The Air Quality Model Evaluation International Initiative (AQMEII) Christian Hogrefe 1, Stefano Galmarini 2, Efisio Solazzo 2, Ulas Im 3, Marta Garcia Vivanco 4,5, Augustin Colette 4, and AQMEII modeling groups 1 Computational Exposure Division, U.S. Environmental Protection Agency, RTP, NC, USA 2 European Commission, Joint Research Centre (JRC), Directorate for Energy, Transport and Climate, Ispra (VA), Italy 3 Aarhus University, Department of Environmental Science, Frederiksborgvej399, 4000 Roskilde, Denmark 4 INERIS, Institut National de l'environnement Industriel et des Risques, Parc Alata, 60550 Verneuil-en-Halatte, France 5 CIEMAT. Avda. Complutense, 40. 28040. Madrid, Spain Workshop on Measurement-Model Fusion for Global Total Atmospheric Deposition (MMF-GTAD) February 28 March 2, 2017 World Meteorological Organization, Geneva, Switzerland
Air Quality Model Evaluation International Initiative AQMEII Initiated in 2009 by researchers from North America and Europe with support from EPA, Environment Canada, and the European Commission s Joint Research Center AQMEII s goal is to bring together the North American (NA) and European (EU) regional-scale air quality modeling communities to Exchange expert knowledge in regional air quality modeling Evaluate regional air quality models to increase the knowledge on atmospheric processes and to support the use of models for policy development Prepare coordinated research projects and model intercomparison exercises
37 Groups 17 Countries 18 Models Participation in AQMEII (All Phases)
AQMEII Phase 1: 2009 2012 Focus on evaluating offline regional air quality models Build AQMEII infrastructure: Common inputs for emissions and boundary conditions Assembling of surface and upper air observations for model evaluation (focus on routine operational networks, e.g. AIRBASE, EMEP, NAPS, AQS) Annual simulations over NA and EU for 2006 23 participating groups from 15 countries in NA and EU Selected key findings: All models showed better performance for ozone than for PM 2.5 and PM 10 Surface wind speeds were generally overestimated by all participating models, likely contributing to a general underestimation of PM 10 concentrations Phase 1 database available for ongoing research
AQMEII Phase 2: 2012 2014 Focus on evaluating coupled meteorology/chemistry models Feedback effects from meteorology to air pollution include modulation of mixing (dilution), clouds (radiation & initiation of photochemistry, scavenging, temperature), and winds (transport) 18 participating groups from 10 countries Annual simulations for 2006 and 2010 over NA and EU Selected key findings: During the summer and during intense emission events (e.g. Russian forest fires of 2010), feedbacks can have a noticeable impact on air quality simulations The aerosol indirect effect tends to dominate over the aerosol direct effect, but also is very uncertain due to challenges in simulating clouds For annual statistics, model-to-model variations in performance are greater than the within-model variations associated with feedbacks Phase 2 database available for ongoing research
AQMEII Phase 3: 2014-2017 Contribute to the modeling and analysis activities performed under the umbrella of the Task Force on Hemispheric Transport of Air Pollution - HTAP2 Apply and compare modeling techniques for assessing the impact of longrange transport on regional air quality Focus year of analysis: 2010 (utilizes emission and observation databases compiled for Phase 2) 14 participating groups from 10 countries Joint HTAP, AQMEII and MICS special issue in ACP (deadline June 1, 2017)
AQMEII Timeline 007 2008 2009 2010 2011 2012 2013 2014 2015 2016 Planning for Future Collaborative Multi- Model Evaluation and Intercomparison 2017 Dennis Et Al, 2010: A Framework for Atmospheric Environment Studies Evaluating Regional-Scale Numerical Phase 1 Special Issue Photochemical Modeling Systems ACP Special Issue for Rao et al., 2011: AQMEII -Advancing the State of EM Special Issue -Phase 1 HTAP, MICS-Asia, and the Science in Regional Photochemical Modeling Highlights and Next Steps AQMEII Phase 3 and Its Applications Alapaty et al., 2012: New Directions: Atmospheric Environment Understanding Interactions of Air Quality Phase 2 Special Issue and Climate Change at Regional Scales
Focus Areas of AQMEII Research Model evaluation against data (mostly) from routine surface and upper air monitoring networks Model ensemble analysis Model intercomparison Model sensitivity simulations and diagnostic error analysis Most work to date has focused on ozone, CO, NO/NO 2, and particulate matter (PM) total mass Some analyses have focused on PM composition, aerosol/radiation interactions, meteorological fields, and (to a small extent) deposition The following slides provide examples of past and present AQMEII analyses
AQMEII Data Management and Analysis Approach: ENSEMBLE Platform Model Simulations Observations AQS NAPS CASTNET IMPROVE CSN EMEP AIRBASE WOUDC Ozonesondes MOZAIC
AQMEII Phase 1 Model-to-Model Comparison for Dry and Wet Deposition Substantial model-to-model variability in simulated deposition over EU Similar results for NA Solazzo, et al., Operational model evaluation for particulate matter in Europe and North America in the context of AQMEII, Atmos. Env., 55, 2012, http://dx.doi.org/10.1016/j.atmosenv.2012.02.045.
AQMEII Phase 2 Model-to-Model Intercomparison for Inorganic Aerosol Indicator Ratios (Winter) Multi-Model Mean Normalized Std. Dev. Total Nitrogen / Total Sulfur (TN/TS) TS = SO 4 2- TN HNO 3 + NO 3 - TA NH 3 + NH 4 + Adjusted Gas Ratio (AdjGR) AdjGR NH 3F /TN NH 3F TA -DSN x TS DSN ([NH 4+ ] -[NO 3- ])/[SO 4 2- ] Campbell et al., A multi-model assessment for the 2006 and 2010 simulations under the Air Quality Model Evaluation International Initiative (AQMEII) phase 2 over North America: Part I. Indicators of the sensitivity of O3 and PM2.5 formation regimes, Atmos. Env., 115, 2015, http://dx.doi.org/10.1016/j.atmosenv.2014.12.026.
Examples of Ongoing AQMEII3 Precipitation and Deposition Evaluation Model Simulations (mm) Annual Total Precipitation 3000 2500 2000 1500 1000 500 0 0 500 1000 1500 2000 2500 3000 Observations (mm) FI1_HTAP_bas DK1_HTAP_GLO TR1_MACC_bas UK1_MACC_bas Model Simulations (kg/km2) 5000 4000 3000 2000 1000 0 Wet Deposition of NO 3 FI1_HTAP_bas FI1_MACC_bas FI1_MACC_bas2 IT2_MACC_bas IT1_MACC_bas TR1_MACC_bas UK2_HTAP_bas ES1_MACC_bas 0 1000 2000 3000 4000 5000 Observations (kg/km2) Model Simulations (kg/km2) 2000 1800 1600 1400 1200 1000 800 600 400 200 DK1_HTAP_bas 0 Wet Deposition of NH 4 FI1_HTAP_bas FRES1_MACC_bas FI1_MACC_bas FI1_MACC_bas2 IT2_MACC_bas IT1_MACC_bas TR1_MACC_bas UK2_HTAP_bas DK1_HTAP_bas 0 500 1000 1500 2000 Observations (kg/km2) Analysis of AQMEII3 deposition fields is ongoing (lead: M. Garcia Vivanco)
Six-Model Comparison of Precipitation and Wet Deposition (Example Taken From EURODELTAIII Project) Precipitation (mm) Oxidized Nitrogen (ugn/m2) Oxidized Sulfur (mgs/m2) Reduced Nitrogen (ugn/m2) The EURODELTAIII project is conducted under UNECE Task Force on Measurements and Modeling (TFMM) for Europe Participation by six regional models (EMEP, CHIMERE, LOTOS- EUROS, CAMx, CMAQ, MINNI) Simulations for four month-long EMEP measurement campaigns during 2006, 2007, 2008 and 2009 Overlap between EURODELTAIII and AQMEII modeling groups and participants similar analyses are planned for existing and/or future AQMEII datasets Vivanco et al., Joint analysis of deposition fluxes and atmospheric concentrations of inorganic nitrogen and sulphur compounds predicted by six chemistry transport models in the frame of the EURODELTAIII project, Atmos. Env., 151, 2017, http://dx.doi.org/10.1016/j.atmosenv.2016.11.042.
Example of AQMEII3 Sensitivity Simulations: Variations in Boundary Conditions, Emissions, and Processes AQMEII3 Base Case Simulated by All Modeling Groups Sensitivity Simulations Leveraging HTAP2 Global Simulations Boundary Conditions C-IFS H-CMAQ GEOS-CHEM AM3 Zero Boundary Conditions BC Zero Photochemical Modeling (12km) CMAQ 5.0.2 Offline CMAQ 5.0.2 Offline No O3 Dry Deposition Meteorological Modeling (12km) WRF 3.4 Emissions AQMEII2 (Pouliot et al., 2015) Zero Anthropogenic Emissions EM Zero Additional Bounding Simulations
Impact of Boundary Conditions on Ozone: Comparison of Observed and Simulated Daily Maximum 8-hr O 3 Daily Observations and Model Simulations BC C-IFS BC H-CMAQ Daily Bias BC AM3 BC GEOSCHEM Observations BC C-IFS BC H-CMAQ BC AM3 BC GEOSCHEM Monthly Mean Bias BC C-IFS BC H-CMAQ BC AM3 BC GEOSCHEM The choice of boundary conditions can have a significant impact on model bias for daily maximum 8-hr O 3, especially for the non-summer months
Use of Sensitivity Simulations and Spectral Analysis for Error Diagnostics Ozone Total Mean Square Error and Error Components for Base and Sensitivity Simulations Modeled Observed Ozone (Top) and Morlet Wavelet of Differences Mean Square Error (ppb^2) Sensitivity simulations help associate processes with types of errors (e.g. zeroing out boundary conditions manifests itself mostly in increased bias error while zeroing out emissions has the largest impact on the variance error) Strong long-term signature of model-observation differences impact of boundary conditions?
AQMEII3 Emission Perturbation Scenarios (Coordinated with HTAP Experiments) Regional models simulated the effects of Global, Home Continent, and Upwind 20% emission reduction scenarios Europe North America GLO EUR NAM GLO NAM EAS NAM EUR EAS Emissions -20% -20% Base -20% -20% Base Boundary Conditions C-IFS GLO C-IFS EUR C-IFS NAM C-IFS GLO C-IFS NAM C-IFS EAS
Multi-Model Analysis of Ozone Changes in Emission Perturbation Scenarios Analysis of simulated ozone and PM 2.5 responses to perturbation scenarios as well as estimated health impacts associated with these responses is ongoing (lead: U. Im)
Impact of Emission Reductions on Wet Deposition Simulated by AQMEII3 Regional Models Changes in oxidized nitrogen wet deposition simulated by CHIMERE as a results of reducing North American anthropogenic emissions by 20% Analysis of deposition fields from perturbation scenarios for other modeling systems is ongoing (lead: M. Garcia Vivanco)
Multi-Model Ensemble Analysis and Its Use for Impact Studies Impact of Model Selection on Ensemble Performance (Ozone from 11 AQMEII1 Models Over Europe) Worst n-model Ensemble Calculated Wheat Production Loss Due to Ozone Pollution: Impact of Ozone Estimate 11-Model Mean Solazzo et al., Model evaluation and ensemble modelling of surface-level ozone in Europe and North America in the context of AQMEII, Atmo. Env., 53, 2012, http://dx.doi.org/10.1016/j.atmosenv.2012.01.003. Best n-model Ensemble Calculated Loss using O 3 (multi-model mean) / Calculated Loss using O 3 (best ensemble) Ongoing work by Solazzo et al.
Summary AQMEII is an active community of regional-scale modeling groups from Europe and North America pursuing research in model evaluation and intercomparisons In planning our next activities under AQMEII, how could these activities be most useful to MMF-GTAD? Modeling time period(s) and domains? (Additional) species to save? Additional data sources for model evaluation? Temporal resolution of outputs daily, quarterly, annual? New methods for quantification and qualification of model error and intermodel variability? Design of model sensitivity simulations?