High-Resolution Air Quality Modeling of New York City to Assess the Effects of Changes in Fuels for Boilers and Power Generation

Transcription

1 High-Resolution Air Quality Modeling of New York City to Assess the Effects of Changes in Fuels for Boilers and Power Generation Final Report June 23, 2014 Submitted to: New York City Department of Health and Mental Hygiene Mayor s Office of Long Term Planning and Sustainability Prepared by ICF International 101 Lucas Valley Road, Suite 260 San Rafael, CA

2 blank page

3 Table of Contents 1 Introduction Background & Objectives Overview of the Modeling Study Application and Evaluation of WRF WRF Application Procedures Description of the WRF Model WRF Modeling Domain and Simulation Period WRF Model Configuration WRF Model Inputs Use of WRF Output Files for Air Quality Modeling and Emissions Processing Model Performance Evaluation Methodology Overview of the WRF Model Performance Evaluation Methodology Qualitative Assessment of Synoptic- and Regional-Scale Meteorological Patterns Qualitative and Quantitative Assessment of Site-Specific Conditions WRF Modeling Results Synoptic- and Regional-Scale Weather Patterns Vertical Profiles of Wind, Temperature and Moisture Temporal Variations in Key Parameters for the High-Resolution Modeling Grids Wind Direction Frequency Distributions for Selected Monitoring Sites Statistical Measures of Model Performance Emission Inventory Preparation Emissions Data and Processing Procedures Emissions Data Emissions Inventory Preparation Methodology Quality Assurance Procedures Emission Summaries Heating Oil Scenarios EGU Fuel Scenarios Application and Evaluation of CMAQ for the Base Year Overview of the CMAQ Modeling System CMAQ Application Procedures Model Performance Evaluation Methodology Air Quality Data Graphical and Statistical Analysis CMAQ Modeling Results Summary of Model Performance for Ozone Summary of Model Performance for PM Summary of Model Performance for PM 10, NO x, SO 2 and CO Modeling Results for the Heating Oil Scenarios Scenario # Scenario # Modeling Results for the EGU Fuel Scenarios Scenario # Scenario # ICF International i New York City Department of Health and Mental Hygiene

4 7 Summary of Results References List of Figures Figure 1-1. CMAQ Modeling Domain for the NYC Modeling Analysis... 2 Figure 2-1. WRF Modeling Domain for High-Resolution Air Quality Modeling for New York City... 6 Figure 2-2. Observed (Left) and Simulated (Right) 500-mb Geopotential Heights (m) and Winds for the Continental U.S. and the NYC 45-km Grid Figure 2-2. Observed (Left) and Simulated (Right) 500-mb Geopotential Heights (m) and Winds for the Continental U.S. and the NYC 45-km Grid (Concluded) Figure 2-3. Observed (Left) and Simulated (Right) Sea Level Pressure (mb) for the Continental U.S. and the NYC 45-km Grid Figure 2-3. Observed (Left) and Simulated (Right) 500-mb Geopotential Heights (m) and Winds for the Continental U.S. and the NYC 45-km Grid (Concluded) Figure 2-4. Observed (Left) and Simulated (Right) Monthly Precipitation Totals (in) for the Continental U.S. and the NYC 45-km Grid Figure 2-5. Simulated and Observed Vertical Profiles of Potential Temperature, Humidity, Wind Speed, and Wind Vectors (Shown in the Order Listed) for 1200 GMT (0700 EST) for Brookhaven, NY (Observed Values are in Red; Simulated Values are in Blue) Figure 2-6. Average Observed (Obs) and Simulated (WRF) Surface Wind Speed (m/s) for the NYC 5-km Grid Figure 2-7. Average Observed (Obs) and Simulated (WRF) Surface Wind Direction (degrees) for the NYC 5-km Grid Figure 2-8. Average Observed (Obs) and Simulated (WRF) Surface Temperature (K) for the NYC 5-km Grid Figure 2-9. Average Observed (Obs) and Simulated (WRF) Surface Water Vapor Mixing Ratio (g/kg) for the NYC 5-km Grid Figure Average Observed (Obs) and Simulated (WRF) Surface Wind Speed (m/s) for the NYC 1-km Grid Figure Average Observed (Obs) and Simulated (WRF) Surface Wind Direction (degrees) for the NYC 1-km Grid Figure Average Observed (Obs) and Simulated (WRF) Surface Temperature (K) for the NYC 1-km Grid Figure Average Observed (Obs) and Simulated (WRF) Surface Water Vapor Mixing Ratio (g/kg) for the NYC 1-km Grid Figure Comparison of Observed and Simulated Wind Direction Frequency for the LaGuardia Airport Monitoring Site (LGA) Figure Comparison of Observed and Simulated Wind Direction Frequency for the JFK Airport Monitoring Site (JFK) Figure Comparison of Observed and Simulated Wind Direction Frequency for the Central Park Tower Monitoring Site (NYC) Figure Comparison of Observed and Simulated Wind Direction Frequency for the Newark International Airport Monitoring Site (EWR) Figure Comparison of Observed and Simulated Wind Direction Frequency for the Teterboro Airport Monitoring Site (TEB) Figure Comparison of Observed and Simulated Wind Direction Frequency for the Westchester Airport Monitoring Site (HPN) Figure Comparison of Observed and Simulated Wind Direction Frequency for the Brookhaven Airport Monitoring Site (HWV) Figure 3-1. Monthly and Diurnal Variations in New York City Boiler Emissions Figure 3-2a. Daily VOC Emissions for 15 July 2008: 2008 Base Year, NYC 15-km Grid Figure 3-2b. Daily NO x Emissions for 15 July 2008: 2008 Base Year, NYC 15-km Grid Figure 3-2c. Daily SO 2 Emissions for 15 July 2008: 2008 Base Year, NYC 15-km Grid Figure 3-2d. Daily PM 2.5 Emissions for 15 July 2008: 2008 Base Year, NYC 15-km Grid Figure 3-3a. Daily VOC Emissions for 15 July 2008: 2008 Base Year, NYC 5-km Grid Figure 3-3b. Daily NO x Emissions for 15 July 2008: 2008 Base Year, NYC 5-km Grid Figure 3-3c. Daily SO 2 Emissions for 15 July 2008: 2008 Base Year, NYC 5-km Grid Figure 3-3d. Daily PM 2.5 Emissions for 15 July 2008: 2008 Base Year, NYC 5-km Grid Figure 3-4a. Daily VOC Emissions for 15 July 2008: 2008 Base Year, NYC 1-km Grid Figure 3-4b. Daily NO x Emissions for 15 July 2008: 2008 Base Year, NYC 1-km Grid Figure 3-4c. Daily SO 2 Emissions for 15 July 2008: 2008 Base Year, NYC 1-km Grid ICF International ii New York City Department of Health and Mental Hygiene

5 Figure 3-4d. Daily PM 2.5 Emissions for 15 July 2008: 2008 Base Year, NYC 1-km Grid Figure 3-5a. Difference in Daily NO x Emissions for 15 July: Scenario #1 Minus Base for the NYC 1-km Grid Figure 3-5b. Difference in Daily SO 2 Emissions for 15 July: Scenario #1 Minus Base for the NYC 1-km Grid Figure 3-5c. Difference in Daily PM 2.5 Emissions for 15 July: Scenario #1 Minus Base for the NYC 1-km Grid Figure 3-6a. Difference in Daily NO x Emissions for 15 July 2008: Scenario #2 Minus Scenario #1 for the NYC 1-km Grid Figure 3-6b. Difference in Daily SO 2 Emissions for 15 July: Scenario #2 Minus Scenario #1 for the NYC 1-km Grid Figure 3-6c. Difference in Daily PM 2.5 Emissions for 15 July: Scenario #2 Minus Scenario #1 for the NYC 1-km Grid Figure 3-7a. Difference in Daily NO x Emissions for 15 July: Scenario #3 Minus Base for the NYC 1-km Grid Figure 3-7b. Difference in Daily SO 2 Emissions for 15 July: Scenario #3 Minus Base for the NYC 1-km Grid Figure 3-7c. Difference in Daily PM 2.5 Emissions for 15 July: Scenario #3 Minus Base for the NYC 1-km Grid Figure 3-8a. Difference in Daily NO x Emissions for 15 July 2008: Scenario #4 Minus Scenario #1 for the NYC 1-km Grid Figure 3-8b. Difference in Daily SO 2 Emissions for 15 July: Scenario #4 Minus Scenario #1 for the NYC 1-km Grid Figure 3-8c. Difference in Daily PM 2.5 Emissions for 15 July: Scenario #4 Minus Scenario #1 for the NYC 1-km Grid Figure 4-1. Simulated Daily Maximum 8-Hour Ozone Concentration (ppb) for Selected Days for the CMAQ 15-km Grid Figure 4-2. Comparison of Simulated and Observed Daily Maximum 8-Hour Average Ozone Concentration (ppb) for the 15-km Grid (April through October) Figure 4-3. Simulated Daily Maximum 8-Hour Ozone Concentration (ppb) for Selected Days for the CMAQ 5-km Grid Figure 4-4. Normalized Bias (%) and Normalized Mean Error (%) Based on Daily Maximum 8-Hour Average Simulated and Observed Ozone Concentrations for April through October for the CMAQ 5-km Grid Figure 4-5. Comparison of Simulated and Observed Daily Maximum 8-Hour Average Ozone Concentration (ppb) for the 5-km Grid (April through October) Figure 4-6. Simulated Daily Maximum 8-Hour Ozone Concentration (ppb) for Selected Days for the CMAQ 1-km Grid Figure 4-8. Comparison of Simulated and Observed Daily Maximum 8-Hour Average Ozone Concentration (ppb) for the 1-km Grid (April through October) Figure 4-9. Simulated Monthly Average PM 2.5 Concentration (µg/m 3 ) for the CMAQ 15-km Grid Figure Simulated Annual Average PM 2.5 Concentration (µg/m 3 ) for the CMAQ 15-km Grid Figure Comparison of Simulated and Observed 24-Hour Average PM 2.5 Concentration (µg/m 3 ) for the 15-km Grid (All Months) Figure Simulated Monthly Average PM 2.5 Concentration (µg/m 3 ) for the CMAQ 5-km Grid Figure Simulated Annual Average PM 2.5 Concentration (µg/m 3 ) for the CMAQ 5-km Grid Figure Fractional Bias (%) and Fractional Error (%) Based on 24-Hour Average Simulated and Observed PM 2.5 Concentrations for CMAQ 5-km Grid (All Months) Figure Comparison of Simulated and Observed 24-Hour Average PM 2.5 Concentration (µg/m 3 ) for the 5-km Grid (All Months) Figure Simulated Monthly Average PM 2.5 Concentration (µg/m 3 ) for the CMAQ 1-km Grid Figure Simulated Annual Average PM 2.5 Concentration (µg/m 3 ) for the CMAQ 1-km Grid Figure Comparison of Simulated and Observed 24-Hour Average PM 2.5 Concentration (µg/m 3 ) for the 1-km Grid (All Months) Figure Comparison of Simulated and Observed 24-Hour Average PM 10 Concentration (µg/m 3 ) for the 5-km Grid (All Months) Figure Comparison of Simulated and Observed Hourly Average NO 2, SO 2, and CO Concentrations (ppb) for the 5- km Grid (All Months) Figure Comparison of Simulated and Observed 24-Hour Average PM 10 Concentration (µg/m 3 ) for the 1-km Grid (All Months) Figure Comparison of Simulated and Observed Hourly Average NO 2, SO 2, and CO Concentrations (ppb) for the 1- km Grid (All Months) Figure 5-1. Difference in Simulated Daily Maximum 8-Hour Ozone Concentration: Scenario #1 Minus Base Figure 5-2. Difference in Simulated Monthly Average PM 2.5 Concentration: Scenario #1 Minus Base Figure 5-3. Difference in Simulated Annual Average PM 2.5 Concentration: Scenario #1 Minus Base Figure 5-4. Difference in Simulated Daily Maximum 1-Hour NO 2 Concentration: Scenario #1 Minus Base Figure 5-5. Difference in Simulated Daily Maximum 1-Hour SO 2 Concentration: Scenario #1 Minus Base Figure 5-6. Difference in Simulated Daily Maximum 8-Hour Ozone Concentration: Scenario #2 Minus Base Figure 5-7. Difference in Simulated Monthly Average PM 2.5 Concentration: Scenario #2 Minus Base Figure 5-8. Difference in Simulated Annual Average PM 2.5 Concentration: Scenario #2 Minus Base Figure 5-9. Difference in Simulated Daily Maximum 1-Hour NO 2 Concentration: Scenario #2 Minus Base Figure Difference in Simulated Daily Maximum 1-Hour SO 2 Concentration: Scenario #2 Minus Base Figure 6-1. Difference in Simulated Daily Maximum 8-Hour Ozone Concentration: Scenario #3 Minus EGU Base ICF International iii New York City Department of Health and Mental Hygiene

6 Figure 6-2. Difference in Simulated Monthly Average PM 2.5 Concentration: Scenario #3 Minus EGU Base Figure 6-3. Difference in Simulated Annual Average PM 2.5 Concentration: Scenario #3 Minus EGU Base Figure 6-4. Difference in Simulated Daily Maximum 1-Hour NO 2 Concentration: Scenario #3 Minus EGU Base Figure 6-5. Difference in Simulated Daily Maximum 1-Hour SO 2 Concentration: Scenario #3 Minus EGU Base Figure 6-6. Difference in Simulated Daily Maximum 8-Hour Ozone Concentration: Scenario #4 Minus EGU Base Figure 6-7. Difference in Simulated Monthly Average PM 2.5 Concentration: Scenario #4 Minus EGU Base Figure 6-8. Difference in Simulated Annual Average PM 2.5 Concentration: Scenario #4 Minus EGU Base Figure 6-9. Difference in Simulated Daily Maximum 1-Hour NO 2 Concentration: Scenario #4 Minus EGU Base Figure Difference in Simulated Daily Maximum 1-Hour SO 2 Concentration: Scenario #4 Minus EGU Base List of Tables Table 2-1. Vertical Layer Structure for the WRF Modeling Domain for High-Resolution Air Quality Modeling for New York City... 8 Table 2-2. Summary of Parameter Specification for WRF Modeling of New York City... 9 Table 2-3. Definition and Description of Measures/Metrics for WRF Model Performance Evaluation for High Resolution Air Quality Modeling of New York City Table 2-4. Statistical Benchmarks for Evaluating Meteorological Model Performance Table 2-5. Statistical Summary of WRF Model Performance for the NYC 15-km Modeling Grid Table 2-6. Statistical Summary of WRF Model Performance for the NYC 5-km Modeling Grid Table 2-7. Statistical Summary of WRF Model Performance for the NYC 1-km Modeling Grid Table 3-1. Emissions Totals (tons/year) by Source Sector for the 2008 Base Year for the 15-km Grid Table 3-2. Emissions Totals (tons/year) by Sector for the 2008 Base Year for the 5-km Grid Table 3-3. Emissions Totals (tons/year) by Sector for the 2008 Base Year for the 1-km Grid Table 3-5. New York City Boiler Emission Summary for the 2008 Base Year and Heating Oil Scenarios Table 3-6a. EGU Emissions Summary for the EGU Base and EGU Fuel Scenarios: NYC 15-km Grid, Total Table 3-6b. EGU Emissions Summary for the EGU Base and EGU Fuel Scenarios: NYC 15-km Grid, Inside of New York City Table 3-6c. EGU Emissions Summary for the EGU Base and EGU Fuel Scenarios: NYC 15-km Grid, Outside of New York City Table 3-7a. EGU Emissions Summary for the EGU Base and EGU Fuel Scenarios: NYC 5-km Grid, Total Table 3-7b. EGU Emissions Summary for the EGU Base and EGU Fuel Scenarios: NYC 5-km Grid, Inside of New York City Table 3-7c. EGU Emissions Summary for the EGU Base and EGU Fuel Scenarios: NYC 5-km Grid, Outside of New York City Table 3-8a. EGU Emissions Summary for the EGU Base and EGU Fuel Scenarios: NYC 1-km Grid, Total Table 3-8b. EGU Emissions Summary for the EGU Base and EGU Fuel Scenarios: NYC 1-km Grid, Inside of New York City Table 3-8c. EGU Emissions Summary for the EGU Base and EGU Fuel Scenarios: NYC 1-km Grid, Outside of New York City Table 3-9. Facility Level EGU Emissions Comparison for the NYC 5 Boroughs: Scenario 4 vs EGU Base Case Table 4-1. Statistical Measures Used for the CMAQ Model Performance Evaluation for High Resolution Air Quality Modeling of New York City Table 4-2. Summary Model Performance Statistics for Ozone for the 15-km Modeling Grid Table 4-3. Summary Model Performance Statistics for Ozone for the 5-km Modeling Grid Table 4-4. Summary Model Performance Statistics for Ozone for the 1-km Modeling Grid Table 4-5. Summary Model Performance Statistics for PM 2.5 for the 15-km Modeling Grid Table 4-6. Summary Model Performance Statistics for PM 2.5 for the 5-km Modeling Grid Table 4-7. Summary Model Performance Statistics for PM 2.5 for the 1-km Modeling Grid Table 4-8. Summary Model Performance Statistics for PM 10, NO x, SO 2 and CO for the 5-km Modeling Grid Table 4-9. Summary Model Performance Statistics for PM 10, NO x, SO 2 and CO for the 1-km Modeling Grid Table 7-1. Summary of CMAQ Modeling Results for the New York City Portion of the 1-km Grid (the 5-Borough Area) Using Key NAAQS-Related Metrics Table 7-2. Summary of CMAQ Modeling Results for the New York City Portion of the 1-km Grid (the 5-Borough Area) Using Key NAAQS-Related Metrics ICF International iv New York City Department of Health and Mental Hygiene

7 Executive Summary This report summarizes an air quality modeling study for New York City designed to quantify the air quality and public health benefits attributable to recent changes in fuel use in the heating and power sectors. The key questions addressed by this analysis are: What is the impact of changes in heating oil use that have occurred since 2010 on air pollutant concentrations in NYC, including at the neighborhood level? What is the impact of changes in fuel use in the electric power generation sector since 2005 on air pollutant concentrations in NYC, including at the neighborhood level? The air quality assessment focused on ozone, fine particulate matter (PM 2.5 ), nitrogen dioxide (NO 2 ), and sulfur dioxide (SO 2 ). The modeling tools that were used for this assessment include the Weather Research and Forecasting (WRF) model for the preparation of meteorological inputs and the Community Multiscale Air Quality (CMAQ) model for quantifying the air quality changes for the different emission scenarios. The WRF and CMAQ modeling domains consist of multiple nested grids, with varying horizontal resolution and were designed to accommodate both regional and subregional influences as well as to provide a detailed representation of the emissions, meteorological fields, and pollutant concentration patterns over New York City and its neighborhoods. The domain includes a 15-km resolution outer grid encompassing the northeastern U.S.; a 5-km resolution intermediate grid over the greater New York area; and a 1-km resolution inner grid over the City. The models were applied for an annual simulation period, using meteorological inputs for a base year of The CMAQ model was applied for a base year scenario and several alternative emission scenarios in which the emissions were modified to reflect the changes in fuels. An evaluation of model performance was conducted for the base year scenario. Good model performance was achieved for both WRF and the CMAQ models. Surface wind speeds and wind directions are especially well represented by WRF at the 1-km scale. For the CMAQ model, the best performance is achieved for ozone and PM 2.5. Statistical measures of model performance for these pollutants are within standard model performance goals and indicative of good model performance. Statistical measures for the remaining pollutants are also within the model performance goals. On average, NO 2 concentrations are underestimated and SO 2 concentrations are overestimated by the CMAQ model. Two scenarios examined the effects of local changes in heating fuels. The heating oil control scenarios are as follows: Scenario #1: Partial implementation of the rule on heating oil, the emissions changes reflect the reduction in emissions achieved by accounting for partial phase out of residual heating oil and implementation of New York State 15 parts per million (ppm) No.2 heating oil requirements. Scenario #2: Full implementation of City and State heating oil rules (complete phase out of Number (No.) 4 and No. 6 heating oil and reduction of No. 2 heating oil sulfur content to 15ppm). ICF International v New York City Department of Health and Mental Hygiene

8 For Scenario #1, the emissions of NO x, SO 2 and PM 2.5 emissions associated with the New York City boilers are reduced by 8, 69 and 67 percent from the 2008 levels, respectively. The reduction in NO x and SO 2 emissions results in simulated decreases in PM 2.5 concentration over New York City. The greatest decreases are simulated for the winter months when heating fuel consumption is greatest. Annual average PM 2.5 concentrations within the 1-km grid are reduced by as much as 4.3 µg/m 3. Decreases occur over Manhattan, Brooklyn, Queens, and the Bronx, with the greatest decreases over Manhattan. Simulated daily maximum 1-hour NO 2 and SO 2 concentrations for this scenario are lower than the baseyear concentrations, both over New York City and beyond. Large decreases in SO 2 concentration are consistent with a 69 percent reduction in overall SO 2 emissions. The reduction in NO x emissions leads to some small simulated increases in ozone concentration (less than 1 ppb), especially over New York City. The response of the CMAQ model to the changes in emissions is influenced by the complex photochemistry represented by the model. Under certain conditions (usually for urban areas characterized by a low VOC-to-NO x ratio), decreases in NO x emissions can lead to increases in ozone. For Scenario #2, NO x, SO 2 and PM 2.5 emissions are reduced by 26, 98 and 89 percent, respectively. Full implementation of the heating-oil rule results in greater simulated decreases in PM 2.5 concentration over New York City, compared to Scenario #1. Annual average PM 2.5 concentrations within the 1-km grid are reduced by as much as 5.5 µg/m 3. The simulation results are also characterized by large decreases in simulated daily maximum 1-hour NO 2 and SO 2 concentrations, substantially greater decreases compared to those for Scenario #1. Consistent with Scenario #1, Scenario #2 showed increases in ozone concentration due to decreases in NO x emissions. Two additional scenarios examined the effects of changes in Electric Generating Unit (EGU) fuel between 2005 and the present. The year 2005 was selected because significant reductions in EGU emissions occurred between 2005 and The two EGU scenarios are as follows: Scenario #3: Adjustment of EGU emissions to reflect changes in fuel use at Title V EGUs outside of the five boroughs of New York City, using emission estimates for these sources from continuous emissions monitoring (CEM) data for 2012 to represent present day. Scenario #4: Adjustment of EGU emissions to reflect changes in fuel use at EGUs located within the five boroughs, using 2012 CEM emissions data for these sources to represent present day. For Scenario 3, for EGUs located outside of the New York City boroughs, the emissions of all criteria pollutants except CO are lower for all three grids, compared to the corresponding base simulation. The SO 2 emissions are 70, 90 and 99 percent lower for the 15-, 5- and 1-km grids, respectively. The NO x emissions are 57, 63 and 84 percent lower. The reduction in NO x and SO 2 emissions outside of the five-borough area leads to simulated decreases in PM 2.5 concentration, including over New York City. Annual average PM 2.5 concentrations within the 1-km grid are reduced by as much as 2.6 µg/m 3, with fairly uniform decreases on the order of 1 to 2 µg/m 3 over all five boroughs. Daily maximum 1-hour NO 2 concentrations are characterized by a mix of ICF International vi New York City Department of Health and Mental Hygiene

9 decreases and increases while SO 2 concentrations are decreased. Reductions in SO 2 emissions outside of the five-borough area influence parts of New York City, depending on wind direction. The reduction in NO x emissions outside of New York City but within the 1-km grid leads to both increases and decreases in ozone concentration throughout the 1-km grid, including over the City. For Scenario 4, the changes occur in New York City and the EGU emissions for the five boroughs are lower by 64 and 97 percent for NO x and SO 2, respectively. The reduction in NO x and SO 2 emissions in the five-borough area leads to small simulated decreases in PM 2.5 concentration over New York City. Annual average PM 2.5 concentrations within the 1-km grid are reduced by as much as 0.4 µg/m 3. The changes in simulated daily maximum 1-hour NO 2 concentrations for Scenario #4 are a mix of increases and decreases, both over New York City and beyond and the CMAQ results show primarily decreases in simulated daily maximum 1-hour SO 2 concentration within the 1-km grid. Reductions in SO 2 within the five-borough area influence areas beyond New York City, depending on wind direction. Again, the reduction in EGU NO x emissions leads to some increases in ozone concentration over New York City. In summary, the modeling results indicate that changes in heating oil use that have occurred since 2010 have resulted in improved air quality in New York City, with large reductions in PM 2.5 and SO 2 concentrations. Full implementation of the heating oil rule results in a 20 percent reduction in simulated annual average PM 2.5 concentration and a 65 percent reduction in simulated daily maximum 1-hour SO 2 concentration. These are key metrics related to the National Ambient Air Quality Standards (NAAQS) for PM 2.5 and SO 2. The modeling results also indicate that changes in fuel use in the electric power generation sector since 2005 have also resulted in reductions in ozone, PM 2.5 and SO 2 concentrations (on the order of 2 to 4 percent relative to key NAAQS metrics). ICF International vii New York City Department of Health and Mental Hygiene

10 1 Introduction This report summarizes an air quality modeling study for New York City designed to quantify the air quality and public health benefits attributable to recent changes in fuel use in the heating and power sectors. This study was sponsored by the New York City Department of Health and Mental Hygiene (DOHMH) and the Mayor s Office of Long Term Planning and Sustainability (OLTPS). 1.1 Background & Objectives A key objective of the air quality modeling analysis was to provide data for quantitatively assessing the air quality related benefits from local changes in heating oil (specifically within New York City) and regional changes in fuel for electric generation units (EGUs). The air quality assessment focused on the following criteria pollutants: ozone, fine particulate matter (PM 2.5 ), nitrogen dioxide (NO 2 ), and sulfur dioxide (SO 2 ). The analysis was specifically designed to examine the benefits at the local scale for New York City (NYC) neighborhoods. To achieve this, the modeling was conducted using a high-resolution modeling grid with 1-kilometer (1-km) spacing. NYC DOHMH maintains health data for a wide range of morbidity and mortality endpoints for the 42 neighborhoods within New York City. Morbidity rates vary widely across the City s 42 neighborhoods, and, thus, the modeling study was conducted using a spatial resolution sufficient for this level of analysis. The key questions addressed by this analysis are: What is the impact of changes in heating oil use that have occurred since 2010 on air pollutant concentrations in NYC, including at the neighborhood level? What is the impact of changes in fuel use in the electric power generation sector since 2005 on air pollutant concentrations in NYC, including at the neighborhood level? The air quality modeling examined a variety of scenarios reflecting changes in fuel-related emissions and the modeling results provide the basis for the calculation of air quality related health benefits associated with the changes in fuel-related emissions for NYC. 1.2 Overview of the Modeling Study The air quality modeling assessment was designed to examine and quantify the effects of changes in heating oil and fuel use in the power sector on air quality in NYC. Key components of this assessment included the preparation of meteorological inputs for a base year of 2008, preparation of emission inputs for the base year and the various alternative emission scenarios, application of the Community Multiscale Air Quality (CMAQ) model, and assessment of the air quality impacts/benefits. The primary tools that were used for this assessment include: Weather Research and Forecasting (WRF) model (version 3.4) for the preparation of meteorological inputs; ICF International 1 New York City Department of Health and Mental Hygiene

11 Introduction Sparse-Matrix Operator Kernel Emissions (SMOKE) processing tool (version 3.1) for the preparation of model-ready emissions; Community Multiscale Air Quality (CMAQ) model (version 5.0) for quantifying the air quality changes for the different emission scenarios; and Atmospheric Model Evaluation Tool (AMET) to evaluate the CMAQ modeling results. These tools are widely used for conducting air quality analysis. The CMAQ modeling domain for this analysis is presented in Figure 1-1 and consists of multiple nested grids, with varying horizontal resolution. The modeling domain was designed to accommodate both regional and subregional influences as well as to provide a detailed representation of the emissions, meteorological fields, and pollutant concentration patterns over New York City and its neighborhoods. The domain includes a 15- km resolution outer grid encompassing the northeastern U.S.; a 5-km resolution intermediate grid over the greater New York area; and a 1-km resolution inner grid over the City. Figure 1-1. CMAQ Modeling Domain for the NYC Modeling Analysis 15-, 5- and 1-km Grids ICF International 2 New York City Department of Health and Mental Hygiene

12 Introduction 5- and 1-km Grids The regional extent of the modeling domain is intended to provide realistic boundary conditions for the primary area of interest and thus avoid some of the uncertainty introduced in the modeling results through the incomplete and sometimes arbitrary specification of boundary conditions. The outer (15- and 5-km) grids also allow the direct simulation of the effects of changes in fuel use at Electric Generating Units (EGUs) outside of NYC. The use of 1-km grid resolution over the primary area of interest is consistent with the neighborhood-scale analysis of air quality in NYC. The CMAQ modeling domain also includes 34 vertical layers. Boundary conditions for the 15-km domain were derived from a national-scale simulation for 2008 that was run using a 36-km resolution continental U.S. (CONUS) modeling domain. The CMAQ model was applied for an annual simulation period, using meteorological inputs for a base year of This year is consistent with the latest available National Emission Inventory (NEI) and thus accommodated the use of up-to-date regional-scale emissions data. In addition, 2008 precedes required changes in heating oil use (the required phase out of residual oil in NYC and institution of 15 ppm sulfur distillate oil (NY State) began in 2010) and therefore provides an appropriate baseline for assessing the changes in air quality due to changes in heating fuel. Boiler emissions for NYC were provided by DOHMH. The CMAQ model was applied for a base year scenario and several alternative emission scenarios in which the emissions were modified to reflect the changes in fuels. An evaluation of model performance was conducted for the base year scenario. ICF International 3 New York City Department of Health and Mental Hygiene

13 Introduction The first set of scenarios examined the effects of local changes in heating fuels. The heating oil control scenarios are as follows: Partial implementation of the rule on heating oil, the emissions changes reflect the reduction in emissions achieved by the end of the winter heating season (April 2013). Full implementation of the rule on heating oil (complete phase out of Number (No.) 4 and No. 6 heating oil). Both heating oil scenarios also included implementation of a 15 ppm sulfur limit to No. 2 heating oil. The second set of scenarios examined the effects of changes in EGU fuel between 2005 and the present. The year 2005 was selected because significant reductions in EGU emissions occurred between 2005 and 2008, and NYC DOHMH desired to assess the impacts of EGU reductions, starting with 2005 emission levels. As a first step in the EGU fuel assessment, CMAQ was run using a hybrid emission inventory consisting of the 2008 base year emissions for all sources except EGU sources, and 2005 emissions for EGU sources. This alternative baseline scenario (EGU Base) provided a basis for assessing the effects of EGU emission changes for the following two EGU scenarios: Adjustment of EGU emissions to reflect changes in fuel use at Title V EGUs outside of the five boroughs of New York City, using emission estimates for these sources from continuous emissions monitoring (CEM) data for 2012 to represent present day. Adjustment of EGU emissions to reflect changes in fuel use at EGUs located within the five boroughs, using 2012 CEM emissions data for these sources to represent present day. The CMAQ modeling analysis is summarized in the remainder of this report. The preparation of meteorological inputs is summarized in Section 2. The preparation of emission inputs is summarized in Section 3. Base year modeling and model performance evaluation is presented in Section 4. The modeling results for the heating oil scenarios are presented in Section 5. The modeling results for the EGU fuel scenarios are presented in Section 6. The CMAQ-derived, gridded, high-resolution pollutant concentration fields for all scenarios have been provided to NYC DOHMH for use in in conducting healtheffects analyses. ICF International 4 New York City Department of Health and Mental Hygiene

14 2 Application and Evaluation of WRF The CMAQ model requires hourly, gridded input fields of several meteorological parameters including wind, temperature, mixing ratio, pressure, solar radiation, fractional cloud cover, cloud depth, and precipitation. The prescribed meteorological conditions influence the transport, vertical mixing, and resulting distribution of the simulated pollutant concentrations. Certain of the meteorological parameters, such as water vapor mixing ratio, temperature, and solar radiation can also influence the simulated chemical reaction rates. Rainfall and near-surface meteorological characteristics govern the wet and dry deposition, respectively, of the simulated atmospheric constituents. Meteorological input fields for the CMAQ model for the NYC air quality assessment were prepared using the WRF meteorological model. Specifically, version 3.4 of the Advanced Research WRF (ARW) model was used. WRF is a state-of-the-science atmospheric modeling system designed for use in simulating meteorological fields for a broad range of scales and applications. The ARW version of the WRF model contains data assimilation capabilities that are integral to the use of the model for air quality modeling of historical simulation periods. The WRF/ARW model is currently maintained by the Mesoscale and Microscale Meteorology Division of the National Center for Atmospheric Research (NCAR), located in Boulder, Colorado. The WRF model was applied for the calendar year The modeling domain includes four nested grids with approximately 45-, 15-, 5-, and 1-kilometer (km) horizontal resolution. The 15-, 5- and 1-km grids encompass the corresponding CMAQ grids with buffer regions around the CMAQ grids to minimize WRF boundary influences. The 1-km grid is centered over New York City. The simulation parameters and options are detailed in Section 2.1 (Table 2-2). The WRF simulation results were evaluated using graphical and statistical analysis. The output from WRF was then processed for input to the CMAQ model using the Meteorology-Chemistry Interface Processor (MCIP). 2.1 WRF Application Procedures Description of the WRF Model WRF is a state-of-the-science atmospheric modeling system designed for use in simulating meteorological fields for a broad range of scales and applications. Version 3.4 of the Advanced Research WRF (ARW) model, as used for this analysis, includes numerous features to support both idealized and real-data simulations. Key features for this application include: a terrain following vertical coordinate system; fully compressible non-hydrostatic equations (for the simulation of the effects of terrain on airflow); two-way grid nesting; variable vertical grid spacing; full physics options for the surface layer, planetary boundary layer, atmospheric and surface radiation, microphysics and cumulus convection; and data assimilation (both analysis and observation nudging). The data assimilation capabilities, through which observed data are used to guide the simulation, are integral to the use of the model to represent historical simulation periods. ICF International 5 New York City Department of Health and Mental Hygiene

15 Application and Evaluation of WRF WRF Modeling Domain and Simulation Period The WRF modeling domain is designed to accommodate regional, sub-regional, and local meteorological influences and to provide a detailed representation of the meteorology over the area of interest. The modeling domain includes four nested grids with approximately 45-, 15-, 5- and 1-kilometer (km) horizontal resolution. The 45-km resolution grid covers the continental U.S. The 15-, 5- and 1-km grids are illustrated in Figure 2-1. The 1-km grid encompasses the five boroughs of New York City. The left and bottom axes indicate the number of grid cells for the outermost grid shown. The right and top axes show the distance from the center of the same grid based on the Lambert Conformal Conic (LCC) map projection. Figure 2-1. WRF Modeling Domain for High-Resolution Air Quality Modeling for New York City 15-, 5- and 1-km Grids Note that the WRF grids encompass and extend beyond the corresponding CMAQ grids with the same horizontal resolution. This is to ensure that regional-scale weather systems influencing the area of interest are represented along the boundaries of the domain. The modeling grids are based on a LCC map projection. In the vertical dimension, the modeling domain includes 34 layers. The top of the model domain is 100 millibars. The thickness of the layers increases with height above ground. The thinner layers near the surface are designed to provide enhanced resolution of the meteorological parameters and dispersion characteristics within the lowest part of the atmosphere (where they tend to be most variable) and to ICF International 6 New York City Department of Health and Mental Hygiene

16 Application and Evaluation of WRF delineate the depth of the planetary boundary layer (PBL). Representation of the near surface meteorological characteristics and PBL depth is critical to accurate simulation of pollutant dispersion and transport. The vertical layers are presented in Table 2-1. For each layer, the table lists the sigma value (this corresponds to the internal sigma-based, or terrain-following, coordinate system), the approximate pressure at the top of the layer, the estimated height of the top of the layer (based on standard atmospheric conditions), and the estimated depth of the layer. Units are millibars (mb) for pressure and meters (m) for layer height and depth. ICF International 7 New York City Department of Health and Mental Hygiene

17 Application and Evaluation of WRF Table 2-1. Vertical Layer Structure for the WRF Modeling Domain for High-Resolution Air Quality Modeling for New York City Layer Sigma Pressure (mb) Height (m) Depth (m) ,662 1, ,822 1, ,356 1, ,127 1, , , , , , , , , , , , , , , , , , Ground The WRF model was applied for the calendar year In applying WRF, the annual simulation period was divided into multiple sub-periods. Each sub-period included a 12-hour initialization period followed by a 5-day simulation period (a total of 5.5 days). Each successive initialization period overlapped the ICF International 8 New York City Department of Health and Mental Hygiene

18 Application and Evaluation of WRF previous simulation period by 12 hours. This resulted in a total of approximately 75 multi-day simulations to complete the application for the entire year WRF Model Configuration The WRF configuration is summarized in Table 2-2. General Information Table 2-2. Summary of Parameter Specification for WRF Modeling of New York City Parameter Specification WRF model version 3.4 Modeling Domain Number of grids 4 Grid resolution 45, 15, 5 & 1 km Number of grid cells (45-km grid) 131 x 103 Number of grid cells (15-km grid) 135 x 126 Number of grid cells (5-km grid) 120 x 135 Number of grid cells (1-km grid) 115 x 115 Nesting approach Two-way nesting Number of vertical layers 34 Meteorological Datasets Initial and boundary conditions Sea surface temperature Observed data for surface analysis nudging Observed data for observational ( obs ) nudging Data Assimilation NCEP/NARR 32-km ds608.0 NARR NCEP/ADP ds461.0 NCAR ds472.0 Analysis nudging aloft (45-km grid) u, v, T, q * Analysis nudging aloft (15-km grid) Analysis nudging aloft (5-km grid) Analysis nudging aloft (1-km grid) Analysis nudging coefficient (aloft; all nudged parameters) Analysis nudging (surface) (45-km grid) Analysis) nudging (surface) (15-km grid) Analysis nudging (surface) (5-km grid) Analysis nudging (surface) (1-km grid) Analysis nudging coefficient (surface; all nudged parameters) Obs nudging (45-km grid) Obs nudging (15-km grid) Obs nudging (5-km grid) u, v, T, q None None 3x10-4 u, v u, v None None 3x10-4 None None Obs nudging (1-km grid) u, v Obs nudging coefficient (all nudged parameters) 6x10-4 u, v ICF International 9 New York City Department of Health and Mental Hygiene

19 Application and Evaluation of WRF Physics Parameter Specification Microphysics WRF single-moment 3-class scheme Long wave radiation Rapid Radiative Transfer Model (RRTM) Short wave radiation Dudhia Surface layer physics Eta similarity Land-surface model Noah LSM PBL Mellor-Yamada-Janic scheme Cumulus parameterization (45 & 15-km grids only) Grell-Devenyi ensemble scheme Surface fluxes Yes Snow cover effects Yes Cloud cover effects Yes Number of soil layers 4 Urban physics Yes Dynamics Vertical velocity damping No Diffusion Simple diffusion Eddy coefficient 2-D deformation Sixth-order numerical diffusion No Base sea-level temperature (K) 290 Upper-level damping No Horizontal diffusion coefficient (m 2 /s) 0 Vertical diffusion coefficient (m 2 /s) 0 Non-hydrostatic Yes Moist advection Positive-definite advection Scalar advection Positive-definite advection Boundary Condition Controls Number of rows for boundary value nudging 5 Number of points in specified zone 1 Number of points in relaxation zone 4 * u = east-west wind component, v = north-south wind component, T = temperature, q = water vapor mixing ratio (moisture) For this application, surface temperature and moisture were characterized using the Noah Land Surface Model (LSM) which has been recently updated (in WRF ARW) to better represent processes over ice and snow covered areas. For the coarser grids, the Grell-Devenyi ensemble cumulus parameterization scheme was used to parameterize the effects of convection on the simulated environment. This feature was not employed for the 5- and 1-km grids where the model can explicitly resolve convection. The WRF-ARW model supports four-dimensional data assimilation (FDDA), a procedure by which observed data are incorporated into the simulation. Analysis nudging, in which the simulation variables are relaxed or nudged toward an objective analysis that incorporates the observed data, was used for all parameters ICF International 10 New York City Department of Health and Mental Hygiene

20 Application and Evaluation of WRF (wind, temperature, moisture) for the outer modeling grids. Analysis nudging of temperature and moisture was applied only for layers that are above the planetary boundary layer. Observational ( obs ) nudging, in which the simulation parameters are nudged directly toward selected observations, was used for surface winds within the high-resolution 5- and 1-km grids, using NCAR ds472 data WRF Model Inputs Inputs required for application of the WRF model include topographic, land-use, and vegetation information for each grid cell (and modeling grid), initial and boundary conditions, and meteorological analysis fields and observed data (for use in the data assimilation schemes). For this application, high-resolution data for preparation of the terrain, land-use, and vegetation input files were obtained from NCAR. The WRF input files were prepared using the preprocessor programs that are part of the WRF modeling system (NCAR, 2010). Meteorological data were obtained from the National Center for Environmental Prediction (NCEP) and NCAR and specifically from the Research Data Archive (RDA), which is maintained, by the Computational and Information Systems Laboratory (CISL) at NCAR: The NCEP/North American Regional Reanalysis (NARR) 32-km model analyses (ds608.0) were used to specify the initial and boundary conditions as well as sea surface temperatures for WRF. The analysis fields were also used for the analysis nudging (above the PBL) for the 45- and 15-km grids. In addition, the analysis fields were combined with the NCEP Automated Data Processing (ADP) surface observational weather data (ds461.0) (in little_r format) and used for the surface analysis nudging (also for the 45- and 15-km grids). The NCAR ds472.0 datasets contain surface and upper-air wind, temperature, moisture, and pressure data for all routine monitoring sites within the domain and these data were used for the observational ( obs ) nudging. These data are from a variety of monitoring sites including National Weather Service (NWS) Automated Surface Observing System (ASOS) monitoring sites and routine aviation weather reporting stations. The ds472.0 dataset includes both surface and upper-air data. Approximately 20 surface and three upper-air meteorological monitoring sites are located within the WRF 5-km modeling grid and approximately six surface sites and no upper-air sites are located within the WRF 1-km grid Use of WRF Output Files for Air Quality Modeling and Emissions Processing The WRF output files were postprocessed to correspond to the CMAQ modeling domain and the units and formats required by the modeling system using the MCIP postprocessing software. MCIP also outputs directly the meteorological fields needed for emissions processing. These include: Temperature, surface pressure, radiation/cloud cover, rainfall, soil temperature, soil moisture and soil type for the calculation of the biogenic emissions; and ICF International 11 New York City Department of Health and Mental Hygiene

21 Application and Evaluation of WRF Temperature and relative humidity for the calculation of motor vehicle emissions. 2.2 Model Performance Evaluation Methodology The WRF model performance evaluation methodology was designed to examine whether the WRF model (configured and applied as discussed in Section 2.1) was able to reproduce the observed meteorological conditions of the (historical) simulation period, especially those features that are important in air quality modeling (and influence the transport, chemical transformation and deposition processes). Key components of the evaluation include: 1) the qualitative assessment of the ability of the WRF model to represent the synoptic- and regional-scale spatial patterns and temporal variations of wind, temperature, water vapor mixing ratio, and precipitation of the simulation period and 2) the qualitative and quantitative assessment of the ability of the WRF model to represent site-specific conditions including vertical profiles and diurnal profiles of wind, temperature, and mixing ratio for sites located within the 5- and 1-km modeling grids. A variety of graphical and statistical methods were used to examine the WRF results, consistent with EPA guidance on the preparation of meteorological inputs for air quality modeling (EPA, 2007) Overview of the WRF Model Performance Evaluation Methodology For the outermost (45-km) grid, examination of the WRF output focused on representation of the regional-scale meteorological features and airflow patterns. Plots of selected meteorological fields were prepared using the NCAR Command Language (NCL) (NCAR, 2013) and compared with weather maps. For the intermediate (15-km) grid, the evaluation also included an examination of the regional-scale patterns and a comparison with weather maps. In addition, statistical measures comparing the simulated values of wind, temperature and mixing ratio (moisture) with observed values were calculated using the METSTAT program (Environ, 2012). A more detailed evaluation of the results was performed for the 5- and 1-km grids, emphasizing representation of the observed data, terrain-related airflow features, vertical temperature structure of the lower atmosphere, PBL heights, and vertical mixing parameters. The modeling results were compared with observed data. Statistical measures comparing the simulated values of wind, temperature and water vapor mixing ratio with observed values were calculated. For all grids, the ability of the WRF model to reproduce observed precipitation patterns was qualitatively assessed by comparing the simulated and observed precipitation patterns (based on NWS data). A detailed analysis of the timing and amount of the precipitation was not performed Qualitative Assessment of Synoptic- and Regional-Scale Meteorological Patterns As a starting point in the evaluation, spatial plots of selected meteorological parameters were prepared for each grid for a date near the 15 th of each month (the exact date varied according to run segmentation and was chosen so that each plot was of the fifth day of a 5-day simulation period). The ICF International 12 New York City Department of Health and Mental Hygiene

22 Application and Evaluation of WRF plots included surface temperature, surface pressure, surface specific humidity, surface wind speed and wind direction, 700 mb temperature, 700 mb wind speed and wind direction, 500 mb wind speed and wind direction, monthly total precipitation amounts, and snow cover. These were inspected in several ways including: comparison of the WRF results with NWS analysis products, comparison of the results for all grids, and inspection of the WRF results relative to the terrain and land-use features Qualitative and Quantitative Assessment of Site-Specific Conditions Methods and Tools Statistical measures were used to quantify model performance for the 15-, 5-, and 1-km grids. Statistical measures were calculated using the METSTAT program. METSTAT was applied for each grid, for each month, and for the following parameters: wind speed, wind direction, temperature, and mixing ratio. The calculated metrics and statistics are summarized in Table 2-3. Table 2-3. Definition and Description of Measures/Metrics for WRF Model Performance Evaluation for High Resolution Air Quality Modeling of New York City Metric Definition # of data pairs The number of observation/simulation data pairs Mean observation value Mean simulation value Mean bias (Bias) Mean error (Gross error) Root mean-squared error (RMSE) Index of agreement (IOA) The average observed value The average simulated value N 1 ( S l O l ) N l 1 where N is the number of data pairs, and S l and O l are the simulated and observed values at site l, respectively, over a given time interval. 1 N N l 1 1 N N l 1 S l O l 2 1 / 2 ( S l O l ) A measure of how well the model represents the pattern of perturbation about the mean value; ranges from 0 to 1. In calculating the statistical measures, METSTAT pairs the WRF model output with the observed data for the appropriate locations and time intervals. The statistical measures were examined and potential biases in the meteorological inputs were identified, with emphasis on those that could affect the use of the meteorological fields in simulating air quality. The statistical measures were also compared with benchmarks derived from prior simulations (Tesche et al., 2002). Since data assimilation was used for selected parameters, a comparison with the observed data (for those parameters) primarily serves as a check on the data assimilation. ICF International 13 New York City Department of Health and Mental Hygiene

23 Application and Evaluation of WRF Additional plots and summaries were prepared to facilitate the overall evaluation of the results for the 4-km grid. These include: Time-series plots comparing simulated and observed values for a variety of meteorological parameters for a) all monitoring sites with the grid (average of all site) and b) selected monitoring sites. Comparison wind frequency diagrams for selected sites and time periods. Plots of simulated and observed vertical temperature and wind profiles for selected sites and time periods, prepared using EPA s Atmospheric Model Evaluation Tool (AMET) model evaluation software (UNC, 2008). Meteorological Data The routine surface and upper-air meteorological data used for model evaluation (ds472.0) are described in Section The dataset contains hourly airways data for approximately 2,000 monitoring sites. These include approximately 20 surface and three upper-air meteorological monitoring sites located within the WRF 5-km modeling grid and six surface sites (and no upper-air sites) located within the WRF 1-km grid. Statistical Benchmarks There are no specific criteria as to what constitutes an acceptable set of meteorological inputs. Nevertheless, many studies refer to a set of statistical benchmarks (Tesche et al., 2002) that can be used to support a finding of acceptable model performance. These benchmarks were developed based on evaluation of approximately 30 applications of the MM5 meteorological model (a predecessor to the WRF model) for specific multi-day simulation periods. The benchmarks are summarized in Table 2-4. Metric Table 2-4. Statistical Benchmarks for Evaluating Meteorological Model Performance. Wind Speed (m/s) Wind Direction (degrees) Temperature (K) Water Vapor Mixing Ratio (g/kg) Bias Gross Error -- < RMSE < IOA Note that not all metrics are applicable to all parameters. 2.3 WRF Modeling Results Synoptic- and Regional-Scale Weather Patterns Qualitative analysis was used to examine how well the WRF modeling results for this application represent the synoptic- and regional-scale weather patterns and key meteorological features (such as high and low-pressure systems, frontal systems, precipitation and snow cover) that characterize the annual simulation period. Plots of the WRF simulation results were compared with weather maps and standard weather analysis products available from the National Oceanic and Atmospheric ICF International 14 New York City Department of Health and Mental Hygiene

24 Application and Evaluation of WRF Administration (NOAA, 2008a and b). For ease of comparison with available national-scale weather products, this analysis focused on the 45-km grid. The assessment considered representation of the weather patterns both near the surface and aloft (500 mb). Note that the 500 mb level (typically about 5,500 m above sea level [asl]) is a standard pressure level used extensively for synoptic-scale weather forecasting and analysis. With the exception of precipitation and snow cover, the comparison was done for a date near the 15 th of each month (the exact date varied according to run segmentation and was chosen so that each plot was of the fifth day of a 5-day simulation period). This was done for consistency and to ensure that the evaluation focused on the simulation results (and not the initial conditions). Precipitation was examined on a monthly basis and snow cover was examined for selected winter days. The qualitative assessment of synoptic- and regional-scale weather patterns is summarized in the remainder of this section. Upper-Air Weather Patterns Accurate simulation of synoptic-scale (large-scale) weather patterns is important since the synopticscale weather patterns determine the range of regional and local conditions that can occur within their region of influence (in this case, within the higher-resolution grids). In the following figures, the WRFderived synoptic-scale weather patterns are compared with standard NWS 500-mb charts that illustrate the synoptic-scale weather patterns for the selected analysis days. Specifically, the 500-mb charts depict the location of pressure ridges and troughs (areas of relatively high and low pressure, respectively), frontal systems, and airflow patterns at the 500-mb level. The 500-mb level is a standard constantpressure level used for meteorological analysis and forecasting; the average height of the 500 mb surface is approximately 5,500 m asl. The plots are for 1200 GMT (0700 EST). The comparison was done for one day for each month as a check on the reasonableness of the WRF-derived synoptic-scale patterns. Plots for January, April, July, and October are presented to illustrate the comparison. For ease of reading, the plots are presented at the end of this subsection. Figure 2-2 compares simulated and observation-based 500-mb charts for 1200 GMT (0700 EST) for January 15, April 14, July 13, and October 16, The observation-based 500-mb analyses (shown on the left) were obtained from the Daily Weather Map Archive for 2008 (NOAA, 2008a). The simulated 500-mb patterns are very similar to the observed patterns. The patterns show an upper-level low influencing the northeastern U.S. for the January, April and July days resulting in westerly to southwesterly winds. October is characterized by more zonal flow. The synoptic-scale weather patterns aloft for the selected days are very well represented by WRF. ICF International 15 New York City Department of Health and Mental Hygiene

25 Application and Evaluation of WRF Figure 2-2. Observed (Left) and Simulated (Right) 500-mb Geopotential Heights (m) and Winds for the Continental U.S. and the NYC 45-km Grid January 15 April 14 ICF International 16 New York City Department of Health and Mental Hygiene

26 Application and Evaluation of WRF Figure 2-2. Observed (Left) and Simulated (Right) 500-mb Geopotential Heights (m) and Winds for the Continental U.S. and the NYC 45-km Grid (Concluded) July 13 October 16 ICF International 17 New York City Department of Health and Mental Hygiene

27 Application and Evaluation of WRF Surface Weather Patterns The regional-scale surface weather patterns further define the prevailing wind and dispersion conditions that affect the air quality within a given area. For many areas in the U.S., episodes of poor air quality are often characterized relative to regional-scale meteorological high- and low-pressure systems and specifically the presence of a surface-based high-pressure system within the area of interest. The location, persistence, and strength of the high-pressure system can be important determinants of air quality. In the following figures, the WRF-derived surface weather patterns are compared with standard NWS surface analysis charts for the selected analysis days. The surface analysis charts depict the location of high and low pressure systems, frontal systems, and airflow patterns at the surface level. The plots are for 1200 GMT (0700 EST). The comparison was done for one day for each month as a check on the reasonableness of the WRF-derived surface weather patterns. Plots for January, April, July, and October are presented to illustrate the comparison. For ease of reading, the plots are presented at the end of this subsection. Figure 2-3 compares simulated and observation-based surface analysis charts for 1200 GMT (0700 EST) for January 15, April 14, July 13, and October 16, The observation-based surface analyses (shown on the left) were obtained from the Daily Weather Map Archive for 2008 (NOAA, 2008a). For all four days, the WRF-derived pressure patterns show many of the same features as the surface analysis, indicating that the synoptic-scale weather patterns near the surface are well represented at different times of the year. ICF International 18 New York City Department of Health and Mental Hygiene

28 Application and Evaluation of WRF Figure 2-3. Observed (Left) and Simulated (Right) Sea Level Pressure (mb) for the Continental U.S. and the NYC 45-km Grid January 15 April 14 ICF International 19 New York City Department of Health and Mental Hygiene

29 Application and Evaluation of WRF Figure 2-3. Observed (Left) and Simulated (Right) 500-mb Geopotential Heights (m) and Winds for the Continental U.S. and the NYC 45-km Grid (Concluded) July 13 October 16 ICF International 20 New York City Department of Health and Mental Hygiene

30 Application and Evaluation of WRF Precipitation The timing and amount of cloud cover and precipitation can affect both air quality and deposition, through both direct and indirect effects on various meteorological and air quality processes. For example, cloud cover directly affects incoming (and outgoing) radiation, temperature, and stability, which, in turn, affect the dispersion characteristics of the atmosphere. Similarly, precipitation affects the moisture content of the soil, which affects the moisture content and stability of the atmosphere, which, in turn, affect various atmospheric chemistry and dispersion processes. In addition, pollutants are removed from the atmosphere through wet deposition. Because they are the combined result of numerous meteorological and geographical factors (including the presence or absence of weather systems, terrain, and land use) and feedback mechanisms, the accurate simulation of cloud cover and precipitation is challenging. In the following figures, WRF-derived monthly precipitation totals are compared with observed precipitation totals. The comparison was done for each month as a check on the reasonableness of overall timing and distribution of the WRF-derived precipitation. For ease of reading, the plots are presented at the end of this subsection. Figure 2-4 compares simulated and observed total precipitation for January, April, July and October. Note that the scales vary from month to month. The plots of observed precipitation (shown on the left) were obtained from the High Plains Regional Climate Center (HPRCC, 2011). The WRF-derived regionalscale precipitation patterns are generally similar to the observed patterns, but overall the precipitation amounts tend to be lower than observed. Over the New York area, total precipitation is underestimated for April, overestimated for July, and well represented for January and October. Overall, the simulation of precipitation is qualitatively better than much of the rest of the 45-km domain. Feedback from the two-way interactive nested grid likely enhances WRF model performance over the area of interest. Figure 2-4. Observed (Left) and Simulated (Right) Monthly Precipitation Totals (in) for the Continental U.S. and the NYC 45-km Grid January ICF International 21 New York City Department of Health and Mental Hygiene

31 Application and Evaluation of WRF April July October ICF International 22 New York City Department of Health and Mental Hygiene

32 Application and Evaluation of WRF Vertical Profiles of Wind, Temperature and Moisture For air quality modeling purposes, reasonable representation of the vertical profiles of wind, temperature, and moisture is required so that the meteorological fields are able to represent the dispersion characteristics of the modeled atmosphere and the dispersion of pollutants within the modeling domain. The vertical profiles determine the stability of the atmosphere and affect vertical mixing, vertical diffusion, horizontal and vertical transport, and deposition of pollutants. Quantitative analysis was used to examine how well the WRF modeling results for this application represent the vertical structure of the atmosphere for the nearest upper-air meteorological monitoring site, at Brookhaven, New York. This site is located in the 5-km grid and is approximately 100 km east of New York City. The WRF-derived vertical profiles were compared with observed data from the twice-daily upper-air sounding for Brookhaven. Radiosondes are released from this site at 0000 and 1200 GMT (1900 and 0700 EST) and the simulation results for both times were compared with the sounding data. As for the spatial weather patterns, the comparison was done for a date near the 15 th of each month (such that each plot was of the fifth day of a 5-day simulation period). Plots for January, April, July and October are presented at the end of this section. Figure 2-5 compares simulated and observed vertical profiles of potential temperature (left), humidity (center left), wind speed (center right), and wind vectors (right) for 1200 GMT (0700 EST) for a middle day for each month for Brookhaven, NY. The observed morning temperature profiles for the selected days are well represented in the simulation results, especially for the upper levels. While the simulated values are generally reasonable, the vertical variations in the humidity and wind speed profiles are not as well represented in the WRF modeling results. The wind vectors show good agreement between the simulated and observed wind directions. The findings for 0000 GMT (1900 EST) and selected days for other months (not shown) are similar. ICF International 23 New York City Department of Health and Mental Hygiene

33 Application and Evaluation of WRF Figure 2-5. Simulated and Observed Vertical Profiles of Potential Temperature, Humidity, Wind Speed, and Wind Vectors (Shown in the Order Listed) for 1200 GMT (0700 EST) for Brookhaven, NY (Observed Values are in Red; Simulated Values are in Blue) January 15 April 14 ICF International 24 New York City Department of Health and Mental Hygiene

34 Application and Evaluation of WRF July 13 October Temporal Variations in Key Parameters for the High-Resolution Modeling Grids A more detailed examination of how well the WRF modeling results for this application represent observed meteorological conditions was obtained by comparing the simulated values and temporal variations of wind speed, wind direction, temperature and water vapor mixing ratio with observed data for surface monitoring sites located within the high-resolution (5- and 1-km) grids. For an overall comparison, average simulated values (averaged over all sites in the grid) were compared with average observed values for each hour and each simulation day, using NCAR ds472.0 data. The comparison was done for all months. To illustrate the results, grid-average time series plots for January, April, July, and October are presented in this section. Note ICF International 25 New York City Department of Health and Mental Hygiene

35 Application and Evaluation of WRF that the scale may vary from month to month to capture the range of observed and simulated values. For ease of reading, all plots are presented at the end of this subsection. Figures 2-6 through 2-9 show the results for the 5-km grid. Figure 2-6 compares hourly average values of simulated and observed wind speed for each hour and day of the selected months. Each time-series plot displays the values for one month. The averages (both observed and simulated) are for all sites located in the 5-km modeling grid. The plots illustrate that, for all months, average surface wind speeds and the day-to-day variations in wind speed are well represented by the WRF model. The model tends to overestimate wind speeds. Figure 2-7 compares hourly average values of simulated and observed wind direction for each hour and day of the selected months. Again, each time-series plot displays the values for one month and the averages (both observed and simulated) are for all site locations in the 5-km modeling grid. Note that the wind direction plots can sometimes indicate large differences that are not indicative of poor model performance (this can occur when the simulated and observed values are close to 360 degrees but one has a more westerly component and the other a more easterly component (for example 355 and 5 degrees are only 10 degrees apart but would show up as a large difference on the plots). Surface wind directions and especially the changes in wind direction over time are very well represented by the WRF model. The use of obs nudging for the surface winds in the 5-km grid likely contributes to the good representation of surface wind speed and directions. Note that the same observations used for the model performance evaluation are also used for the nudging. Figure 2-8 compares hourly average values of simulated and observed temperature for each hour and day of the selected months, for the 5-km modeling grid. The diurnal, day-to-day, multi-day, monthly, and seasonal variations in temperature are well represented by the WRF model. Performance is better for the warmer months as indicated by the plot for July. For January, there is a tendency for the model to underestimate the maximum observed temperatures, especially for the warmer days in early January. For April and October, there is a tendency for the model to overestimate the minimum observed temperatures. Figure 2-9 compares hourly average values of simulated and observed mixing ratio for each hour and day of the selected months, for the 5-km modeling grid. The time-series plots for water vapor mixing ratio indicate less skill in simulating this parameter, compared to wind speed, wind direction, and temperature. Overall, humidity is well represented by WRF. Multi-day and monthly variations are captured by the model. The model has some difficulty simulating the diurnal and day-to-day variations in moisture, especially during the summer months. Figures 2-10 through 2-13 show the results for the 1-km grid. The results are similar but there is a slightly greater mix of under- and overestimation of wind speeds, compared to the 5-km grid. ICF International 26 New York City Department of Health and Mental Hygiene

36 Application and Evaluation of WRF Figure 2-6. Average Observed (Obs) and Simulated (WRF) Surface Wind Speed (m/s) for the NYC 5-km Grid January Wind Speed (m/s) / 1 1/ 2 1/ 3 1/ 4 1/ 5 1/ 6 1/ 7 1/ 8 1/ 9 1/10 1/11 1/12 1/13 1/14 1/15 1/16 1/17 1/18 1/19 1/20 1/21 1/22 1/23 1/24 1/25 1/26 1/27 1/28 1/29 1/30 1/31 Obs WRF April Wind Speed (m/s) / 1 4/ 2 4/ 3 4/ 4 4/ 5 4/ 6 4/ 7 4/ 8 4/ 9 4/10 4/11 4/12 4/13 4/14 4/15 4/16 4/17 4/18 4/19 4/20 4/21 4/22 4/23 4/24 4/25 4/26 4/27 4/28 4/29 4/30 Obs WRF July Wind Speed (m/s) Obs 7/ 1 7/ 2 7/ 3 7/ 4 7/ 5 7/ 6 7/ 7 7/ 8 7/ 9 7/10 7/11 7/12 7/13 7/14 7/15 7/16 7/17 7/18 7/19 7/20 7/21 7/22 7/23 7/24 7/25 7/26 7/27 7/28 7/29 7/30 7/31 WRF Wind Speed (m/s) October 10/1 10/2 10/3 10/4 10/5 10/6 10/7 10/8 10/9 10/10 10/11 10/12 10/13 10/14 10/15 10/16 10/17 10/18 10/19 10/20 10/21 10/22 10/23 10/24 10/25 10/26 10/27 10/28 10/29 10/30 10/31 Obs WRF ICF International 27 New York City Department of Health and Mental Hygiene

37 Application and Evaluation of WRF Figure 2-7. Average Observed (Obs) and Simulated (WRF) Surface Wind Direction (degrees) for the NYC 5-km Grid Wind Direction (Deg.) January 1/ 1 1/ 2 1/ 3 1/ 4 1/ 5 1/ 6 1/ 7 1/ 8 1/ 9 1/10 1/11 1/12 1/13 1/14 1/15 1/16 1/17 1/18 1/19 1/20 1/21 1/22 1/23 1/24 1/25 1/26 1/27 1/28 1/29 1/30 1/31 Obs WRF Wind Direction (Deg.) April 4/ 1 4/ 2 4/ 3 4/ 4 4/ 5 4/ 6 4/ 7 4/ 8 4/ 9 4/10 4/11 4/12 4/13 4/14 4/15 4/16 4/17 4/18 4/19 4/20 4/21 4/22 4/23 4/24 4/25 4/26 4/27 4/28 4/29 4/30 Obs WRF Wind Direction (Deg.) July 7/ 1 7/ 2 7/ 3 7/ 4 7/ 5 7/ 6 7/ 7 7/ 8 7/ 9 7/10 7/11 7/12 7/13 7/14 7/15 7/16 7/17 7/18 7/19 7/20 7/21 7/22 7/23 7/24 7/25 7/26 7/27 7/28 7/29 7/30 7/31 Obs WRF Wind Direction (Deg.) October 10/1 10/2 10/3 10/4 10/5 10/6 10/7 10/8 10/9 10/10 10/11 10/12 10/13 10/14 10/15 10/16 10/17 10/18 10/19 10/20 10/21 10/22 10/23 10/24 10/25 10/26 10/27 10/28 10/29 10/30 10/31 Obs WRF ICF International 28 New York City Department of Health and Mental Hygiene

38 Application and Evaluation of WRF Figure 2-8. Average Observed (Obs) and Simulated (WRF) Surface Temperature (K) for the NYC 5-km Grid January Temperature (K) / 1 1/ 2 1/ 3 1/ 4 1/ 5 1/ 6 1/ 7 1/ 8 1/ 9 1/10 1/11 1/12 1/13 1/14 1/15 1/16 1/17 1/18 1/19 1/20 1/21 1/22 1/23 1/24 1/25 1/26 1/27 1/28 1/29 1/30 1/31 Obs WRF 310 April Temperature (K) / 1 4/ 2 4/ 3 4/ 4 4/ 5 4/ 6 4/ 7 4/ 8 4/ 9 4/10 4/11 4/12 4/13 4/14 4/15 4/16 4/17 4/18 4/19 4/20 4/21 4/22 4/23 4/24 4/25 4/26 4/27 4/28 4/29 4/30 Obs WRF 310 July Temperature (K) / 1 7/ 2 7/ 3 7/ 4 7/ 5 7/ 6 7/ 7 7/ 8 7/ 9 7/10 7/11 7/12 7/13 7/14 7/15 7/16 7/17 7/18 7/19 7/20 7/21 7/22 7/23 7/24 7/25 7/26 7/27 7/28 7/29 7/30 7/31 Obs WRF 300 October Temperature (K) /1 10/2 10/3 10/4 10/5 10/6 10/7 10/8 10/9 10/10 10/11 10/12 10/13 10/14 10/15 10/16 10/17 10/18 10/19 10/20 10/21 10/22 10/23 10/24 10/25 10/26 10/27 10/28 10/29 10/30 10/31 Obs WRF ICF International 29 New York City Department of Health and Mental Hygiene

39 Application and Evaluation of WRF Figure 2-9. Average Observed (Obs) and Simulated (WRF) Surface Water Vapor Mixing Ratio (g/kg) for the NYC 5- km Grid January Humidity (g/kg) / 1 1/ 2 1/ 3 1/ 4 1/ 5 1/ 6 1/ 7 1/ 8 1/ 9 1/10 1/11 1/12 1/13 1/14 1/15 1/16 1/17 1/18 1/19 1/20 1/21 1/22 1/23 1/24 1/25 1/26 1/27 1/28 1/29 1/30 1/31 Obs WRF April Humidity (g/kg) / 1 4/ 2 4/ 3 4/ 4 4/ 5 4/ 6 4/ 7 4/ 8 4/ 9 4/10 4/11 4/12 4/13 4/14 4/15 4/16 4/17 4/18 4/19 4/20 4/21 4/22 4/23 4/24 4/25 4/26 4/27 4/28 4/29 4/30 Obs WRF July Humidity (g/kg) / 1 7/ 2 7/ 3 7/ 4 7/ 5 7/ 6 7/ 7 7/ 8 7/ 9 7/10 7/11 7/12 7/13 7/14 7/15 7/16 7/17 7/18 7/19 7/20 7/21 7/22 7/23 7/24 7/25 7/26 7/27 7/28 7/29 7/30 7/31 Obs WRF October 12 Humidity (g/kg) Obs 10/1 10/2 10/3 10/4 10/5 10/6 10/7 10/8 10/9 10/10 10/11 10/12 10/13 10/14 10/15 10/16 10/17 10/18 10/19 10/20 10/21 10/22 10/23 10/24 10/25 10/26 10/27 10/28 10/29 10/30 10/31 WRF ICF International 30 New York City Department of Health and Mental Hygiene

40 Application and Evaluation of WRF Figure Average Observed (Obs) and Simulated (WRF) Surface Wind Speed (m/s) for the NYC 1-km Grid 12 January Wind Speed (m/s) Obs 0 1/ 1 1/ 2 1/ 3 1/ 4 1/ 5 1/ 6 1/ 7 1/ 8 1/ 9 1/10 1/11 1/12 1/13 1/14 1/15 1/16 1/17 1/18 1/19 1/20 1/21 1/22 1/23 1/24 1/25 1/26 1/27 1/28 1/29 1/30 1/31 WRF April Wind Speed (m/s) / 1 4/ 2 4/ 3 4/ 4 4/ 5 4/ 6 4/ 7 4/ 8 4/ 9 4/10 4/11 4/12 4/13 4/14 4/15 4/16 4/17 4/18 4/19 4/20 4/21 4/22 4/23 4/24 4/25 4/26 4/27 4/28 4/29 4/30 Obs WRF July 9 8 Wind Speed (m/s) / 1 7/ 2 7/ 3 7/ 4 7/ 5 7/ 6 7/ 7 7/ 8 7/ 9 7/10 7/11 7/12 7/13 7/14 7/15 7/16 7/17 7/18 7/19 7/20 7/21 7/22 7/23 7/24 7/25 7/26 7/27 7/28 7/29 7/30 7/31 Obs WRF 12 October Wind Speed (m/s) Obs 0 10/1 10/2 10/3 10/4 10/5 10/6 10/7 10/8 10/9 10/10 10/11 10/12 10/13 10/14 10/15 10/16 10/17 10/18 10/19 10/20 10/21 10/22 10/23 10/24 10/25 10/26 10/27 10/28 10/29 10/30 10/31 WRF ICF International 31 New York City Department of Health and Mental Hygiene

41 Application and Evaluation of WRF Figure Average Observed (Obs) and Simulated (WRF) Surface Wind Direction (degrees) for the NYC 1-km Grid Wind Direction (Deg.) January 1/ 1 1/ 2 1/ 3 1/ 4 1/ 5 1/ 6 1/ 7 1/ 8 1/ 9 1/10 1/11 1/12 1/13 1/14 1/15 1/16 1/17 1/18 1/19 1/20 1/21 1/22 1/23 1/24 1/25 1/26 1/27 1/28 1/29 1/30 1/31 Obs WRF Wind Direction (Deg.) April 4/ 1 4/ 2 4/ 3 4/ 4 4/ 5 4/ 6 4/ 7 4/ 8 4/ 9 4/10 4/11 4/12 4/13 4/14 4/15 4/16 4/17 4/18 4/19 4/20 4/21 4/22 4/23 4/24 4/25 4/26 4/27 4/28 4/29 4/30 Obs WRF Wind Direction (Deg.) July 7/ 1 7/ 2 7/ 3 7/ 4 7/ 5 7/ 6 7/ 7 7/ 8 7/ 9 7/10 7/11 7/12 7/13 7/14 7/15 7/16 7/17 7/18 7/19 7/20 7/21 7/22 7/23 7/24 7/25 7/26 7/27 7/28 7/29 7/30 7/31 Obs WRF Wind Direction (Deg.) October 10/1 10/2 10/3 10/4 10/5 10/6 10/7 10/8 10/9 10/10 10/11 10/12 10/13 10/14 10/15 10/16 10/17 10/18 10/19 10/20 10/21 10/22 10/23 10/24 10/25 10/26 10/27 10/28 10/29 10/30 10/31 Obs WRF ICF International 32 New York City Department of Health and Mental Hygiene

42 Application and Evaluation of WRF Figure Average Observed (Obs) and Simulated (WRF) Surface Temperature (K) for the NYC 1-km Grid January Temperature (K) / 1 1/ 2 1/ 3 1/ 4 1/ 5 1/ 6 1/ 7 1/ 8 1/ 9 1/10 1/11 1/12 1/13 1/14 1/15 1/16 1/17 1/18 1/19 1/20 1/21 1/22 1/23 1/24 1/25 1/26 1/27 1/28 1/29 1/30 1/31 Obs WRF 310 April Temperature (K) / 1 4/ 2 4/ 3 4/ 4 4/ 5 4/ 6 4/ 7 4/ 8 4/ 9 4/10 4/11 4/12 4/13 4/14 4/15 4/16 4/17 4/18 4/19 4/20 4/21 4/22 4/23 4/24 4/25 4/26 4/27 4/28 4/29 4/30 Obs WRF Temperature (K) July 7/ 1 7/ 2 7/ 3 7/ 4 7/ 5 7/ 6 7/ 7 7/ 8 7/ 9 7/10 7/11 7/12 7/13 7/14 7/15 7/16 7/17 7/18 7/19 7/20 7/21 7/22 7/23 7/24 7/25 7/26 7/27 7/28 7/29 7/30 7/31 Obs WRF 310 October Temperature (K) /1 10/2 10/3 10/4 10/5 10/6 10/7 10/8 10/9 10/10 10/11 10/12 10/13 10/14 10/15 10/16 10/17 10/18 10/19 10/20 10/21 10/22 10/23 10/24 10/25 10/26 10/27 10/28 10/29 10/30 10/31 Obs WRF ICF International 33 New York City Department of Health and Mental Hygiene

43 Application and Evaluation of WRF Figure Average Observed (Obs) and Simulated (WRF) Surface Water Vapor Mixing Ratio (g/kg) for the NYC 1-km Grid Humidity (g/kg) January 1/ 1 1/ 2 1/ 3 1/ 4 1/ 5 1/ 6 1/ 7 1/ 8 1/ 9 1/10 1/11 1/12 1/13 1/14 1/15 1/16 1/17 1/18 1/19 1/20 1/21 1/22 1/23 1/24 1/25 1/26 1/27 1/28 1/29 1/30 1/31 Obs WRF 12 April Humidity (g/kg) / 1 4/ 2 4/ 3 4/ 4 4/ 5 4/ 6 4/ 7 4/ 8 4/ 9 4/10 4/11 4/12 4/13 4/14 4/15 4/16 4/17 4/18 4/19 4/20 4/21 4/22 4/23 4/24 4/25 4/26 4/27 4/28 4/29 4/30 Obs WRF Humidity (g/kg) July 7/ 1 7/ 2 7/ 3 7/ 4 7/ 5 7/ 6 7/ 7 7/ 8 7/ 9 7/10 7/11 7/12 7/13 7/14 7/15 7/16 7/17 7/18 7/19 7/20 7/21 7/22 7/23 7/24 7/25 7/26 7/27 7/28 7/29 7/30 7/31 Obs WRF 14 October Humidity (g/kg) /1 10/2 10/3 10/4 10/5 10/6 10/7 10/8 10/9 10/10 10/11 10/12 10/13 10/14 10/15 10/16 10/17 10/18 10/19 10/20 10/21 10/22 10/23 10/24 10/25 10/26 10/27 10/28 10/29 10/30 10/31 Obs WRF ICF International 34 New York City Department of Health and Mental Hygiene

44 Application and Evaluation of WRF Wind Direction Frequency Distributions for Selected Monitoring Sites Good representation of the near surface airflow patterns is an important factor when using the wind fields to drive an air quality model. Good representation of the magnitude, location, and timing of peak concentrations and episodic conditions (as well as a meaningful assessment of short-term impacts) requires an accurate representation of the wind flow parameters. Use of an air quality model to estimate seasonal and annual air quality and deposition (and especially seasonal and annual impacts on air quality and deposition) also requires that the distribution of wind speeds and directions is similar to observed for the period of interest. To examine how well the WRF simulation results capture the observed frequency of wind directions in and around New York City, the simulated frequency of surface wind direction was compared with that for the observed data for three sites located within the City (five-borough area) and four additional sites located outside of the City. The three NYC sites are located at LaGuardia Airport (LGA) in Queens, at John F. Kennedy Airport (JFK) also in Queens, and in Central Park (NYC) in Manhattan. The four additional sites are located at Newark International Airport (EWR) in Newark, NJ, Teterboro Airport (TEB) in Bergen County, NJ, Westchester Airport (HPN) in Westchester County, NY, and Brookhaven Airport (HWV) on Long Island. These sites are located to the south, west, north, and east, respectively, of NYC. EWR and TEB are located within the 1-km grid, while HPN and HWV are located just outside of the 1-km grid (in the 5-km grid). The wind direction frequency plots are presented in the remainder of this section. For ease of reading, all plots are presented at the end of this section. Figure 2-14 through 2-20 summarize the frequency of occurrence of winds from eight principal wind directions. Each wind direction represents the 45 degree sector centered on the direction (e.g., N winds range from to 22.5 degrees, NE winds range from 22.5 to 67.5 degrees), where the wind direction is the direction from which the winds blow. This information is then graphically displayed in a radar diagram, such that each ring moving outward from the center represents a percent increase in the frequency of occurrence of the wind from a given direction. The wind diagrams are based on hourly observed and simulated wind directions for the full annual simulation period. For all sites, the observed predominant wind directions and the distributions of wind direction are reasonably well represented by the simulated surface winds. The agreement between the simulated and observed winds is worst for the Central Park site, and this is expected, given the proximity of this site to New York s skyscrapers and tall buildings, which likely affect airflow patterns throughout Manhattan. Although the WRF model does account for an increased roughness length and adjusts other land-use parameters over the urban area, these are applied on a grid-cell by grid-cell basis and the NYC skyline and its effects on the wind patterns are not fully resolved by WRF. The WRF results for LGA are characterized by a greater frequency of northerly winds (and a less frequent northeasterly and northwesterly winds), compared to the observations. The results for JFK, NYC, HPN, and HWV are characterized by slightly less frequent southeasterly winds, compared to the observations. ICF International 35 New York City Department of Health and Mental Hygiene

45 Application and Evaluation of WRF Figure Comparison of Observed and Simulated Wind Direction Frequency for the LaGuardia Airport Monitoring Site (LGA) Annual Observed Simulated WD % WD % 40 N 8 30 N 12 NE NE 13 E 3 10 E 4 SE 9 0 SE 8 S 13 S 14 SW 12 SW 15 W 15 W 18 NW 20 NW Figure Comparison of Observed and Simulated Wind Direction Frequency for the JFK Airport Monitoring Site (JFK) Annual Observed Simulated WD % WD % 40 N N 12 NE NE 8 E 5 10 E 6 SE 13 0 SE 9 S 16 S 17 SW 15 SW 14 W 13 W 18 NW 15 NW Figure Comparison of Observed and Simulated Wind Direction Frequency for the Central Park Tower Monitoring Site (NYC) Annual Observed Simulated WD % WD % 40 N 7 30 N 13 NE NE 16 E 3 10 E 5 SE 11 0 SE 6 S 4 S 11 SW 19 SW 12 W 29 W 22 NW 4 NW ICF International 36 New York City Department of Health and Mental Hygiene

46 Application and Evaluation of WRF Figure Comparison of Observed and Simulated Wind Direction Frequency for the Newark International Airport Monitoring Site (EWR) Annual Observed Simulated WD % WD % 40 N N 15 NE NE 9 E 4 10 E 4 SE 10 0 SE 8 S 10 S 11 SW 19 SW 17 W 17 W 18 NW 16 NW Figure Comparison of Observed and Simulated Wind Direction Frequency for the Teterboro Airport Monitoring Site (TEB) Annual Observed Simulated WD % WD % 40 N N 15 NE 9 20 NE 7 E 4 10 E 4 SE 6 0 SE 7 S 16 S 19 SW 15 SW 10 W 14 W 19 NW 23 NW Figure Comparison of Observed and Simulated Wind Direction Frequency for the Westchester Airport Monitoring Site (HPN) Annual Observed Simulated WD % WD % 40 N 7 30 N 11 NE 7 20 NE 7 E 8 10 E 8 SE 13 0 SE 7 S 7 S 10 SW 6 SW 8 W 24 W 22 NW 28 NW ICF International 37 New York City Department of Health and Mental Hygiene

47 Application and Evaluation of WRF Figure Comparison of Observed and Simulated Wind Direction Frequency for the Brookhaven Airport Monitoring Site (HWV) Annual Observed Simulated WD % WD % 40 N 9 30 N 9 NE 3 20 NE 5 E 4 10 E 4 SE 13 0 SE 8 S 23 S 22 SW 14 SW 18 W 15 W 17 NW 18 NW Statistical Measures of Model Performance Statistical measures were used to quantify model performance for the 15-, 5-, and 1-km grids. METSTAT was applied for each grid, for each month, and for the following parameters: wind speed, wind direction, temperature, and mixing ratio. METSTAT uses surface data only. Monthly model performance statistics are presented in Table 2-5 through 2-7, for the 15-, 5-, and 1-km grids, respectively. For comparison purposes, the statistical benchmarks (goals) presented in Table 2-4 are included in the table, as appropriate. ICF International 38 New York City Department of Health and Mental Hygiene

48 Application and Evaluation of WRF Table 2-5. Statistical Summary of WRF Model Performance for the NYC 15-km Modeling Grid Metric/Month Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Goal Wind Speed Mean Obs (m/s) Mean Sim (m/s) Bias (m/s) Gross Error (m/s) RMSE (m/s) <2 IOA Wind Direction Mean Obs (deg) Mean Sim (deg) Bias (deg) Gross Error (deg) Temperature Mean Obs (K) Mean Sim (K) Bias (K) <30 Gross Error (K) RMSE (K) IOA Mixing Ratio (g/kg) Mean Obs (g/kg) Mean Sim (g/kg) Bias (g/kg) Gross Error (g/kg) RMSE (g/kg) IOA ICF International 39 New York City Department of Health and Mental Hygiene

49 Application and Evaluation of WRF Table 2-6. Statistical Summary of WRF Model Performance for the NYC 5-km Modeling Grid Metric/Month Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Goal Wind Speed Mean Obs (m/s) Mean Sim (m/s) Bias (m/s) Gross Error (m/s) RMSE (m/s) <2 IOA Wind Direction Mean Obs (deg) Mean Sim (deg) Bias (deg) Gross Error (deg) Temperature Mean Obs (K) Mean Sim (K) Bias (K) <30 Gross Error (K) RMSE (K) IOA Mixing Ratio (g/kg) Mean Obs (g/kg) Mean Sim (g/kg) Bias (g/kg) Gross Error (g/kg) RMSE (g/kg) IOA ICF International 40 New York City Department of Health and Mental Hygiene

50 Application and Evaluation of WRF Table 2-7. Statistical Summary of WRF Model Performance for the NYC 1-km Modeling Grid Metric/Month Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Goal Wind Speed Mean Obs (m/s) Mean Sim (m/s) Bias (m/s) Gross Error (m/s) RMSE (m/s) <2 IOA Wind Direction Mean Obs (deg) Mean Sim (deg) Bias (deg) Gross Error (deg) Temperature Mean Obs (K) Mean Sim (K) Bias (K) <30 Gross Error (K) RMSE (K) IOA Mixing Ratio (g/kg) Mean Obs (g/kg) Mean Sim (g/kg) Bias (g/kg) Gross Error (g/kg) RMSE (g/kg) IOA ICF International 41 New York City Department of Health and Mental Hygiene

51 Application and Evaluation of WRF Discussion of Statistical Model Performance The statistical measures indicate that model performance is similar for all three grids, and consistently good for all parameters and months. Focusing on the 1-km grid, the wind speed bias is within 0.5 m/s for all months. The bias is negative (simulated winds are slower than observed) for the winter and early spring months ((Jan-Apr; Nov-Dec) and positive for the remaining months (May-Oct). The corresponding RMSE is less than 2 m/s for all months. The IOA ranges from 0.71 to Overall, surface wind speeds within the 1-km grid are very well represented. Results are similar for the 5- and 15-km grids, with the exception that the bias increases with increasing grid cell size and is positive for nearly all months. The wind direction bias is within 5 degrees for all months. The gross error ranges from approximately 18 to 32 degrees, indicating excellent agreement with surface wind directions, on average, for the 1-km grid. The results are similar for the 5- and 15-km grids. The temperature bias is within 0.5 K for six months (May-Oct) and in the range of 2 K for the remaining months, with the exception of January and December. The gross error is greater and ranges from 1.6 to 2.9 K, indicating that the error includes both over- and underestimation of the hourly temperatures (as confirmed by the time-series plots). Colder than observed wintertime temperatures account for much of the errors (this was also apparent in the time series plots). The IOA ranges from 0.7 to 0.9. The results are similar for the 5- and 15-km grids (except that the IOA is better for the coarser grids). The mixing ratio bias is within 1 g/kg for all months. The gross error ranges from approximately 0.4 to 1.2 g/kg. The IOA ranges from approximately 0.6 to 0.8 g/kg the lower values reflecting that the model has some difficulty simulating the diurnal and day-to-day variations in moisture, especially during the summer months, as indicated in the time-series plots. On average, the mixing ratio is well represented for the 1-km grid. The results are similar for the 5- and 15-km grids (except that the IOA is better for the coarser grids). Overall, WRF model performance is good to very good for this application and the meteorological fields were processed for use in the CMAQ modeling. ICF International 42 New York City Department of Health and Mental Hygiene

52 3 Emission Inventory Preparation This section describes the data, methods, and procedures that were used to prepare the model-ready emission inventories for the 2008 base-year scenario, the alternative EGU base scenario, and the heating oil and EGU fuel scenarios for application of CMAQ for the New York City air quality assessment. A detailed summary of the emission inventories is also provided. 3.1 Emissions Data and Processing Procedures Emissions Data The emission inventories were prepared using SMOKE version 3.1. The data sources for the base-year and various emission scenarios are listed below: 2007 emissions for all sectors included in the EPA s 2008-based modeling platform (2007v5) (EPA, 2012a) 2005 and 2014 emissions included in the EPA s 2005-based modeling platform version 4 (EPA, 2011) 2012 continuous emissions monitoring (CEM) data for EGUs from EPA s Clean Air Market reporting system New York City boiler emissions provided by the New York City Department of Health & Mental Hygiene (Kheirbek, 2013) The modeling inventories include the following pollutants: volatile organic compounds (VOC), oxides of nitrogen (NO x ), carbon monoxide (CO), sulfur dioxide (SO 2 ), fine particulates (PM 2.5 ), coarse particulates (PM 10 ), and ammonia (NH 3 ). More detailed descriptions of EPA s 2008-based modeling platform, the EGU emissions from EPA s based platform, and 2011 New York City boiler data are provided below: EPA s 2008-Based Modeling Platform The emissions included in the EPA s 2008-based modeling platform are based on data from the 2008 NEI Version 2. Source categories include: Electric Generating Unit (EGU) point sources (estimated with the Integrated Planning Model [IPM ]) Other point sources (non-ipm) excluding the NYC No. 2 distillate, and No. 4 and No. 6 residual boilers Non-point (area) sources excluding the NYC No. 2 distillate, and No. 4 and No. 6 residual boilers On-road motor vehicles Non-road motor vehicles Average-year wildfires and prescribed fires Fugitive dust Agricultural sources ICF International 43 New York City Department of Health and Mental Hygiene

53 Emission Inventory Preparation Class 1 & 2 commercial marine vessel and non-rail maintenance locomotives Category 3 commercial marine vessels Canadian and offshore emissions for point, non-point and on-road sectors Biogenic sources Oceanic gaseous chlorine emissions The 2008 base-year inventory and heating oil scenarios include 2008 NEI emissions for all of the sectors listed above. The alternative EGU base and EGU fuel scenarios include 2008 NEI emissions for all of the sectors listed above, with the exception of the EGU emissions. EGU Emissions in the EPA s 2005-Based Modeling Platform The 2005 EGU emissions included in the alternative EGU Base inventory were based on version 4.2 of EPA s 2005-based modeling platform. The 2005 EGU emissions and the projected 2014 EGU emissions with the Clean Air Interstate Rule (CAIR) controls were used for the EGU fuel scenarios for those EGU sources located in outlying states. EGU Emissions Derived from Continuous Emissions Monitoring (CEM) Data in EPA s Clean Air Market Reporting System Emissions data for 2012 representing present day levels for the majority of EGU s within the modeling domain were obtained from EPA s Clean Air Market reporting system. These estimates reflect any changes (e.g., shut-downs, new sources, emission controls) that may have been imposed on EGU sources since New York City Boiler Emissions The New York City data provides the emissions and locations of all boilers from heating systems in commercial, residential, institutional, and industrial buildings in New York City that use residual oil (No. 6 or No. 4) as their primary fuel, and emissions and locations of all No. 2 oil burning boilers over 350,000 BTUs in NYC subject to Department of Environmental Protection (DEP) permitting. No. 4 and No. 6 Residual Oil Boilers The list of approximately 8000 residual oil boilers was prepared based on permit data from the NYC DEP and boiler conversion tracking information compiled by ICF. Emissions for primary PM 2.5, SO 2, NO x and CO were generated using information of heat throughput of each boiler and emissions factors from NYC DEP, derived from EPA s AP-42 emissions factors for stationary sources. Emissions for the residual boilers were estimated for three time periods: Base (2008): Base-year emissions prior to any regulation affecting residual oil boilers. These emissions were incorporated into the 2008 base-year inventory. Present day (2013): Present day emissions representing partial phase-in of rules affecting No. 4 and No. 6 boilers. All boilers that have either completed conversion or were under construction for ICF International 44 New York City Department of Health and Mental Hygiene

54 Emission Inventory Preparation conversion in 2013 to cleaner heating fuels were adjusted to account for reduced emissions when burning the cleaner heating fuel. At all No. 4 boilers where no conversions occurred, emissions were adjusted to account for City Council rules that require 1500 ppm Low Sulfur No. 4. These emissions were incorporated into the first heating oil scenario (Scenario #1). Future year (2030): This scenario accounts for full implementation of heating oil rules in If conversion occurred prior to 2013, emissions remained the same. For all unconverted boilers, new emissions were estimated using natural gas emissions factors. These emissions were incorporated into the second heating oil scenario (Scenario #2). No. 2 Distillate Oil Boilers The list of distillate boilers was developed from NYC DEP s list of certifications and registrations from boilers listed as using a primary fuel of No. 2 oil. Emissions for primary PM 2.5, SO 2, NO x and CO were generated using information of heat throughput of each boiler, emissions factors from NYC DEP, derived from EPA s AP-42 emissions factors for stationary sources, and data from Brookhaven National Laboratory report on emissions factors from ULS No. 2 fuel. Emissions for the distillate boiler were estimated as follows: Base (2008): Base-year emissions prior to any regulation affecting residual oil boilers. These emissions were incorporated into the 2008 base-year inventory. Present day (2013): All emissions from boilers reflecting new state rules requiring the use of Ultra- Low Sulfur, 15 ppm S fuel. These emissions were incorporated into the first heating oil scenario (Scenario #1). Future year (2030) emissions for No. 2 distillate oil boilers are the same as for 2013 and were incorporated into the second heating oil scenario (Scenario #2). The New York City DEP database only contains data for boilers above 350,000 BTU that require registration or certification. The 2008 emissions for the distillate oil boilers operating in the City below the 350,000 BTU threshold (the remainder of No. 2 emissions) were obtained by subtracting out the total emissions from the distillate oil boilers over 350,000 BTU from the EPA s 2008-based platform estimates of No. 2 heating oil emissions in New York City. The remainder emissions were allocated to grid cells where buildings containing the smaller boilers that use No. 2 home heating oil are located, using information provided by NYC DOHMH. The allocation used weighted area values, which are calculated as the ratios of the No. 2 oil heating space of each building over the total No. 2 oil heating space not accounted for by permits. These emissions were incorporated into the 2008 base-year inventory. The present day (2013) remainder emissions were estimated by reducing 98 percent of PM 2.5 and 99 percent of SO 2 from the base-year emissions to reflect new state rules requiring the use of Ultra-Low Sulfur, 15 ppm Sulfur fuel. These emissions were incorporated into the first heating oil scenario (Scenario #1). The future year (2030) remainder emissions were kept at the same level as These emissions were incorporated into the second heating oil scenario (Scenario #2). ICF International 45 New York City Department of Health and Mental Hygiene

55 Emission Inventory Preparation Emissions Inventory Preparation Methodology As noted above, SMOKE, version 3.1 was utilized to process the emissions with the in-line point-source emissions feature to prepare CMAQ-ready inputs for the base-year and associated scenarios using source sector files obtained from EPA and boiler data provided by NYC DOHMH. Emission files were prepared for the 15-, 5- and 1-km resolution CMAQ grids, and included processing of all source sectors using various SMOKE programs and inputs, and review and quality assurance checks. The general procedures followed in preparing the modeling inventories, using various programs included with SMOKE, are the following: Perform chemical speciation to transform input criteria pollutants into the CB-05 chemical mechanism species, as required by CMAQ. The speciation of PM 2.5 includes the CMAQ required additional species generated using the EPA provided speciation profiles, which are based on the updated speciation profiles in SPECIATE 4.3. Perform temporal distribution to convert annual/monthly emissions into hourly emissions. Perform spatial distribution of the emissions to the 15-, 5- and 1-km resolution modeling grids. Merge emissions from all source categories into the CMAQ model-ready files. Review and quality assure the processing steps and resulting emissions. Additional details for each source sector are provided in the following sections. Point Sources The point source emissions were processed using SMOKE with the in-line point source option, and EPA-provided speciation/temporal profiles and associated cross reference files. The New York City boilers were modeled as low-level point sources allocated at various building locations throughout New York City with monthly and diurnal variations (relative to a flat weekly profile) as shown in Figure 3-1. Figure 3-1. Monthly and Diurnal Variations in New York City Boiler Emissions 20% 10% 0% Monthly Variation: New York City Boiler residual oil boilers Distillate oil boilers 7% 6% 5% 4% 3% 2% 1% 0% Diurnal Variation New York City Boiler ICF International 46 New York City Department of Health and Mental Hygiene

56 Emission Inventory Preparation Non-Point Sources Emissions for all major area source categories were obtained from the EPA s data; subcategories include industrial processes, miscellaneous area sources, mobile sources (marine vessels, aircraft, railroads, paved roads, etc.), average fires, agriculture, solvent utilization, stationary source fuel combustion, storage and transport, and waste disposal, treatment, and recovery. The area source emissions were processed with SMOKE using EPA-provided speciation/temporal/surrogate profiles and associated cross reference files. The gridded surrogates used for spatially allocating anthropogenic emissions for the 15- km, 5-km and 1-km grids were prepared using the Spatial Allocator 4.0 and associated shape files. On-road and Non-road Mobile Sources Estimates for on-road emissions were prepared by combining the emission factors generated using EPA s Motor Vehicle Emissions Simulator MOVES2010b, activity data, and WRF-derived 2008 meteorological data, to produce gridded, hourly emissions. There are three sets of emission factors for the non-refueling part of on-road sources: Rate per Distance (RPD): modeling the on-network emissions, including vehicle exhaust, evaporation, evaporative permeation, break wear, and tire wear; Rate per Vehicle (RPV): modeling the off-network emissions, including vehicle exhaust, evaporative, and evaporative permeation; and Rate per Profile (RPP): modeling the off-network emissions for parked vehicles, including the vehicle evaporative (fuel vapor venting). There are two sets of emission factors for the refueling parts of on-road sources: RPD and RPV. The emissions for non-road sources were estimated with the NONROAD model (NONROAD, 2008). Biogenic Emissions Biogenic emissions were estimated using the BEIS3.14 with land-use data based on the BELD3 database and the WRF-derived 2008 meteorological inputs for the 15-, 5- and 1-km grids Quality Assurance Procedures The emissions inventory processing quality assurance (QA) procedures included the preparation and examination of tabular emissions summaries and graphical display products. Tabular summaries were used to examine emissions totals for various steps of the emissions processing. Summaries for input emissions are based on the input inventory data: monthly emissions for the onroad and non-road sectors, and annual emissions for other sectors for criteria pollutants. Summaries for the emissions are based on the SMOKE output reports, which include daily emissions for each CB-05 species for each sector. The output daily emissions are summed over all days in the year and the CB-05 species are summed for the criteria pollutants. The emissions summaries were made for each scenario by state and sector, and comparisons were made between the input emissions and output emissions for each sector to assure consistency. In addition to the tabular summaries, various graphical displays were prepared for one day of each month to examine the spatial distribution and temporal variation for each sector and the final merged emissions using a graphical plotting package. ICF International 47 New York City Department of Health and Mental Hygiene

57 Emission Inventory Preparation 3.2 Emission Summaries Tables 3-1 through 3-3 summarize the 2008 base-year emissions used for the CMAQ modeling. These tables summarize anthropogenic emissions by source sector and pollutant for the 15-, 5-, and the 1-km grids. Emission totals are provided for the following species: volatile organic compounds (VOCs), oxides of nitrogen (NO x ), carbon monoxide (CO), sulfur dioxide (SO 2 ), fine particulate matter (PM 2.5 ), coarse particulate matter (PM 10 ), and ammonia (NH 3 ). The units are tons/year (tpy). Table 3-1. Emissions Totals (tons/year) by Source Sector for the 2008 Base Year for the 15-km Grid Sector VOC (tpy) NO x (tpy) CO (tpy) SO 2 (tpy) PM 10 (tpy) PM 2.5 (tpy) NH 3 (tpy) EGU 16,647 1,504, ,867 5,225, , ,068 6,378 Non-EGU 324, ,000 1,328, , , ,508 13,299 Nonpoint 1,908, ,953 2,743, , , , ,597 Nonroad 1,367,492 1,002,980 7,851,710 86,313 80,502 76, Onroad 1,189,693 2,603,137 13,831,060 13, , ,724 50,761 Total 4,807,075 6,210,478 25,927,814 6,264,019 1,349, , ,927 Table 3-2. Emissions Totals (tons/year) by Sector for the 2008 Base Year for the 5-km Grid Sector VOC (tpy) NO x (tpy) CO (tpy) SO 2 (tpy) PM 10 (tpy) PM 2.5 (tpy) NH 3 (tpy) EGU 2, ,069 39, ,258 34,374 30,369 2,607 Non-EGU 42, , ,174 74,770 26,593 20,205 3,408 Nonpoint 613, , , , ,543 90, ,541 Nonroad 402, ,424 2,545,112 35,192 23,287 21, Onroad 328, ,743 3,729,778 4,480 39,736 29,864 17,906 Total 1,388,718 1,462,072 7,060, , , , ,710 Table 3-3. Emissions Totals (tons/year) by Sector for the 2008 Base Year for the 1-km Grid Sector VOC (tpy) NO x (tpy) CO (tpy) SO 2 (tpy) PM 10 (tpy) PM 2.5 (tpy) NH 3 (tpy) EGU ,806 4,280 8,980 3,509 3, Non-EGU 6,925 42,364 21,084 20,181 3,267 3, Nonpoint 133,408 28,112 17,002 4,276 8,080 5,656 1,137 Nonroad 36,945 49, ,390 8,781 3,600 3, Onroad 34,242 69, , ,633 3,191 2,107 Total 211, , ,985 42,761 23,090 18,776 4,239 To illustrate and check the reasonableness of the spatial distribution of emissions throughout the modeling domain, daily emission density plots for selected days were prepared and examined. Three sets of daily low-level emission density plots for a representative summer day (July 15) for VOC, NO x, SO 2 ICF International 48 New York City Department of Health and Mental Hygiene

58 Emission Inventory Preparation and PM 2.5 are shown in Figures 3-2, 3-3, and 3-4 for the 15-, 5- and 1-km resolution grids, respectively. The date and time given on this and all subsequent figures refer to the meteorological base year (2008) and start hour for the selected day or averaging period. The minimum and maximum values for any location within the domain are also provided, along with their grid cell (x,y) locations. A summer day was selected for display because it is included in both the ozone season and the annual simulation period. Figure 3-2a. Daily VOC Emissions for 15 July 2008: 2008 Base Year, NYC 15-km Grid Figure 3-2b. Daily NO x Emissions for 15 July 2008: 2008 Base Year, NYC 15-km Grid ICF International 49 New York City Department of Health and Mental Hygiene

59 Emission Inventory Preparation Figure 3-2c. Daily SO 2 Emissions for 15 July 2008: 2008 Base Year, NYC 15-km Grid Figure 3-2d. Daily PM 2.5 Emissions for 15 July 2008: 2008 Base Year, NYC 15-km Grid ICF International 50 New York City Department of Health and Mental Hygiene

60 Emission Inventory Preparation Figure 3-3a. Daily VOC Emissions for 15 July 2008: 2008 Base Year, NYC 5-km Grid Figure 3-3b. Daily NO x Emissions for 15 July 2008: 2008 Base Year, NYC 5-km Grid ICF International 51 New York City Department of Health and Mental Hygiene

61 Emission Inventory Preparation Figure 3-3c. Daily SO 2 Emissions for 15 July 2008: 2008 Base Year, NYC 5-km Grid Figure 3-3d. Daily PM 2.5 Emissions for 15 July 2008: 2008 Base Year, NYC 5-km Grid ICF International 52 New York City Department of Health and Mental Hygiene

62 Emission Inventory Preparation Figure 3-4a. Daily VOC Emissions for 15 July 2008: 2008 Base Year, NYC 1-km Grid Figure 3-4b. Daily NO x Emissions for 15 July 2008: 2008 Base Year, NYC 1-km Grid. ICF International 53 New York City Department of Health and Mental Hygiene

63 Emission Inventory Preparation Figure 3-4c. Daily SO 2 Emissions for 15 July 2008: 2008 Base Year, NYC 1-km Grid Figure 3-4d. Daily PM 2.5 Emissions for 15 July 2008: 2008 Base Year, NYC 1-km Grid ICF International 54 New York City Department of Health and Mental Hygiene

64 Emission Inventory Preparation Heating Oil Scenarios As stated in Section 1, two scenarios were modeled to examine the effects of local changes in heating fuels. The heating oil control scenarios are as follows: Scenario #1: Partial implementation of the rule on heating oil, the emissions changes reflect the reduction in emissions achieved by the end of the winter heating season (April 2013). Scenario #2: Full implementation of the rule on heating oil (complete phase out of No. 4 and No. 6 heating oil). Both heating oil scenarios also included implementation of a 15 ppm sulfur limit to No. 2 heating oil. Table 3-5 provides the NYC boiler s emission summary for the 2008 base year and the two heating oil scenarios by fuel type. For residual boilers, the emissions of NO x, SO 2 and PM 2.5 are reduced 20, 35 and 24 percent from the 2008 level, respectively, in Scenario #1 reflecting the partial implementation of the rule on heating oil. The emissions of NO x, SO 2 and PM 2.5 are reduced by 69, 97 and 75 percent, respectively, in Scenario #2 reflecting the full implementation of the rule on heating oil. For distillate boilers, the emissions of SO 2 and PM 2.5 are reduced by 99 and 98 percent from the 2008 level, respectively, in both scenarios due to implementation of a 15 ppm sulfur limit to No. 2 heating oil. Overall, for Scenario #1, the emissions of NO x, SO 2 and PM 2.5 emissions associated with the New York City boilers are reduced by 8, 69 and 67 percent from the 2008 levels, respectively, and CO emissions are increased by 5 percent. For Scenario #2, NO x, SO 2 and PM 2.5 emissions are reduced by 26, 98 and 89 percent, respectively, and CO emissions are increased by 25 percent. Table 3-5. New York City Boiler Emission Summary for the 2008 Base Year and Heating Oil Scenarios Scenario NYC No. 4 & No. 6 Residual Boilers NYC No. 2 Distillate Boilers NYC Boilers Total NO x (tpy) CO (tpy) SO 2 (tpy) PM 2.5 (tpy) NO x (tpy) CO (tpy) SO 2 (tpy) PM 2.5 (tpy) NO x (tpy) CO (tpy) SO 2 (tpy) PM 2.5 (tpy) Base Year 9, , ,234 3,736 9,375 1,056 24,760 4,641 17,654 1,803 Scenario #1 7,600 1,134 5, ,234 3, ,835 4,870 5, Scenario #2 2,979 2, ,234 3, ,214 5, The emission changes between the 2008 base year and Scenario #1 are shown in the emission density difference plots for a representative summer day (July 15) in Figures 3-5a through 3-5c for NO x, SO 2 and PM 2.5 for the 1-km grid. The emission changes between Scenario #2 and Scenario #1 are shown in the emission density difference plots for a representative summer day (July 15) in Figures 3-6a through 3-6c for NO x, SO 2 and PM 2.5 for the 1-km grid. The figures show the spatial extent and magnitude of the ICF International 55 New York City Department of Health and Mental Hygiene

65 Emission Inventory Preparation emission reductions in the New York City area for Scenario #1 compared to the 2008 base, and further reductions for Scenario #2 reflecting the full implementation of the rules affecting heating oil boilers. Figure 3-5a. Difference in Daily NO x Emissions for 15 July: Scenario #1 Minus Base for the NYC 1-km Grid Figure 3-5b. Difference in Daily SO 2 Emissions for 15 July: Scenario #1 Minus Base for the NYC 1-km Grid ICF International 56 New York City Department of Health and Mental Hygiene

66 Emission Inventory Preparation Figure 3-5c. Difference in Daily PM 2.5 Emissions for 15 July: Scenario #1 Minus Base for the NYC 1-km Grid Figure 3-6a. Difference in Daily NO x Emissions for 15 July 2008: Scenario #2 Minus Scenario #1 for the NYC 1-km Grid ICF International 57 New York City Department of Health and Mental Hygiene

67 Emission Inventory Preparation Figure 3-6b. Difference in Daily SO 2 Emissions for 15 July: Scenario #2 Minus Scenario #1 for the NYC 1-km Grid Figure 3-6c. Difference in Daily PM 2.5 Emissions for 15 July: Scenario #2 Minus Scenario #1 for the NYC 1-km Grid ICF International 58 New York City Department of Health and Mental Hygiene

68 Emission Inventory Preparation EGU Fuel Scenarios As stated in Section 1, two additional scenarios were modeled to examine the effects of changes in EGU fuel between 2005 and the present day. A hybrid emission inventory was prepared consisting of the 2008 base-year emissions for all sectors except EGUs, and 2005 emissions for EGU sources. This alternative baseline scenario (EGU Base) provided a basis for assessing the effects of EGU emission changes for the following two EGU scenarios: Scenario 3: Adjustment of EGU emissions to reflect changes in fuel use at Title V EGUs outside of the five boroughs of New York City, i.e., replace the 2005 emissions outside of New York City boroughs with the present day emission estimates for the EGU sources. Scenario 4: Adjustment of EGU emissions to reflect changes in fuel use at EGUs located within the five boroughs, i.e., replace the 2005 emissions inside of New York City boroughs with the present day emission estimates for the EGU sources. The present day NO x and SO 2 emission estimates are based on the 2012 Continuous Emissions Monitoring (CEM) hour-specific and daily-specific EGU emission data obtained from the EPA s Clean Air Market reporting system (EPA, 2012b) for New York State and the other 16 states that are completely contained in the CMAQ 15-km modeling grid (States of New Jersey, Pennsylvania, Massachusetts, Connecticut, Delaware, Indiana, Ohio, Virginia, West Virginia, Michigan, Maine, Maryland, Rode Island, New Hampshire, Vermont and Kentucky). Since the emissions of VOC, CO, PM 10, PM 2.5 and NH 3 are not available in the 2012 CEM data, the present day EGU emissions for these pollutants are based on the 2014 estimates obtained from the EPA s 2005-based platform. The CEM data only include the EGU facility name, ORIS ID, Boiler ID and associated emissions, not the facility location and stack parameters that are needed for the emission inventory preparation. This information was obtained for the majority of the EGUs included in the CEM data by matching the facility name, ORIS ID and Boiler ID in the EPA s 2005-based platform through an automated process using a utility program. For those facilities for which matches could not be found using the automated approach, facility locations were obtained from the Internet (sometimes by converting the facility ZIP code to latitude and longitude) and SMOKE default stack parameters were assigned. For the five states located on the boundary of the NYC 15-km grid and only partially contained in the modeling domain (States of Wisconsin, Illinois, Missouri, Tennessee and North Carolina), the present day emission for all pollutants are based on the 2014 estimates obtained from the EPA s 2005-based platform. Tables 3-6a through 3-8c summarize the EGU emissions for the alternative EGU Base and the two EGU fuel scenarios for the 15-, 5- and 1-km grids. Separate summaries are provided for the areas inside and outside of the New York City boroughs along with the total emissions for each of the grids for easy comparison. The New York City boroughs are completely contained in the 1-km grid, and therefore the emission values are the same for the inside portion of the New York City boroughs for the 15-, 5- and 1- km grids, as provided in Tables 3-6b, 3-7b and 3-8b. ICF International 59 New York City Department of Health and Mental Hygiene

69 Emission Inventory Preparation Table 3-6a. EGU Emissions Summary for the EGU Base and EGU Fuel Scenarios: NYC 15-km Grid, Total Scenario VOC (tpy) NO x (tpy) CO (tpy) SO 2 (tpy) PM 10 (tpy) PM 2.5 (tpy) NH 3 (tpy) EGU Base 16,411 1,661, ,646 6,099, , ,705 6,836 Scenario 3 16, , ,691 1,851, , ,380 11,494 Scenario 4 16,441 1,652, ,743 6,087, , ,240 6,842 Table 3-6b. EGU Emissions Summary for the EGU Base and EGU Fuel Scenarios: NYC 15-km Grid, Inside of New York City Scenario VOC (tpy) NO x (tpy) CO (tpy) SO 2 (tpy) PM 10 (tpy) PM 2.5 (tpy) NH 3 (tpy) EGU Base ,571 4,642 12, Scenario ,571 4,642 12, Scenario ,663 3, Table 3-6c. EGU Emissions Summary for the EGU Base and EGU Fuel Scenarios: NYC 15-km Grid, Outside of New York City Scenario VOC (tpy) NO x (tpy) CO (tpy) SO 2 (tpy) PM 10 (tpy) PM 2.5 (tpy) NH 3 (tpy) EGU Base 16,283 1,646, ,004 6,087, , ,226 6,553 Scenario 3 16, , ,049 1,838, , ,901 11,211 Scenario 4 16,283 1,646, ,004 6,087, , ,226 6,553 Table 3-7a. EGU Emissions Summary for the EGU Base and EGU Fuel Scenarios: NYC 5-km Grid, Total Scenario VOC (tpy) NO x (tpy) CO (tpy) SO 2 (tpy) PM 10 (tpy) PM 2.5 (tpy) NH 3 (tpy) EGU Base 3, ,830 31, ,859 47,829 40,585 2,217 Scenario 3 2,217 86,048 61,813 88,406 19,978 16,601 3,342 Scenario 4 3, ,922 30, ,774 47,248 40,120 2,223 Table 3-7b. EGU Emissions Summary for the EGU Base and EGU Fuel Scenarios: NYC 5-km Grid, Inside of New York City Scenario VOC (tpy) NO x (tpy) CO (tpy) SO 2 (tpy) PM 10 (tpy) PM 2.5 (tpy) NH 3 (tpy) EGU Base ,571 4,642 12, Scenario ,571 4,642 12, Scenario ,663 3, ICF International 60 New York City Department of Health and Mental Hygiene

70 Emission Inventory Preparation Table 3-7c. EGU Emissions Summary for the EGU Base and EGU Fuel Scenarios: NYC 5-km Grid, Outside of New York City Scenario VOC (tpy) NO x (tpy) CO (tpy) SO 2 (tpy) PM 10 (tpy) PM 2.5 (tpy) NH 3 (tpy) EGU Base 3, ,259 26, ,442 47,229 40,107 1,934 Scenario 3 2,089 70,477 57,172 75,989 19,378 16,122 3,059 Scenario 4 3, ,259 26, ,442 47,229 40,107 1,934 Table 3-8a. EGU Emissions Summary for the EGU Base and EGU Fuel Scenarios: NYC 1-km Grid, Total Scenario VOC (tpy) NO x (tpy) CO (tpy) SO 2 (tpy) PM 10 (tpy) PM 2.5 (tpy) NH 3 (tpy) EGU Base 1,059 28,253 7,534 37,546 3,592 2, Scenario ,585 9,081 12,616 1,321 1, Scenario 4 1,089 18,345 6,631 25,461 3,011 2, Table 3-8b. EGU Emissions Summary for the EGU Base and EGU Fuel Scenarios: NYC 1-km Grid, Inside of New York City Scenario VOC (tpy) NO x (tpy) CO (tpy) SO 2 (tpy) PM 10 (tpy) PM 2.5 (tpy) NH 3 (tpy) EGU Base ,571 4,642 12, Scenario ,571 4,642 12, Scenario ,663 3, Table 3-8c. EGU Emissions Summary for the EGU Base and EGU Fuel Scenarios: NYC 1-km Grid, Outside of New York City Scenario VOC (tpy) NO x (tpy) CO (tpy) SO 2 (tpy) PM 10 (tpy) PM 2.5 (tpy) NH 3 (tpy) EGU Base ,682 2,892 25,130 2,992 2, Scenario ,014 4, Scenario ,682 2,892 25,130 2,992 2, For Scenario 3, for EGUs located outside of New York City boroughs, the emissions of all criteria pollutants except CO are lower for all three grids, compared to the EGU Base. The greatest decreases are for SO 2, followed by NO x, PM 2.5, and VOC. The SO 2 emissions are 70, 90 and 99 percent lower for the 15-, 5- and 1-km grids, respectively. The NO x emissions are 57, 63 and 84 percent lower. The PM 2.5 emissions are 11, 60 and 76 percent lower. The VOC emissions are 1, 36 and 84 percent lower. The CO emissions, on the other hand, are 72, 114 and 53 percent higher for Scenario 3 compared to the EGU Base for the 15-, 5- and 1-km grids, respectively. ICF International 61 New York City Department of Health and Mental Hygiene

71 Emission Inventory Preparation For Scenario 4, the changes occur in New York City and the EGU emissions for the five boroughs are lower by 64, 19, 97 and 97 percent for NO x, CO, SO 2 and PM 2.5, respectively, compared to the EGU Base. The VOC emissions are 23 percent higher. Table 3-9 provides the facility level emission comparisons between Scenario 4 and the EGU Base for the New York City borough EGUs. The facilities included in the Scenario 4 and EGU Base inventories are not completely the same because some of the facilities were decommissioned around 2005, and new plants opened up after 2005, although there are some slight discrepancies between the information provided in the EPA s data base and actual EGU operations. For example, according to the information provided by NYC DOHMH: Hell Gate and Harlem River Yard were both operational in 2005, but the emissions are not provided in the EPA s 2005-based platform. The Riverbay plant at Co-op City opened in 2008, and the Con-Ed Waterside plant was decommissioned around Although the Bronx Zoo source appears to have been operational since 2012, the emissions are not included in the CEM data, so present day emission estimates for this facility were obtained from the 2014 emissions provided in the EPA s based platform. There are also facilities included in the 2012 CEM data but not included in the EPA s 2014 estimates, and, therefore, there are no emissions estimates for VOC, CO, PM 10, PM 2.5 and NH 3 available for these few sources. Table 3-9. Facility Level EGU Emissions Comparison for the NYC 5 Boroughs: Scenario 4 vs EGU Base Case Scenario 4 EGU Base Plant Name VOC NO x SO 2 CO PM 2.5 VOC NO x SO 2 CO PM 2. (tpy) (tpy) (tpy) (tpy) (tpy) (tpy) (tpy) (tpy) ( tpy) (tpy) 23rd and 3rd th Street th Street Arthur Kill Astoria Energy , Astoria Gas Turbine Power , Astoria Generating Station , , Bayswater Peaking Facility Bronx Zoo Brooklyn Navy Yard Cogeneration Con-Ed Waterside Station East River Far Rockaway Gowanus Harlem River Yard Hell Gate Hudson Avenue KIAC Cogeneration Narrows North 1st Poletti 500 MW CC , , , Pouch Terminal Ravenswood Generating Station 0.8 1, , , Ravenswood Steam Plant Riverbay Corp. Co-Op City Vernon Boulevard Total 158 5,663 3, ,571 4,642 12, ICF International 62 New York City Department of Health and Mental Hygiene

72 Emission Inventory Preparation The emission changes between Scenario 3 and the EGU baseline are shown in the emission density difference plots for a representative summer day (July 15) in Figures 3-7a through 3-7c for NO x, SO 2 and PM 2.5 for the 1-km grid. The emission changes between Scenario 4 and the EGU baseline are shown in the emission density difference plots for a representative summer day (July 15) in Figures 3-8a through 3-8c for NOx, SO 2 and PM 2.5 for the 1-km grid. Note that although the overall emissions are lower than the EGU Base for both scenarios, there are also some sources for which the emissions are higher (primarily due to the use of a different emission datasets between the EGU Base and the scenarios). This will be discussed further in the analysis of the results (Section 6). Figure 3-7a. Difference in Daily NO x Emissions for 15 July: Scenario #3 Minus Base for the NYC 1-km Grid ICF International 63 New York City Department of Health and Mental Hygiene

73 Emission Inventory Preparation Figure 3-7b. Difference in Daily SO 2 Emissions for 15 July: Scenario #3 Minus Base for the NYC 1-km Grid Figure 3-7c. Difference in Daily PM 2.5 Emissions for 15 July: Scenario #3 Minus Base for the NYC 1-km Grid ICF International 64 New York City Department of Health and Mental Hygiene

74 Emission Inventory Preparation Figure 3-8a. Difference in Daily NO x Emissions for 15 July 2008: Scenario #4 Minus Scenario #1 for the NYC 1-km Grid Figure 3-8b. Difference in Daily SO 2 Emissions for 15 July: Scenario #4 Minus Scenario #1 for the NYC 1-km Grid ICF International 65 New York City Department of Health and Mental Hygiene

75 Emission Inventory Preparation Figure 3-8c. Difference in Daily PM 2.5 Emissions for 15 July: Scenario #4 Minus Scenario #1 for the NYC 1-km Grid ICF International 66 New York City Department of Health and Mental Hygiene

76 4 Application and Evaluation of CMAQ for the Base Year 4.1 Overview of the CMAQ Modeling System Version 5.0 of the CMAQ model was used for the NYC high-resolution modeling analysis. The CMAQ model is a state-of-the-science, regional air quality modeling system that can be used to simulate the physical and chemical processes that govern the formation, transport, and deposition of gaseous and particulate species in the atmosphere (Byun and Ching 1999). The CMAQ tool was designed to improve the understanding of air quality issues (including the physical and chemical processes that influence air quality) and to support the development of effective emission control strategies on both the regional and local scales. The CMAQ model was designed as a one-atmosphere model. This concept refers to the ability of the model to dynamically simulate ozone, particulate matter, and other species (such as mercury) in a single simulation. In addition to addressing a variety of pollutants, CMAQ can be applied to a variety of regions (with varying geographical, land-use, and emissions characteristics) and for a range of space and time scales. The latest version of CMAQ includes state-of-the-science advection, dispersion and deposition algorithms, the latest version of the Carbon Bond (CB) chemical mechanism (CB05), and diagnostic tools for assessing source apportionment. Numerous recent applications of the model, for both research and regulatory air quality planning purposes, have focused on the simulation of ozone and fine particulate matter (PM 2.5 ). The CMAQ model was used by EPA to support the development of the Clean Air Interstate Rule (CAIR) (EPA, 2005). It was also used by EPA to support the second prospective analysis of the costs and benefits of the Clean Air Act (CAA) (Douglas et al., 2008). The CMAQ model numerically simulates the physical processes that determine the magnitude, temporal variation, and spatial distribution of the concentrations of ozone and particulate species in the atmosphere and the amount, timing, and distribution of their deposition to the earth s surface. The simulation processes include advection, dispersion (or turbulent mixing), chemical transformation, cloud processes, and wet and dry deposition. The CMAQ science algorithms are described in detail by Byun and Ching (1999). The CMAQ model requires several different types of input files. Gridded, hourly emission inventories characterize the release of anthropogenic, biogenic, and, in some cases, geogenic emissions from sources within the modeling domain. The emissions represent both low-level and elevated sources and a variety of source categories (including, for example, point, on-road mobile, non-road mobile, area, and biogenic). The amount and spatial and temporal distribution of each emitted pollutant or precursor species are key determinants to the resultant simulated air quality values. The CMAQ model also requires hourly, gridded input fields of several meteorological parameters including wind, temperature, mixing ratio, pressure, solar radiation, fractional cloud cover, cloud depth, and precipitation. A full list of the meteorological input parameters is provided in Byun and Ching (1999). The meteorological input fields are typically prepared using a data-assimilating prognostic meteorological model, the output of which is processed for input to the CMAQ model using the Meteorology-Chemistry Interface Processor (MCIP). The prescribed meteorological conditions influence ICF International 67 New York City Department of Health and Mental Hygiene

77 Application and Evaluation of CMAQ for the Base Year the transport, vertical mixing, and resulting distribution of the simulated pollutant concentrations. Certain of the meteorological parameters, such as mixing ratio, can also influence the simulated chemical reaction rates. Rainfall and near-surface meteorological characteristics govern the wet and dry deposition, respectively, of the simulated atmospheric constituents. Initial and boundary condition (IC/BC) files provide information on pollutant concentrations throughout the domain for the first hour of the first day of the 10-day spin-up period for the simulation, and along the lateral boundaries of the domain for each hour of the simulation. Photolysis rates and other chemistryrelated input files supply information needed by the gas-phase and particulate chemistry algorithms. 4.2 CMAQ Application Procedures In running the model, the simulation period was divided into two parts covering January through June and July through December, respectively. Each part of the simulation also included an additional 10 start-up simulation days, which were intended to reduce the influence of uncertainties in the initial conditions on the simulation results. Preparation of the meteorological and emissions inputs was discussed in detail in Sections 2 and 3, respectively. Several other input files are required for the application of CMAQ. Initial and boundary conditions (IC/BC) files provide information on pollutant concentrations throughout the domain for the first hour of the first day of the simulation, and along the lateral boundaries of the domain (each grid) for each hour of the simulation. Gridded land-use information is required for the calculation of deposition and is used by other physical and numerical process algorithms. Photolysis rates and other chemistry related input files supply information needed by the gas-phase and particulate chemistry algorithms. For this analysis, boundary conditions for the 15-km domain were derived from a national-scale simulation for 2008 that was run using a 36-km resolution continental U.S. (CONUS) modeling domain. Boundary conditions for the inner grids were generated as part of the CMAQ application and derived from the modeling results for the next larger outer grid within which the inner grid is nested. Land use, photolysis rates, and other chemistry related inputs were prepared using standard CMAQ procedures and pre-processing programs. 4.3 Model Performance Evaluation Methodology An integral component of all modeling studies is the evaluation of model performance for the base-year or (base-case) simulation. For this study, diagnostic testing was used to ensure that the modeling system exhibits a reasonable response to changes in the inputs (and that the inputs do not contain significant biases or compensating errors). In addition, the CMAQ modeling results were compared with observed data, using a variety of graphical and statistical analysis products Air Quality Data Air quality data for the evaluation of model performance were obtained from EPA s Air Quality System (AQS) database. Ozone, NO 2, SO 2, CO, PM 10 and PM 2.5 data for all monitoring sites within the 15-, 5- and ICF International 68 New York City Department of Health and Mental Hygiene

78 Application and Evaluation of CMAQ for the Base Year 1-km grids were obtained and processed for use by AMET. Statistics were calculated using hourly concentrations, daily maximum 1-hour concentrations, daily maximum 8-hour average concentrations, and 24-hour average concentrations. For consistency with the NAAQS, the evaluation for ozone focuses on model performance for daily maximum 8-hour average ozone and the evaluation for PM 2.5 focuses on 24-hour average PM Graphical and Statistical Analysis The overall objective of a model performance evaluation is to establish that the modeling system can be used reliably to predict the effects of changes in emissions on future-year air quality. This was primarily accomplished by comparing the modeling results with observed data, using a variety of graphical and statistical analysis products. In addition, the ability of the model to simulate month-to-month differences in concentration levels and patterns, simulate the concentrations (or at least the frequency distribution of concentrations) associated with different types of meteorological conditions; and perform consistently and reasonably across a range of concentrations were also examined. The evaluation focused on the 5- and 1-km resolution grids. Analysis of results for the 15-km resolution grid emphasized representation of the regional-scale concentration levels and patterns, as well as month-to-month variations in regional-scale ozone air quality. A more detailed analysis of the results was performed for the 5- and 1-km grids. This included analysis of the magnitude and timing of sitespecific concentrations and calculation of model performance statistics. For extraction of the model output and matching with the station values, concentration information was taken from the grid cell in which the monitoring site is located. Graphical Analysis to Support Model Performance Evaluation AMET generates a wide variety of graphical analysis products to facilitate the evaluation of CMAQ model performance. Plots and graphics were used to assess the reasonableness of the results. The graphical analysis included the following: Spatial plots of the simulated concentrations and bias and error values were used to assess the ability of the model to emulate the spatial concentration patterns. Scatter plots were used to graphically compare the simulated and observed concentration. Statistical Analysis to Support Model Performance Evaluation AMET also calculates a variety of statistical measures to facilitate the evaluation of CMAQ model performance. Table 4-1 summarizes key statistical measures that were used to provide insight into model performance. ICF International 69 New York City Department of Health and Mental Hygiene

79 Application and Evaluation of CMAQ for the Base Year Table 4-1. Statistical Measures Used for the CMAQ Model Performance Evaluation for High Resolution Air Quality Modeling of New York City Metric Definition # of data pairs The number of observation/simulation data pairs Mean observation value Mean simulation value Mean bias The average observed concentration The average simulated concentration N 1 ( S l O l ) N l 1 Normalized bias ( S O ) O 100% where N is the number of data pairs, and S l and O l are the simulated and observed values at site l, respectively, over a given time interval. 1 N Normalized mean bias ( S O ) O 100% N l 1 1 N Fractional bias ( S O ) 0.5( S O ) 100% Mean error 1 N l 1 l N l 1 N N l 1 1 N l l l l N l 1 l S l O l Normalized error S O O 100% Normalized mean error S O O 100% N N l 1 l l l 1 l 1 1 N Fractional error S O 0.5( S O ) 100% N l 1 2 Correlation N SO S O N S S Index of agreement l l N l l l l l l l l l l N O O A measure of how well the model represents the pattern of perturbation about the mean value; ranges from 0 to 1. In calculating the statistical measures, AMET pairs the CMAQ model output with the observed data for the appropriate locations and time intervals. ICF International 70 New York City Department of Health and Mental Hygiene

80 Application and Evaluation of CMAQ for the Base Year Model Performance Criteria In keeping with current EPA guidance on model performance evaluation for ozone, a weight-ofevidence approach was employed to determine whether model performance is good enough for use in future-year modeling and air quality assessment. In other words, an integrated assessment of the above information was used to document and qualitatively and quantitatively assess whether an acceptable base-case simulation is achieved. To the extent practicable, the statistical measures for certain of the pollutants were compared with model performance goals and criteria used for prior studies, as suggested in EPA guidance (EPA, 2007). For ozone, these include recommended ranges for the normalized bias and normalized error from prior (ca. 1990) EPA guidance (these are still widely used for urban- and regional-scale model performance evaluation). 4.4 CMAQ Modeling Results Summary of Model Performance for Ozone CMAQ model performance for ozone focused on the typical ozone season months of April through October and is summarized in the remainder of this section. 15-km Grid Spatial Concentration Patterns Spatial plots of the simulated ozone concentration patterns for the 15-km grid for selected days throughout the simulation period were plotted and examined. Figure 4-1 illustrates the simulated ozone concentration patterns for the 15 th of each month (April October). Consistent with the National Ambient Air Quality Standard (NAAQS) for ozone, daily maximum 8-hour average ozone concentration is displayed. The units are parts per billion (ppb). ICF International 71 New York City Department of Health and Mental Hygiene

81 Application and Evaluation of CMAQ for the Base Year Figure 4-1. Simulated Daily Maximum 8-Hour Ozone Concentration (ppb) for Selected Days for the CMAQ 15-km Grid April 15th/May 15 th June 15th/July 15th ICF International 72 New York City Department of Health and Mental Hygiene

82 Application and Evaluation of CMAQ for the Base Year August 15th/September 15th October 15th For many of the selected days, the simulated ozone concentration patterns show moderate to high 8- hour average ozone concentrations along the mid-atlantic coast, including New York. As expected, high ozone concentrations are more widespread during the summer months and the highest ozone concentrations occur in June, July and August. Comparison of Simulated and Observed Concentrations A scatter plot comparing simulated and observed daily maximum 8-hour ozone concentrations for the 15-km grid for April through October is presented in Figure 4-2. The scatter plot provides a visual representation of how well the simulated values match the observations, and can reveal biases toward over- or underestimation of the observed values. Also included on the scatter plot is some statistical ICF International 73 New York City Department of Health and Mental Hygiene

83 Application and Evaluation of CMAQ for the Base Year information further summarizing model performance. Note that these statistical measures are calculated using the 8-hour average ozone concentrations. Figure 4-2. Comparison of Simulated and Observed Daily Maximum 8-Hour Average Ozone Concentration (ppb) for the 15-km Grid (April through October) There is a general tendency for CMAQ to overestimate ozone concentrations (especially those greater than 40 ppb). However, there is good correlation overall as indicated by an index of agreement of Statistical Measures of Model Performance Summary metrics and statistical measures calculated using hourly ozone concentrations for the 15-km grid are presented in Table 4-2. The recommended ranges for the normalized bias and normalized error shown in this table are no longer a part of current EPA guidance but are still widely used for urban- and regional-scale model performance evaluations (EPA, 2007). A lower bound of 40 ppb was used in calculating the normalized bias and error statistics (only observation-simulation pairs for which the observed value is greater than or equal to 40 ppb were included in the calculations). ICF International 74 New York City Department of Health and Mental Hygiene

84 Application and Evaluation of CMAQ for the Base Year Table 4-2. Summary Model Performance Statistics for Ozone for the 15-km Modeling Grid Metric Apr May Jun Jul Aug Sep Oct Apr -Oct Goal Number of Data Pairs 92,607 94,514 92,826 93,863 65,557 39,586 14, ,250 Mean Observed (ppb) Mean Simulated (ppb) Mean Bias (ppb) Normalized Bias (%) ± 15 Normalized Mean Bias (%) Fractional Bias (%) Mean Error (ppb) Normalized Error (%) Normalized Mean Error (%) Fractional Error (%) Correlation (unitless) Index of Agreement (unitless) The statistical measures indicate underestimation of ozone for April and May, overestimation for the remaining months, and very good agreement, on average, between the simulated and observed concentrations for all months. The normalized bias is within ±15 percent and the normalized error is well within 35 percent for all months. Using a lower bound value of 60 ppb, the normalized mean bias for the multi-month period (April October) is -4.8 percent and the normalized mean error is 15.8 percent. Ozone Model Performance for the 5-km Grid Spatial Concentration Patterns Spatial plots of the simulated ozone concentration patterns for the 5-km grid for selected days throughout the simulation period were plotted and examined. Figure 4-3 illustrates the daily maximum 8-hour average ozone concentration patterns for the 15 th of each month (April October). Units are parts per billion (ppb). ICF International 75 New York City Department of Health and Mental Hygiene

85 Application and Evaluation of CMAQ for the Base Year Figure 4-3. Simulated Daily Maximum 8-Hour Ozone Concentration (ppb) for Selected Days for the CMAQ 5-km Grid April 15th/May 15 th June 15th/July 15th ICF International 76 New York City Department of Health and Mental Hygiene

86 Application and Evaluation of CMAQ for the Base Year August 15th/September 15th October 15th For many of the selected days, the simulated ozone concentration patterns indicate relatively low ozone concentrations over the greater New York City area likely due to high NO x emissions in the area, which can suppress local ozone formation. Among the selected days, the highest concentrations occur over New York City on June 15 th and August 15 th. On these days, the simulated daily maximum 8-hour ozone concentration is in the range of 80 to 90 ppb. Figure 4-4 depicts the average bias and error for all sites in the 5-km modeling domain, based on daily maximum 8-hour ozone concentrations for the ozone season months (April through October). For the normalized bias, gray shaded circles indicate that the bias is within ± 15 percent; blue and green shading indicates underestimation of the observed concentrations and yellow, orange, and red shading indicates overestimation. For the normalized mean error, blue and green shading represent the smaller errors, ICF International 77 New York City Department of Health and Mental Hygiene

87 Application and Evaluation of CMAQ for the Base Year while red indicates an error greater than 35 percent. A lower bound of 40 ppb was used in calculating the normalized bias and error statistics. Note that the plotted area is slightly larger than the 5-km grid, but that information is presented only for sites within the 5-km grid. Figure 4-4. Normalized Bias (%) and Normalized Mean Error (%) Based on Daily Maximum 8-Hour Average Simulated and Observed Ozone Concentrations for April through October for the CMAQ 5-km Grid Normalized Bias/Normalized Mean Error Model performance is consistent throughout the 5-km grid and no distinct spatial patterns emerge. For most monitoring sites, the normalized bias is within ± 15 percent (as indicated by the gray shading). The normalized mean error is less than 35 percent for all sites. Comparison of Simulated and Observed Concentrations A scatter plot comparing simulated and observed daily maximum 8-hour ozone concentrations for the 5- km grid for April through October is presented in Figure 4-5. Again, note that the statistical measures given on the plot are calculated using the 8-hour average ozone concentrations. ICF International 78 New York City Department of Health and Mental Hygiene

88 Application and Evaluation of CMAQ for the Base Year Figure 4-5. Comparison of Simulated and Observed Daily Maximum 8-Hour Average Ozone Concentration (ppb) for the 5-km Grid (April through October) There is a general tendency for CMAQ to overestimate 8-hour average ozone concentrations and good correlation overall as indicated by an index of agreement of Statistical Measures of Model Performance Summary metrics and statistical measures calculated using hourly ozone concentrations for the 5-km grid are presented in Table 4-3. A lower bound of 40 ppb was used in calculating the normalized bias and error statistics. ICF International 79 New York City Department of Health and Mental Hygiene

89 Application and Evaluation of CMAQ for the Base Year Table 4-3. Summary Model Performance Statistics for Ozone for the 5-km Modeling Grid Metric Apr May Jun Jul Aug Sep Oct Apr -Oct Goal Number of Data Pairs 25,756 24,808 26,219 26,225 13,847 6,978 1, ,276 Mean Observed (ppb) Mean Simulated (ppb) Mean Bias (ppb) Normalized Bias (%) ± 15 Normalized Mean Bias (%) Fractional Bias (%) Mean Error (ppb) Normalized Error (%) Normalized Mean Error (%) Fractional Error (%) Correlation (unitless) Index of Agreement (unitless) The statistical measures for the 5-km grid also show underestimation of ozone for April and May, and overestimation of ozone for the remaining months. The normalized bias is within ±15 percent and the normalized error is well within 35 percent for all months and for the ozone season. Using a lower-bound value of 60 ppb, the normalized mean bias for the multi-month period (April October) is -2.3 percent and the normalized mean error is 15.4 percent and are also well within the model performance goals. 1-km Grid Spatial Concentration Patterns Spatial plots of the simulated ozone concentration patterns for the 1-km grid for selected days throughout the simulation period were plotted and examined. Figure 4-6 illustrates the daily maximum 8-hour average ozone concentration patterns for the 15 th of each month (April October). Units are parts per billion (ppb). ICF International 80 New York City Department of Health and Mental Hygiene

90 Application and Evaluation of CMAQ for the Base Year Figure 4-6. Simulated Daily Maximum 8-Hour Ozone Concentration (ppb) for Selected Days for the CMAQ 1-km Grid April 15th/May 15 th June 15th/July 15th ICF International 81 New York City Department of Health and Mental Hygiene

91 Application and Evaluation of CMAQ for the Base Year August 15th/September 15th October 15th The simulated ozone concentration patterns for the 1-km grid consistently show lower ozone concentrations over Manhattan and the Bronx and higher concentrations over Staten Island, Brooklyn, and Queens. The relatively low ozone concentrations over Manhattan are likely due to high NO x emissions in the area (primarily from traffic). A high NO x to VOC ratio can suppress local ozone formation. Among the selected days, the highest concentrations occur over New York City on June 15 th and August 15 th. On these days, the simulated daily maximum 8-hour ozone concentration is in the range of 80 to 90 ppb. ICF International 82 New York City Department of Health and Mental Hygiene

92 Application and Evaluation of CMAQ for the Base Year Comparison of Simulated and Observed Concentrations A scatter plot comparing simulated and observed daily maximum 8-hour ozone concentrations for the 1- km grid for April through October is presented in Figure 4-8. Again, note that the statistical measures given on the plot are calculated using the 8-hour average ozone concentrations. Figure 4-8. Comparison of Simulated and Observed Daily Maximum 8-Hour Average Ozone Concentration (ppb) for the 1-km Grid (April through October) There is a general tendency for CMAQ to overestimate the 8-hour average ozone concentrations and good correlation overall as indicated by an index of agreement of Statistical Measures of Model Performance Summary metrics and statistical measures calculated using hourly ozone concentrations for the 1-km grid are presented in Table 4-4. A lower bound of 40 ppb was used in calculating the normalized bias and error statistics. ICF International 83 New York City Department of Health and Mental Hygiene

93 Application and Evaluation of CMAQ for the Base Year Table 4-4. Summary Model Performance Statistics for Ozone for the 1-km Modeling Grid Metric Apr May Jun Jul Aug Sep Oct Number of Data Pairs , ,037 Apr - Oct Goal Mean Observed (ppb) Mean Simulated (ppb) Mean Bias (ppb) Normalized Bias (%) ± 15 Normalized Mean Bias (%) Fractional Bias (%) Mean Error (ppb) Normalized Error (%) Normalized Mean Error (%) Fractional Error (%) Correlation (unitless) Index of Agreement (unitless) The statistical measures for the 1-km grid vary more from month-to-month than those for the 5- and 15-km grids. This is in part due to fewer observed values greater than 40 ppb (as is especially apparent for October). The errors are generally comparable, but in a few cases greater (for example, for April and October). The normalized bias is within ±15 percent for all months but April, and the normalized error is within 35 percent for all months. Using a lower-bound value of 60 ppb, the normalized mean bias for the multi-month period (April October) is -2.3 percent and the normalized mean error is 16.6 percent, both well within the model performance goals and indicative of good model performance, on average, for the higher ozone values Summary of Model Performance for PM km Grid Spatial Concentration Patterns Spatial plots of the monthly average simulated PM 2.5 concentration patterns for the 15-km grid are illustrated in Figure 4-9. The units are micrograms per cubic meter (µg/m 3 ). ICF International 84 New York City Department of Health and Mental Hygiene

94 Application and Evaluation of CMAQ for the Base Year Figure 4-9. Simulated Monthly Average PM 2.5 Concentration (µg/m 3 ) for the CMAQ 15-km Grid January/February March/April ICF International 85 New York City Department of Health and Mental Hygiene

95 Application and Evaluation of CMAQ for the Base Year May/June July/August ICF International 86 New York City Department of Health and Mental Hygiene

96 Application and Evaluation of CMAQ for the Base Year September/October November/December The simulated monthly average PM 2.5 concentrations are generally less than 12 µg/m 3. The concentration patterns reveal localized areas of higher PM 2.5, including over the major urban areas of Canada. For the New York City area, the simulated PM 2.5 concentrations are highest during the winter months. Figure 4-10 displays the annual average simulated PM 2.5 concentration pattern for the 15-km grid. ICF International 87 New York City Department of Health and Mental Hygiene

97 Application and Evaluation of CMAQ for the Base Year Figure Simulated Annual Average PM 2.5 Concentration (µg/m 3 ) for the CMAQ 15-km Grid The plots reveal regional-scale concentrations that range from about 4 to 16 µg/m 3 across most of the 15-km grid, with localized areas of PM 2.5 concentrations greater than 16 µg/m 3. The maximum simulated annual average PM 2.5 concentration is 81 µg/m 3, and is located near Montreal, Canada. Comparison of Simulated and Observed Concentrations Scatter plots comparing simulated and observed 24-hour PM 2.5 concentrations for AQS sites within the 15-km grid for the annual simulation period are presented in Figure ICF International 88 New York City Department of Health and Mental Hygiene

98 Application and Evaluation of CMAQ for the Base Year Figure Comparison of Simulated and Observed 24-Hour Average PM 2.5 Concentration (µg/m 3 ) for the 15- km Grid (All Months) The scatter plot shows a good deal of under- and overestimation of the observed PM 2.5 concentrations. There is an overall tendency for the model to overestimate observed concentrations, although some of the higher concentrations are underestimated. Statistical Measures of Model Performance Summary metrics and statistical measures calculated using 24-hr PM 2.5 concentrations for the 15-km grid are presented in Table 4-5. The recommended ranges for the fractional bias and fractional error are based on Boylan (2005) and are widely used for regional-scale model performance evaluation for PM 2.5. No lower bound was applied in calculating the statistics. ICF International 89 New York City Department of Health and Mental Hygiene

99 Application and Evaluation of CMAQ for the Base Year Table 4-5. Summary Model Performance Statistics for PM 2.5 for the 15-km Modeling Grid Metric Jan Mar Apr Jun Jul Sep Oct Dec Annual Goal Number of Data Pairs 12,662 12,396 12, ,266 Mean Observed (µg/m 3 ) Mean Simulated (µg/m 3 ) Mean Bias (µg/m 3 ) Fractional Bias (%) ± 60 Mean Error (µg/m 3 ) Fractional Error (%) Correlation (unitless) Index of Agreement (unitless) On average, PM 2.5 concentrations at the AQS monitors (Table 4-5) are overestimated for three of four quarterly periods and slightly underestimated for the third quarter (July September). The lowest bias and error values and thus the best model performance are achieved for the third quarter, when observed PM 2.5 concentrations are relatively high. The statistical measures for fractional bias and fractional error are well within the model performance goals for all periods. 5-km Grid Spatial Concentration Patterns Spatial plots of the monthly average simulated PM 2.5 concentration patterns for the 5-km grid are illustrated in Figure The units are micrograms per cubic meter (µg/m 3 ). ICF International 90 New York City Department of Health and Mental Hygiene

100 Application and Evaluation of CMAQ for the Base Year Figure Simulated Monthly Average PM 2.5 Concentration (µg/m 3 ) for the CMAQ 5-km Grid January/February March/April ICF International 91 New York City Department of Health and Mental Hygiene

101 Application and Evaluation of CMAQ for the Base Year May/June July/August ICF International 92 New York City Department of Health and Mental Hygiene

102 Application and Evaluation of CMAQ for the Base Year September/October November/December The highest simulated PM 2.5 concentrations (12 µg/m 3 and above) occur over the New York City area, central New Jersey, and southeastern Pennsylvania, with peak concentrations near or greater than 40 µg/m 3 for several of the days. The concentration patterns reveal localized areas of higher PM 2.5, including over the major urban areas of Canada. For the New York City area, the simulated PM 2.5 concentrations are highest during the winter months. Figure 4-13 displays the annual average simulated PM 2.5 concentration pattern for the 5-km grid. ICF International 93 New York City Department of Health and Mental Hygiene

103 Application and Evaluation of CMAQ for the Base Year Figure Simulated Annual Average PM 2.5 Concentration (µg/m 3 ) for the CMAQ 5-km Grid The simulated annual average PM 2.5 concentrations are less than 35 µg/m 3 throughout the 5-km grid. The highest concentrations occur over New York City. Because the observed PM 2.5 concentrations can be quite small and there is no accepted minimum threshold, fractional bias and error are better suited to characterizing model performance. To illustrate the agreement between the simulated and observed values, Figure 4-14 depicts the fractional bias and fractional error statistics for the 5-km modeling domain. The statistics are calculated using 24-hour average PM 2.5 concentrations and are calculated using data for the annual simulation period. Each monitoring site is represented by a circle and the shading of the circle provides information about how well the 24-hour observed PM 2.5 concentrations are represented by the simulation results, on average. For the fractional bias, gray shaded circles indicate that the fraction bias is within ± 20 percent and, in general, values within ±60 percent (lighter colors) correspond to acceptable model performance. Blue and green shading indicates underestimation of the observed concentrations and yellow, orange, and red shading indicates overestimation. For the fractional error, blue and green shading represent the smaller errors, while red indicates an error greater than 100 percent. Values less than 75 percent are considered to represent reasonable model performance for PM 2.5. ICF International 94 New York City Department of Health and Mental Hygiene

104 Application and Evaluation of CMAQ for the Base Year Figure Fractional Bias (%) and Fractional Error (%) Based on 24-Hour Average Simulated and Observed PM 2.5 Concentrations for CMAQ 5-km Grid (All Months) Fractional Bias/Fractional Error For most monitoring sites, the fractional bias is within the range of -40 to 60 percent (as indicated by the green, gray, yellow and orange shading). However, there are several sites, mostly in Pennsylvania, where the bias (overestimation) is greater. The greatest fractional errors also occur over Pennsylvania. There is clearly a regional pattern to the model performance statistics, and the poorer performance over Pennsylvania could be related to the inconsistencies in the quality of the emissions or meteorological inputs for that area or uncertainties in the boundary conditions (that would tend to affect Pennsylvania more than the greater NY area). Some of the best performance is in the NY/NJ/CT tri-state area. Comparison of Simulated and Observed Concentrations Scatter plots comparing simulated and observed 24-hour PM 2.5 concentrations for AQS sites within the 5-km grid for the annual simulation period are presented in Figure ICF International 95 New York City Department of Health and Mental Hygiene

105 Application and Evaluation of CMAQ for the Base Year Figure Comparison of Simulated and Observed 24-Hour Average PM 2.5 Concentration (µg/m 3 ) for the 5-km Grid (All Months) The scatter plot shows both under- and overestimation of the observed PM 2.5 concentrations and a tendency for overestimation. Statistical Measures of Model Performance Summary metrics and statistical measures calculated using 24-hr PM 2.5 concentrations for the 5-km grid are presented in Tables 4-6. The recommended ranges for the fractional bias and fractional error are based on Boylan (2005) and are widely used for regional-scale model performance evaluation for PM 2.5. No lower bound was applied in calculating the statistics. ICF International 96 New York City Department of Health and Mental Hygiene

106 Application and Evaluation of CMAQ for the Base Year Table 4-6. Summary Model Performance Statistics for PM 2.5 for the 5-km Modeling Grid Metric Jan Mar Apr Jun Jul Sep Oct Dec Annual Goal Number of Data Pairs 4,012 3,851 4,099 4,068 32,282 Mean Observed (µg/m 3 ) Mean Simulated (µg/m 3 ) Mean Bias (µg/m 3 ) Fractional Bias (%) ± 60 Mean Error (µg/m 3 ) Fractional Error (%) Correlation (unitless) Index of Agreement (unitless) On average, PM 2.5 concentrations are overestimated for the 5-km grid, especially during the first and fourth quarters. The fractional bias and error values are within the model performance goals for all periods. 1-km Grid Spatial Concentration Patterns Spatial plots of the monthly average simulated PM 2.5 concentration patterns for the 1-km grid are illustrated in Figure The units are micrograms per cubic meter (µg/m 3 ). Figure Simulated Monthly Average PM 2.5 Concentration (µg/m 3 ) for the CMAQ 1-km Grid January/February ICF International 97 New York City Department of Health and Mental Hygiene

107 Application and Evaluation of CMAQ for the Base Year March/April May/June ICF International 98 New York City Department of Health and Mental Hygiene

108 Application and Evaluation of CMAQ for the Base Year July/August September/October ICF International 99 New York City Department of Health and Mental Hygiene

109 Application and Evaluation of CMAQ for the Base Year November/December The simulated monthly average PM 2.5 concentrations are generally highest over Manhattan and during the winter months. The peak values range from 25 µg/m 3 for August to 57 µg/m 3 for February. Figure 4-17 displays the annual average simulated PM 2.5 concentration pattern for the 1-km grid. Figure Simulated Annual Average PM 2.5 Concentration (µg/m 3 ) for the CMAQ 1-km Grid The concentration patterns reveal regional scale concentrations that range from about 12 to 24 µg/m 3 across most of the overland portion of the 1-km grid, with lower concentrations offshore. There are a few grid cells over south central Manhattan with PM 2.5 concentrations greater than 24 µg/m 3. The maximum simulated annual average PM 2.5 concentration is 37 µg/m 3. ICF International 100 New York City Department of Health and Mental Hygiene

110 Application and Evaluation of CMAQ for the Base Year Comparison of Simulated and Observed Concentrations Scatter plots comparing simulated and observed 24-hour PM 2.5 concentrations for AQS sites within the 1-km grid for the annual simulation period are presented in Figure Figure Comparison of Simulated and Observed 24-Hour Average PM 2.5 Concentration (µg/m 3 ) for the 1-km Grid (All Months) The scatter plot reveals a tendency for overestimation of the observed PM 2.5 concentrations, especially concentrations greater than approximately 40 µg/m 3. Statistical Measures of Model Performance Summary metrics and statistical measures calculated using 24-hr PM 2.5 concentrations for the 1-km grid are presented in Table 4-7. The recommended ranges for the fractional bias and fractional error are based on Boylan (2005) and are widely used for regional-scale model performance evaluation for PM 2.5. No lower bound was applied in calculating the statistics. ICF International 101 New York City Department of Health and Mental Hygiene

111 Application and Evaluation of CMAQ for the Base Year Table 4-7. Summary Model Performance Statistics for PM 2.5 for the 1-km Modeling Grid Metric Jan Mar Apr Jun Jul Sep Oct Dec Annual Goal Number of Data Pairs ,707 Mean Observed (µg/m 3 ) Mean Simulated (µg/m 3 ) Mean Bias (µg/m 3 ) Fractional Bias (%) ± 60 Mean Error (µg/m 3 ) Fractional Error (%) Correlation (unitless) Index of Agreement (unitless) On average, PM 2.5 concentrations are reasonably well represented for the warmer months (second and third quarters) and overestimated by quite a lot for the remainder of the annual period. The fractional bias and error values are within the model performance goals for all periods Summary of Model Performance for PM 10, NO x, SO 2 and CO Model performance for PM 10, NO x, SO 2 and CO was examined with emphasis on quarterly and annual average concentrations. Observed concentrations of these criteria pollutants are generally expected to represent local rather than regional scale concentrations. This is due to the fact that these pollutants are directly emitted into the atmosphere and also because the monitoring sites are typically located in urban areas and near roadways. A grid-based model like CMAQ may not be able to capture the sub gridscale variations in concentration reflected in the data that are due to local emissions sources and thus may not agree with the observed data unless the data are representative of area encompassed by a grid cell. Thus, model performance for these species was examined only for the 5- and 1-km grids. 5-km Grid Comparison of Simulated and Observed Concentrations Scatter plots comparing simulated and observed 24-hour PM 10 concentrations for AQS sites within the 5- km grid for the annual simulation period are presented in Figure Units for PM 10 are µg/m 3. ICF International 102 New York City Department of Health and Mental Hygiene

112 Application and Evaluation of CMAQ for the Base Year Figure Comparison of Simulated and Observed 24-Hour Average PM 10 Concentration (µg/m 3 ) for the 5-km Grid (All Months) Scatter plots comparing simulated and observed hourly NO x, SO 2, and CO concentrations for AQS sites within the 5-km grid for the annual simulation period are presented in Figure Units for the gaseous species are ppb. ICF International 103 New York City Department of Health and Mental Hygiene

113 Application and Evaluation of CMAQ for the Base Year Figure Comparison of Simulated and Observed Hourly Average NO 2, SO 2, and CO Concentrations (ppb) for the 5-km Grid (All Months) NO 2 /SO 2 CO As expected, agreement between the simulated and observed values for these species is not very good. PM 10 concentrations are mostly underestimated, but there is a lot of scatter about the 1:1 line. High observed values tend to be underestimated while the low observed values are both under- and overestimated. Model performance for 1-hour NO 2, SO 2, and CO concentrations is characterized by a good deal of scatter about the 1:1 line and a tendency for underestimation of the higher observed values and overestimation of the low values. ICF International 104 New York City Department of Health and Mental Hygiene

114 Application and Evaluation of CMAQ for the Base Year Statistical Measures of Model Performance Summary metrics and statistical measures for PM 10, NO x, and SO 2 for the 5-km grid are presented in Table 4-8. No lower bound was applied in calculating the statistics; fractional bias and error are emphasized. Table 4-8. Summary Model Performance Statistics for PM 10, NO x, SO 2 and CO for the 5-km Modeling Grid Metric PM 10 (µg/m3) NO 2 (ppb) SO 2 (ppb) CO (ppb) Number of Data Pairs 5, , , ,113 Mean Observed Mean Simulated Mean Bias Fractional Bias (%) Mean Error Fractional Error (%) The statistics suggest better model performance than the scatter plots. A fractional bias within ±67 percent indicates that the simulated values are, on average, within a factor of two of the observed values. This is achieved for all four pollutants. However, as indicated by the scatter plots and confirmed by the greater errors, the relatively low bias values for PM 10, NO 2, and CO are due to a mix of under- and overestimation and not necessarily to good model performance. 1-km Grid Comparison of Simulated and Observed Concentrations Scatter plots comparing simulated and observed 24-hour PM 10 concentrations for AQS sites within the 1- km grid for the annual simulation period are presented in Figure Units for PM 10 are µg/m 3. ICF International 105 New York City Department of Health and Mental Hygiene

115 Application and Evaluation of CMAQ for the Base Year Figure Comparison of Simulated and Observed 24-Hour Average PM 10 Concentration (µg/m 3 ) for the 1-km Grid (All Months) Scatter plots comparing simulated and observed hourly NO 2, SO 2, and CO concentrations for AQS sites within the 1-km grid for the annual simulation period are presented in Figure Units for the gaseous species are ppb. ICF International 106 New York City Department of Health and Mental Hygiene

116 Application and Evaluation of CMAQ for the Base Year Figure Comparison of Simulated and Observed Hourly Average NO 2, SO 2, and CO Concentrations (ppb) for the 1-km Grid (All Months) NO 2 /SO 2 CO Similar to the results for the 5-km grid, the higher PM 10 concentrations are underestimated in the 1-km grid. Note, however, that there are very few valid data available from the AQS database. Model performance for 1-hour NO 2, SO 2, and CO concentrations is characterized by a good deal of scatter about the 1:1 line. For SO 2 and CO there is a tendency for overestimation of some of the low to intermediate concentrations. ICF International 107 New York City Department of Health and Mental Hygiene

117 Application and Evaluation of CMAQ for the Base Year Statistical Measures of Model Performance Summary metrics and statistical measures for PM 10, NO x, and SO 2 for the 1-km grid are presented in Table 4-9. No lower bound was applied in calculating the statistics; fractional bias and error are emphasized. Table 4-9. Summary Model Performance Statistics for PM 10, NO x, SO 2 and CO for the 1-km Modeling Grid Metric PM 10 (µg/m3) NO 2 (ppb) SO 2 (ppb) CO (ppb) Number of Data Pairs 59 38,207 61,670 51,319 Mean Observed Mean Simulated Mean Bias Fractional Bias (%) Mean Error Fractional Error (%) The statistical measures suggest that, on average, PM 10, NO 2 and CO concentrations are underestimated and SO 2 concentrations are overestimated by the CMAQ model. Despite the higher grid resolution, model performance for these pollutants is about the same for the 1-km as the 5-km grid. Using the PM 2.5 goals as a reference point, the fractional bias and fraction error for all four pollutants are within the goals (although the SO 2 error comes close). ICF International 108 New York City Department of Health and Mental Hygiene

118 5 Modeling Results for the Heating Oil Scenarios The first set of CMAQ scenarios examined the effects of local changes in heating fuels. The heating oil control scenarios are as follows: Partial implementation of the rule on heating oil, the emissions changes reflect the reduction in emissions achieved by the end of the winter heating season (April 2013). Full implementation of the rule on heating oil (complete phase out of No. 4 and No. 6 heating oil). Both heating oil scenarios also included implementation of a 15 ppm sulfur limit to No. 2 heating oil. The CMAQ modeling results for these scenarios are summarized in the remainder of this section. 5.1 Scenario #1 Scenario #1 reflects partial implementation of the New York City heating oil rule through For residual boilers, emissions of NO x, SO 2 and PM 2.5 are reduced by 20, 35 and 24 percent, respectively, from 2008 levels. For distillate boilers, emissions of SO 2 and PM 2.5 are reduced by 99 and 98 percent, respectively, due to implementation of a 15 ppm sulfur limit to No. 2 heating oil. Overall, for Scenario #1, emissions of NO x, SO 2 and PM 2.5 are reduced by 8, 69 and 67 percent, respectively, from 2008 levels, and CO emissions are increased by 5 percent. Figure 5-1 illustrates the simulated differences in daily maximum 8-hour ozone concentration for the 15 th of each month for April, July, and October (every third month during the ozone season) for the 5- and 1-km grids. The differences are calculated as Scenario #1 minus 2008 base. ICF International 109 New York City Department of Health and Mental Hygiene

119 Modeling Results for the Heating Oil Scenarios Figure 5-1. Difference in Simulated Daily Maximum 8-Hour Ozone Concentration: Scenario #1 Minus Base April (5-km Grid)/April (1-km Grid) July (5-km Grid)/July (1-km Grid) ICF International 110 New York City Department of Health and Mental Hygiene

120 Modeling Results for the Heating Oil Scenarios October (5-km Grid)/October (1-km Grid) The reduction in NO x emissions leads to some small simulated increases in ozone concentration, especially over New York City. The greatest increases range from approximately 0.7 to 0.9 ppb for the 1- km grid for the dates shown and are located over Manhattan. The response of the CMAQ model to the changes in emissions is influenced by the complex photochemistry represented by the model. Under certain conditions (usually for urban areas characterized by a low VOC-to-NO x ratio), decreases in NO x emissions can lead to increases in ozone. Figure 5-2 illustrates the simulated differences in monthly average PM 2.5 concentration for every third month (January, April, July, and October) for the 5- and 1-km grids. The differences are calculated as Scenario #1 minus 2008 base. ICF International 111 New York City Department of Health and Mental Hygiene

121 Modeling Results for the Heating Oil Scenarios Figure 5-2. Difference in Simulated Monthly Average PM 2.5 Concentration: Scenario #1 Minus Base January (5-km Grid)/January (1-km Grid) April (5-km Grid)/April (1-km Grid) ICF International 112 New York City Department of Health and Mental Hygiene

122 Modeling Results for the Heating Oil Scenarios July (5-km Grid)/July (1-km Grid) October (5-km Grid)/October (1-km Grid) Figure 5-3 illustrates the simulated differences in annual average PM 2.5 concentration for the 5- and 1- km grids. The differences are calculated as Scenario #1 minus 2008 base. ICF International 113 New York City Department of Health and Mental Hygiene

123 Modeling Results for the Heating Oil Scenarios Figure 5-3. Difference in Simulated Annual Average PM 2.5 Concentration: Scenario #1 Minus Base Annual (5-km Grid)/Annual (1-km Grid) The reduction in NO x and SO 2 emissions leads to simulated decreases in PM 2.5 concentration over New York City. The greatest decreases in monthly average PM 2.5 concentrations range from approximately 1.3 to 9.5 µg/m 3 for the 1-km grid for the dates shown. As expected, the decreases are greatest for the winter months when heating fuel consumption is greatest. Annual average PM 2.5 concentrations within the 1-km grid are reduced by as much as 4.3 µg/m 3. Decreases occur over Manhattan, Brooklyn, Queens, and the Bronx, with the greatest decreases over Manhattan. Figure 5-4 illustrates the simulated differences in daily maximum 1-hour NO 2 concentration for the 15 th of every third month (January, April, July, and October) for the 5- and 1-km grids. The differences are calculated as Scenario #1 minus 2008 base. ICF International 114 New York City Department of Health and Mental Hygiene

124 Modeling Results for the Heating Oil Scenarios Figure 5-4. Difference in Simulated Daily Maximum 1-Hour NO 2 Concentration: Scenario #1 Minus Base January (5-km Grid)/January (1-km Grid) April (5-km Grid)/April (1-km Grid) ICF International 115 New York City Department of Health and Mental Hygiene

125 Modeling Results for the Heating Oil Scenarios July (5-km Grid)/July (1-km Grid) October (5-km Grid)/October (1-km Grid) Simulated daily maximum 1-hour NO 2 concentrations for Scenario #1 are lower than the base-year concentrations, both over New York City and beyond. The greatest decreases in 1-hour NO 2 concentrations range from approximately 1 to 2.6 ppb for the 1-km grid for the dates shown. While the maximum decreases tend to occur over Manhattan, for several of the days, the influence of the NO x reductions is also seen to the north of the City and in New Jersey. ICF International 116 New York City Department of Health and Mental Hygiene

126 Modeling Results for the Heating Oil Scenarios Figure 5-5 illustrates the simulated differences in daily maximum 1-hour SO 2 concentration for the 15 th of every third month (January, April, July, and October) for the 5- and 1-km grids. The differences are calculated as Scenario #1 minus base. Figure 5-5. Difference in Simulated Daily Maximum 1-Hour SO 2 Concentration: Scenario #1 Minus Base January (5-km Grid)/January (1-km Grid) April (5-km Grid)/April (1-km Grid) ICF International 117 New York City Department of Health and Mental Hygiene

127 Modeling Results for the Heating Oil Scenarios July (5-km Grid)/July (1-km Grid) October (5-km Grid)/October (1-km Grid) The CMAQ results show large decreases in simulated daily maximum 1-hour SO 2 concentration, ranging from approximately 11 to 70 ppb for the 1-km grid for the dates shown. These results are consistent with a 69 percent reduction in overall SO 2 emissions. ICF International 118 New York City Department of Health and Mental Hygiene

128 Modeling Results for the Heating Oil Scenarios 5.2 Scenario #2 Scenario #2 reflects full implementation of the New York City heating oil rule. For residual boilers, emissions of NO x, SO 2 and PM 2.5 are reduced by 69, 97 and 75 percent, respectively, from 2008 levels. For distillate boilers, the emissions of SO 2 and PM 2.5 are reduced by 99 and 98 percent, respectively, due to implementation of a 15 ppm sulfur limit to No. 2 heating oil. Overall, for Scenario #2, NO x, SO 2, and PM 2.5 emissions are reduced by 26, 98 and 89 percent, respectively, from 2008 levels, and CO emissions are increased by 25 percent. Figure 5-6 illustrates the simulated differences in daily maximum 8-hour ozone concentration for the 15 th of each month for April, July, and October (every third month during the ozone season) for the 5- and 1-km grids. The differences are calculated as Scenario #2 minus base. Figure 5-6. Difference in Simulated Daily Maximum 8-Hour Ozone Concentration: Scenario #2 Minus Base April (5-km Grid)/April (1-km Grid) ICF International 119 New York City Department of Health and Mental Hygiene

129 Modeling Results for the Heating Oil Scenarios July (5-km Grid)/July (1-km Grid) October (5-km Grid)/October (1-km Grid) The greater reduction in NO x emissions for Scenario #2 results in greater increases in ozone concentration for this scenario, over New York City and beyond. The greatest increases range from approximately 2.3 to 3.4 ppb for the 1-km grid for the dates shown. ICF International 120 New York City Department of Health and Mental Hygiene

130 Modeling Results for the Heating Oil Scenarios Figure 5-7 illustrates the simulated differences in monthly average PM 2.5 concentration for every third month (January, April, July, and October) for the 5- and 1-km grids. The differences are calculated as Scenario #2 minus 2008 base. Figure 5-7. Difference in Simulated Monthly Average PM 2.5 Concentration: Scenario #2 Minus Base January (5-km Grid)/January (1-km Grid) April (5-km Grid)/April (1-km Grid) ICF International 121 New York City Department of Health and Mental Hygiene

131 Modeling Results for the Heating Oil Scenarios July (5-km Grid)/July (1-km Grid) October (5-km Grid)/October (1-km Grid) Figure 5-8 illustrates the simulated differences in annual average PM 2.5 concentration for the 5- and 1- km grids. The differences are calculated as Scenario #2 minus 2008 base. ICF International 122 New York City Department of Health and Mental Hygiene

132 Modeling Results for the Heating Oil Scenarios Figure 5-8. Difference in Simulated Annual Average PM 2.5 Concentration: Scenario #2 Minus Base Annual (5-km Grid)/Annual (1-km Grid) The greater reduction in NO x and SO 2 emissions for Scenario #2 results in greater simulated decreases in PM 2.5 concentration over New York City. The greatest decreases in monthly average PM 2.5 concentrations range from approximately 2.2 to 11.8 µg/m 3 for the 1-km grid for the dates shown. Annual average PM 2.5 concentrations within the 1-km grid are reduced by as much as 5.5 µg/m 3. Decreases occur over Manhattan, Brooklyn, Queens, and the Bronx, with the greatest decreases over Manhattan. Figure 5-9 illustrates the simulated differences in daily maximum 1-hour NO 2 concentration for the 15 th of every third month (January, April, July, and October) for the 5- and 1-km grids. The differences are calculated as Scenario #2 minus 2008 base. ICF International 123 New York City Department of Health and Mental Hygiene

133 Modeling Results for the Heating Oil Scenarios Figure 5-9. Difference in Simulated Daily Maximum 1-Hour NO 2 Concentration: Scenario #2 Minus Base January (5-km Grid)/January (1-km Grid) April (5-km Grid)/April (1-km Grid) ICF International 124 New York City Department of Health and Mental Hygiene

134 Modeling Results for the Heating Oil Scenarios July (5-km Grid)/July (1-km Grid) October (5-km Grid)/October (1-km Grid) The greater reduction in NO x emissions for Scenario #2 results in greater simulated decreases in daily maximum 1-hour NO 2 concentration over New York City and beyond. The greatest decreases in 1-hour NO 2 concentrations range from approximately 3.6 to 8.6 ppb for the 1-km grid for the dates shown. The substantial decreases in simulated NO 2 concentration are the result of a 26 percent decrease in NO x emissions within the five-borough area. ICF International 125 New York City Department of Health and Mental Hygiene

135 Modeling Results for the Heating Oil Scenarios Figure 5-10 illustrates the simulated differences in daily maximum 1-hour SO 2 concentration for the 15 th of every third month (January, April, July, and October) for the 5- and 1-km grids. The differences are calculated as Scenario #2 minus 2008 base. Figure Difference in Simulated Daily Maximum 1-Hour SO 2 Concentration: Scenario #2 Minus Base January (5-km Grid)/January (1-km Grid) April (5-km Grid)/April (1-km Grid) ICF International 126 New York City Department of Health and Mental Hygiene

136 Modeling Results for the Heating Oil Scenarios July (5-km Grid)/July (1-km Grid) October (5-km Grid)/October (1-km Grid) The CMAQ results show large decreases in simulated daily maximum 1-hour SO 2 concentration, ranging from approximately 21 to 113 ppb for the 1-km grid for the dates shown. The decreases are greater than for Scenario #1 and consistent with a 98 percent reduction in overall SO 2 emissions within the five-borough area. ICF International 127 New York City Department of Health and Mental Hygiene

137 Modeling Results for the Heating Oil Scenarios blank page ICF International 128 New York City Department of Health and Mental Hygiene

138 6 Modeling Results for the EGU Fuel Scenarios The second set of scenarios examined the effects of changes in EGU fuel between 2005 and the present. As a first step in the EGU fuel assessment, CMAQ was run using a hybrid emission inventory consisting of the 2008 base year emissions for all sources except EGU sources, and 2005 emissions for EGU sources. This alternative baseline scenario (EGU Base) provided a basis for assessing the effects of changes in EGU emissions over time. The simulated concentrations and patterns are similar to those for the 2008 Base and are not shown here. The additional EGU scenarios are as follows: Adjustment of EGU emissions to reflect changes in fuel use at Title V EGUs outside of the five boroughs of New York City, using 2012 CEM emission data for these sources to represent present day. Adjustment of EGU emissions to reflect changes in fuel use at EGUs located within the five boroughs, using 2012 CEM emission data for these sources to represent present day. The CMAQ modeling results for these scenarios are summarized in the remainder of this section. 6.1 Scenario #3 For Scenario 3, the greatest decreases compared to the EGU base are for SO 2, followed by NO x, PM 2.5, and VOC. The SO 2 emissions are 70, 90 and 99 percent lower for the 15-, 5- and 1-km grids, respectively. The NO x emissions are 57, 63 and 84 percent lower. The PM 2.5 emissions are 11, 60 and 76 percent lower. The VOC emissions are 1, 36 and 84 percent lower. The CO emissions, however, are increased relative to the EGU Base and are 72, 114 and 53 percent higher for the 15-, 5- and 1-km grids, respectively. For reference, the emission differences between Scenario 3 and the EGU baseline were illustrated in Figure 3-7 (for the 1-km grid). Note that although the overall emissions are lower for the scenario, there are also some sources for which the emissions are higher (primarily due to the use of a different emission sources between the EGU Base and the scenario). This complicates the analysis of the results. Figure 6-1 illustrates the simulated differences in daily maximum 8-hour ozone concentration for the 15 th of each month for April, July, and October (every third month during the ozone season) for the 5- and 1-km grids. The differences are calculated as Scenario #3 minus EGU Base. ICF International 129 New York City Department of Health and Mental Hygiene

139 Modeling Results for the EGU Fuel Scenarios Figure 6-1. Difference in Simulated Daily Maximum 8-Hour Ozone Concentration: Scenario #3 Minus EGU Base April (5-km Grid)/April (1-km Grid) July (5-km Grid)/July (1-km Grid) ICF International 130 New York City Department of Health and Mental Hygiene

140 Modeling Results for the EGU Fuel Scenarios October (5-km Grid)/October (1-km Grid) Results for the 5-km grid are characterized by both increases and decreases in daily maximum 8-hour ozone concentrations for the sample days included in Figure 6-1. The reduction in NO x emissions outside of NYC but within the 1-km grid leads to simulated increases in ozone concentration throughout the 1- km grid, including over New York City. The greatest increases range from approximately 5 to 8.5 ppb for the days shown and are located in New Jersey, just west of Manhattan. The response of the CMAQ model to the changes in emissions is influenced by the complex photochemistry represented by the model. Under certain conditions (usually for urban areas characterized by a relatively low VOC-to-NO x ratio), decreases in NO x emissions can lead to increases in ozone. Figure 6-2 illustrates the simulated differences in monthly average PM 2.5 concentration for every third month (January, April, July, and October) for the 5- and 1-km grids. The differences are calculated as Scenario #3 minus EGU Base. ICF International 131 New York City Department of Health and Mental Hygiene

141 Modeling Results for the EGU Fuel Scenarios Figure 6-2. Difference in Simulated Monthly Average PM 2.5 Concentration: Scenario #3 Minus EGU Base January (5-km Grid)/January (1-km Grid) April (5-km Grid)/April (1-km Grid) ICF International 132 New York City Department of Health and Mental Hygiene

142 Modeling Results for the EGU Fuel Scenarios July (5-km Grid)/July (1-km Grid) October (5-km Grid)/October (1-km Grid) Figure 6-3 illustrates the simulated differences in annual average PM 2.5 concentration for the 5- and 1- km grids. The differences are calculated as Scenario #3 minus EGU Base. ICF International 133 New York City Department of Health and Mental Hygiene