOL Roy M. Harrison University of Birmingham and National Centre for Atmospheric Science.
Evaluation of Secondary Organic Aerosols in Atlanta
Christian Seigneur AER San Ramon, CA
Source apportionment of the carbonaceous aerosol – Quantitative estimates based on 14 C- and organic tracer analysis 1.Norwegian Institute for Air Research.
Air Quality Impacts from Prescribed Burning Karsten Baumann, PhD. Polly Gustafson.
Recent Finnish PM studies / 2 examples. Characterizing temporal and spatial patterns of urban PM10 using six years of Finnish monitoring data Pia Anttila.
BRAVO - Results Big Bend Regional Aerosol & Visibility Observational Study Bret Schichtel National Park Service,
Fossil vs Contemporary Carbon at 12 Rural and Urban Sites in the United States Bret A. Schichtel (NPS) William C. Malm (NPS) Graham Bench (LLNL) Graham.
IMPROVE Corrects OC and EC for a Positive Artifact The positive artifact correction causes the organic and elemental carbon to approach zero as fine mass.
Illumination Independent Aerosol Optical Properties n Extinction Scattering Absorption n Volume scattering function (phase) n Transmittance.
(work funded through the Great Lakes Restoration Initiative)
WRAP Status + Fire Emissions Inventory Protocol for Regional Air Quality Analysis and Planning Support in the WRAP regionWRAP Tom Moore WRAP/Western Governors’
J. Zhou 1, X. Zhu 1, T. Wang 1, and X. Zhang 2 J. Zhou 1, X. Zhu 1, T. Wang 1, and X. Zhang 2 1 College of Resources and Information Tech., China University.
NATURAL AND TRANSBOUNDARY INFLUENCES ON PARTICULATE MATTER IN THE UNITED STATES: IMPLICATIONS FOR THE EPA REGIONAL HAZE RULE Rokjin J. Park ACCESS VII,
Air Quality Impact Analysis 1.Establish a relationship between emissions and air quality. AQ past = a EM past + b 2.A change in emissions results in an.
Angeliki Karanasiou Source apportionment of particulate matter in urban aerosol Institute of Nuclear Technology and Radiation Protection, Environmental.
Preparation of Fine Particulate Emissions Inventories Lesson 1 Introduction to Fine Particles (PM 2.5 )
U.S. aerosols: observation from space, effects on climate Daniel J. Jacob and funding from NASA, EPRI with Easan E. Drury, Tzung-May Fu Loretta J. Mickley,
MODELS3 – IMPROVE – PM/FRM: Comparison of Time-Averaged Concentrations R. B. Husar S. R. Falke 1 and B. S. Schichtel 2 Center for Air Pollution Impact.
Contribution from Natural Sources of Aerosol Particles to PM in Canada Sunling Gong Scientific Team: Tianliang Zhao, David Lavoue, Richard Leaitch,
WRAP COHA Update Seattle, WA May 25, 2006 Jin Xu.
Causes of Haze Update Prepared by Marc Pitchford for the 5/24/05 AoH conference call.
Properties of Particulate Matter Physical, Chemical and Optical Properties Size Range of Particulate Matter Mass Distribution of PM vs. Size: PM10, PM2.5.
Application of Combined Mathematical and Meteorological Receptor Models (UNMIX & Residence Time Analysis) to IMPROVE Aerosol Data from Brigantine.
The Use of Source Apportionment for Air Quality Management and Health Assessments Philip K. Hopke Clarkson University Center for Air Resources Engineering.
Estimating the Contribution of Smoke and Its Fuel Types to Fine Particulate Carbon using a Hybrid- CMB Model Bret A. Schichtel and William C. Malm - NPS.
Analysis Examples and Issues: Identifying Sources Policy Analysis Tools for Air Quality and Health A workshop hosted by NERAM and Pollution Probe Jeffrey.
VISTAS Emissions Inventory Overview Nov 4, VISTAS is evaluating visibility and sources of fine particulate mass in the Southeastern US View NE from.
Causes of Haze Assessment (COHA) Update. Current and near-future Major Tasks Visibility trends analysis Assess meteorological representativeness of 2002.
1.
Model Evaluation Comparing Model Output to Ambient Data Christian Seigneur AER San Ramon, California.
Regional Air Quality Modeling Results for Elemental and Organic Carbon John Vimont, National Park Service WRAP Fire, Carbon, and Dust Workshop Sacramento,
Source Attribution Modeling to Identify Sources of Regional Haze in Western U.S. Class I Areas Gail Tonnesen, EPA Region 8 Pat Brewer, National Park Service.
OVERVIEW OF ATMOSPHERIC PROCESSES: Daniel J. Jacob Ozone and particulate matter (PM) with a global change perspective.
Bret A. Schichtel Center for Air Pollution Impact and Trend Analysis (CAPITA) Washington University St. Louis, MO, Presented at EPA’s National Exposure.
Office of Research and Development National Exposure Research Laboratory, Atmospheric Modeling and Analysis Division 16 October 2012 Integrating source.
WRAP Regional Modeling Center, Attribution of Haze Meeting, Denver CO 7/22/04 Introduction to the the RMC Source Apportionment Modeling Effort Gail Tonnesen,
AoH/MF Meeting, San Diego, CA, Jan 25, 2006 WRAP 2002 Visibility Modeling: Summary of 2005 Modeling Results Gail Tonnesen, Zion Wang, Mohammad Omary, Chao-Jung.
NPS Source Attribution Modeling Deterministic Models Dispersion or deterministic models Receptor Models Analysis of Spatial & Temporal Patterns Back Trajectory.
Attribution of Haze Report Update and Web Site Tutorial Implementation Work Group Meeting March 8, 2005 Joe Adlhoch Air Resource Specialists, Inc.
Breakout Session 1 Air Quality Jack Fishman, Randy Kawa August 18.
Fairbanks PM 2.5 Source Apportionment Using the Chemical Mass Balance (CMB) Model Tony Ward, Ph.D. The University of Montana Center for Environmental Health.
CHARACTERIZING IMPACTS OF WILD AND PRESCRIBED FIRES ON AMBIENT FINE PARTICLE CONCENTRATIONS CSU Atmospheric Science Department National Park Service/CIRA.
BACKGROUND AEROSOL IN THE UNITED STATES: NATURAL SOURCES AND TRANSBOUNDARY POLLUTION Daniel J. Jacob and Rokjin J. Park with support from EPRI, EPA/OAQPS.
Workshop on Air Quality Data Analysis and Interpretation Evaluation of Emission Inventory.
Source apportionment of submicron organic aerosols at an urban site by linear unmixing of aerosol mass spectra V. A. Lanz 1, M. R. Alfarra 2, U. Baltensperger.
V:\corporate\marketing\overview.ppt CRGAQS: CAMx Sensitivity Results Presentation to the Gorge Study Technical Team By ENVIRON International Corporation.
EPA’s SPECIATE 4.4 Database: Development and Uses Office of Research and Development National Risk Management Research Laboratory, Air Pollution Prevention.
Properties of Particulate Matter
October 1999PM Data Analysis Workbook: Glossary1 Glossary and Acronyms This section provides definitions of acronyms and terms used in the workbook. A.
Garfield County Air Quality Monitoring Network Cassie Archuleta Project Scientist Board of County Commissioners – Regular Meeting.
Mobile Source Contributions to Ambient PM2.5 and Ozone in 2025
Air Pollution and Stratospheric Ozone Depletion
Source Apportionment of Water Soluble Elements, EC/OC, and BrC by PMF
Sunil Kumar TAC, COG July 9, 2007
Aerosol chemistry studies at the SMEARIII station in Kumpula
Svetlana Tsyro, David Simpson, Leonor Tarrason
On-going developments of SinG: particles
Status of data from EMEP intensive period 2008/2009
RECEPTOR MODELLING OF AIRBORNE PARTICULATE MATTER
Data Analysis Techniques
Svetlana Tsyro, David Simpson, Leonor Tarrason
Presentation transcript:

C OUPLING C HEMICAL T RANSPORT M ODEL S OURCE A TTRIBUTIONS WITH P OSITIVE M ATRIX F ACTORIZATION : A PPLICATION TO TWO IMPROVE SITES IMPACTED BY WILDFIRES Sturtz et. al ATMS 790 seminar Ashley Pierce

O UTLINE Background Source Apportionment Positive Matrix Factorization (PMF) Chemical Transport Model (CTM) Hybrid The study Model comparison and evaluation Implications 2

B ACKGROUND Particulate matter (PM 2.5 ): mixture of small particles and liquid drops Aerosol: PM suspended in a gas (e.g. air) Volatile Organic Compounds (VOCs): variety of chemicals (benzene, isoprene) Secondary Organic Aerosols (SOA) Carbonaceous aerosols – major component of fine particulate mass IMPROVE: Interagency Monitoring of Protected Visual Environments (1985) National Ambient Air Quality Standards (NAAQS) 3

S OURCE A PPORTIONMENT 4 Source (Anthropogenic combustion, biomass burning, biogenic emissions from plants) Receptor Affected site/organism or measurement area Source-Receptor relationship: determining the role of meteorology and physical/chemical effects linking source emissions to receptor concentrations Source profiles: the experimentally determined unique proportion of species concentrations (or speciated PM or speciated aerosol) from a source Ex. Biomass burning – OC, EC, levoglucosan Feature: A factor profile (source factor) unique proportion of speciated aerosols determined by the PMF analysis Primary pollutants Emission Primary & secondary pollutants Transport, transformation, & removal processes Primary & secondary pollutants Transport, transformation, & removal processes

P ARTICULATE CARBON Fossil carbon: coal, oil, gas fuels Biogenic carbon: biomass burning, meat cooking, Secondary organic aerosols (SOAs) Upper bound for biomass burning contribution 5

6 More volatile, lower light absorption Higher light absorption

I MPORTANCE Adversely affect health, contribute to haze, affect radiation balance Biogenic sources of carbonaceous aerosols % of fine particulate carbon in rural areas ~50% in some urban areas Carbonaceous species often largest contributor to haze and PM 2.5 Smoke thought to be large contributor (W and SE U.S.) Difficult to apportion smoke from other emissions or between smoke types >50% of smoke particulate mass can be secondary organic aerosol (SOA) Similar to SOA composition formed from gases emitted by plant respiration Biomass burning emissions inventories likely overestimate PM emissions, underestimate VOC emissions from biomass combustion and biogenic release 7

P OSITIVE M ATRIX F ACTORIZATION (PMF) 8

PMF DISADVANTAGES True source profiles not known (no emission info) Requires assumptions that are not always true Can’t apportion secondary organic aerosols to source types Factor profiles can have large errors and may correspond to a mixture of source types Uncertainties in measurements are not always known or well-defined 9

PMF ADVANTAGES 10

C HEMICAL T RANSPORT M ODEL (CTM) CAPITA Monte Carlo Lagrangian CTM Direct simulation of atmospheric pollutants Each emitted quantum contains a fixed quantity of mass for various pollutants based on the source emission rate Individual particles subjected to transport, transformation, and removal processes 6-day back trajectories of air masses using meteorological data from the Eta Data Assimilation System (EDAS) Non-fire emissions: Western Regional Air Partnership (WRAP) 2002 emissions inventory Biomass burning: MODIS inventory Source profiles compiled from burns 11

12

CTM DISADVANTAGES Large information requirements Chemical mechanisms are incomplete Large errors and biases Particularly with wildfires Driven by emissions inventory Overall higher root mean square error (RMSE) than PMF and Hybrid 13

CTM A DVANTAGES Identification and separation of different source types based on emissions inventory Primary and secondary carbonaceous fine particles can be identified from source types biomass combustion, biogenic, mobile, area, oil, point, other 14

H YBRID Source-oriented Measured data used to constrain CTM Direct incorporation of measured data into model Post-processing of model results Receptor-oriented CTM results constrain receptor model (PMF) using Multilinear Engine-2 (ME-2) 15

MASS BALANCE 16

T HE S TUDY Goal: Distinguish source contributions to total fine particle carbon Biogenic sources Biomass combustion due to wildfires Using a receptor-oriented hybrid model 17

S ITES Speciated PM 2.5 from Monture and Sula Peak Montana Three year: Sula Monture

S PECIES Species with 0.2 ≤ S/N < 2.0 were down weighted by factor of 3 Removed species: S/N ratio <0.2 below detection limit missing > 50% samples Mass reconstruction outside IMPROVE limits 8% samples from Monture 25% samples from Sula Looked at 23 species 19

S OURCES 20

21 Missoula paper mill & mining Gold, cobalt and Molybdenum mines Dry soils, Long-range Transport, Fires?

C OMPARISON 22

S EASONS (CTM AND H YBRID ) Winter (Dec Jan Feb) Spring (Mar Apr May) Summer (Jun Jul Aug) Autumn (Sep Oct Nov) 23

M ODEL EVALUATION 24

M ODEL EVALUATION 25

M ODEL E VALUATION PMF γ = 0 CTM γ = 1 Monture γ = 0.83 Sula γ =

H YBRID D ISADVANTAGES Still unable to distinguish between primary and secondary biomass combustion impacts CTM model predictions were highly correlated Equation 2 should account for multiplicative bias but does not work with high correlation and no tracer species Requires experts to run model in current form 27

H YBRID A DVANTAGES Complementary attributes from PMF and CTM Directly applying the CTM predictions to the PMF model allows for resolution of sources not identified by the PMF alone Biogenic vs. biomass combustion Theoretically primary and secondary features should be distinguishable 28

I MPLICATIONS Accurate identification of relevant sources and impact on receptors will guide control policy lower costs better results Ability to better distinguish sources to prove pollution events are due to exceptional events such as wildfires 29

R EFERENCES Norris, G., & Vedantham, R. (2008). EPA Positive Matrix Factorization (PMF) 3.0 Fundamentals & user guide. Paatero, P. (1999). The Multilinear Engine—A Table-Driven, Least Squares Program for Solving Multilinear Problems, Including the n-Way Parallel Factor Analysis Model. Journal of Computational and Graphical Statistics, 8 (4), doi: / Polissar, A. V., Hopke, P. K., Paatero, P., Malm, W. C., & Sisler, J. F. (1998). Atmospheric aerosol over Alaska: 2. Elemental composition and sources. Journal of Geophysical Research: Atmospheres, 103 (D15), doi: /98JD01212 Ramadan, Z., Eickhout, B., Song, X.-H., Buydens, L. M. C., & Hopke, P. K. (2003). Comparison of Positive Matrix Factorization and Multilinear Engine for the source apportionment of particulate pollutants. Chemometrics and Intelligent Laboratory Systems, 66 (1), doi: (02) http://dx.doi.org/ /S (02) Schichtel, B., Fox, D., Patterson, L., & Holden, A. Hybrid Source Apportionment Model: an operational tool to distinguish wildfire emissions from prescribed fire emissions in measurements of PM2.5 for use in visibility and PM regulatory programs. Schichtel, B. A., & Husar, R. B. (1997). Regional Simulation of Atmospheric Pollutants with the CAPITA Monte Carlo Model. Journal of the Air & Waste Management Association, 47 (3), doi: / Schichtel, B. A., & Husar, R. B. (1997). The Monte Carlo Model: PC-Implementation. Retrieved 02/16/15, 2015, from Schichtel, B. A., Malm, W. G., Collett, J. L., Sullivan, A. P., Holden, A. S., Patterson, L. A.,... Barna, M. G. (2008). Estimating the contribution of smoke to fine particulate matter using a hybrid-receptor model. Paper presented at the Air and Waste Management aerosol and atmospheric optics. Sturtz, T. M., Schichtel, B. A., & Larson, T. V. (2014). Coupling Chemical Transport Model Source Attributions with Positive Matrix Factorization: Application to Two IMPROVE Sites Impacted by Wildfires. Environmental Science & Technology, 48 (19), doi: /es502749r 30

EXTRA 31

M ULTILINEAR E NGINE -2 (ME-2) performs iterations via a preconditioned conjugate gradient algorithm until convergence to a minimum Q value Conjugate gradient algorithm: algorithm for the numerical solution of particular systems of linear equations, usually a symmetric, positive-definite matrix 32

33

34