1 Prepared by: Neil Wheeler Sonoma Technology, Inc. Petaluma, CA Presented to: CCOS Technical Committee Sacramento, CA August 15, Central California Ozone Study (CCOS) Improvement of Air Quality Model Aloft Performance

2 Overview Goals and Objectives Scope of Work Technical Approach Data and Modeling Issues Closing Discussion

3 Goals for this Meeting Clarify objectives for the project Discuss technical approach Exchange information Make decisions about episodes and models to be used

4 Project Objectives The objective of Phase 1 is to improve air quality model aloft performance to the extent possible in a time frame that will allow the results to be employed in planned 2006 SIP updates for central and northern California. As appropriate, subsequent improvements will be implemented in Phase 2.

5 Process Characterization Evaluation Identification of model performance issues Diagnosis of the possible causes Correction

6 Scope of Work (Phase 1) Task 1 – Acquire CAMx and CMAQ Models and Data Sets Task 2 – Conduct Analyses of Surface and Aloft Aerometric Data Task 3 – Evaluate CAMx and CMAQ Aloft Performance and Diagnose the Causes of Performance Problems Task 4 – Implement Modifications to Models and Inputs, Reevaluate Model Performance, and Diagnose the Causes of Remaining Aloft Performance Problems Task 5 – Deliver Phase 1 Model Software and Input and Output Files Task 6 – Prepare Interim Study Documentation and Meet With CCOS Technical Committee

7 Scope of Work (Phase 2) Task 7 – Implement Further Modifications to Models and Inputs, Reevaluate Model Performance, and Diagnose the Causes of Any Remaining Aloft Performance Problems Task 8 – Deliver Phase 2 Model Software and Input and Output Files Task 9 – Prepare Study Documentation and Meet With CCOS Technical Committee

8 Technical Approach

9 Task 1 – Acquire Models and Data Sets Identify two modeling episodes for analysis and improvement of aloft model performance. Acquire CAMx and CMAQ software and input and output files for the episodes identified. Install software and replicate appropriate existing simulations.

10 Task 2 – Conduct Analyses of Surface and Aloft Aerometric Data (1 of 2) Review existing data analysis results and conceptual models describing key physical and chemical phenomena thought to influence ozone formation in central and northern California. Review available findings of CCOS-sponsored meteorological analyses carried out by Technical & Business Systems, Inc. Acquire surface and aloft meteorological and ambient air quality data collected during the CCOS summer 2000 field measurement program. Identify existing graphical and tabular displays of ozone and precursor vertical profiles and spatial maps and develop additional displays as appropriate. Examine the consistency of surface and aloft ozone and precursor observations. Identify the characteristics (including timing, depths, and horizontal spatial extents) of aloft layers of ozone and precursors.

11 Task 2 – Conduct Analyses of Surface and Aloft Aerometric Data (2 of 2) Calculate ozone and precursor fluxes across defined planes where suitable measurements exist. Identify any special aloft phenomena, such as evidence of layers of ozone and precursors caused by wild fires. Identify likely mechanisms responsible for the formation of significant aloft layers of ozone and precursors. Assess the potential importance and timing of the fumigation of ozone and precursors from aloft layers on surface air quality observations.

12 Analysis (1 of 5) Create spatial and vertical spiral plots of aloft air quality data collected by aircraft and ozonesondes and average these data to match the model vertical resolution and horizontal grid. Calculate ozone and NO y flux through several planes for each episode day several times a day as the data permit. Use mixing height estimates to create hourly spatial maps of mixing heights, observed winds at several levels, and observed ozone.

13 Analysis (2 of 5) Create multivariate time series plots of meteorological and air quality data at several locations, including upwind and downwind of major source areas and in rural areas. Calculate transport statistics at several heights using RWP data. The statistics will include vector and scalar transport distance, direction, and recirculation factor for day and nighttime.

14 Analysis (3 of 5) Descriptive statistics - Metrics such as minimum, maximum, mean, and standard deviation will be calculated for available species and selected ratios of species, and summarized spatially and temporally. Potential ratios of species include VOC/NOx (or specific VOC species), CO/NOx, CO/NOy, O3/NOy, O3/NOz and O3/HNO3. Three-dimensional visualization.

15 Analysis (4 of 5) Chemical composition - The analysis of chemical composition can provide a means for identifying sources contributing to, the age of ozone precursors in, and ozone formation potential of an air sample. Our analyses would involve calculating, summarizing, and plotting species and species ratios of NOx and VOC tracers. Indicator species ratios - Several ratios of species have been good indicators of whether peak ozone formation is likely to be more sensitive to changes in VOC or NOx. The ratios, O3/NOy, O3/NOz and O3/HNO3, are more amenable to measurement and analyses.

16 Analysis (5 of 5) Biogenic hydrocarbons - Measurements of isoprene alone are often ambiguous because isoprene reacts so rapidly (20- to 60-minute half-life in midday). The combination of isoprene and its oxidation products is a far more stable measure of observed loadings of isoprene-related compounds than isoprene alone and, therefore, provides a more robust basis for evaluating the adequacy of the role of biogenic hydrocarbons in modeled ozone formation. Variations in chemical composition - Conceptually, we would expect the atmosphere’s chemical composition to vary over three regions: (1) air entering at the upwind boundary of the CCOS domain would contain little NOx or highly reactive VOC; (2) the composition of air immediately downwind of major emission sources (e.g., cities) would be dominated by ozone precursors; and (3) further downwind NOx and VOC oxidation products would dominate the composition. We will analyze predominant wind flows during key sampling periods and summarize the chemical composition relative to the regions identified above.

17 Task 3 – Evaluate Aloft Performance and Diagnose the Causes of Problems (1 of 2) Review model configuration and inputs. Review the initial findings of the NOAA and T&B Systems companion studies addressing MM5 aloft performance issues. Using available surface and aloft measurements made during the summer 2000 field measurement program, evaluate CAMx and CMAQ aloft model performance. Identify specific model aloft performance problems. Diagnose the possible causes of such problems. Indicate those problems that may be caused by shortcomings in meteorological inputs and those problems that may be caused for other reasons (such as systematic bias in emissions, boundary concentrations, and other inputs).

18 Task 3 – Evaluate Aloft Performance and Diagnose the Causes of Problems (2 of 2) Design a program to rectify air quality aloft performance issues associated with problems other than those attributed to meteorological inputs. Prepare a draft work plan discussing the proposed approach for improving air quality model aloft performance. Participate in a one-day meeting with the TC to discuss the proposed approach for improving aloft performance of CAMx and CMAQ and to obtain comments on the draft work plan. Address the comments provided by the TC and prepare and submit a final work plan for Phase 1 of the study.

19 Model Performance Evaluation (1 of 3) Peak performance statistics by level, location, and day; Bias and error statistics by level, location, and day; Scatter plots of observations versus predictions by level and location; Quantile-quantile plots; Vertical profiles of observed and predicted ozone and ozone precursor concentrations for balloon soundings and aircraft spirals; Time series comparisons of predicted and observed concentrations from transverse aircraft flights and tall towers; and Spatial plots of predicted concentrations at selected levels aloft with observations over-plotted.

20 Model Performance Evaluation (2 of 3) In comparing aircraft data with model predictions, we will average the aircraft measurements over the extent of the model grid cells through which the flight passed. For vertical profiles of the measurements, we will plot un-averaged concentrations with model predictions plotted at the midpoint of vertical cells along with maximum and minimum values within the surrounding cells to show gradients in the model predictions. In addition to traditional analyses, we will also duplicate key analyses from Task 2 using model input and output data. We will focus on the analyses that illustrated physical and chemical processes important to ozone formation.

21 Model Performance Evaluation (3 of 3) For example, we might replicate the flux- plane calculations or transport statistics from Task 2 with model data and compare them to the observation-based calculations. The advantage of flux-plane analysis is that it is based on a metric derived from both physical and chemical data and, therefore, tests the modeling system’s ability to integrate physical and chemical processes properly at a particular place and time.

22 Identification of Performance Problems The historical performance criteria for SIP ozone modeling will be used initially to assess the adequacy of ozone model performance. A threshold concentration of 20 ppb will be used in calculating performance statistics. However, model performance cannot be judged by statistical measures alone; more subjective criteria will also be used to assess model performance for ozone and other modeled species. These criteria may include the spatial extent of high concentrations, the ability to replicate diurnal and day-to-day concentration variations, and the ability to predict the locations of minimum and maximum observed concentrations. We expect that this aloft model performance evaluation will identify performance problems similar to those in past evaluations. However, by using more directed analyses, we expect to describe the performance problems more clearly and lay a better foundation for exploring the causes of these problems.

23 Diagnosis of the Causes of Performance Problems Investigative analysis Diagnostic simulation Process analysis Hypothesis testing

24 Hypotheses (1 of 2) Emission inventories are biased low. Boundary concentrations are inaccurately specified. Grid resolution is insufficient. Meteorological models are unable to capture stagnant air conditions caused by terrain blocking of the flow. Vertical mixing of pollutants in the planetary boundary layer are overestimated resulting in relatively clean air aloft being mixed downward Mixing of high ozone surface air to the interior of the convective boundary layer is inadequate.

25 Hypotheses (2 of 2) Recirculation of upslope flow from the Sierra Nevada Mountains and Coastal Range over the Central Valley are inadequately represented. Chemical mechanisms underestimate ozone production efficiency at low precursor concentrations Dispersion in areas of significant and steep terrain in central California are treated inadequately Photolysis rates at higher elevations are treated inadequately Wildfires are not adequately represented

26 Task 4 – Implement Modifications, Reevaluate, and Diagnose Remaining Problems Based on the findings of Task 3, implement modifications to the air quality model and its inputs. Obtain updated meteorological inputs from the companion MM5 study. Using the CCOS database, evaluate CAMx and CMAQ surface and aloft model performance for ozone and its precursors. Examine air quality model performance issues that may arise from discrepancies in vertical grid structures used in the meteorological and air quality models. Identify any remaining model performance issues. Diagnose the possible causes of remaining model performance problems.

27 Task 5 – Deliver Phase 1 Model Software and Input and Output Files Identify a mutually agreeable means for transmitting all appropriate modeling software and input and output files developed during Phase 1 of the study to the ARB and local districts currently performing modeling exercises for the selected episodes. Provide modified models, model input and output files, and documentation describing the content, directory structures, and file-naming conventions for all information delivered.

28 Task 6 – Prepare Interim Study Documentation and Meet With CCOS Technical Committee Prepare and submit a draft interim report that documents the analyses of the CCOS aerometric data and the initial efforts to evaluate, diagnose, and rectify CAMx and CMAQ aloft performance. If efforts are needed to further improve air quality model aloft performance, prepare a draft work plan for proposed Phase 2 activities. If no further model improvement work is needed, prepare a draft manuscript for submittal to a peer-reviewed journal summarizing the findings of the study. Participate in a one-day meeting with the TC in Sacramento, California, to describe the initial findings of the study, to receive comments on the draft interim report (and draft journal manuscript if appropriate), and to discuss any further model improvement work needed in Phase 2 of the study. Respond to comments provided by the TC and submit a final interim report (or a final report and manuscript if no further work is to be performed) and, if appropriate, a final work plan for Phase 2 of the study.

29 Task 7 – Implement Further Modifications, Reevaluate, and Diagnose Remaining Problems Implement the proposed program to rectify remaining air quality aloft performance issues. Incorporate any further improvements to meteorological inputs developed in the companion MM5 study. Conduct such further analyses as may be needed to identify the causes of remaining CAMx and/or CMAQ aloft performance issues. Implement appropriate modifications to the models and their inputs. Reevaluate model performance, both at the surface and aloft, and assess the adequacy of model performance. Identify remaining air quality aloft performance problems. Provide possible explanations of the causes of any remaining performance issues.

30 Task 8 – Deliver Phase 2 Model Software and Input and Output Files Identify a mutually agreeable means for transmitting all appropriate modeling software and input and output files developed during Phase 1 of the study to the ARB and local districts currently performing modeling exercises for the selected episodes. Provide modified models, model input and output files, and documentation describing the content, directory structures, and file-naming conventions for all information delivered.

31 Task 9 – Prepare Study Documentation and Meet With CCOS Technical Committee Prepare and submit a draft final report that documents the efforts to evaluate, diagnose, and rectify aloft air quality model performance. Prepare a draft manuscript summarizing the results of the study suitable for submittal to a peer-reviewed journal. Participate in a one-day meeting with the TC in Sacramento, California, to discuss the findings of the study and to receive comments on the draft final report and journal manuscript. Respond to comments provided by the TC and submit a final report for the study. Modify the draft journal manuscript, as appropriate, and submit it for review and publication in a peer-reviewed journal.

32 Discussion

33 Items for Discussion Episodes July 8 -15, 1999 July 29 – August 2, 2000 September 16 – 21, 2000 Models CAMx (Version modifications?) – MM5 meteorology – MM5/CALMET hybrid meteorology CMAQ-APT (July-August 2000 only)

34 Closing Discussion Obtaining Model Input/Output files ARB Others MM5 Schedule Deliverables

35 Schedule

36 Deliverables